Hostname: page-component-745bb68f8f-l4dxg Total loading time: 0 Render date: 2025-02-05T23:42:58.508Z Has data issue: false hasContentIssue false
  • English
  • Français

Forecasting Presidential Nominations or, My Model Worked Just Fine, Thank You

Published online by Cambridge University Press:  15 April 2003

William G. Mayer
William G. Mayer
Affiliation:
Northeastern University
Rights & Permissions [Opens in a new window]

Abstract

An abstract is not available for this content. As you have access to this content, full HTML content is provided on this page. A PDF of this content is also available in through the ‘Save PDF’ action button.
Type
SYMPOSIUM
Information
PS: Political Science & Politics , Volume 36 , Issue 2 , April 2003 , pp. 153 - 157
Copyright
© 2003 by the American Political Science Association

For most Americans, the 2000 elections will be remembered as the Year of the Great Recount in Florida. Contemporary political science being what it is, political scientists are more likely to remember 2000 as the Year when the Forecasting Models Went Crash. Of the seven major presidential forecasting models presented at the 2000 American Political Science Association Annual Meeting, all seven forecast a comfortable Gore victory, with a median prediction that Gore would win 55% of the two-party popular vote.

Yet, not every forecasting model went bust in 2000. Seven years ago, I published an article presenting what was then the only model ever developed for forecasting presidential nomination races (Mayer 1996).1

Several other models have since been put forward. See, in particular, Adkins and Dowdle 2000; and Steger 2000.

Three contested nomination races have occurred since that time, and my model has worked well in every instance. The purpose of this article is two-fold: to examine the model's performance in 1996 and 2000; and to discuss what the model reveals about the nature of contemporary presidential nominations.

The Basic Model

The original model was based on data from seven contested nomination races that took place between 1980 and 1992. One of these, the Republican race in 1992, was never seriously in doubt; no one reasonably expected Pat Buchanan to defeat incumbent President George Bush. But the other six races each featured a serious battle for a major party's presidential nomination, with considerable uncertainty (and a lot of inaccurate predictions) as to how it would turn out. Were there any indicators that might have enabled us to predict the winning candidates before the actual voting and delegate selection commenced? As it turns out, there were two such indicators, both available in January or early February of the election year, which pointed out the winner in five of the six contested races (six of seven if one includes the Bush-Buchanan battle).

The first of these indicators was the candidates' relative standing in polls of the national party electorate. For at least a year before the first caucus or primary takes place, polling organizations routinely ask national samples of Democrats and Republicans whom they would prefer as their party's next presidential candidate. (In everything that follows, I use data drawn from the Gallup polls, partly because the Gallup organization has a very good track record for doing careful, unbiased survey work, and partly because Gallup is the only organization I know of that has asked questions on this topic continuously over the last two decades.) My interest here is on the last such poll before the start of delegate selection activities—meaning, in most years, the last poll before the Iowa caucuses. As shown in the top half of Table 1, the candidate who was leading in this poll went on to win the nomination in five of the six cases.

The second indicator is the candidates' relative success in raising campaign money. Under the federal campaign finance laws that have been in effect since 1974, all active candidates for the presidential nomination are required to submit periodic reports detailing how much money they have raised and spent. Here, my focus is on the total amount of money that each candidate had raised (technically, their net receipts) by December 31 of the year before the election. And again, in five of the six cases, the leading money raiser went on to win the nomination (see Table 1).

To allow for easier comparison with general election forecasting efforts, I then combined these two indicators into a regression equation in order to generate a numerical prediction of each candidate's success in the actual primaries. The dependent variable in this equation is the percentage of the total vote won by each candidate in all presidential primaries held by that candidate's party during the entire nomination season. In the 1992 Democratic nomination contest, for example, 39 different primaries were held at which a total of 20,239,385 presidential preference votes were cast. Bill Clinton received 10,482,411 of these votes, or 51.8%; Jerry Brown had 4,071,232 votes, or 20.1%; Paul Tsongas received 18.1%; and so on.

Two independent variables were used to predict these primary vote shares. The first is the percentage of party identifiers who supported each candidate in the last national Gallup poll before the Iowa caucuses. The second is the total amount of money each candidate raised before the election year, divided by the largest amount of money raised by any candidate in that party's nomination race.2

For example, in the 1992 Democratic nomination contest, the largest fund-raiser during 1991 was Bill Clinton, who raised $3,304,000; Clinton thus received a score of 100 on this variable. Paul Tsongas, who raised $2,630,000, received a score of 79.6. Jerry Brown, with total receipts of $1,034,000, was assigned a value of 31.3.

