The Basic Dynamics of the Presidential Nomination Process: Putting the 2000 Races in Perspective

Article excerpt

How did George W. Bush and Al Gore come to win the 2000 presidential nominations of America's two major political parties? Of all the dozens or even hundreds of persons who might plausibly have aspired to serve as the nation's forty-third president, how was it that the choice was effectively narrowed down to these two individuals? The thesis of this article is that at least during the nomination phase of the campaign, the 2000 election was pretty much business as usual. Over the past two decades, the presidential nomination process, once widely thought of as an almost uniquely turbulent and unpredictable enterprise, has in fact usually operated in a quite regular and predictable fashion. And while every election has its new wrinkles and novel situations, most of what transpired during the 2000 nomination contests has clear and obvious parallels in the races that preceded them.

The analysis in this article thus proceeds along two tracks. In the first place, I attempt to describe and document some of the basic dynamics of the contemporary nomination process. In particular, I focus on the issue of how it is that the process finally settles on a candidate: how and why it is that one candidate rather than another finally wins the nomination. With that as background, I then examine the particular circumstances of 2000, showing how much of that election was simply the latest remake of a quite familiar plotline but also highlighting those aspects of the 2000 campaign that genuinely were exceptional.

An Initial Model

The starting point for this analysis is a nomination forecasting model that I first developed almost ten years ago. (1) Though presidential nomination races often seemed to be chaotic and unpredictable, there were, I argued, two indicators, both available before any of the primaries and caucuses had taken place, that correctly predicted which candidate would win his party's presidential nomination in six of the past seven contested nomination races. (2)

The first of these indicators was the candidates' relative standing in polls of the national party electorate. Well before any of the national convention delegates are actually selected-in fact, quite soon after the conclusion of the previous presidential election-pollsters regularly ask national samples of Democrats and Republicans whom they would like to see nominated as their parties' candidate for president. As shown in Table 1, if one focuses on the last poll taken before the start of delegate selection activities--meaning, in most years, the last poll before the Iowa caucuses--the candidate leading in that poll went on to win the nomination in six of the seven contested nomination races held between 1980 and 1992.

The second indicator was the candidates' relative success in raising money. Under the campaign finance laws enacted in 1974, every active presidential candidate is required to make periodic reports to the Federal Election Commission indicating how much money he or she has raised and spent for the campaign. As is also shown in Table 1, the leading money raiser during the preprimary campaign--more precisely, the candidate who had raised the largest amount of money by December 31 of the year before the election--went on to win the nomination in six of seven cases.

Since that article was published, the country has seen three contested presidential nomination races, and both indicators worked perfectly in every case. The relevant data for the 2000 election are shown in Table 2. On the Democratic side, Al Gore had a large lead over Bill Bradley in the polls; Gore was also his party's leading preelection year fund-raiser, though by a surprisingly narrow margin.

As for George Bush, to say that he was the Republican front-runner in 2000 is a distinct understatement. In fact, Bush had perhaps the most successful invisible primary season in the modern history of presidential elections. The most publicized measure of his success was his fund-raising. During 1999, the Bush campaign took in more than $67,000,000 in contributions. As shown in Table 3, even after adjusting for inflation, this is more than double the amount raised by any other presidential candidate since the Federal Election Commission began keeping records. Equally impressive, however, was Bush's performance in the public opinion polls. Through the final six months of 1999, Bush was supported, on average, by 63 percent of the country's Republican identifiers, as compared to just 13 percent for his nearest competitor (originally Elizabeth Dole; after her withdrawal, John McCain). As the data in the second half of Table 3 indicate, except for Bush's father in 1992, who was the incumbent president and faced fairly weak opposition, no other presidential aspirant of the past two decades has come close to achieving that level of dominance over the rest of the field.

To put these results in a more mathematically tractable form, and to allow for easier comparison with work on general election forecasting, I then combined these two indicators into a regression model that generates a numerical prediction of each candidate's performance in the presidential nomination race. 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 nomination season. In the 2000 Republican race, for example, 17,156,117 votes were cast in forty-two different primaries. George Bush received 10,844,129 of those votes, or 63.2 percent; John McCain had 5,118,187, or 29.8 percent; and so on.

Two independent variables are 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 taken before the start of delegate selection activities. 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. Again using the figures from the 2000 Republican race as an example, the most successful fund-raiser in that contest was George Bush, who raised $67,630,541. Bush thus receives a score of 100 on this variable. Steve Forbes, who raised $34,150,997, gets a score of 50.5; McCain, with net receipts of $15,532,082, is given a score of 23.0. (3)

The equation used to forecast the 2000 race, employing only the data from 1980 through 1996, is shown in the middle column of figures in Table 4. (4) The actual forecasts are shown in the third column of Table 2. The fund-raising variable is not statistically significant, but the poll standings variable clearly is and accounts for about 70 percent of the variance in the final primary vote shares. For reasons that will be explored later in this article, both Gore and McCain did somewhat better than the model forecast; Bradley fared less well than the model predicts. Yet the forecasts do mirror the general shape of the race as it ultimately turned out; in particular, the model does correctly forecast the winner in both contests.

Much of the rest of this article involves attempts to extend and elaborate this basic model. Before turning to that endeavor, however, I wish to call attention to two important conclusions that derive from this model, conclusions that plainly contradict what was, at least until quite recently, the almost universally accepted wisdom about how the presidential nomination process worked.

First, the contemporary presidential nomination process has gradually evolved into a system that is highly favorable to front-runners. In the first two nomination races that were held after the Democrats rewrote their delegate selection rules in the early 1970s, the Democratic Party nominated candidates who were definitely not the preelection front-runners. George McGovern in 1972 and Jimmy Carter in 1976 both began the primary and caucus season as decided long shots. Neither, in fact, was ever supported by more than 6 percent of his party's identifiers in any national poll conducted prior to the Iowa caucuses; nor did either candidate rank among his party's top money raisers. When George Bush Sr. came close to repeating this feat in 1980, a consensus quickly emerged that the new process was unusually well disposed to the political fortunes of outsiders, insurgents, and relative unknowns.

It wasn't just the press and the pundits who reached this conclusion. The same verdict was pronounced in some of the very best academic writing on the subject. One of the best-known models of the primary election process is that of Henry Brady and Richard 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" (p. 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 openness 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.
   (P. 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. (P. 95) (5)

Whatever may have happened in 1972 or 1976, a very different sort of outcome has occurred in every race since then. In the past ten contested nomination races, the candidate who won was the front-runner-or at least one of the front-runners-before any of the delegates were selected. To say the least, this is not a system characterized by its openness to new faces.

Second, if the prerace front-runner usually wins, then momentum is not the overwhelming force that it is frequently portrayed to be. This is not to say that momentum is dead or no longer occurs. Later in this article, in fact, I will provide at least one plausible way of estimating its net effect. But over the past two decades, the limits on momentum seem much more striking than its potency.

Momentum can be compared to a roller coaster ride: it provides a lot of thrills and excitement, but in the end, it leaves us pretty much where we started out. And so it is in presidential politics: after all the effects of momentum have come and gone, the person who starts out ahead almost always finishes ahead. Put another way, if you are interested in figuring out why John McCain went from 15 percent to 34 percent 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 determining who finally gets nominated and who does not, momentum just is not of very much help. Not since Jimmy Carter in 1976 has a momentum-driven candidacy been successful.

Candidate Emergence during the Invisible Primary

However successful this model is in forecasting the nomination winner, it obviously is not the whole story. To the contrary, the preceding analysis positively begs at least one major question. If major party presidential nominations are usually won by the person who is the front-runner at the end of the invisible primary, how does one become such a front-runner? How is it that one candidate breaks out of the pack of declared presidential candidates?

To answer this question, I have conducted an intensive analysis of the surveys of national party identifiers that are conducted throughout the invisible primary period. (6) Because of their accessibility, reliability, and methodological consistency, I have tended to rely on polls conducted by the Gallup Organization, but I have also made use of similar surveys conducted by a wide variety of other survey research firms. (7) I have defined the invisible primary in the broadest possible terms, beginning on the day after the preceding presidential election. In the 2000 election cycle, in other words, the invisible primary extended from November 6, 1996, to January 23, 2000 (the Iowa caucuses took place on January 24). For those who want a more detailed look at the data, the Gallup results for both 2000 nomination races are shown in Table 5 (complete data for all races will be published in Mayer forthcoming).

The ten nomination races being considered here break down rather nicely into four categories, which are set out in Table 6.

The first category consists of three nomination races in which one candidate led all of his competitors from wire to wire: one candidate, that is to say, was clearly ahead in the polls throughout the entire invisible primary period. The three candidates who fit this description are George Bush in 1988, Bush again in 1992, and A1 Gore in 2000. Each was ahead in the first poll conducted after the preceding election, was still ahead in the final poll before the Iowa caucuses, and then triumphed in the primaries and at the convention.

So far as I can determine, the first national poll conducted about the 2000 Democratic nomination race was a survey by Yankelovich Partners on March 11 to 12, 1997. That poll showed Gore already dominating the field. Offered a choice among six potential candidates--Al Gore, Dick Gephardt, Bob Kerrey, John Kerry, Bill Bradley, and Jesse Jackson--41 percent of the country's Democrats chose Gore; his nearest competitor, Jesse Jackson, had just 15 percent. The first Gallup survey on the race, in early September 1997, had it Gore 47 percent and Jackson 13 percent. Gallup asked the same question twenty more times over the next twenty-eight months, varying the list of candidates as different contenders tested the waters or dropped out, and Gore led every time, by an average margin of twenty-nine percentage points. Though Bradley did narrow the race somewhat in the closing months of 1999, Gore's lead over the former New Jersey senator was never smaller than twelve percentage points.

