Did 15 Years of War Win the Presidency for Donald Trump?

Did the promise of peace deliver the nuclear button to Donald Trump? A recent study by Douglas L. Kriner, a political scientist at Boston University, and Francis X. Shen, a law professor at the University of Minnesota, argues that Trump's victory last November can be directly attributed to "the casualty divide"—their term for the division "between communities whose young people are dying to defend the country, and those communities whose young people are not."

Specifically, they think the evidence shows

a significant and meaningful relationship between a community's rate of military sacrifice and its support for Trump. Our statistical model suggests that if three states key to Trump's victory—Pennsylvania, Michigan, and Wisconsin—had suffered even a modestly lower casualty rate, all three could have flipped from red to blue and sent Hillary Clinton to the White House.

Exposure to the costs of war is not equal across America, as Kriner and Shen note: "seven states have suffered casualty rates of thirty or more deaths per million residents. By contrast, four states have suffered casualty rates of fifteen or fewer deaths per million. As a result, Americans living in these states have had different exposure to the war's human costs through the experiences of their friends and neighbors and local media coverage."

The varying effects get more pronounced on the county level: "more than a quarter of counties had experienced a casualty rate more than 3.5 times greater [than the national average], and 10% of counties had suffered casualty rates of more than 7 deaths per 100,000 residents. Voters in such communities increasingly abandoned Republican candidates in a series of elections in the 2000s."

The authors believe they have found a robust association between a state's casualty rates and Trump's excess votes over Mitt Romney's four years earlier. Had Pennsylvania, Michigan, and Wisconsin all had casualties rates equal to the lower ones in nearby New York, they conclude, then Trump's narrow victory margin in those states would have disappeared and Hillary Clinton would be president. Doing the same analysis on the more granular county level leads to even stronger-seeming support for their thesis. (They see a similar historical pattern going back even further: "in the Civil War, Korea, Vietnam, and Iraq, constituencies that have suffered the highest casualty rates have proven most likely to punish the ruling party at the polls.")

Kriner and Shen controlled for some obvious possible confounding variables, including income and educational level, ethnic mix, rural/urban population balance, and percentage of veterans. "Even after including all of these demographic control variables," they write, "the relationship between a county's casualty rate and Trump's electoral performance remains positive and statistically significant."

Before you rush to assume that correlation indicates causation, there's a mystery you have to consider: What exactly did voters know (or think they knew) about Trump's foreign policy, and how important was that to them? In Michigan, for example, only 13 percent of voters in a CNN exit poll rated foreign policy as their most important topic, and only 34 percent of those privileging foreign policy went for Trump. In Pennsylvania those numbers were a very similar 12 percent and 31 percent; in Wisconsin, they were 11 and 38. It is possible, perhaps even likely, that military casualty families were overrepresented in those small percentages who both privileged foreign policy and voted for Trump, but the authors don't know this for sure.

As for what voters believed about his foreign policy: Many of a non-interventionist bent were excited about those portions of Donald Trump's campaign rhetoric that suggested he had strong doubts about the wisdom of some past U.S. wars, such as the one in Iraq, and might be less inclined to get us into new ones. Writing here at Reason, Matt Welch, Sheldon Richman, and I had doubts about Trump's peacenik bonafides. In any case, voter ignorance of a politician's stated stances is rampant, and it's an open question what any given voter thought Trump would do in office. Kriner and Shen admit that Trump's rhetorical record on peace has been decidedly mixed; in their words, he "promised a foreign policy that would be both simultaneously more muscular and more restrained."

Exactly why Trump won over Clinton may be highly overdetermined. In other words, attributing it directly to any one cause may be saying too much. See, for example, very similar correlative analysis from Brian Flaxman, an economics Ph.D. student at the University of Colorado Boulder, who looks at likely former Obama votes that went for Trump and concluded that Obama/Trump support seems highly correlated to worse poverty and unemployment rates. So it's difficult for a noninterventionist to enthusiastically embrace the Kriner/Shen study's conclusion that "If Trump wants to maintain his connection to this part of his base, his foreign policy would do well to be highly sensitive to American combat casualties," as sweet as it is to believe that.

