Copyright 1995 by Political Science Quarterly. Readers may redistribute this article to other individuals for noncommercial use, provided that the text, all html codes, and this notice remain intact and unaltered in any way. This article may not be resold, reprinted, or redistributed for compensation of any kind without prior written permission from Political Science Quarterly. If you have any questions about permissions, please contact The Academy of Political Science, 475 Riverside Drive, Suite 1274, New York, NY 10115-1274, e-mail: Preferred Citation: Paul R. Abramson, John H. Aldrich, Phil Paolino, and David W. Rohde, "Third-Party and Independent Candidates in American Politics: Wallace, Anderson, and Perot," Political Science Quarterly 110, 3 (Fall 1995 [psabra.html]). Acknoweledgments

Third-Party and Independent Candidates in American Politics: Wallace, Anderson, and Perot

PAUL R. ABRAMSON is professor of political science at Michigan State University.
JOHN H. ALDRICH is professor of political science and chairman of the department at Duke University.
PHIL PAOLINO is assistant professor of government at the University of Texas at Austin.
DAVID W. ROHDE is the University Distinguished Professor in the department of political science at Michigan State University.

Between 1852 and 1992 the United States held thirty-six presidential elections. The Democrats won fifteen, the Republicans won the other twenty-one. This duopoly has been threatened from time to time, and Perot's 1992 challenge was remarkably successful in winning popular votes. But no third-party or independent candidate has come close to winning the presidency, largely because the electoral rules in the United States create barriers that third parties and independent candidates have been unable to surmount. [1]

We review U.S. elections since 1832, for these barriers to new political parties were in place even then. The Republicans displaced the Whigs between 1854 and 1860, despite these structural barriers, by changing the issue agenda over slavery. [2] Fortunately for the United States, but unfortunately for third parties, since the Civil War no issue as divisive as slavery has restructured American politics. [3]

We focus on three recent efforts to challenge the two major parties -- George C. Wallace's American Independent Party candidacy in 1968, John B. Anderson's independent candidacy in 1980, and H. Ross Perot's independent candidacy in 1992. Several earlier third-party candidacies are as interesting, but for these recent contests we have high quality survey research data that allow us to test propositions about individual-level behavior. Lastly, we discuss presidential elections in the French Fifth Republic, showing how different electoral rules, along with a different political culture, lead to substantially different outcomes.


The basic rules for electing the president are spelled out in Article II, Section 1 of the Constitution, as modified by the Twelfth Amendment. The most crucial feature of these rules is that the winning candidate must have a majority of the electoral votes. Under the original provisions for electing the president, each elector cast two votes, both of which were for president, and one of these votes had to be for a candidate from outside the elector's state. If no candidate won a majority of the electoral votes, the U.S. House of Representatives -- voting by state delegations -- chose among the five candidates with the largest number of electoral votes. If two candidates both had a majority, and if they both had the same number, the House chose between these two majority-winners.

The Twelfth Amendment, ratified shortly before the 1804 election, prevents the outcome that occurred in 1800, when Thomas Jefferson and Aaron Burr both won a majority of the electoral vote and both won the same number of votes. This tie forced the Federalist-controlled House to choose between Jefferson and Burr, both Democratic-Republicans. Under the Twelfth Amendment, each elector casts a separate vote for president and for vice-president (one of whom must be from outside the elector's state). If no candidate receives a majority of the electoral vote, the House now chooses among the top three candidates -- once again voting by state delegations. However, ties among two majority-winners could have been prevented in 1800 or after without constitutional change, for parties could have instructed one or more presidential electors to cast a strategic vote to assure that the party's presidential candidate had more votes than the party's vice presidential candidate. In fact, the Democratic-Republicans in 1800 were aware of the dangers of a tie, but believed that their party would win in Rhode Island and that one of the state's four electors would withhold his vote from Burr. [4] The winning party has had electoral votes to spare in every presidential election except 1876, when Rutherfold B. Hayes won by a single electoral vote, and as the Republicans held the Senate (which could have chosen the vice president between the top two candidates, with each senator casting an individual vote) they might have still won the vice presidency.

Reducing the number of eligible candidates from five to three may have affected the outcome of the 1824 election, the only other election decided by the House. Although there was only one political party, the nominee of the Democratic-Republican Congressional Caucus, William Crawford, was challenged, leading to a four-way contest among Crawford, Andrew Jackson, John Quincy Adams, and Henry Clay. Jackson won ninety-nine electoral votes, Adams gained eighty-four, Crawford received forty-one, and Clay won thirty-seven. Clay was thus ineligible for election. As Speaker of the House, he had considerable influence, but could not use it to win his own election. Instead, Clay's support proved crucial in securing Adams's election, and his efforts were apparently rewarded when Adams selected him to be secretary of state.

For the next three years, Jackson campaigned for the presidency claiming that a "corrupt bargain" between Adams and Clay had denied him election. These charges helped undermine the Adams presidency. Although presidential election by the House is a legitimate way to gain office, few candidates begin a campaign with election by the House as their goal.

The preferred strategy for gaining the White House is to win a majority of the electoral votes and to be elected without incurring any additional obligations. But some parties and candidates have followed alternative strategies. In 1836, the Whigs ran three presidential candidates against Martin Van Buren, each candidate appealing to a different region. It is not clear what their next step would have been had this strategy led to denying Van Buren a majority of the electoral vote. The strategy of running different presidential candidates in different regions of the country was never tried again.

Wallace was on the ballot in every state in 1968, but he had no realistic chance of winning the 270 electoral votes needed to be elected. Rather, his declared goal was to win enough electoral votes to deny either Richard M. Nixon or Hubert H. Humphrey a majority. All of Wallace's electors were technically unpledged. In principle, Wallace might have bargained with Nixon for his electors, for Nixon wanted to avoid having the House of Representatives determine the outcome. It is difficult to imagine just what sort of bargain Nixon could have made with Wallace that would not have undermined his presidency. As Nixon won a clear majority of the electoral vote, Wallace had no bargaining power.

