The Battle of the Data Nerds, featuring Wasserman and Cohn and Silver, that I mentioned yesterday rages on at Twitter. I mentioned that Wasserman and Cohn agree that Trump's Electoral College advantage is growing, and that to win in 2020 the Democratic candidate probably will need to win the national vote by somewhat more than Clinton's 2.9 million margin in 2016. Silver is agnostic—he thinks it's too early to gauge. Challenging Wasserman, he asks whether he'd predict Trump is reelected if it were now known that the Democratic candidate, whoever it is, will win the popular vote by 3 million. Wasserman answers that he'd "lean" toward Trump's reelection in that scenario. In this analysis, one cannot predict Trump's defeat on the theory that he won't draw to an inside straight two hands in a row. To make that bet, you'd have to know that the deck is standard, and the Electoral College is not a standard deck.
For those interested in this kind of stuff, it's fun to observe the kinds of considerations that inform the analysis. The problem for the Democratic candidate is that she (or he) will likely make large gains in big states—think, California and Texas—without winning a single additional electoral vote: not in California, because you don't get a bonus for winning by 5 million instead of 4 million, and not in Texas, because the Democrat can narrow the gap by several hundred thousand without getting over the top. People are skeptical about this since it seems impossible that the Democrat should make "big gains" in California and Texas without gaining anything in, say, Wisconsin, which Trump won by fewer than 25,000 ballots (about 0.7%). But it's not impossible—for it's exactly what happened in 2016, when Clinton, compared to Obama in 2012, did much better in California and Texas while falling back in Pennsylvania, Ohio, Michigan, and Wisconsin.
The details are probably surprising to more casual observers of American presidential politics. Nevertheless it's true that in 2012 Obama carried California by 3 million votes—and then Clinton, in 2016, won it by 4.2 million. Meanwhile, Obama in 2012 lost Texas by 1.2 million, and Clinton, four years later, made up 400,000 to lose by only 800,000. But, in the same election in which the Democrat gained a total of around 1.6 million votes in California and Texas, that same Democratic candidate lost ground in Pennsylvania (340,000), Michigan (460,000), and Wisconsin (220,000).
Here's how the Electoral College "processed" these changing fortunes. In 2012, Obama got 101 electoral votes in the five states to just 38 for Romney. Then in 2016 Clinton outperformed Obama by more than half a million—up 1.6 million in the two sunny states, down about a million in the three Great Lakes states—but, instead of winning the electoral vote tally by 101-38, she lost by 55-84. Up 570,000 in actual votes, down 46 in the Electoral College. All Wasserman is saying is that the trend will likely persist. The Democrat in 2020, whoever it is, could very well be up another million in California, and a half million in Texas, and still lose after Trump ekes it out again in Wisconsin.
Comments