Sudden Stops, Financial Crises, and Leverage

Emerging-market crises of the late 1990s combined three regularities that standard open-economy models cannot produce together: a sharp reversal of international capital flows, a collapse in production and absorption, and a correction in asset prices. In a frictionless model, a household hit by a bad output shock borrows abroad to smooth consumption – the opposite of what the data show, where the external accounts swing to surplus exactly when output and consumption fall. This paper adds to an otherwise standard stochastic small-open-economy business-cycle model (driven by shocks to total factor productivity, the world interest rate, and the world price of imported inputs) a collateral constraint that caps total debt – one-period bonds plus within-period working-capital loans – at a fraction of the marked-to-market value of physical capital, so the constraint is an upper bound on the economy’s leverage ratio. Because the bound moves with the endogenous price of capital, it binds only occasionally: normal cyclical dynamics carry leverage up during expansions until an ordinary one-standard-deviation shock trips it, and then Irving Fisher’s debt-deflation mechanism takes over – agents fire-sell capital to meet margin calls, the price of capital falls, the constraint tightens further, and credit, asset prices and investment spiral down together, while the loss of collateralised working capital cuts labour and imported-input demand and hence output in the same period. Calibrated to Mexican data, the model reproduces the observed Sudden Stop frequency of about 3.3 percent with a leverage ceiling of about one-fifth, and its simulated event windows match the pre-crisis expansion, the size of the GDP and consumption declines, the current-account reversal and the slow recovery; it also generates a wedge in which the Solow residual overstates the fall in true productivity by roughly 30 percent. Two scope conditions are load-bearing. First, the debt-deflation mechanism is non-monotonic in the leverage ceiling and vanishes at both extremes, so credit markets must permit leverage, but only partially. Second, the collateral constraint must bite on working capital: with working capital removed, the model produces no amplification at all in GDP or factor allocations and no pre-crisis expansion. The model’s main quantitative shortfall is the asset-price collapse, which comes out at about 40 percent of the size observed in the data, and the paper is explicit that true productivity still has to fall for the output drop to be realistic – why it falls in a Sudden Stop is left as an open question.

Enrique G. Mendoza | Macro Paper Warehouse

Sudden Stops, Financial Crises, and Leverage