The results, based on data from 1980 through 1992, are shown in the first column of figures in Table 2. (For a fuller discussion, see Mayer 1992, 51–53.) As I noted in my original article, a number of factors would seem to make presidential nomination races inherently more difficult to predict than general elections: the larger number of serious candidates in the contest; the sequential nature of the process; and the absence of such stabilizing forces as party identification.3

As Adkins and Dowdle (2000) have shown, however, the accuracy of these forecasts can be considerably improved simply by adding in the results of the New Hampshire primary.

That said, the model does a quite commendable job of accounting for the variation in how candidates fared in the presidential primaries. And most importantly, the model correctly forecasts the winning candidate in five of the six races.

Forecasting the 1996 and 2000 Races

So much for the past. As an old Chinese proverb notes, “To prophesy is extremely difficult—especially about the future.” The acid test for models of this sort is how well they fare after publication.

Table 3 provides the relevant data for 1996: the results of the last national poll of Republican party identifiers before the Iowa caucuses; and how much money each Republican candidate had raised as of December 31, 1995. Both predictors (and the equation-based forecasts) point unambiguously to Robert Dole; and after a number of stumbles, and in spite of a rather lackluster campaign, Dole did indeed win the 1996 Republican presidential nomination.4

Though the publication date for my article is listed as 1996, it is important to note that the book in which it was published actually appeared in late 1995. In addition, an all-but-identical version of the book chapter was presented as a paper at the 1994 meeting of the Northeastern Political Science Association. Thus, the results shown here for the 1996 Republican nomination race are a genuine prediction, not an after-the-fact rationalization.

Table 4 provides comparable information on the 2000 nomination races. On the Democratic side, Al Gore led Bill Bradley in both the polls and total in fundraising (though rather narrowly in the latter case). And, of course, Gore also eventually won the nomination. George W. Bush had an even more daunting lead over the rest of the Republican field. Though the Texas governor made the race closer than it might otherwise have been by running a remarkably ill-conceived campaign in New Hampshire, it was ultimately he, and not John McCain, who became the Republican standard-bearer.

What It Means

What, in general, can we learn from election forecasting models? In many cases, I would argue, the answer is: not much. Far too often, forecasting work seems to be low on theory and high on crude empiricism. Variables are added or modified just because they lead to higher R2 values or lower standard errors of estimate—exactly the sorts of things we all tell our introductory methods classes not to do. (For a striking exception—and a wonderful example of what can be done with these models, see Campbell 2000.)

In contrast, I developed my model with the primary purpose of making a series of arguments about the basic dynamics of the contemporary nomination process. And it is these arguments, I think, that have held up particularly well through the 1996 and 2000 election cycles.

1. The contemporary presidential nomination process is unusually favorable to frontrunners

If this point now seems widely accepted, it was most certainly not that way when I first formulated the model. To the contrary, for at least two decades after the rules were re-written in the early 1970s, the conventional wisdom was precisely the opposite: that the new system offered great advantages to outsiders and insurgents. As Robert Scheer of the Los Angeles Times put it, “There's a special problem with the drawn out system of primaries and caucuses. What it allows is for an unknown to get in” (Foley et al. 1980, 44).

But it wasn't just those dreaded journalists and pundits who reached this conclusion. The same verdict was pronounced in some of the very best academic writing on the subject. One of the bestknown models of the primary election process is that of Brady and Johnston (1987). As they conclude in the very last sentence of their article: “The lesson, then, of this analysis is that being the favorite is a mixed blessing, and one might better be a newcomer with media appeal and a little luck in Iowa or New Hampshire” (184). Larry Bartels (1988) ended his award-winning book on presidential primaries on a similar note. Characterizing the current system as one with a “remarkable open-ness to new candidates,” he added, “In many political systems positions of party leadership are earned through decades of toil in the party organization. In contemporary American politics the same positions are sometimes seized, almost literally overnight, by candidates with negligible party credentials and very short histories as national public figures” (287). James Ceaser (1982), who approached these issues from a more historical and theoretical perspective, pronounced a similar verdict: “When the effects of sequence in the primaries and the influence of the media are taken into consideration, the nominating campaign often becomes not simply a test among established national contenders, but an occasion for outsiders to make their reputation during the campaign itself. In this respect, the current system is more open than the pure convention system and the mixed nominating system” (95).

From the late 1970s through the early 1990s, academic writing about the presidential nomination process was saturated with this perspective. (In addition to the sources just quoted, see Sundquist 1980, 193–94; Mann 1985, 35–36; Keeter and Zukin 1983, 190–93; and King 1981.) Even in the mid- and late 1990s, many scholars continued to argue, as Mackenzie (1996) did in his book on political reform, that “[The contemporary presidential nomination] process has a strong tendency to promote the candidacies of outsiders who have little or no Washington experience and who are often strangers to the leading members of their own party” (50).