In many respects, Gore was simply revisiting the path trod by George Bush Sr. twelve years earlier. The resemblance is, of course, no accident: both were sitting vice presidents, serving under relatively popular presidents who were, because of the Twenty-Second Amendment, legally unable to run for reelection. In late December 1984, less than two months after Ronald Reagan was reelected, Penn and Schoen asked a national sample of Republican identifiers which of five candidates was their "first choice" for the 1988 nomination. Bush won 40 percent of the votes; his nearest competitor, Howard Baker, received just 20 percent. In the first Gallup sounding, in June 1985, Bush's vote was little changed, but Baker, who had chosen not to run for reelection in 1984, saw his standing in the polls enter into a sharp decline. (8) By January 1987, the Grand Old Party race was shaping up as a two-man contest, with Bush clearly in the lead, Bob Dole in second place, and everybody else lagging well behind. One year later, on the eve of the Iowa caucuses, that was still the case.

The survey record on the 1992 Republican nomination is considerably thinner, but the data that are available clearly suggest that this race, too, belongs in the first category. The problem, of course, is that up until quite late in the invisible primary period, the 1992 Republican race was expected to be uncontested and, thus, hardly a fit subject for extensive polling. Indeed, through the first two and a half years of the Bush presidency, I have been able to locate just one question about the 1992 nomination, from a March 1990 Gallup poll. Its results conform nicely, however, with what most observers would probably have expected. Simply put, Bush was the overwhelming favorite of the Republican rank and file. Fully 65 percent wanted him to be their party's presidential candidate in 1992; no other of the eight candidates mentioned received more than 7 percent.

In late 1991, however, two candidates who were granted at least some measure of credibility by the media entered the race against Bush: television commentator and former presidential speechwriter Pat Buchanan and former Louisiana state representative and Ku Klux Klan leader David Duke. At least in the polls, neither even came close to denting Bush's lead. In three separate Gallup surveys conducted between December 1991 and early February 1992, Bush was supported, on average, by 85 percent of Republicans, as compared to 9 percent for Buchanan and 4 percent for Duke.

The second category in Table 6 includes four races in which the eventual nominee also ran strong in the polls almost from the very beginning. In each of these cases, however, the nominee-to-be did face serious competition from one other potential candidate. In every case, though, that chief rival ultimately decided not to run for president, leaving the eventual nominee as the clear front-runner. Perhaps the best way to describe this category, then, is to say that the eventual nominee established an early lead in the polls among those candidates who actively sought the nomination and then maintained that lead through the rest of the invisible primary. The four candidates in this category are Ronald Reagan in 1980, Walter Mondale in 1984, Robert Dole in 1996, and George Bush in 2000.

A good example of this pattern is the Republican nomination race of 1996. The first survey on this race was a Hart-Breglio poll for NBC News and The Wall Street Journal, conducted between December 12 and 15, 1992. The leader in that poll, with 20 percent of the vote, was Jack Kemp, one of the few members of the outgoing Bush administration who had received positive notice during Bush's final year in office. But right behind Kemp--in a statistical tie, in fact--was Robert Dole, at 19 percent. James Baker, the former secretary of state who was then serving as Bush's chief of staff, had 12 percent, while Vice President Dan Quayle received 10 percent. Over the next year and a half, however, Kemp, Baker, and Quayle were sitting on the sidelines, while Dole emerged as one of the most visible leaders of the opposition to Bill Clinton. By August 1993, a poll conducted for U.S. News & World Report had Dole solidly in first place, eleven percentage points ahead of his nearest competitor. Kemp, by contrast, had slipped to third place, sixteen points behind Dole.

By the early spring of 1995, after Kemp and Quayle had both announced that they would not be candidates for the 1996 nomination, Dole had opened up a huge lead over every other announced candidate. An April 1995 Gallup survey found that 46 percent of the nation's Republicans said that Dole was the candidate they would be "most likely to support" for their party's next presidential nomination, as compared to just 13 percent for Phil Gramm and 8 percent for Pat Buchanan.

There was one candidate, however, who ran competitively with Dole through at least part of the 1996 invisible primary season: former chairman of the Joint Chiefs of Staff Colin Powell. How well Powell fared against Dole varied over the course of the campaign. In the second half of 1993 and through most of 1994, polls showed Dole slightly ahead of Powell--by about six percentage points, on average. (Dole's lead was not just a product of sampling error, however: he was ahead of Powell in every one of the six surveys conducted during this period.) In early 1995, Dole opened up a much larger lead over the former general. Surveys taken between late March and early June all had Dole trouncing Powell by about twenty-five points. In the fall of 1995, however, Powell benefited from a huge wave of publicity surrounding the release of his autobiography and the accompanying book tour. With just one exception, polls taken between mid-September and early November--Powell finally declared that he would not be a candidate on November 8--showed Dole with, at best, a single-digit lead over Powell. In two polls, Powell was actually ahead of Dole.

All told, I have been able to locate twenty-six polls from the 1996 invisible primary period that matched Dole and Powell as rival candidates for the 1996 Republican nomination (in every case, the question also named a number of other potential candidates). Dole led Powell in twenty-three of these surveys (though in some cases the difference was less than the margin of sampling error), Powell led Dole twice, and in one survey they were tied. There is, then, no compelling reason to think that Powell would have bested Dole even if he had entered the nomination race. The important point is that Powell was the only candidate who even threatened to dislodge Dole from the front-runner's position. Without Powell in the race, the Senate majority leader had, by early 1995, established a large lead over every other announced candidate and then maintained that lead until the beginning of the actual caucus and primary season.

In a similar way, the only candidate who challenged Ronald Reagan during the 1980 invisible primary was eventual noncandidate Gerald Ford. Reagan generally (though not invariably) beat Ford in questions that named the full field of potential Republican candidates. But Gallup also asked a series of "showdown" questions in which respondents were asked to say whom they would support if the 1980 Republican contest "narrows down to Gerald Ford and Ronald Reagan." In this type of question, Ford generally beat Reagan. Whichever question one considers a better measure of voter sentiment on this race, the essential point, of course, is that Ford never did get into the race and that among those candidates who did get in, none came close to challenging Reagan's hold on the Republican rank and file.

For the first two years of the 1984 invisible primary period, the clear front-runner in polls about the Democratic nomination race was Senator Edward Kennedy, who had also run for his party's nomination in 1980. In an April 1982 survey, for example, Gallup found that 45 percent of all Democratic identifiers wanted to see Kennedy as their party's next presidential candidate. The second-place finisher, Walter Mondale, was supported by just 12 percent. On December 1, 1982, however, Kennedy announced that he would not be running for president in 1984. Less than two weeks later, a new Gallup survey showed that Mondale was now the front-runner. The former vice president won the backing of 32 percent of his fellow partisans; in second place, with 14 percent, was John Glenn. Every other potential candidate was in single digits. Fourteen months later, on the eve of the Iowa caucuses, that was still pretty much the pecking order.

Against this background, the presidential odyssey of George Bush in 2000 seems, in many ways, strikingly familiar. Like Bob Dole in 1996, Bush was not the leader in the very first poll conducted on the 2000 nomination race. A Hart-Teeter survey conducted between December 5 and 9, 1996, found Colin Powell at the head of the pack with 37 percent, followed by Jack Kemp, the party's 1996 vice presidential candidate, at 20 percent and Bush at 19 percent. A Harris survey in January 1997 showed Bush in fourth place, backed by just 7 percent of all Republicans, trailing Powell, Kemp, and even Dan Quayle. By the early spring of 1997, however, Bush was already starting to move up in the polls. In late April of that year, a Hart-Teeter survey that did not include Powell's name in its list of prospective candidates had Bush now leading the field, with 24 percent of the vote, followed by Kemp with 17 percent and Elizabeth Dole with 16 percent. Surveys through the rest of 1997 that used the same set of candidates produced highly similar results. Bush was always in the lead, supported by 20 percent to 25 percent of Republican identifiers, with Kemp, Dole, and Quayle all clustered about five to ten percentage points behind him.

It was during 1998 that George Bush clearly began to put some distance between himself and the rest of the Republican field. A May 1998 Gallup survey found Bush now supported by 30 percent of the nation's Republicans, while Elizabeth Dole was still stuck at 14 percent, and Kemp and Quayle had slumped to 9 percent. In an October Gallup poll, it was Bush 39 percent, Dole 17 percent, and Quayle 12 percent (Kemp's name was not included in the question). A September 1998 Hart-Teeter survey showed Bush with an even larger lead. The Texas governor was favored by 41 percent of all Republicans, while Kemp had the support of just 11 percent, and Dole had 10 percent.

As the preceding discussion has already suggested, the candidate who ran best against Bush during this period--the only candidate who ever really threatened his position as the early front-runner--was Colin Powell. Given the widespread assumption that Powell would probably not be a presidential candidate in 2000, I have been able to locate only six polls from the 2000 invisible primary period that included both Bush's and Powell's names. The results for all six are shown in Table 7. In late 1996 and early 1997, Powell ran well ahead of every other Republican hopeful, Bush included. By early 1998, however, Powell's star had started to fade, while Bush's was in the ascendant. A March 1998 Opinion Dynamics poll had Bush beating Powell, 29 percent to 15 percent, though a July Harris survey had Powell narrowly besting Bush, 24 percent to 20 percent. Unfortunately, that seems to have been the last time that a question of this type was asked. As in the 1996 race, then, the available evidence does not provide much ground for confidence that Powell would have won even if he had thrown his hat in the ring. His decision to stay out, however, clearly did remove one more obstacle in the path of the Bush juggernaut.

