The prospect of Donald Trump becoming president is unconsciously affecting bridge players so much that they become “subtly deranged,” researchers reported on Friday, with players succeeding with “no trump” games more often than before the real estate developer entered politics.
The study is real. The researchers are respected statisticians. The meeting where they presented the discovery is legitimate. The “no trump effect” is bogus.
In the same vein as John Oliver’s recent takedown of “scientific” studies, statisticians analyzed data from bridge tournaments to see if Trump’s march to the White House has “irrationally changed the behavior” of elite players.
They compared hands played in 1999 and 2015 at one of bridge’s premier events, the Vanderbilt Knockout Tournament. Although the 2015 tournament was played before Trump announced his candidacy, the chattering class expected him to. That “makes our results conservative,” Andrew Gelman of Columbia University and Jonathan Falk of NERA Economic Consulting write, “and we would expect even stronger effects as his path to the nomination gained strength.”
(Readers for whom “bridge” and the game’s “no trump” bids are foreign concepts are invited to consult Wikipedia.)
The researchers adhered to good scientific practice by starting with a hypothesis — namely, that the idea of a President Trump would make bridge players (whom they believe to lean Democratic, or why would a liberal super PAC be called American Bridge?) unconsciously try harder to win their no trump bids. The bidders would unconsciously focus and strategize more intently, and their opponents would also be so unconsciously “deranged” at the possibility of a no trump bid failing that they don’t put up much of a fight.
Data supported that hypothesis. In the 1999 Vanderbilt tournament, 12.60 percent of hands were successful no trump bids; in 2015, 19.97 percent were. There was an even greater increase in the percent of no trump bids that succeeded: 48.50 percent in 1999 and 69.32 percent in 2015.
“Back in 1999, one could afford to be nonchalant about declaring no trump since the decision to do so had no overt political implications,” the researchers wrote. But “in 2015, if you’re going to state the word ‘Trump’ aloud, you need extra confidence that you can succeed, therefore inoculating yourself against a charge of political bias.”
The possibility of what Gelman and Falk found occurring by chance (known as the p value — and yes, we know that’s not the precise definition, but please hold your irate emails) is 0.0492. Most research journals require study findings to have a p value below 0.05, so the association between Trump’s political rise and a higher rate of success for no trump games easily ducked under that bar.
The Trump effect on bridge resonates with one player. “My defense against no-trump bids is distracted because I’m worrying that my hand is too small,” said Al Lewis, a Boston-based healthcare entrepreneur and cofounder of Quizzify.
The purpose of the paper, Gelman said, was to call attention to studies “that report dramatic and silly claims based on statistically significant p-values.” Among his favorites: that obesity is contagious and that scores of everyday foods either increase or decrease the risk of cancer. Such claims go up in smoke with better research.
“We’ve written this paper in an attempt to draw some attention to the prevalence of junk science,” Gelman said. In particular, cherry-picking data, massaging statistics (called “p hacking”), and publishing only results that find eye-popping associations mislead consumers, doctors, policy makers, and scientists.
“If you’re not aware of p hacking, you’re flying blind,” said Gary Schwitzer, who watchdogs such things at HealthNewsReviews.
No study should end without a call for further research. For the no trump effect, that might include examining how people’s subliminal association between “Trump” and the words “deal” and “hands” (small ones) affects poker and other card games.