Econometrica | Macro Paper Warehouse

Can Deficits Finance Themselves?

Mon, 01 Jan 0001 00:00:00 +0000

The paper asks whether a government can run a deficit today — issuing “stimulus checks” — and allow debt to return to its initial level without any future tax hike or spending cut. In environments combining (i) nominal rigidity and (ii) a violation of Ricardian equivalence (due to finite lives or liquidity constraints), this is possible through two complementary self-financing channels: (a) a Keynesian boom in real activity that expands the tax base and automatically raises revenue at existing tax rates; and (b) a surge in inflation that erodes the real value of outstanding nominal government debt. The paper’s headline result is that self-financing increases monotonically as fiscal adjustment is delayed, converging to full self-financing in the limit: if monetary policy does not lean too heavily against the fiscal stimulus, the initial deficit eventually returns debt to trend with no required future adjustment. Calibrated to empirical evidence on intertemporal MPCs, the speed of fiscal adjustment, the Phillips curve slope, and the monetary reaction, the model finds self-financing up to ν ≈ 0.95 — with the tax base channel dominant and inflation contributing negligibly.

Environment (Section 2): Baseline is a perpetual-youth overlapping-generations (OLG) version of the textbook New Keynesian model. Households survive from one period to the next with probability ω ∈ (0,1]; when ω=1 the model reduces to the standard PIH-RANK benchmark in which Ricardian equivalence holds and no self-financing occurs. When ω<1, two properties of consumer demand emerge: (i) consumers discount future disposable income at a rate higher than the interest rate (“discounting”), so a distant future tax hike barely affects today’s spending; (ii) consumers spend transfers relatively quickly (“front-loading”), so the Keynesian boom plays out before the promised tax hike arrives. The supply block is exactly the standard NKPC. Fiscal policy follows a rule in which taxes respond to income through a fixed tax rate τy (tax base channel) and to debt through a speed-of-adjustment coefficient τd ∈ (0,1) (with τd→0 meaning indefinitely delayed adjustment). Monetary policy keeps (expected) real rates constant in the baseline — a “neutral” benchmark that neither offsets nor amplifies the fiscal stimulus.

Self-financing result (Sections 3–4): Starting from a date-0 deficit shock ε0 (lump-sum transfer of 1% of steady-state output), define the degree of self-financing ν as the fraction of ε0 financed by the tax base and debt erosion channels; 1−ν equals the discounted present value of future tax hikes required to stabilize debt. The central results are:

Theorem 1 (baseline, φ=0): If ω<1 and τy>0, ν increases monotonically as τd→0, with ν→1 in the limit. Intuition via two-period analogy: when cumulative short-run MPC → 1, the Keynesian multiplier → 1/τy, and the induced tax revenue → 1 — exactly financing the original ε0.
Proposition 3: For any given τd or delay H, ν is strictly decreasing in ω: larger departures from permanent income (smaller ω) deliver faster and larger Keynesian booms and hence greater self-financing.
Theorem 2 (general monetary policy): Under a general real rate rule rt = φ·yt, there exists a threshold φ̄ ∈ (0, τy/(β·D^ss/Y^ss)) such that: if φ<φ̄, full self-financing is achieved in the limit; if φ>φ̄, ν is bounded strictly below 1 by ν̄(φ). If the monetary authority perfectly stabilizes output and inflation (φ→∞), ν=0 by construction.
Theorem 3 (general aggregate demand): With generalized demand ct = Md·dt + My·(yt−tt) + δ·Et[Σ(βω)^k(yt+k−tt+k)], self-financing holds whenever (i) ω<1 and (ii) Md>1−β and My·(1 + δ·βω/(1−βω)) ≥ 1. This nests the baseline OLG model, hybrid spender-OLG models, and approximately represents quantitative HANK models.

Distinction from FTPL: The Fiscal Theory of the Price Level (Cochrane) breaks Ricardian equivalence through equilibrium selection in a PIH-RANK setting; the self-financing here operates under the conventional equilibrium, with an active monetary authority and passive fiscal authority. The inflation channel is not the focal mechanism — the tax base channel is dominant.

Calibration (Table 1, hybrid OLG-spender model, quarterly frequency):

Consumer spending: share of hand-to-mouth (HtM) spenders µ = 0.073; OLG survival rate ω = 0.865; jointly matched to average MPC = 0.2 and short-run MPC slope from Fagereng, Holm, and Natvik (2021)
Fiscal adjustment: τd ∈ {0.085, 0.026, 0.004} (fast to slow; from Galí et al. 2007, Bianchi-Melosi 2017, Auclert-Rognlie 2020 respectively; equivalent to H ∈ {12, 23, 43} quarters under the non-Markovian rule)
Monetary policy: real rate feedback φ = 0 (neutral baseline)
Nominal rigidities: NKPC slope κ = 0.0062 (Hazell et al. 2022 point estimate)
Standard parameters: EIS σ=1 (log utility); β = 0.998 (1% annual real rate); tax feedback τy = 0.33 (DeLong-Summers benchmark: 33 cents of surplus per dollar of output); liquid wealth D^ss/Y^ss = 1.04 (Kaplan et al. 2018)

Quantitative results (Figure 3, Table 2):

For empirically calibrated τd range, ν reaches up to 0.95, nearly full self-financing in the most realistic (slow adjustment) specification
Virtually all self-financing (≈95–100%) occurs through the tax base channel — the flat NKPC (κ=0.0062) limits inflation and debt erosion to a negligible share; with steeper NKPC (κ=0.1), about 20% of self-financing comes through date-0 inflation
The quantitative fiscal multiplier at τd=0.085 is 1.11, consistent with Ramey (2011) empirical estimates for transfers with relatively quick adjustment
Table 2 (νmax as function of monetary ψ and NKPC κ): Full self-financing (νmax = 1) is attainable when ψ ≤ 1.25 and κ = 0.0062; drops to νmax = 0.63 at ψ=1.5 and κ=0.0062; drops to νmax = 0.22 with κ=0.1 and ψ=1; approaches 0 with both aggressive monetary and flexible prices. Key lesson: moderate monetary reaction combined with flat NKPC (consistent with evidence) supports near-full self-financing.

Robustness:

HANK model: same conclusions as hybrid spender-OLG; intertemporal MPCs nearly identical (Wolf, 2021; Auclert et al., 2023)
Distortionary fiscal adjustment: negligible impact, since the required adjustment itself vanishes in the limit
Government purchases: same self-financing logic applies (Keynesian boom raises tax revenue)
Investment: Keynesian cross applies to consumption; net of investment aggregate demand follows the same law of motion — self-financing result unchanged

Scope conditions: Self-financing requires Ricardian equivalence to fail (ω<1); in the PIH-RANK benchmark (ω=1), neither self-financing channel is operative. Monetary accommodation is assumed neutral or weak; aggressive offsetting (φ>φ̄) prevents full self-financing. The paper is purely positive: whether deficits are optimal is a separate normative question. Results are log-linearized dynamics; the quantitative conclusions depend on discipline from empirical MPC evidence, NKPC estimates, and fiscal adjustment speed. The self-financing mechanism operates through aggregate demand and is not driven by r<g or by seigniorage from a convenience yield.

In depth

Q1. What is the two-period intuition for full self-financing?

In a two-period economy with fully myopic consumers (MPC=1), a date-0 transfer of ε stimulates output by y = MPC/(1−MPC·(1−τy)) · ε, generating tax revenue τy·y; with MPC→1 the output multiplier converges to 1/τy and tax revenue converges to exactly ε — full self-financing via the tax base. The infinite-horizon economy with ω<1 mirrors this intuition when fiscal adjustment is delayed far enough: the “short run” cumulative MPC approaches 1 (by discounting and front-loading), the Keynesian cross delivers a multiplier of 1/τy, and the additional tax revenue precisely repays the deficit, with no future tax hike needed.

Q2. Why does the degree of self-financing ν increase as fiscal adjustment is delayed?

As the gap H between the date-0 transfer and the promised future tax hike widens, two effects amplify the Keynesian boom: (i) near-term demand is less dampened by anticipation of the future tax hike (discounting makes far-ahead taxes nearly irrelevant to today’s spending); and (ii) the general equilibrium income feedback — the Keynesian cross — has more time to play out before being curtailed by the eventual tax hike, amplifying the total output and revenue response. The longer the delay, the larger the short-run cumulative MPC, and the larger the fraction of the deficit self-financed through the tax base.

Q3. Why does aggressive monetary policy block self-financing?

If the monetary authority raises real interest rates in response to the fiscal boom (φ>0), it discourages household spending, slowing and shrinking the Keynesian boom; above the threshold φ̄, the real rate increase is strong enough to counteract the tax base feedback before the cumulative MPC can converge to 1, meaning full self-financing becomes impossible and some future fiscal adjustment is always required. Conversely, monetary accommodation (φ<0) accelerates the boom and permits full self-financing with less delay, while perfectly stabilizing output and inflation (φ→∞) entirely shuts down both self-financing channels.

Q4. What is the role of the NKPC slope in determining which channel operates?

When the NKPC is flat (κ=0.0062, the Hazell et al. 2022 estimate), a large output boom generates negligible inflation, so debt erosion contributes almost nothing and the tax base channel carries essentially all the self-financing; when the NKPC is steep (κ=0.1, consistent with supply-constrained post-COVID), the same boom generates materially more inflation, shifting the financing split so that ~20% comes through debt erosion while ~80% still comes through the tax base. The overall degree of self-financing ν is affected only through the monetary response: a steeper NKPC triggers a more aggressive real rate response, moderating the boom, but this is captured in the analysis of Theorem 2 and Table 2.

Q5. How does this paper relate to and differ from the Fiscal Theory of the Price Level (FTPL)?

The FTPL (Cochrane) achieves deficit financing through inflation in a PIH-RANK environment by abandoning the Taylor principle and exploiting equilibrium selection; this paper requires no such departure — both monetary and fiscal policy follow conventional active/passive assignments, and the equilibrium studied is the unique bounded one. The key difference is in the consumer block: Ricardian equivalence fails here through finite lives or liquidity constraints (empirically grounded), not through equilibrium selection. Moreover, while FTPL highlights the debt erosion (inflation) channel, this paper finds the tax base (real activity) channel is dominant under empirically calibrated flat Phillips curves.

Q6. What new conditions on aggregate demand ensure self-financing extends beyond the OLG baseline?

Theorem 3 identifies two sufficient conditions: (1) “positive geometric discounting” (ω<1 in the generalized demand block), ensuring that far-ahead future taxes have negligible effect on current demand; and (2) “sufficient front-loading” (Md > 1−β and My·(1 + δ·βω/(1−βω)) ≥ 1), ensuring that income is spent quickly enough for the Keynesian feedback to deliver self-financing before debt explodes. The classical PIH-RANK fails condition (1); the spender-saver model with any margin of PIH consumers fails condition (2); the OLG baseline satisfies both; and the hybrid spender-OLG (the quantitative workhorse) satisfies both for any ω<1.

Q7. Is a margin of truly PIH consumers fatal for self-financing?

Yes — introducing any strictly positive mass of PIH consumers breaks self-financing entirely, creating a discontinuity: ν=0 whenever µ_PIH > 0, no matter how small. The intuition is that PIH consumers never fully spend any income received in finite time (they smooth it across their infinite horizon), so the cumulative MPC never reaches 1 and the Keynesian boom cannot fully finance the deficit. However, the discontinuity is fragile: replacing literal PIH consumers with “near-PIH” consumers (finite but large ω) restores ν→1 in the limit as H→∞ and is consistent with empirical evidence on high MPCs for liquid households.

Key concepts

fiscal self-financing : the property that a deficit-financed government transfer raises output and inflation sufficiently to replenish government revenue (via the tax base channel) and reduce the real debt burden (via the inflation/debt erosion channel), allowing debt to return to steady state without future tax increases; the degree ν ∈ [0,1] measures what fraction of the initial deficit is self-financed.

tax base channel : the mechanism by which a Keynesian boom in real activity — triggered by the deficit-financed transfer — automatically raises tax revenue (by τy dollars per dollar of additional output) without any change in tax rates; dominant over the debt erosion channel whenever the NKPC is flat (empirically, κ ≈ 0.006).

discounting and front-loading : the two consumer demand properties necessary for self-financing; “discounting” (ω<1) means far-ahead future taxes barely affect current spending, allowing the deficit to stimulate demand even with a promised future tax hike; “front-loading” means the income response is spent quickly, so the Keynesian boom plays out before the delayed tax hike arrives, raising tax revenue sufficiently to finance the deficit.

speed of fiscal adjustment (τd) : the quarterly feedback from public debt to tax revenue in the fiscal rule; τd→0 means indefinitely delayed adjustment and maximum self-financing; empirically disciplined values range from τd=0.085 (fast, Galí et al. 2007) to τd=0.004 (slow, Auclert-Rognlie 2020), with νmax ≈ 0.95 across this range under neutral monetary policy and flat NKPC.

hybrid spender-OLG model : the paper’s quantitative workhorse, combining a fraction µ of hand-to-mouth spenders with OLG perpetual-youth consumers; jointly calibrated to match the impact and short-run MPCs from Fagereng et al. (2021), while also providing a close proxy for aggregate demand in quantitative HANK models (Auclert et al. 2023; Wolf 2021).

Can Trade Policy Mitigate Climate Change?

Mon, 01 Jan 0001 00:00:00 +0000

Overview

Farrokhi and Lashkaripour (2025) study the interaction between trade policy and climate change. The central research question is whether and how countries can use trade policy — specifically import tariffs — to address carbon leakage arising from domestic carbon pricing. When a country prices carbon domestically, production and emissions can shift to countries without carbon pricing, partially offsetting domestic emissions reductions. The paper asks how optimal import tariffs should be designed to internalize this leakage, how they relate to standard terms-of-trade tariffs, and what additional gains multilateral coordination can deliver.

Methodology and Data. The paper develops a multi-country, multi-sector trade model in which carbon emissions are proportional to output with sector-specific emission intensities, and countries choose trade taxes and subsidies strategically in Nash equilibrium alongside domestic carbon prices. The model is calibrated to 43 countries and 56 sectors using the 2014 baseline from the World Input-Output Database (WIOD 2016) for trade flows and input-output linkages, IEA data for sector-level carbon emissions, and GTAP for trade elasticities.

Main Findings. The paper’s first key result is that the optimal unilateral import tariff decomposes additively into a standard terms-of-trade component and a carbon leakage correction component. The carbon leakage correction is proportional to the emission intensity of imports from the exporting country in that sector and to the gap between the social cost of carbon and the actual domestic carbon price in the exporting country, divided by the import price. This decomposition implies that countries have incentives to impose import tariffs beyond those justified by standard terms-of-trade arguments, specifically to correct for the carbon embodied in imports from countries with insufficient carbon pricing.

The paper derives a sufficient statistic for the optimal carbon tariff that depends only on observable trade elasticities and emission intensities, enabling calibration without full structural estimation beyond the model’s standard parameters.

Quantitative Magnitudes. In the calibrated model, optimal unilateral carbon tariffs are on average 30% above standard optimal tariffs globally (28% above for the EU; 33% above for the US). The excess is largest in carbon-intensive sectors: petroleum products (41% above standard optimal), cement and non-metallic minerals (45% above standard optimal), basic metals (38% above standard optimal), and chemicals (32% above standard optimal). Imposing the optimal unilateral carbon tariff yields a welfare gain of +0.8% consumption equivalent for the imposing country, with trading partners losing on average 0.3%, and a net global gain of +0.4%.

Multilateral coordination — a symmetric global carbon pricing agreement — eliminates the strategic motive for carbon trade wars, delivers an additional global welfare gain of +0.6% above the unilateral optimum, and eliminates 85% of the carbon leakage remaining under unilateral policy.

CBAM Analysis. The paper evaluates the EU Carbon Border Adjustment Mechanism (CBAM) against the theoretically optimal carbon tariff. The EU CBAM as currently implemented — covering only direct emissions — captures 60% of the theoretically optimal carbon tariff. Extending coverage to indirect (supply-chain) emissions would capture 85% of optimal. The welfare gain to the EU from CBAM relative to no border adjustment is +0.4%.

Scope Conditions and Robustness. Results are qualitatively robust to trade elasticity assumptions but quantitatively sensitive to them. Optimal carbon tariffs are regressive with respect to developing countries; multilateral coordination mitigates this distributional effect via income transfers. General equilibrium labor market effects reduce welfare gains by approximately 20% but do not change the qualitative ranking of policies.

In depth

Q1. What is the formal structure of the optimal unilateral import tariff in the presence of carbon externalities?

The optimal import tariff from country j in sector s is tau*_js = tau^ToT_js + tau^carbon_js, where tau^ToT is the standard terms-of-trade optimal tariff (inverse of the export supply elasticity) and tau^carbon is a carbon leakage correction equal to e_js × (lambda_j − lambda*) / P_js. Here e_js is the emission intensity of country j in sector s, lambda_j is the social cost of carbon in the importing country, lambda* is the actual domestic carbon price in the exporting country, and P_js is the import price. Countries therefore have two distinct and additive incentives to impose import tariffs: the classical terms-of-trade motive and a novel carbon leakage correction motive.

Q2. What is the sufficient statistic result and why does it matter for implementation?

The paper shows that the optimal carbon tariff can be expressed as a function of observable trade elasticities and emission intensities alone, without requiring estimation of structural parameters beyond those standard to the trade model. This sufficient statistic result matters because it means regulators can in principle calculate and implement the theoretically optimal carbon border adjustment using data that are already collected — sectoral emission intensities and trade elasticities — rather than relying on unobservable structural primitives.

Q3. By how much do optimal carbon tariffs exceed standard optimal tariffs in the aggregate and in the most carbon-intensive sectors?

Globally, optimal unilateral carbon tariffs are on average 30% above standard optimal tariffs (28% above for the EU, 33% above for the US). The excess is largest in highly carbon-intensive sectors: cement and non-metallic minerals (45% above), petroleum products (41% above), basic metals (38% above), and chemicals (32% above). These are precisely the sectors where emission intensities are highest, consistent with the carbon leakage correction being proportional to emission intensity.

Q4. What are the welfare effects of unilateral optimal carbon tariff policy?

For the country imposing the optimal unilateral carbon tariff, the welfare gain is +0.8% in consumption-equivalent terms relative to no carbon tariff. Trading partners lose on average 0.3%. The net global welfare gain is +0.4%. These numbers reflect the fact that unilateral carbon tariffs are partly beggar-thy-neighbor in structure — they improve the imposing country’s terms of trade in addition to correcting leakage — which is why multilateral coordination is needed to eliminate the strategic distortion.

Q5. What additional gains does multilateral coordination deliver over unilateral policy?

Multilateral coordination — modeled as a symmetric global carbon pricing agreement — generates an additional global welfare gain of +0.6% above the unilateral optimum. It also eliminates 85% of the carbon leakage that persists under unilateral policy. The mechanism is that coordination removes the strategic motive for trade wars over carbon policy: under unilateral policy, each country has an incentive to impose carbon tariffs partly for terms-of-trade reasons, but under a coordinated agreement these beggar-thy-neighbor components are internalized.

Q6. How well does the EU’s CBAM as actually implemented capture the theoretically optimal carbon border adjustment?

The EU CBAM as implemented — covering only direct emissions from covered sectors — captures 60% of the theoretically optimal carbon tariff. Extending the CBAM to include indirect emissions embedded in supply chains would raise this to 85% of optimal. The remaining gap (15% under the extended CBAM) reflects the difficulty of accounting for all upstream emission intensities across complex global supply chains.

Q7. What is the welfare gain to the EU from CBAM relative to no border adjustment?

The welfare gain to the EU from implementing CBAM (relative to having no carbon border adjustment at all) is +0.4% in consumption-equivalent terms. This figure corresponds to the direct CBAM as implemented, covering only direct emissions.

Q8. How sensitive are the results to trade elasticity assumptions, and what are the distributional implications for developing countries?

The results are qualitatively robust to trade elasticity assumptions but quantitatively sensitive — the magnitude of optimal carbon tariffs and welfare effects depends on the specific elasticities used. On distributional grounds, optimal carbon tariffs are regressive with respect to developing countries, meaning developing economies bear disproportionate costs from carbon border adjustments. Multilateral coordination partially mitigates this distributional concern through income transfers implied by the symmetric global agreement.

Q9. How do general equilibrium labor market effects alter the conclusions?

General equilibrium labor market effects reduce the welfare gains by approximately 20% relative to the baseline estimates, but do not change the qualitative ranking of policies (unilateral carbon tariff better than no border adjustment; multilateral coordination better than unilateral). This suggests that the core policy conclusions are robust to incorporating labor market general equilibrium effects, even if the precise magnitudes are somewhat smaller.

Key Concepts

Carbon Leakage. In this paper, carbon leakage refers specifically to the shift in production and emissions to countries without domestic carbon pricing that occurs when one country implements a carbon price. It is the mechanism by which domestic carbon pricing is partially offset, motivating the use of trade policy as a complementary instrument.

Carbon Leakage Correction (tau^carbon). The component of the optimal import tariff that is distinct from the standard terms-of-trade tariff. It equals emission intensity × (social cost of carbon − domestic carbon price in exporter) / import price. It corrects for the fact that imports from countries with insufficient carbon pricing embody unpriced carbon externalities.

Terms-of-Trade Tariff (tau^ToT). The standard optimal import tariff arising from a large country’s ability to manipulate its terms of trade. Equal to the inverse of the export supply elasticity of the trading partner. The paper establishes that carbon tariffs add to — rather than replace — this classical component.

Sufficient Statistic for Optimal Carbon Tariff. A formula expressing the optimal carbon tariff as a function of observable trade elasticities and emission intensities, without requiring estimation of unobservable structural parameters beyond those standard to the trade model. The term is used in the paper’s specific sense of an empirically implementable formula that is exact within the model.

Emission Intensity. Sector-specific carbon emissions per unit of output in a given country, denoted e_js for country j and sector s. Used as the key observable that scales the carbon leakage correction component of the optimal tariff.

Multilateral Coordination. Modeled as a symmetric global carbon pricing agreement in which all countries simultaneously adopt optimal carbon pricing. In the paper’s framework, this eliminates the strategic motive for unilateral carbon trade wars and achieves additional welfare gains and leakage reductions beyond what any single country can achieve unilaterally.

Carbon Border Adjustment Mechanism (CBAM). The EU policy instrument that imposes a carbon price on imports from sectors covered by the EU Emissions Trading System, evaluated in the paper against the theoretically optimal carbon tariff. The paper distinguishes between the direct-emissions-only CBAM as implemented (capturing 60% of optimal) and a hypothetical full CBAM including indirect supply-chain emissions (capturing 85% of optimal).

Cap‐and‐Trade and Carbon Tax Meet Arrow–Debreu

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1 — Overview

Research Question

Anderson and Duanmu (2025) ask how general equilibrium (GE) interactions — factor reallocation across sectors, capital misallocation under climate uncertainty, and the distributional incidence of damages — alter the social cost of carbon (SCC) relative to the partial equilibrium (PE) estimates embedded in standard integrated assessment models (IAMs). The paper also characterizes conditions for Pareto improvements through climate policy and derives the optimal carbon tax in second-best environments with pre-existing distortions.

Framework

The authors build a dynamic Arrow-Debreu economy with L goods, K capital stocks (including climate stocks), and T periods. The climate module specifies that the carbon stock evolves as S_{t+1} = S_t + sum_j e_j(q_j) − alpha·S_t, and climate damage functions D_j(S_t) = 1 − d_j·(S_t − S_0) reduce sector-specific production possibilities sets. Firms and households take the climate trajectory as given and do not internalize their own emissions’ impact, generating the externality. Under standard regularity conditions, the authors prove existence of a competitive equilibrium and establish that it is inefficient: output is too high and climate-intensive sectors are too large relative to the social optimum.

General Formula for the SCC

The paper derives a general SCC formula — SCC_t = Sum_{tau >= t} beta^(tau−t) · [dW/dS_tau / (dW/dY_t)] — that decomposes into four components: (1) the standard direct productivity-loss term, (2) a GE factor-reallocation term capturing inefficient reallocation as damages shift relative prices, (3) a capital-misallocation term reflecting distortions in investment from climate uncertainty, and (4) a distribution term reflecting the welfare losses from the regressive incidence of climate damages. All three correction terms are positive under standard conditions, so the GE SCC exceeds the PE SCC. The paper shows that this formula nests existing IAM frameworks as special cases.

Quantitative Findings

Calibrating to three leading IAMs, the authors find that general equilibrium interactions raise the SCC by 15–40% above standard PE estimates:

DICE-calibrated: GE correction of 18% above the PE estimate.
FUND-calibrated: GE correction of 15% above the PE estimate.
PAGE-calibrated: GE correction of 40% above the PE estimate, the largest correction owing to greater sector heterogeneity in that model.
Median calibration: a PE SCC of $51/tCO₂ rises to a GE SCC of $62/tCO₂.

Decomposing the aggregate GE correction: factor reallocation across sectors accounts for 55%, capital misallocation due to climate uncertainty for 30%, and the distributional regressivity of damages for 15%.

Second-Best Policy and Uncertainty

In environments with pre-existing distortions, the optimal carbon tax deviates from the SCC: revenue recycling through labor tax cuts generates additional welfare gains of 10–15% of carbon tax revenue; undertaxed capital implies the optimal carbon tax should be set above the SCC (double dividend); and in monopolistically competitive sectors the optimal carbon tax is below the SCC because the carbon tax amplifies monopoly distortions. Under climate uncertainty, the SCC carries a risk premium proportional to the variance of damage estimates times the coefficient of relative risk aversion, estimated at +$8–15/tCO₂ (15–25% of the base SCC).

Scope Conditions

The quantitative corrections are calibrated to DICE, FUND, and PAGE and therefore inherit those models’ parameterizations of damage functions and discount rates. The GE factor-reallocation and capital-misallocation channels are larger when sectors are more heterogeneous in damage exposure — as is explicit in the PAGE result. Second-best corrections depend on the sign and magnitude of pre-existing distortions (labor taxes, capital taxes, market structure).

In depth

Q1. What is the core inefficiency result, and what does it imply about the competitive equilibrium?

The paper’s efficiency theorem establishes that the competitive equilibrium is Pareto inefficient because firms and households take the climate trajectory as given and do not internalize the impact of their own emissions on the carbon stock. As a consequence, output is too high and climate-intensive sectors are too large relative to the social optimum. This externality is the fundamental justification for climate policy in the model.

Q2. How does the paper’s general SCC formula extend existing approaches, and what are the novel terms?

The general formula SCC_t = Sum_{tau >= t} beta^(tau−t) · [dW/dS_tau / (dW/dY_t)] nests standard IAM SCC formulas as special cases. The novel terms relative to partial equilibrium are: (i) a GE reallocation term capturing losses from inefficient factor reallocation as climate damages change relative prices across sectors; (ii) a capital-misallocation term capturing distortions in investment arising from climate uncertainty; and (iii) a distribution term capturing welfare losses from the regressive incidence of damages. All three terms are positive under standard conditions, implying GE SCC > PE SCC in all calibrations.

Q3. How are the quantitative GE corrections decomposed, and which channel dominates?

Of the total GE correction above the PE baseline, factor reallocation across sectors contributes 55%, capital misallocation due to climate uncertainty contributes 30%, and the distributional regressivity of damages contributes 15%. Factor reallocation is the dominant channel because, as climate damages alter relative prices, production shifts toward less-damaged sectors in ways that are distorted by the original carbon externality — generating second-order losses absent from PE damage functions.

Q4. Why does the PAGE calibration produce a larger GE correction (40%) than DICE (18%) or FUND (15%)?

The paper attributes PAGE’s larger GE correction to greater sector heterogeneity in that model’s parameterization. When damage exposure is more heterogeneous across sectors, the relative-price effects of marginal carbon are larger, amplifying the factor-reallocation channel. DICE and FUND, with more uniform sector-level damage structures, exhibit smaller reallocation corrections.

Q5. What is the median-calibration implication for the SCC in dollar terms?

In the median calibration, a PE SCC of $51/tCO₂ rises to a GE SCC of $62/tCO₂, an increase of roughly $11/tCO₂ or approximately 22%. This figure is directly computable from observable trade elasticities and sector-level damage estimates.

Q6. How should the carbon tax be adjusted when pre-existing labor market distortions are present, and what is the magnitude of the welfare gain from revenue recycling?

When labor taxes create a pre-existing wedge, using carbon tax revenue to reduce labor taxes generates additional welfare gains of 10–15% of total carbon tax revenue — the double dividend in the labor market dimension. The optimal carbon tax in this case includes the SCC plus a correction term for the labor-market distortion.

Q7. How do capital market distortions alter the optimal carbon tax relative to the SCC?

If capital is undertaxed (a pre-existing distortion in capital markets), the optimal carbon tax is set above the SCC. The intuition is that a higher carbon tax partially offsets the under-taxation of capital by raising the effective cost of carbon-intensive investment, capturing a double-dividend in the capital market.

Q8. How does monopolistic competition modify the optimal carbon tax?

For monopolistically competitive sectors, the optimal carbon tax is below the SCC. The reasoning is that applying a carbon tax to these sectors amplifies existing monopoly markups and associated distortions, so the social cost of the carbon tax exceeds the raw SCC in those sectors. The optimal policy trades off carbon correction against monopoly amplification.

Q9. What is the risk premium in the SCC under climate uncertainty, and how is it estimated?

The paper adds a term to the SCC proportional to the variance of damage estimates times the coefficient of relative risk aversion. Using empirical estimates of damage uncertainty, this risk premium is estimated at +$8–15/tCO₂, representing 15–25% of the base SCC. This term is absent from deterministic SCC calculations and constitutes a further reason standard PE estimates understate the true social cost.

Q10. What is the paper’s claim regarding computability of the GE correction?

The paper states that the novel GE terms are computable from observable trade elasticities and sector-level damage estimates, implying the GE correction is not merely a theoretical construct but can be implemented in quantitative policy analysis using data sources already available to researchers and policymakers.

Key Concepts

Social Cost of Carbon (General Equilibrium Formula) Defined in the paper as SCC_t = Sum_{tau >= t} beta^(tau−t) · [dW/dS_tau / (dW/dY_t)], the present discounted value of the marginal welfare loss from an additional unit of carbon, expressed relative to the marginal utility of current output. The paper’s version adds GE reallocation, capital-misallocation, and distributional terms absent from standard PE formulations.

GE Adjustment Factor The ratio of the general equilibrium SCC to the partial equilibrium SCC, expressed as GE/PE = 1 + phi_realloc + phi_capital + phi_distribution. Under standard conditions all three phi terms are positive, so the GE SCC strictly exceeds the PE SCC.

Climate Damage Function (Sector-Specific) Specified as D_j(S_t) = 1 − d_j·(S_t − S_0), a sector-specific multiplicative reduction in the production possibilities set as the carbon stock rises above the pre-industrial level S_0. Heterogeneity in d_j across sectors is the driver of the factor-reallocation GE correction.

Carbon Stock Evolution S_{t+1} = S_t + sum_j e_j(q_j) − alpha·S_t, where alpha is the natural decay rate of atmospheric carbon and e_j(q_j) is sectoral emissions as a function of output. Firms and households treat S_t as exogenous, generating the externality.

Double Dividend In second-best environments, a carbon tax can generate two welfare gains simultaneously: correcting the carbon externality and reducing the deadweight loss from a pre-existing distortion (labor or capital tax). The paper finds revenue recycling via labor tax cuts yields 10–15% of carbon tax revenue as additional welfare gain; undertaxed capital implies the optimal carbon tax is set above the SCC.

Risk Premium in the SCC An additive term in the SCC under climate uncertainty, proportional to the variance of damage estimates times the coefficient of relative risk aversion. Empirically estimated at +$8–15/tCO₂, representing 15–25% of the base SCC.

Second-Best Optimal Carbon Tax Written as tau*_carbon = SCC + CORRECTION, where the correction depends on the sign and magnitude of pre-existing distortions. The correction is positive under undertaxed capital (raise above SCC), negative under monopolistic competition (lower below SCC), and augmented by revenue-recycling gains when labor taxes are present.

Comment on 'Asset Bubbles and Overlapping Generations' by Tirole

Mon, 01 Jan 0001 00:00:00 +0000

Tirole (1985) studied an overlapping generations model with capital accumulation and showed that the emergence of asset bubbles can resolve the capital over-accumulation problem when the economy is dynamically inefficient. His Proposition 1(c) claims that a bubble can emerge if and only if the dividend growth rate exceeds the bubbleless steady-state interest rate. This comment identifies an error in that proposition: the stated condition is necessary but not sufficient for bubble existence. The paper constructs an explicit counterexample in which the dividend growth rate exceeds the bubbleless interest rate but no bubble equilibrium exists, and separately constructs a case in which a bubble exists even when the condition in Proposition 1(c) fails. Corrected necessary and sufficient conditions for bubble existence are derived, and the implications of the correction for the welfare results and the relationship between dynamic inefficiency and bubbles are characterized.

In depth

Q1. What is the error in Tirole’s Proposition 1(c)?

The error is in the sufficiency direction: Tirole argued that whenever the dividend growth rate exceeds the bubbleless interest rate, a bubble equilibrium exists; Pham and Toda construct a parameter configuration satisfying this condition where no bubble equilibrium exists, because the continuity argument used in Tirole’s proof fails at boundary parameter values. The necessity direction — that bubble existence requires this rate comparison — is not challenged.

Q2. How do the corrected conditions change the interpretation of dynamic inefficiency?

Tirole’s original result linked bubbles tightly to dynamic inefficiency (r < g), providing a clean condition for when bubbles are both feasible and welfare-improving by absorbing excess saving. The correction weakens this link: bubble existence requires additional structural conditions beyond the rate comparison, meaning dynamic inefficiency is a necessary but not sufficient condition for bubbles in the Tirole framework. Policy prescriptions based on the r < g condition for bubble welfare analysis need qualification.

