Monetary financing produces neither high inflation nor miraculous fiscal multipliers

Layer 1: Overview

When central banks pay interest on reserves — as the Federal Reserve has done since October 2008 and as is standard operating procedure today — does financing fiscal stimulus by permanently expanding the central bank’s balance sheet produce higher output than debt-financed stimulus? Van der Kwaak (2024) argues the answer is no in most model configurations, and only modestly yes in a specific extension.

The motivation is practical: with government debt at high levels in many advanced economies, the private sector may be unable or unwilling to absorb additional bonds needed to fund fiscal stimuli. One alternative is monetary financing — the central bank permanently purchases the extra bonds issued to fund the stimulus (as proposed by Gali 2020b for COVID-era policy). A prior key paper (Gali 2020a) found money-financed stimuli to be substantially more effective than debt-financed ones, but that result was derived in a model where the central bank does not pay interest on reserves, so the policy rate becomes endogenous under money financing. Van der Kwaak shows this assumption is at odds with how modern central banks operate: post-GFC balance sheet expansions by the Federal Reserve and ECB have been financed almost entirely by interest-bearing reserves, with non-interest-paying currency showing no meaningful deviation from trend.

The paper employs a New Keynesian DSGE model with labor as the sole production factor, a central bank that holds government bonds funded by non-interest-paying money and interest-paying reserves (with the composition endogenous), financial intermediaries subject to a Gertler-Kiyotaki (2010) / Gertler-Karadi (2011) incentive-compatibility leverage constraint on bond holdings, and a standard active Taylor rule bounded by the ZLB. Fiscal stimulus takes the form of either (i) a lump-sum tax cut or (ii) an increase in government spending, each equal to 1% of steady-state output. Money financing is modeled as the central bank acquiring the additionally issued bonds and retaining them permanently in nominal terms.

The central analytical result (Proposition 1) is a proof of “extended Ricardian equivalence”: the consolidated government’s funding mix among money, reserves, government bonds, and lump-sum taxes has zero effect on inflation and the equilibrium allocation in the real economy. This holds whether or not the incentive-compatibility constraint of financial intermediaries is binding — that is, even when bonds and reserves are not perfect substitutes and money financing genuinely reduces the government’s funding costs. The key mechanism: because the central bank pays interest on reserves, the deposit rate equals the policy rate in equilibrium, and the policy rate is the sole endogenous variable on which households’ deposit return depends. As a result, household consumption-savings decisions are completely decoupled from the financing mix; inflation and real quantities are pinned down entirely by the standard NK equilibrium conditions plus the Taylor rule. Proposition 2 further shows that net cash flows between households and the government/financial sector ultimately just finance exogenous government expenditures, so changes in bond prices and lump-sum taxes produce no net wealth effects on households.

This irrelevance result is shown to extend analytically to: (i) the ZLB regime (since the central bank still controls the policy rate under money financing), (ii) any maturity structure of government debt, (iii) the ECB’s two-tiered reserve system (where minimum reserves earn zero and excess reserves earn the policy rate), (iv) ex ante sovereign default risk, (v) an alternative leverage constraint form (deposits capped relative to reserves plus a fraction of bonds), and (vi) a model with physical capital when corporate securities are held by unconstrained households.

The irrelevance breaks only when balance-sheet-constrained financial intermediaries also hold corporate securities financing the physical capital stock (Section 4.2 / Sims-Wu 2021 extension). In that case, central bank bond purchases under money financing compress bond yields, which via the intermediaries’ portfolio-choice condition also compresses expected returns on corporate securities, stimulating investment. The quantitative difference between money- and debt-financed stimuli, measured by the discounted cumulative fiscal multiplier over 1,000 quarters, is 0.26 — substantially smaller than the 0.50 difference found by Gali (2020a). For the spending stimulus, the debt-financed multiplier is 0.9103 and the money-financed multiplier is 1.1719, giving a money-over-debt advantage of 0.2616. For the tax cut, the debt-financed multiplier is -0.0219 and the money-financed multiplier is 0.2397, again a difference of 0.2616. The smaller advantage relative to Gali (2020a) reflects the fact that in Gali’s framework the policy rate is not controlled by the central bank under money financing, so households’ saving return falls endogenously and consumption expands sharply — an effect that is entirely absent here because the central bank retains full control of the policy rate.