The Exceptions

The most striking conclusion to emerge from this analysis is how early in the election cycle most eventual nominees were established as their party's front-runner. In seven of the ten cases considered here, the nominee-to-be had opened up a sizable lead over every other eventual candidate by, at the latest, one month after the preceding midterm election--more than a year, in other words, before the start of the actual delegate selection activities and at least a year and a half before the opening of the national conventions.

To appreciate the challenge that faced Bill Bradley and John McCain in 2000, however, it is also worth taking a close look at the exceptions: the three cases in which the eventual nominee was not the early front-runner.

Two of these exceptions--Jimmy Carter in 1980 and Bill Clinton in 1992--did not become front-runners until relatively late in the invisible primary. Up until the final two months before Iowa, Carter and Clinton had quite unimpressive polling numbers, but both then leaped ahead of the competition in December of the year preceding the election and January of the election year.

It is a remarkable commentary on Jimmy Carter's tenuous grip on his own party and the rocky course of his tenure in the White House that by late March 1978, barely fourteen months into his presidency, Carter was, at least according to the polls, no longer the frontrunner for the next Democratic presidential nomination. Matched up against five other leading Democrats, Carter was supported by just 29 percent of his fellow partisans. The leader in that poll was Edward Kennedy, and over the next year or so, Kennedy's lead grew significantly larger, to the point that by the summer of 1979, he was preferred to Carter by almost a three-to-one margin both in head-to-head matchups and when other Democratic hopefuls such as Jerry Brown were included in the question.

In retrospect, it seems clear that Kennedy's lead was artificially, misleadingly high. On one hand, he was benefiting from the aura that had long surrounded his family's name, an aura that, at the time, largely transcended partisan and ideological categories. In a July 1979 survey, for example, Gallup found Kennedy beating Carter within every major demographic category in the Democratic Party, with strikingly little variation. Even among southern Democrats, Kennedy led by a 58 percent to 37 percent margin. In an October 1979 Gallup poll, 36 percent of all Democrats described Kennedy's "political position" as right of center, as against just 44 percent who located him (correctly) to the left of center. The early dynamics of the 1980 campaign also aided Kennedy. Throughout the first ten months of 1979, Carter's weaknesses and shortcomings were at the center of popular and press attention, while the principal storyline involving Kennedy concerned how popular he was and how many people wanted him to run for president.

There is good reason, then, to think that once Kennedy became an official candidate and his real policy views became more widely known, his lead would decline somewhat. (9) But there was nothing inevitable about how far and how fast Kennedy's numbers fell. In mid-October 1979, the nation's Democrats preferred Kennedy to Carter, 60 percent to 30 percent. By early January 1980, Carter was leading Kennedy, 51 percent to 37 percent. The principal reason for this abrupt turnaround is not difficult to identify. In the final two months of 1979, Carter's political fortunes benefited from two major foreign policy crises. On November 4, the American embassy in Teheran was seized by Iranian students who took sixty-five Americans hostage; on December 27, the Soviet Union invaded Afghanistan. In the predictable way that the American public rallies around the president during a time of international stress, Carter's approval ratings, which had been hovering in the high twenties and low thirties since May 1979, soared to 58 percent in mid-January 1980. His standing vis-a-vis Kennedy closely parallels his approval ratings. Absent events in Iran and Afghanistan, it is difficult to believe that Carter would have beaten back Kennedy's challenge. (10)

The rise of Bill Clinton in late 1991 and early 1992 is less easy to explain. The Gulf War of January and February 1991 made the lead-up to the 1992 Democratic nomination contest quite different from every other out-party race considered here. On one hand, the active campaign simply started a lot later. Whereas most contemporary presidential candidates officially launch their campaigns in the late winter or early spring of the year before the election, by early September 1991, the Democrats had only one declared candidate. Even more striking was the fact that all of the putative front-runners--every Democrat who ranked among the leaders in early polls of the national rank and file--ultimately decided not to make the race. Mario Cuomo, Jesse Jackson, Lloyd Bentsen, Dick Gephardt, A1 Gore, Bill Bradley, George McGovern, and Sam Nunn all decided they had other commitments or thought there was too little chance of beating George Bush.

The result was that when the Democratic nomination race finally began to take shape in the fall of 1991, it lacked a real front-runner. The leader in the polls was Jerry Brown, the only candidate who had previously run for national office, but compared to almost every other race considered here, Brown had, at best, a weak grip on the top spot. In two different Gallup polls--one in mid-September and the other in early November--only 21 percent of the nation's Democrats wanted Brown to be their next presidential standard bearer, a level of support far below that enjoyed by Kennedy in 1979, Mondale in 1983, or Gore in 1999. The other five candidates each received between 5 percent and 10 percent of the vote.

Between mid-October 1991 and late January 1992, however, Bill Clinton gradually emerged from this tightly clustered pack. By early January, Clinton was all but tied with Brown, and by late January, the Arkansas governor had opened up a twenty-percentage-point lead over all the other declared candidates. From October through mid-January, Clinton's gradual rise in the polls is probably attributable to the fact that (1) he was receiving more press coverage than any other candidate, and (2) much of that coverage was strikingly favorable. What finally sent Clinton soaring past all his competitors, however, was the controversy that erupted in late January over his marital infidelities. The episode began on January 17, when a tabloid called The Star published a front-page article claiming that Clinton had had at least five extramarital affairs while serving as governor of Arkansas. The story was quickly picked up by the mainstream media and soon began to dominate coverage of the campaign. On January 26, in an attempt to stem the damage, Clinton and his wife appeared on a special 60 Minutes that was aired immediately after the Super Bowl, to answer questions about the matter.

On January 16, just before the original Star article appeared, a Yankelovich Clancy Shulman poll showed Brown and Clinton in a dead heat, each holding on to 22 percent of the Democratic electorate. By early February, four separate polls showed Clinton leading the rest of the field by margins of between sixteen and twenty-six percentage points. How this controversy came to work to Clinton's advantage is unclear, especially since in New Hampshire it seems to have had quite the opposite effect, sending the Clinton campaign into a sharp decline while pouring new life into that of Paul Tsongas. In part, Clinton's soaring numbers in the national polls may have been a simple matter of publicity and name recognition: by the time the dust had settled, Clinton was far and away the best known of the Democratic candidates. Clinton may also have gotten some credit for the resilience and determination he showed in continuing his campaign and perhaps some support from those who saw him as the victim of media excess. Whatever the exact set of causes, the result is clear: by the eve of the Iowa caucuses, Clinton had a large lead in polls of Democratic identifiers.

The final category in Table 6 is the one envisioned in the classic accounts of momentum: a candidate who is not the front-runner at the end of the invisible primary but manages to do well in some of the early primaries and caucuses, which then propels him to a lead in the national polls, further success in the primaries, and ultimately the nomination. What is striking is that this last category includes just one candidate: Michael Dukakis in 1988.

The early poll leader in the 1988 Democratic nomination race was Gary Hart, who had narrowly lost the nomination to Walter Mondale in 1984. In April 1987, about two months after Mario Cuomo had officially declared that he would not be a candidate, a Gallup poll showed Hart supported by 46 percent of Democratic identifiers, as compared to 18 percent for Jesse Jackson and single-digit numbers for everybody else. In early May, however, The Miami Herald revealed that Hart had spent a night in his Washington, DC, townhouse with a sometime model named Donna Rice. Five days later, Hart dropped out of the race.

For the next seven months, the 1988 Democratic race, like the party's 1992 contest, lacked a clear front-runner. In first place, according to four different Gallup polls, was Jesse Jackson, but like Jerry Brown in 1992, Jackson was never supported by more than 22 percent of his fellow partisans. The rest of the vote was divided among six other announced Democratic candidates, (11) none of whom seems to have been very well known outside of his own home state, with an unusually large number of respondents (about 40 percent) saying they were undecided. To further confuse matters, on December 15 Hart reentered the race and, probably just because he was better known than all the other candidates except Jackson, immediately resumed his lead in the polls. Yet the Donna Rice affair had clearly done its damage: in nine different polls conducted during January 1988, Hart was favored, on average, by just 22 percent of the nation's Democrats, about half the support he had had before exiting the race.

So Dukakis was not the clear front-runner in the 1988 Democratic nomination race on the eve of the Iowa caucuses. Yet neither was he a prohibitive long shot. To begin with, Dukakis was, as noted in Table 1, his party's leading fund-raiser during 1987. In fact, he raised almost twice as much money during the invisible primary as any of his competitors. Even in the polls, Dukakis generally finished third, just behind Hart and Jackson. (12) If, as was widely assumed at the time, Hart would falter once the actual caucus and primary season began and Jackson's support was intense but quite narrow, Dukakis was well positioned to pick up the pieces.