Key concepts

dynamic inefficiency : the OLG condition in which the interest rate falls below the growth rate, making intergenerational transfers from young to old welfare-improving; related to but not sufficient for bubble existence under the corrected Tirole conditions.

bubble existence condition : the necessary and sufficient conditions under which an asset bubble can emerge and persist in the Tirole OLG model; the corrected version requires more than the dividend-growth-rate-exceeds-interest-rate comparison of the original Proposition 1(c).

Designing Disability Insurance Reforms: Tightening Eligibility Rules or Reducing Benefits?

Mon, 01 Jan 0001 00:00:00 +0000

This paper develops a sufficient statistics framework for the welfare analysis of disability insurance (DI) policy reforms and applies it to two reform episodes in Austria. The framework derives social optimality conditions for the two main DI policy instruments — eligibility rules and benefit levels — expressed in terms of estimable reduced-form objects (fiscal multipliers and insurance losses). The fiscal multiplier of a DI policy instrument is defined as the ratio of total fiscal cost savings to the mechanical (counterfactual-behavior-held-fixed) fiscal cost savings; it measures how much the program shrinks per dollar mechanically removed, and values above 1 indicate behavioral crowd-out of DI enrollment. The paper then evaluates two Austrian reforms: (1) a 2013 increase in the Rehabilitation Stricter Assessment (RSA) age threshold from 57 to 58 (and separately to 59), which tightened eligibility for DI applicants aged 57 by requiring them to demonstrate inability to be retrained for alternative work; and (2) a 2003 reform that reduced DI benefit generosity for workers aged 30–60 as a side effect of a pension reform. Using difference-in-differences with cohorts just above and below the relevant thresholds, the paper finds that the RSA reform generated a fiscal multiplier of 2.50 (RSA to 58) and 2.05 (RSA to 59), while the benefit reduction generated a multiplier of only 1.41 (ages 57–60) and 1.36 (ages 30–56). The large gap implies that for a given mechanical cost saving, tighter eligibility rules generate 1.8 times more total fiscal savings than benefit cuts. The paper further provides empirical evidence that the insurance losses associated with stricter eligibility rules are, in all likelihood, smaller than those from benefit reductions, strengthening the dominance of eligibility tightening over benefit cuts as a DI reform instrument.

In depth

Q1. What is the sufficient statistics framework and what does it deliver?

The paper derives social optimality conditions for two DI policy instruments — tighter eligibility rules and lower benefits — in terms of two sufficient statistics: the fiscal multiplier (total fiscal savings / mechanical fiscal savings) and the insurance loss (marginal utility of consumption of the affected recipients). An eligibility reform that tightens the threshold θ from θ* to θ* + dθ is welfare-improving if and only if the fiscal multiplier exceeds the social value of one dollar in the hands of the marginally excluded applicant; a benefit cut from b to b − db is welfare-improving if the multiplier exceeds the social value of one dollar held by the average current DI recipient. Because the fiscal multiplier is estimable from reduced-form variation and the insurance loss gives the welfare benchmark, the framework converts the welfare question into: “Is the multiplier large enough relative to the insurance value?”

Q2. How is the fiscal multiplier decomposed, and why does this decomposition matter?

The fiscal multiplier equals 1 + B/M where B is the behavioral fiscal effect (savings from deterred applications and enrollment) and M is the mechanical fiscal effect (savings from unchanged behavior on the affected population); the multiplier exceeds 1 whenever the behavioral response amplifies the direct savings. The decomposition matters because the behavioral effect operates through marginal applicants (who apply only under lenient rules) while the mechanical effect operates through always-applicants (who apply regardless). These groups have different characteristics: in the Austrian data, marginal applicants are more likely to be employed at age 56 (73% vs. 60% for always-applicants) and more likely to be blue-collar workers with musculoskeletal impairments, while always-applicants are more likely to be on sick leave — a proxy for genuine disability. Confusing the two groups would misidentify who bears the insurance loss.

Q3. How is the mechanical fiscal effect identified when marginal and always-applicants cannot be directly observed?

The paper exploits previously-rejected DI applicants (those who filed applications between ages 50–56 and were rejected) as a proxy group for always-applicants: these individuals qualify as always-applicants by revealed preference (they applied under strict rules), and the paper shows that their DI benefit receipt and net fiscal expenditures after a simulated age-57 application are statistically indistinguishable from those of all-age-57 applicants in the whole population. The mechanical fiscal effect per always-applicant in the whole population is estimated as M = 5,585 Euro × 0.070 = 391 Euro per capita (the product of the mechanical effect among pre-57 re-applicants and the share of always-applicants in the population, πAA = 0.070). The behavioral fiscal effect is then computed as the residual: B = total fiscal savings – M = 976 − 391 = 585 Euro, yielding the multiplier of 2.50.

Q4. What are the reduced-form effects of the RSA reform on DI enrollment and employment?

A one-year increase in the RSA from age 57 to 58 reduces DI inflow by approximately 21 percentage points at age 57 (relative to a control-group mean), increases employment by roughly 15 percentage points, increases other benefit receipt (unemployment insurance and social assistance) by about 16 percentage points, and generates net fiscal cost savings of approximately 976 Euro per person per year in the two years after the reform. The pattern of employment and benefit substitution shows that the behavioral response is substantial: a large share of those deterred from DI enrollment at 57 transition to employment or other social benefits rather than remaining without any income support, which is why the fiscal multiplier of 2.50 substantially exceeds 1.

Q5. What are the effects of the 2003 DI benefit generosity reduction?

A 1-percentage-point reduction in DI benefit generosity for the ages 57–60 cohort produces a behavioral fiscal effect of 18.69 Euro per year (through reduced DI inflow and application), a mechanical fiscal effect of 45.16 Euro per year (1% of the pre-reform mean benefit expenditure of 4,516 Euro among those aged 57–60), and a total fiscal effect of 63.85 Euro — yielding a multiplier of 1.41. For the younger cohort (ages 30–56), the multiplier is 1.36, with a behavioral effect of 1.18 Euro and mechanical effect of 3.24 Euro per year. The lower multipliers for benefit cuts relative to eligibility tightening reflect the fact that benefit reductions affect all current recipients uniformly (generating large mechanical savings) rather than targeting a group with a strong behavioral response at the margin.

Q6. How does the paper compare the insurance losses of the two DI instruments?

The paper derives a sufficient condition under which the insurance loss from tighter eligibility rules is smaller than the insurance loss from benefit cuts: it requires that the per-dollar income loss borne by the marginally excluded applicant (upper-bounded by their DI benefit minus available social welfare benefits) is weakly smaller than the per-dollar income loss of current recipients (lower-bounded by the benefit reduction itself), evaluated within each income quintile. Implementing this condition empirically using the Austrian income data, the paper finds that the income losses borne by marginally excluded applicants fall short of those borne by current recipients at all income quintile comparisons — meaning tighter eligibility rules both generate higher fiscal multipliers and impose smaller insurance losses than benefit cuts, making eligibility tightening the dominant instrument when the goal is to reduce DI program costs.

Q7. What welfare conclusion follows from combining the fiscal multipliers with the insurance benchmark?

Combining the multiplier estimates with hand-to-mouth CRRA assumptions as an upper-bound calculation, the RSA increase to 58 is welfare-improving if the coefficient of relative risk aversion of affected DI recipients is below 2.8; the RSA increase to 59 is welfare-improving if risk aversion is below 2.2. The corresponding critical risk aversion level for the benefit cut (ages 57–60) is 1.1 — below the range typically estimated in the literature for low-income individuals — suggesting the benefit cut was likely welfare-reducing while the eligibility reform was likely welfare-improving. For a given dollar of mechanical budget reduction, stricter eligibility rules generate 1.8 times the total fiscal savings (= 2.50 / 1.41) relative to benefit cuts.

Q8. What is the complier analysis and what does it reveal about who is affected by each instrument?

Using the complier analysis method adapted for difference-in-differences settings, the paper estimates that the RSA-58 reform affects three types of individuals in the age-57 population: marginal applicants (πMA = 0.014, who apply only under lenient rules), always-applicants (πAA = 0.070, who apply regardless), and never-applicants (πNA = 0.916, who never apply). Marginal applicants differ from always-applicants in that they are more likely to be employed at age 56 (73% vs. 60%) and less likely to be on sick leave; they are more likely to apply with musculoskeletal impairments than with mental impairments — consistent with these workers facing the largest relaxation in disability eligibility when reaching the RSA and being on the borderline of eligibility under strict rules.

Key concepts

fiscal multiplier of a DI instrument : total fiscal cost savings divided by mechanical fiscal cost savings from that instrument; equals 1 + B/M where B is the behavioral savings (from deterred applications) and M is the mechanical savings (from unchanged behavior); values above 1 indicate behavioral crowd-out and are the policy-relevant benchmark against which insurance losses must be compared.

mechanical fiscal effect (M) : the fiscal cost savings that would accrue if DI application and enrollment behavior were held fixed at pre-reform levels; for eligibility tightening, this equals the DI benefits that would have been paid to always-applicants who are now rejected; identified using the subpopulation of previously-rejected DI applicants as a proxy for always-applicants.

behavioral fiscal effect (B) : the additional fiscal savings generated by deterred applications and enrollment that result from the reform; equals total fiscal savings minus the mechanical fiscal effect; operates through marginal applicants who adjust their application behavior in response to stricter rules or lower benefits.

always-applicants : individuals who apply for DI regardless of whether eligibility rules are strict or lenient; they bear the mechanical cost of eligibility tightening (being rejected under stricter rules); in the Austrian data, their population share at age 57 is estimated at 7.0%.

rehabilitation stricter assessment (RSA) age : the Austrian policy threshold above which DI applicants are evaluated under more lenient standards that do not require demonstration that the applicant can be retrained for alternative work; increasing the RSA age from 57 to 58 subjects the age-57 cohort to stricter evaluation criteria.

insurance loss : the welfare cost to DI recipients or excluded applicants from the income reduction caused by a DI reform; the right-hand side of the social optimality condition; the paper bounds it using income losses by income quintile rather than requiring utility function assumptions.

Double Robustness of Local Projections and Some Unpleasant VARithmetic

Mon, 01 Jan 0001 00:00:00 +0000

This paper provides formal theoretical results on the relative robustness of local projection (LP) and vector autoregression (VAR) confidence intervals for impulse response inference when the data generating process (DGP) is locally misspecified. The research question is whether the widely held belief that LP estimators are more robust to misspecification than VARs is theoretically justified, and if so, precisely under what conditions and with what consequences for VAR inference.

The analytical framework models the DGP as a stationary structural VARMA(1, ∞) that is local to an SVAR(1), of the form y_t = Ay_{t-1} + H[I + T^{-ζ}α(L)]ε_t, where the MA component T^{-ζ}α(L)ε_t represents misspecification that vanishes at rate T^{-ζ} as sample size T grows. The key rate parameter is ζ ∈ (1/4, 1/2), which corresponds to misspecification large enough to be detected with probability approaching 1 by conventional Hausman-type specification tests, yet small enough that the bias-variance trade-off between LP and VAR remains non-trivial asymptotically. The framework encompasses under-specification of lag length, omitted variables, temporal aggregation, measurement error, and failure of shock invertibility — essentially all sources of dynamic misspecification relevant to linearized DSGE models.

The main finding on LP is a “double robustness” result: the conventional LP confidence interval achieves correct asymptotic coverage for all ζ > 1/4, even when misspecification is large enough to be detected with certainty. The mechanism is that the omitted-variable bias in the LP regression is of order T^{-2ζ} = o(T^{-1/2}) when ζ > 1/4, because both the direct effect of omitted lags on the outcome and the covariance of the residualized regressor with omitted lags are each of order T^{-ζ}, so their product is negligible relative to the T^{-1/2} standard deviation. This is formally analogous to double robustness in partially linear regression and debiased machine learning: LP is consistent if either the outcome-equation controls or the first-stage controls are correctly specified.

In stark contrast, the VAR estimator carries asymptotic bias of order T^{-ζ}, which is non-negligible relative to its T^{-1/2} standard deviation for ζ ≤ 1/2. This causes the conventional VAR confidence interval to severely undercover: for ζ ∈ (1/4, 1/2) the coverage converges to zero, and for ζ = 1/2 it converges to a level strictly below the nominal level.

The “no free lunch” result formalizes the trade-off. Setting ζ = 1/2 and bounding the noise-to-signal ratio at M²/T, the worst-case scaled VAR bias equals M√(aVar(β̂_h)/aVar(δ̂_h) − 1). This worst-case bias is small if and only if the VAR asymptotic variance is close to that of LP. When the VAR standard error is less than half that of LP — which is typical in applied practice — worst-case coverage falls below 48% even for M = 1. Moreover, the least favorable misspecification takes the form of exponentially decaying MA coefficients peaking at horizon h, a pattern consistent with standard economic theories of adjustment costs, learning, or overshooting, and is difficult to rule out on prior grounds. The Hausman test also provides weak protection: when M = 1, the odds of the test failing to reject are nearly 3-to-1 at the 10% significance level.

Simulations using the Smets and Wouters (2007) model with T = 240 observations confirm these results. With lag length selected by AIC (median selected p = 2), VAR confidence intervals materially undercover at all but very short horizons while LP achieves close to nominal coverage throughout. Increasing lag length to p = 4 or p = 8 ameliorates VAR undercoverage at short horizons but at the cost of making VAR confidence intervals essentially as wide as LP intervals, with substantial undercoverage persisting at longer horizons. For p = 4 the total misspecification measure is M ≈ 3.23; for p = 8, M ≈ 1.89.

Scope conditions: results are pointwise asymptotic in fixed model parameters and horizon; they abstract from order-T^{-1} small-sample biases from persistence or the nonlinearity of the impulse response transformation. The LP robustness result requires controlling for lags that are strong predictors of the outcome or impulse variables; omitting lags with small-to-moderate predictive power does not threaten coverage.

Q: What is the precise sense in which LP confidence intervals are “doubly robust”?

A: LP is doubly robust in the sense of partially linear regression: its bias from misspecified MA dynamics is the product of two errors, the estimation error in the outcome-equation lag controls γ̂ − γ_0 and the estimation error in the first-stage lag controls ν̂ − ν_0. In the local-to-SVAR model each error is of order T^{-ζ}, so their product is of order T^{-2ζ} = o(T^{-1/2}) whenever ζ > 1/4, making the omitted-variable bias negligible relative to the T^{-1/2} standard deviation. This means the asymptotic distribution of the LP estimator is completely invariant to the misspecification parameters α(L) and ζ.

Q: How large does misspecification need to be before LP coverage is threatened?

A: The LP double robustness result holds for all ζ > 1/4 regardless of the magnitude parameter M of the MA misspecification. Misspecification with ζ ∈ (1/4, 1/2) can be detected with probability approaching 1 asymptotically by standard specification tests — in particular, the Hausman test is consistent for this range — yet LP coverage remains exactly correct. There is no threshold M below which LP fails; robustness is structural, not contingent on misspecification being small.

Q: Under what conditions does the VAR estimator have zero asymptotic bias?

A: The VAR asymptotic bias is zero if and only if the lagged shocks ε_{j*,t-ℓ} for ℓ = 1, …, h lie in the span of the lagged data used for estimation. Two sufficient conditions from Corollary 3.2 are: (i) the true model is SVAR(p_0) and the estimation lag length p satisfies h ≤ p − p_0, so the extra lags absorb the residual MA structure; or (ii) the shock of interest is directly observed and ordered first, and h ≤ p. In these cases the VAR estimator is asymptotically equivalent to LP, with equal variance.

Q: What is the “no free lunch” result for VARs?

A: For ζ = 1/2 and noise-to-signal ratio bounded by M²/T, the worst-case scaled VAR bias equals M√(aVar(β̂_h)/aVar(δ̂_h) − 1) (Proposition 4.1). This quantity is small if and only if aVar(δ̂_h) ≈ aVar(β̂_h), meaning the VAR has little efficiency advantage over LP. Put differently, the only way to guarantee robust VAR coverage is to include enough lags that the VAR confidence interval becomes as wide as the LP interval. There is no procedure that simultaneously offers narrower intervals than LP and reliable coverage.

Q: How severe is the worst-case undercoverage of conventional VAR confidence intervals?

A: From Corollary 4.3, even for M = 1 (a noise-to-signal ratio of just 1/T), worst-case VAR coverage falls below 48% whenever the VAR asymptotic standard deviation is less than half that of LP — a configuration typical in applied practice. For larger M the undercoverage is worse: the formula 1 − r(M√(aVar(β̂_h)/aVar(δ̂_h) − 1); z_{1-α/2}) can approach zero. Furthermore, the worst-case probability that VAR fails to cover AND the Hausman test fails to reject misspecification simultaneously exceeds 46% when the VAR standard deviation is less than half that of LP (Corollary 4.4).

Q: Can the researcher detect the problematic misspecification using a Hausman test before it causes undercoverage?

A: Only weakly. When M = 1, the Hausman test fails to reject misspecification with probability approximately 74% (odds of nearly 3-to-1) at the 10% significance level, since r(1; z_{0.95}) = 26%. At the 5% level the odds of non-rejection are nearly 5-to-1, since r(1; z_{0.975}) = 17%. The least favorable misspecification also cannot be ruled out on economic-theory grounds: the least favorable MA polynomial has exponentially decaying coefficients peaking at horizon h, consistent with adjustment costs, learning, or overshooting.

Q: Does using a bias-aware critical value (Armstrong-Kolesár approach) resolve the VAR undercoverage problem?

A: The bias-aware VAR confidence interval CI_B(δ̂_h; M) achieves correct asymptotic coverage by inflating the critical value based on the known bound M on misspecification. However, the bias-aware VAR interval tends to be wider than the LP interval. Specifically, M must be quite small — apparently below 1 — for the bias-aware VAR to dominate LP in width regardless of DGP and horizon. For M ≥ 2 (noise-to-signal ratio above 4/T), bias-aware VAR is dominated by LP in interval width. The practical conclusion is that the simpler LP interval is preferable in most empirically relevant settings.

Q: What does the minimax model-averaging result say about optimal weighting of LP and VAR?

A: From Corollary 4.2, the minimax optimal weight on LP when estimating a convex combination of LP and VAR estimators is M²/(1 + M²). For M = 1 (equal noise-to-signal threshold), the optimal weight is 50% on each. For M = 2, the LP estimator receives 80% weight. In the Smets and Wouters simulations, M ≈ 3.23 for p = 4 lags, corresponding to an optimal LP weight of approximately 91%, and M ≈ 1.89 for p = 8 lags, giving an optimal LP weight of approximately 78%.

Q: What do the Smets and Wouters simulations show about AIC-selected VARs?

A: In 5,000 simulated samples of T = 240 observations from the Smets and Wouters (2007) model, the AIC selects a median lag length of p = 2. At all but very short horizons, VAR confidence intervals materially undercover while LP confidence intervals throughout achieve close to nominal coverage. A bootstrap correction for VARs somewhat improves coverage but leaves large distortions. Increasing lag length to p = 4 or p = 8 moves coverage closer to nominal at short horizons (h ≤ p) but makes VAR confidence intervals essentially as wide as LP, and substantial VAR undercoverage persists at longer horizons.

Q: Is the no-free-lunch result specific to univariate impulse responses?

A: No. Proposition 4.2 extends the result to simultaneous inference on multiple impulse responses. For any k × 1 linear combination R of the impulse response vector, the worst-case squared bias is M² λ_max(R[aVar(β̂) − aVar(δ̂)]R’), where λ_max denotes the largest eigenvalue. Because VAR impulse response estimates are often highly correlated across horizons, undercoverage can be particularly severe in the multivariate (joint confidence ellipsoid) case. The no-free-lunch principle holds: the VAR ellipsoid offers non-negligible worst-case bias as long as it offers any efficiency gain relative to LP for any linear combination of horizon-specific impulse responses.

Q: What is the practical recommendation for lag selection in LP and VAR?

A: The paper offers three practical guidelines. First, LP researchers should control for those lags of the data that are strong predictors of the outcome or impulse variables, using conventional information criteria (such as AIC) applied to a VAR in all variables to select the number of lags for LP control — omitting lags with small-to-moderate predictive power does not threaten coverage. Second, VAR researchers should increase the lag length until the VAR confidence interval is no longer substantially narrower than the corresponding LP interval. Third, conventional specification tests do not suffice to guard against VAR coverage distortions.

Local Projection (LP) Estimator: The LP estimator for the impulse response at horizon h is the OLS coefficient on the shock variable y_{j*,t} in a direct regression of y_{i*,t+h} on y_{j*,t}, the variables ordered before it, and lagged data. It is a “direct” estimator in that it does not iterate a one-step VAR forward.

Double Robustness: A property of LP whereby its asymptotic bias from MA misspecification equals the product of two estimation errors — in the outcome-equation lag controls and in the first-stage residualization controls — each of order T^{-ζ}, making their product of order T^{-2ζ} = o(T^{-1/2}) for ζ > 1/4. This is the LP analogue of the double robustness of partially linear regression estimators in debiased machine learning.

Local-to-SVAR Misspecification: A DGP of the form y_t = Ay_{t-1} + H[I + T^{-ζ}α(L)]ε_t in which the MA term T^{-ζ}α(L)ε_t represents misspecification that vanishes at rate T^{-ζ}. The rate parameter ζ governs the magnitude; ζ ∈ (1/4, 1/2) is the empirically relevant range where bias is detectable by specification tests yet the bias-variance trade-off between LP and VAR remains non-trivial.

No Free Lunch (for VARs): The result that the worst-case scaled VAR bias equals M√(aVar(β̂_h)/aVar(δ̂_h) − 1), implying that the VAR confidence interval has reliable (robust) coverage if and only if the VAR asymptotic variance is close to that of LP — i.e., there is no way to simultaneously have shorter confidence intervals than LP and guaranteed coverage robustness.

Noise-to-Signal Ratio: The quantity T^{-1}||α(L)||² = trace{Var(T^{-1/2}α(L)ε_t) Var(ε_t)^{-1}}, which measures the total magnitude of the MA misspecification relative to the variance of the shocks. The paper bounds this at M²/T and uses M as the sufficient statistic for worst-case bias and coverage.

Bias-Aware Critical Value: An inflated critical value cv_{1-α}(b) solving r(b; cv_{1-α}(b)) = α, used to construct a VAR confidence interval CI_B(δ̂_h; M) that achieves correct asymptotic coverage by accounting for the worst-case bias M√(aVar(β̂_h)/aVar(δ̂_h) − 1). The paper shows this approach typically produces intervals at least as wide as LP for M ≥ 2.

Asymptotic Bias of VAR (aBias): The scaled bias term T^{ζ}E[δ̂_h − θ_{h,T}] converging to aBias(δ̂_h) = trace{S^{-1}Ψ_h H Σ_{ℓ=1}^∞ α_ℓ D H’(A’)^{ℓ-1}} − e’{i*,n} Σ{ℓ=1}^h A^{h-ℓ} H α_ℓ e_{j*,m}. This term is structurally absent from the LP asymptotics due to the double robustness mechanism.

Dynamic Concern for Misspecification

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1 — Overview

Research Question

This paper asks how an agent who fears that none of their probabilistic models is the correct description of the data-generating process (DGP) should update that fear as evidence accumulates, and what long-run behavior such an agent exhibits. The central contribution is making the concern for misspecification endogenous: the better the agent’s structured models explain past observations, the less concerned the agent becomes.

Decision Criterion

The agent posits a finite-dimensional parametric set of structured models Θ, holds a prior µ over Θ, and evaluates each action according to an average robust control criterion. This criterion takes a weighted average (over models) of robust control assessments, where each assessment penalizes expected utility for probability distributions that deviate from the structured model in terms of relative entropy, scaled by a misspecification concern parameter λ > 0. A standard subjective expected utility maximizer is the limiting case as λ → 0 (no concern), and a maxmin agent is approached as λ → ∞.

Endogenous Misspecification Concern

The concern parameter λ is updated each period as a function of the likelihood ratio test (LRT) statistic of the structured models against unstructured alternatives, scaled by a time-normalizing sequence βₜ: λ(hₜ) = LRT(hₜ, Θ) / (2βₜ). The sequence βₜ determines how demanding the agent is in evaluating model fit.

Taxonomy of Agent Types

Three types emerge based on the speed of βₜ:

Statistician type (βₜ = ct, linear): applies a time scaling that keeps the LRT asymptotically informative about the degree of misspecification. This is the unique type satisfying both safety (long-run average payoff at least ε-close to the maxmin guarantee, almost surely) and consistency under almost correct specification (no ε-regret when misspecification is small).
Lenient type (t = o(βₜ)): attributes unexplained evidence to sampling variability; corresponds to the Law of Large Numbers intuition.
Demanding type (βₜ = o(t)): overly penalizes small discrepancies, analogous to the Law of Small Numbers fallacy (Tversky and Kahneman, 1971).

Standard SEU maximization fails safety; robust control with an invariant λ (Hansen and Sargent, 2001; 2022) fails consistency under almost correct specification.

Long-Run Convergence Results (Theorem 1)

For a misspecified agent (no θ ∈ Θ with qθ_{a*} = p*_{a*}), the nature of the limit action a* depends on the agent type:

Lenient type: a* is a Berk-Nash equilibrium — an SEU best reply to beliefs supported on the models with minimum relative entropy from the true DGP.
Demanding type: a* is a maxmin equilibrium — a worst-case best reply to all models absolutely continuous with respect to the true DGP.
Statistician type: if behavior converges, a* is a c-robust equilibrium — a robust control best reply to beliefs on the relative entropy minimizers, with the concern for misspecification endogenously set at minθ R(p*{a*} || qθ{a*}) / c.

For a correctly specified agent (Proposition 2), every limit action is a self-confirming equilibrium, regardless of the agent type.

Cycles and Limit Frequency (Section 4, Theorem 2)

The statistician type’s behavior need not converge. In natural settings, the agent cycles between actions: playing a “safe” action whose consequences are well-explained by Θ reduces concern for misspecification, eventually leading to a riskier action whose poorly-explained consequences raise concern again, inducing a return to the safe action. The paper proves that every limit frequency (empirical distribution over actions) is a mixed c-robust equilibrium — a generalization that allows mixing while tying the concern for misspecification to the frequency-weighted average relative entropy of each action.

Empirical Applications

Monetary policy cycles (Sargent 1999, 2008): In a central bank model where the true DGP includes increased inflation variability under aggressive policy (a feature absent from the bank’s structured models), no pure c-robust equilibrium exists for small c. The model predicts persistent cycles between conservative and aggressive policy. The frequency of the conservative policy is increasing in the strength of the exploitable inflation-unemployment trade-off (θ₁π + θ₁a).
Labor supply under complex tax schedules (Rees-Jones and Taubinsky, 2020): Agents with a “schmeduling” heuristic (linearizing the tax schedule) are misspecified. Berk-Nash equilibrium predicts these agents exert excess effort, with the bias increasing in the complexity (convexity) of the tax code. The c-robust equilibrium attenuates this bias: conditional on the equilibrium, minθ R(p*_a || qθ_a) > 0, so agents maintain positive concern for misspecification and pull back from the biased recommendation. The paper rationalizes the empirical finding that approximately 40% of agents hold the schmeduling belief but only about 20% fewer agents act on it — consistent with endogenous concern reducing the behavioral impact of the biased model.

Axiomatization (Section 5)

The paper axiomatizes the static average robust control criterion (Theorem 3) using: a Variational Axiom (from Maccheroni, Marinacci, and Rustichini, 2006a), a Structured Savage axiom (Sure-Thing Principle for bets on the model identity), an Intramodel Sure-Thing Principle (STP for bets conditional on the model), and Uniform Misspecification Concern (the agent is equally concerned about misspecification regardless of which model is identified as best-fitting). Three additional dynamic axioms characterize preference evolution: Constant Preference Invariance (utility index stable over time), Dynamic Consistency over Models (Bayesian updating over structured models), and Q-Likelihood (misspecification concern increases in the LRT). A novel Asymptotic Frequentism axiom characterizes the statistician type: preferences must become arbitrarily similar (in a precise quantitative sense) after sufficiently long histories with the same outcome frequency.

In depth

Q1. What is the average robust control criterion and how does it generalize prior decision criteria?

A: An agent evaluates action a by averaging over structured models θ a robust control assessment: for each θ, minimize expected utility over probability distributions within relative entropy distance (penalized by 1/λ) of qθ_a, then integrate over θ with prior µ. This nests SEU (λ → 0, perfect trust in models), standard robust control of Hansen and Sargent (2001) (µ is Dirac, single benchmark model), and maxmin expected utility of Gilboa and Schmeidler (λ → ∞). The key extension is allowing µ to be nondegenerate, so the agent is simultaneously uncertain about the best-fitting model and about whether any model is exact.

Q2. What is the role of the likelihood ratio test statistic in driving misspecification concern?

A: The LRT statistic compares the maximum likelihood of the structured models against the best unstructured alternative. It diverges almost surely when the agent is misspecified, regardless of how close the structured models are to the true DGP. The concern parameter λ(hₜ) = LRT(hₜ, Θ) / (2βₜ) uses a time-scaling sequence βₜ to keep this statistic interpretable. Without scaling, a misspecified agent’s concern would always explode to infinity.

Q3. Why does linear time scaling (βₜ = ct) uniquely characterize the statistician type as rational?

A: Proposition 1 establishes two properties: (1) ε-safety — every βₜ = ct-optimal policy achieves average payoff at least ε below the maxmin guarantee, almost surely; (2) ε-consistency under almost correct specification — for DGPs sufficiently close to Θ, the agent avoids long-run regret. Part 2 of Proposition 1 shows that no βₜ with βₜ = o(t) or t = o(βₜ) satisfies both properties simultaneously. SEU fails safety; invariant-λ robust control fails consistency.

Q4. What is a c-robust equilibrium and how does it differ from a Berk-Nash equilibrium?

A: A Berk-Nash equilibrium (Esponda and Pouzo, 2016) requires the action to be an SEU best reply to beliefs supported on the relative entropy minimizers of the true DGP. A c-robust equilibrium requires the same support condition but with the best reply taken under the average robust control criterion, where the concern for misspecification λ equals minθ R(p*{a*} || qθ{a*}) / c — that is, the minimum relative entropy scaled by 1/c. The endogenous λ is positive whenever the agent is misspecified, so the agent does not fully trust even the best-fitting model.

Q5. How does the paper explain that misspecified lenient types converge to Berk-Nash while demanding types converge to maxmin?

A: For the lenient type (t = o(βₜ)), the time scaling makes the concern for misspecification converge to 0 (the LRT grows slower than βₜ relative to t), so the agent effectively behaves as an SEU maximizer with beliefs on the KL-minimizing models — the Berk-Nash condition. For the demanding type (βₜ = o(t)), the LRT diverges relative to βₜ, so λ → ∞ and the agent’s preferences converge to worst-case evaluation over all models absolutely continuous with the true DGP — the maxmin condition. These are Theorem 1, parts 1 and 2.

Q6. Why does the statistician type exhibit cycles rather than convergence?

A: Section 4 and Corollary 1 show in the monetary policy application that no pure c-robust equilibrium exists for small c. Intuitively, the conservative policy (a=0) is a best reply to a high misspecification concern, but it produces outcomes well-explained by Θ, which drives concern down. The aggressive policy (a=1) is a best reply to a low concern, but it generates increased inflation variability not captured in Θ, which drives concern up sharply. There is no fixed point that is self-sustaining, so the agent cycles. Theorem 2 shows that the empirical frequency of actions still converges to a mixed c-robust equilibrium.

Q7. What are the quantitative comparative statics for the monetary policy cycles?

A: Corollary 1 establishes that there exists a threshold c̄ > 0 such that for all c ≤ c̄: (1) no pure c-robust equilibrium exists; (2) a mixed c-robust equilibrium exists; and (3) in the maximal and minimal equilibria, the frequency of the conservative policy α*(0) is increasing in θ₁π + θ₁a — a larger exploitable trade-off between inflation and unemployment implies more time spent on the aggressive policy.

Q8. How does the model rationalize the Rees-Jones and Taubinsky (2020) labor supply finding?

A: Rees-Jones and Taubinsky (2020) find that approximately 40% of agents have incentive-compatible beliefs consistent with the schmeduling heuristic (linearizing a convex tax schedule), but approximately 20% fewer agents act according to that heuristic. In a Berk-Nash equilibrium, the schmeduling agent exerts excess effort relative to the optimum; the more convex the tax code, the larger the excess. In a c-robust equilibrium, the agent retains a positive misspecification concern proportional to the deviation between the convex tax schedule and the linear approximation. Higher effort levels are more exposed to uncertainty in the marginal rate (the misspecified term θ+ε multiplies a higher average income z), so the concern for misspecification provides a natural force that reduces effort below the Berk-Nash prediction. The paper notes this finding is also consistent with an alternative interpretation in Rees-Jones and Taubinsky where all agents hold schmeduling beliefs but under-respond behaviorally.

Q9. What is the mixed c-robust equilibrium and why does it always exist?

A: A mixed c-robust equilibrium is a mixed action α* ∈ Δ(A) such that beliefs ν are supported on the relative entropy minimizers Θ(α*) — computed as the parameter minimizing the α*-weighted average relative entropy across actions — and every action in the support of α* is a best reply under the average robust control criterion with λ = minθ Σ_a α*(a) R(p*_a || qθ_a) / c. Proposition 3 proves existence by mapping this fixed-point condition to a Nash equilibrium in an auxiliary game between the agent and two adversarial Nature players, then invoking Reny (1999) on that game. A pure c-robust equilibrium need not exist, but mixing over actions allows the concern for misspecification to be calibrated to the frequency of poorly-explained actions.

