Evaluating option-expiration weeks as index regimes

Three week types on the option calendar

Listed index options stopped trading on the third Friday of each month. The calendar week that contains that close is the option-expiration-week. The week immediately after it is the post-expiration-week. Any week that is neither of those is a nonexpiration-affected-week.

That expiration week was framed as often stronger, with the following week often reversing lower. Successive expirations are not a fixed interval, and the spacing can change from year to year. A seasonal or model input therefore needs week-type-encoding: a categorical mark for the three states, rather than a constant lag.

What was compared

The evaluation used weekly-point-change, the index move measured in points rather than percent, because listed strikes are set on absolute index levels. A two-tailed-mean-test compared average weekly point changes without assuming which week type was larger. Up-week-frequency compared the share of weeks whose index change was positive. The same questions were asked of medians and of weekly variance.

A sample-split-check then repeated the mean, frequency, and variance tests on a later window to see whether an earlier contrast still held.

Means, medians, and up weeks

In the full sample, the average weekly-point-change was positive overall, larger in option-expiration weeks, and negative in post-expiration weeks. The two-tailed-mean-test rejected equal average weekly changes for expiration versus post-expiration weeks. A comparison of medians matched that mean test.

Expiration weeks finished higher more often than post-expiration weeks. Nonexpiration-affected-weeks sat between those two rates. A two-proportion comparison flagged the expiration versus post-expiration frequency gap as the significant contrast.

Expiration weeks were not significantly stronger than other weeks in the full sample. Post-expiration weeks were significantly weaker than other weeks in the full sample. A crash week in the sample was noted as a possible skew.

Of the sample's cumulative rise, expiration weeks accounted for most of the total, while the week after expiration contributed a cumulative decline.

The later window and variance

In the later window, expiration weeks still finished higher more often than the following week, and other weeks remained in between. That more recent frequency gap fell below conventional statistical significance.

Expiration weeks were stronger than other weeks in the later window. Post-expiration weeks were not significantly weaker than other weeks in that window.

Tests of weekly variance did not confirm the expected pattern that expiration and post-expiration weeks were systematically more volatile than other weeks. The only significant volatility contrast was post-expiration versus other weeks, and only in the later sample.

Using the tag without a fixed clock

Because successive expirations can sit at different spacings from year to year, a rule that assumes a constant week lag will mislabel some weeks. Week-type-encoding keeps the three states as categories.

Editorial conclusion: the historical workflow supports treating option expiration as a weekly regime mark, then testing that mark. It does not, on its own, establish a trading rule. Mean, up-week-frequency, and variance still need to be rechecked on the window under study, and volatility should not be assumed from the calendar label.