Clearly, the House procedure is not seen as a runoff among the three top candidates. From 1828 through 1992 a single presidential candidate has gained a majority of the electoral vote. But even if a third-party or independent candidate did become eligible for election by the House, a legislative body dominated by Democrats and Republicans would be unlikely to turn to an independent. On 16 July 1992, Perot announced that he would not run for president, because he feared that his candidacy would throw the election to the House and that this would be disruptive. Ed Rollins, who had resigned from the Perot campaign days before, claimed that there was a strategy to deal with the eventuality of a House election, but he provided no hint about what that strategy was. The majority-win election rule, then, appears to have promoted stability in presidential elections by encouraging the formation of two major parties to prevent the potentially harmful consequences arising from the election of a president unable to claim a mandate from the majority of the electorate.

It is not just the need for an electoral vote majority that thwarts third-party candidates. The way that states choose their electors also undermines their chances. The Constitution allows each state to select its electors in a manner determined by the state's legislature. Since the Civil War, every state has chosen its electors by popular vote. Every state except Maine and Nebraska chooses its electors by voting for a statewide slate. Maine and Nebraska choose two of their electors by a statewide vote, and then select one elector from each of the state's congressional districts. All of these elections are based upon plurality-vote-win rules. Even states that require runoffs for other offices require no runoffs for choosing presidential electors. In principle, states could allocate their electors by proportional representation, although no state has ever done so. The plurality-vote system chosen by most states creates an additional obstacle to third-party and independent candidacies.


The U.S. plurality-vote system can be seen as a confirmation of Duverger's law. According to Maurice Duverger, "le scrutin majoritaire a un seul tour tend au dualisme des partis." [5] Barbara North and Robert North translated this as: "the simple-majority single-ballot system favours the two party system." [6] Clearly, by

"majoritaire" Duverger meant what we would call a plurality, or what North and North call a "simple majority." By "un seul tour" Duverger meant a system without runoff elections, which for all intents and purposes is an outcome of the plurality-vote decision rule. Duverger explicitly saw the U.S. system for electing presidents as confirming his law. As he writes, "The American procedure corresponds to the usual machinery of the simple-majority [plurality] single-ballot system. The absence of a second ballot and of further polls, particularly in the presidential election, constitutes in fact one of the historical reasons for the emergence and the maintenance of the two-party system." [7]

According to Duverger, his law applies for two reasons. First, the plurality-vote-win system has a "mechanical" effect. Third parties or independent candidates may win many votes nationwide, but gain a plurality of the vote in very few electoral units. Second, the plurality-vote-win system has a "psychological" effect. Some voters who prefer a candidate or party that they think cannot win will cast a vote for their first choice among the major-party candidates. This behavior is often called "sophisticated" or "strategic" voting, and is widely known as "tactical" voting in Britain, another democracy with a plurality-vote-win system. William H. Riker defines strategic voting as "voting contrary to one's immediate tastes in order to obtain an advantage in the long run." [8]

The mechanical effect of the U.S. electoral system is easy to demonstrate. Between 1832 and 1992 a total of thirteen third-party or independent presidential candidates received either 5 percent of the popular vote or 5 percent of the electoral vote. [9] Ten of these thirteen candidates received a larger share of the popular vote than they received of the electoral vote. Perot won 18.9 percent of the popular vote without gaining a single electoral vote, because his support was distributed relatively evenly across the country. [10] He finished third in forty-eight states as well as the District of Columbia. He finished second in two states, but in neither did he come close to a plurality, losing by 8 points to Bill Clinton in Maine and by 16 points to George Bush in Utah. Perot came closest to winning a vote in Maine's second congressional district, but even there he trailed Clinton by 5 points. Anderson received a far smaller share of the popular vote (6.6 percent), finishing third in forty-nine states and the District of Columbia and fourth behind Ed Clark (the Libertarian candidate) in Alaska. Wallace did have a regional base, but he too was underrepresented in the electoral college. He gained 13.5 percent of the popular vote, but only 8.5 percent of the electoral vote (including one faithless elector from North Carolina).

Two third-party candidates were overrepresented in the electoral college. In 1860, John C. Breckinridge, the Southern Democrat, won 18.1 percent of the popular vote, but gained 23.8 percent of the electoral vote (including eight electoral votes from South Carolina, where the electors were chosen by the state legislature). Sixty-seven percent of his popular vote came from the slave states, which had somewhat lower turnout among their politically eligible population than states outside the South. Moreover, under the three-fifths rule, the slave population contributed to a state's representation in the U. S. House of Representatives, and, therefore, to representation in the electoral college. In the same election, John Bell, the candidate of the Constitutional Union Party, the last remnant of the Whigs, was neither under nor overrepresented, receiving 12.6 percent of the popular vote and 12.9 percent of the electoral vote.

In 1948, J. Strom Thurmond, the States' Rights Democrat, won only 2.4 percent of the popular vote, but he won 7.3 percent of the electoral votes. Thurmond won 55 percent of his popular vote in the four states that he carried (Alabama, Louisiana, Mississippi, and South Carolina), all of which had low turnout, partly because the large African-American population was disfranchised. He received no popular votes at all in thirty-one of the forty-eight states.

The psychological effects of the plurality-vote system are more difficult to document. The argument that voters should avoid wasting their votes on a candidate who has little chance of winning is well known to politicians and has been used frequently by them. For example, in the 1932 presidential elections in the Weimar Republic, the three major candidates were Paul von Hindenberg, Adolf Hitler, and the Communist candidate, Ernst Thaelmann. Campaign posters for Hindenberg did not emphasize his merits. Instead, Hindenberg posters in working-class neighborhoods warned that "a vote for Thaelmann is a vote for Hitler," whereas posters aimed at conservative voters warned that "a vote for Hitler is a vote for Thaelmann." [11] In that same year, supporters of Franklin D. Roosevelt urged Americans not to waste their vote on Norman Thomas, the Socialist candidate. Thirty-six years later, Thurmond purportedly helped hold South Carolina for Nixon by arguing that "a vote for Wallace is a vote for Humphrey." [12] And in 1992, Perot countered the wasted vote argument with the slogan, "Don't waste your vote on politics as usual."

