clock menu more-arrow no yes mobile

Filed under:

Campaign polls are still variable. Votes are still predictable.

Andrew Burton/Getty Images

The heat of a presidential campaign is always a delicate time for the relationship between political scientists and political journalists. While there's been more dialogue and understanding between these two types of analysts in recent years (much of it prompted by the ease of communication between the two worlds on the internet), a basic division remains. The biggest difference in political outlook is that journalists are always more willing to say that things have changed in recent years and the old rules don't apply. Political scientists are much more likely to say that politics works similarly to how it did in the past.

Even very smart authors, who appreciate the contributions of both journalists and political scientists, can fall into this trap of debating among ourselves when the craziness of a presidential campaign splashes all over the news. In the past couple of weeks, my Vox colleagues Ezra Klein and Andrew Prokop have argued that changes in the media, technology, and campaign finance, and the rise of new and contentious issues, may mean that party insiders have lost control of presidential nominations.

The environment certainly has changed in many ways in recent years. I agree with Ezra and Andrew that it is certainly possible for things to change in politics to make the patterns of the past no longer relevant. But is that the case here? The evidence so far doesn't convince me for several reasons.

The Party Decides makes predictions about the nomination outcome

Donald Trump's strong showing in national and state polls, and Bernie Sanders's strong showing in Iowa and New Hampshire polls, might indicate that party insiders have lost control of presidential nominations. The most well-known statement that party insiders are able to select presidential nominees (or at least veto unacceptable candidates) comes in the book The Party Decides by Marty Cohen, David Karol, Hans Noel [my colleague at Georgetown and Mischiefs of Faction], and John Zaller.

I've read several political commentators asking some form of the following question: How well do outsider candidates like Trump and Sanders have to do before their success contradicts The Party Decides? In my reading of the book, the answer is simple. One of them needs to win his party's nomination. If that happens, the theory will need to be rethought. The changing political environment may have moved the power center of US political parties.

But nothing that has happened so far contradicts the thesis of The Party Decides. The main point of the theory is that support from party insiders and organized party factions, rather than early polling or early fundraising, determines who wins the nomination. It wouldn't surprise any advocate of the theory to hear that a candidate without insider support led in early polls or even in fundraising. As Seth Masket notes, many losing candidates have led in early polls for the nomination, such as Rick Perry in 2012, Hillary Clinton and Rudy Giuliani in 2008, Richard Gephardt in 2004, and Ted Kennedy in 1980. The Party Decides' statistical analysis in chapter nine finds that even polls conducted right before the Iowa caucuses are a bit less predictive of total national delegate share than party insider endorsements. (Other variables like campaign funds are even less predictive of final delegate counts.)

I don't think it would surprise most mainstream political scientists that there are temporary surges in media coverage and poll support for outsider candidates, especially several months before the primary and caucus season even begins. I remember the surge in media coverage and poll standing for Steve Forbes in the 1996 Republican campaign. In the last week of January 1996, Forbes appeared simultaneously on the covers of Time and Newsweek. But he came in a distant fourth in Iowa just two weeks later. He ended up winning only the Arizona and Delaware primaries before suspending his campaign.

Credit: The Meelypops Shop and

(The Meelypops Shop/

In John Sides and Lynn Vavreck's book on the 2012 campaign, The Gamble, one of the main themes of their chapter on the primary season was the boom-and-bust cycle of Rick Perry and Herman Cain. Both experienced temporary surges in media coverage and big rises in the polls, which later evaporated. In the end, Cain dropped out before the Iowa caucuses, while Perry had almost no success in caucuses and primaries. When were these ephemeral surges in these candidates' fortunes? Wait for it ... in August through November of the year before the election — basically right now. The surge in media coverage and polling support for Trump, while larger and longer-lasting than the examples from 2012, appears now to have peaked and begun its likely decline.

General election polls will be volatile too, especially before the conventions

One theme of political science scholarship on both presidential nomination contests and general elections is summed up in the inspired title of Andrew Gelman and Gary King's 1991 article "Why Are American Presidential Election Campaign Polls So Variable When Votes Are So Predictable?" Early polls are often highly volatile, but in the end, who is nominated and the percentage of the general election vote for each party is fairly predictable.