Q10. How does Theorem 2 formally connect cycles to mixed c-robust equilibria?

A: Theorem 2 states that if βₜ = ct for all t and α* is a βₜ-limit frequency (i.e., the empirical action distribution converges to α* with positive probability under some optimal policy), then α* is a mixed c-robust equilibrium. The intuition is that when α* places weight on both a well-explained action and a poorly-explained action, the time-averaged relative entropy stabilizes at a fixed level, producing a stable endogenous concern for misspecification that makes the agent asymptotically indifferent between the actions in the support — sharply reducing the incentive to break the cycle.

Q11. What does the axiomatization contribute beyond the learning results?

A: The axiomatization (Section 5, Theorem 3) provides behavioral foundations observable from choices, without assuming the internal LRT mechanism. Two primary axioms pin down the average robust control criterion within the variational class: Structured Savage (Sure-Thing Principle for bets over model identity) and Uniform Misspecification Concern (equal concern for misspecification regardless of which model is revealed as best-fitting). Dynamic Consistency over Models pins down Bayesian updating. Q-Likelihood axiomatizes that the concern for misspecification is ordinally increasing in the LRT. The novel Asymptotic Frequentism axiom (Axiom 9) pins down the quantitative speed of adjustment: long histories with the same empirical frequency must induce asymptotically similar preferences, and Proposition 5 shows this implies λ_{hₜ} / (LRT(hₜ, Q) / (2tₙ)) converges to a finite limit — exactly the statistician type’s linear scaling.

Q12. What is the correlation between behavioral biases that the model predicts?

A: The paper derives three novel empirical predictions about the cross-sectional and time-series correlation of uncertainty attitudes: (1) long-run uncertainty aversion positively correlates with initial misspecification and with belief in the Law of Small Numbers; (2) these correlations are causal — repeated model failures and overly demanding evaluation induce a shift toward cautious behavior; (3) even holding misspecification and probability reasoning fixed, limit uncertainty attitudes are stochastic, depending on whether the limit action’s outcomes are well-explained by the structured models.

Q13. How does Example 2 (Correlation Neglect) show that endogenous concern can amplify rather than attenuate biases?

A: In a double auction, a buyer who mistakenly treats their own valuation and the ask price as independent (Correlation Neglect, Esponda, 2008) bids below the optimum in Berk-Nash equilibrium. In a c-robust equilibrium, the positive correlation between valuations and prices produces a strictly positive minθ R(p*{a*} || qθ{a*}), so the agent maintains misspecification concern. Since lower bids are accepted with lower probability (and thus are less sensitive to model misspecification), the endogenous concern drives the agent to bid even lower — amplifying the bias rather than attenuating it. This example illustrates that the direction of the correction depends on the geometry of how the misspecification interacts with the payoff structure.

Key Concepts

Average Robust Control Criterion: The decision criterion proposed in the paper. An agent evaluates action a by taking the expectation over structured models θ (with prior µ) of min_{p_a ∈ Δ(Y)} [E_{p_a}[u(a,y)] + (1/λ) R(p_a || qθ_a)]. This is a weighted average of robust control assessments, each penalizing distributions that deviate from a structured model in relative entropy. The parameter λ > 0 governs the intensity of misspecification concern, with SEU as the limit at λ → 0 and maxmin at λ → ∞.

Endogenous Misspecification Concern: Unlike prior robust control models where λ is fixed or set externally, here λ(hₜ) = LRT(hₜ, Θ) / (2βₜ) is a function of how well the structured models explain the observed history hₜ via the likelihood ratio test statistic. The better the models explain past data, the smaller λ becomes and the less the agent hedges.

Statistician Type: An agent who scales the likelihood ratio test statistic with a linear time sequence βₜ = ct for some c > 0. This is the unique agent type satisfying both ε-safety (guaranteed long-run average payoff above the maxmin guarantee minus ε) and ε-consistency under almost correct specification (no long-run regret when misspecification is small). The statistician type’s linear scaling is the only one for which the LRT statistic retains asymptotic informativeness about the degree of misspecification.

c-Robust Equilibrium: A fixed-point concept for the long-run behavior of the statistician type. Action a* is a c-robust equilibrium if it is an average robust control best reply to beliefs supported on Θ(a*) = argmin_θ R(p*{a*} || qθ{a*}), with misspecification concern λ = minθ R(p*{a*} || qθ{a*}) / c. This generalizes Berk-Nash equilibrium by incorporating an endogenous hedging motive proportional to the minimum relative entropy between the true DGP and the best structured model.

Mixed c-Robust Equilibrium: A generalization of c-robust equilibrium to mixed actions α* ∈ Δ(A) for environments where no pure equilibrium exists. The beliefs are supported on the models minimizing the α*-weighted average relative entropy, and the misspecification concern is tied to that average entropy. Every βₜ-limit frequency is a mixed c-robust equilibrium (Theorem 2). This concept characterizes the long-run time-average behavior when the statistician type cycles.

Law of Small Numbers (LSN) Type / Demanding Type: An agent for whom βₜ = o(t), meaning the time scaling grows sub-linearly. This agent is excessively sensitive to early model failures (analogously to the Law of Small Numbers fallacy of Tversky and Kahneman, 1971, where short-run frequencies are treated as the long-run norm). The long-run behavior of such a type converges to maxmin behavior rather than robust control.

Asymptotic Frequentism (Axiom 9): A novel axiom requiring that conditional preferences after sufficiently long histories with the same empirical outcome frequency must be arbitrarily similar (in a quantitative sense defined by measuring rods x, y, E) to a limiting preference. This axiom axiomatically pins down the statistician type’s linear time scaling: it implies that the ratio λ_{hₜ} / (LRT(hₜ, Q) / (2t)) converges to a finite limit c, exactly characterizing βₜ = ct.

Berk-Nash Equilibrium: The equilibrium concept (Esponda and Pouzo, 2016) that describes the long-run behavior of lenient (SEU) agents learning under misspecification. An action a* is a Berk-Nash equilibrium if it is an SEU best reply to beliefs supported on Θ(a*) — the KL-minimizing models — without any additional hedging against misspecification. The current paper shows that lenient types converge to Berk-Nash equilibria, while statistician types converge to c-robust equilibria that differ by incorporating a positive misspecification concern.

Endogenous Production Networks Under Supply Chain Uncertainty

Mon, 01 Jan 0001 00:00:00 +0000

This paper studies how firms’ optimal technique choices under productivity uncertainty endogenously shape the structure of production networks and aggregate macroeconomic outcomes. Each sector chooses input shares (production techniques) before observing sector-specific TFP realizations. Techniques are selected to maximize a risk-adjusted expected log GDP measure — expected log GDP minus a risk-aversion-scaled variance term — with endogenous productivity shifters that favor balanced use of inputs. When uncertainty about sector TFP rises, firms shift toward suppliers with lower expected productivity but lower variance — a “flight to safety” in input sourcing. The key aggregation result is that the contribution of each sector to aggregate welfare depends on its endogenous Domar weight (expenditure share times adjustment factor), which itself responds to changes in beliefs. The paper establishes propositions characterizing how Domar weights respond to changes in mean (μ) and variance (Σ) of TFP beliefs: higher mean raises a sector’s Domar weight; higher variance lowers it when inputs are gross substitutes, but can lower it even with complementary inputs through belief adjustment. A basic calibration to 37 US BEA sectors (1948–2020) finds that the flexible-network economy has expected log GDP 2.1% higher than a fixed-network alternative. During the Great Recession, elevated uncertainty caused firms to shift toward safer, lower-productivity suppliers, reducing expected log GDP by 0.25% but reducing GDP variance by 2.4% and improving actual realized GDP outcomes by 2.7% relative to a “no uncertainty” benchmark.

In depth

Q1. What is the model’s core setup and how does technique choice generate an endogenous network?

Each of n sectors chooses input shares αi = (αi0, αi1, …, αin) — where αi0 is the labor share — before observing sectoral TFP realizations εt, subject to a convex cost function Ai(αi) that penalizes deviation from fixed-proportion baseline techniques; the equilibrium network α is then the solution to a social planner’s problem that maximizes expected welfare W = E[y] − (ρ/2)V[y], where y is log GDP and ρ is the coefficient of relative risk aversion.* Because cost functions Ai are jointly determined by the input shares chosen and by Hessian matrices Hi that govern substitutability/complementarity of inputs, sectors can substitute or complement in the production of any given good, and the equilibrium network balances expected log GDP gains from choosing more productive suppliers against the variance reduction from choosing safer suppliers.

Q2. What role do Domar weights play, and how do they generalize to the endogenous-network case?

In the standard fixed-network case (Hulten’s theorem), each sector i’s contribution to aggregate log GDP equals its Domar weight ωi = (expenditure on sector i’s output)/(total GDP) — a sufficient statistic for first-order productivity effects. The paper extends this: with an endogenous network, the social planner’s optimality conditions imply that equilibrium Domar weights equal the shadow value of relaxing each sector’s resource constraint, and these shadow values respond to changes in beliefs (μ, Σ) through the induced changes in α.* Lemmas 3–5 characterize these responses: a sector’s Domar weight increases in its expected log TFP (μi), and changes in variance Σij propagate through the network via the adjustment terms in the first-order conditions, so that a single sector’s volatility change affects the Domar weights of all connected sectors.

Q3. What are the key propositions about how beliefs affect aggregate welfare and GDP?

Proposition 6 (monotone welfare response to mean beliefs): welfare W is increasing in each sector’s mean log TFP μi, and the marginal effect equals the sector’s Domar weight; this holds even though expected log GDP E[y] may non-monotonically respond to μi when inputs are gross substitutes, because the variance-reduction benefit of adjusting away from the now-more-productive but higher-variance sector can temporarily dominate. Proposition 7 (variance increases hurt expected log GDP): for substitutable inputs, a rise in Σii decreases E[y] because firms shift away from the more volatile sector toward less productive alternatives; for complementary inputs, the same shift also reduces E[y] because complementary inputs move together. Corollary 4 shows that welfare W always falls when uncertainty rises, combining these effects.

Q4. How does the flight-to-safety mechanism work in a multi-sector economy?

When uncertainty about sector i’s productivity rises, the optimal technique response is to reduce αji for all downstream sectors j that use sector i as an input substitute, and increase labor shares or shares in less volatile inputs; since sectors with lower μ but lower Σ become relatively more attractive on a risk-adjusted basis, the network reconfigures toward “safer” but typically less productive suppliers. The cascading link-destruction example (Section 7) illustrates this: when an industry’s production becomes uncertain, the endogenous deletion of risky links propagates across the network as complementary and substitute linkages amplify or dampen the flight to safety, with the direction depending on whether inputs are gross complements or substitutes in the Hessian Hi.

Q5. What does the calibration to US data find about the quantitative importance of the endogenous network?

The calibrated model with 37 BEA sectors (1948–2020) achieves a cross-sectional correlation between model and data Domar weights of 0.96 (though the model average Domar weight of 0.03 is below the data’s 0.05) and matches the data correlations Corr(ωjt, μjt) = 0.1 and Corr(ωjt, Σjjt) = −0.4 closely (model delivers 0.1 and −0.3 respectively). Comparing the flexible-network baseline to a fixed-network alternative, expected log GDP is 2.1% lower in the fixed-network economy, and welfare differs by a similar 2.1%. This suggests the endogenous reallocation of input shares over the sample period — as some sectors became persistently more productive — delivered substantial gains relative to a static network.

Q6. What happens during high-uncertainty episodes such as the Great Recession?

During the Great Recession (2007–2009), the estimated uncertainty measure Σt spiked sharply; firms responded by shifting techniques toward safer suppliers, resulting in expected log GDP that is about 0.25% lower in the baseline than in a “no uncertainty” (Σ = 0) economy and GDP variance that is about 2.4% lower. The insurance paid off in terms of realized outcomes: realized log GDP in the baseline economy was approximately 2.7% higher than in the “as-if Σ = 0” economy in 2009, because firms had taken out insurance against exactly the kind of bad TFP draws that materialized during the crisis. The perfect-foresight economy (where εt is known before technique choice) outperforms the baseline by up to 3% in realized GDP during the Great Recession — the maximum value of uncertainty resolution.

Q7. What is the key distinction between the effects of mean and variance changes for welfare vs. expected GDP?

Changes in mean beliefs μi and welfare W are co-monotone (Proposition 6), but changes in μi and expected log GDP E[y] can be non-monotone when inputs are substitutes: a small increase in μi for a less productive sector can actually lower E[y] in the short run because firms shift toward that sector at the expense of more productive alternatives, even though this shift reduces variance and raises welfare. The divergence between E[y] and W is the key mechanism: when ρ > 0 (risk-averse households), reducing variance has positive welfare value even when it lowers the level of expected GDP, so the production network adjusts in directions that appear sub-optimal for average productivity but are optimal for welfare. The calibrated relative risk aversion parameter ρ̂ = 4.3 indicates meaningful risk aversion that makes these variance-mean trade-offs quantitatively relevant.

Q8. What is the role of input complementarity versus substitutability in determining network responses?

When inputs i and j are gross substitutes (Hessian element [Hi]ij < 0), an increase in sector j’s uncertainty Σjj induces sectors that use both i and j to shift away from j and toward i, reducing j’s Domar weight and increasing i’s — the network becomes more concentrated in safer inputs. When inputs are gross complements ([Hi]ij > 0), an increase in Σjj also reduces the demand for the complementary input i, because both inputs must be used together and the safe input i becomes jointly less attractive when paired with volatile j; this can cause both E[y] and V[y] to fall simultaneously, resulting in an ambiguous welfare effect that depends on the magnitude of ρ relative to the E[y]-V[y] trade-off (Corollary 4 ensures welfare falls, but the split across E[y] and V[y] depends on complementarity structure).

Key concepts

technique choice : a sector’s endogenous selection of input shares αij prior to observing TFP realizations; the key margin of adjustment in the model through which uncertainty shapes the production network; characterized by convex cost functions Ai that favor balanced input use around baseline shares α°.

endogenous Domar weight : the share of total expenditure on a sector’s output in aggregate nominal GDP, computed in the model’s equilibrium; equals the shadow value of the sector’s resource constraint and responds to changes in beliefs (μ, Σ); in the fixed-network case reduces to the standard Hulten-theorem Domar weight.

flight to safety : the equilibrium response in which firms shift their input shares away from high-mean, high-variance suppliers toward lower-mean, lower-variance alternatives when aggregate uncertainty rises; generates the prediction that network restructuring during recessions reduces GDP volatility while raising expected production costs.

risk-adjusted expected welfare (W) : the social planner’s objective, defined as E[y] − (ρ/2)V[y] where y is log GDP and ρ is the coefficient of relative risk aversion; this non-separable objective function generates the trade-off between expected productivity and risk reduction that drives endogenous network formation.

cascading link destruction : the propagation of reduced sectoral linkages through the network when one sector’s uncertainty rises; in examples with complementary inputs, the reduced demand for a volatile sector also reduces demand for its complements, potentially amplifying the flight to safety beyond the directly affected sector.

Firm Accommodation After Workplace Disability: Labor Market Impacts and Implications for Subsidy Design

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1 — Overview

Research Question

This paper studies (1) how firm accommodation decisions respond to financial incentives in the context of workplace disability under workers’ compensation, (2) what the causal effect of accommodation is on workers’ subsequent labor market outcomes, and (3) whether the equilibrium level of accommodation is socially efficient, and what the welfare implications of wage subsidies for accommodation are.

Empirical Context and Data

The analysis uses the universe of Oregon workers’ compensation claims from 2005 through 2017 — over 131,000 disabling claims — linked to longitudinal quarterly earnings records from the Oregon Employment Department. The setting exploits Oregon’s Employer at Injury Program (EAIP), which subsidizes employers who provide “transitional work” accommodations (primarily through wage subsidies) to workers with temporary workplace disabilities. EAIP accounts for roughly 25 percent of claims on average, with the wage subsidy component representing over 96 percent of EAIP expenses.

Identification Strategy

The authors exploit a policy change in July 2013 that reduced the EAIP wage subsidy rate from 50 percent to 45 percent. They construct a firm-level “exposure” measure — the fraction of a firm’s claims that used EAIP in a baseline period (2005–2009) — and estimate a continuous difference-in-differences specification in which the interaction of exposure and a post-2013 indicator instruments for accommodation. The identifying assumption is strong parallel trends: firms with low baseline exposure are unlikely to respond to the subsidy reduction, while high-exposure firms respond more, generating cross-firm variation in accommodation rates after 2013. An MTE framework (Heckman and Vytlacil 2005) is then used to explore heterogeneous treatment effects along an unobserved resistance-to-treatment dimension.

Main Empirical Findings

The subsidy reduction from 50% to 45% decreased accommodation rates by 2.9 percentage points (9.3 percent) for claims in firms with average exposure, implying a subsidy elasticity of accommodation of 0.9.
The policy change led to a 0.95 percentage point decrease in employment and a $120 decrease in quarterly earnings four quarters after disability for claims in average-exposure firms (roughly 1.3–1.5 percent declines relative to means), with no significant effect on worker turnover to other firms.
IV estimates of the effect of accommodation itself (using predicted EAIP as instrument) show accommodation increases the probability of employment four quarters after disability by 33 percentage points and increases quarterly earnings by approximately $4,100.
The MTE analysis reveals negative selection on gains: workers with workplace disabilities who are least likely to receive accommodation have the highest potential gains from it, driven largely by severe disabilities with high accommodation costs.
Descriptive and IV evidence is consistent with accommodation operating primarily as general human capital investment: accommodation has no statistically significant effect on the probability of moving to a new firm, and earnings gains are not systematically lower for workers who change employers after accommodation.

Structural Model and Counterfactual Findings

A two-period frictional labor market model with risk-averse workers, risk-neutral firms, Nash bargaining, imperfect experience rating in workers’ compensation, and firm accommodation as human capital investment is developed and estimated. Two inefficiency sources are identified: (1) a human capital externality — because accommodation builds general human capital, firms cannot capture the full surplus when workers separate, reducing accommodation incentives; and (2) a fiscal externality — imperfectly experience-rated firms do not fully internalize the workers’ compensation cost savings from accommodation, further depressing it below the efficient level. Counterfactual simulations show:

Eliminating wage subsidies (from 50% to 0%) reduces accommodation rates from 33% to 11%, leading to a 7% decline in post-disability employment and a 15% decline in post-disability quarterly wages (roughly $1,358).
A revenue-neutral reform eliminating wage subsidies reduces average welfare and the welfare of more than 90% of workers.
Welfare gains from the subsidy are larger for low-skilled workers than high-skilled workers.
Conditional on experiencing disability, eliminating wage subsidies decreases welfare by about 10%, while increasing the subsidy to 100% raises welfare for disabled workers by around 30%.
Firm profit is maximized at a subsidy rate around 80%, after which higher taxes offset accommodation gains.

In depth

Q1. What is the Employer at Injury Program (EAIP), and how does it differ from standard workers’ compensation?

A1: EAIP is an optional component of Oregon’s workers’ compensation system that subsidizes employers for the costs of accommodating workers with temporary disabilities during a transitional return-to-work period. Unlike standard workers’ compensation premiums (which are experience-rated at the firm level), EAIP is funded through a flat payroll tax on all firms that is not experience-rated — meaning firms that use EAIP do not pay higher premiums. The wage subsidy component accounts for over 96 percent of EAIP expenses; other reimbursable costs (worksite modifications up to $5,000, retraining up to $1,000, clothing up to $400) are rarely used. Eligible employers must be the employer at which the disability occurred, and accommodation is limited to a transitional period during which workers cannot simultaneously receive time-loss benefits.

Q2. How is firm-level “exposure” constructed, and what is the rationale for using it as an instrument?

A2: Exposure is the fraction of a firm’s workers’ compensation claims that used EAIP during a five-year baseline period from 2005 to 2009 — a separate historical period chosen to reduce volatility and avoid mean-reversion. The rationale draws on prior work (Aizawa et al., 2022) showing that firm fixed effects account for nearly 25 percent of variation in accommodation, far more than worker or disability characteristics (1 and 3 percent, respectively), suggesting permanent firm-level heterogeneity in the relative benefits and costs of accommodation. Firms with zero historical exposure are unlikely to change accommodation behavior in response to a subsidy reduction, while high-exposure firms respond more, creating differential quasi-experimental variation in accommodation rates after July 2013.

Q3. What are the first-stage and reduced-form results from the DID specification?

A3: The first-stage DID coefficient shows that a ten-percentage-point increase in exposure is associated with a one-percentage-point decrease in EAIP take-up after 2013, implying a 2.9 percentage point decrease for claims in firms with average exposure (mean 0.27). The corresponding reduced-form results show a 0.35 percentage point decrease in employment four quarters post-disability and a $45 decrease in quarterly earnings for every ten-percentage-point increase in exposure, scaling to 0.95 percentage points and $120 at average exposure. There is no statistically significant effect on the probability of moving to a new firm. Pre-trend tests show parallel accommodation trends across exposure terciles prior to 2013, supporting the identifying assumption.

Q4. What do the IV estimates imply about the causal effect of accommodation on labor market outcomes?

A4: Under the exclusion restriction that the subsidy change affects labor market outcomes only through accommodation, the IV estimates imply that receipt of accommodation increases the probability of employment four quarters after disability by 33 percentage points (against a mean of 72 percent) and increases quarterly earnings by approximately $4,100 (against a mean of $7,807). There is no significant effect on the probability of working at a new firm four quarters later. The authors note these large estimates reflect local average treatment effects for compliers — workers whose accommodation status was changed by the instrument — who disproportionately have high unobserved resistance to treatment and high accommodation returns, explaining the magnitude.

Q5. What does the MTE framework reveal about the distribution of accommodation effects and selection?

A5: The MTE curves show that workers with the highest unobserved resistance to treatment (least likely to receive accommodation) have the highest potential employment and earnings gains from accommodation. This negative selection on gains arises because these workers tend to have worse employment outcomes in the untreated state, consistent with more severe disabilities commanding higher accommodation costs. IV weights are concentrated at high-resistance values, explaining the large IV estimates. Negative selection on gains is also found along observable dimensions: workers in self-insured firms, healthcare support occupations, women, and those with wounds/cuts/burns show larger gains but lower likelihood of receiving accommodation.

Q6. What evidence supports characterizing firm accommodation as general rather than firm-specific human capital investment?

A6: Three pieces of evidence point toward general human capital. First, the IV estimate shows accommodation has no statistically significant effect on the probability of working at a new firm four quarters after disability. Second, a triple-interaction specification (DID interacted with new-firm indicator) yields suggestive evidence of even larger earnings gains for workers who move to a new firm post-accommodation, though this is not statistically significant — a pattern inconsistent with firm-specific human capital. Third, the subset of claims that receive non-wage EAIP benefits (worksite modifications, retraining) do show lower mobility, but this comprises fewer than 5 percent of the sample, meaning the predominant form of investment in the context is general in nature.

Q7. What are the two sources of market inefficiency in accommodation identified in the model?

A7: The first is a human capital externality operating through worker turnover. Because accommodation builds general human capital that workers carry to new employers, a firm accommodating a worker does not capture the portion of future surplus that accrues to future employers upon separation. In a Nash bargaining framework with lack of commitment, this dynamic inefficiency is larger when industry-wide turnover rates are higher — consistent with the descriptive finding that accommodation rates are strongly negatively associated with industry separation rates. The second is a fiscal externality from imperfect experience rating: firms whose workers’ compensation premiums are not fully linked to their own claim costs do not fully internalize the cost-savings from accommodation (i.e., reduced time-loss benefit payments), leading them to accommodate at inefficiently low rates.

Q8. How is heterogeneity incorporated in the structural estimation, and what do the estimated parameters show?

A8: The model incorporates observed heterogeneity (firm insurance status, worker skill type — measured by pre-disability wages — firm baseline exposure, and pre/post policy change) and unobserved heterogeneity mapped to the MTE framework’s unobserved resistance to treatment. Indirect inference matches cross-sectional accommodation rates, earnings by subgroup, and the DID coefficients. Key findings: net output during the disability period is negative (accommodation is a costly short-run investment), while post-disability output is higher for accommodated workers. Low-skilled workers experience larger productivity gains from accommodation than high-skilled workers. Accommodation cost shock variance is lower for higher unobserved types, meaning high-gain workers are also more sensitive to subsidy changes, consistent with the large IV estimates. The model fits the DID coefficients for accommodation, employment, and wages well.

Q9. What do the counterfactual simulations show about the welfare effects of varying the subsidy rate?

A9: Eliminating wage subsidies from the current 50% rate reduces the accommodation rate from 33% to 11% and lowers post-disability employment by 7 percentage points and post-disability quarterly wages by 15% ($1,358). From a welfare perspective, eliminating subsidies in a revenue-neutral reform reduces average ex-ante worker welfare and lowers welfare for more than 90% of workers. Conditional on experiencing disability, eliminating subsidies reduces welfare by about 10% while raising the subsidy to 100% increases welfare of disabled workers by around 30%. Firm profit is increasing in the subsidy rate up to about 80%, then decreases. Ex-ante worker welfare gains from the current 50% subsidy relative to no subsidy are modest in consumption-equivalent terms (at most 0.6% increase in consumption), partly because the disability probability is low (2.2%) and because unaccommodated workers still receive two-thirds wage replacement through time-loss benefits.

Q10. What distributional implications do wage subsidies have across worker and firm types?

A10: Welfare gains from higher wage subsidies are larger for low-skilled workers than high-skilled workers, so the subsidy has a redistributive dimension beyond efficiency correction. Welfare gains are also larger for workers in imperfectly experience-rated firms, where the fiscal externality creates the greater wedge from the efficient level. Self-insured firms, which already internalize workers’ compensation cost savings and thus accommodate closer to the optimal rate, benefit less from the subsidy and can even be made worse off if subsidies are set very high (since they bear higher flat payroll taxes with smaller marginal accommodation gains). The fraction of worker-firm matches experiencing welfare gains exceeds 90% under the benchmark subsidy level, indicating broad rather than narrowly concentrated gains.

Q11. How do the experience-rating channel and the worker-turnover channel interact in comparative statics?

A11: Model comparative statics show that reducing the job-to-job transition rate of workers with disabilities to one-quarter of its estimated value substantially raises accommodation rates, and this effect is more pronounced for imperfectly experience-rated firms than for self-insured firms. This occurs because self-insured firms already have a strong incentive to accommodate (to reduce workers’ compensation premiums), so turnover is less marginal for them. Forcing all firms to be self-insured (perfect experience rating) would substantially increase accommodation rates in currently imperfectly rated firms. Lowering the accommodation cost during the disability period (increasing net output during the disability period) also raises accommodation rates for both firm types.

Key Concepts

Firm Accommodation (EAIP): In this paper’s specific sense, accommodation refers to a firm’s decision to offer a worker with a temporary workplace disability “transitional work” — alternative tasks, modified duties, or flexible arrangements — during their recovery period, funded in part through Oregon’s Employer at Injury Program wage subsidy. Accommodation is distinct from simple early return to work; it functions as a form of human capital investment by potentially providing skill development opportunities and preventing human capital depreciation.

Exposure (Instrument): A firm-level continuous measure defined as the fraction of a firm’s workers’ compensation claims that used EAIP during a five-year baseline period (2005–2009). Exposure captures permanent, time-invariant firm-level propensity to accommodate, and is used to construct a difference-in-differences instrument for the causal effect of accommodation by interacting exposure with a post-2013 indicator (when the subsidy rate was cut from 50% to 45%).

Imperfect Experience Rating: The degree to which a firm’s workers’ compensation insurance premium adjusts to reflect that firm’s own claims costs, rather than being set at an industry average. Fully experience-rated (self-insured) firms internalize 100% of claim costs and thus have strong incentives to accommodate. Partially experience-rated firms face a fiscal externality: because their premiums do not fully reflect their own time-loss benefit expenditures, they do not capture all the cost savings from accommodating workers, leading to under-accommodation relative to the social optimum.

Human Capital Externality (Dynamic Inefficiency in Accommodation): The mechanism — analogous to Acemoglu and Pischke (1999) and Fang and Gavazza (2011) — by which worker turnover reduces firms’ incentives to invest in general human capital (here, accommodation). When accommodation raises workers’ general productivity, part of the future surplus from this investment accrues to future employers upon job-to-job separation. With Nash bargaining and lack of commitment (re-bargaining in the second period), the accommodating firm cannot capture this surplus, creating a dynamic inefficiency that is more severe in high-turnover industries.

Negative Selection on Gains: The empirical finding, established via the MTE framework, that workers with workplace disabilities who are least likely to receive accommodation (highest unobserved resistance to treatment) have the largest potential employment and earnings gains from accommodation. This pattern arises because workers with more severe disabilities have high accommodation costs (making firms unwilling to accommodate them) but also face far worse counterfactual labor market outcomes without accommodation, creating large potential gains.

Marginal Treatment Effect (MTE): Following Heckman and Vytlacil (2005), the treatment effect of accommodation evaluated at a specific quantile of unobserved resistance to treatment — defined here as the propensity score value at which a worker is indifferent between treatment and non-treatment. The MTE curve maps out the full distribution of treatment effects and reveals who benefits (and by how much), how IV estimates are weighted averages over this distribution, and which compliers drive the large IV estimates.

General vs. Firm-Specific Human Capital (in Accommodation Context): Accommodation is characterized as general human capital investment if the productivity and earnings gains it produces are transferable across employers — i.e., if accommodated workers who move to new firms retain their wage gains. It is firm-specific if gains are tied to the current match. In this paper, general human capital is supported by the null effect of accommodation on new-firm employment probability, suggestive evidence of non-lower (possibly larger) earnings gains for new-firm movers, and the observation that fewer than 5% of claims use non-wage EAIP benefits associated with firm-specific investment.

Revenue-Neutral Counterfactual: A counterfactual policy experiment in which the wage subsidy rate for accommodation is varied while imposing that both the time-loss benefit program and the EAIP wage subsidy program remain budget-balanced. Higher subsidy rates raise firm accommodation, reduce time-loss benefit payouts (lowering base premiums for imperfectly experience-rated firms), but require a higher flat EAIP payroll tax on all firms, some of which is passed through to workers via lower first-period wages.

Genetic Prediction and Adverse Selection

Mon, 01 Jan 0001 00:00:00 +0000

This paper asks how much adverse selection would arise in critical illness insurance (CII) markets if consumers can observe polygenic indexes (PGIs) — genetic risk scores derived from millions of genetic variants — while insurers are legally barred from using genetic information. The authors develop an econometric method that measures selection under current PGI technology, then extends identification to expected future PGI accuracy using heritability bounds, even though future PGIs are not yet observable in data.

The primary dataset is the UK Biobank (UKB), comprising approximately 446,570 genotyped individuals of European-like ancestry linked to NHS electronic health records. The authors study seven single-disease CII contracts (Alzheimer’s disease, breast cancer, coronary artery disease, colorectal cancer, prostate cancer, schizophrenia, and type 2 diabetes) and multiple-disease bundled contracts paying a lump sum upon onset. The econometric model assumes a probit disease probability, Gaussian PGI structure, and identification relies on published heritability estimates to pin down future PGI predictive power. The key selection metric is the implicit tax proposed by Hendren (2013): the percentage markup a marginal consumer must pay above her actuarially fair price due to adverse selection. The authors use the minimum implicit tax up to the 80th percentile of risk (t80) as their summary statistic, with market unraveling benchmarked at t80 between 43% and 83% from prior literature.

The paper reports three main findings, all scoped to a population of 35-year-olds in the standard insurer risk class (those whose predicted risk falls within 0.75–1.25 times the population mean).

First, under current PGI technology with full consumer adoption, selection is noticeable but heterogeneous across diseases. t80 ranges from 17.9% for coronary artery disease to 117.9% for Alzheimer’s disease. Coronary artery disease and colorectal cancer fall in the middle of the no-unraveling range; breast cancer, schizophrenia, and type 2 diabetes fall between the no-unraveling and unraveling ranges; Alzheimer’s disease and prostate cancer (t80 = 59.8%) reach or exceed the unraveling range. The current prostate cancer PGI explains 9.9% of liability variance, adding 8.3 percentage points over the 22.9% explained by non-genetic covariates.

Second, under expected future PGI accuracy — bounded below by SNP heritability and above by twin heritability — selection becomes potentially crippling. Under the lower bound (Scenario 3L), t80 ranges from 57.5% for breast cancer to above 1,000% for Alzheimer’s. Under the upper bound (Scenario 3U), t80 exceeds 100% for all seven single-disease contracts and exceeds 1,000% for three of them. For prostate cancer, the reference case, t80 reaches 86.8% under Scenario 3L and 426.9% under Scenario 3U — far above Hendren’s unraveling benchmarks. For multiple-disease male contracts, t80 = 30.8% under current technology, rising to 54.4% (Scenario 3L) and 243.9% (Scenario 3U).

Third, variation in selection across contracts is driven primarily by: the predictive power of the future PGI, the incremental predictive power over non-genetic covariates, and disease prevalence. Alzheimer’s and schizophrenia — high heritability, low prevalence — display the highest implicit taxes; breast and colorectal cancer — lower SNP heritability, lower incremental R2 — display the lowest.

These findings are corroborated by a calibrated Akerlof-Einav-Finkelstein equilibrium model using HRS data: current PGI availability reduces equilibrium market quantity from 30% to 21.4%; future PGI availability drives equilibrium quantity to zero in a full adverse selection death spiral. Partial take-up robustness checks show that even at 50% consumer adoption, selection remains problematically high under future PGI accuracy for most contracts. The analysis is restricted to individuals of European-like ancestry due to data availability constraints.