At the risk of generalizing on the basis of just three cases, the preceding analysis underlines just how difficult the task was facing McCain and Bradley. To derail an early frontrunner, it would appear, takes a special set of circumstances. Kennedy's defeat is primarily attributable to one of the worst U.S. foreign policy crises in the final quarter of the twentieth century. The 1988 Hart campaign was wrecked by the personal self-destructiveness of the candidate. The 1992 Democratic nomination race, I think it fair to say, never really had an early front-runner. Of the candidates who did decide to run for president that year, none had a clear and commanding lead over the rest of the field. Of course, every presidential campaign has its unique occurrences and unexpected twists. If the McCain and Bradley strategists were hoping that something similar might happen in 1999 or 2000, it was not entirely an exercise in wishful thinking. The point is that they probably did need something well out of the ordinary: merely running a good campaign, history suggests, would not be enough.

Enter Iowa and New Hampshire

Having examined how candidates become--or fail to become--front-runners, I now want to extend the analysis in the opposite direction. As the forecasts at the bottom of Table 2 should suggest, the basic model described in the first section of this article does correctly forecast the eventual nominee and perhaps says something useful about the general shape of the race. Yet some of the individual predictions are clearly rather wide of the mark. (13) What accounts for these errors?

One obvious place to look is the events that lead off the delegate selection calendar. Compared to other types of elections, perhaps the most distinctive feature of presidential primaries is that they occur sequentially. While the states in a general election all vote on the same day, presidential primaries take place over a period of approximately four months. Under such an arrangement, it would not be surprising to find--indeed, it is probably to be expected--that the voting in later primaries is affected by what happens in earlier primaries.

The general term for such an effect, of course, is momentum. Though I have argued earlier that momentum has somewhat more limited effects than has generally been supposed, it clearly does have some impact. In this section, I want to make some attempt to incorporate the effects of early momentum into the basic model, to see what difference it makes and how much total effect it has on the final outcome.

To be a bit more specific: if there is one moment when nomination races genuinely do seem up for grabs, it is at the very beginning of the delegate selection calendar, when two events dominate the proceedings: the Iowa caucuses and the New Hampshire primary. To what extent do these two states rearrange the basic structure of the race? How much more predictable does the race become once these two states have had their say? (14)

To answer these questions properly, one important preliminary issue needs to be addressed. In what form are the results of Iowa and New Hampshire incorporated into the results of later primaries? When we say that voters in states such as California, Illinois, and Ohio are "influenced by" what has happened in Iowa and New Hampshire, what exactly are they reacting to? It seems clear, to begin with, that what matters is not the raw, unadjusted vote percentage received by each candidate. In 2000, for example, Bill Bradley received 45.6 percent of the votes cast in the New Hampshire Democratic primary, a percentage substantially higher than that received by Jimmy Carter in 1976, Gary Hart in 1984, or Paul Tsongas in 1992. Yet all indications are that Bradley received far less benefit from that showing than any of these other candidates. In some fashion, then, these raw percentages are adjusted or discounted because of the special circumstances and conditions that accompany a given election.

Obviously, one thing that matters is the order in which the candidates finish. Carter, Hart, and Tsongas all gained a lot from New Hampshire because they won that state's primary; Bradley's vote totals were seen as much less impressive for the simple reason that Al Gore beat him. In multicandidate races, it also probably helps to come in second, since that at least allows a candidate to project himself as the principal opponent of the winner.

Yet there is clearly a lot more to the story than that. In many cases, what seems to matter is not who wins or loses in an absolute sense but whether a candidate does better or worse than "expected." In 1984, to take a particularly celebrated example of this phenomenon, Walter Mondale posted one of the most dominating victories in the history of the Iowa caucuses yet seems to have received almost no benefit from that showing because it was so widely expected. The candidate whose campaign was buoyed by Iowa in 1984 was Gary Hart, who finished more than 30 percent behind Mondale yet performed better than most observers had anticipated he would.

So expectations matter. The problem, from the perspective of model building, is finding a plausible way to operationalize this hypothesis. We need, that is to say, a reasonable way to determine how well each candidate was expected to run in Iowa or New Hampshire, as a standard against which to compare their actual showing. Yet the more one studies this subject, the more clear it becomes that a quite large number of factors enter into these expectations judgments, many of which are probably idiosyncratic to a particular race and candidate.

To a large extent, expectations seem to be based on how well each candidate is perceived to be doing in the race as a whole. This, for example, explains why Hart's Iowa showing was regarded as such a significant story in 1984. Hart won 16 percent of the delegates to the next round of caucusing at a time when he had just 3 percent in the national polls. By contrast, Mondale's 49 percent in the Hawkeye State exactly matched his standing in Gallup's most recent national poll and thus raised few eyebrows. Similarly, Buchanan's second-place showing in the 1992 New Hampshire primary was widely regarded as a victory because he won 37 percent of the vote, far higher than the 11 percent he was then receiving in the national polls.

Yet there are also lots of exceptions and qualifications to this generalization. Richard Gephardt's 1988 Iowa victory, for example, was substantially discounted because he came from a neighboring state. On the surface, Steve Forbes's showing in the 2000 Iowa caucuses would seem to resemble Gary Hart's situation in 1984: Forbes came in second to George Bush (a quite strong second, in fact), and Bush's victory had long been expected. Yet Forbes, so far as one can tell, got little benefit from that performance, partly because Bush's principal rival in the national polls (John McCain) had skipped Iowa and partly because most reporters and commentators thought that Forbes had little real chance of winning the nomination. In New Hampshire, most of these same factors come into play--but the Iowa results are also taken into account. In 1988, George Bush seems to have gained a lot from his New Hampshire victory: even though he was the early front-runner, his front-runner credentials had been badly tarnished by his third-place finish in Iowa. The same thing happened to Ronald Reagan in 1980.

In the end, I have tried to measure the impact of Iowa and New Hampshire by creating, for each event, three pairs of variables.

1. The first is a pair of dummy variables that designate the winners in Iowa and New Hampshire, respectively.

2. There is a second pair of dummy variables that designate the second-place finishers in Iowa and New Hampshire but only in multicandidate races. As indicated earlier, the logic behind this variable is that in a crowded field, the second-place candidate at least establishes himself as the principal opponent of the front-runner, a designation that may be of considerable value in upcoming primaries.

3. Finally, there is a measure of how well each candidate finished in Iowa and New Hampshire relative to expectations. This variable is derived by taking the candidate's actual showing in each state and subtracting from it the candidate's current standing in the national polls of party identifiers. (15)

The equation that results, estimated for all races except those held in 1992, (16) is shown in Table 8. Three points are worth noting. First, to no one's great surprise, the earliest delegate selection events do have a substantial effect on the general shape of a contested nomination race. Adding in the results of Iowa and New Hampshire increases the adjusted [R.sup.2] in the forecasting equation by .12 (i.e., the new variables explain an additional 12 percent of the total variance). More important, the final forecasts become substantially more accurate. For the original equation, the mean absolute prediction error for all forty-five candidates is 7.9: that is to say, the average forecast was off by 7.9 percentage points. With the New Hampshire and Iowa results added in, the mean absolute error falls to 5.3 points.

Second, what seems to matter, according to Table 8, is how the candidates actually finish, not how they finish relative to prior expectations. There are two ways of interpreting this result. One is, as I have already conceded, that I simply may not be measuring expectations very accurately. Alternatively, it may be, particularly when it comes to New Hampshire, that reporters and commentators have become somewhat more suspicious of expectations judgments and are therefore increasingly reluctant to second-guess the hard, mathematical results that get registered in the voting booths. In 1968 and 1972, the media received a lot of criticism for turning New Hampshire victories by Lyndon Johnson and Edmund Muskie into losses, based on the highly debatable notion that these candidates had run "worse than expected." Just as one might have anticipated, in subsequent nomination races, many campaigns actively tried to manipulate these expectations by downplaying their chances before the primary and then painting their actual showing in the rosiest possible terms. As a result, media reports on New Hampshire have tended in recent years to focus more closely on the actual results: wins are wins and losses are losses. Over the past six election cycles, the only candidate who managed to convert a New Hampshire loss into a perceived victory was Pat Buchanan in 1992. (17) While many other candidates tried to duplicate this feat, there is no evidence that they were successful.

The Role of Iowa

The third and perhaps most surprising result in Table 8 is that the capacity of early delegate selection events to affect the final outcome of a presidential nomination race is due entirely to New Hampshire. The two coefficients that measure the order of finish in New Hampshire are both large and statistically significant. After controlling for everything else, a win in the New Hampshire primary increases a candidate's expected share of the total primary vote by a remarkable 26.7 percentage points. Even a second-place finish in New Hampshire (in a multicandidate race) increases a candidate's final vote totals by 17.1 percent. By contrast, the two coefficients that measure the impact of Iowa are both statistically insignificant. According to one of them, a win in Iowa actually reduces a candidate's expected total vote in the primaries. The general finding that New Hampshire substantially outweighs Iowa, it is worth adding, stands up quite well to changes in specification. For example, if we drop the four dummy variables in Table 8 that designate the first- and second-place finishers in Iowa and New Hampshire, and measure the effects of these two events solely by comparing each candidate's finish relative to expectations, the New Hampshire variable still turns out to be large and statistically significant; the Iowa variable is very small and insignificant.

