Geanakoplos Leverage Cycle Theory

Geanakoplos leverage cycle theory is a general-equilibrium theory of endogenous collateral margins (haircuts), credit leverage, and asset-price boom–bust cycles in business and economics theory developed by American mathematical economist John Geanakoplos across a series of foundational papers (1997, 2003, and his 2010 NBER Macroeconomics Annual treatise "The Leverage Cycle"). Challenging the classical competitive equilibrium paradigm—which assumes that supply and demand for credit are cleared by a single scalar price, the interest rate, while treating borrowing limits or collateral ratios as exogenous institutional parameters—Geanakoplos demonstrates that in a world with default risk and non-recourse collateralized promises, a single competitive supply-and-demand intersection simultaneously pins down both the equilibrium interest rate and the equilibrium loan-to-value (LTV) leverage ratio (or haircut). When investors hold heterogeneous subjective beliefs (optimism h ∈ [0, 1]) about future asset payoffs, periods of low fundamental uncertainty allow lenders to demand small collateral haircuts, empowering the most optimistic investors in the economy to borrow and bid asset prices up to levels reflecting top-percentile optimism. Conversely, when an adverse volatility shock ("scary bad news") strikes, asset prices suffer a three-part crash cascade: (1) bad news lowers every buyer's fundamental valuation; (2) the spike in downside tail risk forces lenders to tighten collateral margins (de-leveraging); and (3) leveraged optimists who bought at the peak are wiped out and forced into fire-sale liquidations, transferring the marginal pricing of the asset down to more pessimistic buyers.

In standard Arrow–Debreu general equilibrium theory, every agent honors financial promises with certainty, so credit contracts are specified by their promised interest rate r. Even in macro-financial credit models such as Kiyotaki–Moore credit cycles (1997) and the Bernanke–Gertler–Gilchrist financial accelerator (1999), borrowers face binding collateral constraints, yet the collateral margin is determined by a deterministic future resale price or an exogenous verification parameter rather than by an explicit menu of competitive collateralized contracts traded in general equilibrium.

At first glance, determining both the interest rate and the collateral haircut by supply and demand for loans appears to violate basic algebra: how can one scalar equation (Aggregate Demand for Loans = Aggregate Supply of Loans) solve for two endogenous variables (the interest rate r and the loan-to-value ratio LTV)? In his 1997 article "Promises, Promises" (with Geanakoplos and Zame, 2014) and his 2003 lecture "Liquidity, Default, and Crashes: Endogenous Contracts in General Equilibrium", Geanakoplos resolved this paradox by introducing Collateral Equilibrium:

1. Instead of trading a single generic loan contract, the competitive market lists a continuum of distinct contracts indexed by j, where each contract j specifies both a promised repayment vector A_j across future states and a physical collateral requirement C_j.
2. Under no-recourse limited liability beyond the pledged collateral C_j, actual delivery on contract j in future state s is the minimum of the promised face value and the market value of the collateral:

1. Because each collateral level C_j defines a distinct traded security with its own equilibrium price π_j (and implied interest rate r_j), every contract j has its own market-clearing equation, and endogenous leverage is determined by which contract(s) j* attract positive trading volume in equilibrium.

Continuum of Heterogeneous Buyers h ∈ [0, 1] and Risky Asset Payoffs

To see why only one collateral contract trades in equilibrium under binomial transitions, consider Geanakoplos's two-period economy (t = 0 and t = 1):

• Endowments: A unit continuum of risk-neutral investors indexed by h ∈ [0, 1] each starts at t = 0 with 1 unit of liquid cash (consumable at t = 0 or stored until t = 1 at zero interest) and 1 unit of a risky asset (aggregate asset supply = 1; aggregate cash = 1).
• Binomial Future States (t = 1): At t = 1, 1 unit of the risky asset pays 1 unit of consumption good in the Up state (U) (s_U = 1) and s_D = s ∈ (0, 1) units in the Down state (D) (for example, s = 0.20).
• Heterogeneous Optimism h ∈ [0, 1]: Investors differ in their subjective probability γ(h) that State U will occur at t = 1, where γ(h) increases in h (h = 1 is the most optimistic buyer; h = 0 is the most pessimistic). Investor h's valuation V(h) of 1 unit of the risky asset at t = 0 is:

Why Binomial Uncertainty Concentrates Borrowing at the Maximum No-Default Promise s

When investors borrow cash at t = 0 by issuing non-recourse promises with face value φ > 0 backed by 1 unit of the risky asset, two regimes arise:

1. Riskless Promises (φ ≤ s): When φ does not exceed the worst-case Down-state payoff s, the collateral covers the debt in both State U and State D. Because there is zero default in every state, all investors h ∈ [0, 1] value the promise at φ/(1 + r) with riskless interest rate r = 0.
2. Default-Prone Promises (φ > s): Each incremental dollar of face value Δφ above s pays off only in State U. Because borrowers at t = 0 are optimists (high h) while cash lenders are pessimists (low h′ < h), any lender h′ values an Up-state-only payout at γ(h′)Δφ while the borrower h values that obligation at γ(h)Δφ > γ(h′)Δφ. Thus, no optimist is willing to sell an Up-state contingent claim to a more pessimistic lender.