The policy implication is that proposals to use monetary financing to achieve “miraculous” multipliers beyond the normal spending multiplier are misguided in modern institutional settings where central banks pay interest on reserves. Money financing avoids increasing private-sector-held debt but does not amplify macroeconomic stimulus relative to conventional debt financing in the baseline case, and offers only a small incremental boost in the more structured extension.

Layer 2: Deep Dive

What is the key analytical result and what is the formal proposition that establishes it?

Proposition 1 proves ’extended Ricardian equivalence’: the consolidated government’s funding mix among money, reserves, government bonds, and lump-sum taxes has zero impact on inflation and the equilibrium allocation in the real economy. The proof works by exhibiting a self-contained subset of equilibrium conditions — households’ first-order conditions for consumption, labor, and deposits; the Taylor rule; firms’ pricing conditions; and market clearing — that uniquely pins down all real quantities and inflation without including any equation governing the government’s or central bank’s financing mix. Because the deposit rate equals the policy rate in equilibrium (due to reserves not being subject to the incentive-compatibility constraint), households’ saving return depends only on inflation and real variables, so the funding mix drops out entirely.

Why does the irrelevance result hold even when the incentive-compatibility constraint of financial intermediaries is binding and bonds and reserves are NOT perfect substitutes?

When the constraint binds, reserves earn a lower return than bonds, so the central bank’s bond purchases do increase bond prices and reduce government funding costs — but these price changes generate no net wealth effects on households. Proposition 2 shows formally that all cash flows between households on one side and the government and financial intermediaries on the other ultimately just finance (exogenous) government expenditures on final goods. Changes in bond prices, intermediary dividends, and households’ bond and deposit returns cancel out in the household budget constraint, so W_t = g_t regardless of the financing mix. The intuition is that the financial sector and government together form a closed circuit relative to households, and because government spending is exogenous, the circuit’s net effect on household wealth is always the same.

How does this result differ from Gali (2020a), and why is the multiplier advantage of money financing larger in that paper?

Gali (2020a) assumes the monetary base consists solely of non-interest-paying money. In that setting, when the central bank permanently expands the monetary base to finance a fiscal stimulus, it cannot simultaneously control the policy rate and the money supply, so the policy rate becomes endogenous and falls relative to a debt-financed stimulus. This endogenous reduction in the rate at which households can save causes a substantial increase in consumption. In van der Kwaak’s framework, the central bank pays interest on reserves and retains full control of the policy rate regardless of whether the stimulus is debt- or money-financed, eliminating this consumption-expansion channel. As a result, Gali finds a money-over-debt multiplier advantage of 0.50, while van der Kwaak finds 0.26 in the one model extension where irrelevance is broken, and zero in the baseline.

In what model extension is the irrelevance result broken, and what is the mechanism?

The irrelevance breaks when balance-sheet-constrained financial intermediaries hold both government bonds and corporate securities (financing the physical capital stock), as in Sims and Wu (2021) and van der Kwaak (2023). In this configuration, the incentive-compatibility constraint links the expected excess returns on bonds and corporate securities through a fixed ratio lambda_b / lambda_k. When money financing causes the central bank to acquire additional bonds, bond prices rise and expected bond returns fall. Via the portfolio-choice optimality condition, this also compresses expected returns on corporate securities, which encourages investment. A direct link thus emerges from the government’s financing mix to the real economy through the financial sector’s balance sheet. Without this channel — whenever corporate securities are held by unconstrained households, or the model has no physical capital — the irrelevance holds exactly.

What are the exact quantitative multiplier results from the numerical exercise?

