The presidential election is less than a month away, and polling shows a tight race between Donald Trump and Kamala Harris. While no one knows who will win, history suggests the answer may depend on how the stock market fares between now and Election Day.

A new report from stock market researcher Leuthold Group says the Dow Jones Industrial Average () may tip the scales. If the Dow, up nearly 12% this year, keeps chugging along, that could lift the vice president to victory. If the Dow falters, it may make the difference for the former president.

A longstanding rule of thumb suggests that, if stocks are up in the three months before Election Day, the incumbent party usually wins. The method, using the S&P 500 () as a measure of the stock market, correctly predicted the outcome in 20 of the past 24 presidential elections going back to 1928, according to Leuthold Group.

The firm decided to adjust the methodology to see if it could find a better predictor. Leuthold looked at the performance of the Dow, rather than the S&P 500, over the 11 weeks before Election Day rather than 12. The result? Leuthold’s number crunching found that the Dow correctly predicted the result in 22 of the past 24 elections. In other words, this method picked the winner 92% of the time, missing only Dwight Eisenhower’s 1956 re-election and Richard Nixon’s 1968 victory over Hubert Humphrey at the height of the Vietnam War.

While these stats may be tantalizing, there are drawbacks. First, since both methods rely on stock market returns through Election Day, their predictive value is limited. The Dow is up 5.5% since Leuthold’s key 11-week window started on Aug. 13. That bodes well for Harris, but anything could happen to stocks between now and Nov. 5. On Monday, for example, the Dow closed down 399 points, or 0.9%.

In addition, while it is easy to see how the state of the stock market might affect voters’ moods, it is hard not to get the impression that an 11-week window is the result of cherry-picking data. For its part, Leuthold admitted as much, calling its refinement of the S&P 500’s three-month rule an “unapologetic curve-fitting exercise” it did in part “for fun.”

All the same, Leuthold does seem to be on to something. The Dow hit a record just last Friday, helped by a strong jobs report. Harris has maintained a small but consistent lead in the polls. And recent polls suggest she is making inroads with some voters on the economy, long a strong talking point for Trump.

Voters could also take a look at the same stock market numbers and reach a different conclusion. Only about 58% of U.S. families own stocks, according to the Pew Research Center. What most voters ultimately care about isn’t the stock market but the economy. While the stock market is usually a good proxy for the economy’s overall health, pundits have been talking for months about a disconnect between the market’s performance and consumer sentiment.

Then there is everything else going on—from foreign wars to immigration to the candidates themselves—that could eclipse the stock market in voters’ minds. While Harris is a member of the incumbent party, she’s currently not serving as president.

To that end, it is difficult to ignore the last time the Dow’s returns failed to predict the outcome of the presidential election. The 1968 election featured a deeply divided electorate, a president who dropped out after deciding he couldn’t win, and a vice president striving to take his place.

Studying market returns provides one more piece of data for anyone handicapping the coming election. But—whatever the Dow says—it is still very much anyone’s race to win.