That politicians make (or refute) the wasted vote argument does not mean that voters act upon it. As the chances that an individual voter will break or create a tie in a presidential election are remote, voters may decide to vote for their first choice, consistent with the logic of John A. Ferejohn and Morris P. Fiorina's minimax-regret model. [13] Indeed, their model predicts that rational voters never engage in strategic voting.

To demonstrate the presence or absence of sophisticated voting one needs survey data, so we cannot determine whether Hindenberg benefited from his campaign posters. [14] And while American third-party and independent presidential candidates from William Wirt in 1832 through H. Ross Perot in 1992 may have been hurt by the wasted vote argument, we are restricted to examining the psychological effects postulated by Duverger for voters in the 1968, 1980, and 1992 elections.

Before attempting to test for these effects, we want to underscore that even if Wallace, Anderson, and Perot all lost some votes as the result of strategic voting, it seems highly unlikely that they lost enough votes to cost them the election. In fact, it seems likely that all three would have finished second in head-to-head contests against either of the major party candidates they faced. Moreover, it seems likely that Richard Nixon, Ronald Reagan, and Bill Clinton were all Condorcet winners. The Marquis de Condorcet (1743-1794) argued that if there is an outcome that would be preferred by a majority over any other alternative, that outcome should be selected. A candidate who would defeat any of his or her opponents in a head-to-head contest meets this criterion. If there is a three-person contest in which candidate A would defeat candidate B in a head-to-head contest and would also defeat candidate C in a head-to-head contest, candidate A meets the Condorcet criterion. If candidate A wins the actual election, he or she is the Condorcet winner. But since in most elections voters can state a single preference for each office, a candidate can meet the Condorcet criterion but lose the actual election. And there may be no outcome that meets the Condorcet criterion. For example, candidate A might defeat candidate B in a head-to-head contest, candidate B might defeat candidate C, and candidate C might defeat candidate A. [15]

To support the conclusion that Nixon, Reagan, and Clinton were Condorcet winners, we rely upon the "feeling" thermometers in which respondents are asked to evaluate candidates. Respondents who have "Very warm or favorable" feelings toward candidates can score them at 100, whereas those who have "Very cold or unfavorable" feelings can score them at 0. A score of 50 indicates "No feeling at all." These "thermometer" measures can be used to create a mock election, since voters are very likely to vote for the candidate whom they score highest on this measure. The thermometer measure was used to evaluate the candidates and other political leaders in the 1968 National Election Studies (NES) postelection survey. The 1980 and 1992 NES measured feelings toward the candidates in both the pre- and postelection surveys.

We begin our analysis by analyzing the head-to-head comparisons among all three candidates in all three elections, and these comparisons are presented in Table 1. As we do not know what turnout would have been if voters could make a series of head-to-head comparisons, we begin by examining rankings

among all respondents. In Part B, we compare rankings among respondents who reported voting for president in the postelection interview. The 1980 NES survey included a vote validation study that checked registration and voting records to determine whether or not respondents actually voted, and in Part C we compare rankings among validated voters.

As Wallace was far more conservative than either Nixon or Humphrey (and thus unattractive to moderates, let alone liberals), it seems highly unlikely that he met the Condorcet criterion. He did not. As the table shows, Wallace would have been badly defeated in a head-to-head contest against either Nixon or Humphrey. When we restrict the analysis to self-reported voters (Part B), Wallace loses by a slightly larger margin. In addition, nearly half the total sample rank Nixon ahead of Humphrey (Part A). Among respondents who report voting, a slight majority rank Nixon ahead of Humphrey (Part B). Nixon, therefore, emerges as the likely Condorcet winner.

A centrist candidate like Anderson is more likely to meet the Condorcet criterion yet lose, because voters discount his chances. But the evidence suggests that Anderson did not meet the Condorcet criterion. Head-to-head thermometer comparisons suggest that he would have finished a distant second to either Ronald Reagan or Jimmy Carter. Ties are reduced when we restrict the analysis to self-reported voters, but the overall ratio of major party supporters to Anderson supporters is not affected (Part B). When we restrict the analysis to validated voters (Part C), the hypothetical race between Carter and Anderson tightens considerably. However, regardless of whether we examine all respondents, self-reported voters, or validated voters, Reagan decisively defeats Anderson in the postelection survey. And Reagan clearly leads Anderson in the pre-election interview.

Table 1 also reminds us of the changing dynamics of the 1980 contest. In the 1980 preelection survey, Carter appears to be the likely Condorcet winner, at least among the entire sample. But when the analysis is restricted to reported voters (Part B), Carter edges out Reagan by only a single percentage point, and when it is restricted to validated voters (Part C) Reagan ekes out a one point margin over Carter. There thus appears to have been no clear Condorcet winner before the election. In the postelection survey, Reagan emerges as the likely Condorcet winner. Almost half the total sample rank Reagan ahead of Carter (Part A), and among both self-reported voters (Part B) and validated voters (Part C), a majority do.

The Perot candidacy is more interesting, because Perot won a far larger percentage of the popular vote than Wallace or Anderson, and also because some critics have argued that he would have won if pre-election polls had not predicted his defeat. Gordon S. Black and Benjamin D. Black reach this conclusion based upon an analysis of the exit polls conducted for the four television networks by Voter Research and Surveys. [16] At the end of the exit poll ballot, respondents were asked to check statements that applied to them. One statement read, "Would you have voted for Perot if you thought he had a chance to win?" (About 4,000 respondents were asked this question.) Black and Black report that 36 percent of the respondents said they would have voted for Perot, whereas another 4 percent who did not answer this question had voted for him. As Larry Hugick points out, it is inappropriate to conclude that many of these Clinton or Bush voters would actually have voted for Perot had his chances been viewed as better. [17] It seems more reasonable to infer that many voted for Clinton or for Bush with little enthusiasm, and answered yes merely to voice their frustration. Moreover, as Hugick points out, Gallup polls found that 5 percent of Perot's supporters said they would not vote for him if they thought he could win.