Using the discounted cumulative multiplier formula summed over 1,000 quarters (Table 2): (i) Debt-financed tax cut: -0.0219. (ii) Money-financed tax cut: 0.2397. Difference: 0.2616. (iii) Debt-financed spending stimulus: 0.9103. (iv) Money-financed spending stimulus: 1.1719. Difference: 0.2616. The money-over-debt advantage is identical (0.2616) for both types of stimulus, though the levels differ substantially. The debt-financed tax-cut multiplier is negative because higher bond issuance generates capital losses on intermediaries’ bond portfolios, tightening the incentive-compatibility constraint and reducing credit provision and investment. Money financing mitigates these losses by having the unconstrained central bank absorb the newly issued bonds, raising bond prices and net worth.

What robustness checks does the paper conduct on the irrelevance result?

The paper proves the irrelevance analytically for: (1) Both binding and slack incentive-compatibility constraints (Section 3.1). (2) Any maturity structure of government debt — the maturity parameter rho drops out of the relevant equilibrium conditions (Section 3.2.1). (3) The ZLB — since the central bank still controls the reserve rate even under money financing (Section 3.2.1). (4) An alternative leverage constraint where deposit capacity depends on reserves plus a discounted fraction of bonds rather than a fixed fraction of bond value (Appendix C.2). (5) The ECB’s two-tiered reserve system, where minimum reserves receive zero interest and excess reserves receive the policy rate; the deposit rate becomes (1-theta)*policy rate instead of the policy rate itself, but is still solely determined by the policy rate (Proposition 3, Section 3.2.2). (6) Models with physical capital when households hold the corporate securities (Proposition 4, Section 4.1). (7) Ex ante sovereign default risk following Corsetti et al. (2013) (Appendix C.1).

What is ’extended Ricardian equivalence’ as defined by the author, and how does it differ from the original Barro (1974) result?

Barro’s (1974) Ricardian equivalence shows that the funding mix between government debt and lump-sum taxes has zero effect on the real economy. Van der Kwaak extends this to include the monetary base — the funding mix among money, reserves, government bonds, and lump-sum taxes has zero impact on inflation and the real equilibrium. This is a strictly more general result because it covers the substitution of money/reserves for bonds (i.e., monetary financing), not just the substitution of debt for taxes. Crucially, the extension holds even when bonds and reserves are not perfect substitutes (when the incentive-compatibility constraint binds), which is the nontrivial part of the contribution.

How is ‘money financing’ modeled in the paper?

A money-financed stimulus is modeled as one in which the government bonds newly issued to fund the additional spending or the tax cut are acquired by the central bank and permanently retained on its balance sheet in nominal terms. For a spending stimulus, the parameter kappa_g = 1 means the central bank’s nominal assets expand by the amount of each period’s additional government purchases (g_t - g_bar). For a tax cut, kappa_tau = 1 means the central bank acquires bonds equal to the tax-cut component tau_tilde_t. Debt financing corresponds to kappa_g = 0 or kappa_tau = 0. The central bank’s dividends (profits net of interest on reserves and seigniorage on currency) are returned to the fiscal authority each period, so central bank net worth is zero. The author notes this is consistent with the legal constraints on central banks (Buiter 2014) since it takes the form of permanent QE rather than overt fiscal transfers.

What is the role of the incentive-compatibility constraint in generating the bond-price spread, and why does the irrelevance result still hold?

The Gertler-Kiyotaki constraint limits the volume of government bonds intermediaries can hold relative to their net worth (chi_t * n_t = lambda_b * q^b_t * s^{b,f}_t when binding). When binding, intermediaries cannot freely expand bond holdings in response to higher bond supply, so an increase in bond supply under a debt-financed stimulus depresses bond prices and creates capital losses. Conversely, the unconstrained central bank buying additional bonds under money financing raises bond prices. So the constraint creates a genuine price and funding-cost differential between money- and debt-financed stimuli. Yet the irrelevance still holds because, as shown in Proposition 2, these bond-price changes, together with changes in intermediary dividends, net out from the household budget constraint — the household sees the same net obligation regardless of financing mix.

How does Corollary 1 relate to the empirical observation about the monetary base composition?