The most important point of my excursion into election forecasting was to assert the opposite: that in almost every recent nomination race, the candidate who ultimately won was the person leading before any of the delegates were selected. Indeed, in seven of the last 10 contested nomination races, the eventual nominee was leading in the national polls for at least a year before the Iowa caucuses. To say the least, this is not a system characterized by its openness to new faces.

2. As a factor in the presidential nomination process, momentum is greatly overrated

If the pre-race frontrunner usually wins, then one must also conclude that momentum is nowhere near the overwhelming force that it is frequently portrayed to be. Momentum can be compared to a roller-coaster ride: it provides a lot of thrills and excitement, but in the end it leaves us exactly where we started. And so it is in presidential politics: After all the effects of momentum have come and gone, the person who started out ahead almost always finishes ahead. Put another way, if you're interested in figuring out why John McCain went from 15% to 34% in the national polls within five days of his New Hampshire victory, momentum clearly provides the best explanation. But if your main interest is in who finally wins the nomination, momentum is really of very little help at all. Not since Jimmy Carter's campaign in 1976 has a momentum-driven candidacy been successful.

3. Frontrunners almost invariably stumble at some point in the process—but these mistakes, for all the media attention they receive, are rarely fatal

If front-runners usually triumph in the end, their road to the nomination is never an entirely smooth one. Bad stuff happens. Bob Dole lost the New Hampshire, Delaware, and Arizona primaries; George Bush lost in New Hampshire and Michigan; and Al Gore spent the better part of 1999 squandering his lead in both the polls and in fundraising (on Gore's early missteps, see Mayer 2001, 23–26). But all of these candidates also had real strengths, that were not entirely overwhelmed by a few weeks of bad campaign-related publicity. Each recovered his footing—rather quickly, in fact—and ultimately racked up a long string of victories in the presidential primaries.

4. The longer we live under the current system, the more we learn about it—and this, too, works to the advantage of frontrunners

There are a number of reasons why Jimmy Carter won the 1976 Democratic nomination against a series of better-known and better-financed opponents, but one of the most important is simply that the nomination process was then new and not very well understood. Carter and his advisors figured out what it took to win in the new environment; many of his chief opponents (in particular, Henry Jackson) were less astute. But the longer the current system endures, the more its basic quirks and tendencies have become a matter of common knowledge. The net effect of all this learning and experience has been to create a fair amount of uniformity in basic campaign strategy—and thus to neutralize the advantage that any one candidate can derive from “playing the game” better than his or her opponents. And if campaign strategy doesn't matter, presidential nomination races will be decided more and more on the basis of such fundamental factors as popularity and money—resources that frontrunners, almost by definition, will have in greater supply than their competitors.

A good example of how campaign learning works against outsiders and insurgents is provided by the contrast in how two early front-runners—Walter Mondale and Robert Dole—dealt with opponents who tried to ride a “better than expected” showing in Iowa to a breakthrough win in New Hampshire. When Mondale scored a resounding victory in the 1984 Iowa caucuses, he and his campaign strategists fully expected that win to propel them to another win eight days later in New Hampshire, where Mondale was already well ahead in the polls. In fact, the candidate with the momentum turned out to be Gary Hart, who had finished 30 percentage points behind Mondale in Iowa but had nevertheless managed to convince the press that he was now Mondale's chief rival. What is noteworthy for our purposes is that the Mondale campaign never really saw it coming until the very end (see Germond and Witcover 1985, 161–68). In particular, they made no sustained attempt to attack Hart until the day after New Hampshire, by which time Hart's impressive win in the Granite State almost completely transformed the character of the race. In the end, it took the Mondale campaign four more months of hard fighting before they finally, narrowly defeated Hart.

In 1996, Dole faced a quite similar predicament. After winning in Iowa (though by a considerably less decisive margin than Mondale), Dole, too, found that victory did little or nothing to help his standing in New Hampshire. Instead, according to all the polls, the momentum went to Pat Buchanan, who finished second in Iowa, and, even more, to Lamar Alexander, who had finished third. In one set of tracking polls, Alexander's support in New Hampshire went from 5% on the day before Iowa to 18% four days after the primary. But here the parallel breaks down, for the Dole campaign knew quite well what had happened to Mondale and, recognizing the peril they faced, did not wait for their opponents' momentum to snowball. The Dole strategists felt that Buchanan was not a long-term threat to win the nomination—but Alexander was. Hence, on the final weekend before the New Hampshire vote, they launched an all-out attack on the former Tennessee governor. And it worked: Alexander's momentum stalled, he finished third in New Hampshire, and he never again was a serious factor in the 1996 Republican nomination race. In the 13 primaries held between New Hampshire and Alexander's withdrawal, he averaged just 9% of the vote (for further details, see Mayer 1997). Instead of having to fight a prolonged battle against a fresh face with a tidal wave of momentum, the Dole campaign stopped Alexander before he ever gained traction.

