The large proportion of relatively lopsided American presidential elections over the past forty years presents a dilemma for the political scientist. In the period from 1940 to the present, the margin of victory has averaged 10.1 percent of the two-party vote. This figure differs dramatically, however, when one compares elections in which one of the candidates is the incumbent President to open elections, elections without an incumbent running. Open elections, as Table One shows, have been much closer over the past fifty years than have elections involving a sitting President. Three elections of this latter type, the elections of 1964, 1972, and 1984, account for much of this disparity.

        This paper investigates the behavior of the losing parties in these anomalous elections. On the one hand, a prospective, spatial model would predict that in a two-party system, both parties' policy positions would converge to the position of the median voter, and the parties would thus split the vote evenly. (Downs 1957: 115) In the elections of 1964, 1972, and 1984, I shall demonstrate that this clearly did not happen. On the other hand, a psychological or retrospective model might allow for rather unbalanced contests to occur occasionally, but it predicts nothing about alterations of candidates' strategies in accord with pre-existing advantages or disadvantages. (Fiorina 1981: 6-9; Key 1966: 2-3) The retrospective model does not, in other words, alter the logic of spatial competition; instead, it portrays elections as referenda on the incumbent party's performance. According to the basic spatial model, there is no advantage to incumbency; according to the retrospective model, the presence of an incumbent may stack the deck in favor of the incumbent, but it should not alter the policy strategy of the challenger.

        This paper is an experiment in turning the conventional wisdom regarding these three elections on its head. Many political scientists and political commentators have attributed the magnitude of these victories to the fact that the losing candidates, all members of the out-party, took "extreme" policy stands, stands which did not reflect the preferences of the majority of the voters and did not conform to the spatial logic of the Downsian model. (Converse et al 1965, Miller and Miller 1976; and in a much more cautious manner, Abramson, Aldrich, and Rohde 1986) I argue instead that an approximation of the eventual result of the election, attained through retrospective evaluation of the incumbent, was known by both candidates prior to the general election campaign, and that this knowledge freed the challenging candidate from the need to run a centrist campaign. Whereas the conventional wisdom argues that positions away from the median lead to electoral disaster, I argue instead that impending electoral disaster can lead to such positions.

        This process can happen in two ways. First, an expected loss for the out-party may either discourage the most viable candidates from running for president or may discourage voters from opting for the most "electable" candidate during the primaries or the party convention. (see Jacobson 1987: 46-48) Second, an expected loss may dissuade the challenger from the moderating his primary positions after winning the party nomination. In both of these cases, winning the election is deemed by voters or by the candidate an unrealistic goal, and alternate goals are introduced. Finding out what these goals are is more of an exercise in hypothesizing than in demonstration, for there is no way to concretely ascertain or prioritize campaign goals. Nevertheless, there are means by which the degree of "extremism" in the campaign can be gauged, and there is much anecdotal and qualitative evidence regarding the motivations behind the activities of candidates, party elites, and the members of the candidate or party campaign organizations.

        This paper, then, is a three-pronged endeavor. First, I present the logic of candidate positioning and a theoretical justification for my notion of the extreme candidate as an actor rationally responding to political circumstances. Second, I place the relevant candidates in each election on a rough policy continuum. I analyze major news articles during the campaign to judge which candidate was perceived as being most extreme in each election and whether candidates' issue emphases changed over the course of the campaign. This content analysis is supplemented by a presentation of candidates' actual voting records as members of Congress and by data culled from the National Election Survey regarding voter perceptions of candidate positions. Third, I analyze qualitative data from campaign documents, funding and scheduling reports, and campaign histories and memoirs that suggest an alteration of campaign strategy from the vote-maximization model, the presence of alternate goals, or an acknowledgement of the low chances of victory for the challenging candidates. I present as possible alternate goals factional dominance within the party, presentation of alternate policy options to the public in order to shift debate, and positioning for future elections.
 
 

        Specification of the actors in a spatial model must not be made carelessly. The primary actor in the model presented here is not the voter; voters' policy preferences are taken to be static at each time stage of the model and known by candidates. The major question, then, is whether the relevant actor here is the candidate or the party. These two concepts, though often used interchangeably, are not the same. In a model where primary or intra-party elections are considered, the actors are of course the individual candidates and their supporters, but in a general election, candidate and party are difficult to separate. Downs presents parties as the actors in question here, although he admits that there are problems with this approach. (Downs 1957: 24-27) A party cannot be said to include all who vote for the party, as this would not imply a unified ordering of preferences. Rather, Downs argues, a party must be internally consistent. One might thus read party to include a set of elites who consent to abide by an established contract, such as a party platform.

        One could avoid this definitional problem by merely considering candidates as the actors. Many spatial theorists writing after Downs have done this. Page (1978) uses party and candidate interchangeably. Riker (1963) and Riker and Ordeshook (1973: 307-337) speak of coalitions, which may be parties or may be subunits of parties. Schlesinger (1975) attributes the different strategies Downs and Riker and Ordeshook advocate to differing conceptions of who the actors are, of who is included in the parties or coalitions that must take positions. Downs, Schlesinger claims, is the theorist of candidate or party behavior, while Riker makes greater allowance for voters' participation as members of coalitions within the party. For the purposes of this discussion, I will employ the Downsian concept of a party, narrowly conceived and bound to a particular candidate as its chosen agent, as the actor except when I explicitly refer to primary elections. I do so because my discussion assumes some continuity from one election to the next, some value placed upon subsequent elections even if a party loses the current election. Hence, I must assume that the actor will be present to contest the next election, as assumption that is often not borne out in the case of individual candidates, but which in the American case has almost without exception been applicable to parties.⁽¹⁾ Following Aldrich (1995), I assume the candidate to be an agent of the party, duly appointed through the nomination process to take positions on behalf of the party.

        The goals set forth in a spatial model are also not uncontroversial. The Downsian approach posits winning as the sole goal of the party; policy concerns are merely a means towards gaining power. (Downs 1957: 27-28) Critics of Downs have shown that the Downsian model can still be used, although with somewhat different results, if winning is seen not as an end in itself but as a means toward making policy. Wittman (1973, 1983), Schlesinger (1975), and Chappell and Keech (1986) have all presented spatial models that counter the Downsian theory of policy appeal to the median voter, on the grounds that there may be more than one winning strategy in an election, and in such cases candidates can maximize their expected policy utility while not damaging their chances of winning.

        The simple model I propose here employs utility calculations regarding winning both for its own sake and for the sake of advancing policy preferences. I assume policy preferences to be manifested in a candidate's past behavior. Because of the constricted nature of the preference distribution both among primary voters and among a candidate's constituency in previously held offices, I assume that actual policy preferences can be more accurately identified in past behavior than in general election behavior for presidential candidates. The model I present relies heavily upon subjective calculations; hence the mathematical specifications entailed in the model are not able to be precisely identified. I propose them rather to give a notion of probability thresholds for different types of behavior.
 

Formal Specification of Game:

1. Players:

        7 Citizens, where C_i represents the ith citizen

        2 candidates, A (the incumbent) and B (the challenger)
 

2. Strategy Sets:

Citizens are arrayed on a policy scale from 1 to 7 such that the preferred, exogenously given policy of C₁ = 1, that of C₂ = 2, etc. Preferred policies are denoted as P*_i for each citizen.

        All C_i vote for either A or B. There is no abstention.

Candidates establish a position along this policy continuum.

        Thus S_A is an element of{1, 2, 3, 4, 5, 6, 7}

        S_B is an element of{1, 2, 3, 4, 5, 6, 7} for each S_A
 

3. Sequence of Moves:

A, the incumbent, already has an established policy position which is known to the voters. As A has played this game previously, A is known to voters, and, in effect, has the first move.

B moves second.
 

