Research chat: Political scientist Michael Traugott on primary polls


University of Michigan political scientist Michael Traugott is an expert on public opinion, media dynamics and U.S. elections. After problems and controversies during the the last presidential primary, in 2008 — crystallized when Hillary Clinton defied all 13 polls prior to the New Hampshire primary and beat Barack Obama — Traugott participated in an inquiry commissioned by the American Association of Public Opinion Research to analyze what had happened. The scholarship that came out of that work provides some the most systematic analysis to date on presidential primary polling.

As part of our “research chat” series, Journalist’s Resource, which is compiling academic studies on polling issues, spoke with Traugott, a former Shorenstein Center fellow, and asked for his thoughts on presidential elections past and future.


Journalist’s Resource: Your study “Dynamics of Poll Performance During the 2008 Presidential Nomination Contest” seems highly relevant to the presidential nominating contest now, in 2012. You demonstrate that there were polling problems throughout the last primary period, and this is related to the difficulty of capturing momentum and “late surges.” What are some things reporters should consider as we again enter the nominating contest season?

Michael Traugott: On a kind of a grand scale, you can’t really interpret polls unless you know how they were conducted. The central issue is that pollsters should be providing a full set of information about sampling, question wording, weighting, and in particular in the primaries, about the likely-voter models that they use. And in the early primaries particularly, there will be a lot of polling done, and there will be a pretty high level of variance in the estimates. But usually you can explain that variance by the specific attributes of the polls. So it’s actually advantageous to journalists to be able to assemble and then compare  the details of each in order to make some sense of what the estimates are.

JR: What are the ways that you’d like to see political journalism improve in this election cycle?

Michael Traugott: One element that is confusing and that sometimes journalists don’t handle very well is the difference between the national polls and the state-level polling as each of the events comes along on the calendar. The nomination process is not a national-level contest, and national-level preferences or estimates are very misleading because the winner is determined by a combination of the local rules for delegate selection and a very substantial “order effect” — the sequence and the rules of the game along the way make a lot of difference. Many reporters who are not experienced in dealing with polls will write as though they are surprised by the outcome in a state, compared to the estimates being produced by the national polls, when it seems to be pretty understandable if you know something about the internals of the specific state-level polls compared to the national polls.

The striking difference, of course, is that in many states, the primaries are closed, and for 2012 you have to be a declared Republican in many states to participate. But the national level polls will be incorporating independents, and they’ll often be running trial heats against Barack Obama, so they’ll include Democrats as well. And party identification is a very strong predictor of candidate support and preference. So the relative mix of Democrats, Republicans and independents is usually a pretty powerful explanation for the differences that appear in the national versus state polls.

JR: The political science world often calls this pre-primary stage the “invisible primary,” and there’s a lot of research on how contests are decided before voters or caucus goers ever participate. What’s your perspective?

Michael Traugott: If we talk about this in the context of 2012, we have an incumbent Democrat who is unchallenged in his party, and therefore we have a relatively large field among the Republicans, who have different levels of name recognition and visibility in the electorate. So there are all of these events that are taking place, with an eye toward collectively establishing credibility for the Republican party, and presenting a possibility or potential for particular candidates to try to separate themselves from the field. The number of debates that the Republicans are holding is extreme. But there can’t be many concrete differences between them; the differences between Democrats and Republicans are much greater than the differences among the GOP candidates alone.

It’s also true that the primary electorates tend to be more hard-core and ideologically extreme than the general election set of voters. So one consequence of the pre-nomination events and the invisible primary is that, in order to appeal to these likely participants in the primary and caucuses, the candidates are forced to take more extreme positions than they might otherwise choose. And the consequences of both of these things — the large number of debates and the need on the Republican side to appeal to extremely conservative voters — is that they are just creating content, fodder, for the Democratic campaign, Obama’s campaign, in the fall.

Tags: polling, presidency, Iowa, New Hampshire, research chat

Last updated: November 29, 2011


We welcome feedback. Please contact us here.