Candidate Competition: Equilibrium or Disequilibrium

The literature on party realignment often represents party alignments as lines of cleavage in a two-dimensional space such as in Figure 1. See Sundquist 1983, for example. This is consistent with the rational choice literature on spatial modeling, where cleavage lines can be derived from a “deterministic” model of rational choice by voters. The most important theore- tical result from the spatial modeling literature is that no party may take a position that cannot be beaten by a challenging party taking some other policy position in the space (McKelvey 1979; McKelvey and Schofield 1986; Schofield 1978, 1983). Given any party position such as D_a in Figure 1, there will in general exist a set of alternatives, called the win set, each of which can defeat D_a. The opposition party, by taking position R_c in the win set of D_a should be able to shift its coalition of support to defeat an incumbent.

Candidates, according to the literature on political parties, are interested in winning elections (Schlesinger 1994). An unconstrained challenging candidate trying to defeat an incumbent identified at location D_a would move the party position from R_a to location R_c. The standard spatial modeling assumption is that the party's position is unilaterally chosen by the candidate. That assumption, together with the assumption that the candidate always wants to win elections, leads to the conclusion that two-party competition should be a chaotic series of ever-shifting party coalitions.

Although the recent literature on realignment (Brady 1988; Carmines and Stimson 1989; Huckfeldt and Kohfeld 1989) acknowledges the possibility of shifting party positions, it has not considered the likelihood of permanently chaotic party configurations. This paper regards party realignments as being a rather controlled manifestation of underlying two-dimensional electoral instability.The “instability theorem” alluded to above assumes that voters choose “deterministically” (by, for example, voting for the candidate nearest the voter “bliss” point). An alternative theory “smooths” voter preferences by assuming that each voter is “stochastic.” That is, each voter is described by a probability vector; the nearer the candidate is to the voter, the higher the probability that the voter will choose that candidate. Under the assumption that candidates choose positions to maximize expected vote shares, and accepting a number of reasonable conditions, it can be shown that candidates converge to the mean of the distribution of the voter ideal points. The existence of such pure strategy Nash equilibria (PSNE) is discussed in Banks and Duggan 1999, Coughlin 1992, Enelow and Hinich 1984, and Lin, Enelow, and Dorussen 1999. This convergence parallels the equilibrium result of Downs (1957) in one dimension.³

In contrast to the assumption of these models that candidates are unconstrained, it is evident that candidates in U.S. presidential elections are, to some degree, constrained by the necessity of winning primaries, raising funds, and mobilizing volunteers. These constraints are associated with party activists, who have different goals from candidates. The next section describes the Aldrich model, in which party “location” in policy space is determined by the decisions of party activists.

Party Activist Equilibrium

As noted by Schlesinger (1994) and Aldrich (1983a, 1983b, 1995), activists are less concerned with winning elections than with maintaining the ideological stance of the party. “The political role [of party activists] is to attempt to constrain the actual leaders of the party, its ambitious office seekers, as they try to become the party-in-government by appealing to the electorate” (Aldrich 1995, 183)

Aldrich (1983a, 1983b, 1995) argues that party activists, not candidates, play the primary role in creating the public's perception of the “location” of the party in ideological space. Voters take the location of the “average” Democratic and Republican activists as their cues; those voters who find that the location of one party's activists is much closer to their own preferences than that of the other party may decide to become activists themselves. For example, if a voter finds that the average Democratic activist has policy preferences very like her own, she is more likely to enlist as an activist. In doing so, she may “move” the location of the average Democratic activist slightly in the direction of her ideal point—which may include a strong prochoice position, for example. In that way, a pro-life activist may find himself at too much of a distance from the gradually moving typical Democratic activist position and resign. Both the enlistment of the pro-choice activist and the deactivation of the pro-life activist have the effect of moving the public's identification of the Democratic Party. At the same time, as the Democratic Party becomes more pro-choice, the strongly pro-life voter may decide to become a Republican activist, initiating a similar adjustment of the Republican party.

