Forget tracking polls, focus groups and kissing babies. The most accurate predictor of election-year winners might be the stock market.

InvesTech, a research firm in the decidedly non-financial hub of Whitefish, Mont., has examined election year cycles over the past 100 years.

The conclusion?

Stock market increases in the two months prior to a presidential election favor the incumbent. Stock market decreases for the same period favor the challenger. The trend has presented itself almost 90% of the time (89% to be exact) in the past 28 elections, failing in just three; 1956, 1968 (though incumbent LBJ did not run) and 2004.

The reason, according to InvesTech, is that an increasing stock market is accompanied by increased consumer sentiment and overall satisfaction, which leads to feelings of comfort, an important psychological gain just prior to the election. Conversely, stock market declines are accompanied by volatility and investor (and voter) concern.

“We’ve found that overall, the market typically performs best in the third year of an election cycle,” says Eric Vermulm, senior portfolio manager with Stack Financial management, InvesTech’s investment arm. “It performs well during the fourth year (election year) as well, just typically not as well as the third year.”

Vermulm declined to comment on the reason for the three outlier years, noting the firm doesn’t analyze political issues.

“We instead spend a lot of time analyzing years and quarters, looking for trends and importantly, warning flags,” he says.

While the election is still months away (eons, politically speaking), it should be noted that the Bloomberg Consumer Comfort Index rose to minus 39.8 in the period ended Feb. 14 from minus 41.7 the previous week. It marked just the third time since April 2008 that the gauge has climbed above minus 40, a reading consistent with recessions or their aftermath. The monthly expectations gauge climbed to minus 7 in February, also a 12-month high.