One of the clear messages of this analysis, then, is that Iowa has substantially less impact than New Hampshire on the final outcome of a contested nomination race. Yet Iowa is not quite as impotent as a superficial reading of Table 8 might suggest. The key to understanding the role of Iowa in the presidential nomination process is to recognize that the coefficients in Table 8 show the effect of Iowa holding New Hampshire constant. Iowa does have an impact on many races, but that impact is mediated through New Hampshire. The general set of relationships that governs the earliest stages of the presidential nomination process is shown in Figure 1. What is important about Iowa, according to this model, is that it sets the table for New Hampshire. Iowa influences New Hampshire, and New Hampshire and the national poll standings, in turn, affect the final distribution of votes in the presidential primaries. Put another way, a candidate who does well in Iowa puts himself in a somewhat better position to do well in New Hampshire. But if that does not occur-if a strong showing in Iowa is not followed up by a good performance in New Hampshire--the Iowa result will probably not help the candidate much in the long run.


How much effect do the Iowa caucuses have on the New Hampshire primary? This question is difficult to answer succinctly. Consider the data in Table 9. For each of the eight nomination races being considered here (again, I exclude 1992), this table shows the results of a poll of New Hampshire voters conducted just before the Iowa caucuses, the results of a second New Hampshire poll conducted shortly after Iowa, and the final results of the New Hampshire primary. It is clear, at a minimum, that New Hampshire listens to Iowa: in every race shown here, the polls in New Hampshire show movement in response to the Iowa caucus results. In a fair number of instances, however, that movement turns out to be short lived. In 1988, for example, Bob Dole's win in Iowa, coupled with a fairly weak showing by front-runner George Bush, vaulted Dole to first place in the Granite State polls. But in the last few days before the primary, Bush waged a strong counterattack that reestablished his lead--and effectively ended any shot Dole had of winning the White House that year. When the two changes are added together, the final New Hampshire results ended up very close to what the polls were showing on the day before Iowa.

Something very similar happened in the 2000 Democratic race. A1 Gore trounced Bill Bradley in the Iowa caucuses, 63 percent to 35 percent; as a result, Gore's lead in New Hampshire, which had been just 5 percent before Iowa, widened to 16 percent by the end of the week. In the last few days before the primary, however, Bradley finally regained his footing and began to respond to the attacks Gore had been making on him for the last several months. That effort significantly narrowed Gore's lead but ultimately fell short. As with the Republicans in 1988, the final Democratic vote totals in 2000 look remarkably like what the polls had been showing just before Iowa.

Yet there clearly are cases where Iowa has had a more lasting effect on New Hampshire. The best example of this occurred during the 1984 Democratic nomination race. In the week before Iowa, a Washington Post/ABC News tracking poll showed Walter Mondale with a substantial lead over the rest of the Democratic field. Mondale had 37 percent of the vote, to 20 percent for John Glenn and just 13 percent for Gary Hart. Immediately after finishing second in Iowa, however, Hart's support began to climb. By election day, Hart had 37 percent of the vote, while Mondale had slipped to 28 percent, and Glenn was down to 12 percent. In 1996, the results in Iowa buoyed the New Hampshire campaigns of Pat Buchanan and Lamar Alexander, while dooming those of Phil Gramm and Steve Forbes.

The erratic, idiosyncratic way that the results in Iowa have influenced the course of events in New Hampshire necessarily makes it difficult to model that process. Though I have been tempted at times to introduce a host of new distinctions and qualifications, I have chosen in the end to estimate a model similar to the one in Table 8. Specifically, the final vote in New Hampshire is hypothesized to be a function of three variables:

1. the candidates's standing in the New Hampshire polls just before Iowa (i.e., the data in the first column of figures in Table 9), (18)

2. a dummy variable that designates the winner of the Iowa caucuses, (19) and

3. a second dummy variable that designates second-place finishers in Iowa but only in multicandidate races. I also exclude from this variable the second-place finisher in the 1980 Republican caucuses, Ronald Reagan, on the grounds that he was the prerace front-runner, for whom a second-place finish was definitely not regarded as a victory.

The result is shown in Table 10. In general, according to this model, candidates who won the Iowa caucuses got little or no benefit in New Hampshire from that showing. The benefits from Iowa have tended to go to the second-place finisher. After controlling for the candidate's pre-Iowa standing in the polls, a second-place finisher could expect to increase his share of the New Hampshire primary vote by about eight percentage points. Since the winner of the New Hampshire primary in these eight races received, on average, just 42 percent of the vote, an eight-point boost is not insignificant. As the preceding discussion has tried to make clear, however, that eight-point figure is an average that masks a considerable amount of variation. Hart's second-place finish in Iowa helped boost his New Hampshire vote by twenty-four percentage points; in 1996, Pat Buchanan gained 11 percent between Iowa and New Hampshire. By contrast, finishing second to George Bush in the 2000 Iowa caucuses brought no apparent gain to Steve Forbes, in New Hampshire or anywhere else.

Overall, even the effect of Iowa on New Hampshire is somewhat weaker than I had initially anticipated. Some candidates in the Granite State have clearly benefited from running strong in Iowa, but for many others, the gains were small and fleeting.


The major purpose of this article has been to describe some basic dynamics of the contemporary presidential nomination process. My findings in that regard can be summarized in the form of three major propositions.

1. Its initial reputation notwithstanding, the contemporary presidential nomination process is actually quite favorable to front-runners. In all of the past ten contested nomination races, the eventual nominee was either (a) leading in the national polls of party identifiers at the beginning of the delegate selection season or (b) had raised more money than any other candidate in the year before the election. In eight of ten cases, the nominee-to-be led in both the polls and the fund-raising derby.

2. Most front-runners establish themselves in that position at a quite early point in the election cycle. Most recent nominees were leading the national polls by at least a year before the first delegates were selected.

3. Though momentum is not the overwhelming force that it is sometimes portrayed to be, strong showings in the earliest caucuses and primaries, particularly the New Hampshire primary, do have a clear and significant effect on a candidate's fortunes.

Against that background, little about the 2000 nomination races seems particularly surprising. Both Al Gore and George Bush were clearly the front-runners for their parties' nominations. In the national polls of party identifiers, Gore led the Democrats from wire to wire; Bush had a clear lead over the rest of the Republican field by mid-1998. On the eve of the Iowa caucuses, both candidates had a commanding advantage in the polls. Gore led Bradley 60 percent to 27 percent; 63 percent of the nation's Republicans wanted Bush to be their party's 2000 presidential candidate, while just 19 percent favored John McCain. Gore and especially Bush also had the monetary resources to wage a long campaign, if that became necessary.

In the end, Gore slightly outperformed Bush in the primaries. The vice president won 76 percent of the votes cast in the Democratic primaries, while Bush received 63 percent of his party's presidential preference votes. Much of this difference, it seems clear, can be explained by the candidates' differing fortunes in New Hampshire. John McCain's decisive victory in the Granite State gave his campaign a substantial boost in media attention, monetary contributions, and voter favorability, which made him a major factor in the Republican race for five weeks after New Hampshire. Bradley's narrow loss in that state (along with, perhaps, the peculiarities of the Democratic schedule) denied him these same advantages. The result was that Bradley never really came close to winning another primary.

To say all this is not to claim that things could not have been different. When a model of this sort forecasts a Gore victory, it does not mean--even assuming the model is correct--that Gore can go on vacation for the next six months. To the contrary, such models assume that the candidates will do their part: that each candidate will run a reasonably competent campaign and exploit whatever advantages he has. All campaigns, of course, will also make mistakes. The assumption, however, is that these mistakes will be evenly distributed between the campaigns and, perhaps, that they matter less than is sometimes claimed.

Of all the strategic and tactical decisions made during the 2000 nomination races, the most intriguing, given the analysis in this article, was the different ways that Bradley and McCain chose to deal with Iowa. Through most of 1999, Bill Bradley concentrated his time and money on New Hampshire, and by the early fall, it had started to pay off. By early September, numerous polls showed the race there tied or Bradley narrowly ahead. Late in 1999, however, Bradley decided to mount a major effort in Iowa, where he had not campaigned much to that point--and where polls showed him trailing Gore by about twenty percentage points.

Iowa had never looked like particularly fertile ground for Bradley's message, and the caucus format would only seem to magnify all the advantages that Gore derived from his support among unions and party officials. But the conventional wisdom going into the 2000 campaign held that both Iowa and New Hampshire were essential proving grounds for the serious national candidate, and Bradley chose not to put this nostrum to the test. Instead, he spent $2.2 million and half of January campaigning in Iowa. The most important penalty Bradley paid for that decision was not the decisive drubbing he suffered in the Hawkeye State but all the time diverted from New Hampshire. There is no way to prove, of course, that Bradley would have won New Hampshire even if he had concentrated his efforts there--but it almost certainly would have helped, and as it was, Bradley only lost the state by six thousand votes.

What gives particular force to such speculations is that John McCain chose the path that Bradley rejected. McCain completely ignored Iowa and instead devoted all of his considerable energy to New Hampshire, holding a total of 114 town meetings. There is, to say the least, no indication that the lack of an Iowa campaign handicapped McCain in New Hampshire. As can be seen in Table 9, McCain's support in New Hampshire did drop a bit in the days immediately after Iowa, but it had recovered by week's end, and the Arizona senator then went on to win the Granite State primary by eighteen percentage points.