Aldrich's contribution is to show that the aggregation of such decisions to join or exit the cadre of party activists may constitute stable, activist equilibrium in which parties maintain a certain distance in the ideological space, as illustrated in Figure 1 with D_a and R_a. Activists preserve their ideological distance and, at the same time, create stability in partisan alignments. In Figure 1, the domains N_D and N_R are used to denote the support of ideal points of those voters who choose, in Aldrich's model, to become activists for the Democrats and Republicans, respectively. Note that the radii of N_D and N_R are determined by parameters of the model, in particular, by the cost, c, that each activist chooses to pay in support of the chosen party. In Aldrich's model, D_a and R_a are, respectively, the average policy positions taken over the prefererred (bliss) points of Democratic and Republican activists. Aldrich's model has the great theoretical virtue of predicting neither chaos nor convergent equilibrium.

The constraint implicit in this model can be illustrated in Figure 1. While a move from R_a to R_c might be favored by the challenging candidate as a winning move, it is one that would be viewed by economic conservatives as “selling out.” The economic conservatives will inevitably play a key role in a party alignment (such as the New Deal alignment), in which parties are differentiated primarily on economic policy. They will feel that a great deal is at stake in the difference between the parties and will be most likely to vote in primaries, donate money, and engage in canvassing. During the forties and fifties, for instance, Republican activists were professional and small business entrepreneurs who were activated by the classic party conflicts over unionization, nationalized medicine, and taxation. Democratic activists in that period were much more likely to be active union members and farmers, with an equal stake in an economically liberal stance by their party.

What control do party activists have over the public perception of the party? First, they will play a major role in selecting the party candidate. Time and again since 1964, we have seen a majority of party activists throw their support for more extreme over more centrist candidates in primary elections. In 1964, it was Barry Goldwater who defeated William Scranton of Pennsylvania. In 1972, George McGovern won the nomination in the primaries with a liberal position that proved disastrous at the polls. In 2000, centrist John McCain's popularity with Independents and moderate Republicans proved insufficient to defeat George W. Bush when traditional conservatives rallied to the latter's side. Furthermore, candidates, once selected, must act to keep party activists happy enough to continue to make contributions of money, time, and effort.

It appears that there is a great deal of evidence for Aldrich's argument. First, there is the empirical observation that in the United States (and other political systems), parties do not converge to the center (Adams and Merrill 1999 or Poole and Rosenthal 1984). Furthermore, party activists in the United States tend to be more ideologically extreme than the average Democratic or Republican voter. Finally, political parties do not constantly shift and realign as would be predicted by a model of candidate vote maximization in two-dimensional space. In fact, the “disciplining” force of party activists can be seen in virtually every election.

However, while Aldrich's model rationalizes nonconvergent activist equilibrium, it fails to do two things. First, it does not contain party candidates—who do want to win elections first and foremost. Second, it does not explain how party activist equilibria are themselves disrupted. For it is clear that occasionally party positions do change, in ways that often dismay traditional party activists and disrupt existing party activist equilibria. In 1964, traditional Republican activists were upset by the nomination of Goldwater; in 1972, many traditional Democratic activists dropped from the rolls of party activists rather than support a more socially liberal Democratic candidate. Evidently, there are destabilizing forces that occasionally disrupt party activist equilibria as described by Aldrich. The purpose of the next section is to provide a formal model, building on Aldrich (1983a, 1983b) and Aldrich and McGinnis (1989), to examine the forces that party candidates may exploit to try to win elections.

