(Jacqueline Roggenbrodt/AP)

With partisan polarization at historic heights, it might seem reasonable to assume that presidential campaigns have become highly partisan affairs. Pundits have claimed that both Democratic and Republican candidates are gearing up for an especially nasty 2016 presidential campaign. And the tense hearings over Democratic frontrunner Hillary Clinton’s role in the Benghazi tragedy – which, Clinton has alleged (and some Republican leaders have acknowledged) have the partisan aim of weakening her candidacy – also seem to point toward this conclusion.

But have presidential campaigns really become more partisan over time? Truth be told, there have been few efforts to investigate this matter in a rigorous way.

Until now. Using a new dataset comprised of every statement mentioning one or both political parties by the Democratic or Republican candidates in general election stump speeches between 1952 and 2012, we examined this question. Here’s how we did it.

First, we gathered every statement mentioning either one or both political parties — more than 8,000 in all. We then coded a random sample of statements, assessing whether each statement included a positive or negative overall tone toward the mentioned party or, alternatively, whether it advocated the transcendence of partisanship. (We also included a category called “not about the parties” to account for the possibility that some of the statements we gathered were not really about the parties.)

Thus, for example, statements in which the presidential candidate personally identified with a party or touted the party’s legislative record were coded as positive. Statements in which the candidate mocked the party or criticized its policy positions were coded as negative. Notably, our coding procedures allowed for the possibility that presidential candidates could criticize their own parties or praise the opposition. (Details on our coding procedures, with examples of coded statements, can be found here).

Because there were so many statements referring to the parties, it wasn’t feasible to do all of the coding by hand. To deal with this problem, we used a powerful algorithm to “learn” our coding rules from the statements we coded, and apply them to the uncoded statements. We did extensive checking to make sure that the algorithm accurately replicated our coding by hand.

Finally, for each presidential candidate, we reconstructed our theoretically important categories from the coded data.

Partisan rhetoric was comprised of statements in which the candidate made positive references about (members of) his own party and also about assertions in which he denigrated (members of) the opposition party. Cross-partisan rhetoric refers both to negative statements about (members of) the candidate’s party and positive declarations about (members of) the opposition party. And bipartisan rhetoric encompasses statements in which both parties were mentioned and in which the candidate criticized the partisanship of both or expressed the desire to overcome party differences.

In our paper, we focused on examining partisan rhetoric. Note that, under our coding scheme, statements that are not partisan are by definition either cross-partisan or bipartisan, and thus imply a much more conciliatory tone. However, only a very small proportion of statements turned out to not be about the parties.

Remarkably, when we looked at the data, we found that presidential campaigns have actually become much less partisan over time.

This is largely due to a dramatic decline in Democratic candidate partisanship. Indeed, Republican presidential candidates have generally refrained from explicit partisan rhetoric over the period we were analyzing. The sharp drop in Democratic partisanship has resulted in a convergence in rhetoric across parties in which explicit partisan appeals now make an infrequent appearance.

Note: Data for the chart are drawn from presidential candidate stump speeches, archived in the Annenberg/Pew Archive of Presidential Election Discourse and the American Presidency Project (americanpresidency.org.)

What’s going on? Have candidates actually embraced the politics of “post-partisanship” so famously advocated by President Obama during the 2008 campaign?

Almost certainly not. Instead, these trends likely reflect candidates’ strategic adjustment to a changing political environment in which explicit partisan appeals have more costs than benefits.

While the rise in partisan polarization over the past four decades has excited extremists in both parties, it remains off-putting to many moderate voters. This puts presidential candidates in a bind.

Realizing the electoral imperative to court moderates, presidential candidates have adjusted their rhetoric to deemphasize explicit partisan appeals in their public campaign statements. Candidates have increasingly presented themselves as compromisers eager to work across the partisan aisle.

This doesn’t mean that presidential candidates have stopped being partisan. They just do it in more subtle ways. After all, presidents and candidates can use – and increasingly have used – less public methods to strengthen their parties, such as fundraising, organization-building and candidate recruitment.

Furthermore, candidates’ embrace of rhetorical bipartisanship hasn’t eliminated partisan appeals from presidential campaigns. Instead, presidential candidates simply handed off responsibility for partisan messaging to party networks, left- and right-wing media, and social movements (such as the civil rights movement on the left and the religious conservative movement on the right).

The story is somewhat different for Democratic and Republican candidates, of course. For Democrats, the 1950s and especially 1960s represented the modern heyday of the party’s brand (as well as the height of its influence in national politics), making partisan appeals a useful campaign strategy.

Further, the parties weren’t very polarized, so invoking the party label was unlikely to evoke a negative response among moderate voters. Thus, Democrats made extensive use of partisan appeals during this period.

In our research, Adlai Stevenson, the Democratic presidential nominee in 1952 and 1956, stands out as an especially partisan candidate. It’s likely that Stevenson’s sharp partisanship in these campaigns reflected an effort to cope with the remarkable personal popularity of his opponent, the general (and World War II hero) Dwight Eisenhower.

While Stevenson couldn’t possibly match Ike’s personal appeal, he could draw on the broad popularity of the Democratic Party – and the relative unpopularity of the GOP – to try to make up the difference.

After Stevenson’s campaigns, however, things gradually began to change.

The Democratic brand was battered by the racial upheavals of the 1960s, the stagflation of the 1970s, and the perception that the party favored government programs over economic growth. Even as Democratic influence in national politics gradually declined, the tremendous growth of partisan polarization made explicit partisan rhetoric a turn-off for moderate voters.

All this has increasingly encouraged Democrats to deescalate partisan appeals in favor of more conciliatory rhetoric.

In contrast, for Republican presidential candidates, explicit partisan rhetoric has never been a compelling campaign strategy. During the 1950s and 1960s, the Republican Party brand was not especially strong, discouraging GOP presidential candidates from featuring partisan appeals.

Since then, intensifying partisan polarization has reinforced Republicans’ disinclination to present themselves as partisan warriors on the campaign trail, even though the party’s prospects have increased substantially.

In our paper we give statistical evidence for our claims, showing that measures of partisan polarization are the strongest predictors of partisan rhetoric by Democratic presidential candidates.

Our analyses show that partisan polarization is closely – and negatively – associated with Democratic candidate partisanship, even when we control for a wide variety of other factors, including the competitiveness of the campaign, incumbency, party control of Congress, and the strength of partisan identification in the electorate.

The results give important evidence for our argument by showing that Democratic candidates have sharply reduced their reliance on partisan rhetoric as partisan polarization has increased.

In the end, our findings suggest that it is unlikely that the 2016 general presidential election campaign will be especially partisan, at least when it comes to the explicit rhetoric of the presidential candidates themselves. Once again the candidates will preach the gospel of conciliation, leaving the partisan dirty work to affiliated media and activist groups.

This will give the 2016 Democratic and Republican nominees the best of both worlds: conciliatory messages for moderate, less involved voters and partisan appeals for the true believers.

Jesse Rhodes is an associate professor and Zachary Albert is a graduate student in the Department of Political Science at the University of Massachusetts Amherst.