5. Money helps, but it's definitely not the whole ball game

Though a great deal of attention is always focused on candidate fundraising, the clear thrust of this model is that, unless a candidate also has a strong base of support among ordinary, rank-and-file party voters, a huge warchest will not get a candidate very far. Indeed, according to all three equations in Table 2, the effect of money on primary voting, holding initial poll standings constant, is essentially zero. The same moral emerges, on a more anecdotal level, from a listing of all the candidates who were very successful fundraisers in the year before the election, but then fared dismally once the actual primary voting began: John Connally in 1980, John Glenn in 1984, Pat Robertson in 1988, Phil Gramm in 1996, Steve Forbes in both 1996 and 2000, and even, to some extent, Bill Bradley.

In a number of ways, these statistics probably understate the true role of money in the presidential nomination process. On the one hand, the model does not include (it was not designed to) the role that money might have played in helping a candidate become the frontrunner or remain in that position. On the other side, there is obviously a strong correlation between these two variables, such that frontrunners are generally guaranteed to be at least reasonably successful fundraisers. As noted in the original article, an interesting test case would be to see how a candidate fared who had a large lead in the polls but couldn't raise a decent warchest. But in the races examined here, there is no such candidate.

And on to 2004

The final column in Table 2 provides the most recent version of my forecasting model, reestimated to include all the data from 1980 through 2000. In the meantime, I am pursuing research that pushes the model both forward and backward in time: forward to see how much of a race's final outcome is determined by the results in Iowa and New Hampshire, backwards to examine the dynamics of candidate support during the invisible primary.

References

Adkins Randall E. Dowdle Andrew J. 2000 Break Out the Mint Juleps? Is New Hampshire the ‘Primary’ Culprit Limiting Presidential Nomination Forecasts? American Politics Quarterly 28 251 269 Google Scholar
Bartels Larry 1988 Presidential Primaries and the Dynamics of Public Choice Princeton, NJ Princeton University Press Google Scholar
Brady Henry E. Johnston Richard 1987 What's the Primary Message: Horse Race or Issue Journalism? Media and Momentum: The New Hampshire Primary and Nomination Politics Orren Gary R. Polsby Nelson W. Chatham, NJ Chatham House Google Scholar
Campbell James E. 2000 The American Campaign: U.S. Presidential Campaigns and the National Vote College Station Texas A&M University Press Google Scholar
Ceaser James W. 1982 Reforming the Reforms: A Critical Analysis of the Presidential Selection Process Cambridge, MA Ballinger Google Scholar
Foley John Britton Dennis A. Everett Eugene B. Jr 1980 Nominating a President: The Process and the Press New York Praeger Google Scholar
Germond Jack W. Witcover Jules 1985 Wake Us When It's Over New York Macmillan Google Scholar
Hagen Michael G. Mayer William G. 2000 The Modern Politics of Presidential Selection: How Changing the Rules Really Did Change the Game In Pursuit of the White House 2000: How We Select Our Presidential Nominees Mayer William G. New York Chatham House Google Scholar
Keeter Scott Zukin Cliff 1983 Uninformed Choice: The Failure of the New Presidential Nominating System New York Praeger Google Scholar
King Anthony 1981 How Not to Select Presidential Candidates: A View from Europe The American Elections of 1980 Austin Ranney Washington, DC American Enterprise Institute Google Scholar
Mackenzie G. Calvin 1996 The Irony of Reform: Roots of American Political Disenchantment Boulder, CO Westview Google Scholar
Mann Thomas E. 1985 Should the Presidential Nominating System Be Changed (Again)? Before Nomination: Our Primary Problems Grassmuck George Washington, DC American Enterprise Institute Google Scholar
Mayer William G. 1996 Forecasting Presidential Nominations In Pursuit of the White House: How We Select Our Presidential Nominees Mayer William G. Chatham, NJ Chatham House Google Scholar
Mayer William G. 1997 The Presidential Nominations The Election of 1996: Reports and Interpretations Pomper Gerald M. Chatham, NJ Chatham House Google Scholar
Mayer William G. 2001 The Presidential Nominations The Election of 2000: Reports and Interpretations Pomper Gerald M. New York Chatham House Google Scholar
Steger Wayne P. 2000 Do Primary Voters Draw from a Stacked Deck? Presidential Nominations in an Era of Candidate-Centered Campaigns Presidential Studies Quarterly 30 (December) 727 753 Google Scholar
Sundquist James L. 1980 The Crisis of Competence in Our National Government Political Science Quarterly 95 183 208 Google Scholar
Figure 0

National Poll Standings and Pre-Election Fund-Raising Results as Predictors of Presidential Nominations

Figure 1

Regression Equations for Predicting Primary Vote Shares

Figure 2

Presidential Nomination Predictors in the 1996 Republican Nomination Race

Figure 3

Presidential Nomination Predictors in the 2000 Nomination Races