Banking System Stability and Stress Tests in Haiti: A Bayesian and Quantile Approach to Systemic Risk

Using seven endogenous variables (real GDP growth, real interest rate, exchange rate, inflation, capital ratio, non-performing loans, return on assets) plus political-instability and natural-disaster indicators, the Bayesian VAR is estimated with Minnesota priors via Gibbs sampling (25 annual observations, 2000-2024), a method suited to Haiti's short, noisy time series. A two-step Bayesian LASSO quantile extension (quantiles 0.05 to 0.95, since no integrated BQVAR package exists) captures tail asymmetries. Key results: return-on-assets elasticity to growth ranges from 0.05 during crises to 0.12 during expansion, rejecting linearity; systemic default probability reaches 15.2 percent under political stress; the estimated capital ratio (7.4 percent) exceeds the 5 percent regulatory minimum; and systemic Value-at-Risk estimates converge across the BVAR and quantile approaches at USD 30.4-32.1 million (95 percent confidence).

The paper concludes Haiti's banking system has a generally satisfactory resilience base but remains structurally exposed, particularly to political shocks and exchange-market tensions. Building on BRH Circular No. 88-1, it recommends: (1) maintaining the 14.5 percent overall solvency requirement (including the conservation buffer) and adding a 1 percentage-point countercyclical buffer to absorb identified political shocks; (2) developing early-warning indicators such as systemic default probability with regime-conditional trigger thresholds; and (3) strengthening foreign-exchange risk management, given its demonstrated adverse effect on asset quality and profitability during crises.