Evaluating safety labels in a crash window

A classroom stress test for construction

Editorial view: a market-wide break is a classroom stress test for portfolio construction. Keep only the traits that stayed inside a defined loss bound, and retire popular quality labels that fail that crash-regime check.

A 1987 crash-window comparison split listed names into two cohorts using a 5 percent price-change cutoff between a pre-shock close and a later close. The resilient cohort is the set of names whose later close stayed at or above 95 percent of the pre-shock close. The decliner cohort is the set of names whose later close sat below 95 percent of the pre-shock close.

A fixed shock window and a loss bound

Editorial framing: scenario risk analysis here means fixing a shock window and a price-loss threshold before asking which holdings stayed inside a bounded drawdown. The 5 percent cutoff is that threshold.

Selection was quasi-random with industry balancing. After that step the decliner cohort had 240 names and the resilient cohort had 85. Incomplete fields were omitted from calculations.

Compare the same statistic across both cohorts

Editorial framing: hypothesis testing here means comparing two crash-window cohorts on the same statistic with a two-sided mean test, instead of ranking names by a favorite fundamental. Mean differences were tested with a two-tailed Student t-test.

Mean dividend yield, price-to-earnings ratio, and return on equity did not differ significantly between the two cohorts. Price-to-earnings means were 19.5 with a 13.2 spread versus 16.0 with a 13.1 spread, and the difference was not significant at p greater than 0.3.

Yield means were not significantly different at p greater than 0.2 because the resilient cohort’s 3.9 percent mean had an 8.4 standard deviation, versus 2.3 percent with a 2.9 spread in the decliner cohort.

Quality labels that failed the crash-window check

Large-index membership, high dividend payout, larger market value, beta, and return on equity did not mark the more resilient cohort in this sample.

Those familiar proxies are quality labels. Editorial view: a quality label that fails a crash-window test should be retired from that use, even if it is widely treated as a safety marker.

Traits that accompanied limited loss

The more resilient cohort had a lower average share price, a smaller price change from the prior January 1, lighter pre-shock trading in shares and as a share of shares outstanding, and smaller capitalization.

Return on assets was lower in the more resilient cohort, which the analysis linked to a possibly higher mix of tangible assets.

No single fundamental or technical statistic stood out as a predictor of subsequent price action. The comparison was framed as a safety contrast rather than a forecast.

A distinct crash regime

Editorial framing: market regime classification here means treating an abrupt cross-market break as a distinct regime and asking which construction traits clustered in names that did not follow the break.

Smaller later market declines produced a similar contrast that was less marked than the 1987 window.