4. Utility Functions:

Citizen preferences are symmetric, unimodal, and monotonic, peaking at the citizen's preferred policy. Citizens evaluate candidates solely on the basis of the candidates' policy positions. They have full information regarding these positions.⁽²⁾ The utility of the policy proposals a candidate x provides to a citizen is thus

         EU_Ci = - (|P*_i - S_x|/n)

Citizen preferences can be graphically represented as follows:

Most importantly, where there is no difference between candidates (S_A = S_B), citizens are assumed to be averse to change, and thus vote for candidate A.⁽³⁾

    Candidates, as well, are assumed to have policy preferences, such that the expected value of a strategy for a candidate is affected by the policy position taken. There is, however, intrinsic value to winning, while there is none to losing. Let us define these two utilities as U^W (the value of winning for its own sake) and U^P (the utility of taking a policy position). If we give winning a nonpolicy utility of 1, the following EV function obtains:

        EV(S_x) = [P(W) * U^W] + U^P = [P(W) * 1] - (|P*_x - S_x|/n)

P(W) is a function of S_A and S_B such that A seeks to maximize P(W) given its expectation of B's optimal strategy; B attempts to maximize P(W) given the strategy A has already adopted.
 

This candidate payoff function is adapted slightly from Riker and Ordeshook (1973: 334-336; Wittman 1983), who define a candidate's expected value as

        EV(S_x) = [P(W at S_x) * EU(W)] + [P(L at S_x) * EU(L)]

I drop the utility of losing by setting it equal to 0 and define EU(W) in both policy and nonpolicy terms. I have presented the policy utility of winning, however, in the same terms as it is for citizens, such that a candidate (who, after all, is also a citizen) has policy preferences entirely independent of the campaign.

For this game, let us assume that candidate A has a preferred policy of 2 and candidate B has a preferred policy of 6. The logic of this game holds, however, for any candidate preferences in a two-player game, even if both have policy preferences on the same side of the median.
 

5. Solution Concept:

The above EV calculation tells us two things:

a) winning is always preferred to losing, regardless of a candidate's policy preferences.

b) In an election where candidates choose positions simultaneously, there is still an incentive to choose the median position. Since this is a full information two-candidate game, we only have three possible probabilities of winning: 0, .5, and 1. No candidate need move more than three spaces to reach the median, so the utility for any candidate of moving toward the median if candidates move simultaneously will be no less than .5 - 3/7, or approximately .08, which is preferable to not reaching the median, in which case a utility of 0 results.

For the game specified in (3), however, where candidate A moves first, the Nash Equilibria are

For A,

        S_A = 4

        EV(S_A) = 1 - 2/7 = 5/7
 

All other strategies for A will result in an expected value of 0 (as A expects B to choose 4 if A does not).
 

For B,

        If S_A 4, B has a variety of best replies:

            If S_A = 1 or 7, S_B = 6

                EV(S_B) = 1 - 0 = 1

            If S_A = 2 or 6, S_B = 5

                EV(S_B) = 1 - 1/7 = 6/7

        If S_A = 3 or 5, S_B = 4

            EV(S_B) = 1 - 2/7 = 5/7

        ELSE if S_A = 4, S_B = 6

            EV(S_B) = 0 - 0 = 0

any other S_B where S_A = 4 will yield a negative EV.

In sum, the expected equilibrium outcome is {4, P*_B}

        There are several implicit assumptions required to yield this result. First, candidates are not vote-maximizers. They seek only to win; the margin of victory (or defeat) is irrelevant.

        Also, the notion that candidates will suffer if they move from their preferred policy positions (even in a losing campaign) is subject to debate. A rather ad hoc justification for this is that candidates for president (as in this paper) have usually already won other elections, presumably at positions closer to their established policy positions. To move from these positions is costly if they wish to maintain strength among their initial constituencies.
 

        Most spatial competition models that incorporate policy preferences of candidates assume that candidates are concerned with policies that are implemented, not policy positions taken. Thus, in two-candidate competition the policies the winning candidate plans to implement are figured into the losing candidate's utility function. I have opted rather to include only a candidate's own positions taken (in relation to the candidate's preferred policies) for two reasons. First, there is the rather simple psychological explanation that a candidate prefers to argue for policies that he totally supports. Otherwise, a sort of cognitive dissonance results which may cause the candidate a sort of personal displeasure with himself. Second, and perhaps more plausibly, candidates must maintain their own coalitions. Unless a candidate is certain that the current election shall be the final one, a candidate may alienate his original coalition of supporters by shifting policy positions. In the case, for example, of a member of Congress running for President, we may assume that this candidate would like, if he is not successful in his quest for the Presidency, to continue to serve (and run for re-election) in Congress. Positions too far afield from the positions which have in the past gotten him re-elected to Congress may have a negative impact upon future re-election attempts in his Congressional district. Likewise, a candidate who would like to be able to run for President again must be certain of having a constituency within the party to return to in order to secure the nomination.

        There are, of course, other concerns on the part of candidates in taking a policy position, though these are difficult to model. I will touch on just one of these formally here.

a) party unity - helping other party office-seekers or reconciling splits in the party.

b) factional dominance within the party. The converse of (a), this concern would have candidates trying to establish the upper hand in party conflict by presenting the position of a faction of the party as that of the party as a whole.

c) future-oriented concern. Chappell and Keech (1986) argue that incumbents' hands are tied in future planning (thus guaranteeing victory to the challenger in multidimensional spatial competition), as they have an overriding imperative to prepare solely for the election at hand. Candidates who do not expect to win, however, may look ahead, as Downs notes:

Some political parties (especially newly founded ones) are more interested in future elections than in present ones because their chances of gaining office are greater in the future. However, a governing party has already gained office; hence its primary concern is retaining its position, i.e., winning the next election. (Downs 1957: 174n)
 

Within our two-party or two-candidate universe, let us conceive of this in the following manner:

        Impose a two-term limit on office-holders (as there is for the US Presidency). While A has the first move for the current election (time 1), A will not be able to run in the next election (time 2). A's ability to secure victory merely by taking a strategy at the median is eliminated, and voters move simultaneously.

        Suppose furthermore, however, that voters have a good memory of the campaigns that candidates have run in the past. Let us move one election beyond the one considered above, and assume that A won that election. In the subsequent election, the new, non-incumbent candidate of A's party (here denoted A') will be judged, in part, by the positions that A has taken. Likewise, the new candidate from B's party (either the original candidate B or a new candidate B') will be judged in part based on the positions B took in the prior election. Finally, assume that the preferences of voters have shifted between elections in the following manner:

            Time 1

            Time 2

        Each voter has shifted one notch to the right during the intervening time interval except for C7, who cannot move any further to the right. The median position in this newly reconfigured electorate is equivalent to the position that C5 held in the previous election.

        If candidate B is able to foresee this shift from a vantage point at time 1, is there any way that candidate B can seize this position before candidate A'? Can B, in effect, reverse the sequence of moves presented earlier and reverse the voter preference for the incumbent that held in the earlier game?

A payoff function for B which takes into account positioning for subsequent elections might take the following EV form:

        EV(S_B) = {[P(W) * U^W] - U^P} at time 1
            +(delta) {[P(W) * U^W] - U^P} at time 2

where delta  is the discount factor.

        This equation can be expanded to cover any number of time periods (that is, we can suspend the two-term rule) as long as the length of A's incumbency is finite and known.

        Let us plug in our original utility calculations. It has already been established that, given a strategy of 4 for A at time 1, B's optimal response is to maintain its most preferred policy position, 6. This yields an EV of 0 for a one-shot game. Suppose, however, that B knows that the preferences of voters will shift before the next election, as shown above, and that B will gain the new median before A' does (thus rendering the strategy of A' unsuccessful) if A' seeks to converge to this new median. In a two-stage game, a strategy of 5 for B will yield

        EV(S_B) = 0 - 1/7 + (1 - 1/7) = -1/7 + [6/7 (delta)]

This strategy beats the one-shot strategy of 6 if

        6/7 (delta) > 1/7
                > .16
 

        Even if we can only assume that candidates would move simultaneously at time 2, a forward-looking strategy of the type mentioned above may be preferable:

        EV(S_B) = 0 - 1/7 + (.5 - 1/7)
        = -1/7 + .36(delta)

which defeats the one-shot strategy of 6 if

        .36(delta) > 1/7
        > .39
 

Because we have not defined any costs to changing one's position between elections, however, this strategy has no real meaning, since forward-looking strategies have no bearing on a future campaigns where a party's past positions do not have some effect on voter calculations.

