Starting to Sort Out the Election

November 29, 2016

There’s lots to sort out. Obviously, that it was a disaster is not something we need to sort out: We know that it was a disaster. The precise contours of the disaster will only become clear over time, but it will be pretty damn bad.

We do need to come to an understanding of what happened, including what we can or cannot know about it. Joe from lowell suggests that the election was analogous to the Iraq war from a punditry (and maybe political science?) perspective: That is, people got it very wrong and those people should have some humility and be treated as at least somewhat unreliable.

I said stuff before the election about the election outcome, so I need to figure out if I’m unreliable (or to what degree I’m unreliable). Note that this isn’t a flagellation exercise, or at least, not intended to be primarily one. If I’m relying on flawed models I should update my models! This election does seem to potentially provide some interesting new data, in any case.

Before the election:

And I really don’t think Karen24’s Llama Drama is because she’s a Rat Rogerer. I think she’s a Nervous Nellie who keeps panicking in the comments here at least partly because she can be so authoritatively reassured by other commenters. For those of us who are freaked out by the reminders that it’s not an automatic total blowout even with Donald Trump on the ballot, because (1) the MSM won’t do its fucking job and (2) so many American voters are still such horrid reactionary dumbshits, I think you underestimate how soothing it is for Bijan Parsia et al. to set us straight.

Here’s an example in that thread of me “setting them straight” with “authoritativeness”:

Look at the pattern though. “Trend” doesn’t mean “current slope”.

And look at several:

It’s annoying, but there’s no evidence that he’s winning or that his line will cross hers. If it does, that will be a surprise.

I think that what I said is consistent with what happened, so I guess I would say it again in relevantly similar situations. Should the people who trusted me before trust me in the future? I…guess? I’m not going to overpredict worst case scenarios, but I don’t think I was overconfident.

One thing that did come out of this election was the idea that many aspects of campaigns matter even less than we thought. The classic view about “fundamentals” based models of elections is that modern campaigns are fairly evenly balanced so tend to cancel each other out. So all that’s left are things like the state of the economy.

One thing that was clear in this election was that the campaigns were very lop sided on some key features like get out the vote operations. This suggested that it would be a powerful natural experiment!

I would guess this range underestimates Clinton’s chances, because the models can’t account for Trump’s unusually unprofessional campaign

I’ve felt this a few times, but then I start to worry that perhaps an unusual (or unusually unprofessional) campaign might not be the sort of drag we’ve thought it would be.

Just consider two factors:

1) Advertising, esp. television
2) GOTV, esp. day of ground game

Up until now, HRC has had the airwaves to herself. It’s not clear that it’s done any good. Or, I’d like to know how it’s done good. Maybe it’s too early. Maybe it just hasn’t shown up in polling. Who knows. In the primaries, the rest of the candidates held off and Trump never collapsed. So we don’t really know what advertising will do to this heavily exposed, extremely strange candidate.

What is the effect of GOTV efforts? I believe that it will have an effect, but it might be pretty small compared to random turn out effects. It’s good bit of body english, but it doesn’t seem to be the makings of a blow out. (I’d love pointers to literature on this.)

Maybe campaigns *really* don’t matter?

wtjs formulated an interpretation of possible outcomes:


But what would show us that the inept vs. ept campaign didn’t matter?

Certainly a Trump win (absent a major exogenous shock like 9/11) would. A win for Clinton along the lines of Obama’s reelection might. A convincing win for Clinton in the popular vote plus a sweep or near-sweep of the toss-up states and a win in one or two “lean R” states in the Electoral College would be pretty good evidence that campaigns do matter.

So, there we go?

One potential confounding factor is “the media” and, frankly, the FBI. One alternative story is that the last minute Comney letters were fairly strong events (given all the media priming on EMAILZ!!!) that worked to the Trump campaign’s advantage.

In general, Democrats seem to have a disadvantage in turn out (as we see dramatically in midterm elections). There are almost certainly loads of factors, but that might start the scales strongly against them. This would explain that the “vote against Trump” effect wasn’t so very strong.

Of course, the simplest explanation is just that fundamentals set the stage and a certain amount of randomness completes the job. If you make it to a major party nomination, you can win. Trump was slightly favored by the fundamentals, strongly hampered by his campaigned, but had a bit of luck.

This isn’t very satisfying! It’s particularly unhelpful for any future planning. (Lots of post mortem analysis is actually quite worthless for future planning. “Nominate a better candidate” doesn’t really help as no one is really in control of the nominations! This conclusion suggests that Bernie would not have done better. If campaign effects are small they are small in both directions. My guess is that he would have been “as likely” to lose…that is, not very likely, but it would be possible.)