This yields Geanakoplos's Binomial No-Default Leverage Theorem: in any two-state binomial transition with heterogeneous beliefs and risk-neutral agents, 100 percent of equilibrium borrowing occurs in the unique contract whose promised face value equals the worst-case collateral payoff φ* = s = s_D, at the riskless interest rate r = 0. When the asset price at t = 0 is P, every borrower borrows φ* = s per unit of asset and posts a cash margin of (P − s), pinning down the equilibrium Loan-to-Value Ratio (LTV), Collateral Haircut (H), and Leverage Ratio (LR):

Solving for the Marginal Buyer h* and the Leveraged Asset Price P*

Because V(h) is increasing in h, there exists a unique marginal buyer h* ∈ (0, 1) who is indifferent between buying the risky asset at price P* and holding cash (V(h*) = P*):

• Optimists (h ∈ (h*, 1]): Prefer the risky asset to cash (V(h) > P*). They pledge all of their initial cash endowment (1 − h*) plus borrow s against all 1 unit of the economy's asset to buy the h* units of the asset sold by the pessimists.
• Pessimists (h ∈ [0, h*)): Prefer cash and riskless lending to holding the risky asset (V(h) < P*). They sell their h* units of the risky asset to the optimists and lend out their cash at the riskless rate r = 0 against collateral promises of s per unit of asset.

Equating the value of assets sold by the pessimists (h*P*) to the cash spent plus total collateralized borrowing by the optimists ((1 − h*) + s) yields the collateral-budget market-clearing equation:

Combining the marginal buyer's valuation equation P* = s + γ(h*)(1 − s) with the market-clearing equation P* = (1 − h* + s)/(h*) uniquely determines both the marginal buyer h* and the equilibrium asset price P*.

Distinguishing "Calm Bad News" from "Scary Bad News" (Tail Volatility)

To derive Geanakoplos's Three-Part Crash Mechanism, extend the model to a three-period binomial tree (t = 0, t = 1, t = 2) in which bad news arrives at the interim date t = 1 (state D at t = 1, followed by DU paying 1 or DD paying s_DD < s at t = 2). Suppose at t = 0 the asset pays 1 in states UU, UD, and DU (so at t = 0 the worst-case price at t = 1 is P_D, allowing large Period-0 borrowing φ₀ = P_D), and pays s_DD = s < 1 only if two consecutive Down shocks occur (DD).

The Three Concurrent Stages of a Leverage Crash

When the first Down shock (D) hits at t = 1, three mechanisms unfold simultaneously:

1. Direct Fundamental Bad News (Stage 1): The probability of reaching the bad final state DD increases from (1 − h)² to (1 − h), lowering every investor's valuation of the asset. For top optimists (h ≈ 0.90), however, this direct effect is modest because they believe the economy will recover in DU with 90 percent probability.
2. De-Leveraging / Margin Tightening (Stage 2): At t = 0, lenders lent up to φ₀ = P_D ≈ 0.70 against each unit of asset because in state D at t = 1, the asset is worth P_D. Once state D arrives at t = 1, however, the next period's worst-case state is DD, where the asset pays s_DD = 0.20. Thus, at t = 1, lenders cut maximum borrowing per unit of asset from φ₀ ≈ 0.70 down to φ₁ = s_DD = 0.20. Geanakoplos calls any shock that widens the gap between expected value and worst-case tail payoff "scary bad news" (increased tail uncertainty/volatility): scary bad news forces a collapse in equilibrium leverage (a spike in collateral haircuts).
3. Bankruptcy of the Optimists and Loss of the Marginal Buyer (Stage 3): The top optimists h ∈ (h₀*, 1] bought 100 percent of the economy's risky assets at t = 0 by borrowing P_D per unit. When the asset price falls to P_D at t = 1, their collateral value P_D equals their promised debt repayment φ₀ = P_D, leaving every optimist h ∈ (h₀*, 1] with zero net worth at t = 1. With the top optimists wiped out at t = 1, the stock of the risky asset must be sold in a fire sale to the surviving moderate buyers h ≤ h₀* who lent cash at t = 0—requiring a new, lower marginal buyer h₁* < h₀* who now faces tighter margin requirements (borrowing s_DD = 0.20 instead of P_D).