        Determining the value of is, again, an empirical matter, and may rest more on subjective expected probabilities than upon any preordained spatial logic.

        American Presidential candidates do not, of course, maneuver within such a small universe of voters. I have used this seven-voter illustration in part because it is a small-scale way of envisioning candidate's calculations. I have also used it because the number of voters accords with one of the most convenient means of categorizing voters and candidates, the NES 7-point scale, which ranges from extremely liberal to extremely conservative, with a ranking of 4 indicating a candidate perfectly between these two extremes. Another important aspect of presenting candidate competition within such a small electorate, however, is that it allows us to make some allowance for voter uncertainty. If we divide the range of possible candidate positions into a small number of blocks, we can allow a corresponding simplification of voter understandings of candidate positions; a "moderate" candidate or a "slightly conservative" candidate can be identified far more easily than can a precise position on a more nuanced scale. That is, a voter may not know the precise nature of a candidate's welfare proposals, but that voter may understand whether a candidate who is "slightly conservative" is likely to look out for her interests.
 

        Before presenting case studies to determine how closely this model accords with reality, a brief review of the techniques I employ here for spatial positioning is necessary. The process for empirically applying this model is a threefold mission. First, it must be ascertained where candidates' preferences lie on the liberal-conservative issue dimension. Second, one must also determine the degree of knowledge candidates have both of the distribution of policy preferences among the electorate and of their chances of winning the election given any particular set of strategies. Third, one must identify clues as to how information regarding both voter preferences and probabilities of winning affect candidate strategy and candidate positioning in the general election.

        Surprisingly little comprehensive research exists which identifies where American candidates have stood on the liberal-conservative dimension.⁽⁴⁾ I shall use as a framework for my own methodology the spatial positioning study of Laver and Hunt (1992: 31-34), which presents four different methods of estimating party or candidate positions on single or multiple issue dimensions: content analysis of policy documents or speeches, mass survey, data, government expenditures, and expert surveys. For my purposes, government expenditures are clearly inapplicable, as we are concerned with candidacies, not the records of governments, and because government expenditures in the American case often may represent a compromise between members of the governing party or parties. Policy documents, likewise, can only be considered insofar as they are policy proposals of the candidates in question, not explanations of policies implemented. The most obvious source of policy proposals for candidates, their party platforms, runs into this problem, as Rosenstone (1983: 173) notes.

        The data I have chosen for each of the remaining aspects are somewhat incomplete. To estimate candidate positions prior to the campaign, I have chosen two options. Americans for Democratic Action (ADA) have rated the voting records of members of Congress since 1947 in terms of their "liberal quotient," with a score of 100 representing the most liberal and 0 representing the most conservative.⁽⁵⁾ ADA ratings thus serve as a sort of proxy for expert surveys. In the manner of Gray's analysis, I have in Table Two measured the relevant candidates in relation to their peers. This sidesteps, somewhat, potential problems caused by the fact that ADA ratings are bimodal, with many members often appearing at each of the extremes of the ADA scale.⁽⁶⁾ ADA ratings do not reflect positions in relation to the preferences of the electorate as a whole, yet one might hypothesize that as all members of Congress are elected based upon their policy appeals to subsets of the electorate, they represent a relatively accurate sample, albeit with a smaller distribution of preferences, of the electorate at large.

        Another problem, however, is that not all candidates have ADA ratings, and ADA ratings do not always reflect change over time. The ratings of candidates who have been out of Congress for some time, such as Richard Nixon or George Bush, should be treated with some circumspection. We do, however, have ADA ratings for the challenging candidates in each of the elections considered; the positions of Goldwater and McGovern were taken immediately precedent to their candidacies, while there is an eight year lag between Mondale's Congressional service and his candidacy. For the six of the eighteen candidacies since 1960 for which we have no ADA measurements, however, one must rely upon other means.

        Each of the remaining candidacies can be filled in using mass survey data culled from the ANES. While NES data are not necessarily comparable with ADA ratings, they also allow us to gauge how centrist candidates are, as perceived by the voters. Since voters are the ones to whom the candidate position matter most, in the end, these may in fact be more relevant data. Since 1976, the NES has asked voters to place the candidates, the political parties, and themselves on a seven-point scale, ranging from "extremely liberal" to "extremely conservative." The responses to these questions are presented in Table 3b. A major drawback of the NES questions, however, is that they are asked in September, after the general election campaign has already begun. They are thus not a "pure" measure of initial candidate positions. Insofar as the incumbent is more of a known quantity to the voters, and hence less likely to attempt to modify positions at the last moment, they may be more reliable for measuring initial incumbent positions than they are for measuring challenger positions. To put this in the context of the game presented above, the incumbent's "first move" has already been made by the time NES questions are asked, while the challenger is still in the process of responding.

        Answering the remaining two questions is a less quantitative endeavor. One cannot get a precise measure of what candidates believe the odds on their candidacies are. I rely instead upon a combination of statements issued from within the campaign, NES data, political science prediction models, and pre-election polls. The NES has for the entire period covered contained a question asking respondents to predict the winner of the Presidential election. The results of this question are shown in Table 3a. One could assume that candidates are surely no less aware than is the average voter of the probability of their victory.

        Pre-election polls, though notorious for their fluctuation (see King and Gelman, 1992; Rosenstone 1983: 24-42), also are presumed to be considered carefully by candidates. These polls do not, however, necessarily provide an accurate barometer of a candidate's true chances. Whereas in two of the elections considered polls throughout the campaign provided somewhat uniform results, in the case of the 1984 election they varied to a significant enough degree that early polls would have been an unreliable indicator of the eventual outcome. Hence, more scientific prediction models are really the main driving force behind this model.⁽⁷⁾

        The third question is the most difficult to conclusively answer. Whether candidates altered their strategies, and their reasons for doing so, is a speculative question, as one cannot expect candidates to be forthcoming about their prospects of winning and their goals should victory prove unlikely. Measuring whether candidates attempt to move toward the center may be done in the case of the 1984 election because we have the two measures of candidate position for Mondale, one preceding the campaign and the other taken during the campaign. This by itself is not conclusive, however. We cannot be sure how much of this movement was planned by Mondale and how much was due to other factors. Again, inference must be made for this question using more qualitative evidence culled from reports of the events of the campaign. Issues discussed and statements made are gleaned from the written record of the campaigns, as made both by participants in the campaign and journalistic observers. An analysis of campaign coverage by the two major news weeklies, data from which are presented in table four, provides a clue to the issues discussed and the reaction of the news media to candidate strategies. Shifts in strategy may be seen in the issues discussed and the regions of the country in which candidates campaign.

        Finally, should the challengers in these elections, as the model predicts, fail to move toward the center, this need not support the future-oriented concerns presented in the model. Alternate explanations may include ineptitude on the part of the candidate, mistaken beliefs about the preferences of the electorate, strategies designed to shift the terms of debate toward the challenging candidates' positions, or limitations placed upon the candidates by the primary process.

        While the details of these explanations will become clear in the case studies, a few words on two of these explanations are in order. Strategies designed to pull the incumbent toward the challenger would not be supported by the model here. First, the incumbent is given only the first move; he cannot subsequently alter his strategy to converge toward the challenger's position. Second, the logic of the game gives him little reason to seek to do so; the median position is in all cases a winning strategy for the incumbent. While a move toward the challenger on the part of the incumbent would increase the incumbent's share of the vote, (a) the model does not use vote-maximization criteria, and (b) such a move would leave the incumbent vulnerable should the challenger be able to make a subsequent move. Such strategies would lead to an expansion of the number of moves allowed in the game, resulting, were there no limit on the number of moves, in convergence toward the center. An aggressive voter registration attempt, which may be found in each of these campaigns and in many other contemporary campaigns, might serve to bring the median, and thus the opponent, closer. This attempt at expansion of the voting population would likely fall more under the heading of long-range concerns.