Turn out is obviously critical, but our understanding of how to goose turnout is really poor. Voter suppression efforts are pure evil, but it’s unclear that they are turning elections (yet). I can’t find the reference, but apparently the Clinton campaign contacted twice the voters that the Trump campaign did. Now, some are suggesting that they ended up turning out Trump voters, but I don’t think we know the actual effects yet. If we go back to Romney-Obama, we see a large gap in effort but not a dramatic advantage:

We estimate that the presidential campaigns increased turnout by more than 10 percentage points among targeted subgroups, indicating that modern campaigns can significantly alter the size and composition of the voting population.

In this paper, we exploit the 2012 presidential campaign to assess the aggregate effects of a large-scale campaign on the size and composition of the voting population. We take advantage of variation in ground campaigning across state boundaries, extensive information on Romney and Obama campaign tactics, and detailed information on every voter in the United States to estimate the effects of the entire campaign. Our results suggest that the aggregate mobilizing effect of a presidential campaign is quite large. We estimate that the 2012 campaign increased aggregate turnout by approximately 7 percentage points

This analysis also allows us to compare the relative effectiveness of the Obama and Romney campaigns. The Obama campaign of 2012 has been championed as the most technologically-sophisticated, evidence-based campaign in history while the Romney campaign was more traditional (e.g., Issenberg 2013). When we began this project, we surveyed 46 academics, and they predicted that Obama’s campaign was almost 3 times as effective as Romney’s in mobilizing supporters.17 Do these perceptions manifest themselves in the data?

As discussed above, this analysis allows us to roughly compare the effectiveness of the Obama and Romney campaigns in mobilizing their respective supporters. Despite the purported technological sophistication of the Obama campaign and its devotion to a data-driven, evidence- based campaign, we see similar mobilization effects on both sides of Figure 3. The two campaigns were roughly comparable in their ability to turn out supporters.

One interpretation of the Romney campaign’s slight advantage with their own partisans is that Democrats are simply harder to mobilize than Republicans. Indeed, previous research suggests that, on average, GOTV interventions are more effective for conservative and high-socioeconomic- status citizens (Enos, Fowler, Vavreck 2014). The amount of effort and resources needed to mobilize Democratic supporters may be greater than that needed to mobilize Republican supporters. With this in mind, even if the Obama campaign was more advanced than the Romney campaign, this difference was not great enough to overcome this structural disadvantage

If this is confirmed in the 2016 data, it points a direction, at least, for the Democratic party: It needs needs needs to solve the GOTV problem. This would solve all sorts of problems including midterm elections, statehouse controls, etc. etc.

The problem is that we really don’t know how.

On a related note, Nate Silver is sorta claiming that his prediction was more accurate because it gave more weight to Trump’s chances of winning (plus some details of why his model gave such weight…roughly, the scenario that played out was one his model could and did contemplate). I thought that 538’s coverage and the Silver models were poorer than alternatives, esp. Sam Wang‘s. Now , Wang’s prediction was much more heavily weighted toward Clinton (i.e., 99% chance vs. ≈70% in Silver’s).

I need to think more about this.

4 Responses to “Starting to Sort Out the Election”

  1. I would caution against treating turnout solely as a function of formal GOTV operations. It’s probably more useful to look at it as a force multiplier for actual enthusiasm in the electorate.

    Yes, the biggest thing the Democrats need to change from this election is turnout among core constituencies (in this case, younger voters of all races and both genders). But that doesn’t mean the answer is better software and more door-knocking.

    • Bijan Parsia Says:

      Oh, I agree and don’t think that classic ground game is the be all. (I also don’t think I suggested that above :)) If anything, the 2016 shows that, absent some revolution, data driven ground game isn’t a salvation. It seemed to have a marginal net effect in 2012 and an insufficient effect in 2016.

      It’s a deep problem.

  2. halfspin Says:

    I see a huge difference between the pundits’ treatment of the Iraq War and the political scientists’ treatment of this election. All the actual expert and historical evidence suggested that the Iraq War would turn out to be a disaster, and it was wishful and misguided and emotional thinking that led people to predict otherwise. The opposite was generally true for this election.

    As for the Wang vs. Silver debate, I preferred Wang because he fully disclosed and justified his methods and data. It turned out that his assumption that inter-poll variance correlated with the difference between polls and election results was incorrect. So was Silver’s trade-secret-sauce prediction model for the GOP primary though.

    (As a side note, nice to hear from you again, jfL!)

  3. […] interpretation of the election is that campaign quality doesn’t matter at all, given Trumps weak traditional campaign. Specifically, campaign quality […]

Leave a Reply

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

You are commenting using your 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

%d bloggers like this: