Capital Flows and the Global Collateral Cycle

Layer 1: Overview

The paper asks why large gross financial flows exist between similarly rich countries (especially the U.S. and Europe), why financial integration raises rather than lowers asset price volatility, and why safe-asset prices rise during crises. The authors argue that cross-country disparities in collateral technology — the capacity to securitize domestic assets into state-contingent tranches — can account for all three phenomena simultaneously, without invoking differences in preferences, endowments, production technologies, or idiosyncratic shocks.

The model is a two-country (Home = U.S., Foreign = Europe) collateral general equilibrium model built on Geanakoplos (2003). Agents within each country are risk-neutral but heterogeneous in beliefs (indexed by optimism parameter i). The only asymmetry across countries is the collateral technology: Home collateral can back any state-contingent promise (tranching), while Foreign collateral can back only non-contingent debt (leverage). Both countries share common shocks. Collateral requirements are endogenously determined in equilibrium. The authors first characterize static autarky and integrated equilibria analytically, then simulate a three-period dynamic model calibrated with dUU = dDU = 1 and dDD = 0.2.

In the static numerical example (dD = 0.2, uniform beliefs γ(i) = i), Foreign autarky yields an asset price of p* = 0.75 with marginal buyer i₁ = 0.69. Home autarky yields a higher asset price of p = 0.83 (marginal buyers i₁ = 0.65, i₂ = 0.10) and a D-tranche price of πT = 0.18. In international equilibrium, the Home price rises further to p̂ = 0.86, the Foreign price falls to p̂ = 0.73, and the D-tranche price rises to π̂T = 0.19. Financial integration moves identical-payoff asset prices further apart (Proposition 2), and the Law of One Price fails with a strictly positive collateral gap Δ̂ = p̂ − p̂* = dD(γ(î₁) − γ(î₂)) (Proposition 1).

In the dynamic three-period model (dDD = 0.2), the Foreign autarky leverage cycle produces a 25% asset price fall from p₀ = 0.96 to pD = 0.72 after scary bad news. The Home autarky securitization cycle produces a larger 39% fall from p₀ = 1.21 to pD = 0.74. Financial integration amplifies both: the Home price in international equilibrium starts higher at p̂₀ = 1.40 and falls 44% to p̂D = 0.79; the Foreign price falls from p̂₀ = 0.91 to p̂D = 0.68 (25%), both crashes exceeding their autarky counterparts. The collateral gap is pro-cyclical, falling from Δ̂₀ = 0.49 at s=0 to Δ̂D = 0.11 at s=D. Gross flows are also pro-cyclical: Home gross inflows drop from 0.266 to 0.173 and gross outflows from 0.378 to 0.215 from the good to the bad state. The trade balance deficit collapses from TBH₀ = 0.12 to TBH_D = 0.04. Meanwhile, the Arrow D security (the negative beta, super-safe tranche) rises in price counter-cyclically from π̂⁰_D = 0.85 to π̂^D_D = 0.96 in international equilibrium, and is always priced higher in international equilibrium than in Home autarky.

Four mechanisms drive the results. First, the collateral value premium: tranching splits cash flows to serve heterogeneous buyers and raises asset prices above the unsecuritized level, producing a law-of-one-price failure. Second, bidirectional gross flows: Foreign investors demand Arrow D tranches available only from Home; Home investors buy cheap Foreign bonds because the basis (price of replicating Arrow portfolio minus price of non-contingent Foreign bond) is positive. Third, a permanent trade deficit for Home: Home’s collateral-driven wealth advantage (Corollary 2) generates higher consumption purchases in every state, and the trade deficit equals eY·Δ̂/(2e_c0 + eY(p̂+p̂*)) in all states. Fourth, the Global Collateral Cycle: scary bad news curtails the feasibility of creating negative beta tranches, making Home’s effective collateral advantage procyclical even though the technology itself is fixed, driving procyclical gross flows and trade imbalances and counter-cyclical safe-asset prices through a supply channel that complements the conventional demand-side flight-to-safety.

