The Democrats are picking their presidential candidate. The primaries and nominating convention are meant to surface that candidate who best represents Democratic ideals and voters. But what is the chance that the nominated candidate will win the electoral college? The answer is captured in the notion of electability, that is, the probability that, if nominated, a candidate will win the presidential election.
Prediction markets offer a clear answer to this question: Former New York City Mayor Michael Bloomberg is the most electable Democrat. Indeed, as of this writing, the largest political market – Betfair, with more than $25 million in play – suggests that if nominated, Bloomberg is the only viable Democrat whose electability crosses 50 percent.
Why believe this when recent polls of hypothetical head-to-head match-ups suggest that more voters would prefer not just Bloomberg but any of former Vice President Joe Biden, former South Bend, Ind., Mayor Pete Buttigieg, Sens. Amy Klobuchar (D-Minn.), Bernie Sanders (I-Vt.) or Elizabeth Warren (D-Mass.) to President Trump? Polls produce snapshots of current voter sentiment by asking, if the election were held today, for whom would you vote? Such polls are horse races that ignore the actual rules of presidential elections, which are won or lost in the electoral college. By contrast, political markets trade contracts that pay off based on actual election results, forcing traders to contemplate future events and reason backwards while taking the rules of the electoral college into account.
While prediction markets are not perfect, a high-volume market such as Betfair provides a credible vehicle for expressing the “wisdom of crowds.” In such a market, the price of an all-or-nothing security that pays off if an event such as “Sanders wins the election” occurs is an estimate for the underlying likelihood of that event.
Deducing a security price from Betfair requires a small intermediate calculation. Consider again the event “Sanders wins the election.” Betfair presently reports decimal odds of 5.9 and 6 for this outcome, depending upon whether a trader chooses to “back” (believes Sanders will win) or “lay” (believes Sanders will not win). Averaging the reciprocals of the “back” and “lay” odds yields the estimated price.
Continuing our example, Sanders’s market-based estimate of winning the election is given by (1/5.9 + 1/6) / 2 or about 17 percent. It is thus a simple matter to obtain estimated prices – probabilities – for various candidates winning the nomination or winning the presidency from Betfair.
Having determined the market prices for winning the election and the nomination as described above, computing a candidate’s electability is a simple matter: Divide the security price for a candidate winning the presidency by the security price for this same candidate winning the nomination. The result estimates the probability that if nominated, that candidate will win the election — the electability.
Among Democratic contenders priced at a 1 percent or better chance of winning the nomination, currently Betfair judges Sanders to have the best chance of winning the nomination (41 percent) and the presidency (17 percent). But his 41 percent electability (17/41) is much lower than Bloomberg’s 53 percent (16/30). Repeating this calculation for the other viable candidates reveals electabilities of 41 percent (Biden), 32 percent (Buttigieg) and 35 percent (Klobuchar), respectively.
Markets are volatile, of course, and political markets are no exception. But for the past several weeks, Bloomberg’s electability has consistently registered above 50 percent, and he is the only candidate who has done so.
The market has spoken: If the Democrats want their candidate to actually win the election, they should nominate Bloomberg the electable.
Edward H. Kaplan is the William N. and Marie A. Beach Professor of Operations Research at the Yale University School of Management. He is an expert in probability and statistics who has published academic research on prediction markets and presidential elections.