        The primary process, and its effect upon candidates' spatial mobility in the general election, has been discussed by Aldrich (1980) and by Page (1978). Aldrich proposes a "winnowing-out" process, through which candidates at the extremes of the preference distribution of their party can win the party nomination. (Aldrich 1980: 171-173) The logic here is similar to that of multi-party competition; with multiple players, convergence to the median is no longer a dominant strategy and candidates can more closely represent their policy preferences. (Downs 1957: 122) Numerous commentators on the primary process (Barber 1973; Grassmuck 1985; Weil 1973: 116-117) have argued that "bruises" from the primaries can linger into the general election. Aldrich, however, finds that in his case study of the 1976 campaign, competition on the issues in the 1976 preconvnetion campaign had remarkably little effect on the issues raised in the general election. (Aldrich 1980: 193)  Issues that seem important in the primaries may fail to motivate the general electorate; candidates will also not be "squeezed out" from both sides in the general election, as there are only two candidates. Character questions may remain, but in a spatial model these are not factors. The main importance of the primary campaign is that it delays the eventual nominee's adaptation to his opponent's strategy. The primaries are in part what gives the incumbent the first move. A striking regularity of Presidential elections in the past fifty years is that no incumbent who has faced a serious intra-party challenge has been re-elected; no incumbent who has run unopposed or close to it has failed to win re-election.⁽⁸⁾ It is easy to infer from this pattern the chronological basis for imputing a first move to the incumbent; I therefore minimize attention to the primaries on the grounds that any claim of constraint supports this initial premise of the model.

        Finally, in each of these cases I hypothesize the existence of long-term concerns on the part of the challenging candidates via the situation of the campaign within the context of developments within the candidate's party and changes made within the party in the years preceding and subsequent to the campaign. While it is tempting to assess the existence of such strategy by gauging the party's success in the subsequent election, the uncertain nature of any such strategy makes this means of assessment problematic, as shall become clear in the following cases studies, and I have therefore tried, not always successfully, to avoid it.
 

        The 1964 contest between Lyndon Johnson and Barry Goldwater is held up by many analysts as one of the foremost examples of an election with clearly differentiated policy positions. (see especially Converse et al 1965) The incumbent Johnson won this election with 61.3 percent of the vote, sweeping every demographic group: all races, levels of education, age groups, religions, occupational groups, and regions of the country. This was the largest margin of victory since Roosevelt's 1936 victory. Goldwater had his strongest showing in the Deep South, picking up the electoral votes of five Southern states in addition to his home state of Arizona. Goldwater's supporters claimed that his campaign would have fared better had he faced his intended opponent, John F. Kennedy. Even measured in relation to Kennedy, however, Goldwater is clearly far from the median.

        In terms of past record, the Republicans could scarcely have chosen a more conservative candidate, nor could the Democrats have chosen a more moderate candidate. While the news reports of the campaign were dominated by references to "extremism," partly due to Goldwater's own use of the word in his acceptance speech at the convention, ADA ratings bear out such claims. Goldwater was the most conservative member of the Senate in each of the ten years leading up to his Presidential campaign; his mean rating of 1.3 reflects three votes he cast in 1953; other than those he attained a consistent rating of 0 for his entire tenure in the Senate prior to 1964. Johnson, by contrast, would have ranked 48th most liberal had he been in the Senate in 1964. Johnson was no doubt associated by voters with the more liberal policies of John F. Kennedy, yet Kennedy had also been more moderate than Goldwater, as Table Two shows, and was perhaps even more popular following his death than he had been while President. Nonetheless, Goldwater's defeat of a two-term Democratic incumbent Senator in 1952 and the fact that as many as 80,000 Arizona Democrats voted for him in his 1958 re-election bid established him as one of the most prominent Republicans in the early 1960s. (Shadegg 1965: 3-5) 1958 had been a bleak year for Republican members of Congress; Goldwater's victory catapulted him to the chairmanship of the RSCC, which was in turn to serve as a springboard for his Presidential campaign.

        His campaign, however, was in the eyes of many observers doomed from the start. Because the party (under Goldwater's tutelage) had reapportioned the seats in the RNC so as to give Southern states greater numerical strength and had increased its organizational activities in the South, Goldwater was able to lock up an insurmountable lead by sweeping Southern and other small state primaries and caucuses early in the primary season. (Klinkner 1994: 53-60) His march to the nomination was slowed, however, by Nelson Rockefeller's strong showing in later primaries, most notably Oregon's and California's, and by the emergence of an "anybody but Goldwater" movement immediately prior to the convention, which featured more established figures such as Pennsylvania Governor William Scranton and Michigan Governor George Romney. While Goldwater won the nomination on the first ballot, the convention was still a bitter one, and several of his opponents refused to endorse him. (White 1965: 397) The convention adjourned with Goldwater trailing in the polls by 28 points; he never rose above this amount until Election Day. (Runyon 1971: 278-280)

        In addition to survey results, there is copious evidence that Goldwater was in a hopeless situation. Of 29 combined issues of campaign coverage in Time and Newsweek, 12 made mention of Goldwater's dreary prospects. Shadegg (1965: 241-244), a member of the campaign leadership, argues that Goldwater spent much of October rationalizing to the voters his impending loss. Karl Hess, another campaign staffer, noted that

if Goldwater could win the candidacy, he would also win control of the Republican National Committee, and this, for the future of the party, would be secondary in meaningfulness only to an electoral victory. (quoted in Klinkner 1994: 72)

        Given these strictures, then, how can one gauge Goldwater's possible reaction? First, there is the matter of issues discussed. While Table Four's issue headings are meant to do little more than tally what the focal points of the campaign are, they do provide some interesting leads. First, it appears that the economy was not a viable issue for Goldwater. With per capita GNP increasing at a rate of 5.6 percent (Tufte 1978: 25), unemployment dropping, and Johnson's approval rating at 76 percent (Runyon 1971: 279), Goldwater had little to gain through references to the economy. Other issues that he did take up, however, seem equally ill-considered. What the table does not show is that much of the discussion of civil rights was couched in terms of the views Goldwater's most fervent conservative supporters may have projected onto him, not the views actually espoused in his speeches. Goldwater went so far, in fact, as to make a not-so-secret deal with President Johnson to avoid making civil rights a campaign issue. (Lamb and Smith 1968: 98-102) This is puzzling because, according to Gray's analysis, civil rights was the issue upon which Goldwater was closest to the center; on foreign policy, internationalism, agriculture, and economic matters, Goldwater was conspicuously farther from the median than was Johnson or Kennedy. On civil rights, Goldwater ranked 55th most liberal in the 85th Congress (1957-1958) as compared to 75th for Johnson and 58th for Kennedy; he ranked 73rd in the 86th Congress (1959-1960) as compared to 7th for Kennedy and 59th for Johnson.

        One could argue that, insofar as Goldwater carried much of the South anyway, there was no need for him to make his views on civil rights explicit, and that he may have been better served by keeping silent in regions where many would have liked to believe him more conservative than he actually was. He was not similarly restrained, however, on far more controversial issues. Foreign policy, on which the President held a clear advantage according to Gray's calculus, was a focal point of his campaign, as was military strength. These issues make up many of the negative references to his views shown in the table. Words such as "extremist" and "dangerous" are frequently found in coverage; Time went so far as to devote several articles to pointing out factual errors Goldwater had made in his arguments regarding defense, a treatment that was not given to Johnson.