A Joint Model of Activists and Candidates

Essentially, the model is a dynamic one based on the willingness of voters to provide support to a candidate. Given current candidate strategies (z), let C(z) = [C₁(z), …, C_p(z)] be the current level of support to the various candidates. This support is important to candidates, we assume, because it allows them to have an impact on the nonpolicy level of support of voters—which has been called valence (Ansolabehere and Snyder 2000; Groseclose 2001; Schofield 2003). We suggest, in contrast to the usual assumptions, that valences consist of two components. For each candidate, j, there is an “innate” valence. We suggest that this is best characterized by a stochastic error term, ε_j. The second component, λ_j, is affected by the money and time that activists make available to candidate j. Essentially, this means that the valence component, λ_j, is a function of the policy choices of all candidates. This implies that we model voter utility by the equation

where x_i is voter's i's preferred policy and z_j is candidate j's strategy, both in a policy space, X. We make an important modification of the spatial model in that the salience of different policy dimensions varies among the electorate. More precisely, we assume that

Here | |_i is an “ellipsoidal” norm giving a metric whose coefficients depend on x_i, as specified further below. The spatial term, A_ij, we call the nonvalence or policy component of voter utility.

Candidates deploy their resources, via television and other media, and this has an effect on the vector λ = (λ₁, …, λ_p) of candidate-dependent valences. We assume that the nonstochastic valence component, λ_j, is a function of C(z) and thus z.

At this point, a voter, i, may choose to be an activist by adding her own contribution c_ij > 0 to candidate j subject to the condition

Aldrich (1983a) considered an equilibrium of this dynamic process between two candidates, 1 and 2, where the candidate's position, z_j, was defined to be the mean of the ideal points of all activists who supported this candidate.

Consider now a group of Republican “economic” activists. The Republican candidate, j, is situated at the position R_a = (x,y), while the activist has a utility function given by

The activist contributes some amount, c_ij < u_ij(x,y). For activists who are more concerned about economic issues, it is natural to assume that a < b. If the activist actually had “bliss point” (s₁, t₁) = R_a, then his indifference curve would be given by the “ellipsoid” centered at R_a, as in Figure 2. Depending on various parameters, there will exist a domain, N_R, say, in X, with the property that every voter whose bliss point (s_i, t_i) belongs to N_R is a contributor to the Republican candidate. It is natural to assume that there is an opposing Democratic candidate, whose position is at D_a, say, and an opposing set, N_D, of Democratic activists.

Aldrich showed, essentially, that these conditions could be satisfied, such that R_a was given by the mean of the bliss points of the set N_R, while D_a was the “mean” of the set N_D. It is obvious that for such an activist equilibrium to exist, it is necessary that λ_j, regarded as a function of campaign contributions, is concave (or as has diminishing returns) in contributions to candidate j.

Because candidates can transform activist contributions into non-policy related votes, they cannot be indifferent to the policy preferences of those who are motivated to be activists in a given existing activist equilibrium. On the other hand, as vote maximizers, they may be willing to trade off current activist support for the support of other potential activists—hence the much-documented tension between party activists and candidates. How might candidates (who are passive in the Aldrich model) go about finding a policy position that induces an optimal corps of activists?

As Figure 2 indicates, a typical socially motivated voter would regard Democratic and Republican candidates during the New Deal as equally unattractive, and tend to be indifferent. Civil rights supporters, for example, were frustrated by FDR's refusal to support an antilynching law and, occasionally, threatened to “sit out” elections. Similarly, Dixiecrats walked out of the only New Deal-era Democratic convention (1948) that took a strong pro-civil rights position.

Let us now suppose that such a socially conservative voter, g, has bliss point (s₂, t₂), say, near the position I_a, with utility function

Let N_C be the set of such “disaffected” social conservatives who would be willing to contribute to a candidate as long as this candidate adopted a policy position close to his bliss point—especially if the social policy position was close to t₂. We suggest that such social conservatives regard social policy to be of greater significance and so e > f in Eq. (5).

Unlike Aldrich, we now suppose that the Republican candidate adopts a position, not at the mean R_a, but at some compromise position between R_a and I_a. It is easy to demonstrate that the “contract curve” between the point (s₁, t₁) and the point (s₂, t₂) is given by the equation

where s = [b²/a²][e²/f²]. We use the term “catenary” to describe this curve.