Given that outcome and (to the extent anyone pays attention to this article) the results in Table 8, it seems likely that other candidates will try to copy McCain's strategy in the future. Yet other campaigns may not be quite so fortunate. McCain could afford to ignore Iowa in 2000 because no other candidate in the Republican nomination race was likely to derive much benefit from the caucuses. Bush was such an overwhelming front-runner that the media were unlikely to make much of his victory there. More important, there was no second-tier candidate who could use a "better than expected" showing in Iowa to make himself the story of the week in New Hampshire, as Gary Hart had done in 1984. Steve Forbes, who actually did finish second, had so many obvious limitations that the media and the voters were unlikely to take him very seriously. And the other candidates--Alan Keyes, Gary Bauer, and Orrin Hatch--seemed even less likely to catch fire. All of the candidates who might plausibly have used Iowa as a real source of momentum-Lamar Alexander, Pat Buchanan, and Elizabeth Dole-had dropped out months earlier. Absent such assurances, one suspects that most candidates in 2004 will take the safer route and mount serious campaigns in both Iowa and New Hampshire.

So Bradley probably made the wrong decision about Iowa. Yet it would be wrong, I think, to put too much emphasis on this one incident. If the Bradley campaign made its share of errors, the Gore campaign was also something less than a flawless operation. Indeed, for the first eight months of 1999, it is almost impossible to find a newspaper or magazine article that offers a favorable review of the vice president's campaign. For whatever combination of reasons, there was a widespread perception that the campaign was floundering, that it lacked a sense of direction, and that it was failing to connect with the voters (for further details, see Mayer 2001).

A Bradley win in New Hampshire would almost certainly have given his campaign a needed boost. But even then, I would argue, the odds would have been stacked against him. What mattered, in the end, was not campaign strategy but all the resources that A1 Gore brought to the race from his many years in politics and his eight years as vice president. The South, in particular, seemed to provide Gore with an almost impregnable firewall. Even in the early fall of 1999, when almost everything seemed to be going right for Bradley and everything seemed to be working against Gore, southern Democrats showed no inclination to climb on the Bradley bandwagon. On March 7, when Bradley averaged 41 percent of the vote in five New England state primaries, he won just 16 percent in Georgia, the only southern state that voted that day.

The same sort of verdict applies to the 2000 Republican race. Given the resources available to him, John McCain ran a remarkable race. Though the senator has sometimes been criticized for a speech he gave in mid-February attacking the influence of the religious right in the Republican Party, I know of no evidence to show that that speech actually changed many votes. McCain was already running poorly among conservative Christian voters, which is why he gave the speech in the first place. Bush won because of a clear consensus among Republican officeholders and contributors-and ultimately among ordinary votersthat he was the best person to represent the party in the fall 2000 campaign. McCain lost, in the most fundamental sense, because of who he was and the kinds of positions he took.


National Poll Standings and Preelection Year Fund-Raising Results as
Predictors of Presidential Nominations

Year                     Party     Leading    Eventual
                                   Candidate   Nominee

Poll standings (a)
  1980                Republican   Reagan      Reagan
  1980                Democratic   Carter      Carter
  1984                Democratic   Mondale     Mondale
  1988                Republican   Bush        Bush
  1988                Democratic   Hart        Dukakis
  1992                Republican   Bush        Bush
  1992                Democratic   Clinton     Clinton
  1996                Republican   Dole        Dole
  2000                Republican   Bush        Bush
  2000                Democratic   Gore        Gore
Fund-raising totals (b)
  1980                Republican   Connally    Reagan
  1980                Democratic   Carter      Carter
  1984                Democratic   Mondale     Mondale
  1988                Republican   Bush        Bush
  1988                Democratic   Dukakis     Dukakis
  1992                Republican   Bush        Bush
  1992                Democratic   Clinton     Clinton
  1996                Republican   Dole        Dole
  2000                Republican   Bush        Bush
  2000                Democratic   Gore        Gore

(a.) Candidate listed is the person leading in the last national poll
before the Iowa caucuses.

(b.) Candidate listed is the person who had raised the most money by
the end of the year before the election.


Presidential Nomination Predictors in the 2000 Nomination Races

                Poll Money      Predicted     Actual
               Standings (a)    Raised (b)     Vote     Vote

  Gore              60          27,847,335     61.2     75.7
  Bradley           27          27,465,950     28.5     19.9
  Bush              63          67,630,541     64.2     63.2
  McCain            19          34,150,997     20.5     29.8
  Forbes             6          15,532,082      7.6      0.8
  Bauer              2           8,761,166      3.6      0.4
  Keyes              1           4,483,505      2.6      5.3
  Hatch              1           2,285,829      2.6      0.1

Source: Poll results are taken from the Gallup poll, survey of January
17 to 19, 2000. Fund-raising totals are derived from the individual
candidate reports submitted to the Federal Election Commission.

(a.) Entries show the presidential nomination preferences of national
party identifiers, as taken from the Gallup Poll, survey of January 17
to 19, 2000.

(b.) Entries show the total net receipts of each candidate through the
end of 1999, as derived from the individual candidate reports submitted
to the Federal Election Commission.


Historical Comparison of George W. Bush and Other Early Front-Runners
in Contested Presidential Nomination Races


                                          Total Net
                                       Receipts in the     Inflation-
                        Year and       Year before the      Adjusted
Top Fund-Raiser           Party         Election ($)      Receipts (a)

John Connally        1980 Republican      9,160,000        21,020,000
Jimmy Carter         1980 Democratic      5,752,000        13,199,000
Walter Mondale       1984 Democratic     11,448,000        19,148,000
George Bush          1988 Republican     19,058,000        27,949,000
Michael Dukakis      1988 Democratic     10,371,000        15,210,000
George Bush          1992 Republican     10,093,000        12,346,000
Bill Clinton         1992 Democratic      3,304,000         4,041,000
Robert Dole          1996 Republican     25,192,000        27,539,000
George W. Bush       2000 Republican     67,631,000        67,631,000
Al Gore              2000 Democratic     27,847,000        27,847,000

                                   National Poll Ratings

                                                         Average Margin
Top Candidate in        Year and           Average        over Nearest
the National Polls        Party        Poll Rating (b)   Competitor (b)

Ronald Reagan        1980 Republican         42              27
Edward Kennedy       1980 Democratic         54              18
Walter Mondale       1984 Democratic         42              21
George Bush          1988 Republican         42              23
Jesse Jackson        1988 Democratic         19               6
George Bush          1992 Republican         86              80 (c)
Jerry Brown          1992 Democratic         21              10
Robert Dole          1996 Republican         47              37
George W. Bush       2000 Republican         63              50
Al Gore              2000 Democratic         56              22

Source: Fund-raising data are taken from Federal Election Commission
reports; poll ratings come from the Gallup poll.

(a.) Data are in 1999 dollars.

(b.) Data are average results for all Gallup presidential nomination
polls taken during the second half of the year before the election.

(c.) This is based on only one survey taken in December 1991.


Baseline Model for Predicting Primary Vote Shares

                                1980 to 1992         1980 to 1996

                             Regression           Regression
                             Coefficient    SE    Coefficient    SE

National poll standings          0.94      0.14       0.99      0.13
Total funds raised               0.02      0.08       0.00      0.07
Constant                         1.31      3.37       1.61      3.00
[R.sup.2]                         .70                  .72
Adjusted [R.sup.2]                .69                  .71
Standard error of estimate      11.93                11.37
n                               38                   45

                                1980 to 2000

                             Coefficient    SE

National poll standings          1.05      0.11
Total funds raised              -0.02      0.06
Constant                         1.75      2.57
[R.sup.2]                         .78
Adjusted [R.sup.2]                .77
Standard error of estimate      10.80
n                               53

Note: The dependent variable is the percentage of the total primary
vote won by a candidate in all presidential primaries held by that
candidate's party during the nomination season.


Presidential Nomination Preferences of National Party Identifiers
during the 2000 Invisible Primary

Republican nomination race: "Next, I'm going to read a list of people
who may be running in the Republican primary for president in the year
2000. After I read all the names, please tell me which of those
candidates you would be most likely to support for the Republican
nomination for president" (asked of Republicans and Republican-leaning


                         April 11-13,    September 6-7,    May 8-10,
Candidate                    1997             1997           1998

Colin Powell                  36               --           --
George W. Bush                14               21           30
Elizabeth Dole                 7               --           14
Jack Kemp                     14               15            9
Dan Quayle                    11               10            9
Steve Forbes                   6                9            7
Christie Whitman               4                9           --
Newt Gingrich                 --                5            6
Fred Thompson                 --                5           --
Pat Buchanan                  --                4            3
John Ashcroft                 --                3            * (a)
Lamar Alexander                2                3            2
Bob Smith                     --                2           --
John McCain                   --               --            4
Gary Bauer                    --               --            1
John Kasich                   --               --            1
Other/none/don't know          8               14           14


                         October 23-25,
Candidate                    1998

Colin Powell                  --
George W. Bush                39
Elizabeth Dole                17
Jack Kemp                     --
Dan Quayle                    12
Steve Forbes                   7
Christie Whitman              --
Newt Gingrich                  4
Fred Thompson                 --
Pat Buchanan                  --
John Ashcroft                  4
Lamar Alexander                4
Bob Smith                     --
John McCain                   --
Gary Bauer                    --
John Kasich                    4
Other/none/don't know          9

                 January         March          April        April 30-
                8-10, 1999    12-14, 1999    13-14, 1999    May 2, 1999

Bush                42            52           53             42
Dole                22            20           16             24
Quayle               6             9            7              6
McCain               8             3            5              4
Forbes               5             1            6              6
Buchanan            --             4            4              5
Kasich               2             3            2              1
Alexander            4             2            * (a)          3
Bauer                2             1            2              3
Hatch               --            --           --             --
Smith                1             1            * (a)          * (a)
  don't know         8             4            5              5