Q: What is the core market failure the paper analyzes? A: The paper analyzes adverse selection arising from an asymmetric information gap: consumers can observe PGI-based disease risk predictions from consumer genetic tests (e.g., 23andMe), while insurers in many jurisdictions are legally prohibited from requesting or using genetic information. This creates a situation where high-risk consumers have private information allowing them to sort into insurance, driving up average claims costs and potentially unraveling the market.

Q: What is a polygenic index (PGI) and why does it differ from classical genetic testing? A: A PGI is a weighted sum of millions of genetic variants (typically over one million) each with individually tiny effects, constructed using effect-size estimates from genome-wide association studies (GWASs). This contrasts with traditional genetic testing focused on rare single-gene mutations (e.g., BRCA for breast cancer or PKD for kidney disease), which are rare, explain small shares of population-level disease variance, and can largely be inferred from family history. PGIs target common polygenic diseases and are the primary driver of the adverse selection concern because they aggregate diffuse genetic signals into a meaningful risk prediction.

Q: What are the current PGI R2 values for the seven diseases studied? A: Estimated on the liability scale in the UKB, current PGI R2 values are: Alzheimer’s disease 7.1%, breast cancer 6.7%, coronary artery disease 2.5%, colorectal cancer 2.2%, prostate cancer 9.9%, schizophrenia 4.9%, and type 2 diabetes 7.4%. These represent the share of liability variance explained by each disease’s current PGI in the study sample.

Q: How does the paper identify the degree of selection under future PGI technology that does not yet exist in the data? A: The identification strategy combines three elements: the normality of PGI distributions, the relationship between current and future PGIs (the current PGI is modeled as a noisy version of the future PGI with an independent Gaussian error), and published heritability estimates that bound the future PGI’s predictive power. Theorem 1 establishes that under five stated assumptions — including a probit disease model and known future R2 from heritability studies — the full joint distribution of loss, current PGI, future PGI, and non-genetic covariates is identified from observed data.

Q: What heritability bounds are used for the future PGI scenarios, and why two bounds? A: Scenario 3L sets future PGI R2 equal to each disease’s SNP heritability (estimated from common genetic variants), which the authors treat as a conservative lower bound because future PGIs will also incorporate rarer variants with better effect-size precision. Scenario 3U sets future PGI R2 equal to twin heritability, treating it as an upper bound since the theoretical maximum predictive power of a PGI is the trait’s narrow-sense heritability. For prostate cancer, these bounds are 18.0% (SNP) and 57.0% (twin); for Alzheimer’s, SNP heritability is 33.1% and twin heritability is 58%.

Q: What is the implicit tax and how is it used as a benchmark? A: The implicit tax t(r) for a consumer with private risk r equals the percentage by which her insurance cost exceeds her own actuarially fair price when she must pool with all consumers of equal or higher risk. It measures how much the marginal buyer overpays due to adverse selection. The authors follow Hendren (2013) in reporting t80, the minimum implicit tax up to the 80th percentile. Hendren’s benchmarks: t80 between 7–35% for markets that did not unravel; t80 between 43–83% for markets that had unraveled.

Q: What are the single-disease contract results under current PGI technology (Scenario 2)? A: With full consumer adoption of current PGI technology, t80 ranges from 17.9% for coronary artery disease to 117.9% for Alzheimer’s disease. Coronary artery disease (17.9%) and colorectal cancer (26.5%) fall in the middle of Hendren’s no-unraveling range. Breast cancer (36.9%), schizophrenia (42.1%), and type 2 diabetes (37.0%) fall between the no-unraveling and unraveling ranges. Alzheimer’s disease (117.9%) and prostate cancer (59.8%) reach or exceed the unraveling range.

Q: What are the single-disease contract results under future PGI technology? A: Under the lower bound (Scenario 3L, R2 = SNP heritability), t80 ranges from 57.5% for breast cancer to above 1,000% for Alzheimer’s disease. Under the upper bound (Scenario 3U, R2 = twin heritability), t80 exceeds 100% for all seven contracts and exceeds 1,000% for three (Alzheimer’s, schizophrenia, and at least one other). These figures substantially exceed Hendren’s unraveled-market benchmarks for virtually all contracts.

Q: What drives cross-disease variation in the implicit tax? A: The authors identify three main drivers: the expected accuracy of future PGI (higher heritability → higher implicit tax), the incremental predictive power of the future PGI over non-genetic covariates observable by insurers (more incremental information → more adverse selection), and disease prevalence (lower prevalence concentrates risk heterogeneity, amplifying selection). Alzheimer’s disease and schizophrenia — high heritability and low prevalence — have the highest implicit taxes. Breast and colorectal cancers — lower SNP heritability and lower incremental R2 — have the lowest.

Q: What do the multiple-disease bundled contract results show? A: For the male multiple-disease contract under Scenario 2 (current PGI), t80 = 30.8%, comparable to Hendren’s no-unraveling range. Under Scenario 3L, t80 = 54.4%; under Scenario 3U, t80 = 243.9%, both in or above the unraveling range. The female contract yields qualitatively similar results. Implicit taxes in bundled contracts are generally lower than in single-disease contracts, suggesting some diversification of genetic risk across diseases.

Q: What does the calibrated equilibrium model find? A: Using an Akerlof (1970) / Einav-Finkelstein-Cullen (2010) supply-and-demand model calibrated to match a 30% market participation rate and a 50% loss ratio in the UK CII market, and using HRS data on individual risk aversion, the model finds that current PGI availability reduces equilibrium quantity from 30% to 21.4%. Future PGI availability (both Scenario 3L and 3U) drives equilibrium quantity to zero — a complete adverse selection death spiral with no trade.

Q: How robust are results to partial consumer adoption of genetic testing? A: At 10% consumer take-up, selection is low regardless of PGI accuracy. At 50% take-up, selection remains problematically high for all single-disease contracts under future PGI accuracy (Scenarios 3L and 3U). For multiple-disease contracts at 50% take-up, t80 falls just below Hendren’s unraveling threshold under Scenario 3L but enters the unraveling range under Scenario 3U. This suggests market problems would materialize once predictive power exceeds the SNP heritability bound and take-up exceeds roughly 50%.

Q: What role do risk preferences play, and do they confound the results? A: The authors test whether risk tolerance correlates with disease risk in the UKB using a self-reported general risk tolerance measure. They find extremely low correlations between risk tolerance and each disease. This is consistent with low correlation between relative risk aversion and disease risk in the HRS calibration, and supports the finding that correlation between risk and risk preferences is unlikely to meaningfully affect the main results.

Q: What is the paper’s assessment of preventive treatment as a mitigating factor? A: The authors acknowledge that genetic testing could enable personalized preventive medicine, which would reduce actual disease incidence among high-risk individuals. However, they argue this is unlikely to substantially affect their main findings because the most commonly covered diseases under CII are cancers, for which preventive behaviors have bounded effectiveness.

Q: What are the paper’s policy implications? A: The paper situates the genetic information problem within the standard regulatory framework for selection markets, distinguishing laissez-faire (allow genetic underwriting — efficient but potentially unfair to high-risk consumers), government provision (unattractive for non-essential CII), and managed competition (community rating combined with subsidies and risk adjustment). The authors argue that a full ban on genetic underwriting — the current policy in many countries — may become untenable as PGI accuracy improves, because it generates potentially crippling adverse selection. Some level of community rating may remain desirable for redistribution, but needs to be paired with subsidies or risk adjustment to prevent market collapse.

Q: What are the main data and scope limitations? A: The analysis is restricted to individuals of European-like ancestry because most large GWASs were conducted in European ancestry samples and PGIs perform poorly across ancestries. The UKB sample was aged 40–69 at recruitment and the analysis adjusts for age-dependent covariates; the HRS replication uses approximately 20,000 individuals. The equilibrium model ignores moral hazard and uses a parsimonious binary loss framework. The paper does not specify a timeline for when PGI accuracy will reach heritability bounds.

Polygenic Index (PGI): A weighted sum of an individual’s genetic variants across the genome (typically over one million variants), constructed using effect-size estimates from a genome-wide association study (GWAS) conducted in an independent sample. It is a noisy proxy for the individual’s true additive genetic factor for a disease, and its predictive power is bounded above by the trait’s narrow-sense heritability.

Implicit Tax: A measure of adverse selection defined by Hendren (2013) as the percentage by which a consumer with private risk r must overpay relative to her own actuarially fair price if she is pooled with all consumers of equal or higher risk. The minimum implicit tax up to the 80th percentile of risk (t80) serves as the paper’s primary summary statistic; t80 above roughly 43% is associated with market unraveling in prior literature.

SNP Heritability: The share of variance in a disease’s liability attributable to the set of common genetic variants (SNPs) used in heritability estimation. Used in this paper as a conservative lower bound on the predictive power of future PGIs, because future PGIs will additionally capture rarer variants.

Twin Heritability: An estimate of a trait’s narrow-sense (additive) heritability computed by comparing resemblance of monozygotic twins (sharing 100% of their genomes) to dizygotic twins (sharing ~50% on average). Used as an upper bound on future PGI predictive power, since heritability is the theoretical maximum R2 for a PGI.

Standard Risk Class: The set of consumers whose predicted disease risk (based on non-genetic covariates observable to insurers) falls between 0.75 and 1.25 times the population-wide average risk, following standard insurance underwriting practice. Insurers charge the same premium to all consumers in this class; any variation in risk within the class due to private genetic information constitutes the source of adverse selection analyzed in this paper.

Private Risk Function: The probability rho(g, w) of contracting the disease conditional on both the consumer’s observed PGI g and non-genetic factors w. Contrasted with the non-genetic private risk function pi(w), which conditions only on non-genetic covariates. The dispersion of the private risk distribution across consumers in the same risk class determines the degree of adverse selection.

Adverse Selection Death Spiral: The Akerlof (1970) mechanism in which high-risk consumers disproportionately purchase insurance, causing insurers to raise premiums, which deters low-risk consumers, which further raises the average risk of purchasers, ultimately driving equilibrium quantity to zero. The paper’s calibrated equilibrium model finds this outcome under future PGI accuracy for the HRS CAD contract.

Minimum Wages, Efficiency, and Welfare

Mon, 01 Jan 0001 00:00:00 +0000

Overview

Research question. Can minimum wages improve welfare through efficiency — by correcting monopsony-driven under-employment — and, if so, by how much? What is the optimal minimum wage, and how much of the welfare gain from a higher minimum wage comes from efficiency versus redistribution?

Model and methodology. The paper develops a tractable general equilibrium oligopsony model with heterogeneous workers (four types: non-high-school, high-school, college workers, and capital owners) and heterogeneous firms (varying in total factor productivity), embedded in a continuum of local labor markets where firms compete strategically in Cournot fashion. Firms face downward-sloping labor supply curves; their market power generates wages below the marginal revenue product of labor (markdowns). The model is calibrated to US data using the Census Longitudinal Business Database (LBD, 2014), the Bureau of Labor Statistics Current Population Survey (CPS, 2019), and the Survey of Consumer Finances (SCF). Key calibration targets include: average firm size of 22.83 workers (LBD), 29 percent of workers earning below $15/hr (CPS), labor and capital income shares, and household-level earnings and capital income ratios. The model is validated by quantitatively replicating four strands of empirical evidence: (i) reallocation effects of the German minimum wage introduction (Dustmann et al., 2021); (ii) employer spillover responses to Amazon’s voluntary $15 minimum wage (Derenoncourt et al., 2021); (iii) wage distribution compression evidence from Brazil (Engbom and Moser, 2021); and (iv) heterogeneous employment effects by market concentration (Azar et al., 2019).

Three channels for efficiency gains. The model identifies three mechanisms through which a minimum wage can improve efficiency under oligopsony: (1) a direct effect in which constrained firms with monopsony markdowns increase wages and expand employment toward the competitive level (Region II firms); (2) a spillover effect in which unconstrained competitor firms narrow their own markdowns in response to constrained firms’ increased wages and market shares; (3) a reallocation effect in which employment is shifted away from low-productivity firms (which enter Region III — constrained on labor demand) toward more productive firms.

Main findings on efficiency versus redistribution. Under the $15.12/hr minimum wage that maximizes social welfare under utilitarian weights (population-share weights), less than 5 percent of the welfare gains come from improved efficiency, while more than 95 percent come from redistribution. When the government is additionally given access to budget-neutral lump-sum transfers that fully address redistribution goals, the efficiency-maximizing minimum wage narrows to a range of approximately $7.50–$10.00 per hour, which is robust across social welfare weight specifications. The welfare gains attributable to efficiency alone are approximately 0.16–0.20 percent in consumption-equivalent terms, representing only about 1–2 percent of the welfare gains achievable in an economy with no labor market power at all (which would be 15.26 percent in consumption-equivalent terms under the same conditions with optimal transfers).

Why efficiency gains are small. Three structural reasons limit efficiency gains: (i) low-productivity firms — which are the firms most affected by a binding minimum wage in Region II — have endogenously narrow markdowns even absent a minimum wage, because they face more elastic labor supply and command small market shares; (ii) the calibrated production function has relatively flat marginal revenue product of labor schedules (decreasing returns parameter α = 0.940), so once firms enter Region III, employment rationing occurs rapidly; (iii) the large, high-productivity firms with the widest markdowns are not materially affected by the minimum wages of their small, low-wage competitors because those competitors have small market shares — making spillovers quantitatively negligible even though the model matches empirical cross-employer wage elasticities.

Optimal minimum wages under alternative frameworks. Without transfers and under utilitarian weights, the optimal minimum wage is $15.12. Without transfers but under Negishi weights (which rationalize the observed competitive equilibrium and load approximately 62 percent of weight on college workers and owners versus their 35 percent population share), the optimal is $6.97. Under a 97 percent weight on high-school graduates, the optimal rises to $18.32. With optimal lump-sum transfers, the optimal collapses to $7.76–$10.11 regardless of social welfare weights — a range robust across Frisch elasticity variants (ϕ ∈ {0.30, 0.62, 0.86}), regional decompositions (low, medium, and high income US states), short-run capital-fixed scenarios (where the optimum declines by approximately $1 under utilitarian weights), and the removal of household heterogeneity entirely (which yields an optimum of $7.74).

Distributional proxies versus welfare. Wage inequality (college–non-college log wage premium, cross-sectional variance of log wages) and the labor income share are monotonically improving as the minimum wage rises, even as welfare is hump-shaped and eventually declining. A rise in the minimum wage from $7.50 to $15 reduces the college–non-college log wage premium from 0.53 to 0.43 (roughly one-fifth), reduces the cross-sectional variance of log wages by nearly half, and raises the aggregate labor income share by approximately 3 percentage points — all while welfare (under utilitarian weights with no transfers) reaches its maximum at $15.12 and then declines. These standard proxies therefore do not reliably indicate welfare.

Scope conditions. All results are long-run steady-state comparisons unless otherwise noted. Results assume no price passthrough and a unit elasticity of substitution between capital and labor. The paper abstracts from capital–labor substitution responses and occupational choice. The redistribution channel quantified here is specific to the utilitarian welfare criterion and to the existing distribution of capital and profit income, in which owners (6 percent of households) earn 92 percent of dividends.

In depth

Q1. What are the three regions of firm behavior in response to a binding minimum wage, and what are their efficiency implications?

A: A firm can be in one of three regions. In Region I the minimum wage is not binding: the firm pays its optimal monopsony wage and employment is inelastically below the competitive level. In Region II the minimum wage binds and exceeds the firm’s optimal monopsony wage, but labor supply at the minimum wage still falls short of labor demand: employment and efficiency improve as the shadow markdown narrows. In Region III the minimum wage exceeds the competitive wage, so unconstrained labor supply would exceed demand: the firm rations employment and the rationing constraint binds, reducing efficiency. At the boundary of Region II and Region III, the shadow markdown equals one and the firm is at its efficient employment level. Only a firm-specific minimum wage targeting each firm’s competitive wage could deliver economy-wide efficiency.

Q2. How does the paper define and use “shadow wages” to characterize equilibrium?

A: The shadow wage for a firm is the effective wage that rationalizes equilibrium employment given rationing constraints. Formally, when a firm rations employment (Region III), households act as if facing a shadow wage equal to the actual minimum wage multiplied by a rationing factor p < 1 (the Lagrange multiplier on the rationing constraint, normalized as a fraction). Shadow wages aggregate across firms into market- and type-level shadow wages via CES aggregation. The key insight is that shadow wages, not observed wages, are allocative: aggregate labor supply for each worker type is determined by the type-level shadow wage, not by the minimum wage that firms actually pay. This allows the paper to express aggregate efficiency via two wedges — the aggregate shadow markdown (capturing average market power) and a misallocation term — without tracking all firm-specific constraints individually.

Q3. What are the two aggregate efficiency wedges and how do they behave as the minimum wage rises?

A: The two wedges are: (i) the aggregate shadow markdown µ̃, which is a productivity-weighted average of firm-level shadow markdowns and measures the extent to which aggregate wages fall short of marginal revenue products; and (ii) the misallocation term ω, which measures whether employment is allocated toward more productive firms and equals one when all shadow markdowns are identical. As the minimum wage rises from zero, µ̃ initially narrows (improving efficiency) because firms in Region II expand toward their competitive employment level and constrained firms’ market shares rise, tightening the residual labor supply of unconstrained competitors and narrowing their markdowns. But as the minimum wage rises further, Region III rationing causes shadow markdowns to widen rapidly — first for low-productivity firms and then progressively for more productive ones — so µ̃ turns back downward. The misallocation term ω first improves as low-productivity firms are pushed out, but then worsens because rationing at intermediate-productivity firms redirects employment from high- to medium-productivity firms.

Q4. What does the model validation exercise on the German minimum wage (DLSUB 2021) show?

A: The paper calibrates the model to the German context by setting a minimum wage of $8.95/hr equivalent to 48 percent of the pre-reform median wage — matching Germany’s 8.50 euro introduction in 2015, where 15 percent of workers earned below the threshold. The model produces employment effects that are slightly positive (consistent with empirical findings of no disemployment), average wage increases consistent with both constrained and unconstrained firms raising wages, a negative elasticity of the number of operating firms with respect to minimum wage exposure (correctly signed, moderately smaller than data), and a positive elasticity of average firm size with respect to exposure (slightly larger than the data). The reallocation direction — small unproductive firms shrinking and workers moving to larger, more productive firms — matches the data qualitatively and within the range of data estimates across specifications.

Q5. What does the Amazon spillover replication (DNWT 2021) show, and what does it imply about the minimum wage spillover channel?

A: Derenoncourt et al. (2021) estimate a cross-employer wage elasticity of 0.26: when Amazon raised wages by approximately 18.1 percent, competitors raised wages by 4.7 percent on average. The model replicates this by treating Amazon as the largest (or second-largest) firm in each market, exogenously narrowing its markdown by a fraction ζ calibrated to deliver an 18.1 percent wage increase. Competitors in the model raise wages through the strategic interaction mechanism: Amazon’s higher wage and market share tightens competitors’ residual supply curves, inducing them to narrow their own markdowns. The model matches the 0.26 cross-employer elasticity when Amazon is the largest firm in markets with at least 36 competitors, or the second-largest in markets with at least 12. Critically, the authors note that this empirical evidence concerns responses to a large firm raising wages; for minimum wages the question is whether large firms respond to their small wage competitors, which the model shows they do not substantially, because small firms have negligible market shares.

Q6. How does the paper separate efficiency from redistribution, and what is the key methodological innovation?

A: The paper gives the government access to budget-neutral, unrestricted lump-sum transfers across households in addition to the minimum wage. With transfers available, the government can use them to meet any redistributive objective encoded in arbitrary social welfare weights. Whatever is left for the minimum wage to do must be purely efficiency-improving. The paper shows (via aggregation theorems) that optimal lump-sum transfers can be computed in closed form for any social welfare weights, and that the social welfare maximizing allocation subject to transfers can be decentralized by transfers that sum to zero across households. Under this framework, the efficiency-maximizing minimum wage lies between $7.50 and $10.00 per hour regardless of whether utilitarian, Negishi, or 97 percent high-school-weighted social welfare functions are used — collapsing the original $0–$31 range to a tight interval.

Q7. How are Negishi weights computed, and why are they important for interpreting the results?

A: The Negishi weights are the social welfare weights under which a planner would choose the observed competitive equilibrium with zero lump-sum transfers. They are computed by inverting the planner’s first-order conditions: for the competitive equilibrium to be optimal under some set of weights, the implied consumption ratios must match observed data. The calibrated Negishi weights assign a combined weight of approximately 62 percent to college workers and owners, who constitute only 35 percent of the population. This means the competitive equilibrium is disproportionately aligned with higher-income households. A utilitarian planner, which weights households by population shares, therefore sees large scope for redistribution toward non-college workers — which is exactly why the utilitarian-optimal minimum wage is $15.12 and why 94 percent of its welfare gains come from redistribution rather than efficiency.

Q8. What are the quantitative welfare gains from the efficiency-maximizing minimum wage, and how small are they relative to the potential gains from eliminating monopsony?

A: With optimal lump-sum transfers, the welfare gains from the efficiency-maximizing minimum wage are approximately 0.16–0.20 percent in consumption-equivalent terms, robust across social welfare weight specifications, Frisch elasticity variations, and regional decompositions. The welfare gains associated with an economy in which all firms’ markdowns are set to one (no labor market power at all), also evaluated with optimal transfers, are 15.26 percent in consumption-equivalent terms. The efficiency-maximizing minimum wage therefore recovers approximately 1–2 percent of the potential welfare gains from eliminating monopsony. Equivalently, the efficiency gains correspond to roughly a 0.1 percent increase in TFP. These gains are small despite the model matching all empirical evidence on the channels through which efficiency gains could occur.

Q9. How do employment effects of minimum wages vary by market concentration, and why?

A: In concentrated markets (upper tercile of HHI), firms have larger monopsony markdowns, so a binding minimum wage pushes them into Region II — where employment expands — over a wider range of minimum wage values before entering Region III. This produces large, positive employment effects in concentrated markets. In less concentrated markets, firms already have narrow markdowns (they are closer to competitive), so even small minimum wage increases push them into Region III, where employment contracts. The model replicates the statistically significant positive effects in high-concentration markets and negative effects in low-concentration markets documented by Azar et al. (2019), for initial minimum wages below approximately $8/hr. At higher initial minimum wages, however, even high-concentration markets exhibit negative employment effects as more firms enter Region III.

Q10. What does the robustness exercise for Mississippi reveal?

A: Mississippi has the lowest per capita income in the US, and a $15 minimum wage would bind for 41.3 percent of its workers (versus 29.4 percent nationally). Despite this, the model finds that Mississippi would benefit from a $15 federal minimum wage under utilitarian weights, and the Mississippi-specific optimal minimum wage is $14.89 — nearly identical to the national optimum. The reason is an offsetting compositional effect: while Mississippi has lower average wages (pushing toward a lower optimal), it has a larger share of high-school graduates (63 percent versus 52.8 percent nationally) who prefer higher minimum wages (around $17 in the model). These two forces wash out, producing a stable optimal close to the national figure.

Q11. What happens to common empirical proxies for inequality and worker power as the minimum wage rises?

A: The college–non-college log wage premium declines from 0.53 to 0.43 (a fall of roughly one-fifth) as the minimum wage rises from $7.50 to $15. The cross-sectional variance of log wages falls by nearly half over this range, driven equally by declining within- and between-type inequality. The aggregate labor income share rises by approximately 3 percentage points, and the share of output created in non-high-school jobs paid to non-high-school workers rises by 7 percentage points. All of these proxies are monotonically improving in the minimum wage throughout, even as aggregate welfare under the model’s social welfare function is hump-shaped and declining past the optimum. The paper concludes that observations of declining inequality or a rising labor share are consistent with falling welfare, so these proxies cannot serve as reliable welfare indicators.

Q12. How does the short-run (fixed-capital) analysis differ from the long-run baseline?

A: In the short run, capital at each firm is fixed at the type-specific level chosen under a zero minimum wage. This creates sharper decreasing returns in labor (parameter γα rather than α̃), overhead costs that can make operation unprofitable, and a narrower range of minimum wages over which firms remain in Region II. The result is that firms in the short run enter Region III at lower minimum wages than in the long run, limiting the range of efficiency gains. Quantitatively, the efficiency-maximizing optimal minimum wage declines by approximately $1 under utilitarian weights (from about $10 to about $9 in the short-run exercise) and by only about $0.20 under Negishi weights. The robustness conclusion is that the difference between short- and long-run optimal minimum wages is modest, and the main finding that efficiency gains are small is preserved.

Key Concepts

Shadow wage (w̃ᵢⱼ): The effective wage that rationalizes a firm’s equilibrium employment in the presence of a minimum wage. When labor is rationed at firm ij (Region III), the shadow wage equals the actual minimum wage multiplied by a rationing factor pᵢⱼ < 1, where pᵢⱼ is derived from the Lagrange multiplier on the household’s rationing constraint. The shadow wage is allocative — it determines labor supply decisions — while the observed minimum wage wage is not. When the rationing constraint is slack (Regions I and II), the shadow wage coincides with the observed wage.

Shadow markdown (µ̃ᵢⱼ): The ratio of a firm’s shadow wage to its marginal revenue product of labor. In Region I (unconstrained), this equals the standard monopsony markdown. In Region II (constrained, on the labor supply curve), the shadow markdown narrows as the minimum wage increases, moving the firm toward its efficient employment level. In Region III (constrained, on the labor demand curve), the shadow markdown equals the rationing multiplier pᵢⱼ and widens, reflecting efficiency losses from rationing. An aggregate shadow markdown µ̃ is computed as a productivity-weighted average of firm-level shadow markdowns across all firms in the economy.

Misallocation wedge (ω): A productivity-weighted measure of how well employment is allocated across firms. In an efficient allocation with identical shadow markdowns, ω = 1. When high-productivity firms have wider markdowns than low-productivity firms (the baseline oligopsony outcome), ω < 1 because employment is directed away from productive firms. A minimum wage can improve ω by shrinking low-productivity firms but worsens it when high-productivity firms enter Region III and are over-rationed relative to medium-productivity firms.

Oligopsony with Cournot competition: The specific form of labor market power in this model. In each local labor market (defined as a NAICS 3-digit industry × commuting zone cell), a finite number of firms compete strategically in employment quantities, taking their competitors’ employment levels as given (Cournot assumption). Each firm has an upward-sloping labor supply curve derived from nested CES household preferences, and exercises a markdown on the marginal revenue product of labor. This differs from monopsony (one firm) or perfect competition (infinitely many firms), and generates both direct effects and spillover effects of minimum wages.

Negishi weights: The vector of social welfare weights under which the observed competitive equilibrium allocation would be the solution to a social planner’s problem with zero lump-sum transfers. In this model, the calibrated Negishi weights assign roughly 62 percent combined weight to college workers and owners (who constitute only 35 percent of the population), reflecting the fact that the market equilibrium allocates a disproportionate share of consumption to high-income households. The Negishi weights are used both to identify the gap between market outcomes and utilitarian objectives (motivating redistribution) and as one alternative normative benchmark.

Efficiency-maximizing minimum wage: The minimum wage that maximizes social welfare when the government additionally has access to budget-neutral lump-sum transfers across households. Because transfers can be optimized to handle any redistributive objective encoded in any arbitrary social welfare weights, the minimum wage under this framework serves solely to improve productive efficiency. In the calibrated model, the efficiency-maximizing minimum wage is approximately $7.50–$10.00 per hour, robust to social welfare weight specifications, Frisch elasticity variations (ϕ ∈ {0.30, 0.86}), and regional income differences.

Rationing constraint (n̄ᵢⱼₖ): A firm-specific, type-specific upper bound on the labor a household may supply to a firm in equilibrium. These constraints are taken as given by households and determined in equilibrium by firms’ labor demand decisions. When the minimum wage is above the firm’s competitive wage (Region III), the firm’s labor demand is less than what households would want to supply at that wage, so the rationing constraint binds. The binding rationing constraint generates the shadow wage discount (pᵢⱼ < 1) and is the mechanism by which high minimum wages reduce efficiency in the model.

Mussa Puzzle Redux

Mon, 01 Jan 0001 00:00:00 +0000

The Mussa (1986) puzzle is the empirical observation of a sharp, simultaneous increase in the volatility of both nominal and real exchange rates following the end of the Bretton Woods fixed exchange rate system in 1973 — a fact commonly interpreted as evidence for monetary non-neutrality. This paper resolves the puzzle by developing a model in which the dominant driver of nominal exchange rate fluctuations is a “financial shock” — a shock to the international demand for a country’s assets that is orthogonal to goods market fundamentals. Under a fixed rate, the central bank offsets financial shocks through reserve intervention, preventing them from moving the exchange rate; under a float, financial shocks freely move the nominal and real exchange rate simultaneously. The same framework also reconciles the Meese-Rogoff disconnect (exchange rates are unpredictable from macro fundamentals), the Backus-Smith puzzle, and the forward premium puzzle within a single unified model, with the financial shock accounting for the dominant share of exchange rate variance in each case.

In depth

Q1. What is the financial shock and how does it differ from standard macro shocks?

The financial shock is an orthogonal disturbance to international portfolio demand — the preference of foreign investors for holding domestic versus foreign assets — that is disconnected from productivity, monetary policy, and goods-market conditions. Because it is uncorrelated with macro fundamentals, it generates exchange rate movements without corresponding movements in output, prices, or interest rate differentials, producing the observed disconnect between exchange rates and macro variables.

Q2. Why does the Mussa pattern arise from regime switching?

Under a fixed rate, the central bank absorbs financial shocks via reserve intervention, sterilizing their exchange rate effects; the real exchange rate is equally insulated because the nominal rate is fixed and prices adjust slowly. Under a float, the same financial shocks freely move the nominal exchange rate, and with sticky prices this passes through to the real exchange rate. The variance of the real exchange rate therefore jumps discontinuously at the regime switch, matching the sharp Mussa empirical finding without requiring any change in the shock process.

Q3. How unified is the resolution across exchange rate puzzles?

A single model with the financial shock, sticky prices, and a standard asset pricing kernel simultaneously matches the Mussa pattern (regime-switching real volatility), the Meese-Rogoff disconnect (exchange rates unpredictable from fundamentals), the Backus-Smith puzzle (exchange rates and relative consumption uncorrelated), and the forward premium puzzle (high-interest-rate currencies appreciate). The financial shock accounts for the majority of exchange rate variance in each application.

Key concepts

Mussa puzzle : the discrete jump in real exchange rate volatility at the Bretton Woods breakdown (1973); resolved in this paper as the change in the central bank’s absorption of financial shocks between fixed and floating regimes.

financial shock : a disturbance to international portfolio demand orthogonal to goods-market fundamentals; the paper’s key mechanism for exchange rate disconnect, the Mussa pattern, and several other exchange rate puzzles.

Private Information and Price Regulation in the US Credit Card Market

Mon, 01 Jan 0001 00:00:00 +0000

The 2009 Credit Card Accountability Responsibility and Disclosure (CARD) Act barred credit card lenders from discretionarily raising borrowers’ interest rates in response to new information. Using the near-universe of US credit card account data (covering roughly 90% of outstanding general-purpose balances across 17–19 large and midsize issuers) together with a large panel of consumer credit reports, the paper documents that the class of rate increases restricted by the Act affected over 50% of borrowing accounts annually before the Act; incidence dropped to nearly zero afterward, and the interquartile range of interest rates on newly-mature accounts compressed immediately by one-third. The paper then estimates a structural model of the credit card market featuring differentiated lenders competing à la Bertrand, consumers with dynamic discrete choice over lenders and borrowing status, private information types identified from equilibrium pricing, and flexible correlation between demand and risk. Imposing the Act’s restrictions in the estimated model reveals that consumer surplus rises at all credit scores — by roughly $600 for subprime and over $1,000 for prime and superprime consumers — despite partial market unraveling among the deepest subprime accounts. The net surplus gains are driven by two forces: (1) a fall in lender markups (pre-CARD-Act median-risk subprime markups exceeded 40 percent), and (2) the insurance value of rate lock-in for borrowers whose credit risk deteriorates over time. Counterfactual analysis shows that if pre-CARD-Act markets had been perfectly competitive (zero markups), the Act would have induced complete market unraveling and consumer surplus would have fallen by $100–$600 per consumer.

In depth

Q1. What did the CARD Act restrict, and how large was the pre-Act practice it curtailed?

Before the CARD Act, lenders raised borrowers’ interest rates in response to new information about risk and demand on 48–54% of borrowing accounts at least once per year; the Act’s repricing restrictions drove this incidence to nearly zero and immediately compressed the interquartile range of interest rates across accounts within a credit-score tier from about 7.5 percentage points to about 5 percentage points — a one-third reduction in price dispersion. The pricing of emergent risk (risk revealed after account opening) was nearly indistinguishable from the pricing of origination risk before the Act (both approximately 30 basis points per 10-point FICO difference); after the Act, emergent risk was priced at only about 7 basis points per 10 FICO points — less than one-third as much — while origination risk remained priced at 26 basis points.

Q2. How are private information types identified and what do they reveal about adverse selection?

Private information types ψ are identified from the equilibrium pricing of mature accounts: the paper exploits portfolio-wide repricing events (where a lender raises rates on all accounts in a segment simultaneously) as quasi-experimental price variation to estimate price sensitivities γ via 2SLS, then inverts the lender first-order conditions to recover residual default-risk types orthogonal to observed credit scores. The Chiappori–Salanié bivariate probit test on simulated borrowing choices and default outcomes confirms adverse selection on new accounts: the estimated correlation of unobservables is 0.104, and within each credit-score group the correlation between private type and borrowing utility is at least 0.4 — meaning the highest-risk private types also have the strongest demand for credit.

Q3. What does the structural model say about markups and their source?

Estimated pre-CARD-Act markups exceed 40% for the median-risk subprime consumer; the high markups reflect a combination of switching costs (setup costs κ averaging $86.5–$89.7 per account for the lowest FICO groups), high borrowing utility δ among high-risk types (averaging 22.9 at FICO 580–599, falling to 3.3 at FICO 780–799), and lenders’ exploitation of private information about which borrowers have low price sensitivity. Setup costs are substantially larger than liquidity costs (costs of paying off existing balances), consistent with prior findings on adjustment costs in financial markets; new-account acquisition costs are increasing in credit score, consistent with lenders making larger offers to attract higher-quality borrowers.

Q4. What happens to prices under the CARD Act restrictions, and why does partial unraveling occur only for deep subprime?

Among deep subprime (FICO 580–599), the Act induces near-complete pooling at roughly 50% APR annualized for all private types; the safest private types exit the market as they are pooled with riskier peers, whose higher costs push prices further up — a textbook Akerlof unraveling spiral — and over 30% of the privately safest subprime borrowers face prices that newly exceed their willingness to pay. At higher credit scores (e.g., FICO 680–699), nearly all private types face lower prices because the mark-up compression dominates; at FICO 780+, all private types face lower or unchanged prices. Average traded prices fall at all credit score levels because: (1) consumers who exit were paying lower prices than those who stay, and (2) borrowers locked into favorable rates retain them as their types worsen.

Q5. What are the consumer surplus gains, and what drives them?

Consumer surplus rises at all credit scores — by approximately $600 for subprime consumers and by over $1,000 for prime and superprime consumers — with the gain coming from two sources of roughly equal magnitude: markup compression (a transfer from lender profits to consumer surplus) and the insurance value of rate lock-in for consumers whose default risk deteriorates. The insurance channel is most important for superprime borrowers, who are most likely to lock in favorable pricing and to experience type migration over the life of a relationship; the direct pecuniary markup gains dominate for subprime borrowers. Even absent any insurance value, the surplus gains from markup compression alone (a few hundred dollars) are comparable to prior reduced-form estimates of the Act’s price effects.

Q6. Why were high pre-CARD-Act markups necessary for the surplus gains?

In a counterfactual where pre-CARD-Act markets are perfectly competitive (marginal-cost pricing), imposing the CARD Act restrictions causes complete market unraveling: all prices exceed 150% APR, virtually no consumers borrow, and surplus per consumer falls by $100–$600 depending on credit score. With only modestly higher price sensitivity (one bootstrapped standard error larger than the point estimate), unraveling occurs at all credit score tiers and total surplus falls throughout the market. The intuition is that with zero markups, lenders have no cushion to absorb adverse selection costs; restricting risk-based repricing immediately makes lending unprofitable for any pooled price, triggering exit cascades that the pre-CARD-Act markup buffer prevented.

Q7. What is the adverse retention finding and its magnitude?

After the Act, lenders face adverse retention on mature accounts: for every 100 basis points by which emergent risk is priced below origination risk, the quarterly hazard of attrition from borrowing falls by 0.7 percentage points — meaning newly risky borrowers are less likely to leave while newly safe borrowers are more likely to leave. This is precisely the dynamics consistent with Akerlof adverse selection: the Act’s restriction on emergent-risk pricing reduces the signal lenders can use to retain safe borrowers selectively, degrading the quality of each lender’s continuing portfolio.

Q8. What private information rents existed before the Act?

Before the CARD Act, lenders exploited private information about borrowers’ demand characteristics: accounts that engaged in over-limit transactions or brief delinquencies (less than 30 days late) received median price increases of 6.9 and 15.5 percentage points annualized respectively, and 12-month revenue yields on these accounts rose by over 50% relative to baseline accounts, generating sustained higher returns despite higher charge-off risk. The revenue yield increase is not transitory; it persists for at least 12 months, and the returns figures confirm these behaviors reveal not just higher costs but also lower price sensitivity — enabling lenders to extract information rents.

Key concepts

private information type (ψ) : in the model’s notation, a consumer’s residual risk and demand characteristic that is known to the consumer and revealed to the incumbent lender over time through account behavior, but not observed by competing lenders or at the time of account opening; identified from equilibrium pricing using the assumption of price-invariant default.

emergent risk : default risk that becomes observable to the lender after account origination, in contrast to origination risk visible at account opening; the CARD Act restricted repricing in response to emergent risk while leaving origination-risk pricing unrestricted.

adverse retention : the post-CARD-Act phenomenon in which borrowers who become higher-risk are less likely to attrite (because their pricing is not raised) while borrowers who become lower-risk are more likely to leave (because their favorable pricing is no longer reinforced); the paper estimates a 0.7 percentage point drop in quarterly attrition hazard per 100 basis points of under-pricing of emergent risk.

insurance value of rate lock-in : the welfare benefit to consumers from knowing that an adverse draw of their risk type will not trigger a price increase; quantified by comparing actual surplus gains with gains in a counterfactual where types are perfectly persistent (eliminating the demand for insurance); roughly equal in magnitude to the direct markup compression gains.

market unraveling : the Akerlof-style exit spiral in which pooling raises prices, inducing exit by low-risk types, which raises the average cost of the remaining pool, which raises prices further; in the paper’s estimates this is severe only among deep subprime (FICO below 620) and would have been universal had pre-CARD-Act markups been zero.

Robust Real Rate Rules

Mon, 01 Jan 0001 00:00:00 +0000

The paper proposes and analyzes real rate rules — monetary policy rules of the form i_t = r_t + φπ_t (φ > 1), where r_t is the current-period real interest rate observed via TIPS yields or inflation swap markets. The central analytical result is that combining this rule with the Fisher equation i_t = r_t + E_t[π_{t+1}] immediately yields E_t[π_{t+1}] = φπ_t, whose unique non-explosive solution is π_t = 0 for all t. This proof uses only the Fisher equation — not the aggregate Euler equation — making the determinacy result robust to household heterogeneity, hand-to-mouth consumers, non-rational household or firm expectations, active fiscal policy, missing transversality conditions, and any specification of intertemporal or nominal-real links. The Fisher equation itself requires only two deep-pocketed, fully-informed, rational agents to arbitrage between nominal and real bonds — a much weaker assumption than aggregate Euler equation rationality. Under the real rate rule, inflation is decoupled from the Phillips curve: causation runs monetary policy → inflation, then inflation → output gap, not the reverse; the Phillips curve determines the output gap residually given already-determined inflation. In a three-equation New Keynesian model with a mark-up shock ζ_t and cost-push shock ω_t, the output gap satisfies x_t = −(ζ_t/(κ(φ − ρ_ζ))) − (ω_t/κ), where the Euler equation plays no role in inflation determination. The rule is globally stable under learning via a contraction argument using Gautschi’s inequality: even if financial market participants hold incorrect prior beliefs, the learning process converges to the target inflation. With a time-varying inflation target π*_t, the modified rule i_t = r_t + φ(π_t − π*_t) implements any target path determinately — π_t = π*_t for all t, including optimal Ramsey paths — making real rate rules observationally equivalent to any other monetary policy specification. The Taylor principle (φ_π > 1) is neither necessary nor sufficient for determinacy in richer models (Bilbiie 2008 TANK; Leeper-Leith 2016 FTPL); the real rate rule achieves determinacy without invoking Euler equation structure. An additional result: with long-maturity government debt, a stable inflation equilibrium always exists under the real rate rule regardless of whether fiscal policy is active or passive — the fiscal theory of the price level fails to produce unique outcomes in this setting.

In depth

Q1. What is the real rate rule, and why does it achieve determinacy without requiring the aggregate Euler equation?

The real rate rule i_t = r_t + φπ_t (φ > 1) combined with the Fisher equation i_t = r_t + E_t[π_{t+1}] immediately gives E_t[π_{t+1}] = φπ_t, whose unique non-explosive solution is π_t = 0 for all t; the proof is complete at this step, requiring no information about how households form expectations or optimize intertemporally. Standard Taylor-rule determinacy proofs rely on the aggregate Euler equation to close the system — the IS curve determines aggregate demand as a function of the real interest rate; deviation from determinacy arises when the Euler equation-Phillips curve system allows self-fulfilling expectation spirals. The real rate rule bypasses this entirely: the Fisher equation alone pins down the inflation path. The Fisher equation is a no-arbitrage condition between nominal and real bonds; it holds as long as two “deep-pocketed, fully-informed, rational agents” can trade both types of bonds — a condition that does not require aggregate household rationality, representative agent assumptions, or any specific consumption theory. Hand-to-mouth households, heterogeneous expectations, learning dynamics, and non-Ricardian fiscal regimes all leave the Fisher equation intact as long as some agents are pricing both asset classes. The consequence is that the Euler equation in the three-equation NK model becomes residual under the real rate rule: it determines the path of real interest rates given already-determined inflation and output gap, but plays no role in choosing among inflation equilibria.

Q2. What does the real rate rule imply about causation between inflation and the output gap?

Under the real rate rule, the Phillips curve operates in reverse relative to standard models: inflation is determined first (by the Fisher equation and the monetary rule), and the Phillips curve then determines the output gap as a residual; cost-push and demand shocks cannot amplify or dampen inflation variance under the rule. In the standard three-equation NK model with a mark-up shock ζ_t (law of motion ζ_t = ρ_ζ ζ_{t-1} + ε_{ζ,t}) and cost-push shock ω_t, the output gap under the real rate rule is x_t = −ζ_t/(κ(φ − ρ_ζ)) − ω_t/κ — a closed-form solution determined entirely by shocks, where the Euler equation does not appear. Inflation is π_t = 0 at all t (zero target): shocks affect the output gap but not inflation. Under an augmented rule that also responds to the output gap (i_t = r_t + φ_π π_t + φ_x x_t), determinacy still holds as long as a Phillips curve linking inflation and the output gap exists and the Taylor principle φ_π > 1 holds — providing additional policy degrees of freedom without sacrificing robustness. The decoupling of inflation from the Phillips curve is consistent with the empirical finding of Dotsey, Fujita, and Stark (2018) that the Phillips curve ceased to forecast inflation after 1984 — compatible with the hypothesis that the Fed’s post-Volcker behavior moved toward more real-rate-rule-like rules, giving the Fisher equation stronger anchor over inflation.

Q3. How does global stability under learning extend the determinacy result beyond local uniqueness?

Equilibrium determinacy is a local result (unique bounded solution near the target); the real rate rule additionally provides global stability under learning — even if financial market participants start with prior beliefs far from zero, the learning process converges to π_t = 0, preventing self-fulfilling sunspot equilibria from taking hold in the first place. The proof (Appendix D, using Gautschi’s inequality) establishes that the mapping from current beliefs to future beliefs is a contraction in the appropriate norm: since E_t[π_{t+1}] = φπ_t with φ > 1 drives realized inflation to zero, agents who update beliefs based on observed prices will progressively correct any initial error. This contrasts with Taylor rules, which are only locally determinate — an economy that starts at a non-zero sunspot inflation level may remain there if the sunspot is self-fulfilling. The global stability result also provides a response to the Cochrane (2022) critique that indeterminate equilibria under standard Taylor rules are “everywhere”: under the real rate rule, the only globally stable equilibrium is the target. The interest rate smoothing variant (Section 1.5) — fully smoothed real rate rule, θ > 0 — provides additional robustness: it requires agents to believe only that the central bank responds positively to inflation (not that φ > 1 specifically), and still generates identical inflation dynamics; this is more credible as a commitment device because the specific magnitude of φ cannot be directly observed.

Q4. How can the real rate rule implement arbitrary inflation dynamics, including optimal policy?

With a time-varying inflation target π_t, the modified rule i_t = r_t + φ(π_t − πt) implements any target inflation path determinately: the Fisher equation gives E_t[π{t+1} − π*_{t+1}] = φ(π_t − π*_t), whose unique solution is π_t = π*_t for all t, so realized inflation tracks the announced target exactly.** The CB must announce π*_t each period; this announcement may respond to the output gap, cost-push shocks, or any other variable. For example, to stabilize inflation while accommodating a cost-push shock, the CB sets π*t as a function of ω_t; realized inflation then follows this target, and the Phillips curve determines the output gap residually. There are two constraints: (1) the CB must be able to compute a reasonable approximation to E_t[π*{t+1}] — achievable via inflation futures, inflation swap markets, or an internal forecasting model; (2) the target path itself must not be explosive (a target that amplifies its own past realizations would generate explosive equilibria). Under these constraints, the paper formally proves (Appendix E.5) that real rate rules with time-varying targets can replicate the outcomes of any other monetary regime. This implies: (a) real rate rules can implement Ramsey-optimal policy, attaining the highest possible welfare; (b) it is empirically impossible to test whether a central bank is following a general real rate rule — any observed inflation and interest rate dynamics are consistent with some choice of π*_t. The Smets-Wouters (2007) estimated rule for the US illustrates: at the posterior mode, the correlation between the rule component z_t and the real interest rate r_t is 0.63, with both variables having standard deviation 0.46%, suggesting the Fed is already approximately two-thirds of the way toward a simple robust real rate rule.

Q5. Why does the Taylor principle fail in richer models, and how does the real rate rule avoid those failures?

The Taylor principle (φ_π > 1) is sufficient for determinacy in the benchmark three-equation NK model with a representative rational agent, but it is neither necessary nor sufficient in richer environments: Bilbiie (2008) shows that with enough hand-to-mouth consumers, higher φ_π can destabilize the economy; Leeper-Leith (2016) shows that following the Taylor principle can generate explosive inflation under the fiscal theory when nominal debt is present. Bilbiie (2008, 2019) inverts the Euler equation for the representative rational household when hand-to-mouth agents dominate: the aggregate consumption Euler equation has a negative intertemporal substitution sign, making the system’s eigenvalues switch. With enough hand-to-mouth agents, φ_π > 1 actually generates explosive equilibria (indeterminacy flips). Under the real rate rule, the Euler equation is disconnected from inflation determination entirely — Bilbiie’s mechanism cannot operate because the inflation equation relies only on the Fisher equation, not on whether the Euler equation has positive or negative sign. Similarly, the paper’s Section 2 result on fiscal robustness: with long-maturity government debt (Appendix B), a stable inflation equilibrium always exists under the real rate rule regardless of whether fiscal policy is active or passive. This implies the fiscal theory of the price level (FTPL) cannot uniquely determine inflation under the real rate rule — there is always a stable solution — so FTPL determinations are not unique, which may be of independent theoretical interest. The proof uses the contracting property of the non-linear real rate rule in the fully non-linear model, showing the target gross inflation Π* is always a solution of the bond-pricing fixed-point equation and that it is approached from all starting points via iteration.

Q6. How is the real rate rule implemented in practice, and what are the policy implications for central bank design?

Implementation uses TIPS yields (Treasury Inflation-Protected Securities) or inflation swap markets as real-time signals for r_t; the central bank sets i_t = TIPS_yield_t + φπ_t without estimating the natural rate (r) or output gap, eliminating the key measurement error in standard rules.* The key operational advantage over standard Taylor-type rules: standard rules require estimating the natural rate r* (now known to be mismeasured; Holston-Laubach-Williams 2017 revisions) and the output gap (subject to large real-time revisions); the real rate rule bypasses both because r_t is directly observable from financial markets (it equals the TIPS yield to a risk premium). The CB must also compute E_t[π*_{t+1}] to set the time-varying target; inflation futures or swap markets provide a forward-looking market price for this purpose. The paper discusses Hall and Reis (2016) “indexed payment on reserve” rules, which use a different mechanism (central bank liability indexation) to achieve similar robustness goals but do not rely on the Fisher equation as directly. Adão, Correia, and Teles (2011) achieve related results via complete nominal bond indexation. The real rate rule is more transparent and simpler to communicate: the CB says “we will raise the policy rate one-for-one with the real rate plus respond to inflation with coefficient φ.” For a smoothed version, communicating “we respond positively to inflation” — without specifying exactly how much — is sufficient for determinacy, and arguably more credible as a commitment. Section 4 (not covered here) develops a ZLB-adapted version of the rule for zero lower bound episodes that rules out explosive inflation equilibria at the bound.

Key concepts

real rate rule : the monetary policy rule i_t = r_t + φπ_t (φ > 1), where r_t is the current real interest rate observed from TIPS or inflation swap markets; achieves equilibrium determinacy via the Fisher equation alone, without invoking the aggregate Euler equation, making it robust to heterogeneous agents, hand-to-mouth consumers, non-rational expectations, and active fiscal policy.

Fisher equation : the no-arbitrage condition i_t = r_t + E_t[π_{t+1}] linking the nominal policy rate, real rate, and expected inflation; in the context of the real rate rule, it is the only structural equation needed for determinacy; requires only two deep-pocketed rational agents to arbitrage between nominal and real bonds — not aggregate household rationality.

inflation decoupling : the property under the real rate rule that the Phillips curve determines the output gap residually given already-determined inflation, rather than operating as a transmission mechanism for cost-push or demand shocks into inflation; implies that only monetary policy shocks and Fisher equation shocks can move inflation — cost-push and demand shocks affect the output gap but not the price level.

Taylor principle failure : the result (Bilbiie 2008) that standard Taylor rules can fail to deliver determinacy in models with hand-to-mouth consumers or heterogeneous agents — because the inverted aggregate Euler equation can flip eigenvalue signs — and (Leeper-Leith 2016) that following the Taylor principle can generate explosive inflation under the fiscal theory of the price level with nominal debt; the real rate rule avoids both failures by relying on the Fisher equation rather than the Euler equation for inflation determination.

global stability under learning : the property that even if financial market participants start with beliefs far from the inflation target, the learning process converges to the target under the real rate rule, proven via a contraction argument using Gautschi’s inequality; stronger than local determinacy (which only guarantees uniqueness near the target), ruling out self-fulfilling sunspot equilibria from any starting point.

fiscal theory robustness : the paper’s finding that with long-maturity government debt, the real rate rule always implies a stable inflation equilibrium regardless of whether fiscal policy is active (non-Ricardian) or passive (Ricardian); equivalently, the fiscal theory of the price level cannot uniquely determine inflation under the real rate rule because a stable solution always coexists with any fiscal regime.

Running Primary Deficits Forever in a Dynamically Efficient Economy: Feasibility and Optimality

Mon, 01 Jan 0001 00:00:00 +0000

Running Primary Deficits Forever in a Dynamically Efficient Economy: Feasibility and Optimality

Research Question

The paper addresses two questions about government debt rollover. First, a positive question: what is the maximum ratio of government bonds to capital that can be sustained forever without any primary budget surpluses? Second, a normative question: among sustainable bond-capital ratios along a balanced growth path, which one maximizes the welfare (steady-state utility) of consumers? The analysis is motivated by Blanchard’s (2019) AEA presidential address and the fiscal responses to the COVID-19 pandemic.

Setting and Mechanism

The baseline environment is a standard two-generation (young and old) overlapping-generations model. Young consumers earn labor income and save; old consumers live off portfolio returns. The production function is Cobb-Douglas, Yt = (GtN)^(1−α) K^α, where G = 1+g is the gross growth rate of labor-augmenting productivity. Uncertainty enters exclusively through a stochastic i.i.d. durability shock ε_t to the depreciation rate of capital (δ − ε_t), so the rate of return on capital r = αk^(α−1) − δ + ε is stochastic even though the capital stock per unit of effective labor k is deterministic along a balanced growth path. Consumers have Epstein-Zin-Weil utility with an intertemporal elasticity of substitution equal to one. Because IES = 1 and labor income is earned only when young, aggregate saving of young consumers is a constant fraction β of their wage income, making total assets (capital plus bonds) non-stochastic.

This structure creates a key wedge: the expected rate of return on capital R can exceed the growth rate g (dynamic efficiency) while the riskfree interest rate rf — determined by the portfolio equilibrium between risky capital and riskless bonds — can remain below g. In deterministic economies these two rates coincide, so dynamic efficiency and the infeasibility of permanent debt rollover always go together. In this stochastic model they can be decoupled.

Main Findings

Positive finding. The maximum sustainable bond-capital ratio, Bmax, is attained precisely when rf = g (equivalently, when the adjusted gross riskfree rate Rf = 1). Starting from a bond-less economy with rf < g (which may itself be dynamically efficient), introducing government bonds crowds out capital, raises the marginal product of capital and the constellation of returns, and drives rf upward toward g. Once rf = g is reached, any further increase in bonds would require rf > g, making rollover infeasible without primary surpluses. The maximum sustainable ratio Bmax is characterized as the unique root of f(Bmax, 1) = 0, and it is finite.

Normative finding. The welfare-maximizing sustainable bond-capital ratio equals Bmax. Proposition 6 establishes that u′(B) ≥ 0 for all B ∈ [0, Bmax] whenever Rf ≤ 1, with strict inequality unless Rf = 1. Proposition 7 therefore concludes that the welfare-maximizing B is the corner solution Bmax. Intuitively, increasing B reduces capital and wages but raises the rate of return on capital. When rf ≤ g, the welfare gain from a higher return on capital in old age dominates the welfare loss from a lower wage when young (via the factor-price frontier and the intertemporal optimality condition E{uo′(co)} ≥ uy′(cy)). When rf = g (at Bmax), a marginal increase in bonds also provides no additional welfare improvement if all seignorage is transferred to young consumers (ζ = 1), but still raises welfare if some seignorage is wasted (ζ < 1). In either case, Bmax is the optimum. Critically, at the optimum the economy is dynamically efficient — even though the government is running permanent primary deficits.

Dual role of bonds. At the optimal bond-capital ratio, government bonds serve two purposes simultaneously: (1) they crowd out any dynamically inefficient overaccumulation of capital that might prevail without bonds, and (2) they supply riskfree assets to risk-averse consumers who would otherwise hold only risky capital, improving risk sharing.

Quantitative Illustration

The paper calibrates a 30-year-period OLG model with α = 0.33, β = 0.353 (annual discount rate 2%), annual productivity growth g = 1% (G = 1.35), and target mean return on unlevered equity m = 3% per year. Risk aversion γ ∈ {1, 3, 8, 10} and annualized standard deviation of capital returns s ∈ {0.02, …, 0.22}. Key results (ζ = 0): at γ = 10 and s = 0.22, Bmax = 0.478 and B∗ (the bond-capital ratio needed just to eliminate dynamic inefficiency) = 0.083, so there is a wide interval [0.083, 0.478] of dynamically efficient, permanently rollable bond-capital ratios. For a capital-output ratio of 2, the debt-GDP ratio corresponding to Bmax = 0.478 is approximately 0.956. Bmax is strictly increasing in both γ and s, and is invariant to ζ (the share of seignorage transferred rather than wasted).

Scope Conditions

Results hold along balanced growth paths with constant g and constant rf; the sustainability characterization is more complex if either rate is stochastic.
The key sufficient condition for Rf to be increasing in B (Proposition 1) is that risk aversion γ < Λ, a model-dependent upper bound that is always positive. All subsequent propositions assume R′f(B) > 0, which is satisfied for a potentially larger set of γ.
The paper focuses on welfare along the balanced growth path; it does not study transition dynamics or welfare during convergence from an initial state.
The No Ponzi Game (NPG) condition is violated by design in the feasible-rollover region (rf ≤ g); the value of government bonds is positive even though the present value of all future primary surpluses is non-positive.

In depth

Q1. Why can an economy be both dynamically efficient and able to roll over government bonds forever, when this is impossible in deterministic models?

In a deterministic economy, the riskfree rate rf and the rate of return on capital r are equal, so the conditions rf < g (feasibility of rollover) and r < g (dynamic inefficiency) are identical. In a stochastic economy, aggregate uncertainty drives a wedge between rf and the expected return on capital. Risk-averse consumers require a premium to hold risky capital over riskless bonds, so rf < E{r}. It is therefore possible that E{ln R} > 0 (the Zilcha sufficient condition for dynamic efficiency holds) while Rf < 1, i.e., rf < g. This decoupling is the central theoretical contribution of the paper.

Q2. What is the formal criterion the paper uses for dynamic efficiency, and how does it relate to the AMSZ criterion?

Abel, Mankiw, Summers, and Zeckhauser (AMSZ, 1989) show that if the rate of return on capital exceeds g in all states (R > 1 always), the economy is dynamically efficient, and since rf < r, the economy has rf > g so rollover is infeasible; conversely if r < g always, the economy is dynamically inefficient. The AMSZ criteria are silent when R sometimes exceeds and sometimes falls short of one. Building on Zilcha (1991), the paper uses E{ln R} ≥ 0 as a sufficient condition for dynamic efficiency. In the five-region diagram (Figure 1), Region E satisfies E{ln R} > 0 (Zilcha-efficient) and Rf < 1 (rollover feasible simultaneously), which is the case of central interest.

Q3. How does the model achieve a deterministic capital stock despite stochastic capital returns?

The durability shock ε_t affects depreciation but is additively separable from the production function. Because (1) IES = 1 and (2) consumers earn income only when young, aggregate saving is the fixed fraction β of wage income, which depends only on capital k (itself non-stochastic). Total assets At+1 = Kt+1 + Bt+1 = St are thus non-stochastic. The stochastic shock to depreciation makes the rate of return on capital r = αkα−1 − δ + ε stochastic even though k is deterministic. Online Appendix B establishes that this model is isomorphic to a model with production function shocks, extending the scope of the results.

Q4. What is the financial market equilibrium condition that pins down the riskfree rate?

Young consumers optimally choose the portfolio share λ in riskfree bonds. The first-order condition for this portfolio problem along a balanced growth path is E{(λRf + (1−λ)R)^(−γ)(Rf − R)} = 0 (equation 20). In equilibrium, λ = B/(1+B) (the bond-capital ratio determines the portfolio share), so the equilibrium riskfree rate Rf satisfies the implicit equation f(B, Rf) = 0 (equation 21). Lemma 1 establishes that Rf = E{R^(1−γ)_a}/E{R^(−γ)_a}, a ratio-of-moments formula analogous to an Euler equation.

Q5. Why is the riskfree rate Rf an increasing function of the bond-capital ratio B, and what is the sufficient condition for this?

Lemma 2 shows ∂f/∂B > 0; intuitively, more bonds reduce capital, raise the marginal product of capital, and raise R, inducing consumers to demand more capital and less bonds, pushing Rf up to restore equilibrium. Lemma 3 provides a sufficient condition for ∂f/∂Rf < 0, namely γ < Λ (where Λ is a positive parameter-dependent bound). Under this condition, the implicit function theorem implies Rf′(B) > 0 (Proposition 1). The condition γ < Λ is sufficient but not necessary, so the results of all downstream propositions hold potentially for a wider parameter range.

Q6. What is the maximum sustainable bond-capital ratio Bmax, and how is it characterized?

By definition, a bond-capital ratio B is sustainable if and only if Rf(B) ≤ 1. If Rf(0) ≥ 1, then Bmax = 0 (no positive amount of bonds is sustainable). If Rf(0) < 1, Bmax is the unique positive root of Rf(B) = 1, i.e., f(Bmax, 1) = 0 (Proposition 4). At Bmax, the riskfree rate exactly equals the growth rate: rf = g. The paper also shows Bmax ≤ (1−α)β/α − 1, an upper bound that depends only on production and preference parameters. Notably, Bmax is invariant to the parameter ζ (the share of seignorage transferred to young consumers rather than wasted), because at Bmax transfers are always zero regardless of ζ.

Q7. Why does the welfare-maximizing sustainable bond-capital ratio equal Bmax rather than some interior value?

Proposition 6 shows that u′(B) ≥ 0 for all B ∈ [0, Bmax] whenever Rf ≤ 1, with strict inequality unless Rf = 1 and (1−ζ)B = 0. Since utility is weakly increasing throughout the feasible set, the optimum is the corner solution Bmax (Proposition 7). The mechanism: increasing B reduces k, lowering wages (bad for utility when young) but raising the marginal product of capital and hence the rates of return on capital and bonds (good for utility when old). The factor-price frontier ensures that the wage reduction equals the income gain accruing to initial capital, and the intertemporal optimality condition uy′(cy) = Rf E{uo′(co)} implies that when Rf ≤ 1 (so E{uo′(co)} ≥ uy′(cy)/Rf ≥ uy′(cy)), the welfare gain in old age dominates.

Q8. How does Proposition 5 square with the optimality of Bmax? Does reducing expected consumption not reduce welfare?

Proposition 5 shows that when ζ = 1, a marginal increase in B at Bmax reduces expected aggregate consumption (dE{c}/dB < 0). However, welfare is not simply expected aggregate consumption: it also depends on the distribution of consumption across states. At Bmax, even though expected consumption falls, the increased risk sharing from holding more riskfree bonds — which smooth consumption between the high-return and low-return states of capital depreciation — is large enough to leave welfare unchanged (u′(Bmax) = 0 when ζ = 1) or to increase it (u′(Bmax) > 0 when ζ < 1). This illustrates that in stochastic economies, the welfare criterion diverges from the aggregate consumption criterion that characterizes dynamic inefficiency in deterministic economies.

Q9. How does the paper’s welfare analysis relate to the No Ponzi Game (NPG) condition and the fiscal theory of the price level?

The standard NPG condition requires that the value of government debt equals the present value of future primary surpluses. In the paper’s feasible-rollover region (rf ≤ g), the NPG condition is violated by design: the present value of future primary surpluses is non-positive (all primary balances are deficits or zero), yet the market value of outstanding bonds is strictly positive. This is possible because, as Santos and Woodford (1997) show, when the present value of aggregate consumption is infinite, the NPG can fail. The market value of the capital stock remains finite (it is the value of profits on a depreciating capital stock approaching zero), but the bubble value of government bonds is positive.

Q10. What does the quantitative calibration reveal about the range of dynamically efficient, permanently rollable bond-capital ratios?

With α = 0.33, β = 0.353, g = 1% per year, G = 1.35, target mean equity return m = 3% per year, and risk aversion γ = 10 with annualized return standard deviation s = 0.22, the paper finds Bmax = 0.478 and B∗ = 0.083 (ζ = 0, Table 1). The interval [B∗, Bmax] = [0.083, 0.478] is the range of bond-capital ratios for which the economy is both dynamically efficient and able to roll over bonds permanently. For an economy with a capital-output ratio of 2, these bond-capital ratios correspond to debt-GDP ratios of up to 0.956. Both Bmax and B∗ are increasing in risk aversion γ and in the standard deviation of capital returns s; Bmax is independent of γ in any given column of the table for the ζ = 0 case (since R is independent of γ there), but rises substantially with γ in the ζ = 1 case.

The parameter ζ governs what the government does with seignorage revenue: transfer it to young consumers (ζ = 1) or waste it (ζ = 0), or some mix. Corollary 1 shows that Bmax is completely invariant to ζ, because at Bmax, rf = g so seignorage (g − rf)Bt = 0 in any case. The value ζ does affect u′(Bmax): if ζ < 1, u′(Bmax) > 0; if ζ = 1, u′(Bmax) = 0. Both configurations yield Bmax as the welfare-maximizing level. The parameter ζ matters for welfare levels and for B∗ (only in the ζ = 1 case, where transfers are positive and boost saving capacity), but not for the main positive or normative results.

Q12. In what sense is the model tractable, and what are its key limitations?

Tractability comes from three design choices: (i) the durability shock is additively separable from the production function, so labor income and aggregate saving are non-stochastic; (ii) IES = 1 with Epstein-Zin-Weil preferences, making saving a constant fraction of income; (iii) along balanced growth paths, g and rf are constant, so sustainability reduces to comparing two constants. Limitations acknowledged by the authors: the paper analyzes only balanced growth paths and does not characterize transition dynamics; the framework does not directly address economies where g or rf are stochastic; and the two-period OLG structure is stylized. The authors pose as an open question whether the result that optimal borrowing equals maximal borrowing generalizes to settings with random g.

Key Concepts

Bond-capital ratio (B): The ratio of outstanding government bonds to the capital stock, Bt/Kt. This is the paper’s central state variable and policy instrument. A value B is “sustainable” if the government can roll over its debt forever at the riskfree interest rate without any primary budget surpluses. The paper distinguishes B from the more commonly reported debt-GDP ratio (which equals B times the capital-output ratio).

Adjusted gross rate of return / riskfree rate (R, Rf): R ≡ (1+r)/G and Rf ≡ (1+rf)/G, where r is the net return on capital, rf is the riskfree interest rate on bonds, and G = 1+g is the gross growth rate. Expressing returns in these “adjusted” gross units scales out balanced growth and simplifies the sustainability condition to Rf ≤ 1 (equivalently, rf ≤ g).

Dynamic efficiency (Zilcha criterion): In the paper’s stochastic setting, the relevant criterion for dynamic efficiency is E{ln R} ≥ 0 (Zilcha 1991, as amended by Rangazas-Russell 2005 and Barbie-Kaul 2009), meaning the geometric mean of the adjusted gross return on capital is at least one. This differs from the deterministic condition r ≥ g. The paper’s Region E in Figure 1 is the key zone where E{ln R} > 0 (dynamically efficient) and Rf < 1 (rollover feasible) simultaneously.

Bmax (maximum sustainable bond-capital ratio): The largest value of B for which the bond-capital ratio is sustainable, defined as the unique root of Rf(B) = 1. At Bmax, the riskfree rate exactly equals the growth rate (rf = g). The paper proves Bmax is finite, invariant to ζ, and equals the welfare-maximizing sustainable bond-capital ratio.

B∗ (dynamic efficiency threshold): The bond-capital ratio at which the economy crosses from Zilcha-inefficiency into Zilcha-efficiency, defined by E{ln R} = 0. For B ∈ [B∗, Bmax], the economy is dynamically efficient and debt rollover is feasible. B∗ < Bmax when risk aversion γ or return volatility s is large enough, defining a non-trivial interval of dynamically efficient, permanently rollable bond levels.

Durability shock (ε): An i.i.d. random variable with mean zero that enters the capital depreciation rate as δ − ε_t. This shock makes the rate of return on capital r = αkα−1 − δ + ε stochastic while leaving the capital stock per unit of effective labor, aggregate wages, and aggregate saving non-stochastic. It is the only source of aggregate uncertainty in the model and is the mechanism that drives a wedge between rf and E{r}.