Sure, economic performance and other situational variables can only predict the general election popular vote with a several-percentage-point margin of error, but we certainly know which years we are likely to see a landslide and which years we won't. And in some years, we may not be able to predict which of several candidates acceptable to party insiders those insiders will settle on. But we know that candidates like Donald Trump, Carly Fiorina, and Ben Carson will not get nominated. Despite their meager standing in current polls, the Republican nominee is highly likely to be Jeb Bush or Mark Rubio, the only two remaining candidates who are broadly acceptable to party insiders. What remains is for insiders to assess who will give them the best chance in November and then coordinate their support for that candidate so that establishment resources aren't divided.

During the nomination contest, there are usually big swings in poll numbers and in which candidates get the most media coverage. This churn continues until the Iowa caucuses. Outsider candidates occasionally get a surge of success during the primaries and caucuses (e.g., Gary Hart in 1984, Pat Buchanan in 1992 and 1996, John McCain in 2000), but these things get much less likely as the process moves along. Poll numbers tend to stabilize as one candidate gets closer to locking up the nomination.

Similarly, general election polls are very volatile early after the nominees are known and get more stable as the election gets closer. The two most complete studies of trends in general election presidential polls during the election year are James Stimson's book Tides of Consent and Robert Erikson and Christopher Wlezien's book The Timeline of Presidential Elections. Both find that even though the ratings of presidential nominating conventions and their space on the network primetime schedule have declined over the decades, they are still a turning point. When both conventions are over, poll results are similar to eventual popular vote. There is movement, but it is small. This is probably because the conventions and the other campaign publicity that occurs around that time remind partisans why they usually vote for their party's candidate, leading them to "come home," and also help publicize the state of the  economy to those with weak party attachments.

Yet prior to the two nominating conventions, it is common to find wild swings in the general election polls. In spring 1980, Jimmy Carter held a wide lead over Ronald Reagan. As late as July 1988, after the Democratic Convention but before the Republicans held theirs, Michael Dukakis opened up a large lead on George Bush.

No past candidate is exactly like Trump. Yet there are a few similarities between Trump's campaign (and the media's reaction to it) and Ross Perot's 1992 third-party candidacy. Beyond some clear similarities in their messages, both are self-funded billionaire candidates who, through their use of new media tactics, caused observers to think that the old rules of presidential elections had changed. Perot first became famous as a potential presidential candidate in early 1992 by going on television talk shows. He went on a number of different shows, including Phil Donohue's, but the one he went on most often, and on which he officially announced his campaign, was Larry King Live on CNN.

Ross Perot, February 20, 1992, on Larry King Live.

CNN was still considered relatively new to politics in 1992. It had a reputation as a cheap, non-serious news network in the 1980s, sometimes labeled early on as the "Chicken Noodle Network." The 1991 Persian Gulf War increased CNN's national political profile considerably. After Perot launched his campaign primarily based on appearances with Larry King, and then led the general election polls by June 1992, only five months before the election, it was reasonable to wonder whether the advent of cable news had completely changed how presidential elections worked. In June 1992, Perot led the race with 39 percent in Gallup polls intending to vote for him, with Bush at 31 percent and Bill Clinton a distant third at 25 percent. Did cable news mean that anyone who had money and was good on TV could run for president? Would this break the lock of the two major parties on the presidency? Yet, while Perot's candidacy was out of the ordinary — winning 19 percent of the popular vote is no small thing — it didn't mark a major change in US presidential elections.

As in other years, after both major parties held their nominating conventions, the 1992 polls stabilized with Clinton in a modest lead he did not relinquish. The post-convention polls and the final result were consistent with political science fundamentals. It gets harder to win with White House the longer your party is in power. After 12 years of Republican presidents, it would have required a very strong economy for Bush to win. The 1992 economy wasn't strong enough.

If Perot had won the presidency in 1992, or if Bush had won in a landslide, it would have upended political science models. Neither happened. After the conventions, things got more predictable. Despite big technological change, presidential elections were still dominated by retrospective voting to choose between the two major parties.

This is why I'll need more evidence before I believe that presidential elections have fundamentally changed. It is certainly possible that parts of the 2016 presidential election cycle will contradict political science theories. One could imagine ways the media and campaign finance environment or several other factors might plausibly change what determines the outcomes of nomination fights and the general election. It might happen. But events so far don't contradict past patterns. Early campaign polls remain variable, yet it is very likely that votes remain predictable.

Sign up for the newsletter Sign up for Vox Recommends

Get curated picks of the best Vox journalism to read, watch, and listen to every week, from our editors.