Parametric Stress Tests

Stress tests apply hypothetical adverse shocks to the latest portfolio holdings and report the resulting one-period loss. They don't use historical data — they use a battery of scenarios designed to capture tail events that may not have occurred in the backtest sample. The Stress Tests sub-pill applies a default battery (market, sector, factor shocks) to your latest portfolio composition.

Why parametric stress tests, not just historical max drawdown

Historical max drawdown reflects what happened in the backtest period. If the backtest started in 2010 it missed 2008–2009, and the max drawdown understates tail risk. Parametric stress tests fill the gap by asking: "if a 2008-style shock occurred today, what would happen to this portfolio?"

The framework is standard in regulatory risk management (Basel banking, Solvency II for insurance) and well-documented in Jorion (2007). FM103 implements a simplified retail-friendly version.

The default scenario battery

Each scenario is a vector of factor / sector / market shocks applied to the latest portfolio:

• Market −15%: Beta-style broad equity decline. Portfolio loss = beta · −15%.
• Market −30%: Severe shock; loss = beta · −30%.
• Tech −25%: Sector shock applied to tech-classified holdings.
• Financials −25%: Sector shock applied to financials.
• Energy −40%: Sector shock applied to energy.
• Value factor −10%: Top-quintile value names underperform bottom-quintile by 10%.
• Momentum factor −15%: Same for momentum (momentum crashes can be 20%+, see Daniel & Moskowitz 2016).
• Rates +100 bps: Duration-style shock applied via duration proxy.

Each scenario is independent — the report shows one shock at a time, not combinations. Real-world crises combine shocks (2008 was credit + equity + rates), and a portfolio that survives each individually may not survive the combination.

The output

For each scenario the sub-pill reports:

• Modelled loss as a percentage of portfolio value
• Top contributors — the 5 names that drove the bulk of the loss
• Loss vs. expected drawdown — how the stress loss compares to the backtest's max drawdown

Interpretation

• Stress loss < backtest max-DD: The backtest period saw worse than the modelled scenarios. The strategy has been tested against history; new stress doesn't add information.
• Stress loss ≈ backtest max-DD: Consistent. The stress scenarios approximate the worst historical period.
• Stress loss > backtest max-DD: The backtest period was benign relative to the modelled scenarios. The strategy is untested against this magnitude of shock.

Limitations

• Linear-only. Loss = beta · shock assumes linear response. In real crises, correlations rise (correlation breakdown) and individual stock losses can be non-linear (gap-down opens).
• Independent shocks. Multi-dimensional shocks (rates + equity + credit) are not modelled. A combined scenario is usually worse than the sum of individual scenarios.
• Static portfolio. Stress is applied to the latest portfolio. The actual strategy would rebalance into the shock, potentially adding losses (buying falling knives) or limiting them (reducing exposure via stop-loss). Stress test ignores this dynamic.
• Sector classifications can mislead. A "tech" company that is really a payments processor will be shocked as tech but behave like financials. Classifications are not perfect.

Using stress test output to design overlays

If a scenario produces a loss greater than your tolerance, the overlay design is:

• If sector shock dominates: cap sector exposure (e.g., max 25% in any one sector).
• If market shock dominates: cap portfolio beta (via inclusion of low-beta names or hedge).
• If factor shock dominates: blend offsetting factors (long value + long quality is more shock-robust than long value alone).