Corollary 1 proves analytically that any expansion of the monetary base under money financing consists entirely of an expansion in interest-paying reserves — non-interest-paying money holdings are unchanged. This is because, in equilibrium, households’ demand for non-interest-paying money depends only on consumption and the nominal deposit rate (via the money-in-utility first-order condition), neither of which changes under money financing (by the irrelevance result). This directly matches the empirical evidence shown in Figures 1 and 4 for the Federal Reserve and ECB respectively: post-GFC balance-sheet expansions were almost entirely in interest-paying reserves, with currency in circulation showing no deviation from trend.

What is the tax-cut mechanism under debt financing in the numerical exercise, and why is the multiplier negative?

Under a debt-financed tax cut (kappa_tau = 0), the fiscal authority must issue more bonds to offset the revenue shortfall. Because financial intermediaries’ incentive-compatibility constraint is binding, they cannot perfectly elastically absorb the additional bond supply; bond prices fall, causing capital losses on intermediaries’ existing holdings. This reduces net worth, tightens the constraint further, and forces intermediaries to reduce lending to the real economy. The capital price and investment therefore fall. The trough in output is at most about 0.03% of steady-state output, but the cumulative multiplier is -0.0219 — negative because the adverse financial amplification from falling bond prices more than offsets any direct effect of the lump-sum transfer on households. This mechanism is similar to van der Kwaak and van Wijnbergen (2017).

What is the calibration strategy, and how closely does it follow Gali (2020a)?

The calibration of the model with financial intermediaries holding corporate securities follows Gali (2020a) for most household and production parameters: discount factor beta = 0.995, risk aversion sigma_c = 1, inverse Frisch elasticity phi = 5, price semi-elasticity of money demand eta = 7, Calvo probability psi_p = 3/4, elasticity of substitution epsilon = 9, labor share = 0.75, steady-state government debt / output = 2.4 (60% of annual GDP), AR(1) for government spending rho_g = 0.5. Deviations from Gali include: government spending share of output set at g_bar/y_bar = 0.2 (consistent with advanced economy averages), steady-state investment share i_bar/y_bar = 0.2, and a monetary base equal to 1/3 of quarterly output (as in Gali) now split into non-interest-paying money (10% of quarterly output) and interest-paying reserves (1.63 times currency). For financial intermediaries: average banker tenure 24 quarters (sigma = 0.9583), adjusted leverage ratio 5, steady-state spread on corporate securities and bonds over deposits = 25 quarterly basis points (100 annual basis points), implying lambda_b = lambda_k. Capital adjustment cost gamma_k = 2.5.

How does the paper relate to Wallace (1981) and when does the neutrality argument break down?

Wallace (1981) first showed that open-market operations are neutral in complete-markets models where all investors can purchase any asset at market prices without binding constraints. Woodford (2012) distills the key conditions: assets are valued only for pecuniary returns, and all investors face the same market prices with no binding position constraints. Van der Kwaak’s irrelevance extends the Wallace neutrality to incomplete markets with binding leverage constraints on bond holdings, which go beyond Woodford’s conditions. The neutrality breaks only when the binding constraint links together multiple asset classes — specifically when the same constraint covers both government bonds and corporate securities, creating a direct transmission from bond prices to the cost of capital.

How does the paper relate to Reis and Tenreyro (2022) on helicopter money?

Reis and Tenreyro (2022) study helicopter drops — direct transfers of newly created central bank liabilities to households — and derive an irrelevance result that applies only when bond and reserve interest rates are equal (perfect substitutes). Van der Kwaak’s irrelevance extends to the case where the return on bonds exceeds that on reserves (binding incentive-compatibility constraint). A second difference is that Reis-Tenreyro focus on helicopter money (a liability-side transfer), while van der Kwaak models money financing as permanent QE (an asset-side expansion). Third, van der Kwaak also studies money-financed government spending stimuli, which Reis-Tenreyro do not.

What are the implications for policy proposals to use monetary financing in high-debt environments?