Layer 2: Deep Dive

What drives gross financial flows in both directions between two otherwise identical countries?

Foreign agents demand Arrow D securities (negative beta tranches) that only Home can produce via its superior collateral technology. This generates gross inflows to Home. Simultaneously, Home agents buy Foreign bonds because the basis is positive — the foreign non-contingent bond trades cheaper than a replicating portfolio of Arrow securities produced at Home. This generates Home gross outflows. Both directions arise purely from the collateral technology disparity, with no role for interest rate differentials, endowment differences, or idiosyncratic shocks.

What is the Law of One Price failure and how is it characterized analytically?

Proposition 1 establishes that in any international equilibrium, the collateral gap Δ̂ = p̂ − p̂* = dD(γ(î₁) − γ(î₂)) > 0. Two assets with identical payoffs trade at different prices because the Home asset can be tranched into state-contingent claims sold to different buyers, generating a collateral value premium, while the Foreign asset can only back non-contingent debt. Corollary 1 shows the basis β = π̂U + π̂D − 1 > 0 and Δ̂ = dD·β, linking both deviations to the degree of collateral technology advantage measured by dD.

Why does Home run a permanent trade deficit and how large is it?

Proposition 5 proves that in the home-biased neutral international equilibrium, Home runs a trade deficit in every state (0, U, D). Because financial integration raises Home asset prices (Proposition 2), Home agents are wealthier in every state (Corollaries 2 and 3). By homotheticity, Home purchases more of every good, including foreign consumption goods. The deficit at s=0 equals eY·Δ̂ / (2e_c0 + eY(p̂+p̂*)) = eY·dD·β / (same denominator). This mechanism does not require Home to have a lower interest rate or higher saving — the collateral advantage directly raises Home’s permanent wealth. In the numerical example, TBH₀ = 0.12.

Why does financial integration increase asset price volatility rather than reduce it through diversification?

Integration raises the collateral value of Home assets at s=0 because Foreign demand for D tranches is added to domestic demand, pushing prices to a higher starting point (p̂₀ = 1.40 vs. p₀ = 1.21 in Home autarky). After scary bad news, the same Securitization Cycle dynamic that would reduce Home prices in autarky now operates from a higher starting point and propagates to Foreign asset prices, because Foreign assets are priced relative to Home assets. Price crashes deepen: Home falls 44% in IE versus 39% in autarky; Foreign falls 25% from a lower s=0 base. The collateral gap and the volume of negative beta assets that can be created both collapse after bad news, reinforcing the price drop.

What is the supply channel for safe-asset price appreciation during crises, and how does it differ from the flight-to-safety demand channel?

The supply channel works through the endogenous collapse in the quantity of Arrow D (negative beta) securities created from Home collateral after scary bad news. Since the collateral’s worst-case payoff worsens at s=D, fewer Arrow D securities can be guaranteed per unit of collateral, even though the technology itself is unchanged. The reduced supply — combined with persistent demand from pessimistic agents — drives up the Arrow D price (from 0.85 to 0.96 in the IE numerical example). This contrasts with the conventional flight-to-safety demand channel, in which agents shift demand toward safe assets due to heightened risk aversion. Both channels operate simultaneously in the model: the wealth redistribution toward pessimists at s=D also raises aggregate effective risk aversion.

How does Home’s collateral technology advantage create exorbitant privilege?

The exorbitant privilege arises because only Home can create negative beta (Arrow D) securities, but both Home and Foreign agents demand them. In international equilibrium the Arrow D price is always higher than in Home autarky — Foreign demand adds to domestic demand while supply remains constrained by Home collateral. This means Home’s collateral generates a rent above the payoff value. In turn, Home is wealthier in every state and can run a permanent trade deficit, receiving more consumption goods from the world in exchange for financial claims that in aggregate pay less (because distinct buyers value distinct tranches more than the aggregate). The collateral gap measuring this privilege is larger in IE than the autarky spread, and it is pro-cyclical — largest in good times.

