Testing bond and stock seasonality before you trade

Put one seasonal idea inside a broader market context

Editorial view: a single seasonal or intermarket idea is easier to misread when a month, a last day, or a crash is treated as a standalone signal. The historical workflow below uses calendar and regime checks that place one bond or stock idea inside a broader market context.

The first step was not a slogan. It was hypothesis testing: asking whether an anecdotal market rule still holds after counting same-direction, opposite-direction, and unchanged cases against an explicit historical baseline.

Count how often stocks and bonds actually move together

A 10-year daily sample showed stocks and bonds moving in the same direction 61.6% of the time, in opposite directions 34.7% of the time, and with one market unchanged 3.7% of the time.

Weekly price-change counts in the same comparison were 57.2% same direction, 41.6% opposite, and 1.2% unchanged. Monthly price-change counts were 60% same direction, 39.1% opposite, and 0.8% unchanged.

Editorial view: the three intervals supply a baseline for a diversified or regime-aware context. Same-direction moves were the majority case at each horizon, so an inverse stock-bond story has to be checked against that count rather than assumed.

Treat a crash as a regime episode, not a lasting inverse rule

Opposite stock-bond moves after the 1987 and 1990 equity crashes were treated as a regime episode to test, not as proof of a lasting inverse relationship.

Editorial view: that is how seasonal analysis is used here. A crash can mark a regime episode in which the two markets travel opposite ways. It does not, by itself, replace the baseline in which same-direction days, weeks, and months were more common than opposite ones.

Stress-test a last-day bond story with a random calendar

A last-day-of-month bond-up rule was checked by placing observed daily changes back into the same calendar so the month-end direction would be random under the null. That randomization was repeated 2,500 times, recording up, down, and unchanged last-day counts on each trial.

That procedure is Monte Carlo simulation in the sense used here: repeated reshuffling of historical daily changes into a new calendar so the observed count of up days can be compared with a distribution of chance outcomes.

In the actual 120-month sample, Treasury bonds rose on 73 last days of the month, and only 24 of the 2,500 random trials produced 73 or more up days.

Editorial view: the last-day story stays a claim until the observed up-day count can be set beside the chance distribution created by the reshuffles. It is not treated as a trade at the slogan stage.

Ask the same of a January claim

The same research program treated a January barometer applied to bonds as another calendar claim that required tabulated year-by-year direction checks rather than an untested seasonal slogan.

Editorial view: a January barometer is the same class of claim as a last-day rule. Both belong in calendar and regime checks, and both remain untested slogans until year-by-year directions are counted against an explicit baseline.