Financial Stress Testing: Scenario Analysis and Regulation

Module: Fintech AI | Difficulty: Advanced

Regulatory Framework

• CCAR (US): Capital stress test
• EBA (EU): EU-wide stress test
• ICAAP: Internal capital assessment

Scenario Generation

where .

Macro Factors

| Factor | Stress Impact | |--------|--------------| | GDP | | | Unemployment | | | Interest Rate | | | Housing | |

import numpy as np

class StressTest:
    def __init__(self, model, cov_matrix):
        self.model = model; self.cov = cov_matrix
    def historical_stress(self, historical_shock):
        return self.model.predict(historical_shock)
    def hypothetical_stress(self, scenario):
        return self.model.predict(scenario)
    def monte_carlo_stress(self, n_simulations=10000):
        shocks = np.random.multivariate_normal(
            np.zeros(len(self.cov)), self.cov, n_simulations)
        losses = [self.model.predict(shock) for shock in shocks]
        return {
            'var_99': np.percentile(losses, 1),
            'es_99': np.mean([l for l in losses if l <= np.percentile(losses, 1)]),
            'max_loss': np.min(losses)
        }

Research Insight: Stress testing reveals model weaknesses under extreme scenarios. The 2008 crisis showed that models assuming normal distributions underestimate tail risk. Fat-tailed distributions and regime-switching models provide more realistic stress scenarios.