If the Republican candidate moves on this locus, then the resulting number of activists will be of the order μ N_R + (1 − μ)N_C, where μ is some constant (less than one) dependent on the position taken by the candidate. Because of the asymmetry involved, the total number of activists may increase, thus increasing overall contributions to the Republican candidate. Clearly, there are plausible conditions under which λ_j increases as a result of such a move by a Republican candidate, thus increasing the effective vote share of the Republican candidate. That is, the Republican candidate's desire to win elections can cause him to make a move that will transform the existing party activist equilibrium.

Determination of the existence of a candidate PSNE depends on continuity and quasi-concavity (or concavity) of the candidate utility functions U_j. While each U_j will be a function of z, its dependence on z will be more complex than the simple relationship implicit in the standard spatial model. It is important to note that this proposed model involves differing voter utility functions. To preserve continuity of voter response, it is necessary that the coefficients of voter policy loss vary continuously with the voter-preferred policy. This can be accommodated by requiring that the function A_j(z_j) : X →R given by

is continuous. This means that the salience parameters change continuously with the bliss point of the voter.

With these assumptions, candidate vote share functions {V_j} will be continuous in (the vector of) candidate strategies. Candidate utility functions are generally derived from the vote share functions, and so quasi-concavity or concavity of the candidate utility functions and thus the existence of PSNE can then be shown (see Schofield [2002, 2003] for the technical argument). It is worth emphasizing that the greater the relative salience (b/a) and (e/f), the greater will be S, and thus, the more significant will be the attraction of building a coalition with dissident activist groups to enhance electoral support.

The mixed activist–voter model suggests why a transformation in activist-generated equilibria can occur. In contrast to Aldrich, we suggest that the policy positions of the two parties are chosen by “expected” vote-maximizing candidates. The voter model itself is of the standard probabilistic variety (Enelow and Hinich 1985; Lin, Enelow, and Dorussen 1999). However, the novel feature is that the voter calculus involves a nonpolicy, valence variable, associated with each candidate. This nonpolicy variable is a monotone increasing, concave function of the respective party activist total contributions. It is an immediate and obvious feature of this model that candidates do not converge in Downsian fashion to the center of the electoral distribution. Instead, a “rational” candidate will choose a policy position so as to “balance” activist contributions and voter responses. (For an empirical analysis based on this formal model, see Schofield, Miller, and Martin 2003) In this variant, potential activists who place high valuation on policy shifts can offer significant contributions and, thus, “affect” rather than “control” candidate locations.

Note, also, that if one of the presidential candidates initiates a flanking move of the kind described, then the opponent should move even farther round the opposing catenary. We can illustrate the existence of the PSNE using Figure 2 as a guide. Let R_a′ and D_a′ be the positions of the Republican and Democratic candidates that maximize their total contributions from activists. These positions will be on the respective catenaries and will be independent of each other (because activist contributions will be determined solely by the respective candidate positions). At these positions the “valence” effects, now labeled λ_R and λ_D, will be maximized. Draw the arc between the position R_a′ and the position D_a′. If the Republican candidate moves on this arc, toward the origin, λ_R will fall, but the overall policy or spatial component of the vote share may rise. Concavity of the vote share function means that there will be a unique point on the arc where the Republican candidate's vote share is maximized. Indeed, by the implicit assumptions of continuity and concavity, the Republican candidate will have a continuous best response to the Democratic candidate's position. Mutual best response defines the PSNE. This PSNE can also be characterized by the partisan cleavage line, which separates those voters with a higher probability of voting Republican than Democrat. Note that if liberal social activists attract the democratic candidate to move toward them on the liberal catenary, then the best response of the Republican will be to move farther toward the conservative social activists. (More details on the existence of PSNE can be found in Schofield, Miller, and Martin 2003)

The catenary or contract curve (and thus the best response position) depends on the ratio of the intensities of relative preferences on the two axes associated with the activist groups. For example, if economic activists care relatively very strongly about the economy, and social activists care greatly about social policy, then the optimum position will be extreme on both axes. In a very intuitive fashion, the model allows for response to the relative numbers of the activist groups, their intensity of preferences, and their “willingness” to contribute. A party's candidate should be more willing to seek to move farther down the catenary in an attempt to enlist the social activists when the social activists are larger in numbers and when the intensity of their preferences induces a greater willingness to contribute.