What is ‘scary bad news’ and why does it create amplified price crashes?

Scary bad news is a shock at s=D that simultaneously (i) worsens expected payoffs and (ii) raises downside variance, so the collateral’s worst-case value from D is much lower (dDD = 0.2 versus dUU = 1). In Foreign autarky this reduces the maximum non-contingent debt that can be collateralized, sharply reducing leverage and hence the price of risky assets beyond what the direct dividend news implies — the Leverage Cycle of Geanakoplos (2003). In Home autarky the same scary news reduces the quantity of Arrow D securities that can be created, causing an even larger asset price crash — the Securitization Cycle of Fostel and Geanakoplos (2012a). In international equilibrium both cycles interact, as the higher collateral values at s=0 unwind more sharply.

What refinement resolves multiplicity in the international equilibrium and what does it imply for gross flows?

Because Home and Foreign consumption goods and Arrow U securities are perfect substitutes under linear utility, the international equilibrium has a continuum of solutions for individual portfolio allocations. The authors introduce a ‘home-biased neutral’ refinement in two steps: first, ’neutrality’ selects the allocation where agents seeking proportional payoffs hold proportional portfolios (this is justified as the limit of small perturbations breaking perfect substitutability); second, ‘home bias’ requires each agent to hold all domestic goods before holding foreign ones, minimizing the scale of gross flows. Even under this most conservative refinement, Propositions 3 and 4 establish that Home is a seller of Arrow D and net seller of Arrow U securities (gross inflows) and a buyer of Foreign bonds (gross outflows), and Proposition 5 establishes the permanent trade deficit.

How does this paper relate to and differ from the prior global imbalances literature?

The standard literature (Caballero-Farhi-Gourinchas 2008, Mendoza-Quadrini-Rios-Rull 2009, Angeletos-Panousi 2011) explains capital flows via differences in insurance capacity or financial development that affect autarkic savings rates and interest rates, generating primarily net capital flows and current account imbalances. Maggiori (2017) assumes Home financiers face weaker borrowing constraints, allowing them to absorb aggregate risk. The present paper differs: (i) all investment returns and insurance possibilities are identical across countries — only the collateral technology differs; (ii) the paper focuses on gross flows, which dwarf net flows; (iii) flows are driven by positive-supply collateral-backed cash flows, not zero-supply Arrow securities; (iv) financial integration increases rather than decreases volatility (contra Mendoza-Quadrini 2010 who find integration attenuates U.S. crisis severity); (v) the mechanism generates violations of the Law of One Price, not just interest rate differentials.

What are the main testable implications and what data would be needed to test them?

Section V lists eight testable implications: (1) securitization raises collateral prices relative to identical unsecuritized foreign collateral, testable via option-adjusted spreads on mortgages versus sovereign bonds across countries; (2) larger securitization gaps predict larger gross flows in both directions, requiring data on cross-border securitization trades; (3) larger securitization gaps predict larger trade imbalances; (4) larger collateral technology gaps increase global asset price volatility in both countries; (5) changes in financial integration affect price volatility; (6) larger technology gaps increase pro-cyclicality of gross and net flows; (7) larger gaps increase counter-cyclicality of super-safe asset prices; (8) changes in financial integration affect flow cyclicality. The authors note that cross-border securitization trade data are currently scarce and call for a taxonomy of collateral structures and volumes by country as a preliminary step.

What scope conditions and extensions are discussed?

