Testing the July-August summer rally as an occurrence count

A two-month story with a frozen definition

A historical equity record beginning in 1885 was examined to test whether July and August show a recurring bullish seasonal bias.

A summer rally was coded as present when August 31 closed higher than June 30. That close-to-close rally rule turns the named summer-rally story into a yes-or-no count for each year.

Where the rise concentrated

July and August were split into six roughly 10-day parts, and each ten-day window was scored by the share of years in which the market rose. That occurrence count treats each slice as a frequency, not as a trade recommendation.

The first ten-day window rose in 69.5% of years, the highest frequency among the six parts. It is the holiday-adjacent window: it includes the early-July holiday, and the session immediately before that holiday was reported as especially bullish in the same tally.

Each of the remaining ten-day windows still showed more rising years than declining years, though those frequencies were described as less striking than the first part.

How often the full window rose

Across the full sample, the study reported about two-to-one historical odds that August 31 closed above June 30.

The 106-year sample was split into eight roughly 13-year era blocks. Occurrence varied by era, including better than four-to-one in 1924-1937 and absence in two-thirds of years in 1964-1977, with no clear trend across blocks.

An election-year overlay

Election years were compared with the full 106-year record and did not show an appreciable difference in July-August occurrence.

Editorial reading: the election-year overlay asks whether that calendar filter changes the two-month rate. It does not ask whether politics explains the seasonal pattern.

Frequency is not size

The archive facts keep the occurrence count apart from the magnitude sketch. Average July-August magnitude was illustrated by adding the historical average rally to a Dow Jones Industrial Average level of 3000 as a scale sketch, with the contemporaneous index level noted as subject to change.

Editorial reading: those two results answer different questions. How often the window rose is not the same claim as how large a typical July-August move was.