Paul Campos writes:
I’m certainly open to the possibility that presidential campaigns are, under contemporary political and cultural conditions, becoming significantly less effective than they were in the past.
Actually, though, the consensus is that campaign effects (for the general election for president) were small in the past too. The classic reference on this is Rosenstone (1983), where he found that election outcomes could be predicted pretty well from economic and political “fundamentals” that were not affected by campaigning. The classic case study is 1988, when Dukakis was ahead of Bush Sr., who looked like a loser in ever way, and then Bush won. The point here is that, even when Dukakis was leading, people in the poli sci community were pretty sure that Bush was going to win, because they’d read Rosenstone. Following that was our 1993 article explaining why elections were predictable even though polls were variable, which featured this table:
The big change since then is that polls are no longer variable.
Regarding 2024: I think it was well understood that campaign effects are small. But with the election forecast being so close, this is the one setting where even small effects of less than 1 percentage point could make a difference. So I think 2024 is consistent with our understanding. The most intersting issue here is the role of inflation and the economy. Arguably the Republicans did well and the Democrats did poorly in the meta-campaign of what the economy was said to be about. Yet another way of saying this is that Trump was kicked out unfairly because of covid, and then, four yers later, Biden/Harris were kicked out unfairly because of covid.
P.S. In comments, Steve B. writes:
With all the idiosyncratic events that occurred this cycle—e.g., a failed assassination, an abrupt shift from Biden to Harris—the structural/fundamental-based models performed really well.
As usual, PS Political Science published forecasts for the 2024 presidential election. Here are some of the more traditional approaches to forecasting the incumbent party’s two-party vote share. I have not included models based on betting markets or more novel approaches. The Time for Change model wasn’t included in this journal. Therefore, I found the published prediction on Crystal Ball and also included my slight variant of that model.
Lewis-Beck & Tien (economic retrospective model): 48%
Lockerbie (economic prospective model): 49.1%
Algara et al (president fav and net party brand model): 47.2%
Woolley adaptation of Abramowitz Time for Change model: 50.1% (this result was found on the Sabato’s Crystal Ball website)
My version of the Time for Change model (with real disposable income instead of GDP): 49.8%
Simple average of all models: 48.8%.Current two-party margin for Harris (the models all forecast the incumbent’s margin, sans 3rd party candidates): 49.2% after rounding up.
Not too shabby!
I question the 2024 conjectures. From economic conditions alone, we should have expected the Republicans to win. Given Trump’s unique characteristics, that is one argument weighing against that result. The campaign ran a very effective (in my opinion) message about inflation, inflation, inflation, migrants, migrants, migrants, and the repeated question: are you better off than 4 years ago? I think the campaign had more than a small effect – it overcame the significant negative feelings about Trump to allow the economic priors to prevail. (I am not ignoring the large MAGA element in the population, but there was an even larger (I think) group harboring quite negative feelings about Trump).
Now that may be a small effect in a close election, as you say. But it depends on what you are comparing the result to – the prior based on economic conditions or the prior based on economic conditions + Trump as candidate. Only one of those is amenable to statistical analysis from past elections. I think the Trump effect is unique and we can’t really say whether the campaign effect was small or large as a result.
Dale –
The campaign ran a very effective (in my opinion) message about inflation, inflation, inflation, migrants, migrants, migrants, and the repeated question: are you better off than 4 years ago?
Don’t forget that the whole trans issue was a big focus of the Trump campaign. They spent a ton of money running the “She’s for the they/them and he’s for you” commercials.
Trump’s progress in the favorable/favorable ratio has been going on for a pretty long time. I doubt that it’s possible to tease out progress along that vector from the recent campaigning relative to his ongoing messaging. On the other hand, it does seem that there was a reversed trend, electorally, from what happened during the 2022 off-year election. So many factors to consider: abortion, inflation, immigration, crime, trans issues, Ukraine war, etc. Looks to me like trying to isolate a particular influence is bound to reflect a “degrees of freedom” kind of problem
“Don’t forget that the whole trans issue was a big focus of the Trump campaign. They spent a ton of money running the “She’s for the they/them and he’s for you” commercials.”
This is the opposite of the trans issue. The point of those commercials was to say that Harris and the Democrats were too focused on cultural issues. Democrats reacting to it as if it was anti-trans served to prove this point even more. I’ve seen several people saying this.
As it happens, I didn’t say it was “anti-trans.”
But clearly it was an effort to weaponize trans issues by exploiting statistically insignificant events – such as undocumented migrants getting sex change operations in prison paid for by the government (under policies initiated during Trump’s first administration).
It was basically the same formula as the “They’re eating the the pets” or the “they’ve got murder in their genes” strategy to weaponize immigration.
If you want to argue that’s not “anti-trans” or “anti-immigrant” have at it, but that doesn’t change that it’s clear scapegoating, fear-mongering demagoguery.
A counter-argument is being made by quite a few people who not that Harris negative swing relative to Biden was considerably smaller in the swing states where they focused the campaign than in other parts of the country.
Not sure I buy it, but it’s worth investigating.
Joshua:
Sure, but these differences are pretty small. I can believe that all sorts of things can have campaign effects in the 0.5-percentage-points-of-the-vote range, and big things (such as Harris being on the left wing of the Democratic party) can have effects in the 1-percentage-point range, and then there’s the idea that Trump in particular may have motivated many people who are not regular voters to vote . . . these things are not nothing, and they can make a difference in a close election, of which we’ve seen several since 2000.