If Gordon and Benjamin Black are correct, Perot should emerge as a winner in a set of head-to-head comparisons using the feeling thermometers. Although voters may be reluctant to "waste" their votes, there is no reason for them to waste their opportunity to score Perot highly. Granted, responses to the feeling thermometer may be affected by how the respondent has voted or (in the case of the pre-election survey) by how he or she plans to vote. To the extent that the respondents' assessments of the candidates' chances affect their ratings on this measure, the measure does not provide an evaluation that is independent of the vote itself. Despite this qualification, the head-to-head rankings in Table 1 strongly suggest that Perot would have lost a two-candidate contest to either Clinton or Bush. Clinton emerges as the likely Condorcet winner in both the pre-election and postelection survey. Steven J. Brains and Samuel Merrill III, who use somewhat different procedures based upon these same data, also conclude that Clinton was the likely Condorcet winner. They also report that Clinton would have been elected if approval voting had been employed. [18]

Although it seems very unlikely that Wallace, Anderson, or Perot lost because of sophisticated voting, all three were probably hurt somewhat by the wasted vote argument. Moreover, even if a relatively small percentage of voters who preferred Wallace voted for Nixon, the net effect of their behavior could have undermined his strategy of preventing either Nixon or Humphrey from winning an electoral vote majority. If Wallace had carried the three states in which he came in second to Nixon (North Carolina, South Carolina, and Tennessee), Nixon could have won only 270 electoral votes, and a faithless elector would have deprived him of a majority. As we will show, there is clearly some evidence consistent with the logic of sophisticated voting.

Direct tests of strategic voting require that, in addition to feeling thermometer evaluations of the candidates, one must also measure the respondent's assessment of each candidate's chances of winning the election. Measures of these voter assessments proved valuable in demonstrating the presence of sophisticated voting in the 1988 presidential nominating contest. [19] But similar measures were not included in the 1992 NES, nor were they used in the 1968 or 1980 surveys. All the same, the candidate thermometer scores provide important indirect evidence about problems facing third-party and independent candidates.

Table 2 reports voting choices in the 1968, 1980, and 1992 elections controlling by the respondents' three-way ranking of the candidates on the postelection feeling thermometers. In all three elections, those who ranked a major party nominee highest overwhelmingly supported that candidate. Bush suffered a larger percentage of defections of any of the major party candidates, but even he won 93 percent of the vote among respondents who ranked him highest. Wallace won the votes of 84 percent of the voters who ranked him highest, but he was less successful in holding his supporters than the major party candidates. [20]

Anderson, by contrast, won the votes of only 57 percent of the voters who ranked him highest. We find similar results among Anderson supporters who were found by the vote validation study to have actually voted. Among validated voters who ranked Anderson first (N = 91), 56 percent voted for him, 18 percent voted for Reagan, and 26 percent voted for Carter. Among validated voters who ranked Reagan first (N = 323), 97 percent voted for him, 2 percent voted for Carter, and 1 percent voted for Anderson. And among validated voters who ranked Carter first (N = 183) 97 percent voted for him, 3 percent voted for Reagan, and none voted for Anderson. Among voters ranking Perot highest, 79 percent voted for him, while 21 percent defected. This finding supports Gordon and Benjamin Black's claim that Perot was hurt by strategic voting. But it seems implausible that he was hurt enough to cost him the election.

One alternative explanation for the findings in Table 2 is basically methodological. Assuming that voters are reporting their behavior accurately, we have a measure of their behavior on election day. We have a measure of their feelings toward the candidates in a survey conducted about a month or more after the election. Feelings toward the candidates can change after the election, and the rankings of third-party candidates may be more changeable than those of the major party candidates. We cannot test this possibility for 1968, but we can compare pre and postelection evaluations in 1980 and 1992.

We have elsewhere presented the joint distribution of individual-level rankings before and after each of these elections. [21] The results for 1980 reveal a substantial individual-level shift away from Carter and toward Reagan, reflecting the aggregate-level shift toward Reagan documented by Table 1. Even so, among respondents who ranked Carter first before the election, 71 percent ranked him first after the election. Among those who ranked Reagan first, 88 percent did so after he was elected. But among respondents who ranked Anderson first before the election, only 53 percent ranked him first in the postelection survey. Relatively few respondents switched to Carter. Among respondents who ranked Carter first after his defeat, 84 percent were Carter supporters before the election. By contrast, only 70 percent of those who ranked Reagan first after his victory had ranked him first before the election. Two-thirds of the respondents who ranked Anderson first after the election had ranked him first before the election.

Our analyses also reveal that evaluations of Perot were less stable than those of Bush and Clinton. Among respondents who ranked Bush first before the election, 74 percent did so after his defeat, while among those who ranked Clinton first, 84 percent did so after his election. But among those who ranked Perot first before the election, only 55 percent ranked him first in the postelection interview. Among respondents who ranked Bush first after his defeat, 83 percent had ranked him first before the election, while among those who had ranked Clinton first, 79 percent had done so before he was elected. Among respondents who ranked Perot first after the election, only 56 percent had ranked him first before the contest. (Perot was not a candidate for the entire period during which the pre-election interviews were conducted. But overall levels of consistent candidate support are only slightly higher among respondents interviewed in October and November than among those interviewed in September.)