                May 23-24,    June 4-5,    June 25-27,
                   1999          1999         1999

Bush                46            46           59
Dole                18            14            8
Quayle               7             9            6
McCain               6             5            5
Forbes               5             5            6
Buchanan             6             6            3
Kasich               2             1            3
Alexander            1             3            2
Bauer                2             1            2
Hatch               --            --            2
Smith                2             1            1
  don't know         5             9            3


                  August        September      October        October
                16-18, 1999    10-14, 1999    8-10, 1999    21-24, 1999

Bush                61             62             60            68
Dole                13             10             11            --
Quayle               6              5             --            --
McCain               5              5              8            11
Forbes               4              5              4             8
Buchanan             3              3              3            --
Bauer                2              2              3             1
Hatch                1              2              2             3
Keyes                1              1              3             2
  don't know         4              5              6             7

                November 4-7,    November 18-21,
                    1999              1999

Bush                 68                63
Dole                 --                --
Quayle               --                --
McCain               12                16
Forbes                6                 6
Buchanan             --                --
Bauer                 2                 4
Hatch                 2                 4
Keyes                 2                 1
  don't know          8                 6


                 November       December       December      January
                18-21, 1999    9-12, 1999     20-21 1999    7-10, 2000

Bush                63             64             60            63
McCain              16             18             17            18
Forbes               6              7              9             5
Bauer                3              2              2             1
Hatch                4              2              1             2
Keyes                2              4              4             2
  don't know         6              3              7             9


                     January        January
                   13-16, 2000    17-19, 2000

Bush                   61             63
McCain                 22             19
Forbes                  5              6
Bauer                   2              2
Hatch                   1              1
Keyes                   3              1
  don't know            6              8

Democratic nomination race: "Next, I'm going to read you a list of
people who may be running in the Democratic primary for president in
the year 2000. After I read all the names, please tell me which of
those candidates you would be most likely to support for the Democratic
nomination for president" (asked of Democrats and Democratic-leaning

                  September    May 8-10,      October       January
                  6-7, 1997      1998       23-25, 1998    8-10, 1999

Al Gore              47           51            41             47
Jess Jackson         13           12            11             11
Bill Bradley         12            8            15             12
Dick Gephardt         6            7            14             13
John Kerry            5            2             4              5
Bob Kerry             3            3             4             --
Paul Wellstone       --            1             1              1
  don't know         14           16            10             11

                  12-14, 1999

Al Gore               58
Jess Jackson          15
Bill Bradley          21
Dick Gephardt         --
John Kerry            --
Bob Kerry             --
Paul Wellstone        --
  don't know           6

Date                            Gore    Bradley          Know

All Democrats
  April 13-14, 1999              54       34              12
  April 30-May 2, 1999           66       23              11
  May 23-24, 1999                59       30              11
  June 4-5, 1999                 63       28               9
  June 25-27, 1999               64       28               8
  August 16-18, 1999             58       31              11
  September 10-14, 1999          63       30               7
  October 8-10, 1999             51       39              10
  October 21-24, 1999            57       32              11
  November 4-7, 1999             58       33               9
  November 18-21, 1999           54       35              11
Democratic registered voters
  November 18-21, 1999           56       34              10
  December 9-12, 1999            54       39               7
  December 20-21, 1999           52       38              10
  January 7-10, 2000             59       30              11
  January 13-16, 2000            59       30              11
  January 17-19, 2000            60       27              13

Source: All data taken from the Gallup poll.

(a.) Less than 0.5 percent.


How Candidates Emerge as Front-Runners in Contested Nomination Races,

1. Eventual nominee leads the polls throughout the invisible primary
   Bush in 1988
   Bush in 1992
   Gore in 2000

2. Eventual nominee establishes an early lead among the declared
     candidates; his strongest potential rival decides not to run
   Reagan in 1980
   Mondale in 1984
   Dole in 1996
   Bush in 2000

3. Eventual nominee becomes the front-runner late in the invisible
     primary period
   Carter in 1980
   Clinton in 1992

4. Eventual nominee becomes the poll leader during the primary and
     caucus season
   Dukakis in 1988


Matchups of George Bush and Colin Powell as Candidates for the
Republican Presidential Nomination during the 2000 Invisible Primary


Organization            Date             Powell   Bush   Kemp   Quayle

Hart and
  Teeter       December 5-6, 1996 (a)      37      19     20    -- (b)
Harris         January 9-13, 1997 (c)      36       7     14    11
Gallup         April 11-13, 1997 (a)       36      14     14    11
Harris         October 15-19, 1997 (c)     28       9     12     8
  Dynamics     March 10-11, 1998 (c)       15      29     5      5
Harris         July 17-21, 1998 (c)        24      20     5      7

(a.) Asked of Republicans and independents who lean toward the

(b.) Quayle's name was not included in this survey.

(c.) Asked of Republicans only.


Primary Forecasting Model with Iowa and New Hampshire Taken into

Independent Variable                    Coefficient    Standard Error

National poll standings                     0.95          0.21 ***
Pre-election year fund-raising             -0.05          0.06
First-place finish in Iowa                -12.46         12.79
Second-place finish in Iowa                -0.42          5.07
Iowa finish relative to expectations        0.05          0.22
First-place finish in New Hampshire        26.70          5.41 ***
Second-place finish in New Hampshire       17.08          9.47 *
New Hampshire finish relative to
  expectations                             -0.07          0.18
Constant                                    0.31          2.18

[R.sup.2] = .90

Adjusted [R.sup.2] = .87

Standard error of estimate = 7.84

n = 45

* p < .05. *** P < .001.


Effect of the Iowa Caucuses on the Vote in New Hampshire (in

                     Poll of New Hampshire     Poll of New Hampshire
Candidate           Voters Just before Iowa    Voters Just after Iowa

1980 Republicans
  Reagan                      43                         39
  Bush                        24                         43
  Baker                       10                          6
  Anderson                     6                          2
1980 Democrats
  Carter                      35                         54
  Kennedy                     31                         36
  Brown                       11                          5
1984 Democrats
  Hart                        13                         22
  Mondale                     37                         39
  Glenn                       20                         14
  Jackson                     10                          7
1988 Republicans
  Bush                        36                         28
  Dole                        26                         32
  Kemp                        13                         13
  DuPont                       9                         11
  Robertson                    6                         10
1988 Democrats
  Dukakis                     43                         40
  Gephardt                    12                         16
  Simon                       13                         15
  Jackson                      6                          8
1996 Republicans
  Buchanan                    16                         22
  Dole                        24                         25
  Alexander                    5                         18
  Forbes                      20                         13
2000 Republicans
  McCain                      43                         36
  Bush                        34                         37
  Forbes                      13                         15
2000 Democrats
  Gore                        50                         56
  Bradley                     45                         40

                      Final New Hampshire
Candidate               Primary Results

1980 Republicans
  Reagan                      50
  Bush                        23
  Baker                       13
  Anderson                    10
1980 Democrats
  Carter                      47
  Kennedy                     37
  Brown                       10
1984 Democrats
  Hart                        37
  Mondale                     28
  Glenn                       12
  Jackson                      5
1988 Republicans
  Bush                        38
  Dole                        28
  Kemp                        13
  DuPont                      10
  Robertson                    9
1988 Democrats
  Dukakis                     36
  Gephardt                    20
  Simon                       17
  Jackson                      8
1996 Republicans
  Buchanan                    27
  Dole                        26
  Alexander                   23
  Forbes                      12
2000 Republicans
  McCain                      48
  Bush                        30
  Forbes                      13
2000 Democrats
  Gore                        50
  Bradley                     46


Regression Model for the Effects of Iowa on New Hampshire

Independent Variable                   Coefficient    Standard Error

New Hampshire poll just before Iowa       0.99           0.07 ***
First-place finish in Iowa                1.14           2.62
Second-place finish in Iowa               8.27           2.83 **
Constant                                  0.36           1.34

[R.sup.2] = .87

Adjusted [R.sup.2] = .86

Standard error of estimate = 5.88

n = 45

Note: Dependent variable is the percentage of the total vote a
candidate received in the New Hampshire primary.

** p < .01. *** p < .001.

(1.) The model is presented in Mayer (1996), but an all-but-identical version of that article was presented as a paper at the 1994 meeting of the Northeastern Political Science Association. In the past few years, several other scholars have developed models that use slightly different data and extend the model's reach in a number of ways but are otherwise fully compatible with my own work. See, in particular, Adkins and Dowdle (2000) and Steger (2000).

(2.) For an explanation as to why I limited the model to races occurring after 1976, see Mayer (1996, 60-63).

(3.) I use this formulation to make the fund-raising data as comparable as possible across elections. Simply using the raw, unadjusted amount of money collected by each candidate would create several problems. Most obviously, such figures would not take into account the effects of inflation. In addition, fund-raising totals are often affected by circumstances unique to each race. In 1992, for example, the fallout from the Persian Gulf War meant that most Democratic candidates did not enter the race until the fall of 1991, about six to nine months later than they probably would have under normal circumstances (see the data in Hagen and Mayer 2000, 21-26). As a result, Democratic presidential fund-raising in 1991 was, by historical standards, unusually modest. Expressing each candidate's fund-raising as a percentage of the leading money raiser also incorporates the idea that the net effect of a candidate's spending in a competitive race is determined not only by what that one candidate spends but also by what his or her opponents are spending.

