The January barometer as a rest-of-year scoring problem

Keep January out of its own score

The January barometer is a seasonal hypothesis that the first calendar month's market direction forecasts later-year direction. In the historical workflow that claim is not scored on the full calendar year. It is scored on the rest-of-year window, the interval from the January close through the December close.

A first-month gain followed by a later decline is counted as a miss even if the full calendar year is still higher. The signal month is not allowed to serve as its own confirmation.

Pooled Dow counts still hide a split

On the Dow Jones Industrial Average, the January-to-rest-of-year rule was wrong more often than it was right from 1900 through 1937. From 1938 through 1989 the same Dow rule was recorded as correct 40 times and incorrect 12 times.

Those pooled counts still mix two different first-month states. After 1937, 29 of 33 January advances on the Dow were followed by a rest-of-year advance, while only 11 of 19 January declines were followed by a rest-of-year decline. The supplied tests therefore keep a bullish January regime apart from a bearish January regime.

Nest the first five sessions

After the month-end rule is stated, first-five-sessions direction is combined with full-January direction to form six mutually exclusive January chart patterns. That six-pattern January map is the complete set of early-January and full-month combinations, each treated as a distinct testable case.

Accuracy of those six patterns is judged on the S&P 500 from January 31 through December 31, not on the Dow used for the simple month-end count.

Rising January held up better than falling January

After 1937, 19 of 21 S&P 500 cases with a first-five-sessions advance that then extended through January were followed by a February-December advance. Combined bearish patterns 1 and 2 were correct on eight of twelve later-year outcomes, while pattern 6 was wrong more often than it was right.

Across the six-pattern ledger, rest-of-year forecasts that began with a rising January were more often correct than those that began with a falling January. In the supplied tests, a bearish January regime is a weaker and less consistent rest-of-year filter than a January advance.