Third Parties

Third parties play a particularly interesting role in this model. They represent voters who are particularly concerned about issue dimensions that are suppressed by the existing party alignment. For example, in 1968, Wallace represented those who were alienated from the Democratic Party by its sponsorship of the 1964 Civil Rights Act and the 1965 Voting Rights Act. In 1935, Franklin Roosevelt was moved to consolidate the economic liberal position of the New Deal to avert a third-party threat from the economic Left. In 1980, Anderson ran a campaign appealing to upper-income, college-educated liberal Republicans on the coasts who were upset about the socially conservative course that Reagan was setting for the GOP. In 2000, Ralph Nader insisted that both major parties were in thrall to corporate elites and represented economic liberals who felt that the “culture wars” competition between Democrats and Republicans had left New Deal liberalism behind.

Once organized, such third parties demonstrate to party elites the electoral advantage of the “flanking maneuver” described in the model above. In the runup to 1972, the attraction of the millions of socially conservative Wallace voters rationalized Nixon's “Southern Strategy” and exerted a powerful tug on Nixon toward point I_a in Figure 2. Anderson's third-party run in 1980 offered the prospect of disaffected socially liberal Republicans whose votes were up for grabs; many of these became Clinton Democrats in the nineties.

The Wallace candidacy of 1968 and the Anderson candidacy of 1980 represented two distinct groups of activists who had quite distinct perspectives about a reorientation of the Republican Party. The Wallace candidacy illustrated what we call a leading third party for the Republicans in the seventies, since it served to attract Nixon and other Republican elites toward a new party position that emphasized conservative social policy rather than the historic economic conservativism. In contrast, the Anderson candidacy represented traditional New Deal Republicans who were dismayed to see the Republican party moving away from their position; consequently, we can call the Anderson candidacy a dragging third party. More recently, the Nader candidacy was quite consciously a dragging third party for the Democrats, as Nader objected that the Democratic Party had in the nineties abandoned the traditional economic liberalism of the New Deal and was, in the opinion of his supporters, indistinguishable from the Republican party in its support of big business.

Summary of the Model

The elements of the model are as follows.

Some voters are motivated by the policy proximity of one set of party activists (and the distance from the other party activists) to join the cadre of activists. Each decision to exit or enter the cadre of activists changes the mean location of the activist core and, thus, the public perception of the locations of the parties. Stable divergence in the parties' public image is the result of what Aldrich calls “activist equilibrium.”

Candidates, unlike party activists, are primarily interested in winning elections; consequently, each candidate has an incentive to try to move the public's perception of his party to a position that maximizes the expected vote.

At any point in time, a candidate is likely to find that current party activists are intensely concerned with one dimension, while a “disaffected” group is more intensely concerned with a different dimension of ideology. Consequently, the candidate may reasonably hope to make a “flanking” move that appeals to one group of disaffected voters without losing too many party activists. When successful, this initiates a shift in partisan activists and, therefore, a redefinition of the public's image of the “relocation” of the parties.

For any given party alignment, “disaffecteds” may try to hasten a realignment by means of a leading third-party attempt. The success of a leading third party is often the excuse that a vote-maximizing party candidate is looking for, to destabilize an existing party activist equilibrium. Party activists who are disgruntled with the previous activist equilibrium may form the core of a dragging third-party attempt.

From this perspective, the partisan realignments of the past century have been fundamentally linked to the multidimensionality of the potential policy space. They have been initiated by “flanking movements” rather than by frontal assaults. Rather than fighting toe-to-toe for the moderates in the exact center of the space, candidates have tried to appeal to disaffected voters in the dimension that has recently not distinguished the two parties.

Of course, an outward thrust on a party's left flank leaves its own right flank exposed and vulnerable, at least in the long run. If this is the case, then each attempt to build a new majority party sows the seeds of the next party realignment.