(4.) As estimated here, this model almost certainly violates one of the standard regression assumptions, which requires the error terms to be uncorrelated across observations. That is to say, if the model underpredicts the vote for one candidate in a given race (say, Gary Hart in 1984), it will probably overpredict the vote for one or more other candidates in the same race (e.g., Mondale or Glenn). Unfortunately, in large, multicandidate fields (which have tended to be the rule in most recent contested nomination races), it is impossible to specify in advance the correlation between the errors for any two candidates and, therefore, difficult to invoke any of the standard remedies for this sort of problem (e.g., generalized least squares). If the uncorrelated errors assumption has, indeed, been violated, the parameter estimates in Table 4 will be unbiased but inefficient. In other words, the estimated values of the coefficients will, over the long run, be equal to the true values, but the variances of these estimates will be larger than they would be if the assumption had not been violated. In particular, the standard errors in Table 4 will likely be underestimated and thus provide an inaccurate measure of the true confidence interval surrounding each coefficient. In defense of my decision to continue using ordinary least squares to estimate this model, I would make two points. First, nothing of substance is really affected by reasonable assumptions about the true variance of the parameter estimates. Even if the real standard errors are three times the size of those listed in Table 4, the first coefficient is still statistically significant and the second one still is not. Second, after each of the past two elections, I have reestimated the model to incorporate the results of the recently concluded nomination contests, which provides a kind of quasi-experimental test of the model's robustness. As can be seen in Table 4, adding these additional races to the model's database has very little effect on the parameter estimates.

(5.) In the interests of full disclosure, I should point out that in my first academic article, published in 1987, I flirted with the same conclusion (see Mayer 1987, 32-33).

(6.) I focus on these polls, rather than the candidates' fund-raising prowess, for two reasons. First, the results in Table 4 clearly indicate that it is the candidates' standing in the polls rather than their fund-raising that has the greatest influence on nomination outcomes. Second, fund-raising data are not available until the candidates have set up their major campaign committees and filed the necessary papers with the Federal Election Commission, which is generally not until the first or second quarter of the year before the election. By contrast, as the following analysis will show, polling data can generally be found from three or even four years prior to the actual election year.

(7.) My attempt to assemble a fairly complete collection of surveys on each of the ten nomination races considered here has been made dramatically easier by POLL, the online database of polling questions maintained by the Roper Center at the University of Connecticut.

(8.) On February 27, 1987, Baker agreed to become Ronald Reagan's chief of staff--and simultaneously announced that he would not be a candidate for the 1988 presidential nomination. According to the data discussed above, even if Baker had made the race, he would have faced a distinctly uphill battle.

(9.) For an interesting argument along these lines, see Popkin (1991, 13840).

(10.) This point is worth more than passing notice, if only because so many Carter supporters would later blame their resounding loss in the 1980 general election on their bad luck in being unable to extricate the hostages--and on the fact that the first anniversary of the hostage seizure came on the day before the election. This explanation ignores the fact that the U.S. economy was also in very bad shape that fall and, in a time of less international stress, might have played an even larger role in determining people's votes. More to the immediate point, blaming Carter's loss on events in Iran overlooks all the benefits that the president's campaign had received from the hostage crisis earlier in the year.

(11.) That number dropped to five in late September 1987, when Delaware Senator Joseph Biden withdrew from the race amid charges that he had plagiarized one of his best speech lines from a leader of the British Labour Party and lied about several items in his personal history.

(12.) Of the nine surveys on the Democratic race that were conducted in January 1988, Dukakis finished second or third in seven of them. In the remaining two, Paul Simon was in third place and Dukakis in fourth.

(13.) These prediction errors, however, need to be put in context. While the absolute errors produced by most general election forecasting models are far smaller, the effective range of the variable being predicted in those models is also much narrower. The dependent variable for the equations in Table 4, the percentage of the total primary vote won by each candidate, varies between 0.1 and 75.7. By contrast, the dependent variable in most general election models is the percentage of the two-party popular vote won by the candidate of the incumbent president's party. Over the thirteen elections between 1952 and 2000, this variable has never risen above 61.8 and never fallen below 44.6. Thus, a general election forecasting model that always predicted the incumbent party to win 53.2 percent of the vote would be virtually guaranteed never to be wrong by more than 8.6 percentage points. Viewed from this perspective, the failure of these models in 2000, when the median error was 4.7 percentage points, is even more remarkable.

(14.) To give credit where it is due, the possibility of adding Iowa and New Hampshire into the model had not occurred to me until I read Adkins and Dowdle (2000).

(15.) For Democrats in Iowa, there are generally two measures of how well each candidate did: the percentage of caucus attendees who preferred that candidate and the percentage of "delegate equivalents" won by the candidate. I use the latter figures (the former are not available in some years). To measure the candidate's current standing in the national polls, I use, for Iowa, the data from the end of the invisible primary that are discussed in the second section of this article. For New Hampshire, I have in almost every case been able to find a national poll conducted between Iowa and New Hampshire that shows the state of the national race with Iowa taken into account.

(16.) I exclude the 1992 Republican nomination race for the simple but dispositive reason that the 1992 Iowa Republican caucuses did not take a poll of caucus attendees, and there is thus no measure available of how well each candidate did in that event (for further details, see Winebrenner 1998, 197-98). On the Democratic side, the 1992 Iowa caucuses were completely transformed by the fact that a favorite son, U.S. Senator Tom Harkin, was a presidential candidate. The result was that none of the other candidates waged an active campaign there, the media almost completely ignored the state, and Harkin wound up with 76.5 percent of the delegate equivalents. While I am loathe, in a model of this sort, to exclude a race or candidate simply because it is "exceptional," the 1992 results in Iowa are such an obvious outlier--and were known to be that at the time--that any attempt to include them in the model would probably distort the meaning and value of every other coefficient.

(17.) Some might add the case of Bill Clinton in 1992. But Clinton's well-publicized claim to be "the comeback kid" because he had finished second in New Hampshire seems, at best, to have prevented further damage to his campaign; it did not convert a loss into a victory. The evidence from the national polling data, in particular, shows that it was Paul Tsongas who received a huge boost from New Hampshire; the most that can be said of Clinton is that his standing in the national polls declined by only a few percentage points.

(18.) Though Table 9 shows poll standings only for the major candidates in each race, comparable figures are available for all the other candidates who were included in the original model.

(19.) Some might argue for creating two dummy variables to designate the winners of the Iowa caucuses, one for winners who were already the front-runners for their parties' nominations (e.g., Dole in 1996, Bush and Gore in 2000) and a second for winners who were not the prerace front-runners (e.g., Bush in 1980, Gephardt and Dole in 1988). But creating separate dummy variables does not alter any conclusions drawn from Table 10. Both variables are small and statistically insignificant: over the past two decades, neither front-runners nor non-front-runners seem to have derived much advantage in New Hampshire from winning in Iowa.


Adkins, Randall E., and Andrew J. Dowdle. 2000. Break out the mint juleps? Is New Hampshire the "primary" culprit limiting presidential nomination forecasts? American Politics Quarterly 28:251-69.

Bartels, Larry. 1988. Presidential primaries and the dynamics of public choice Princeton, NJ: Princeton University Press.

Brady, Henry E., and Richard Johnston. 1987. What's the primary message: Horse race or issue journalism? In Media and momentum: The New Hampshire primary and nomination politics, edited by Gary R. Orren and Nelson W. Polsby, 127-86. Chatham, NJ: Chatham House

Ceaser, James W. 1982. Reforming the rearms: A critical analysis of the presidential selection process Cambridge, MA: Ballinger.

Hagen, Michael G., and William G. Mayer 2000. The modern politics of presidential selection: How changing the rules really did change the game. In In pursuit of the White House 2000: How we select our presidential nominees, edited by William G. Mayer, 1-55. New York: Chatham House.

Mayer, William G. 1987. The New Hampshire primary: A historical overview. In Media and momentum: The New Hampshire primary and nomination politics, edited by Gary R. Orren and Nelson W. Polsby, 9-41. Chatham, NJ: Chatham House.

--. 1996. Forecasting presidential nominations In In pursuit of the White House: How we select our presidential nominees, edited by William G. Mayer, 44-71 Chatham, NJ: Chatham House

--. 2001. The presidential nominations. In The election of 2000: Reports and interpretations, edited by Gerald M. Pomper, 12-45. New York: Chatham House

--. Forthcoming. The basic dynamics of the contemporary nomination process: An expanded view. In The making of the presidential candidates 2004, edited by William G. Mayer New York: Chatham House.

Popkin, Samuel L. 1991. The reasoning voter: Communication and persuasion in presidential campaigns. Chicago: University of Chicago Press

Steger, Wayne P. 2000. Do primary voters draw from a stacked deck? Presidential nominations in an era of candidate-centered campaigns Presidential Studies Quarterly 30:727-53

Winebrenner, Hugh. 1998. The Iowa precinct caucuses: The making of a media event. 2d ed. Ames: Iowa State University Press.

William G. Mayer is associate professor of political science at Northeastern University. Every four years since 1996, he has edited a series of essays on the presidential nomination process. The next book in the series, The Making of the Presidential Candidates 2004, will be published by Chatham House in 2003.

AUTHOR'S NOTE: The author would like to thank Lois Timms Ferrara, Randall Adkins, Tony Corrado, Bruyce Bassett, Steve Ansolabehere, and Amy Logan for their help with this article.