No Ponzi Game (NPG) condition: The condition that the present discounted value of government debt converges to zero (equivalently, debt equals the present value of future primary surpluses). Standard fiscal sustainability analyses assume this condition holds. The paper explicitly violates it: in the feasible-rollover region rf ≤ g, the present value of aggregate consumption is infinite and the NPG fails, yet government bond values are positive and debt rollover is sustainable.

Seignorage (ζ): The revenue the government obtains by issuing new bonds in excess of interest payments on existing bonds, equal to (g − rf)Bt when rf < g. The parameter ζ ∈ [0,1] governs the share transferred to young consumers (as lump-sum transfers τt) versus wasted (captured by the government but yielding no utility). A key finding is that Bmax is invariant to ζ, since seignorage is zero at rf = g regardless of ζ.

Rural Migrants and Urban Informality: Evidence From Brazil

Mon, 01 Jan 0001 00:00:00 +0000

Overview

Research Question. Does rural-urban migration increase or decrease urban informality, and through what mechanisms — and does the answer depend on the time horizon?

Setting and Data. The paper studies internal migration in Brazil over 2000–2010. The empirical analysis combines: (i) two waves of the Decennial Population Census (2000 and 2010) covering working-age adults (ages 15–64) across 3,548 Minimum Comparable Areas (MCAs); (ii) the universe of formal firms and workers from the matched employer-employee administrative dataset RAIS (1997–2018); (iii) the ECINF informal firm survey (2003); and (iv) the annual National Household Survey (PNAD, 2001–2009) for year-on-year short-run analysis in 700 identifiable municipalities. Internal immigration to the average urban destination was large: 17.6 percent overall over the decade, 7 percent for state-to-state migration.

Empirical Design. The authors use a shift-share instrumental variable (IV) design. The shares are pre-existing migration networks (migrant flows by origin-destination pair, 1995–2000). The shifts are drought shocks constructed from the Standardized Precipitation-Evapotranspiration Index (SPEI) interacted with agricultural crop calendars and the value share of each crop in each origin municipality — accumulated over the 2000–2010 decade. A second independent instrument uses international commodity price shocks as push factors (following a China-analogous construction); the two instruments are nearly uncorrelated across origins (0.007) and only weakly correlated across destinations (-0.3), providing an independent validation.

Long-Run Findings (decadal changes, 2000–2010). A one-percentage-point increase in the immigration rate (equal to 18.5 percent of a standard deviation):

Increases the share of workers in formal wage employment by 0.27 percentage points (a 1.2 percent increase from the mean of 23 percent).
Decreases the share in informal wage employment by 0.29 percentage points (a 2.9 percent decrease from the mean of 10 percent).
Has no effect on overall wage employment, unemployment, or self-employment — the formalization effect is a reallocation from informal to formal jobs, not net job creation.
Reduces formal sector wages by 0.6 percent, with no effect on informal wages.
Increases the number of formal establishments by 1.6 percent and the number of formal jobs by 2 percent.
Raises gross firm entry by 2.8 percent and gross firm exit by 3 percent (higher churn), with effects stable or slightly increasing through 2017–18.

These firm-creation effects are not driven by migrants starting businesses: migrants are not more likely to be business owners in high-immigration municipalities.

Short-Run Findings. Using year-on-year specifications with the PNAD (2001–2009), the authors replicate the results in the prior literature: municipalities receiving more migrants experience a reduction in formal wage employment, with no change in informal employment or non-employment — so the share of informal jobs rises. These short-run informality-increasing effects coexist with the long-run formalization results, and are not a sample artifact (the long-run results are unchanged when restricted to the same 700 PNAD municipalities).

Mechanism — Downward Nominal Wage Rigidity (DNWR). DNWR in the formal sector is the key mechanism reconciling short- and long-run effects. In Brazil, nominal wage cuts were illegal, and the national minimum wage rose regularly during the 2000s. Two municipality-level DNWR proxies are used: (i) the Kaitz index (national minimum wage / municipality median wage in 2000); (ii) the share of workers with negative year-on-year nominal wage changes (from RAIS, 1997–2000). In municipalities with higher DNWR: the positive formalization effects of immigration are smaller or fully muted; non-employment increases; and formal wages decline less. These cross-sectional patterns echo the Harris-Todaro-Fields prediction, and are consistent with DNWR being more binding in the short run (when nominal rigidities bind) than in the long run (when inflation and worker turnover allow real wage adjustment).

Model. The paper develops and estimates a dynamic model of firm dynamics and informality, extending the canonical Hopenhayn framework with (i) two margins of informality — the extensive margin (whether a firm registers) and the intensive margin (whether a registered formal firm hires workers formally) — and (ii) heterogeneous long-run productivity parameters (nu) that generate firm-specific life-cycle growth profiles. Formal firms cannot revert to informality; informal firms can formalize by paying the cost differential between formal and informal entry costs.

Counterfactuals. A simulated once-and-for-all 10 percent labor supply shock (approximately the 80th percentile of observed immigration shocks) produces: a 4.1 percent decline in the share of informal workers (IV: 7.5 percent); a 16.1 percent increase in formal firms (IV: 21.1 percent); and a 3.4 percent wage decline (IV: 5 percent). Of the increase in formal firms, 40 percent is accounted for by formalization of previously informal firms, highlighting the stepping-stone role of informality that a static or dual-economy model would miss. Average firm productivity declines by 1.4 percent due to worsening firm composition (the share of formal firms in the lowest productivity quartile rises by more than 4 percentage points). A counterfactual that nearly eliminates the extensive margin of informality (via steep enforcement costs) raises total output by 8.6 percent vs. 7 percent in the baseline shock, and increases average firm productivity by 2.1 percent vs. a decline of 1.4 percent — at the cost of displacing the least productive informal firms.

Scope Conditions. Results pertain to internal (not international) migration; drought-induced migrants do not change the skill composition of the labor force at destination, justifying a homogeneous worker assumption. The formalization effects hold for migrants and non-migrants separately, and for high- and low-skilled workers separately. The model is calibrated to the average urban destination in Brazil, not a spatial general equilibrium.

In depth

Q1. What is the identification strategy, and what are the key threats to validity the authors address?

The authors use a shift-share IV where shifts are drought shocks at origin municipalities (constructed from SPEI x crop calendar x crop revenue share, accumulated over 2000–2010) and shares are pre-2000 migration networks. Threats addressed: (i) pre-trends — no evidence of differential pre-trends in firm outcomes between 1997–98 and 1999–2000; (ii) demand channel — controlling for local drought shocks and distance-weighted neighboring shocks leaves results unchanged; (iii) capital reallocation — adding a bank-network-based shift-share control (following prior literature) does not change results; (iv) agricultural processing linkages — results hold after excluding agricultural firms and food/beverage/tobacco manufacturers; (v) migration persistence — controlling for baseline log population and 1995–2000 migration rates leaves results unchanged. The commodity-price-shock instrument provides an independent validation, yielding similar results despite near-zero cross-origin correlation with drought shocks and only -0.3 correlation across destinations.

Q2. How do the authors reconcile the long-run formalization result with the short-run informality-increasing result, and what role does DNWR play?

DNWR is the key mechanism. Nominal wage cuts are illegal in Brazil’s formal sector, and the minimum wage rose through the 2000s, making DNWR binding especially in the short run. In the year-on-year specification (PNAD, 2001–2009), immigration reduces formal wage employment with no change in informal employment, raising the informal share — consistent with prior literature. Over the decade, inflation and worker turnover permit real formal wage adjustment, enabling formal sector expansion. Cross-sectional heterogeneity confirms this: in municipalities with above-median Kaitz index or below-median share of negative wage changes, the formalization effect of immigration is smaller or zero, and non-employment rises — precisely the Harris-Todaro-Fields prediction for rigid-wage environments.

Q3. What is the exact magnitude of the firm-level effects and how persistent are they?

A one-percentage-point increase in the immigration rate increases formal establishments by 1.6 percent, formal jobs by 2 percent, firm entry by 2.8 percent, and firm exit by 3 percent — all decadal effects (1999–2000 to 2011–12). Effects on firms, entry, exit, and jobs remain stable or slightly increasing through 2017–18 as estimated using RAIS panel data, with no evidence of pre-trends (effects near zero in 1997–98 to 1999–2000 period). The effect on firm-level average wages is negative (consistent with the worker-level wage effect) but not statistically significant.

Q4. Are migrants themselves the source of new formal firm creation?

No. The authors directly test and reject this channel. Migrants are not more likely to be business owners — either of small firms (fewer than 5 employees) or larger firms (6 or more employees) — in municipalities that receive more immigration. The increase in formal firm entry is driven by non-migrants responding to cheaper labor.

Q5. What are the two margins of informality in the model, and why does the intensive margin matter for the migration-formality nexus?

The extensive margin is whether a firm registers formally (firm-level binary). The intensive margin is whether a formally registered firm hires workers without formal labor contracts (worker-level, within formal firms). The intensive margin is crucial because it links formal firms to migrants: newly arrived migrants may take informal jobs within formal firms, allowing formal firm creation to respond to the immigration shock even before the labor market fully formalizes. In the transition dynamics after an immigration shock with DNWR, new formal firms tend to be small and lower-productivity, and hire a substantial fraction of their workforce informally — so labor informality hovers near its initial level for several years even as firm informality declines quickly.

Q6. What fraction of the increase in formal firms in the counterfactual comes from stepping-stone formalization versus new formal entry?

In the baseline 10 percent labor supply counterfactual, approximately 40 percent of the increase in the number of formal firms comes from formalization of previously informal firms across their life cycles. The remaining 60 percent comes from new formal firm creation. A static framework would miss the stepping-stone channel entirely and substantially underestimate total formalization.

Q7. How does the model’s calibration pin down the cost structure of informal vs. formal firms?

The model is calibrated using a two-step minimum distance procedure. First-step parameters include the persistence of formal firms’ productivity process (estimated from RAIS: rho_f = 0.92), and statutory tax rates (payroll tax tau_w = 0.375; revenue VAT tau_y = 0.293). Second-step parameters (12 total, including entry costs, exogenous death rates, productivity dispersion, and cost-function curvatures for both margins of informality) are estimated by minimizing the distance between simulated and observed moments from RAIS (2003 cross-section for static moments; 2000–2011 panel for growth moments) and ECINF (informal firms with up to 5 employees, 2003). Key calibrated values: formal entry costs are more than twice informal entry costs and correspond to over 30 times the 2003 monthly national minimum wage; the informal sector exogenous death rate (delta_i = 0.148) is more than twice the formal rate; productivity variance and persistence are similar across sectors.

Q8. What happens to firm productivity and output per worker in the long-run counterfactual?

Average firm productivity declines by 1.4 percent despite lower informality. The composition of formal firms worsens: the share of firms in the lowest productivity quartile rises by more than 4 percentage points, while the share in the top quartile falls by about 3 percentage points. Total output and tax revenues increase (7 and 8.6 percent, respectively), but both decline in per capita terms. The authors note these are likely lower bounds because the model assumes no technological differences between formal and informal sectors and no differential capital access.

Q9. What does the enforcement counterfactual reveal about the dual role of informality?

When the extensive margin of informality is nearly shut down (by making the informal cost function very steep), a 10 percent labor supply shock produces: output increase of 8.6 percent (vs. 7 percent with informality present); average firm productivity increase of 2.1 percent (vs. decline of 1.4 percent); much higher tax revenues due to greater formality. However, this comes at the cost of a sizable reduction in total firm count as the least productive informal firms are displaced. This illustrates the dual role: in the short run, the informal sector acts as an employment buffer and stepping-stone, which is more important when formal wage rigidity is stronger; but in the long run, it dampens aggregate economic benefits from immigration by sheltering low-productivity firms.

Q10. Do the results hold for both migrants and non-migrants, and across skill levels?

Yes. Appendix results show similar employment and wage effects for migrants and non-migrants separately, though formal wage declines are more pronounced for non-migrants. Results are also similar for high- and low-skilled workers — which the authors attribute to the fact that drought-induced migration does not change the skill composition of the workforce at destination (confirmed empirically). Price-shock-induced migrants differ: they are more likely to be young and male, and do change workforce composition, providing a different set of compliers that strengthens external validity.

Q11. How does the paper relate to the “startup deficit” literature on demographic decline?

The paper’s findings are the mirror image of the US startup deficit literature, which argues that demographic slowdown reduced firm entry, labor reallocation, and employment growth. The magnitudes are comparable in scale: the US startup deficit corresponds to a 5-percentage-point decline in firm entry between 1980 and 2012, while the rural-urban migration shocks studied here produce first-order effects on firm entry of similar or larger magnitude (2.8 percent per percentage point of immigration rate), suggesting labor supply growth is a primary driver of formal firm dynamics in both directions.

Key Concepts

Downward Nominal Wage Rigidity (DNWR). In the paper’s usage, the binding constraint that formal sector wages cannot be cut in nominal terms — in Brazil, both legal prohibition of nominal wage cuts and a rising national minimum wage. DNWR is the paper’s central mechanism explaining why immigration increases informality in the short run (wages cannot adjust) but reduces it over the decade (inflation and turnover permit real adjustment). Measured empirically via the municipality-level Kaitz index (national minimum wage / local median wage) and via the share of workers with negative year-on-year nominal wage changes in RAIS.

Extensive Margin of Informality. Whether a firm is registered with the government (formal) or not (informal). In the model, informal firms can avoid taxes but face a size-increasing cost of informality and the option to formalize by paying the difference in entry costs. This margin captures the firm’s legal registration status.

Intensive Margin of Informality. Whether a formally registered firm hires individual workers with or without formal labor contracts (signed work booklet, carteira de trabalho). Formal firms face increasing costs for informal hiring but exploit this margin for lower-cost labor, especially when small or young. This margin is critical because it links formal firms to migration-induced informal labor supply and allows formal firms to absorb migrants before full wage adjustment occurs.

Stepping-Stone Role of Informality. The paper’s term for the dynamic channel through which the informal sector facilitates transitions to formality for both firms and workers. Informal firms accumulate productivity experience and formalize when productivity crosses the formalization threshold; informal workers within formal firms transition to formal contracts as firms grow. In the counterfactuals, 40 percent of the increase in formal firms following a labor supply shock is attributable to this channel. The stepping-stone role is most valuable during the short-run period of formal wage rigidity.

Shift-Share Instrumental Variable. The identification design combining pre-existing migration network shares (fraction of prior migrants to destination d from each origin o, computed 1995–2000) with exogenous push shocks at origin (drought shocks or commodity price shocks). The instrument predicts which destination municipalities receive more migrants based purely on exogenous origin-level shocks, purging the endogeneity from migrants self-selecting into prosperous cities.

Minimum Comparable Area (MCA). The paper’s geographic unit of analysis: a harmonized aggregation of Brazilian municipalities whose administrative borders changed during the study period, yielding 3,548 stable units covering all urban destinations studied. The authors call these “municipalities” for convenience.

Harris-Todaro-Fields Framework. The theoretical benchmark against which the paper’s results are compared — the view (from Harris and Todaro 1970 and Fields) that rural-urban migration increases urban unemployment or informality because DNWR prevents the formal sector from absorbing migrants, who instead queue for formal jobs or enter the informal sector. The paper shows this prediction holds in the short run and in high-DNWR municipalities, but not in the long run where real wage adjustment occurs.

Search Frictions and Product Design in the Municipal Bond Market

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1 — Overview

Research Question

This paper investigates whether intermediaries in the U.S. municipal bond market strategically exploit product design to increase search frictions and, through that channel, capture rents. Specifically, it asks: do underwriters who negotiate bond design with local governments have an incentive to add nonstandard provisions that raise their own competitive advantage in subsequent secondary-market intermediation, even at the expense of issuing governments and their taxpayers?

Setting and Data

The study focuses on tax-exempt general obligation and revenue bonds issued via negotiated sales by local governments (counties, cities, school districts, and other special-purpose governments) from 2010 to 2013, tracking all secondary-market transactions through 2014. The final sample comprises 13,118 bond issues with a total face value of $266.9 billion. Bond attribute data come from Mergent; transaction data come from the Municipal Securities Rulemaking Board (MSRB). Issuer financial health, demographics, and economic conditions are drawn from the Census and American Community Survey; state revolving-door regulations are compiled from the National Conference of State Legislatures database. Structural estimation uses a subsample of 927 bonds concentrated in the five states that enacted revolving-door regulations during the study period and neighboring border counties.

Identification Strategy

A core empirical challenge is that unobserved factors may jointly determine bond complexity and market outcomes. The authors exploit panel variation in state-level revolving-door regulations — laws that restrict former public officials from taking employment at firms regulated by their former agencies for a “cool-off” period — as an instrument for bond complexity. Between 2010 and 2013, three states (Arkansas 2011, Indiana 2010, Maine 2013) enacted new legislation covering state officials, and two states (New Mexico 2011, Virginia 2011) extended existing regulations to cover local officials. A difference-in-differences regression, with county and year-month fixed effects, shows that adopting revolving-door regulations covering local officials reduces bond complexity by 6% on average (coefficient −0.064, p < 0.01). Regulations targeting only state officials, who are not directly involved in bond negotiations, yield smaller and statistically fragile effects. Placebo checks on auctioned bonds, where underwriters cannot influence design, show no effect, and there is no evidence of pre-existing trends in complexity.

Main Findings

Flexibility vs. liquidity trade-off: A 1% increase in the bond complexity index lowers the number of negative credit-watch events (a proxy for default risk) by 0.002, a 3% decrease relative to the mean of 0.074, confirming that nonstandard provisions provide genuine financial flexibility. However, increasing the complexity index from its mean (1.46) to the 75th percentile (1.69) raises the intermediation spread — the cost for an investor to buy and immediately sell a bond — by 17 basis points (a 14% increase over the average of 120 basis points), confirming that complexity raises trading frictions. For context, the average intermediation spread of 120 basis points is large relative to the 30–60 basis point bid-ask spread of corporate bonds in 2010–2013.
Underwriter incentive to complicate: Increasing complexity from the mean to the 75th percentile raises the underwriter’s market share in secondary-market intermediation by 1.4 percentage points, an 11% increase over the average underwriter share of 12.2%. The underwriter’s gross profits from intermediation also increase with complexity.
Structural estimates — search costs: For a median bond, average dealer search costs amount to 10% of monthly gross profits ($2,625 per month). The underwriter’s exclusive initial sales generate a client network that lowers its effective search costs by 21% relative to an average dealer, more than offsetting its initial geographical disadvantage (for 72% of bonds, the underwriter’s baseline search cost exceeds the median dealer’s). Nonstandard provisions increase both the initial search cost parameter (φ₀) and the network-effect parameter (φ₁): a 1% increase in the complexity index increases φ₀ by 3.79% and φ₁ by 1.66%, implying complex bonds raise search costs broadly but amplify the advantage of a large client network — a position the underwriter occupies via exclusive primary-market sales.
Investor demand: Nonstandard provisions do not substantially change the average investor valuation but substantially increase the dispersion: the standard deviation of investor valuations is 0.003 for simple bonds and 0.013 for complex bonds, consistent with complex bonds being niche products that investors “either love or loathe.”
Government cost: The marginal cost of paying debt obligations is convex in complexity, reaching a minimum at an interior level of provisions; the government’s marginal financial cost increases by 42% when a median bond is stripped of all nonstandard provisions, reflecting the value of payment flexibility.
Conflict of interest: The estimated weight that government officials place on underwriter payoffs in the absence of revolving-door regulations (ψ₀) is 0.34, implying the underwriter’s value accounts for 6.7% of the government official’s payoff under the median unregulated issuer. With revolving-door regulations in place, ψ₁ is essentially zero.

Counterfactual Policies (on representative bond: face value $6.45 million, maturity 7.7 years)

Standardization mandate (ban on all nonstandard provisions): The coupon rate falls from 2.81% to 2.16% (−23%), average dealer search costs fall 47%, and investor surplus rises 13.3%. However, the marginal financial cost (c₀) rises by 41% (from 0.615 to 0.871), so the issuer’s total debt payment cost — principal plus interest, weighted by c₀ — rises by 35%, from $5.13 million to $6.96 million. The standardization policy harms issuers even while saving 7.8% of raw principal-and-interest payments ($8,349K to $7,997K), because the loss of flexibility more than offsets the liquidity gain.
Issuer-driven design (issuer sets complexity to minimize its own debt payment cost, then negotiates the coupon): Complexity falls 19% to 1.14, the interest rate falls to 2.37%, total issuer cost falls 1.5%, investor surplus rises 6%, and the underwriter’s secondary-market payoff falls 19.9%.
Underwriter intermediation ban (underwriter excluded from trading after six months): Complexity falls 5.7% to 1.33, the coupon falls to 2.59%, issuer cost falls 1.5%, but investor surplus falls 1.84% and even other dealers are worse off by 3.97%, because the underwriter’s information on primary-market buyers is lost, offsetting the liquidity gains from lower complexity.

In depth

Q1. What are the five nonstandard bond features tracked as proxies for complexity, and how are they combined into a single index?

Following Harris and Piwowar (2006), the paper focuses on five features that are particularly difficult for investors to price: (i) multiple or serial bonds per issue (as opposed to a single bond), (ii) call provisions allowing early redemption, (iii) sinking fund provisions requiring periodic debt retirement, (iv) nonstandard interest payment frequencies (other than semiannual), and (v) variable or floating interest rates. The complexity index is constructed as the simple average of the latter four provisions across bonds within an issue, plus a dummy for whether the issue contains multiple bonds.

Q2. Why do revolving-door regulations that target local officials reduce complexity more than those targeting state officials?

State officials are not directly involved in bond origination negotiations — they can only indirectly influence local governments through budget allocations. Local officials negotiate directly with underwriters and are thus the proximate counterparties whose incentives the regulations alter. Accordingly, revolving-door regulations covering local officials reduce complexity by 6% (coefficient −0.064, p < 0.01 with full controls), whereas regulations targeting only state officials produce a smaller effect (approximately 2%) that loses statistical significance once issuer financial health controls are added.

Q3. How does the paper validate that revolving-door regulations are a valid instrument for bond complexity?

The paper provides three pieces of evidence. First, the regulations have no effect on the credit ratings of bonds issued prior to their enactment, on the annual amount of bond issuance, or on the maturity length and sale method conditional on issuance — confirming the regulations do not alter governments’ risk management or underlying financing needs. Second, the regulations have no effect on complexity for competitively auctioned bonds, where underwriters cannot influence design — a direct placebo test. Third, a pre-trend analysis (Figure A1) finds no differential trend in complexity in states that subsequently adopted regulations.

Q4. What is the mechanism by which underwriters benefit from adding nonstandard provisions, and why does this advantage not diminish over time?

Underwriters purchase and distribute the entire bond issue at origination, giving them an exclusive network of investors who initially purchased the bonds. In the secondary market, knowing who owns a bond allows the underwriter to locate buyers and sellers with lower search effort. For complex bonds, this advantage is amplified: nonstandard provisions make investor education and persuasion more costly, increasing the value of pre-existing client relationships. The network-effect parameter φ₁ — which governs how rapidly search costs fall as a dealer’s cumulative trades grow — itself rises with complexity (by 1.66% per 1% increase in the complexity index), so the underwriter’s head start in client network accumulation translates into a persistently larger cost advantage precisely for the most complex bonds.

Q5. How large is the underwriter’s search cost advantage in equilibrium, and what drives it?

At the equilibrium meeting rate, the underwriter’s effective search cost of maintaining a given meeting rate is 21% lower than that of an average dealer. This advantage arises despite the underwriter having a higher initial search cost type (φ₀ of $3,609 vs. $3,216 for the average dealer at λ = 1), because for 72% of bonds the underwriter has less local trading experience than the median dealer. The advantage is entirely driven by the underwriter’s network: its exp(−φ₁ log(b)) cost discount factor averages 0.34, 32% lower than the average dealer’s 0.50. The underwriter meets investors 20% more frequently than the average dealer (0.23 vs. 0.19 per month), despite higher absolute search expenditures ($3,045 vs. $2,625 per month).

Q6. How does bond complexity affect investor demand — mean or dispersion of valuations?

Structural estimates show that increasing the complexity index by 1% increases the standard deviation of investor valuations (γ₂) by 4.60% but has no statistically significant effect on the mean valuation (coefficient −0.085, standard error 0.561). This pattern is consistent with complex bonds being niche products — they attract a subset of investors with specific preferences for the embedded features (e.g., certain tax or cash-flow attributes), while being unappealing to most investors. The standard deviation of valuations is 0.003 for a low-complexity bond (25th percentile) and 0.013 for a high-complexity bond (75th percentile).

Q7. What does the structural estimate of ψ₀ imply about the degree of collusion between government officials and underwriters?

The estimated collusion parameter without revolving-door regulations (ψ₀ = 0.34) implies that, for the median unregulated issuing government, the underwriter’s value from secondary-market trading accounts for 6.7% of the government official’s objective function. This is a substantial weight: it means officials act partly as agents for the underwriter rather than purely for taxpayers. With revolving-door regulations (ψ₁ ≈ 0), this collusive weight is essentially eliminated, explaining the empirical reduction in complexity found in Table 2.

Q8. What are the effects of a full standardization mandate on each class of market participant, and why does the issuer lose overall despite paying a lower coupon?

Under standardization, the coupon falls 23% (from 2.81% to 2.16%) and the raw principal-plus-interest payment falls 7.8% (from $8,349K to $7,997K). However, the marginal financial cost c₀ rises 41% (from 0.615 to 0.871), reflecting the loss of payment flexibility previously provided by call provisions and other features; the total issuer cost — c₀A(1 + rT) — rises by 35% (from $5.13 million to $6.96 million). Investors gain 13.3% in surplus because they value liquidity and, on average, do not value nonstandard features. The underwriter loses 36.6% of its secondary-market value while other dealers gain 36.1%, as standardization erodes the underwriter’s network advantage.

Q9. Why does the issuer-driven design scenario outperform standardization in terms of total issuer cost, even though complexity does not fall to zero?

Under issuer-driven design, the government minimizes its total cost of debt payment c₀A(1 + rT), accounting for both the flexibility value of provisions and their effect on the negotiated coupon. The optimal complexity index is 1.14 — positive, but 19% below the current baseline of 1.41 — because some provisions genuinely lower c₀ by allowing flexible debt service. The cost of search frictions (and hence the liquidity premium embedded in the coupon) falls 32% and the negotiated coupon falls to 2.37%, sufficient to reduce total issuer cost by 1.5%. By contrast, full standardization imposes a complexity of zero, which overshoots: c₀ rises more than the coupon savings compensate, increasing total costs by 35%.

Q10. What are the net welfare effects of the underwriter intermediation ban, and why is investor surplus negative despite lower complexity?

The ban reduces complexity by 5.7%, lowering the coupon to 2.59% and reducing issuer costs by 1.5%. However, the underwriter’s client network — built during exclusive initial sales — is a productive resource that improves match quality in the secondary market; banning the underwriter from trading after six months wastes this information. Average dealer search costs rise 1.2% and the meeting rate falls 1.7%, net of the complexity reduction. Investors face bonds with lower coupons and higher effective search frictions, so their surplus falls 1.84%. Non-underwriter dealers also lose 3.97% because lower coupons reduce the rents extractable from intermediation.

Q11. How is the structural model estimated, and what role do revolving-door regulations play in the estimation?

Estimation proceeds in three steps. In Step 1, bond-specific trading market parameters (investor demand, dealer search costs, meeting rates, bargaining parameters) are recovered separately for each bond by minimizing squared differences between observed and simulated trading prices, quantities, and transaction timing. In Step 2, IV regressions using revolving-door regulations and their interactions with county/state attributes as instruments for endogenous complexity map Step 1 parameters to bond attributes, addressing the endogeneity of complexity in determining search costs and investor demand. In Step 3, GMM moment conditions derived from Nash bargaining first-order conditions for the equilibrium complexity and coupon rate identify government preference parameters (θ_c, ψ₀, ψ₁), using the orthogonality condition that unobserved financing cost shocks are mean-zero conditional on observed attributes, regulations, and bond supply from neighboring counties.

Q12. Does the underwriting market show signs of concentration that might amplify the conflict-of-interest problem?

Yes. The mean state-level Herfindahl-Hirschman Index (HHI) for underwriting is 0.12, with the top three firms covering 45% of the market on average. For smaller deals (under $10 million), concentration is markedly higher: mean HHI of 0.24 and top three firms covering 64% of the market. Repeat relationships are common — 41% of bonds issued in 2011–2017 were underwritten by a firm that had underwritten a prior bond for the same issuer within five years — reflecting both informational advantages of local presence and potentially entrenched relationships that may increase government officials’ susceptibility to underwriter influence.

Key Concepts

Complexity index (nonstandard provisions): A bond-level measure computed as the simple average, across bonds within an issue, of four nonstandard features — call provisions, sinking fund provisions, nonstandard interest payment frequency, and variable/floating interest rates — plus a dummy for whether the issue contains multiple bonds. Used as the primary measure of bond complexity in all regressions and the structural model.

Revolving-door regulation: A state-level law restricting former public officials or employees from engaging in lobbying or taking employment at regulated firms for a specified “cool-off” period (typically one to two years) after leaving office. The paper uses the presence and scope of such regulations (whether they cover state officials, local officials, or both) as a source of exogenous variation in government officials’ incentives to align with underwriter interests.

Intermediation spread: The logarithm of the average dealer-to-investor sale price minus the logarithm of the average dealer-from-investor purchase price for a given bond. Used as the empirical measure of trading frictions; the sample average is 120 basis points.

Network effect in search (φ₁): The parameter governing how a dealer’s cumulative prior trades with investors in a given bond reduce its cost of meeting new investors for that bond. A higher φ₁ means a larger client network translates into steeper cost savings. The paper estimates that φ₁ itself increases with bond complexity, so complex bonds amplify the advantage of dealers (especially the underwriter) who accumulate large client networks.

Marginal cost of debt payment (c₀): A bond- and issuer-specific parameter capturing the effective cost to the government of repaying each dollar of principal and interest, net of the flexibility benefits provided by nonstandard provisions. Normalized to one for a bond with zero nonstandard provisions at average issuer characteristics; estimated to be convex in complexity with an interior minimum, implying some nonstandard provisions are beneficial from the government’s perspective.

Collusion weight (ψ): The weight a government official places on the underwriter’s secondary-market value from trading when negotiating bond design. Estimated at ψ₀ = 0.34 in the absence of revolving-door regulations (implying the underwriter’s interest accounts for 6.7% of the official’s objective) and at ψ₁ ≈ 0 when such regulations are present.

Underwriter dual role: The institutional arrangement in which the same investment bank (i) negotiates and purchases the entire bond from the issuing government at origination, and (ii) subsequently acts as a dealer in the bond’s secondary market. This dual role creates an incentive to design complex bonds that strengthen the underwriter’s competitive advantage in secondary intermediation via network effects in search.

Issuer-driven design: A counterfactual policy scenario in which the government sets the complexity level to minimize its total cost of debt payment — accounting for both the flexibility value of provisions and the anticipated effect on the negotiated coupon rate — before bargaining with the underwriter only over the coupon. This policy allows some nonstandard provisions (complexity index 1.14 vs. baseline 1.41) and reduces total issuer cost by 1.5% relative to the baseline.

Tell Me Something I Don't Already Know: Learning in Low- and High-Inflation Settings

Mon, 01 Jan 0001 00:00:00 +0000

This paper uses randomized control trials (RCTs) applied over time in multiple countries to study whether the economic environment — specifically the level of inflation — affects how agents learn from new information. The main finding is that as inflation rose in advanced economies, both households and firms became more attentive to and informed about publicly available news about inflation, causing them to respond less to exogenously provided information about inflation and monetary policy in the RCT treatments. When agents are already well-informed about the current inflation environment (because high inflation makes it salient), additional information provision moves their beliefs less — the marginal value of information is decreasing in prior attentiveness. Complementary evidence from Uruguay (persistently high inflation) and New Zealand (persistently low inflation) confirms the cross-sectional pattern: agents in high-inflation environments have stronger priors and respond less to information treatments. The results imply that central bank communication interventions are more effective during low-inflation periods when agents are less informed.

In depth

Q1. How does the RCT design identify the effect of the inflation environment?

The paper exploits the within-country time variation in inflation — running comparable RCT information treatments before, during, and after the 2021-2023 surge in multiple countries — to identify whether the same information treatment has different effects depending on the prevailing inflation level. By holding the content of the information treatment constant and varying the macroeconomic environment, the paper isolates how environment-driven changes in agent attentiveness mediate the treatment effect.

Q2. Why do agents respond less to information when inflation is high?

When inflation is high and salient, agents actively monitor publicly available inflation news, reducing their prior uncertainty; the Bayesian posterior shift from a given information signal is smaller when the prior is more concentrated. This mechanism — decreasing marginal value of information with prior precision — means that information provision campaigns are subject to diminishing returns as the macroeconomic environment itself provides more signal.

Q3. What does this imply for central bank communication strategy?

Central bank forward guidance and information campaigns are most effective when agents have diffuse priors — i.e., during periods of low, stable inflation when inflation is not salient to households and firms. During high-inflation episodes, agents become more sophisticated but also more resistant to updating on any individual information signal, making communication a weaker tool precisely when the inflation challenge is most acute.

Key concepts

prior attentiveness : the degree to which agents actively monitor publicly available information about inflation; the paper shows this rises with inflation level, reducing the marginal value of additional information provided by RCT treatments.

decreasing marginal value of information : the property that additional information moves beliefs less when agents already have precise priors; the mechanism explaining why RCT treatment effects are smaller in high-inflation environments.