Andrew –
I’ve heard people talking about a 4 or 5 point difference how well Harris did in the swing states relative to the rest of the country (in terms of a rightward swing). I’ll point out that those who are making that argument tend to be the same people who are defending the campaign decisions. That said, those who are blaming campaign decisions for the loss to Trump are basically, imo, engaging in facile arguments based on counterfactual assumptions. For example, they’ll blame campaigning with Cheney or not distancing from Biden without taking seriously the downsides of other choices (like Biden leaking that Harris never criticized his polices prior to getting the nomination).
Friends of mine have engaged in a spirited discussion of whether the results in the non-swing states can be used as a control group to measure the effect of the campaigns. There are many problems with this, including the obvious problem that so much campaigning is national. The fallback position is that the results in non-swing states are a control group for the effectiveness of swing-state advertising. This too faces the problem that, even in non-swing states, there may be individual Congrssional races that get so much campaigning that it contaminates the vote totals at the top of the ticket. The naysayers among us think that while there is probably some signal in there of the efficacy of swing-state expenditure (and of course reducing the campaigns to their expenditure is yet another problem — and when you don’t have a better variable for “effort” than that maybe you’d be better off not trying) there’s still going to be a ton of noise.
I’ve seen analyses that say, for example, “New Jersey and Virginia weren’t swing states, and they were really close, therefore Harris expenditures in swing states must have been really effective or she would have done as badly in North Carolina as she did (relatively) in NewJersey.” Note that this use of (relatively) in the previous sentence now sets up an additional complication…. an implicit difference-in-difference methodology that assumes that Biden 2020 ~= Harris 2024, and even that Trump 2020 ~= Trump 2024, among other dubious assumptions.
I’m not convinced that the argument holds up – that the better performance in swing states is an indicator of effective campaigning. However, I’m also kinda skeptical that it’s just coincidence that she did better in basically all the swing states than in basically all the non-swing states. I’m not sure what the explanation might be, but if it’s not a coincidence, then it is an interesting question as what the explanation might be.
As for your conjecture – it could be as simple as that they designed their campaign to target the swing states specifically, which would counter the argument that the campaign wasn’t differentiated by state. It would also more or less cancel out the Congressional campaigning factor, as that could be essentially irrelevant to whether they designed the campaign to target swing states.
But the main issue I would have is thst I can’t really figure out what it would mean to design a campaign for the swing states. Why, for example, would campaigning with Cheney to create attempt to create the appearance of a moderate be expected to have more of an impact on swing states as opposed to non-competitive states? Maybe because they think that the swing states are more likely to “swing” on the vote of moderates?
Back in 1983, I ran in a hotly contested election for a seat on the board of the Monterey Peninsula Water Management District (I’m not kidding; water limits growth in the area, so this was a big deal election, and we made the headline in the local paper when my coalition won). In the course of the campaign, I knocked on lots of doors and talked to lots of people, and learned that most people were ignorant of matters that had been well covered in the local paper (that was back in the days when everybody got the paper), even if they had strong feelings on the general topic. I think we won because our campaign was in sync with the local mood at the time, not because voters made a reasoned choice.
I think is consistent with the recent Data for Progress poll showing that the less attention people paid to the news, the more likely they were to support Trump (https://www.dataforprogress.org/insights/2024/11/14/what-political-news-engagement-tells-us-about-donald-trumps-victory)
There are a multitude of explanations for why the 2024 election resulted the way it did. The one explanation very few are willing to admit is what may be called, “The dung beetle explanation.” Although TED Talks are not necessarily generally looked upon favorably on this blog, have a look at this humorous and thought-provoking video, entitled “The Dance of the Dung Beetle”L
https://www.youtube.com/watch?v=qPjI6YFFgA4
The video is amusing and insightful because the behavior of the dung beetle is both tenacious and so different from us that it is inexplicable until one envisions the worldview of the dung beetle. Likewise, readers of this blog may be incapable of explaining Trump’s attraction to the overall American public.
Even granting that elections can largely be predicted from ” economic and political fundamentals” that doesn’t mean campaigns have little effect. It just means the net effect of both campaigns is small.
To see the size effect of campaigning it would be helpful if we could talk Americans into doing an experiment where for one election cycle, one side doesn’t campaign at all.
Anon:
Yes, I think we discuss this in our 1993 paper, that we’re talking about net campaign effects. And there are some local elections and referenda where the campaign is very one-sided, and I think the consensus is that in such settings of unequal campaigns, the net effect can be large.
Regarding your experiment: 2020 was pretty close to that, no? Biden didn’t do much campaigning.
With all the idiosyncratic events that occurred this cycle—e.g., a failed assassination, an abrupt shift from Biden to Harris—the structural/fundamental-based models performed really well.
As usual, PS Political Science published forecasts for the 2024 presidential election. Here are some of the more traditional approaches to forecasting the incumbent party’s two-party vote share. I have not included models based on betting markets or more novel approaches. The Time for Change model wasn’t included in this journal. Therefore, I found the published prediction on Crystal Ball and also included my slight variant of that model.
Lewis-Beck & Tien (economic retrospective model): 48%
Lockerbie (economic prospective model): 49.1%
Algara et al (president fav and net party brand model): 47.2%
Woolley adaptation of Abramowitz Time for Change model: 50.1% (this result was found on the Sabato’s Crystal Ball website)
My version of the Time for Change model (with real disposable income instead of GDP):49.8%
Simple average of all models: 48.8%.
Current two-party margin for Harris (the models all forecast the incumbent’s margin, sans 3rd party candidates): 49.2% after rounding up.
Not too shabby!
Adding the Ray Fair two-party vote model forecast using Cook Report vote totals:
Harris: 0.4947
Actual: 0.4921
https://fairmodel.econ.yale.edu/vote2020/indeane2.htm