Predicting the General from Primaries

There is a widespread inclination to treat the primaries as an entirely separate phenomenon from the general election.  In a sense, this is wholly justified for reasons empirical and theoretical.  As the fivethirtyeight.com post covered the empirical aspects of this justification, I will offer the theoretical:  the choices facing the voters are different in the primaries and the general.  People choose to spend their time on voting for a reason, and that reason is supplied by the candidates and the circumstances under which the election takes place, even if the voters themselves remain the same.  If the candidates and the circumstances change, there is no reason to expect that the same voters would continue to participate.

This is hardly a new observation:  one reason Gary Hart was thought to be the more formidable candidate in 1984, at least until the scandal involving Donna Rice came out–and changed the circumstances of the election, so to speak–was the recognition that he activated a different set of voters from Walter Mondale and many if not most of those voters were expected to remain in support of Hart candidacy in the general but not necessarily Mondale.  The fall of Hart candidacy, of course, ensured that this could not be evaluated, of course, and furthermore underscores the difficulties of evaluating the impact of “unusual” candidates on turnout by different demographics:  candidates who try to draw on “unusual voters” to win the nomination rarely succeed in winning the nomination.

On top of these, we just don’t have that many presidential elections to draw data from and analyze to the satisfaction of statistical significance.  This limits us to two responses:  we can refuse to believe the data that we have since they are not close to being conclusive, or we can try to learn what we can from the data, but with the recognition that the data is not very good and the conclusions drawn from them largely unreliable.  Science demands that we should stick to the former, if we want to gain good, reliable insights.  But we are trying to gain some insights, not necessarily good and reliable ones.  With appropriate caveats, I see little reason not to.

The search for a historical analogue to 2016 brings us to 1980:  the primary turnout was high on the Republican side, due to many voters enthused by Ronald Reagan as well as by many opposed to him.  Indeed, the #NeverReagan movement within the Republican Party gained enough support that John Anderson ran a third party candidacy.  Reagan was considered, despite his tenure as governor of California, considered a political incompetent who did not know what he was doing and an out of touch extremist who might bring the Republican Party by bleeding disenchanted voters, while he claimed that he was drawing new voters in.  It is premature to conclude that Trump would necessarily win because he is like Reagan, for the opposition to Reagan within the Republican Party was nowhere as strong as Trump’s today.  Reagan’s support was so large than he carried so many primaries by huge margins after a few hiccups in the early going.  It certainly helped that, for all his non-mainstream appeal, he was also an experienced electoral politician who did, after all, nearly derail a sitting president’s (Gerald Ford’s) campaign in 1976 in addition to his time as a large state’s two-term governor.  Trump, in contrast, seems to have hit a hard ceiling around 40-45% even within the Republican Party.

Reagan, presumably, would have activated similar kinds of demographics in the early primaries in both 1976 and 1980.  In 1976, with Ford, rather than Reagan as the nominee, did these voters stay home, while, in 1980, with Reagan at the top of the ticket, did the same voters stay onboard?  The demographics of Reagan and Ford electorates do seem different enough, with Ford’s being much more upscale in terms of income and more “conservative” (in practice, this just means that Reagan drew a lot more voters who did not fit liberal-conservative scale neatly, not that he drew necessarily “moderate” voters–the same reason Trump draws many “moderate” voters), although Reagan’s coalition in 1980 does not look quite like his 1984 coalition.  Since the exit polls themselves from these years are, unfortunately inaccessible for me at this time, the more detailed analysis will have to wait.  But they suggest that some aspects of the general election coalition may be predictable from the primary results:  IF Reagan’s early success was dependent on blue collar, apolitical voters, they seem to have carried over to his general election coalition.  (It is also worth noting that head-to-head matchups in 1980 showed Carter defeating Reagan handily until August, at least)

After Reagan, there was another primary presidential candidate who drew out unusual voters and won primaries unexpectedly riding on them, and mostly kept them in the general–Barak Obama in 2008.  Other than these, we just don’t have primary polling data that is readily available (Roper Center’s data goes back only to 1976).