Even though we can measure respondents' feelings about the candidates before and after the election, we cannot use the NES surveys to measure their evaluations on election day itself. All the same, respondents who ranked a candidate first both before and after an election were more likely to have ranked him first on election day itself than respondents who switched their rankings. In Table 3 we present the reported vote choice among respondents according to both their pre and postelection rankings. Among candidates who change their rankings between the pre- and postelection interviews, defection rates are substantially higher than those who do not. As our results make clear, the postelection rankings provide a better predictor of voting choices than the pre-election rankings. Among voters who rank a candidate highest in both the pre- and postelection interview, the vast majority voted for that candidate. Even so, the percentage of consistent Anderson supporters who voted for him is less than the percentage of consistent major party supporters who voted for their first choice. All 275 voters who ranked Reagan first both before and after the election reported voting for him, while 97 percent of the respondents who ranked Carter first both before and after the election voted for Carter. On the other hand, three out of ten respondents who ranked Anderson first both before and after the election voted for either Reagan or Carter. Moreover, similar results obtain if we restrict these analyses to validated voters. Among validated voters who ranked Anderson first both before and after the election (N = 60), 70 percent voted for Anderson, 10 percent voted for Reagan, and 20 percent voted for Carter. By contrast, all 220 validated voters who ranked Reagan first both before and after the election reported voting for Reagan. Among the 143 validated voters who ranked Carter first in both the pre- and postelection interview, 97 percent voted for Carter, 3 percent voted for Reagan, and none voted for Anderson.

Perot was more successful in winning the votes of his consistent supporters than Anderson was in holding his. Even so, he was less successful than the major party candidate with consistent support. Among respondents who ranked Clinton highest both before and after his election, 98 percent reported voting for him, while 96 percent of the respondents who ranked Bush highest before and after his defeat voted for his reelection. But even among voters who ranked Perot first both before and after the election, one out of eight reports voting for one of the major party candidates.

These results strongly suggest that at least some voters responded to the wasted vote argument. Even so, the mechanical effects of the plurality-vote-win system had more of an impact than its psychological effects. In both 1968 and 1992, the system's mechanical effects were strong enough to transform a plurality-vote winner in the popular vote into a majority-vote winner in the electoral college. And while Reagan won a popular vote majority of 50.7 percent in 1980, the mechanical effect transformed that vote into a huge 90.9 percent electoral vote majority. Moreover, as the apparent Condorcet winner won all three elections, it would be difficult to argue that the electoral system led to a pernicious result. Still, the psychological effect appears to have some genuine impact. Riker argues that elites also act on the psychological effect (as noted above in several ways, including explicit campaign appeals). [22] That Wallace held nearly the same poll standing in September as Humphrey, but then lost support rapidly while Humphrey's increased, for example, suggests that the institutional base of a major party made elite manipulation more effective for the major parties, and thus reinforced the status-quo, two-party system.


Although presidential systems are widely used in Latin America and although many newly created democracies have presidential systems, Arend Lijphart classifies only three Western democracies as presidential: the United States, Finland, and France. [23] As Finland has had only one popular vote election for president (held in 1994), it can provide little evidence about the effects of electoral laws upon presidential voting. However, the six popular vote elections of the French Fifth Republic provide insights about the effects of direct popular vote election.

Differences between the French and the U.S. party systems cannot be explained by institutional differences alone. As Philip E. Converse and George Dupeux pointed out, differences between French and American political socialization processes lead to weaker partisan loyalties in France than in the United States. [24] The relative importance of ideology and partisanship in French elections is a subject of debate, [25] but ideological factors are more important in French elections than in most U.S. elections. And there are great differences between the American republic that has been politically stable since 1865 and the French republic in which civilian leaders abandoned power in the face of a threatened military coup as recently as 1958.

But institutional differences are also important in contributing to the French multiparty system, as Converse and Dupeux recognized. [26] As Joseph A. Schlesinger and Mildred Schlesinger demonstrate, the run-off system in legislative elections has not only shaped electoral outcomes but the strategies pursued by the political parties. [27]

Nine of the ten National Assembly elections of the Fifth Republic have employed a run-off system, and French presidential elections employ a run-off system as well. To be elected on the first ballot, a candidate needs a majority of the votes cast, and if no candidate receives a majority a run-off is held among the two top candidates. By contrast, to win in the first round of a legislative election, one needs a majority of the votes cast in a single-member district; any candidate with 12.5 percent of the registered voters is eligible to compete in the run-off. There have been only seven presidential elections in the Fifth Republic, and Charles de Gaulle's election in 1958 was by an electoral college of elected officials. In 1962, the Fifth Republic adopted direct presidential elections in a referendum approved by only 62 percent of the voters, and by only 46 percent of those registered to vote. The first direct election was in 1965, when de Gaulle sought a second seven-year term. Even de Gaulle could not win a majority of the vote and was forced into a run-off. After winning the runoff, de Gaulle became France's second directly elected president. The first was Louis Napoleon Bonaparte, who was elected president of the Second Republic in 1848.

French presidential elections have many candidates, partly because it is easy to get on the ballot. As France is a unitary system, one qualifies for placement on a national ballot, and does not need to qualify separately in fifty-one electoral units. After the 1974 election, in which a record twelve candidates qualified for the presidential ballot, the requirements for gaining ballot access were tightened somewhat. A candidate must receive signatures from a total of 500 elected officials, and those officials must come from at least thirty of the ninety-six departments. As a further requirement, no more than 10 percent of the signatures can come from any one department. But there are many elected officials in France, [28] and these requirements can be met by relatively obscure candidates. Ten presidential candidates qualified for the ballot in 1981, and nine qualified in 1988 and 1995. Political parties have little control over access to the ballot, but parties do officially support candidates.

The French electoral system may appear to be ideally designed for independent candidates, but no one has ever been elected without party support. [29] De Gaulle claimed to be above the parties, but a Gaullist party (the Union for the New Republic) supported him. In 1969, Georges Pompidou was the leader of the Union of Democrats for the Republic (the new name of the Gaullists), and in 1974 Valery Giscard d'Estaing was president of the Independent Republicans. In 1981 Francois Mitterrand was first secretary of the Socialist party. Jacques Chirac, elected in his third presidential race in 1995, reorganized the Gaullists in 1976. He was president of the Rally for the Republic (the current name of the Gaullists) when he announced his candidacy in November 1994.