The core message for policy is nuanced. On the fiscal side, monetary financing does achieve its main stated goal: it prevents private-sector-held government debt from rising, since the additional bonds are absorbed by the central bank. On the stimulus effectiveness side, however, money financing has no macroeconomic advantage over debt financing in the baseline model (and in most extensions). The one setting where there is an advantage — intermediaries holding both bonds and corporate securities — yields only a modest multiplier boost of 0.26 relative to debt financing, compared to the 0.50 suggested by Gali (2020a). This smaller number reflects the fundamental institutional difference: with interest-on-reserves, the policy rate stays fixed under money financing, eliminating the consumption-expansion channel. The paper also implies there is no inflationary danger from money financing in this setup — the irrelevance result holds for inflation as well as real variables — directly contradicting fears that monetary financing inherently produces high inflation.

What happens to inflation under money financing compared to debt financing in the analytical result?

The extended Ricardian equivalence result covers inflation explicitly: the path of inflation is identical under money financing and debt financing in all the analytical baseline cases. This is because inflation is pinned down by the New Keynesian Phillips curve and the Taylor rule, neither of which depends on the financing mix. The central bank retains full control of the policy rate under money financing (because it pays interest on reserves), so the Taylor rule continues to govern inflation dynamics. This directly contradicts the claim that monetary financing is inherently inflationary; in the model, it is neither inflationary nor expansionary relative to debt financing.

Key Concepts

Extended Ricardian equivalence: The author’s label for the proposition that the consolidated government’s funding mix among money, reserves, government bonds, and lump-sum taxes has zero effect on both inflation and the equilibrium allocation in the real economy. It extends Barro (1974)’s original Ricardian equivalence (which covered only debt vs. taxes) to include the monetary base, and holds even when bonds and reserves are not perfect substitutes due to binding intermediary leverage constraints.

Money-financed fiscal stimulus: In this paper’s modeling: a fiscal stimulus (tax cut or spending increase) in which the additional government bonds issued to fund it are acquired by the central bank and permanently retained on its balance sheet in nominal terms. This is equivalent to a permanent expansion of the monetary base equal to the size of the stimulus, and is distinct from helicopter drops (which involve direct transfers rather than bond purchases).

Incentive-compatibility constraint (binding case): A Gertler-Kiyotaki (2010) / Gertler-Karadi (2011) constraint limiting financial intermediaries’ bond holdings relative to net worth: chi_t * n_t = lambda_b * q^b_t * s^{b,f}_t when binding. When binding, it creates a spread between bond and reserve returns, meaning bonds and reserves are not perfect substitutes. The paper’s irrelevance result holds whether or not this constraint binds, which is the nontrivial analytical contribution.

Interest-paying reserves (interest on reserves): Central bank liabilities that pay a nominal interest rate set by the central bank, distinct from non-interest-paying currency (‘outside money’). The paper argues this is the empirically relevant form of modern monetary base expansion: post-GFC balance-sheet growth by the Fed and ECB was almost entirely in interest-paying reserves. Paying interest on reserves allows the central bank to simultaneously control the policy rate and the size of its balance sheet, which is the feature that drives the irrelevance result.

Cumulative (discounted) fiscal multiplier: As computed in the paper following Gali (2020a): the ratio of the sum of output deviations from steady state over 1,000 quarters to the sum of the fiscal instrument deviations over the same horizon. The relevant multiplier here is the difference between money- and debt-financed versions: 0.26 in the extension with corporate securities held by intermediaries, compared to 0.50 in Gali (2020a).

Two-tiered reserve system: The ECB framework (in operation since July 2023) under which intermediaries must hold minimum reserves equal to a fixed fraction of deposits (currently 1%) at zero interest, while excess reserves earn the policy rate. The paper proves (Proposition 3) that extended Ricardian equivalence carries over to this system: the nominal deposit rate becomes (1-theta)*policy rate, but since the policy rate remains the sole endogenous variable determining the deposit rate, the irrelevance result is unaffected.

Source-text-origin note: The working paper title reads ‘Monetary financing does not produce miraculous fiscal multipliers’; the published EJ title adds ’neither high inflation nor’ — the summary uses the published title as given in the task, which also reflects the paper’s second finding (no inflationary effect).