The Policy Choice After the Boom

Financial liberalization can create a boom before it raises tail risk because a direct reform changes a bank’s lending capacity first, while debt, collateral dependence, and funding fragility accumulate slowly. The final policy question is not whether to choose growth or stability in the abstract. It is whether a specific reform’s gains justify its distribution of outcomes, and whether safeguards reach the institutions, currencies, and funding markets that carry the exposure. That judgment requires policy directness, credit allocation, funding structure, evidence strength, and the full time horizon to remain in the same frame.

A welfare trade-off, not a free lunch

The central evidence supports a conditional conclusion. Removing a binding credit constraint can deepen intermediation and raise activity. In the Federal Reserve study, the reforms that directly expand credit supply are followed by stronger real private-credit growth and an initial rise in real GDP per person. They are also followed, after a delay, by higher crisis probability and more severe downside outcomes. The question is not whether the first effect is real and the second effect is real. It is how they should be valued together.

The study makes that comparison through a welfare exercise rather than through a single GDP average. It considers a representative agent with constant-relative-risk-aversion utility, a 15-year evaluation horizon, and a discount factor of 0.96. Under the no-crisis path, deregulation produces higher consumption. The exercise then adds estimated crisis hazards and an empirical distribution of crisis losses drawn from 82 systemic banking crises with usable consumption data. Because utility is more sensitive to a large loss than a simple average is, the probability and severity of the downside affect the conclusion disproportionately.

The resulting threshold is a coefficient of relative risk aversion of 7.2. Below that value, the estimated gains in normal times outweigh the utility losses from the additional risk in the exercise. Above it, the insurance value of the regulated status quo is larger. The threshold is useful because it makes the disagreement explicit: two decision makers can see the same estimated growth path and reach different welfare judgments if they assign different weight to a low-probability, high-cost crisis.

But 7.2 is not a universal policy cutoff. It is a result of the model and calibration. Change the horizon, discounting, crisis-loss distribution, crisis hazard, or preference specification, and the threshold can move. The exercise also treats the economy through a representative agent, which is a way to aggregate outcomes rather than a complete account of who gains and who loses. A borrower with a newly available productive loan, a bank exposed to a later default, and a worker affected by a downturn need not experience the same path even if the aggregate data are combined into one consumption series.

The result also depends on which reform is being considered. The study does not say that every change labeled liberalization has the same credit response. A policy that directly removes a lending ceiling is closer to the estimated treatment than a reform that changes risk-sharing, bank activities, or international capital allocation without directly releasing loan supply. Nor does the historical evidence directly estimate the effect of relaxing the intensive margin of post-2008 capital and supervisory rules. A current proposal must first be classified before the historical estimate can inform it.

The time horizon is equally decisive. An evaluation restricted to the first three years will capture much of the early GDP response and little of the estimated crisis-risk increase. An evaluation that extends to eight or fifteen years sees the later spread widening, output reversal, and tail exposure more clearly. Neither horizon is “the” correct one independent of the decision.