Partly because there are so many candidates, a run-off becomes very likely and is widely anticipated by the electorate. In fact, several candidates have qualified for the run-off with a low percentage of the popular vote. In 1981, Mitterrand won only a fourth of the first ballot votes, but overtook Giscard d'Estaing in the run-off. In both 1988 and 1995, Chirac qualified for the run-off with only a fifth of the vote. Chirac was soundly defeated by Mitterrand in the run-off in 1988, but he overtook Lionel Jospin, the Socialist candidate, seven years later.

In addition, a relatively large number of third-or-worse place finishers receive a substantial share of the vote. The contrast with the United States is striking. In the forty-one U.S. presidential elections between 1832 and 1992, only thirteen third-or-worse place finishers received 5 percent or more of the popular vote. In the six French presidential elections between 1965 and 1995, fourteen third-or-worse place finishers won 5 percent or more of the first round vote.

Strategic voting may occur in first-round presidential elections. A voter may recognize that his or her first choice has little chance of qualifying for the run-off, but may have a less favored candidate who has a good chance of qualifying. On the other hand, recognizing that a run-off will be held, voters may choose to vote sincerely on the first ballot, which may account for the large percentage of votes for candidates who have very little chance of winning.

The French experience raises questions that proponents of direct presidential elections in the United States should consider. How would the presidential nomination process work? Would a run-off be required if no candidate gained an absolute majority of the popular vote? Would a lower percentage be sufficient, and, if so, what would that percentage be? Does direct election make a "wrong winner" less likely than the current U.S. system? [30 ] We will not answer these questions here, but one conclusion seems warranted. Directly electing the president by popular vote would threaten the Republican and Democratic presidential duopoly.

As of today, major electoral change seems unlikely, although some states may adopt the district system employed by Maine and Nebraska. But the plurality-vote-win system does not guarantee major-party dominance, as the Progressive Conservative party in Canada learned in October 1993. Ultimately, no electoral system can protect major political parties from the electorate. But the U. S. system clearly offers considerable protection for the major parties. The Wallace, Anderson, and Perot candidacies illustrate that both the mechanical effects of the electoral system, as well as the psychological effects of the system on individual voters, create serious problems for candidates who challenge two-party dominance.


[1] [2] [3]


Comparative Thermometer Ratings of the Candidates in 1968, 1980, and 1992: Head-to-Head comparisons in percentages.

    Republican            Republican           Democrat
    versus                versus               versus
    Democrat              Independent          Independent
Part A. All Respondents
1968 Postelection Survey: Candidate Rated First*
    Nixon       48        Nixon      78        Humphrey   72
    Tie         13        Tie         7        Tie         7
    Humphrey    39        Wallace    16        Wallace    20
    Total      100        Total     101        Total      99
     (N =    1,302)       (N =    1,290)       (N =    1,293)
1980 Pre-election Survey:   Candidate Rated First
    Reagan      41        Reagan     50        Carter     50
    Tie         10        Tie        16        Tie        15
    Carter      48        Anderson   34        Anderson   35
    Total       99        Total     100        Total     100
    (N =     1,521)       (N =    1,378)       (N =    1,384)
1980 Pre-election Survey:   Candidate Rated First
    Reagan      48        Reagan     59        Carter     49
    Tie         13        Tie        16        Tie        16
    Carter      39        Anderson   25        Anderson   35
    Total       100       Total     100        Total     100
    (N =      1,370)      (N =    1,273)       (N =    1,275)
1992 Preelection Survey: Candidate Rated First
    Bush        40        Bush       52        Clinton   56
    Tie         12        Tie        12        Tie       14
    Clinton     49        Perot      35        Perot     30
    Total      101        Total      99        Total     100
    (N =     2,414)**     (N =    2,295)**     (N =    2,293)
1992 Postelection Survey: Candidate Rated First
    Bush        36        Bush       47        Clinton    57
    Tie         11        Tie        14        Tie        15
    Clinton     53        Perot      39        Perot      28
    Total      100        Total      100       Total     100
       (N  = 2,222)       (N =     2,178)**    (N =    2,175)
Part B. Respondents who Reported Voting for President
1968 Postelection Survey: Candidate Rated First*
     Nixon      51         Nixon      80        Humphrey   74
     Tie        10         Tie         5        Tie         6
     Humphrey   39         Wallace    15        Wallace    20
     Total     100         Total     100        Total     100
     (N =      975)        (N =      969)       (N =      970) 

1980 Pre-election Survey: Candidate Rated First
     Reagan     45         Reagan     51        Carter     47
     Tie         9         Tie        15        Tie        14
     Carter     46         Anderson   34        Anderson   39
     Total     100         Total     100        Total     100
      (N =     947)          (N =    886)          (N =   889)

1980 Postelection Survey: Candidate Rated First
     Reagan     52         Reagan     61        Carter     46
     Tie        11         Tie        14        Tie        15
     Carter     37         Anderson   26        Anderson   39
     Total     100         Total     101        Total     100
       (N =    980)         (N =     940)       (N =      937)

1992 Pre-election Survey: Candidate Rated First
     Bush       41         Bush       52         Clinton   57
     Tie         9         Tie        12         Tie       13
     Clinton    50         Perot      36         Perot     30
     Total     100         Total     100         Total    100
     (N =    1,660)**      (N =     1,615)**      (N =   1,614)** 
1992 Postelection Survey: Candidate Rated First
     Bush       37         Bush       47          Clinton  57
     Tie         9         Tie        13          Tie      14
     Clinton    54         Perot      40          Perot    29
     Total     100         Total     100          Total   100
       (N =  1,665)**      (N =     1,651)**      (N    1,647)**
Part C.  Validated Voters (1980 Only) ***
1980 Pre-election Survey: Candidate Rated First
     Reagan     46         Reagan     52           Carter    46
     Tie         8         Tie        15           Tie       14
     Carter     45         Anderson   33           Anderson  40
     Total      99         Total     100           Total    100
       (N = 814)              (N = 772)             (N =   775)
1980 Postelection Survey: Candidate Rated First
     Reagan     53         Reagan     61            Carter   44
     Tie        11         Tie        14            Tie      16
     Carter     36         Anderson   25            Anderson 40
     Total      100        Total     100            Total   100
          (N = 767)             (N = 736)            (N = 734)