In his 2009 and 2010 papers (and collaborative work with Ana Fostel in the American Economic Review, 2008 and 2015), Geanakoplos applied the leverage cycle to explain the 2000–2006 U.S. housing and non-agency mortgage-backed securities (MBS) boom and the 2007–2009 financial crisis:

1. The Upswing (2000–2006) and Credit Default Swaps (CDS): From 2000 to early 2006, down payments on subprime and Alt-A mortgages fell from ~15–20% to under 3–5% (LTV ratios rising toward 97%), while repo haircuts on AAA/AA-rated non-agency MBS fell to 1.5–5%. Then, in late 2005 and 2006, the standardization of Credit Default Swaps (CDS) on subprime MBS bonds (the ABX index) allowed pessimists (h near 0) to leverage their negative beliefs by buying CDS protection without owning the underlying bonds, compressing MBS prices in 2006–2007 even before peak mortgage defaults arrived.
2. The Crash (2007–2009) and Macroprudential Policy: As delinquency news turned into "scary bad news" in 2007–2008, repo lenders raised haircuts on non-agency securities from 5% to 40–100% (reinforced by the Brunnermeier–Pedersen liquidity spiral model), bankrupting leveraged funds and investment banks. Because traditional monetary policy targets the riskless short-term interest rate r while leaving collateral haircuts H and LTV ratios unregulated, Geanakoplos argues that central banks should monitor and regulate economy-wide collateral leverage—imposing countercyclical LTV caps during booms and providing direct collateral lending facilities during de-leveraging crashes.

How can a single supply-and-demand intersection determine both the interest rate and leverage in Geanakoplos's model?

In Geanakoplos's Collateral Equilibrium, the market lists a menu of distinct contracts j, each defined by a promised payment φ_j and a physical collateral requirement C_j, each priced by its own supply-and-demand equation. In a two-state binomial tree with heterogeneous beliefs, optimistic borrowers never wish to issue defaultable debt (φ > s) to pessimistic lenders because pessimists discount the Up-state payoff. As a result, equilibrium borrowing concentrates in the unique maximum riskless collateral contract φ* = s (the worst-case Down-state payoff of the collateral), which simultaneously pins down the loan interest rate at the riskless rate (r = 0) and the equilibrium loan-to-value ratio at LTV = s/P*.

Why do asset prices crash by much more than fundamental bad news in the leverage cycle?

In a leveraged economy, the asset price P* = V(h*) reflects the valuation of the marginal buyer h*, who is a high-percentile optimist because the top (1 − h*) optimists use borrowed money to buy the asset supply. When "scary bad news" hits, three forces strike at once: (1) every investor lowers their valuation; (2) increased downside tail uncertainty (s′ < s) forces lenders to raise collateral margins (de-leveraging); and (3) the optimists h > h* who bought on margin at t = 0 go bankrupt and lose their wealth. With the original optimists wiped out and leverage curtailed, the asset must be sold to more pessimistic buyers (h′* ≪ h*), causing the market price to fall beyond the change in any single investor's valuation.

Why did the introduction of Credit Default Swaps (CDS) help trigger the 2007–2008 subprime crisis?

Before 2005, physical real estate and non-agency mortgage-backed securities (MBS) could be bought on high leverage by optimists via repo loans, but could not easily be shorted by pessimists. Consequently, asset prices reflected the beliefs of the most optimistic buyers (h near 1). As Fostel and Geanakoplos (2012, 2015) proved, the standardization of CDS on subprime MBS in 2005–2006 allowed pessimistic investors (h near 0) to take leveraged short positions against mortgage bonds, shifting the balance of leveraged purchasing power from optimists toward pessimists.

How does Geanakoplos's leverage cycle differ from the Kiyotaki–Moore (1997) credit cycle?

In Kiyotaki–Moore credit cycles, the future path of land prices after a one-time shock is deterministic, so borrowers borrow up to 100 percent of next period's discounted land price (q_t+1/R), and the percentage loan-to-value margin does not widen due to endogenous uncertainty. By contrast, in Geanakoplos's leverage cycle, future asset payoffs are stochastic, investors have heterogeneous beliefs, and the percentage collateral haircut is endogenous to tail volatility: when uncertainty widens, haircuts spike (LTV falls), amplifying the price crash.

1. Geanakoplos, J. (2010). "The Leverage Cycle." NBER Macroeconomics Annual, 24(1), 1–66. https://doi.org/10.1086/648285
2. Geanakoplos, J. (1997). "Promises, Promises." In W. B. Arthur, S. N. Durlauf, & D. A. Lane (Eds.), The Economy as an Evolving Complex System II (pp. 285–320). Addison-Wesley. https://doi.org/10.1201/9780429496639-11
3. Geanakoplos, J. (2003). "Liquidity, Default, and Crashes: Endogenous Contracts in General Equilibrium." In M. Dewatripont, L. P. Hansen, & S. J. Turnovsky (Eds.), Advances in Economics and Econometrics: Theory and Applications, Eighth World Congress (Vol. 2, pp. 170–205). Cambridge University Press. https://doi.org/10.1017/CBO9780511610257.007
4. Fostel, A., & Geanakoplos, J. (2008). "Leverage Cycles and the Anxious Economy." American Economic Review, 98(4), 1211–1244. https://doi.org/10.1257/aer.98.4.1211
5. Geanakoplos, J., & Zame, W. R. (2014). "Collateral Equilibrium, I: A Basic Framework." Economic Theory, 56(3), 443–492. https://doi.org/10.1007/s00199-013-0797-4