        The issues deployed in Goldwater's campaign follow a similar trajectory to other campaigns discussed here: there is an initial sally on a variety of issues, done concurrently with a somewhat unsuccessful attempt to iron out intra-party differences and enlist Goldwater's primary opponents in his campaign. Once this stage, from roughly July through early September, is over, however, Goldwater's views on the military are brought even more to the front, as are his "philosophical" appeals, his sharpening of the moral difference between himself and Johnson, or more broadly between Republicans and Democrats. Furthermore, the location of his campaigning shifts away from what one would expect. Whereas the 1960 Nixon campaign concentrated upon the Midwest and Northeast, Goldwater completely wrote off the Northeast, spending only five days campaigning there, and concentrated his energies almost exclusively on the Plains states, the South, and the Southwest. (see Runyon 1971: 156-168 for campaign itineraries) Goldwater surely had more success in these areas than elsewhere, so he may have been concentrating upon areas where he had the greatest chance of winning, yet these areas put together did not have the electoral votes to win the election. As Miles (1980: 298) points out, Goldwater carried few of the racially moderate areas of the Midwest and Upper South where Eisenhower had won in the 1950s; with the exception of his victory in Arizona and in the more conservative counties surrounding Chicago and in Southern California, even at the county level he won nowhere outside the Deep South. And all of this from a candidate who refused to make his more centrist views on civil rights an issue!

        This may, as Hess's quote indicates, seem part of a quest for long-term dominance within the party. None of the five Southern states Goldwater carried in 1964 had been carried by Nixon in 1960;⁽⁹⁾ no non-Southern Democrat has carried any of them since. While Goldwater's appointees to the RNC were summarily replaced following the election, the changes in RNC composition that he had pushed through remained, and the Republican cells that his campaign had established in the South were reactivated with relative ease by Richard Nixon. Nixon carried both of the Carolinas in 1968 while running second to George Wallace in several other Southern states. He went on to sweep the South in 1972, topping 70 percent in all the Goldwater states except Louisiana.

        None of this ensures that Goldwater actually knew what effect his campaign would have on this regional realignment, nor that it would not have happened without Goldwater.⁽¹⁰⁾ Lamb and Smith (1968) argue that Goldwater's campaign was crippled by a relatively inflexible decision-making structure, which had been put in place before John F. Kennedy's death. Had Goldwater been facing Kennedy instead of Johnson, he might have made even greater inroads in the South and West. The maintenance of this hierarchical structure after Kennedy's death, however, may support the hypothesis that the Republicans backing Goldwater regarded the election as a foregone conclusion and were preparing for a future election in which there would be no centrist Southerner such as Johnson to compete with.

        Throughout the election, Goldwater was to some degree a creature of his handlers. Shadegg explains many of the shortfalls of the campaign with a complaint that a cabal of staffers ran the campaign rather than Goldwater himself. Goldwater was never the chief architect of his campaign; he had decided to run as a result of a draft movement begun in 1960. This movement was composed of conservative activists whose sole goal was not victory in 1964; they sought a broader role in the party and in national politics. It would be fallacious to assess Goldwater's judgement by the results of the 1968 campaign. Yet it is somewhat more sound to argue that through the Goldwater campaign a conservative movement within the party solidified, a movement which was to wield great strength in subsequent years. Goldwater, who returned to the Senate in 1968, was able to forestall efforts by Northern Republicans to regain control by threatening a third-party candidacy. (Klinkner 1994: 75) Of those who worked on his campaign, Richard Nixon was one of the few elite Republicans who may have seen any future benefit for the party and for himself in the campaign; he actively campaigned for Goldwater, in return for which Nixon received Goldwater's support four years later. (Miles 1980: 302)

        In sum, Goldwater's candidacy loosely represents the predictions of the model in several ways. Goldwater shows no signs of moderating as the general election approaches, save for his initial attempt to reconcile party leaders to his candidacy. His emphasis on philosophical differences towards the close of the campaign may be seen as an attempt to sharpen the ideological differences while rallying Republican partisans. Numerous attributes of his campaign may be seen as elements of a long-term strategy, though the results are hardly conclusive. Goldwater evidently sought to aid in the shifting of the electorate median in his emphasis upon the "silent majority." While Miller (1965) and others (Polsby 1966: 88-92) argued flatly that this majority did not exist, that Goldwater had been deceived, the dramatic increase in voters in the South shows that there were enough "silent" conservatives to allow the party to build its base, assuming liberal and moderate Republicans in the rest of the country would forgive the party.⁽¹¹⁾ Republican turnout rose above 1964 levels in 1972, and was even above the 1964 figures in raw numbers, if not in percentages, for Gerald Ford in 1976, indicating that a sizable new population of voters had entered the party ranks. Whether or not Goldwater foresaw or caused this, his campaign fits in with overall Republican strategy in the 1960s and 1970s.
 

        No other campaign gives the model as good ex post facto support as does the 1964 campaign. In many ways, however, George McGovern's campaign serves as a mirror image of Goldwater's. McGovern lost by an even greater margin than did Goldwater; Richard Nixon was re-elected with 61.8 percent of the two-party vote and lost the electoral votes of only Massachusetts and the District of Columbia. Like Johnson, Nixon scored majorities in almost every demographic category (with the exception of nonwhites). Nixon had only token opposition in the Republican primaries, while McGovern, who began his campaign from obscurity far greater than any recent nominee, fought through several bruising primaries and was already being accused of being a radical by his own party members before the general election campaign had even begun. The Democratic Convention of 1972, while not as divisive as the party's 1968 convention, finished with a less than ringing endorsement from party leaders for McGovern. Like Goldwater, McGovern was the leader of a faction within the party that, through organizational reforms, was suddenly catapulted into party control.

        McGovern was not as ideologically extreme, however, as was Goldwater, though he may have been a less logical choice of the party. McGovern's placement by voters, shown in Table 3b, does place him as the most liberal candidate of those considered; this data should be treated with caution, however, as it is based upon a question asked during the 1976 ANES, and thus may not reflect views voters had during the campaign. If anything, it may indicate that McGovern succeeded in moving voter perceptions of him to the left during the campaign. His mean ADA rating of 78.7 places him closer to the center than other candidates shown in the ADA data, most notably Mondale, Goldwater, and Humphrey. Indeed, McGovern's post-1972 Senate ratings are generally in the seventies, consistently lower than Humphrey's. Page (1978), in addition, shows that McGovern agreed with the majority of voters on 14 of the 20 major issues of the 1972 election. The NES data, however, show that voters had a different picture of McGovern. Page's CPS issue placement likewise shows that voters believed Nixon was closer to them on six of the seven major campaign issues. (Page 1978: 92-93) News coverage casts McGovern as a "radical" to almost exactly the same degree as it did Goldwater. Virtually every campaign report in Time and Newsweek makes mention of McGovern's dim chances; the only positive references to his chances or the popularity of his positions are confined to particular groups within the electorate -- youth, non-whites, and other groups expected to vote Democratic.

        There appears, then, to be a significant discrepancy between McGovern's presentation of himself and the voters' interpretation of his positions. One explanation of this is that McGovern, like Goldwater, may have been identified with his most fervent supporters. While McGovern was not "drafted" as Goldwater was, McGovern was also a sort of "movement" candidate, representing an insurgency of new participants in the Democratic Party. McGovern had chaired a party committee which reformed delegate selection rules for the 1972 convention; like Goldwater, McGovern was an early beneficiary of the changes he had wrought. The new rules, specifying representation quotas in state delegations for women and minorities, were used to unseat members of delegations from South Carolina, Illinois, and California at the convention; established party leaders such as Chicago Mayor Richard Daley and AFL-CIO boss George Meany (who subsequently refused to endorse McGovern) were denied a vote in the convention.⁽¹²⁾ While the 1972 convention was stage-managed so as to minimize the amount of dissension presented to television viewers, there was sufficient conflict that, like Goldwater, McGovern was forced to spend precious time early in his campaign courting party leaders and mending rifts.

        Richard Nixon was not as seemingly invincible as Johnson had been going into the campaign, according to news reports, but he did have a more moderate record while in Congress and was portrayed in the media as more of a moderate, with no references to unpopularity or radicalism attached to discussions of his campaign. As in 1964, economic growth was up, to the tune of 3.3 percent of per capita GNP. (Tufte 1978: 25) Nixon was able to parlay his powers as President into increasing his advantage over McGovern, negotiating military settlements with China and announcing shortly before the election that peace would soon be at hand in Vietnam. McGovern had no similar platform from which to advance his own promises, and though the Vietnam settlement may have been conveniently timed by Nixon for political ends, it nonetheless pushed all other campaign issues, save Watergate, off the agenda during the final four weeks of the campaign.