Source:      The results are based upon our analyses of the
National Election Studies.
* Candidate thermometers were not used in the 1968 preelection
** The numbers for 1992 are weighted using the time-series
*** There was no vote validation study in 1968 or 1992.  Our
procedures for classifying respondents as validated voters or
nonvoters follow those described in Paul R. Abramson and William
Claggett, "Race-Related Differences in Self-Reported and
Validated Turnout," Journal of Politics 46 (August 1984): 719-38,
except that we classified four validated voters who said they did
not vote for president as nonvoters.


Candidate Thermometer Rankings of the Candidates and the Vote: 1968, 1980, and 1992 (in percentages).

          Presidential Vote in 1968
First Place in
                     Nixon  Humphrey  Wallace   Total   (Number)
Nixon                 96        2        2      100     (418)
Humphrey               2       97        1      100     (353)
Wallace               15        1       84      100     (107)
Nixon-Humphrey tie    39       60        2      101    (67)
Nixon-Wallace tie    [4]       [0]      [5]             (9)
Wallace-Humphrey tie [0]       [3]      [0]             (3)
Three-way tie        [3]       [4]      [2]             (9)

           Presidential Vote in 1980 
First Place in
Thermometer Ratings*  
                     Reagan   Carter  Anderson  Total (Number)
Reagan                97        2        1      100    (409)
Carter                 3        7        0      100    (253)
Anderson              18       25       57      100    (111)

Reagan-Carter tie     40       60        0      100    (41)
Reagan-Anderson tie   88        3        9      100    (34)
Carter-Anderson tie    7       67       26      100    (27)
Three-way tie         24       64       12      100    (25)

        Presidential Vote in 1992
First Place in
Ratings*             Bush    Clinton   Perot    Total  (Number)  
Bush                 93         2        5      100    (481)**
Clinton               2        95        3      100    (674)**
Perot                 9        12       79      100    (233)**

Bush-Clinton tie     50        42        8      100     (72)**
Bush-Perot tie       45         4       51      100     (51)**
Clinton-Perot tie     4        60       36      100     (80)**
Three-way tie        23        40       37      100     (27)**

Source: The results are based upon our analyses of the National
Election Studies.
* Based upon the postelection thermometers.
* * The numbers for 1992 are weighted using the time-series


Candidate Thermometer Rankings of the Candidates in the Pre- and Postelection Interviews and the Vote: 1980 and 1992 (in percentages).

First in  | First in   |Presidential Vote in 1980
Pre-Elect.| Post-elect.|
Ratings   | Ratings    |Reagan Carter Anderson Total (Number)
Reagan      Reagan      100      0       0      100    (275)
Reagan      Carter      [1]     [0]     [0]              [1]
Reagan      Anderson    [4]     [0]     [2]              [6]
Carter      Carter       3      97       0      100    (196)
Carter      Reagan      90      10       0      100     (21)
Carter      Anderson    20      70      10      100     (10)
Anderson    Anderson    11      19      70      100     (73)
Anderson    Reagan      89       4       7      100     (28)
Anderson    Carter      [0]     [8]     [0]              [8]

First in | First in | Presidential Vote in 1992
Pre-El.  | Post-El. |
Ratings  | Ratings  |  Bush   Clinton   Perot   Total  (Number)
Bush       Bush         96       1       3      100     (407)**
Bush       Clinton      38      53       9      100      (24)**
Bush       Perot        33       3      64      100      (29)**
Clinton    Clinton     ***      98       2      100     (524)**
Clinton    Bush        [3]      [2]     [0]               [5]**
Clinton    Perot         3      22      74      99       (31)**
Perot      Perot         3       9      88     100      (126)**
Perot      Bush         69       4      27     100       (22)**
Perot      Clinton       3      86      10      99       (29)**
Source:   The results are based upon our analyses of the National
Election Studies.
* Cases in which respondents tied candidates for first place have
been excluded.
* * The numbers for 1992 are weighted using the time-series
* * * Less than one percent.


1. This article emphasizes the reasons third parties fail. For an extensive discussion of the reasons voters do support third-party candidacies, see Steven J. Rosenstone, Roy L. Behr, and Edward H. Lazarus, Third Parties in America: Citizen Response to Major Party Failure (Princeton, NJ: Princeton University Press, 1984).

2. See John H. Aldrich, Why Parties? The Origin and Transformation of Political Parties in America (Chicago: University of Chicago Press, 1995); William H. Riker, Liberalism Against Populism: A ConfrontationBetween the Theory of Democracy and the Theory of Social Choice (San Francisco: W. H. Freeman, 1982); and Riker, The Art of Political Manipulation (New Haven: Yale University Press, 1986).

3. Race-related issues led to substantial shifts in party support since World War II (see Edward G. Carmines and James A. Stimson, Issue Evolution: Race and the Transformation of American Politics, (Princeton, NJ: Princeton University Press, 1989), but not to the success of a new political party. For a recent critique of the Carmines and Stimson thesis, see Alan I. Abramowitz, "Issue Evolution Reconsidered: Racial Attitudes and Partisanship in the U.S. Electorate," American Journal of Political Science 38 (February 1994): 1-24.

4. See Ralph Ketcham, James Madison: A Biography (Charlottesville: University of Virginia Press, 1990).

5. Maurice Duverger, Les partis politiques, 3rd ed. (Paris: Armand Colin, 1958), 247.

6. Maurice Duverger, Political Parties: Their Organization and Activity in the Modem State, trans. Barbara North and Robert North (New York: Wiley, 1963), 217.

