This post was originally published on this site
Be on your guard this election season for those using the stock market to predict who will win the White House in November. That’s because the historical data can be tortured to say almost anything.
This isn’t to say that there is no relationship between stock market performance, consumer confidence, voters’ moods and the incumbent’s chances of staying in the White House. But, as Campbell Harvey, a Duke University finance professor, told me in an email: “The sample [of past Presidential elections] is too small and there are hundreds of variables.”
Given this, it’s difficult to come up with any conclusion that has genuine statistical significance. Consider the notion that is currently making the rounds that the stock market’s direction between Labor Day and Election Day predicts whether the incumbent party wins. On the surface this appears to enjoy strong statistical support, at least since the late 1890s, when the Dow Jones Industrial Average DJIA, -0.78% was created.
However, as I pointed out in a column a couple of months ago, there is no statistically significant correlation between the incumbent’s re-election chances and the Dow’s return over the trailing one month prior to Election Day — or the trailing six months. Why should the market over a two-month period be a good election forecaster but not over a one- or six-month period?
These are statistical red flags, and they’re not the only ones. As I also pointed out in that earlier column, the Dow’s trailing nine-month return is strongly correlated with the incumbent’s re-election odds, but not the Dow’s trailing eight- or trailing 10-month returns.
This discussion brings up a broader statistical lesson we should always keep in mind: To properly assess the significance of a given correlation, we must take into account the number of hypotheses that were analyzed during the process of “discovering” that correlation. As the number of tested hypotheses increases, the odds also increase of reaching a conclusion that appears to be statistically significant but isn’t really — a false positive, so to speak.
Given the widespread availability of historical databases and the pressure on both academics as well as Wall Street analysts to discover “uncanny” correlations, the number of hypotheses that get tested is huge. My review of research into alleged correlations between the stock market and election outcomes found hypotheses covering periods as short as one month to as long as four years. It would be surprising if a false positive were not to emerge.
There is another reason to question the stock market’s ability to forecast this upcoming election: Even if it ever did have forecasting ability, it won’t this time around because keeping the bull market going has in effect become a policy goal of the U.S. government.
This ironic result is known in economic circles as Goodhart’s Law. Named for a British economist, Charles Goodhart, this law holds that when a measure becomes a target, it ceases to be a good measure.
Vincent Deluard, head of global macro strategy at investment firm StoneX, recently explained the origin of Goodhart’s Law in a note to clients: “The most famous examples of Goodhart’s Law were the Soviet factories which, when given targets on the basis of numbers of nails, produced many tiny useless nails, and when given targets on basis of weight produced a few giant nails. Numbers and weight both correlated well in a pre-central plan scenario. After they are made targets (in different times and periods), they lose that value.”
“ ‘The stock market, which used to be a measure of the economy, has become a policy target. As a result, it has lost its value as a measure of the real world.’ ”
“In our case,” Deluard continued, “the stock market, which used to be a measure of the economy, has become a policy target. As a result, it has lost its value as a measure of the real world.”
To nip in the bud the arguments that my discussion is politically motivated, let me remind you that it cuts both ways. If the stock market declines between now and Election Day, the lack of a correlation will be taken as being pro-Republican. And if the market instead rises over the next two months, that lack will be taken as being pro-Democrat.
Neither is my intent. The truth is that, when the data are tortured to say something, the conclusions tell us more about the torturers’ biases than about any objective election odds.
Mark Hulbert is a regular contributor to MarketWatch. His Hulbert Ratings tracks investment newsletters that pay a flat fee to be audited. He can be reached at mark@hulbertratings.com
More:An ‘extreme’ August on the stock market might be telling us something about the November election
Plus: Biden blasts Trump for having ‘fomented’ violence, as presidential race tightens