        Interestingly, Nixon's most conspicuous move towards the center during the campaign is not tied to any particular issues, but is rather his separation of his own campaign from the general Republican effort of 1972. As each of the years for which NES data exists on voter placement of the parties on a liberal/conservative scale will attest, the positions imputed to candidates are quite similar to the positions imputed to their political parties. Nixon, in establishing the Committee to Re-elect the President (CREEP) was able to minimize his connections to the Republican Party. He did precious little campaigning for Republican members of Congress, as a means of limiting this association somewhat. (White 1973: 276-279) This did not help Republican prospects in Congress; while Nixon was triumphant in his re-election bid, the party picked up only 12 House seates and lost two Senate seats. While CREEP was eventually tied to the Watergate scandal, pre-election news coverage indicates that the initial findings regarding Watergate had a minimal effect upon the voters' decisions.

        Were McGovern's chances in the 1972 election for all practical purposes equal to zero? Miller and Miller (1976) lay the blame for the magnitude of the Democratic loss in 1972 at McGovern's feet, yet again Rosenstone's model casts such a claim into doubt. Rosenstone claims that a more conservative Democrat would have run only 3.8 percentage points ahead of McGovern nationally; such a candidate would have fared 12 points better than McGovern in the South, but this would still have not brought about a significantly better chance of victory. (Rosenstone 1980: 132-136) Again, it must be stressed that in my model parties are not vote maximizers; McGovern was no more of a liability for the party in terms of probability of winning than any other candidate would have been. Whether one considers the stock of available candidates, the state of the economy, or Nixon's popularity ratings, there is little to show that the Democrats ever had any chance of winning the 1972 election.

        Some anecdotal evidence suggests that McGovern and his campaign staff were aware of this. While reports of the campaign revel in second guesses and charges of ineptitude on the part of the campaign managers, they also acknowledge the hopeless odds that McGovern confronted. McGovern himself reports to Newsweek on October 30 that

if we don't [win], maybe some other, more effective political leader will come along and market these ideas better than I can.

        McGovern's reaction to these circumstances is not as readily discernible as is Goldwater's reaction. McGovern's pre-campaign activities on the party reform front indicate that this election was a transitional one for the party, aimed at altering the structure of the party. Following the 1968 election, the Democrats had failed to make accommodations for the Wallace supporters within the existing party structure (Klinkner 1994: 102-103); although McGovern reached out to these supporters following the attempt on Wallace's life that derailed Wallace's 1972 candidacy, he made no policy concessions. (White 1973: 343) McGovern remained active in the party following the election, as did many of his campaign workers,. most conspicuously Campaign Manager Gary Hart, who went on to make a successful bid for Congress in 1974. The Democrats did little to alter their policy proposals following the 1972 election; whether this was through neglect or calculation is impossible to tell, since the policy reasons for Jimmy Carter's 1976 victory are difficult to separate from the electorate's reaction to Watergate. (Klinkner 1994: 130-132)

        McGovern's own strategy during the campaign, however, shows a noticeable lack of effort to engage Richard Nixon on his own terms, to move towards the median on Nixon's issues. As Table Four shows, Nixon emphasized the economy, the military, and foreign policy. McGovern's economic focus was not on national economic performance but on distributional issues; a linchpin of his campaign was his proposal for reorienting the tax code. McGovern concentrated upon the Vietnam War, civil rights and busing (where he held, according to Page's analysis, a minority view), and on social issues such as abortion. The focus on ethics or morals in the closing weeks of the campaign is surely due in part to revelations regarding Watergate, but it bears a striking resemblance to the trajectory of the Goldwater and Mondale campaigns.⁽¹³⁾ In the closing weeks of each of these campaigns, the challengers actually move to sharpen ideological differences, casting the campaign in terms of good and evil.⁽¹⁴⁾

        It may be best to remember an earlier point when ascertaining the strategy adaptations of the McGovern campaign, however. McGovern's supporters included many with no past history in, and hence no stake in the short-run success of, the Democratic Party.⁽¹⁵⁾ News reports reflect this; in news reports, significant numbers of volunteers in the McGovern campaign reflect disappointment over McGovern's attempts to move towards the center, to reconcile differences with party leaders, and to recover from the largest blunder of his campaign, his initial choice, and subsequent dismissal, of Thomas Eagleton as his running mate.

        McGovern's choices for the party leadership were, like Barry Goldwater's, dismissed soon after the election. The rules altering the composition of the 1972 convention, however, endured. Many of the McGovern volunteers went on to be active in the 1976 primaries, many working for the Mo Udall candidacy. (Hadley 1976: 75) Were it not for the upheaval caused by Watergate, we would have a clearer view of the long-term effects McGovern's candidacy, whether consciously or not, had upon the Democratic Party's prospects.
 

        The 1984 election results parallel those of 1972 in many ways. Incumbent Ronald Reagan defeated the Democratic challenger Walter Mondale 59.2 percent to 40.8 percent, sweeping the entire electoral college save for the District of Columbia and Mondale's home state of Minnesota. Mondale received a majority of votes from non-whites, union members, and Jewish voters, but was outpolled by Reagan in every other category. In sharp contrast to the 1964 and 1972 elections, the incumbent's margin of victory was close to uniform throughout the country.⁽¹⁶⁾ What further separates the 1984 election from the previous two elections considered is that Mondale was not seen as an extremist by members of his own party or by the media; Mondale was, in terms of political credentials, arguably the most qualified candidate his party could field. Mondale may have been disadvantaged because of his service as Vice President in the unpopular Carter administration, but he began the race with far higher name recognition and higher national popularity than did either McGovern or Goldwater. While the 1984 was not quite as lopsided as either of these cases, it may be considered an even more devastating landslide precisely because the excuse of having fielded an extreme candidate could not be used by the losing party, nor could the excuse that political neophytes were running the campaign.

        The data in Tables 3b and 4 support the claim that Mondale was not as extreme as the candidates previously considered. His Senate voting record is farther from the Senate median than are those of any candidate save Goldwater, Bush, and Humphrey, but it is roughly in the range of McGovern's. His placement by the voters, however, is closer to the median voter than that of McGovern, of Carter's first candidacy, and of Reagan's first candidacy; it is also not significantly farther from the median than is Reagan's second candidacy. Mondale was actually placed closer to the median by voters than was the Democratic Party. In addition, his positions were, in the eyes of the voters, closer to the median than they had been in 1976.⁽¹⁷⁾ This is a consistent trend in the NES data; all candidates who appear on two or more years' surveys show some degree either movement towards the center.

        The media, while it dwelt extensively on the low chances of a Mondale victory, likewise made few negative (or positive) references to either Mondale's views or Reagan's views in relation to those of the general public. All available evidence points to the conclusion that Mondale was not running an extremist campaign; he was, in fact, closer to the median than Reagan on many of the individual campaign issues, as Abramson, Aldrich, and Rohde (1986: 170) point out. Abramson and his co-authors also show, however, that while voters were more able to see differences in candidate positions in 1984 than they had been in the past three elections, they were less likely to vote based upon their spatial positioning of the candidates and of themselves than they had been in the previous elections. (Abramson, Aldrich, and Rohde 1986: 174-177)

        Does this evidence invalidate the hypothesis of this paper? In part, it does. From a purely spatial perspective, Mondale was handed a golden opportunity by Reagan.⁽¹⁸⁾ Since Reagan's first move was not to claim the median position, Mondale should have staked a claim to that position. Mondale could have even taken a position to the left of the median and still won.⁽¹⁹⁾ This makes it difficult to assess strategic alteration by Mondale, since Mondale did in fact choose a position of 3 when one considers the NES median, not the mean, where Mondale is very slightly farther than is Reagan. Even this tiny and statistically insignificant difference (.05) in the mean might suggest that Mondale either did not position himself carefully enough or ran a risky campaign at the edge of the margin where his P(W) would still be 1.