The model abstracts from production and investment, so results apply to the trade balance not the current account. The authors conjecture that adding production (cf. Fostel-Geanakoplos 2016) would reinforce Home’s current account deficit via collateral-driven over-investment. There are no exchange rates; the conjecture is that differentiated goods would imply a stronger Home currency, connecting to the exorbitant privilege literature (Gourinchas-Rey 2022, Jiang-Krishnamurthy-Lustig 2024). All agents are risk-neutral, which makes equilibria tractable but rules out curvature-based risk-sharing motives; the authors interpret heterogeneous optimism as a proxy for heterogeneous risk aversion or hedging mandates. Shocks are common, not idiosyncratic; idiosyncratic shocks would add further risk-sharing motives on top of the collateral channel but the authors argue their mechanism is conceptually distinct. Partial correlation of asset payoffs across countries is considered in an appendix extension and shown to reinforce the main results.

How does the paper handle the relationship between the collateral technology and the quantity of safe assets in the cycle?

The key insight is that while the collateral technology (the set of contracts J available) is fixed across the cycle, the amount of negative beta assets that can actually be created varies endogenously with the collateral’s payoff characteristics. At s=0, with a worst-case payoff dD = p*D = 0.72 for the dynamic problem, substantial Arrow D securities can be created. At s=D, the worst-case payoff is dDD = 0.2, drastically curtailing the feasible quantity of Arrow D securities per unit of collateral. This procyclical variation in effective securitization capacity, driven by scary bad news, is what generates the Global Collateral Cycle — the collateral technology itself is constant but the ‘room’ to use it varies with macroeconomic conditions.

Key Concepts

Collateral technology: The legally enforceable set J of financial contracts that can be created using a domestic asset as collateral; in the paper it determines whether an asset can back state-contingent (tranching, Home) or only non-contingent (leverage, Foreign) promises, and it applies only to domestic collateral because enforcement depends on domestic courts and legal infrastructure.

Negative beta asset (super safe asset): A financial asset whose price typically rises when aggregate conditions worsen; in the model this is the Arrow D security (a tranche promising payment only in the bad state D), whose real-world analogues include AAA securitization tranches and U.S. Treasuries. In the paper’s static model, the D-tranche price rises from 0.74 to 0.92 in Home autarky after bad news, and from 0.85 to 0.96 in international equilibrium.

Collateral gap (Δ̂): The equilibrium price difference p̂ − p̂* between identical-payoff assets in Home and Foreign arising purely from the difference in collateral technologies; always strictly positive in international equilibrium and equal to dD(γ(î₁) − γ(î₂)), measuring the collateral value premium of the Home asset. In the dynamic model it falls pro-cyclically from 0.49 at s=0 to 0.11 at s=D.

Basis (β): The premium of a replicating portfolio of Arrow securities over a non-contingent bond with the same aggregate payoff: β = π̂U + π̂D − 1; always positive in international equilibrium and equal to Δ̂/dD, reflecting that contingent claims backed by Home collateral command a higher combined price than their non-contingent Foreign equivalent.

Scary bad news: A negative shock that simultaneously lowers expected payoffs and raises downside variance, so that the collateral’s worst-case value from the bad state is lower than from the initial state; following Geanakoplos (2003, 2010), this type of news causes endogenous collapses in leverage and securitization volume beyond what the fundamental payoff news alone would imply, generating amplified asset price crashes and the leverage/securitization cycle dynamics.

Global Collateral Cycle: The international financial cycle generated by the interaction of disparate collateral technologies and scary bad news: in the down phase, the feasible quantity of Home-created negative beta assets falls (supply contraction), the collateral gap shrinks, gross flows collapse, trade imbalances narrow, risky asset prices crash further than in autarky in both countries, and safe-asset prices rise above their autarky levels.

Collateral value: The component of a risky asset’s equilibrium price that exceeds its expected payoff value and arises from the asset’s capacity to serve as collateral backing contingent financial promises; it is positive when heterogeneous buyers are willing to pay a combined premium for distinct tranches relative to what a single buyer would pay for the undivided asset, as in the floater/inverse-floater securitization example described in the paper.