The Cost of Consumer Collateral: Evidence From Bunching

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1: Overview

This paper estimates the shadow cost that consumers assign to pledging their primary residence as collateral, using administrative loan application and performance data from the U.S. Federal Disaster Loan (FDL) Program, which offers low-interest loans to households following natural disasters. A loan amount threshold — set at $10,000 from 2005–2007, $14,000 from 2008–2013, and $25,000 from 2014–2018 — separates uncollateralized from collateralized borrowing, with no other loan terms changing at the threshold; this sharp, discontinuous design allows the paper to use bunching estimation to identify collateral aversion. Roughly one-third of all program borrowers, and 38% of those with losses above the threshold, choose exactly the maximum uncollateralized loan amount, generating sharp mass at the threshold. Traditional bunching estimates, corroborated by two alternative approaches using household-level damage data and originally requested loan amounts, consistently find that the median borrower is willing to forgo 40–47% of their potential loan amount to avoid pledging their home as collateral, equivalent in demand terms to a 200 basis point interest rate increase. The paper also exploits threshold variation over time as an instrument for collateralization and finds that posting collateral causally reduces default rates by approximately 35%, an effect comparable in magnitude to a 100-point increase in borrower credit score, establishing that collateral substantially mitigates moral hazard in consumer lending.

In depth

Q1. What is the setting and why does it cleanly identify the collateral shadow cost?

The Federal Disaster Loan Program creates a quasi-experimental threshold because collateral is the only loan term that changes at the $25,000 boundary — interest rate, maturity, approval probability, and all other terms remain identical on both sides. Households with uninsured disaster damages (median $51,000) can borrow up to their loss amount at a fixed 2.5% rate; those requesting above the threshold must post their home as collateral. Unlike mortgage or auto loan markets, which always require collateral, and credit card markets, which never do, this program generates a setting where the binary collateral requirement is the borrower’s own choice subject only to the threshold, eliminating the standard endogeneity between contract terms and borrower risk. The authors use administrative data covering over 1 million applications from 2005 to 2018 across all 50 states.

Q2. How do the bunching estimates identify the distribution of collateral aversion?

The traditional bunching estimator fits a polynomial to the distribution of loan amounts below the threshold, extrapolates it above the threshold to construct a counterfactual without the collateral requirement, and attributes the excess mass at the threshold (the “missing” mass above) to collateral aversion. The bunching region spans from the threshold to the upper loan amount beyond which essentially no borrower would give up to avoid collateral; for the $25,000 threshold, this region extends to $49,900. Across all three threshold regimes, 73–78% of borrowers within the bunching region move to the threshold, and the estimated mean private value of collateral ranges from $7,944 (at the $10,000 threshold) to $18,268 (at the $25,000 threshold), representing 37–44% of ideal loan amounts. The median borrower’s collateral aversion — 39–47% depending on the threshold — is robust across all three estimation methods.

Q3. What alternative bunching methods does the paper develop and what do they find?

Two alternative estimation approaches — a difference-in-bunching (DiB) estimator that compares borrowers with identical damage levels across different threshold regimes, and an originally-requested-loan estimator that uses the amount households requested before collateral salience increased — both yield median collateral aversion estimates consistent with the traditional method (40–47%), while suggesting considerably wider heterogeneity in the tails. The DiB approach exploits the fact that for a household with $20,000 in damages, the same loan amount was uncollateralized under the $25,000 regime but required collateral under the $10,000 regime, enabling consumer-level identification. The originally-requested-loan method shows that 70% of eventual bunchers initially requested an amount above the threshold, with the majority of the shift occurring after meeting with a loan officer when the collateral requirement became salient. Both methods are immune to the standard counterfactual mis-specification concern of the traditional bunching approach, and they suggest the traditional estimator substantially under-predicts the proportion of highly collateral-averse borrowers at the upper end of the distribution.

Q4. What is the mechanism behind collateral aversion and what does heterogeneity reveal?

Collateral aversion in this setting is driven by both financial incentives and behavioral/preference factors, as evidenced by the finding that roughly 30% of borrowers already underwater on existing mortgages — who have no real equity to lose — still bunch at the threshold to avoid adding a lien on their home. More creditworthy borrowers (higher credit scores, higher incomes) are actually more likely to bunch, consistent with an “advantageous selection” interpretation in which borrowers who are confident in their repayment capacity are especially averse to the stigma and risk of pledging their home. Interest rates also matter: borrowers bunch more when program interest rates are higher, suggesting that financial incentives amplify the existing aversion. The magnitude of bunching — giving up thousands of dollars of subsidized low-interest disaster recovery loans — indicates that the perceived cost of a lien on the primary residence extends far beyond the financial value of potential foreclosure.

Q5. How does collateral causally reduce default rates?

Using time variation in the collateral threshold as an instrument for whether a borrower’s loan is collateralized — borrowers are more likely to collateralize when the threshold is low (so their ideal loan amount exceeds the low threshold) than when it is high — the paper estimates that collateral causally reduces default rates by about 35%. This local average treatment effect applies to borrowers who would collateralize under the $10,000 threshold but not under the $25,000 threshold, and the magnitude is comparable to a 100-point FICO score improvement. The finding implies that collateral requirements address genuine moral hazard in consumer lending: when a primary residence is pledged, borrowers take repayment obligations substantially more seriously, reducing the probability of strategic or precautionary default. This provides causal evidence for the classic prediction of models like Bester (1985) and Chan and Thakor (1987) that collateral mitigates information asymmetries and expands efficient credit access.

Q6. What are the aggregate implications and contribution to methodology?

In aggregate, borrowers in the program have given up more than $1.1 billion in disaster recovery loans to avoid posting collateral, indicating that collateral requirements — standard in most large consumer credit markets — impose large implicit costs that are not captured in stated interest rates. The methodological contribution is threefold: the paper is among the first to apply bunching to consumer (rather than corporate) collateral; it develops two consumer-level alternative estimators that relax assumptions of the standard method and reveal greater heterogeneity in collateral aversion; and it separately identifies the moral hazard effect of collateral from adverse selection by exploiting threshold variation, extending bunching beyond tax compliance and into household finance.

Key Concepts

collateral shadow cost : the implicit cost a borrower assigns to pledging collateral beyond the direct financial cost; measured in this paper as the maximum loan amount a borrower forgoes to avoid posting their home, identified from bunching mass at the collateral threshold in the FDL program.

bunching estimator : an estimation strategy that infers a structural parameter — here collateral aversion — from the excess density of agents at a policy threshold, using a polynomial-extrapolated counterfactual distribution to identify the missing mass above the threshold.

difference-in-bunching (DiB) estimator : a consumer-level alternative to traditional bunching estimation that uses borrowers with identical damage levels but facing different threshold regimes over time to construct a within-person counterfactual for the ideal loan amount, avoiding assumptions about the counterfactual distribution shape.

advantageous selection : the pattern in which more creditworthy, higher-income borrowers are the ones most likely to avoid collateral requirements, the reverse of the adverse selection typically assumed in collateral models; consistent with these borrowers having strong repayment intent independent of the lien.

The Hitchhiker's Guide to Markup Estimation: Assessing Estimates From Financial Data

Mon, 01 Jan 0001 00:00:00 +0000

Using matched data combining physical production records (EAP survey) and financial statements (FARE) for 147,403 firm-years across 18 two-digit manufacturing industries in France (2009–2019), the paper audits the production-approach markup estimator — μ = α × (PY/WV) — when α must be estimated from revenue rather than quantity data. The central analytical result is that revenue-based markup estimates are positively correlated with true markups whenever the estimated output elasticity falls strictly between zero and the revenue elasticity (the Bond et al. 2021 knife-edge case). For firms sharing the same production function, the correlation is exactly one — revenue markups rank firms identically to quantity markups up to an additive constant. Monte Carlo simulations in an Atkeson–Burstein (2008) oligopoly model with translog production confirm this: correlation between revenue and true markups is 0.94 (SD 0.05) across 200 simulations with 1,600 firms, 180 markets, and 40 periods. In the French matched data, within-sector Pearson correlations between quantity and revenue markup estimates are 0.61 in levels and 0.80 in first differences; rank correlations reach 0.62 and 0.83 respectively, with medians above 0.65 and 0.84 across sectors. Cross-sectional relationships between markups and profit rates, labor shares, material shares, and market shares are qualitatively robust across estimation methods (Table VI). However, average log markup levels differ sharply: 0.37 (quantity) vs 0.13 (revenue); aggregate French manufacturing markups average 1.45 (quantity) vs 1.08 (revenue). The policy implication: revenue-based financial data are adequate for studying markup dispersion, trends, and firm-level correlates; they cannot identify markup levels.

In depth

Q1. What is the core analytical result about when revenue data can identify true markups?

The production-approach markup estimator μ_it = α_it × (P_it Y_it / W_t V_it) recovers true markups from revenue data if and only if the estimated output elasticity α̂ equals the true quantity elasticity; when α̂ equals the revenue elasticity instead, estimated markups are identically one; in the empirically relevant intermediate case where α̂ lies strictly between these extremes, revenue markups are positively correlated with true markups. The key insight is that standard estimators (ACF, LP, OLS) applied to revenue data produce an α̂ that lies strictly between zero and the revenue elasticity (which is less than the true output elasticity for firms with market power), so revenue markup estimates are neither correct nor uninformative. For firms sharing the same production technology — the common case in Cobb-Douglas with industry-level coefficients — revenue markups equal true markups up to an additive constant, implying a within-industry correlation of exactly one. The Bond et al. (2021) result (revenue markups ≡ 1) arises only in the knife-edge case where ACF converges exactly to the revenue elasticity; under the empirically dominant case of partial revenue elasticity contamination, the correlation is high but the level is biased downward.

Q2. How does the paper relate to the Bond et al. (2021) critique, and when does that critique apply exactly?

Bond et al. (2021) showed that if the ACF procedure is applied to revenue data and converges to the revenue elasticity, then estimated markups are identically one regardless of true markups; the present paper demonstrates this is a knife-edge special case, not a generic property of revenue-based estimation. The Bond et al. result requires that the estimated α̂ equals exactly the revenue elasticity r = α/(1+α·(μ−1)/μ) × (1/price index correction) — a relationship that holds only under specific functional form and market structure assumptions. In the Atkeson–Burstein Monte Carlo (translog production, oligopolistic competition), the ACF estimator applied to revenue data produces α̂ = 0.29 against true βv = 0.32 and revenue elasticity that is strictly lower — placing the estimate in the intermediate zone. The general principle: when α̂ is strictly between zero and the revenue elasticity, revenue markups are informative; when α̂ converges to the revenue elasticity, they are not.

Q3. What do Monte Carlo simulations in the Atkeson–Burstein model show about the reliability of revenue markup estimates?

Across 200 simulations with 1,600 firms organized in 180 markets over 40 periods, using a translog production function, the correlation between revenue-based ACF markup estimates and true quantity-based markups is 0.94 (SD 0.05), and all distributional moments of the markup distribution are well-recovered from financial data. Table I reports the estimated variable input elasticity: true βv = 0.32, quantity estimate = 0.32, revenue-ACF estimate = 0.29 — the small downward bias in the elasticity generates a small downward bias in the markup level but does not destroy the cross-sectional ranking. Table II reports the markup distribution correlations: revenue vs. true = 0.94 (SD 0.05) vs. quantity vs. true = 1.00 (SD 0.01). Cross-sectional regression coefficients of revenue markups on quantity markups are 0.85 in levels and 0.99 in first differences; all moments (mean, standard deviation, median, interquartile range) are well-estimated from revenue data. The simulations thus confirm the analytical result: revenue data preserves the cross-sectional ranking of markups and their time-series variation, even though the level is biased downward.

Q4. What does the matched French production-and-financial-statement data show about within-sector correlations?

In the French manufacturing data (EAP physical production records matched to FARE financial statements, 2009–2019, 147,403 observations, 18 two-digit industries), within-sector Pearson correlations between quantity and revenue markup estimates are 0.61 in levels and 0.80 in first differences; rank correlations reach 0.62 (levels) and 0.83 (first differences), with medians above 0.65 and 0.84 respectively across sectors. Table III documents the divergence in average output elasticities: 0.54 (quantity) vs. 0.40 (revenue), with quantity exceeding revenue in 16 of 18 industries and an average gap of 38%. Table IV shows the implied log markup averages: 0.37 (quantity, SD 0.23) vs. 0.13 (revenue, SD 0.16) — a large level gap but smaller dispersion gap. Table V reports within-sector correlations: Pearson 0.61 (levels) and 0.80 (first differences); rank 0.62 and 0.83. Binned scatter regressions confirm: slope 0.89 (levels) and 0.92 (first differences). These within-sector correlations are what matter for studying which firms have relatively higher or lower markups, and they are high enough to support revenue-based research on markup dispersion, cyclicality, and firm-level correlates.

Q5. Are the relationships between revenue markup estimates and other firm characteristics reliable?

Table VI shows that the associations of markup estimates with profit rate, labor share, material share, and market share are qualitatively robust across revenue and quantity estimation methods — the sign and relative magnitude of correlations are preserved, even though the level of markups differs. This result validates using financial data for the dominant research application of production-approach markup estimation: studying the firm-level determinants and correlates of markups. The positive correlation between markups and profit rates, the negative correlation with labor shares, and the positive correlation with market shares all survive the switch from quantity to revenue-based estimation. The key caveat is that the magnitudes of these correlations are attenuated under revenue estimation due to the downward bias in markup levels, but the qualitative patterns are preserved with statistical significance.

Q6. What do aggregate French manufacturing markup trends show, and what is the paper’s policy conclusion?

Aggregate French manufacturing markup trends are qualitatively well-captured by revenue-based estimates (Figure 5), but the average level differs substantially: 1.08 (revenue) vs. 1.45 (quantity) — a gap of 37 log points. The time-series evolution of the markup distribution (aggregate mean, trends, cyclical variation) is reliably estimated from financial statements. This implies that the empirical literature documenting rising markups over the past decades — primarily using revenue-based financial data — has correctly identified the direction and approximate magnitude of markup trends, even if the levels are underestimated. The paper’s policy conclusion is precise: revenue data are adequate for studying markup dispersion across firms, trends over time, and correlates with other firm characteristics; they are not adequate for measuring the absolute level of markups or for calibrating models that require accurate markup levels (e.g., models of aggregate price-setting or optimal tax policy).

Key concepts

production-approach markup estimator : the method pioneered by Hall (1986, 1988) and formalized by De Loecker and Warzynski (2012) that estimates firm-level markups as μ = α × (PY/WV), where α is the output elasticity of the variable input V, P is output price, Y is output, and W is input price; requires physical quantity data in principle but is widely applied to revenue (PY) data from financial statements.

revenue elasticity : the elasticity of firm revenue with respect to variable input, which equals the true output elasticity divided by a markup adjustment term; equals the output elasticity only for perfectly competitive firms; lies strictly below the output elasticity for firms with market power; when the estimated α̂ converges to the revenue elasticity, the production-approach estimator returns markups identically equal to one (the Bond et al. 2021 result).

Bond et al. (2021) knife-edge : the special case in which the ACF production function estimator applied to revenue data converges to the revenue elasticity — causing estimated markups to be identically one and uninformative about true markups; the present paper shows this is not a generic property but a specific functional-form restriction.

within-sector cross-sectional rank : the ordering of firms within an industry by their estimated markups; the statistic most reliably preserved when using revenue rather than quantity data, because within-industry production function homogeneity causes revenue markups to equal true markups up to an additive constant, so ranks are identical analytically.

level vs. dispersion distinction : the paper’s central empirical finding that revenue markup estimates reliably measure within-industry dispersion, trends, and rankings (the first moment of changes and cross-sectional variation) but cannot recover the absolute level of markups (the mean log markup is 0.13 from revenue vs. 0.37 from quantity in French manufacturing, and aggregate average is 1.08 vs. 1.45).

The Impact of Incarceration on Employment, Earnings, and Tax Filing

Mon, 01 Jan 0001 00:00:00 +0000

Overview

This paper estimates the causal effect of incarceration on employment, wage earnings, self-employment, and tax filing behavior using administrative criminal justice data linked to Internal Revenue Service (IRS) records for approximately half a million felony defendants in two U.S. states: North Carolina and Ohio. The study period covers cases filed from the early 2000s through 2014, with outcomes tracked through 2020 using IRS W-2 and 1040 records.

Research Question

The central question is whether incarceration itself — as distinct from arrest, conviction, and other criminal justice interactions that precede or accompany it — causes lasting reductions in defendants’ labor market outcomes. The paper explicitly holds fixed upstream interactions (conviction, arrest) to isolate the effect of the incarceration sentence.

Data and Sample

Criminal justice records from Ohio (Common Pleas courts in Franklin, Cuyahoga, and Hamilton counties, covering Columbus, Cleveland, and Cincinnati) and North Carolina (Administrative Office of the Courts and Department of Public Safety) are linked to de-identified IRS records via name, date of birth, sex, address, and partial Social Security Numbers. Match rates are 92% in Ohio and 95% in North Carolina. The sample is restricted to defendants aged 18–50 at time of offense with cases filed 2002–2014. IRS records include employer-reported W-2 wages (regardless of individual tax filing), self-employment income from Schedule C/SE, non-employee compensation (1099-MISC), and gig-economy earnings from 1099 returns. All dollar figures are adjusted to 2016 dollars using the PCE deflator.

Empirical Strategy

Two independent quasi-experimental research designs are used:

North Carolina — Sentencing guideline discontinuities: North Carolina’s structured sentencing guidelines map offense class (E through I, the five least severe felony classes) and prior record points (a numerical criminal history score) into permissible punishment types (incarceration vs. probation) and sentence lengths. Allowable punishment types change discretely at five cell boundaries, generating discontinuities in incarceration sentences for otherwise similar defendants. The paper uses these five boundary discontinuities as excluded instruments in a parameterized regression discontinuity design stacked across offense classes. First-stage F-statistic = 115.
Ohio — Random assignment to judges: Cases are randomly assigned by computer to judges at arraignment in the three counties studied. Judge leave-out mean sentence length is used as an instrument for individual sentence length. The design follows Norris et al. (2021) and yields F-statistic = 321. The instrument shifts sentences along both the extensive margin (any vs. no incarceration) and intensive margin (longer vs. shorter sentences).

Both designs produce complier populations for whom at least 37–45% are shifted on the extensive margin (from no incarceration to some incarceration), based on partial identification bounds using linear programming.

Main Findings

The paper’s central finding is that incarceration generates large short-run reductions in labor market activity during the incapacitation period, but no detectable long-run reductions in annual employment or earnings once defendants have been released and the incapacitation effects have dissipated.

In the first year after case filing, when incarceration rates peak (roughly 75–100 additional days incarcerated for a 12-month sentence), employment falls by approximately 10 percentage points and total W-2 earnings contract commensurately.
Within 3–4 years of filing, employment effects return to near zero and are statistically insignificant in both states.
Five to nine years after filing, when effects on contemporaneous incarceration have dissipated, the estimated effect of a 12-month sentence on annual earnings is positive or near zero in both states. The combined 95% confidence interval rules out reductions in annual wages greater than $231 (approximately 5% of the untreated complier mean) and rules out any adverse employment effects.
Despite no long-run level effects, losses during incapacitation are never recouped. A one-year sentence reduces cumulative earnings over five years by approximately $2,914 — a 13% reduction relative to the complier mean.
Effects on self-employment, independent contracting, 1040 filing, adjusted gross income, EITC take-up, and interstate migration are similarly null in the long run.

Incapacitation vs. Post-Release Scarring

The paper provides two tests for whether short-run earnings losses reflect incapacitation alone or also post-release scarring (e.g., human capital depreciation, employer discrimination, or discouragement effects):

A “visual IV” regression of year-t earnings effects on year-t days-incarcerated effects yields an R² of 0.83–0.85 across states, with the intercept near zero (positive and small), indicating that virtually all dynamic earnings impacts flow through contemporaneous incapacitation and not through a post-release channel.
Constructed outcomes that impose the null of pure incapacitation (scaling pre-case average earnings or covariate-predicted earnings by the share of the year free from prison) closely track actual earnings effects in both states, further confirming that incapacitation is the dominant mechanism.

Pre-existing Labor Market Detachment

A key scope condition is defendants’ severe labor market disadvantage prior to their case. Fewer than 50–60% of defendants are employed in the year before filing; average pre-case W-2 earnings (including zeros) are below $6,000. Among employed defendants, only 10% earn more than $22,000 per year. Untreated complier means for earnings in the year after case filing are below $4,000, with virtually no earnings or employment growth over the following nine years. The paper concludes that returning to pre-filing earnings levels is sufficient for incarcerated defendants to match their non-incarcerated peers — a low bar that is readily met.

Policy Implications

Back-of-envelope aggregation implies incapacitation losses of approximately $6.16 billion per year in foregone earnings for the U.S. prison population, concentrated in communities heavily affected by incarceration. However, a marginal reduction in incarceration rates would increase average earnings by only $51 for white men and $213 for black men, suggesting incarceration’s direct contribution to labor market inequality is modest relative to the $21,100 black-white earnings gap estimated by Bayer and Charles (2018). The paper concludes that upstream factors — other criminal justice interactions, human capital deficits, and broader socioeconomic disadvantage — are more plausibly responsible for low earnings among the formerly incarcerated.

In depth

Q1. What is the exact treatment variable, and what is the counterfactual?

The treatment variable is months of incarceration sentenced in the focal case (a continuous, weakly positive ordered treatment). The counterfactual for non-incarcerated defendants in North Carolina is probation (all defendants are convicted by construction under structured sentencing guidelines). In Ohio, the authors cannot reject that all compliers who do not receive a prison sentence are still convicted, implying the counterfactual is also conviction and probation. All compliers therefore acquire a criminal record regardless of sentence. The treatment effect is thus the effect of incarceration conditional on conviction, holding fixed the criminal record.

Q2. How are effects interpreted given multiple instruments and continuous treatment?

Under a “weakly positive ordered treatment” assumption and standard LATE conditions, the 2SLS estimates can be interpreted as Average Causal Responses (ACRs) — weighted averages of the marginal dose effects (12 vs. 11 months, 6 vs. 5 months, 1 vs. 0 months, etc.) for complier subgroups shifted by each instrument. In North Carolina with five parameterized RD instruments, the estimate averages ACRs weighted by first-stage strength. In Ohio with a leave-out mean instrument, the estimate is a convex average of ACRs under the assumption that the linear first-stage model is a good approximation. Dosage weights for both states put mass on a wide range of sentence lengths including both extensive and intensive margins, though Ohio’s weights are more skewed toward shorter sentences.

Q3. How large are the first-stage effects, and how strong is the instrument?

In North Carolina, sentences jump by 50% or more at sentencing guideline cell boundaries where allowable punishment types change to include incarceration. The first-stage F-statistic is 115. In Ohio, defendants assigned to the most severe judge receive incarceration sentences approximately six months longer than those assigned to the least severe judge (roughly 30% of the average non-zero sentence), with a slope of approximately 0.8 in the first-stage regression; F-statistic = 321. At least 37% of compliers in North Carolina and 45% in Ohio are shifted on the extensive margin (from no incarceration to some positive incarceration), with upper bounds as high as 95%.

Q4. What evidence supports instrument validity (exclusion restriction and independence)?

Instrument validity is tested by estimating 2SLS “effects” on pre-case outcomes measured 2–4 years before the focal case. In both states, the instruments show no relationship with pre-case employment, W-2 wages, total days previously incarcerated, or binary severe prior incarceration. The probability of being matched to IRS records and the quality of the match are also uncorrelated with the instruments. In Ohio, potential exclusion restriction violations from judges affecting conviction (not just sentence) are addressed empirically: nearly 90% of defendants are convicted, the most severe judge is only 0.7 p.p. more likely to convict than the least severe judge (t-stat = 1.53), and the estimated conviction rate among untreated compliers is 0.972 (s.e. 0.018), so one cannot reject that all non-incarcerated compliers are convicted.

Q5. How does the paper test for the incapacitation mechanism against post-release scarring?

Two complementary exercises are conducted. First, a “visual IV” plot regresses year-t earnings effects on year-t days-incarcerated effects across all post-filing years. If incapacitation is the sole channel, all points should lie on a line through the origin. The R² is 0.83 in North Carolina and 0.85 in Ohio, the estimated intercept is near zero (positive and small) in both states, and the slope (earnings lost per day incarcerated) is approximately $12. This implies cumulative earnings losses of $12 × 268 days = $3,216, very close to the directly estimated $2,914. Second, constructed outcomes that scale pre-case earnings or covariate-predicted earnings by the share of the year not incarcerated closely track actual earnings effects throughout the post-filing period, and both converge to zero as incapacitation effects fade — consistent with pure incapacitation and no net scarring.

Q6. What are the long-run (5–9 years) earnings and employment estimates, and how precisely are null effects ruled out?

Averaged across both states using inverse-variance weights, the estimated effect of a 12-month sentence on annual W-2 earnings five to nine years after filing is positive but statistically indistinguishable from zero. The 95% confidence interval rules out reductions in annual wages greater than $231 (approximately 5% of the untreated complier mean of roughly $4,500–$5,000). The 95% CI also rules out any adverse employment effects. The untreated complier mean for employment 5–9 years post-filing is approximately 40% in North Carolina and slightly above 40% in Ohio.

Q7. What happens to cumulative earnings over five years despite null long-run level effects?

Even though long-run annual earnings are unaffected, earnings losses during incapacitation are never made up. A one-year sentence reduces cumulative employment (measured as years with any W-2) and cumulative earnings over five years by approximately $2,914 — a 13% reduction relative to the complier mean. This reflects the mechanical loss of earnings during the period of physical incapacitation, without a subsequent compensating period of higher earnings after release.

Q8. Do defendants with stronger pre-case labor market attachment show different long-run patterns?

The sample is split between defendants employed in at least 2 of the 4 years prior to the case (53–57% of the sample across states) and those less attached. Both groups show zero long-run earnings and employment effects. Previously employed defendants experience much larger short-run earnings drops — more than three times larger in the first year post-filing — and their earnings recover more slowly, reaching zero effect approximately six years after filing (vs. three years for the previously unemployed). For a stricter cut (pre-case average earnings above $15,000, representing only 12–15% of the sample), the long-run earnings effect is −$1,426 (8% of the untreated complier mean), significant only at the 10% level, and partly attributable to residual incapacitation (19.6 additional days incarcerated 5–9 years post-filing). For defendants with pre-case earnings below $15,000, incarceration slightly increases long-run employment (2.4 pp, p = 0.01) and earnings ($400, p = 0.03), possibly reflecting rehabilitative benefits (GED or educational programs) for labor-market-detached individuals.

Q9. Does first-time incarceration (extensive-margin exposure) have larger long-run effects than repeat exposure?

The paper tests this by splitting the sample into defendants with and without prior incarceration history. Among defendants with no prior incarceration, the instruments generate large differences in lifetime exposure: a 12-month sentence increases the probability of ever being incarcerated over the next 5–9 years by 26 p.p. (North Carolina) and 41 p.p. (Ohio). Among those not receiving a sentence, 48% (North Carolina) and 19% (Ohio) are eventually incarcerated anyway, implying treatment causes a 52 and 81 p.p. increase in lifetime incarceration probability for extensive-margin compliers. Despite these large differences in lifetime exposure, long-run earnings and employment effects remain small and statistically insignificant in both subsamples. The difference in long-run effects between previously and never incarcerated defendants is not statistically significant (p = 0.29 for employment, p = 0.82 for earnings).

Q10. Are there heterogeneous effects by race, sex, or criminal history?

There is no evidence of long-run scarring for any demographic or criminal history subgroup. Effects for black and non-black defendants are both positive for long-run earnings and employment. Non-black defendants show somewhat larger cumulative losses (consistent with marginally higher counterfactual earnings), but differences are not statistically significant. Estimates for women are imprecise due to small sample size. Among defendants with and without prior felony charges in the four years preceding the case, there are neither economically nor statistically significant long-run earnings or employment effects. Cumulative losses are somewhat larger for defendants without prior felony charges (p = 0.07), reflecting their higher pre-case earnings.

Q11. How does the paper handle potential migration bias in outcomes?

Tax filing and W-2 receipt in the state of sentencing are used to proxy for whether defendants remain in the same state. Among untreated compliers, 88% of those with a tax footprint maintain it in the state of sentencing. No statistically significant effects of incarceration on migration (measured as filing or receiving a W-2 in North Carolina or Ohio) are detected, suggesting prior studies of recidivism measured within-state are unlikely to be severely biased by migration responses.

Q12. What effect does incarceration have on mortality?

Incarceration reduces five-year mortality by approximately 0.8 percentage points (about 20% of the untreated mean). The authors note this is too small to explain the null long-run labor market effects: even if all defendants whose death was averted were employed, removing them from the employment count would reduce the employment effect of a 12-month sentence only to approximately zero.

Q13. How do the paper’s findings compare to prior studies, particularly Mueller-Smith (2015)?

Mueller-Smith (2015) finds large and persistent negative incarceration effects on labor market outcomes in Texas using a structural decomposition and Lasso-based judge-covariate interactions as instruments. The paper argues methodological differences are the likely explanation: the Lasso-selected interacted instruments can be susceptible to many-weak instruments bias toward OLS. It notes that Mueller-Smith’s simpler 2SLS specifications (analogous to those used here) show no statistically significant earnings effects. North Carolina and Ohio are documented to be broadly similar to Texas (and the U.S. average) in rehabilitation program participation, recidivism rates, and incarceration rates, reducing the likelihood that genuine geographic heterogeneity explains the divergence.

Q14. What is the paper’s aggregate extrapolation of incapacitation earnings losses?

Scaling the estimated $2,914 cumulative loss per 12-month sentence by the ratio of days exposed to total days in a year gives a per-day loss of approximately $12. Applied to the 1,435,500 people incarcerated in U.S. prisons on any given day in 2019 (excluding the more than 700,000 in jail), the implied aggregate yearly earnings loss from incapacitation is approximately $6.16 billion.

Key Concepts

Incapacitation effect: The mechanical reduction in earnings and employment that occurs while a defendant is physically confined in prison and unable to work, as distinct from any post-release scarring effect. The paper shows this is the dominant — and essentially sole — causal channel through which incarceration affects labor market outcomes in their sample.

Post-release scarring: Persistent reductions in earnings or employment that persist after a defendant is released from prison, caused by mechanisms such as employer discrimination based on incarceration history, human capital depreciation, loss of job contacts, or psychological discouragement effects. The paper finds no evidence of scarring in either state.

Average Causal Response (ACR): The weighted average of the marginal dose effects of incarceration (e.g., effect of 12 vs. 11 months, 1 vs. 0 months) for groups of defendants whose sentence lengths are shifted by a given instrument. Contrasted with a binary LATE, the ACR averages across the full dosage distribution for compliers.

Complier: An individual whose incarceration sentence is shifted by the instrument — either from zero to some positive sentence (extensive margin) or from a shorter to a longer sentence (intensive margin). Counterfactual outcome means for compliers sentenced to zero months provide the baseline for evaluating effect magnitudes.

Sentencing guideline discontinuity: The discrete jump in permissible punishment types and minimum sentence lengths at specific criminal history score thresholds within North Carolina’s structured sentencing grid. Defendants just above a threshold are more likely to be incarcerated than otherwise similar defendants just below, generating quasi-experimental variation exploited as an instrument.

Leave-out mean judge instrument: In Ohio, each defendant’s assigned judge’s average incarceration sentence length computed over all other cases that judge handles (excluding the defendant’s own case), residualized on court-by-month fixed effects. Because judges are randomly assigned to cases, this measure is conditionally independent of defendant potential outcomes and serves as an instrument for sentence length.

Control complier mean: The estimated mean potential outcome for compliers under the counterfactual of receiving zero months of incarceration. Used as a benchmark to evaluate the magnitude of treatment effects and to characterize how low the earnings baseline is for the population driving the causal estimates.

Extensive vs. intensive margin of incarceration: The extensive margin refers to the binary shift from receiving no prison sentence to receiving any prison sentence; the intensive margin refers to increasing sentence length conditional on some incarceration. The paper argues that neither margin appears to produce long-run labor market scarring, and uses linear programming bounds to estimate that at least 37–45% of compliers in each state are shifted on the extensive margin.

The Margins of Trade

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1 — Overview

Research Question

Eaton and Fieler seek to reconcile two literatures that have advanced in parallel but remained at odds: (i) general equilibrium models of bilateral trade flows (the “gravity” tradition) and (ii) empirical work on the margins of trade — the decomposition of bilateral trade into the extensive margin (number of products traded), the quantity margin (physical volumes), and the unit-value (price) margin. Standard GE models cannot accommodate two of the most robust empirical regularities: that richer importing countries pay higher unit values for the same product, and that richer exporting countries charge higher unit values. The paper builds a framework that captures all three margins jointly while still delivering the standard gravity equation and the Arkolakis-Costinot-Rodriguez-Clare (ACR) welfare formula.

Data

The analysis uses UN COMTRADE bilateral merchandise trade data for the 50 largest economies by GDP in 2007, the most disaggregated 6-digit HS product classification (HS6). The working sample covers 2,611,700 importer-exporter-HS6 triads representing US $9.62 trillion of trade. Country characteristics (GDP, population) come from the World Development Indicators; geographical variables from CEPII.

Empirical Regularities Addressed

In a standard gravity decomposition of bilateral trade, the elasticities of total trade value with respect to importer and exporter GDP are both approximately 1.1 and the distance elasticity is approximately −0.81. Decomposing total value into its extensive, quantity, and price margins reveals: (i) the extensive margin of exporters rises strongly with exporter GDP (elasticity 0.76) but the corresponding importer extensive margin is much smaller (0.34), contrary to what the Eaton-Kortum (2002) model predicts; (ii) the unit-value margin rises with both importer GDP per capita (elasticity approximately 0.13 in product-level regressions controlling for exporter-product fixed effects) and exporter GDP per capita (elasticity approximately 0.22 controlling for importer-product fixed effects); (iii) there is no significant interaction between importer and exporter per capita income in bilateral trade values (coefficient 0.002, statistically insignificant), rejecting the Linder-type prediction from one-dimensional quality models that rich countries disproportionately sell to other rich countries.

