The Policy Rules That Kept Losses Contained

Chile's experience allows a more useful policy question than whether crisis credit should be generous or restrictive. What happens to credit quantity, borrower quality, and public exposure when one rule changes while the others remain in place? This chapter compares the study's empirical counterfactuals with its endogenous model. The comparison helps identify which rules appear to contain risk, but it must be read as a set of mechanism tests—not as a forecast or a welfare ranking—because the empirical scenarios hold much of firm and bank behavior fixed while the model relies on calibration and simplifying assumptions.

Changing the rules in a partial-equilibrium lens

The empirical counterfactuals begin with a combined-policy baseline. Program credit is 3.6 percent of GDP, the credit-weighted default probability is 7.4 percent, and expected loss is 3.7 percent of baseline credit. These figures use consistent comparison bases within the exercise. Credit is reported relative to GDP, while expected loss is reported relative to the baseline credit of eligible formal firms. Keeping those denominators separate is essential: a change in the amount of credit can alter the average risk simply by changing which firms receive weight in the portfolio.

The first counterfactual reduces the maximum loan from three months of sales to one month. Credit falls from 3.6 to 2.5 percent of GDP, and the average default probability falls from 7.4 to 7 percent. Expected loss declines from 3.7 to 2.5 percent of baseline credit. The result is not just a smaller loan book. Because the firms that lose the most weight under the smaller cap are riskier, the credit-weighted average risk also falls. A loan cap therefore acts on quantity directly and on the composition of the portfolio indirectly.

That does not make a smaller cap an unqualified improvement. It reduces the exposure created by the program, but it can also leave a viable firm with less working capital than it needs to survive a temporary interruption. The counterfactual does not measure which firms fail because the cap is lower, how many jobs are preserved, or whether private credit replaces the missing amount. It isolates the risk effect of a tighter limit under the study's assumptions. A policymaker would still have to weigh that contained exposure against the unmeasured benefit of allowing more borrowing.

The second counterfactual removes the rule that excludes firms with more than 30 days past due. Credit expands to 7.6 percent of GDP, more than twice the combined-policy baseline in the reported GDP comparison. The average default probability rises to 9.2 percent, and expected loss reaches 9.9 percent of baseline credit, nearly three times the 3.7-percent baseline. This is a nonlinear response: the eligibility change increases the number of potential borrowers and admits firms with a risk profile that is materially worse under the study's approximation.

The contrast with the smaller cap shows why “access” is not one dimension. A cap controls how much an eligible firm can borrow. A repayment-status gate controls which firms can enter the pool at all. Relaxing the cap for the existing pool and relaxing the eligibility rule for delinquent firms may both be described as more generous credit, but they change quantity and borrower quality through different channels.