So, two data points, from two very unusual presidential candidates, Reagan in 1980 and Obama in 2008, that MIGHT, upon further analysis, suggest that who turns out in primaries might give us useful clues, although not definite guidance, about who would turn up in the general election–although the overall turnout might well be a red herring.  In 2016, we have another strange election brewing, with another candidate drawing out strange voters.  The information we have is almost completely unreliable and I would not suggest for a minute that there is much that is reliable here.  However, we also know that where the data is plentiful, there is no useful insight to be had.  So between reliable but useless information and totally unreliable but perhaps insightful information, what shall it be?

Advertisements

3 thoughts on “Predicting the General from Primaries

  1. frozengarlic March 20, 2016 / 6:24 pm

    You are paying a lot of attention to the idea of the missing white voter. Do you really think Trump will be able to scale up his appeal to this group in the general election? I think it is more likely that he has already gotten all the “extra” voters that he can attract. With all the negative publicity that he is getting and will continue to get, if you haven’t already decided you are for him, it doesn’t seem very likely to me that he will be able to convince you that he is “presidential” or that he can somehow lead America to “win.”

    Like

    • anon9999 March 20, 2016 / 6:40 pm

      That is the great big question that everyone is pondering about. The striking development in the last few election cycles is that even as the overall turnout has been rising (well, it fell sharply in 2014, even for a midterm), the variance in turnout across different demographic groups has been growing as well, mirroring the kind of inequality seen in wealth and such. The working class whites (and apparently, younger, low-income black males as well) showed up in unusually low numbers in 2012 and 2014, continuing the trend that has been ongoing since 1990s, in particular. My sense is that this is the consequence of polarization and excessive agenda setting by party insiders in both parties: if all the only alternatives allowed to be left standing are those that are equally bad, even if they might be total “ideological opposites” (on the dimensions of little or no interest to a given set of voters), why should they think there’s a dime’s worth of difference between them?

      The primaries are in fact showing that all sorts of voters are coming out of woodworks, both to support and oppose Trump and Sanders. The remaining question, the really big one, is whether the hypothetical head to head polls are capturing the true extent of their support in the electorate. On the latter, I have no idea. It is pretty clear that the main reason Trump seems to do so poorly against either of the Democrats, compared to all other Republicans, is that Republican respondents are not willing to back Trump, at least on polls. The absence of additional support for Trump among the Independents could easily mean, I suppose, that there is no real “missing white voters” that remain to be counted–and if a chunk of Sanders supporters not available to Clinton (those who give him a bigger advantage against all Republicans) are liable to swing behind Trump as I keep suspecting, surely, they’d show up in the polls–which they don’t.

      so the short version of the answer: I have no idea. I keep thinking that there should be something there because it makes me really uncomfortable. But there really is no evidence that I can see that these voters are, in fact, “missing” presently.

      Like

      • anon9999 March 20, 2016 / 7:12 pm

        To elaborate further, we know, for general election results, that a significant demographic segment is missing because we can estimate turnout by each segment separately. So we know that there were missing white voters and that they were growing in number since 1990s onward, mostly in relatively safe Republican Congressional districts in the Midwest and the South. What is not clear is that the polls being taken in 2016 are accounting for them accurately, and there’s no obvious reason to suspect that that is so. Whether Republican regulars will stick with Trump or defect to the Democrats should he be the nominee is the more obvious question, since that is apparent in the polling results.

        As far as I can tell, the main filter that most polls I’ve seen are applying is whether the respondents are registered to vote already, and I have trouble seeing whether there will be an upsurge in registration that will take place between now that will bring these suspected “missing” voters into the electorate over next 7-8 months. One possibility that we can explore, if we had the raw data, is to explore whether the states that already had primaries (where, presumably, people had to register to participate in primaries) necessarily indicate something different from those that have not yet held primaries. All sorts of selection bias will be at work, though, in what is already a small N sample.

        Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s