        Even this narrow discrepancy, however, would still not predict Reagan's landslide victory. In this sense, the model need not be disregarded. The odds against Mondale were high from the beginning. Reagan's voter approval scores never dipped below fifty percent in 1984; the economy was booming, and by all non-spatial measures Reagan was far ahead. (Schneider 1985) Lewis-Beck's model predicts that Reagan would win a tremendous majority of the electoral vote, and, by implication, a majority of the popular vote. (Lewis-Beck 1992: 54) All polls taken during the course of the general election save one (taken immediately after the Democratic Convention) put Reagan in front by up to twenty points. (Schneider 1985: 203) Mondale would not win this election, and he and his campaign staff knew this. (Hunt 1985: 145-150) Given this knowledge, we may ignore Reagan's first move and concentrate only on Mondale's response to this information. His probability of winning would be zero regardless of the strategy he took, so even if his preferences had been much farther from the median than they actually were, he would have still been best served by refraining from courting the median voter.

        As we have done for the Goldwater and McGovern candidacies, let us ask first whether Mondale acknowledged his chances in the campaign, and ask second what his response to this understanding was. Mondale seems initially to have attempted to engage Reagan on each of the issues, as Table Four shows. Reagan was not, however, campaigning on the issues; his campaign was, as Mondale put it in a moment of frustration, a campaign of "picket fences and puppy dogs." (Gillon 1992: 378) Early in the campaign, Mondale attempted to return the campaign to policy issues. Table Four shows that Mondale concentrated his energies on the military and the economy, but he also attacked the President on the area where news reports found Reagan most vulnerable, social issues. As has been the trend in above elections, however, as Mondale's chances sink in, he changes course. Mondale is quoted on September 14 as saying "I may lose this election, but at least I will say something in defeat." (Gillon 1992: 380) Numerous accounts of the campaign assert that Mondale made the most inspired speeches of his campaign after he had acknowledged he could not win. (Gillon 1992: 386-388; Hunt 1985: 162-163; Abramson, Aldrich, and Rohe 1986: 60-61) There is no conspicuous change in issue focus according to Table Four, but campaign accounts refer to Mondale's presentation of a "liberal litany" in the waning weeks of the campaign -- economic fairness and opportunity, civil rights and liberties, and arms control. (Hunt 1985: 162; see also Gillon 1992: 386) Earlier in the campaign, Mondale's economic focus had been on the deficit, which was not a traditional Democratic issue. (Quirk 1985: 174) Now, he reverted to themes designed to rally the constituency. While a certain loser at the Presidential level is rarely a welcome campaign companion for Congressional candidates, Mondale devoted much of the final two weeks of his campaign to assisting embattled Democrats running for the House and Senate. In this regard, he may have been more successful than Reagan -- while Reagan won easily, his party lost two Senate seats and gained only 14 House seats. According to Abramson et al. (1986: 61), none of the candidates for whom Reagan campaigned won.
 

        Whatever foresight there was in Walter Mondale's campaign cannot be judged by the ex post facto analysis that some might apply to the campaigns of Goldwater or McGovern. The Democrats did not win the next election, nor does their 1992 victory owe any obvious debt to Mondale. It does, however, point to a pattern in these three losing campaigns. None of these three candidates show great desire to moderate their views. Where they do show signs of doing such, their efforts are confined to the early weeks of the campaign, before they have given up realistic hope of winning. Each of these campaigns accentuates policy differences in the closing weeks, drawing out philosophical differences between the parties, investigating the long-term prospects for the party, or making claims about the morality of different positions. The candidates do not present themselves to the voters as certain losers, and in fact may for a time have been convinced by non-scientific measures of success such as crowd size or enthusiasm that it was not certain that they would lose. Each candidate acknowledges his slim chances as the campaign progresses, however, and each campaign subsequently shows signs of deviating from the strict Downsian logic.

        Too many analysts of contemporary political campaigns, however, have used the lopsided nature of elections such as these three to dispense entirely with spatial logic. In each of these elections, the challenging candidates had no realistic probability of winning. While I have not in this paper considered reasons for the large incumbency advantage in American politics, I have argued that accommodation for that advantage can be made within a spatial model, and that such modification entails only small modifications within the logic of Downs' theory. Downs saw this, but many subsequent election analysts have not. When winning the present election is no longer a viable option, policy preferences of the candidates are brought to the forefront. This assertion may not satisfy those who see spatial competition as a means of vote maximization, but it does allow us to gain a more realistic understanding of the causes for what may otherwise seem unorthodox and counterproductive campaign strategies on the part of the losers. Not all elections are confrontations between evenly matched opponents, but savvy candidates, and savvy political parties, can and often do attempt to maximize benefits from any campaign.

2. This is a controversial assumption, but not an entirely unprecedented one. Rosenstone (1983: 50) uses it as the basis of his relatively successful prediction model.

3. See Austen-Smith and Banks (1989) for a similar conception.

4. Among the better examples of research in this vein are Rosenstone's (1983: 155-180) expert survey, the Guttman scaling research of Gray (1965), and the measures of issue centrism made by Page (1978: 29-61). These studies demonstrate that (a) candidate positions on issues are highly correlated and thus can have meaning when conflated to one dimension, and (b) the expected convergence to the center does increase as candidate fortunes in the general election increase.

5. For a summary of the method the ADA uses to devise its ratings, see Sharp 1988.

6. See Table Two for an example of the distribution of ADA ratings.

7. We cannot assume that candidates are in all cases familiar with such models. Models have existed, however, since Pool, Abelson, and Popkin's Simulmatics model of the late 1950s; more recent models, each of which makes minimal reference to the spatial positions or other attributes of the challenger, include those of Kramer (1971), Tufte (1978), Rosenstone (1983), and Lewis-Beck (1992). I thus assume that candidate knowledge of economic and other determinants of voting should increase as one moves closer to the present, an assumption also made by Jacobson (1987).

8. Eisenhower and Reagan won renomination unopposed; Nixon faced a challenge from Representative John Ashbrook, but Ashbrook only won 5.0 percent of the primary vote, compared with Nixon's 86.9 percent. The 1964 primary results show numerous "favorite son" candidates for states where Johnson's name was not on the ballot, but none of these candidates mounted a convention challenge to Johnson.

9. Kennedy carried Louisiana, Georgia, South Carolina, and two of Alabama's electoral votes; Virginia Democrat Harry Byrd carried six of Alabama's votes and all of Mississippi's.

10. Skowronek (1993: 339) argues, however, that Goldwater was able to present Johnson as the primary defender of civil rights, thus pushing the Democrats away from the Southern median.

11. Nationally, there was a 2.6% increase in the number of voters from 1960 to 1964; in the Southern states Goldwater carried, however, the increases were:

Alabama 21.0% Georgia 55.4%

Louisiana 10.9% Mississippi 37.2%

South Carolina 35.7%

(source: Congressional Quarterly, Presidential Elections Since 1789)

12. Accounts of this include White (1972)and Mailer (1972).

13. Analyzing Goldwater's campaign poses a similar problem, insofar as many of the ethical issues brought up by Goldwater involve relatively late-breaking findings concerning the conduct of Johnson staffer Walter Jenkins and of Johnson's financial dealings with Bobby Baker and Billy Sol Estes.

14. See the Newsweek issue of October 16 for discussion of this in the McGovern campaign.

15. For a discussion of the role of these types of activists in campaigns, see Stone and Abramowitz 1983.

16. Abramson, Aldrich, and Rohde (1986: 75-85) conclude that the 1984 election had the lowest state-by-state variance of any American election.

17. In the NES candidate placement question for Mondale in 1976, Mondale had a median placement of 3 and a mean placement of 3.02. (ANES 1976)

18. Reagan also made the unorthodox move of accepting Mondale's challenge to debates on the issues, a move which neither Johnson or Nixon made; such a move would actually, argue Light and Lake (1985: 100-101), highlight the contrasting positions of the two candidates.

19. Given the above logic, a position of 3 by Mondale, in reaction to Reagan's position of 6 in the NES model, would have optimized U^P while maintaining a P(W) of 1.