Model

Building on the Ricardian EK framework with a continuum of varieties, CES aggregation, and perfect competition, the paper introduces two dimensions of quality:

Vertical quality (q) complements quantity: as spending on a variety increases, both physical quantity and vertical quality rise. This drives the positive relationship between importer per capita income and unit values, because buyers in richer (higher-wage) countries optimally demand higher vertical quality.
Horizontal quality (Q) perfectly substitutes for quantity and is determined by the producing country’s endowment of intermediates per worker. Because a better-equipped worker produces higher horizontal quality, this dimension rises with the exporter’s wage, explaining why richer countries charge higher unit values.

The model uses Fréchet-distributed productivities as in EK. Despite the non-homothetic intricacies introduced by the two quality dimensions, the trade-share equation is identical to EK’s homothetic formulation, and the welfare formula takes the standard ACR form with the elasticity of real income with respect to the home trade share equal to −1/(α̃θ).

To accommodate the extensive margin, the paper introduces stochastic minimum shipment sizes: small-value flows are observed probabilistically, generating zeros in the trade matrix. Products are treated as bundles of varieties, and the number of varieties per product follows a discretized Weibull distribution.

Estimation and Key Parameter Values

Multilateral resistance terms (Φ) are estimated from bilateral trade flow regressions. Using product-level unit values and wages/Φ estimates, the authors estimate three structural parameters: γ = 0.13 (cost elasticity of vertical quality, governing how spending splits between quantity and price), ν = 0.22 (elasticity of horizontal quality with respect to intermediate use), and θ = 4 (Fréchet shape parameter, calibrated from the literature as it is imprecisely identified from prices). From IV regressions of product-level spending on unit values — instrumenting a given destination’s price with the same exporter’s average price to other destinations — the implied demand elasticity with respect to price is −2.83, and the corresponding β (governing the distribution of the structural error across varieties) is estimated at 0.65. Three shipment-size parameters (λ₁ = 2.26×10⁻⁷, λ₂ = 0.042, λ₃ = 0.48) are fitted to match the observed bilateral extensive margins (R-squared 0.79).

Main Findings

A simulation of five million varieties, aggregated into approximately 3,807 traded products, reproduces the key margins of trade in the data. The model with only seven parameters (γ, ν, θ, β, λ₁, λ₂, λ₃) captures: (i) the positive relationship between unit values and both importer and exporter per capita income; (ii) the concave relationship between GDP and the extensive margin (leveling off for large countries); (iii) the standard gravity elasticities of bilateral trade on GDP and distance. Two discrepancies remain: the model understates the effects of per capita income on the extensive margin (shifting them toward the quantity margin), and it does not generate the Alchian-Allen distance effect on unit values.

Disaggregation to the level of 15 HS sections confirms the pooled results: 80% of HS6 products show positive importer-income elasticities and 94% show positive exporter-income elasticities of unit values. Although section-specific γ and ν estimates are formally rejected to be equal (χ²(28) = 1,249 against a critical value of 41), the implied improvement in model fit is only 3.6% of the total sum of squared residuals, vindicating the aggregate approach.

In depth

Q1. What are the two dimensions of quality in the model, and why does the paper require both?

A1: Horizontal quality (Q) substitutes perfectly for physical quantity and is valued identically by all users — a worker equipped with more intermediates produces goods of higher horizontal quality. Vertical quality (q) complements quantity: the isobeneft surface requires a CES aggregator with ρ < 0 (elasticity of substitution below one between effective quantity and vertical quality) so that a buyer spending more on a variety optimally raises both the amount and the vertical quality dimension. One dimension of quality is insufficient: with a single dimension, if rich countries both produce and prefer higher-quality goods, market shares of rich exporters should be systematically higher in rich importing destinations than in poor ones. No such interaction is found in the data (the interaction coefficient in bilateral trade regressions is 0.002, statistically insignificant), requiring the two-dimension structure to break the link.

Q2. How does the model predict that unit values rise with importer per capita income?

A2: The optimal spending on vertical quality for any variety is governed by the elasticity γ/(1+γ): as a buyer’s wage rises, spending on any variety rises, and the fraction of that spending that goes to higher unit values (price) has elasticity γ/(1+γ) with respect to spending. The structural elasticity of unit values with respect to the importer wage is δ_{w,M} = γ/(1+γ). With γ = 0.13, this equals approximately 0.115, close to the empirically estimated importer per capita income elasticity of 0.12–0.13 from product-level regressions with exporter-product fixed effects. Richer importers also face a higher price index Φ (lower competition), contributing an additional negative elasticity δ_{Φ,M} = −1/[θ(1+γ)] on Φ, reinforcing the unit-value–income gradient.

Q3. How does the model predict that unit values rise with exporter per capita income?

A3: In the model, horizontal quality Q is produced by equipping workers with intermediates: Q = m^ν where ν > 0 and m is intermediate use per worker. Because m is determined by the optimal factor mix and rises with the wage (w), horizontal quality rises endogenously with the exporter’s wage. The structural elasticity of unit values with respect to the exporter wage is δ_{w,X} = ν/(1+γ). With ν = 0.22 and γ = 0.13, this equals approximately 0.195, consistent with the estimated exporter per capita income elasticity of 0.20–0.22. The exporter Φ contributes δ_{Φ,X} = ν/[θ(1+γ)] > 0.

Q4. Why does the model still deliver a standard gravity equation despite the non-homothetic quality structure?

A4: The key result is that the trade-share equation — the fraction of varieties that destination n sources from country i — takes the same Fréchet-based form as in Eaton-Kortum (2002): πni = Ti(dni C̃i)^{−θ} / Φn, where C̃i = Ci/Qi is the horizontal-quality-adjusted unit cost. Although quality is non-homothetic in individual variety demands, the distribution of the maximum effective inverse cost across sources conditional on country i being the cheapest is independent of the source country — the key aggregation property inherited from the Fréchet structure. As a result, country i’s share in total absorption by n equals its share in the number of varieties sourced from i, and aggregate bilateral trade flows satisfy a standard log-linear gravity equation. The gains from trade are given by the standard ACR formula Un = constant × (Tn d^{−θ}_{nn} / πnn)^{1/(α̃θ)}.

Q5. How does the paper handle the extensive margin empirically, and what does the data show?

A5: The extensive margin is defined as the fraction of HS6 product categories that destination n imports from source i. In panel regressions, the elasticity of the extensive margin with respect to exporter GDP is 0.76 (much less than the total trade value elasticity of 1.16, implying an intensive margin of 0.38), while the importer extensive margin elasticity is 0.34. Both elasticities display a concave relationship with GDP in levels: the range of products both exported and imported expands rapidly for small countries but levels off at high GDP. The standard EK model predicts an importer extensive margin elasticity that is zero or negative (larger importers source more domestically), inconsistent with the positive 0.34 found in the data.

Q6. How does the paper model the extensive margin, and what are the parameter estimates?

A6: The extensive margin arises from stochastic minimum shipment sizes. Trade flows for individual varieties exist according to model-implied values, but are only observed in a given year if the flow exceeds the stochastic shipment size drawn from an exponential distribution H(x) = 1 − exp(−λ₁x). Products are treated as bundles of varieties drawn from a discretized Weibull distribution f(M) parameterized by λ₂ and λ₃. The three parameters are estimated by minimizing squared differences between model-predicted and observed bilateral extensive margins across all country pairs. The estimates are λ₁ = 2.26×10⁻⁷ (SE 1.21×10⁻⁷), λ₂ = 0.042 (SE 0.020), λ₃ = 0.48 (SE 0.10), with an R-squared of 0.79. These imply a mean shipment size of $4.42 million (median $3.07 million) and a mean number of varieties per product of 1,597 (median 344).

Q7. How is β estimated, and what does it govern?

A7: β governs how spending is distributed across varieties — specifically, the elasticity of spending on a variety with respect to its effective inverse cost. It also governs the elasticity of physical demand with respect to price: the model implies that log spending on a product equals log value minus (β/(1−β)) times log unit price. To identify β, the authors regress product-level trade values on unit prices, instrumenting a given importer’s price for a product with the same exporter’s average price of that product to all other destinations. The IV estimate of −β/(1−β) is −1.83 (SE 0.019), compared with the OLS estimate of −0.25, indicating substantial simultaneity bias. The implied price elasticity of demand is −2.83 and the implied β is 0.65.

Q8. What are the price regularities documented at the product level, and how does the two-quality model explain price overlaps across country pairs?

A8: Regressions of unit values at the importer-exporter-HS6 level show that individual exporters charge systematically higher prices to richer importers for the same product (elasticity 0.12 with exporter-product fixed effects), and that buyers pay systematically higher prices for products from richer exporters (elasticity 0.22 with importer-product fixed effects). A one-dimensional quality model would predict no overlap between prices charged by a rich and a poor exporter across destinations: even Japan’s lowest-priced sales should exceed Malaysia’s highest-priced sales for the same product. Back-of-envelope calculations using the regression coefficients predict a Malaysian product should sell in Norway at 0.3 log points above a Japanese product in Pakistan — systematic overlap. The paper documents this overlap in the raw data using two HS6 examples: motorcycle hubs (HS871493) and washing machines under 10kg (HS845011). The two-quality model resolves this by making horizontal quality an exporter attribute that raises prices proportionally but leaves market share determination to the EK gravity equation, allowing rich and poor country exporters to coexist in all markets.

Q9. How does the model address the absence of a Linder-type income interaction effect in aggregate trade flows?

A9: In one-dimensional quality models (e.g., Fajgelbaum, Grossman, and Helpman 2011), rich countries produce high-quality goods appealing primarily to high-income households, so rich-to-rich bilateral trade flows should be systematically higher than rich-to-poor flows. A gravity regression of bilateral trade values on importer and exporter fixed effects, distance, and an interaction of log importer GDP per capita × log exporter GDP per capita yields a coefficient of 0.0020 (SE 0.016), which is small and statistically insignificant. The two-quality model is consistent with this: horizontal quality enters as an exporter fixed effect (it affects prices proportionally for all destinations) and the demand system is structured so that all destinations spend the same share of absorption on a given source’s varieties, regardless of income level.

Q10. How robust are the results to disaggregation by industry?

A10: For each of 4,786 HS6 products with more than 20 country pairs, separate price regressions are estimated. Across all products, 80% have positive importer per capita income elasticities and 94% have positive exporter per capita income elasticities. The 15 broad HS sections account for only 10% of the variance in importer-income elasticities and 13% of the variance in exporter-income elasticities across HS6 products, suggesting high within-industry heterogeneity. A quasi-likelihood ratio test formally rejects equal γ and ν across sections (χ²(28) = 1,249 against a critical value of 41), but the reduction in total sum of squared residuals from allowing section-specific parameters is only 3.6%, and the R-squared increases from 0.353 to 0.376. The authors conclude the aggregate approach is vindicated for the purpose of characterizing common patterns.

Q11. How does the model simulate trade, and how many products does it generate?

A11: The simulation draws productivities for 5,000,000 varieties across 50 countries using the estimated model parameters (θ = 4, γ = 0.13, ν = 0.22, β = 0.65, and the estimated gravity fixed effects). For each variety, the cheapest source is determined; trade values and unit values are computed using equations (28) and (29); censoring due to stochastic shipment sizes generates zeros. Varieties are aggregated into products by partitioning sequentially using the estimated Weibull distribution. The simulation yields 3,842 total simulated products of which 3,807 are traded between at least one country pair, compared with 4,973 HS6 products in the COMTRADE data. A Monte Carlo exercise confirms that the estimation procedure recovers parameter values close to the true values when applied to simulated data.

Key Concepts

Vertical quality (q): A dimension of quality that complements physical quantity. In the paper’s utility specification, vertical quality and effective quantity enter a CES aggregator with elasticity of substitution below one (ρ < 0). A buyer spending more on a variety raises both quantity and vertical quality simultaneously, in proportions governed by γ. Vertical quality rises endogenously with the importer’s wage because higher-income buyers optimally demand it; it is the mechanism behind the positive relationship between importer per capita income and unit values.

Horizontal quality (Q): A dimension of quality that substitutes perfectly for physical quantity (enters the aggregator multiplicatively with quantity). All buyers value an increase in Q equivalently regardless of income level, so it does not generate Linder-type income-matching in trade flows. Horizontal quality is produced by the exporter: better-equipped workers produce higher horizontal quality (Q = m^ν), so it rises with the exporter’s wage. It is the mechanism behind the positive relationship between exporter per capita income and unit values.

Extensive margin (E_{ni}): In the paper’s empirical framework, the fraction of HS6 product categories that destination n imports from source i in a given year. The paper shows this margin rises with both importer and exporter size but in a concave, nonlinear fashion. It is generated in the model by stochastic minimum shipment sizes that probabilistically censor small-value variety flows.

Intensive margin: Total bilateral trade value divided by the extensive margin. The paper further decomposes the intensive margin into a quantity margin and a unit-value (price) margin. The paper’s key contribution is to generate all three margins jointly from one parsimonious framework.

Stochastic minimum shipment size: A modeling device, drawn from distribution H(x) (parameterized as exponential with parameter λ₁), that determines whether a given variety’s trade flow is observed in any year. If the annual flow x_{ni}(ω) exceeds the drawn minimum size x̄, the shipment is observed with certainty; otherwise, it is observed with probability x_{ni}(ω)/x̄. This mechanism generates the concavity of the extensive margin with respect to GDP without departing from the standard gravity framework.

Effective inverse cost (v_{ni}): Defined as Z_i(ω)/[C̃_i d_{ni}], where Z_i is country i’s Fréchet-distributed productivity for variety ω, C̃_i = C_i/Q_i is the horizontal-quality-adjusted unit cost, and d_{ni} is the iceberg trade cost. A buyer in n sources variety ω from the country maximizing v_{ni}. This formulation ensures that horizontal quality differences across exporters are absorbed into the effective cost, preserving the EK aggregation result.

γ (vertical quality cost elasticity): The parameter governing how spending on a variety divides between physical quantity and vertical quality. Spending has elasticity 1/(1+γ) with respect to quantity and elasticity γ/(1+γ) with respect to unit price. The paper estimates γ = 0.13 from product-level unit value regressions.

ν (horizontal quality elasticity): The parameter governing how horizontal quality rises with intermediate use per worker: Q = m^ν. Combined with γ, it determines the structural elasticity of unit values with respect to exporter per capita income: δ_{w,X} = ν/(1+γ). The paper estimates ν = 0.22.

The Social Tax: Redistributive Pressure and Labor Supply

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1 — Overview

Research Question

This paper asks whether informal redistributive pressure — the social obligation to share earned income with kin and social networks — distorts labor supply in low-income communities. The authors conceptualize such pressure as a “social tax” on earnings and develop the first direct causal test of whether it reduces labor supply, output, and earnings among full-time workers.

Setting and Sample

The study works with 474 full-time piece-rate factory workers (464 of whom are women) employed in cashew processing plants run by Olam in Côte d’Ivoire. Workers are paid biweekly in cash entirely through piece rates for individual nut-peeling output, creating a direct mapping between labor supply and income. At baseline, workers report transferring 25–35% of their income to individuals outside their household, with 77% having made at least one transfer in the previous 3 months. Workers also strongly believe that earning more triggers more transfer requests: 77% agree that if someone starts earning more by working harder, people will ask that person more often for financial help.

Intervention

The authors introduce a blocked savings account into which workers can deposit any earnings above a self-chosen threshold (set at least as high as their own baseline average earnings). Earnings above the threshold are automatically deposited by the factory directly into the account with the Banque Populaire de Côte d’Ivoire; the cash component of pay is unchanged. Funds cannot be withdrawn until the end of the blocked period (9 months in Phase 1; 3 months in Phase 2). The key design feature is that the account reduces the effective social tax rate only on earnings increases above baseline, thereby eliminating income effects and generating only a pure substitution effect — an unambiguous positive prediction on labor supply if a social tax exists.

Experimental Design

Workers are randomized into three conditions: (1) Control (no account); (2) Private account (existence unknown to anyone outside the worker); (3) Non-private account (existence and forthcoming unblock date revealed to network members via promotional text messages). The contrast between Private and Non-private isolates the role of redistributive pressure specifically — holding constant all other features of the blocked account product. The experiment runs in two cross-randomized phases conducted between 2018 and 2019.

Main Findings

Take-up of blocked accounts is dramatically higher when accounts are private: 60% in Phase 2 (Private) versus 14% (Non-private), a 77% decline (p<0.001). Among workers who declined Non-private accounts, 96% cite anticipated increases in transfer requests as an important factor.

Being offered a Private account sharply raises labor supply. Pooling both phases, the Private arm increases average daily earnings by 175.9 FCFA, or 11.4% (p=0.012), relative to Control or Non-private arms. This is accompanied by a 6.2 percentage point (9.7%) increase in daily work attendance (p=0.023), with the entire attendance effect driven by reduced absenteeism rather than turnover. Effects in Phase 1 (Private vs. Control: +11.3%, p=0.032) and Phase 2 (Private vs. Non-private: +11.5%, p=0.043) are nearly identical in magnitude, indicating the results are not sensitive to cross-phase design. The treatment effect magnitude is equivalent to each worker working an additional 1.19 days in every two-week paycycle. Because 89% of workers have no income outside the factory, these constitute increases in total earned income.

Heterogeneity is consistent with the hypothesized mechanism: among workers who report difficulty saving due to redistributive pressure, the Private treatment increases earnings by 15.0% (p=0.018); among those not reporting such difficulty, the estimated effect is near zero and insignificant (p=0.95). Among workers who report transfers to acquaintances (the most likely social-tax-motivated transfers), the effect is 17.5% (p=0.014). Workers without a partner — for whom intra-household redistribution is irrelevant — experience a 15.8% earnings increase (p=0.017), indicating that extra-household pressure drives the results.

Outgoing transfers do not decline. The design leaves cash-on-hand unchanged by construction, and consistent with this, there is no significant change in the likelihood or amount of transfers from treated workers to their networks. Total outgoing transfers are if anything higher among Private account workers (p=0.049), suggesting no loss in redistribution to the network.

Social Tax Rate Estimation

Combining the 11.4% treatment effect on output with a labor supply elasticity estimated from an end-of-experiment piece-rate randomization (intensive-margin elasticity of 0.17; total elasticity of approximately 1.11), the authors estimate the social tax rate for the average worker in the sample at 9–14%. For the subset who actually take up Private accounts, the implied social tax rate is 19–23%.

Scope Conditions

Results pertain to full-time female piece-rate workers in formal cashew processing plants in Côte d’Ivoire, with average tenure of 1.7 years. Because the intervention lowers the tax only on earnings above baseline (not on all earnings), the estimates do not directly capture the total distortion from eliminating all redistributive pressure. Alternative confounds — fairness/morale effects, self-control, privacy concerns, goal-setting — are each tested and ruled out as primary drivers.

In depth

Q1. What is the theoretical basis for predicting that Private accounts unambiguously increase labor supply?

The authors model redistributive pressure as a social tax rate τ₁ on gross earnings. The blocked account reduces this tax to τ₂ < τ₁ only on earnings above baseline labor supply e₁, creating a kink in the budget constraint. Starting from e₁, the worker faces only a pure substitution effect (no income effect) when τ₂ falls, because her net earnings at e₁ are unchanged. Equation (2) in the paper shows formally that the income effect term drops out, and the derivative of labor supply with respect to τ₂ is unambiguously negative (i.e., reducing τ₂ increases effort). This “clean” prediction — no income effect, no ambiguity — is the central design advantage relative to simply shielding existing earnings.

Q2. How do take-up rates differ between Private and Non-private accounts, and what do workers say explains the difference?

In Phase 2, take-up of Private accounts is 60% versus only 14% for Non-private accounts — a 77% reduction (p<0.001). Among workers who declined a Non-private account, 96% cite the anticipation of increased transfer requests from network members knowing about the account as an important factor in their decision. Only 5% cite any other reason. This pattern is strong direct evidence that the fear of redistribution — not other features of the accounts — drives take-up differences.

Q3. What are the treatment effects on earnings and attendance, and how consistent are they across phases and subsamples?

Pooled across both phases, the Private arm raises daily earnings by 175.9 FCFA (11.4%, p=0.012) and attendance by 6.2 percentage points (9.7%, p=0.023). In Phase 1 alone (Private vs. Control), earnings rise 11.3% (p=0.032). In Phase 2 alone (Private vs. Non-private), earnings rise 11.5% (p=0.043). Restricting to workers not previously treated in Phase 1, the effect is 12.8% (p=0.034); restricting further to workers new to the study in Phase 2 only, the effect is 17.3% (p=0.020). The authors cannot reject that effects across these three Phase 2 subsamples are statistically the same (p=0.427), ruling out sensitivity to the cross-randomized design.

Q4. How does treatment effect heterogeneity support the redistributive pressure mechanism?

Workers who report difficulty saving because “someone else will need it for something urgent” see earnings increase by 15.0% (p=0.018) from the Private treatment; those not reporting this difficulty see near-zero, insignificant effects (p=0.95). Workers who make transfers to acquaintances — transfers especially unlikely to reflect altruism — see earnings rise 17.5% (p=0.014). Workers with below-median baseline earnings, potentially those facing the strongest relative disincentive to work, see larger effects. Each of these heterogeneous patterns is in the direction predicted if the social tax is the operative mechanism.

Q5. Do the treatment effects reflect substitution away from outside earnings or genuine total income gains?

No. The paper finds no treatment effects on earnings outside the factory. At baseline, 89% of workers report zero outside earnings, and on average 93% of total income comes from factory wages. Consequently, the 11.4% earnings increase represents a near-one-for-one increase in total earned income.

Q6. Do Private accounts reduce transfers to the network?

No. The design ensures that cash-on-hand is unchanged by construction — workers receive the same or slightly higher take-home cash pay (the difference is positive but insignificant). Consistent with this, neither the probability of making transfers (p=0.37) nor transfers to family (p=0.35) or non-family (p=0.93) change significantly. Total outgoing transfers in the endline survey are if anything higher in the Private arm (p=0.049, though this may partly reflect redistribution of unblocked savings). The net transfer amount is positive but insignificant (p=0.32). The authors conclude the intervention did not make others in workers’ networks worse off.

Q7. How do the authors rule out morale or fairness effects as an explanation?

Treatment assignment was conducted by lottery with ID numbers drawn in front of workers, clearly dissociating it from employer favoritism. More directly, the authors test for morale effects using the 3–4 week “announcement period” between treatment disclosure and account activation. If disgruntlement among non-Private workers drove results, output should fall during this period — but estimated announcement effects are near zero (0.8% of control mean, p=0.859 in Phase 2). In contrast, effects arise immediately in the first active paycycle: earnings jump 11.4% (p=0.082) even before workers have seen any deposits occur. The fairness story also cannot explain why effects are concentrated precisely among workers who report more redistributive pressure.

Q8. How do the authors test and rule out self-control as the primary mechanism?

Self-control cannot explain why Non-private accounts — which offer the same commitment benefit — have dramatically lower take-up than Private accounts. Separately, the authors test a core prediction of time inconsistency models by surprising workers with an option to opt out of the next deposit, randomly varying whether the offer comes 4 days before payday or on payday itself. Under quasi-hyperbolic preferences, workers should be more likely to opt out on the payday itself. Counter to this prediction, 94% of workers keep their earnings in the account on payday, compared to 86% four days before — and these means are not statistically distinguishable, with the relative magnitudes actually running opposite to time inconsistency predictions.

Q9. How do the authors address the concern that Non-private accounts may raise the tax rate above the baseline, inflating treatment effect estimates?

The concern is that Non-private SMS alerts could make network members more aware of available cash than under the status quo, pushing the effective comparison above the Control level. The authors note that (a) paydays are already publicly known in this setting and workers regularly face transfer requests around them; (b) workers must physically withdraw savings from a bank after the unblock date, and can even re-block funds; and (c) the magnitude of effects when comparing Private to Control is nearly identical to the effect when comparing Private to Non-private (11.3% vs. 11.5%), suggesting the Non-private condition does not materially raise the tax above the status quo.

Q10. How do the authors rule out privacy concerns (rather than redistributive pressure) as the driver of low Non-private take-up and treatment effects?

Four arguments are provided. First, Phase 1 effects (Private vs. Control, no Non-private arm) are the same magnitude as Phase 2 effects, yet Phase 1 cannot be confounded by privacy concerns. Second, among workers who refused Non-private accounts, 96% cite transfer request anticipation; none volunteer generic privacy concerns. Third, heterogeneity effects — concentrated among high-redistributive-pressure workers — have no obvious connection to privacy preferences. Fourth, two placebo SMS exercises: 95% of Non-private workers grant permission to send generic bank promotional texts, and 88% of workers who had Phase 1 Private accounts grant permission for messages about their past (already-spent) savings — indicating no inherent aversion to having some financial information shared with networks. Since these workers forgo 11.5% of full-time earnings by refusing Non-private accounts, privacy concerns alone are implausible as a full explanation.

The authors combine the 11.4% ITT treatment effect (used as the ratio e₁/e₂) with a compensated labor supply elasticity ζ estimated from an end-of-experiment piece-rate randomization. The piece-rate experiment (varying piece rates over four values from −15% to +30% of baseline over 6 days) yields an intensive-margin elasticity of 0.17. Using the ratio of attendance to intensive-margin effects from Table 3, the implied extensive-margin elasticity is 0.94, giving ζ ≈ 1.11. With this elasticity and assuming τ₂ = 0 (most conservative), the ITT-implied social tax rate is 9%; assuming τ₂ = 5%, it is 14%. For compliers (workers who actually take up Private accounts), the estimated rate is 19–23%. If instead the lower elasticity estimate of 0.32 (comparable to Goldberg 2016) is used, the ITT tax rate would be at least 29%.

Q12. What are the broader implications discussed by the authors?

The authors propose that if redistributive pressure distorts work incentives, it may also distort other costly income-generating actions: technology adoption, human capital investment, and formal sector participation. They note that 74% of workers believe taking a formal job would increase transfer requests, even though network members could also access such jobs. A speculative but highlighted policy implication is that formal safety nets (health or unemployment insurance) could reduce social tax burdens on non-recipients by absorbing demand for redistribution, potentially generating positive productivity externalities.

Key Concepts

Social Tax: The paper’s central concept. Redistributive pressure from kin and social networks is modeled as a tax rate τ₁ on gross earnings — not altruistic transfers, but transfers made under social pressure that workers would prefer to avoid. The “tax” analogy captures that the obligation is proportional to visible income and reduces the private return to earning more. The paper explicitly does not take a stance on the underlying microfoundation (risk-sharing, cultural norms, or a mix).

Blocked Savings Account: A date-based savings account (implemented with Banque Populaire de Côte d’Ivoire) into which any earnings above a worker-chosen threshold are automatically deposited by the factory. Funds are inaccessible until the blocked period ends (3–9 months). Workers cannot withdraw during the period, making deposited earnings unavailable to fulfill transfer requests and therefore effectively reducing the social tax rate on earnings increases.

Private vs. Non-private Treatment: The paper’s key experimental contrast. A Private account’s existence is unknown to anyone in the worker’s network. A Non-private account triggers SMS messages to network members disclosing that the worker is saving and announcing when the unblock date approaches. The contrast isolates whether the shielding of income from social visibility — not the commitment device per se — drives take-up and labor supply.

Substitution Effect without Income Effect: The paper’s design deliberately places the tax reduction only on earnings above baseline, creating a kink in the budget constraint. Starting from the existing labor supply level, there is no change in net earnings at the margin — eliminating the income effect of a tax reduction — so any labor supply response is a pure compensated (substitution) effect. This makes any observed increase in labor supply an unambiguous signal that a distortionary social tax exists.

Intent to Treat (ITT) vs. Treatment on the Treated (ToT): The ITT estimate (11.4% earnings increase) reflects the effect of being offered a Private account on all offered workers, including those who did not take up. The ToT estimate — relevant for workers who actually used the accounts — implies a higher social tax rate (19–23%) because only roughly half of offered workers take up the accounts and only those workers face a materially reduced effective tax rate.

Compensated (Hicksian) Labor Supply Elasticity (ζ): The ratio used to infer the social tax rate from the observed treatment effect. The paper estimates ζ ≈ 1.11 (extensive margin ζₐ ≈ 0.94, intensive margin ζₑ ≈ 0.17) from an end-of-experiment piece-rate randomization. The social tax rate is recovered as τ₁ = 1 − (1−τ₂)(e₁/e₂)^(1/ζ) from Equation (5).

Piece Rate Setting: Workers earn a linear piece rate for every kilogram of cashews peeled, with no fixed pay component. This setting ensures that every unit of additional effort by a worker translates directly into higher earnings, and that any observed earnings changes cleanly reflect labor supply responses rather than hour or schedule effects.

Uniform Priors for Impulse Responses

Mon, 01 Jan 0001 00:00:00 +0000

Structural vector autoregressions (SVARs) identified with sign restrictions are a widely used tool for estimating dynamic causal effects in macroeconomics. Critics—notably Baumeister and Hamilton (2015) and Watson (2020)—have called for caution because the standard practice of using a uniform prior over the set of orthogonal matrices (with respect to the Haar measure) induces non-uniform marginal prior distributions over the identified sets of individual impulse responses. This paper formally challenges that caution: through an if-and-only-if theorem the authors show that the uniform prior over orthogonal matrices is not only sufficient but also necessary to induce a uniform joint prior distribution over the identified set for the vector of impulse responses—a result that holds for any prior distribution over the reduced-form parameters. The paper additionally shows how to conduct posterior inference based on a uniform joint prior for the vector of impulse responses, which requires modifying the prior for the reduced-form parameters away from the standard Minnesota prior while retaining the uniform prior over orthogonal matrices. An application to Watson’s (2020) empirical example finds that joint credible sets under this new prior are similar to, but wider than, those obtained under the standard approach, and that imposing tighter identifying restrictions sharpens inference under both priors.

In depth

Q1. What is the core result and what does it imply for applied work?

The central result is an if-and-only-if theorem: the uniform prior over the set of orthogonal matrices is both sufficient and necessary for the conventional Bayesian approach to induce a uniform joint prior distribution over the identified set for the vector of impulse responses, for any prior over the reduced-form parameters. The critics’ concern about non-uniform individual marginal priors does not extend to the joint object: when inference targets the full vector of impulse responses, the standard Haar prior is exactly appropriate. Practitioners interested in joint inference on the shape and comovement of the impulse response function need not heed the call for caution.

Q2. Why does non-uniformity of individual marginal priors not imply non-uniformity of the joint distribution?

The marginal distribution extracted from a uniform joint distribution over a compact manifold need not be uniform; marginal uniformity and joint uniformity are different properties, and only the latter is required for observationally equivalent vectors to be distinguished solely by the identifying restrictions. Baumeister and Hamilton (2015) and Watson (2020) correctly note that individual impulse responses have non-uniform marginal priors under the Haar measure, but this is not the relevant criterion when the object of interest is the entire impulse response vector. The paper’s theorem shows the joint distribution is uniform, which is the property that ensures the identification restrictions—not the prior—drive the posterior shape.

Q3. How does one implement a uniform joint prior for the vector of impulse responses?

The authors show that a uniform joint prior for the vector of impulse responses requires a modified prior for the reduced-form parameters: one that is independent between (B, Σ) and Q, takes a model-dependent non-standard form for (B, Σ), and retains a uniform prior over orthogonal matrices. The induced reduced-form prior resembles but differs from both the standard Minnesota prior and Uhlig’s (2005) “weak prior.” Because the induced prior for (B, Σ, Q) is still a uniform-normal-inverse-Wishart (UNIW) distribution, the conventional sampling algorithm applies without modification; analysts supply the modified reduced-form prior while continuing to draw Q uniformly from the Haar measure.

Q4. What does the empirical illustration show?

In Watson’s (2020) empirical example, joint credible sets under the uniform-joint-prior approach are similar to but wider than those under the standard Minnesota-prior approach. The widening is consistent with theory: the uniform joint prior spreads probability mass more evenly over the identified set rather than concentrating it toward regions favored by the Minnesota prior. The finding that tighter identifying restrictions sharpen inference under both approaches reinforces the conclusion of Inoue and Kilian (2022b) that many sign restrictions help when the focus is on joint distributions.

Q5. How is the analysis generalized?

The paper extends the results to a broader class of objects of interest—any smooth function of impulse responses, such as combinations of structural elasticities and standard deviations—with an importance-sampling correction when the induced prior over orthogonal matrices is not uniform in the extended case. The generalization exploits the diffeomorphism between IR parameters and orthogonal reduced-form parameters, which allows the change-of-variables formula to apply to any smooth object of interest.

Key concepts

vector of impulse responses : the collection of impulse responses across all variables, shocks, and horizons, treated as a single vector object for joint inference; contrasted with individual impulse responses (the response of one variable to one shock at one horizon).

uniform prior over orthogonal matrices (Haar measure) : the unique probability measure on the set of n×n orthogonal matrices invariant under left and right multiplication; the standard prior used in Bayesian sign-restricted SVARs.

identified set : the set of vectors of impulse responses that are observationally equivalent given the data and the sign restrictions; the conventional approach draws uniformly from this set under the Haar prior.

uniform-normal-inverse-Wishart (UNIW) prior : the joint prior over orthogonal reduced-form parameters consisting of the Haar prior over Q and a normal-inverse-Wishart prior over (B, Σ); conjugate and computationally tractable.