Charles Franklin twitted a plot pointing to the similarity between Hillary Clinton’s and Tim Kaine’s common space DW-Nominate scores. So what the heck is anyone supposed to make out of this?
One reasonably sensible thing to do, which someone raised in the twitter thread, was to somehow compare the states that Clinton and Kaine represented, noting, presumably that Virginia is supposedly much more conservative state than New York. This is a common thing to do, and I am myself guilty of having done this, this is wrong for a very simple reason: Tim Kaine did not represent Virginia, at least not the whole state. Kaine represented the subset of the Virginia electorate that voted for him, most of whom are Democrats. Have we any reason to believe that Virginia Democrats are necessarily much, if any, more “liberal” than their New York counterparts? Given the present trends in demographics, that is highly improbable.
To illustrate this, let us consider two hypothetical electorates. In the first electorate, the makeup is 60% “Democrats” and, in the second, the makeup is 51% “Democrats.” The Democrats in the first electorate, however, have much greater variance: on average, only 80% of the Democrats would vote for the Democratic candidate, while, in the second, 99.9% would. This means that, on average, the Democrat representing the former is assured only of 48% of the electorate just by being a Democrat, while in the latter, she is assured of 51%. 51% wins elections 100% of the time.
The interplay between the conditional variance of the voting behavior by the electorate and the candidates’ characteristics needs to be fleshed out further for this baseline model to gain more traction, but the point is that having a larger partisan voteshare means nothing in face of a more variable voting behavior. It was not uncommon for wrong party’s candidates to win even in seemingly partisan states because they could cobble together bipartisan/non-ideological coalitions. This was not a luxury, but an essential path to survival in face of fickler electorates where just having “Democrats” as the only source of your votes was not sufficient. This is far less the case today, now that electorates are both more partisan, more ideological, and more predictable.
One might say, of course, that Virginia is a swing state where just having the Democrats may not be sufficient to consistently win elections. If you get unlucky and more Republicans come out, would you not lose? Would you not want to gain some pieces of independent or even Republican electorates as an insurance against such possibilities? The short answer is that you would if the insurance were sufficiently cheap. But it is not, in the era of polarization. For a Democrat to gain any appreciable of Republican votes is, I imagine, sufficiently difficult that most of them would simply stick to the party line and take the chance that the turnout winds are sufficiently favorable (or, if able, try to wind up the GOTV machinery–money for those operations costs a lot cheaper than the political price that must be paid to gain crossover votes).
In this sense, then, the collapse (literally, in the statistical sense) of DW-Nominate as an informative measure that meaningfully shows the differences in politicians’ voting record, never mind this nebulous “ideology” nonsense, can be placed in context. Everyone votes the same because they appeal to the same voters, who just happen to be present in greater or fewer numbers in different jurisdictions. Given the low variance in voting patterns, just having a slight majority that is reliable enough is as good as having a big majority. Since money is cheaper than politics, everyone spends lots of the former–even if it doesn’t buy a whole lot. Maybe Sanders is right after all: money in campaigns is at the root of the polarization problem. Make money more expensive than politics, then people will start paying political costs rather than just throw money at vote-getting machinery.