7. Duverger, Political Parties, 218. For a discussion of the historical developments leading to Duverger's law, see William H. Riker, "The Two-party System and Duverger's Law: An Essay on the History of Political Science," American Political Science Review 76 (December 1982): 753-66. For a more recent formulation by Duverger himself, see Duverger, "Duverger's Law Forty Years Later" in Bernard Grofman and Arend Lijphart eds., ElectoralLaws and Their Political Consequences (New York: Agathon Press, 1986).

8. Riker, The Art of Political Manipulation, 78.

9. The complete list of candidates is presented in Paul R. Abramson et al., "The Problem of Third-Party and Independent Candidates in the American Political System: Wallace, Anderson, and Perot in Comparative Perspective" (paper presented at the annual meeting of the American Political Science Association, New York, 1994), table 1.

10. See Paul R. Abramson, John H. Aldrich, and David W. Rohde, Change and Continuity in the 1992 Elections, rev. ed. (Washington, DC: Congressional Quarterly Press, 1995), 76.

11. See Hannah Arendt, The Origins of Totalitarianism, 2nd. enlarged ed. (Cleveland: The World Publishing Company, 1958), 264.

12. Nadine Cohodas, Strom Thurmond and the Politics of Southern Change (New York: Simon & Schuster, 1993), 400.

13. John A. Ferejohn and Morris P. Fiorina, "The Paradox of Not Voting: A Decision Theoretic Analysis," American Political Science Review 68 (June 1974): 525-36; "Closeness Counts Only in Horseshoes and Dancing," American Political Science Review 69 (September 1975): 920-25; and "To P or Not to P? Still Asking After All These Years?" (typescript, Harvard University, 1993).

14. For individual-level evidence suggesting that some voters behave strategically in Britain, see Bruce E. Cain, "Strategic Voting in Britain," American Journal of Political Science 22 (August 1978): 639-55. For evidence on Canada, see Jerome H. Black, "The Multicandidate Calculus of Voting: Application to Canadian Federal Elections," American Journal of Political Science 22 (August 1978): 609-38.

15. For a discussion of Condorcet's theory of social choice, see Duncan Black, The Theory of Committees and Elections (Cambridge, England: Cambridge University Press, 1958).

16. Gordon S. Black and Benjamin D. Black, "'Perot Wins!': The Election That Could Have Been," The Public Perspective 4 (January/February 1993): 15-16.

17. Larry Hugick, "A Response to Gordon and Benjamin Black: Perot's Own Actions Determined His Fate," The Public Perspective 4 (January/February 1993): 17-18.

18. Steven J. Brams and Samuel Merrill III, "Would Ross Perot Have Won the 1992 Presidential Election Under Approval Voting?" PS 27 (March 1994): 39-44. The 1992 NES exaggerates Clinton's support. See Abramson, Aldrich, and Rohde, Change and Continuity in the 1992 Elections. However, it seems unlikely that these biases are so great as to call into question the conclusion that Clinton was the likely Condorcet winner.

19. See Abramson et al., "'Sophisticated' Voting in the 1988 Presidential Primaries," American Political Science Review 86 (March 1992): 55-69.

20. For evidence that Wallace supporters who voted for Nixon were acting strategically, see Richard F. Bensel and M. Elizabeth Sanders, "The Effect of Electoral Rules on Voting Behavior: The Electoral College and Shift Voting," Public Choice 34 (1979): 69-85.

21. See Abramson et al., "The Problem of Third-Party and Independent Candidates," table 4.

22. Riker, "The Two-party System and Duverger's Law."

23. Arend Lijphart, Democracies: Patterns of Majoritarian and Consensus Government in Twenty-One Countries (New Haven: Yale University Press, 1984).

24. Philip E. Converse and Georges Dupeux, "Politicization of the Electorate in France and the United States," Public Opinion Quarterly 26 (Spring 1962): 1-23; see also Converse and Roy Pierce, PoliticalRepresentation in France (Cambridge, MA: Harvard University Press, 1986).

25. Converse and Pierce, 'Comment on Fleury and Lewis-Beck: 'Anchoring the French Voter: Ideology versus Party,"' Journal of Politics 55 (November 1993): 1110-17; Christopher J. Fleury and Michael S. Lewis-Beck, "Anchoring the French Voter: Ideology versus Party," Journal of Politics 55 (November 1993): 1100-09; Fleury and Lewis-Beck, "Deja Vu All Over Again: A Comment on the Comment of Converse and Pierce," Journal of Politics 55 (November 1993): 1118-26; and Ronald Inglehart and Hans D. Klingemann, "Party Identification, Ideological Preference and the Left-Right Dimension among Western Mass Publics" in Ian Budge, Ivor Crewe, and Dennis Farlie, eds., Parry Identification and Beyond: Representations of Voting and Party Competition (London: Wiley, 1976).

26. Converse and Dupeux, "Politicization of the Electorate."

27. Joseph A. Schlesinger and Mildred Schlesinger, "The Reaffirmation of a Multiparty System in France," American Political Science Review 84 (December 1990): 1077-1101; "French Parties and the Legislative Election of 1993," Party Politics 1 (July 1995): 367-78.

28. See Francoise Subileau and Marie-France Toinet, Les chemins de l'absention: une comparaison Franco-Americaine (Paris: Editions la Decouverte, 1993).

29. For the complete results for the five direct presidential elections held between 1965 and 1988, see Abramson et al., "The Problem of Third-Party and Independent Candidates," table 6.

30. See David W. Abbott and James P. Levine, Wrong Winner: The Coming Debacle in the Electoral College (New York: Praeger, 1991).

* The National Election Studies surveys used in this article were provided by the Inter-university Consortium for Political and Social Research. The Consortium is not responsible for our analyses or our interpretation of these data. We are grateful to Joseph A. Schlesinger and Mildred Schlesinger for their assistance and to Abraham Diskin for his suggestions.

 [PSQ Home] /
Copyright 1995 New Prospect, Inc.