On a deeper level, an interesting study forthcoming in the American Political Science Review casts some cynical shadows for all attempts to connect an election's outcome with specifics about what a candidate says about his policies, at least to the extent that those policy communication attempts come via the standard efforts of the political campaign itself.

The paper was written by Joshua L. Kalla, a grad student in political science at Berkeley, and David E. Broockman, who teaches political economy at Stanford Graduate School of Business. They conclude that "the best estimate of the effects of campaign contact and advertising on Americans' candidates choices in general elections is zero."

What exactly do they mean by this, which seems to imply that all paid attempts at messaging about a politicians' stances to potential voters is a big waste of time? The question they strive to answer is: "How susceptible are American voters' choices in general elections to influence from political elites in the form of campaign contact and advertising?" (This question has some relevance to how much panic, moral or otherwise, Americans should have about Russian attempts to influence the election via Facebook ads—though they consider their conclusions less certain when it comes to the online world.)

After analyzing the results from 49 field experiments, Broockman and Kalla conclude:

• The best estimate for the persuasive effects of campaign contact and advertising—such as mail, phone calls, and canvassing—on Americans' candidate choices in general elections is zero. Our best guess for online and television advertising is also zero, but there is less evidence on these modes.

• When campaigns contact voters long before election day and measure effects immediately, campaigns often appear to persuade voters. However, this early persuasion decays before election day and the very same treatments usually cease working close to election day. This suggests political scientists and practitioners should consider whether an experiment was run close to an election when attempting to generalize its findings.

• Campaigns can sometimes identify pockets of persuadable voters, but even this only appears possible in some elections and when campaigns conduct within-cycle field experiments to identify responsive subgroups.

The authors also conducted their own studies in 2015 and 2016, "in partnership with a national door-to-door canvassing operation." Here again they tended to find no impact at all on how people voted. (One experiment did show a small but statistically significant possibility that canvassing in Ohio helped Trump, though they "urge caution when interpreting these results" because of a greater chance of bias under the specific circumstances where that effect was found.)

"Our findings also offer an important caveat to the widespread notion that political elites can easily manipulate citizens' political choices," the authors conclude. "The circumstances in which citizens' political choices appear manipulable appear to be exceedingly rare in the elections that matter most."

"Our evidence is silent on several questions," they note. "It does not speak to the effects of candidates' qualities, positions, or overall campaign 'message.'" But at the very least, this paper sparks a strong suspicion that the professional geniuses who sell themselves as experts in campaign messaging and methods are selling a phantasm to protect their phoney-baloney jobs.

If traditional campaigning doesn't sway voters' choices, what does? In an interview with The Atlantic, Kalla said: "The first order of understanding an election and how people vote is partisan identity. Most people vote based on whether there's a D or an R next to their name."

While the Kalla/Broockman study makes it seem that convincing voters what a candidate thinks or believes via the standard methods of professional campaigns is worthless, their stress on the importance of mere partisan identification raises important considerations relevant to the "Trump won because of foreign policy" study.

That study was mostly trying to figure out why those who had voted for one party in 2012 voted for another in 2016. That seems, in the light of this Kalla/Broockman study, to be the most important place to examine the whys of Trump's victory.

But the "Trump won because of military casualties" argument lacks detailed evidence on what is apparently the most important question when it comes to explaining voter choice: Where and how do voter form their opinions about which party is "theirs"? Which elements of that are vital, and which are ancillary? This seems especially significant as Trump shifts the GOP from the Romney-era consensus in important ways, which may or may not as time goes by include foreign policy.

In any case, whatever voters may think about Trump's views on foreign policy or any other subject, there is strong reason to believe it had nothing to do with paid deliberate efforts on the part of the Trump campaign.