Year  Dem Candidate   Dem Vote %    Rep. Candidate   Rep Vote %   Incumbent's Margin

1940   Roosevelt (I)        55.0                Wilkie                       45.0              10.0
1944   Roosevelt (I)        53.8                Dewey                      46.2                7.6
1948   Truman (I)           52.4                Dewey                      47.6                4.8
1952   Stevenson            44.6                Eisenhower               55.4            - 10.8
1956   Stevenson            42.2                Eisenhower (I)          57.8               15.6
1960   Kennedy              50.1                Nixon                       49.9              - 0.2
1964   Johnson (I)           61.3                Goldwater                38.7               22.6
1968   Humphrey            49.6                Nixon                       50.4              - 0.8
1972   McGovern           38.2                Nixon (I)                  61.8               23.6
1976   Carter                  51.1                Ford (I)                    48.9              - 2.2
1980   Carter (I)             44.7                Reagan                     55.3            - 10.6
1984   Mondale              40.8                Reagan (I)                59.2               18.4
1988   Dukakis               46.1                Bush                         53.9                7.8
1992   Clinton                 53.1                Bush (I)                    46.9              - 6.2

Source:
Calculated from data in
Congressional Quarterly, 1990. Presidential Elections Since 1789.  Washington, DC:
Congressional Quarterly, Inc.

I denotes an incumbent candidate.

Mean margin of victory:  10.1%
Mean margin in elections with incumbent candidates: 13.9%
Mean margin in open elections:  4.9%
Mean margin, 1964, 1972, & 1984:  21.5%

Year Senate Mean Senate Median     Incumbent Mean                          Challenger Mean

1960      44.5                 33                  17.0  (36th most con.)                  70.6 (26th most liberal)*
1964      47.3                 47                  52.2  (48th most liberal)                  1.3  (8th most con.)*
1968      32.9                 29                  84.4  (4th  most liberal)                  17.0 (35th most con.)
1972      38.3                 35                  17.0  (33rd most conservative)     78.7 (15th  most lib.)*
1976      42.9                 45                  22.5  (33rd most conservative)                  No Data
1980      46.3                 44                          No Data                                           No Data
1984      47.9                 50                          No Data                             92.1 (12th most liberal)
1988      48.1                 55                  6.3 (20th most conservative)                     No Data
1992      55.9                 65                  6.3 (8th most conservative)                       No Data

Source:  Sharp, J. 1988.  The Directory of Congressional Voting Scores and Interest Group
  Ratings.  New York:  Facts on File.
Congressional Quarterly. 1990.  Politics in America 1990.  Washington, DC: Congressional
 Quarterly Press.
Congressional Quarterly. 1994.  Politics in America 1994.  Washington, DC:  Congressional
 Quarterly Press.

*  Candidates in question were actually serving in the Senate during the year of their candidacy.
Kennedy received an ADA score of 67 for 1960, tied with 11 others most liberal in the Senate
for the year.  Goldwater scored 0 in 1964, to tie with 6 others for most conservative.  McGovern
received an atypically low score of 45 in 1972.  This does not indicate a moderation of positions;
the score instead reflects the large number of votes for which McGovern was absent.

Incumbent and challenger means are averaged over the length of the candidate's tenure in
Congress.  Incumbents listed are the candidates of the incumbent party, not necessarily the
incumbent President.
 
 

 Year     Incumbent     Challenger         Don't Know         Close Race     Not Close
 1960 42.9% (838)     33.4% (653)     21.6% (423)          80.0  (1197)  14.6 (219)
 1964 81.0   (1272)    8.0   (127)        9.5    (149)            45.2    (633)   44.9 (628)
 1968 23.2     (341)    60.4   (887)     16.4    (241)          23.9    (299)   71.3 (893)
 1972 83.5   (2258)    6.8   (185)        9.5    (259)            35.0    (854)   61.7 (1505)
 1976 41.3   (1185)    43.1 (1234)      15.6    (446)         82.2   (1985)  16.3 (393)
 1980 37.8 (615)        46.4    (750)     12.2    (198)         79.1  (1239)  14.9 (234)
 1984 80.6   (1820)   11.8    (267)       7.3    (165)          47.4   (1070)  46.1 (1040)
 1988 63.5   (1295)   22.8    (466)     13.3    (272)          69.4  (1415)   25.0 (510)
 1992 30.8 (765)       56.3  (1400)     11.0   (273)           78.9   (1960)  17.4  (433)

Source:  ANES 1960-1992.  Figures in parentheses indicate the total number of responses.
Responses of "other" for the prediction-of-winner question are excluded.

Question format:
Who do you think will be elected President in November?
Do you think it will be a close race or will [answer to above] win by quite a lot?
 

Year/Candidate                 1         2         3             4         5           6         7            Median        Mean

1968*
            Humphrey            15.5     29.9     20.2     16.4      9.3       6.5       2.2             3             3.02
1972*
            McGovern            25.2     31.0     16.8     12.1     7.8       5.0       1.7             2             2.70
1976
            Ford                       1.1      5.4       7.4      17.2     31.0     32.8     5.1             5             4.90
            Carter                     6.0     25.8     30.5     20.0     10.4       6.5     0.8             3             3.26
            Republicans            1.0       4.9       8.9      15.8    29.7      33.8     5.8            5             4.93
            Democrats              8.7     33.5     27.9      15.8     7.5         5.8     0.8            3             3.01
            Respondent             2.0     10.1    12.0      37.7    18.5       16.4     3.2            4             4.23
1980
            Carter                     5.9     16.5     21.9     26.8     14.1      11.4     3.4            4             3.74
            Reagan                   1.2       7.2       7.6       9.1     18.6       40.9   15.4            6             5.21
            Democrats              6.0     27.9     26.0     19.8     10.2        8.4      1.6            3             3.32
            Republicans            2.0       4.1       7.4     11.9     25.7       41.2     7.6            5             5.09
            Respondent             2.5      9.3     13.5     30.6     21.0        19.8    3.3            4             4.31
1984
            Reagan                    3.8      8.8       8.7     11.3     14.4       39.5     13.5         6             4.96
            Mondale                  6.8    25.7     22.1     22.3     10.7        9.2       3.2          3             3.45
            Republicans              2.5     8.0       9.2     13.7     21.0       36.3      9.3          5             4.89
            Democrats               7.9     28.2     21.8    19.0     10.5         9.5      3.1          3             3.37
            Respondent             2.3     10.4     12.9    33.4      20.1       18.5     2.3           4             4.24
1988
            Bush                        2.5      4.4       6.7     13.5     22.2        39.9   10.8          6             5.11
            Dukakis                 11.6     28.5     20.2     18.6      9.4          8.3     3.3          3             3.24
            Republicans             2.8       6.8       6.1     13.7    19.6        38.9   12.1          6             5.05
            Democrats               7.1     28.9     23.3     19.0    11.2          7.8     2.7          3             3.33
            Respondent             2.5       7.9      13.1     31.3   21.7         19.5    4.1           4             4.37
1992
            Bush                        3.2       4.5       7.7     15.4    19.3         37.7   12.1          5             5.05
            Clinton                     8.5     26.4     27.4     22.7      8.2           4.8    2.1           3             3.19
            Republicans              2.8      5.2       7.5     14.9     18.6         35.2  15.7          6              5.10
            Democrats                9.3     29.4    25.0     18.0      9.9           5.8    2.6           3              3.17
            Respondent              2.7     11.5     13.4     31.4    20.3         17.1   3.5            4             4.21

Source:  ANES 1976-1992

Responses are scaled such that 1=most liberal, 4=moderate, 7=most conservative.  Answers of
"don't know" or "haven't thought about it much" are excluded.

*  Data for McGovern, Humphrey taken from questions on the 1976 ANES.

                                   Johnson     Goldwater     Nixon     McGovern     Reagan       Mondale

Weeks of Coverage           25                 29            22                27             25              28

Issue Mentions
  Economy                          10                   4              9                13