It is important to note that we hypothesize a voter–activist equilibrium at each election, conditional on the various parameters of the model. However, this does not necessarily imply that the sequence of equilibria will vary smoothly over time. Each equilibrium is determined by candidate calculations over the relative “value” of “activist” and “disaffected” coalitions (and thus by the configurations of the utility functions of such actors). These parameters may shift dramatically as the result of exogenous social and economic shocks.

In the rest of this paper, the implications of this model are used to provide an interpretation of the past century of partisan shifts in ideology.

The Decline of Race and the Rise of Class: 1896–1960

While the period of 1960 to 2000 saw the undoing of the New Deal partisan cleavage, the period from 1896 to 1960 saw the creation of that economic cleavage from a system that had been primarily divided over the issue of race, civil war, and reconstruction. After the Civil War, the Republicans were most clearly associated with the successful emancipation of slaves and the less successful Reconstruction of the South on an integrated basis. The Republican coalition was united on a social policy dimension. Some Republicans, like Lincoln, were already pro-business advocates, but some Republicans were emancipationists who had joined the party for social liberal reasons only. After the Civil War, the Republican Party continued to include liberal supporters of Reconstruction, new black voters, and what historian Eric Goldman calls “patrician dissidents,” also known as “goo-goos” and “mugwumps,” who resisted the increasing influence of industrialists in the Republican Party (Goldman 1956, 16).

The two wings of the Republican Party had no strong differences of opinion on social policy: They did have markedly different positions on economic policy. They were kept in alliance by frequent electoral “waving of the bloody flag”—references to the losses of the Civil War that served to reaffirm the Civil War issue party alignment. As a result, the Reconstruction Republican position was socially liberal but relatively neutral on the economic dimension.

The Democrats had to bear the burden of the Civil War legacy; they also took the racially conservative position that allowed the gradual reimposition of white supremacy in the South. They represented largely rural interests and opposed the tariff, but Democrats, like Republicans, included both pro-business and anti-business forces. Cleveland was a Democrat who gave aid and comfort to business forces. He supported the gold standard, kept tax burdens low for corporations, and fought to keep the government out of economic life, especially where it might benefit low-income groups. He used federal troops to help the Pullman Company put down its desperate workers during their famous strike, which made fellow Democrat William Jennings Bryan charge, “Cleveland might be honest, but so were the mothers who threw their children in the Ganges.” (Goldman 1956, 33) Since each party combined economic liberals and conservatives, the cleavage line between the parties was best thought of as a horizontal line separating the party of Reconstruction from its socially conservative opponent.

Who were the “disaffected” voters in the Reconstruction Era? Both economic liberals and economic conservatives had reason to feel inadequately represented in the Reconstruction party alignment. Economic liberals were necessarily agrarian interests, who formed various agrarian interest groups, culminating in the Populist Party, to fight what they saw as the stranglehold of Eastern banks and railroads on the pocketbook of the small farmers.

The creation of the Republican majority after 1896 carried with it a long-term opportunity for the Democratic Party. After 1896, Republicans were increasingly identified with a pro-business economic position. As that issue became more salient, the Bull Moose Progressives under Theodore Roosevelt were in an increasingly disaffected position. Because the Republicans had made the pro-business position the defining position for the Republicans, the Progressives' advocacy of economic regulation and socially liberal positions (social welfare agencies, public health, pure food and drug) made them the obvious target for Democratic cooptation.

Up until the 1920s, the Democrats had been primarily a rural party and, therefore, an unacceptable alternative for the largely urban Progressives. But during the twenties, Al Smith began to push the Democratic Party to take positions that appealed to Catholics, Jews, immigrants, and the new urban working class. Al Smith and, in particular, FDR were more persuasive wooers of urban liberal Republicans than Bryan had ever been. The 1928 Democratic convention marked the beginning of a new Democratic coalition that combined the Bryan populists (Region A) with urban liberals (Region D).

Although Roosevelt was elected in 1932 as a moderate, and tried for two years to maintain a centrist position, the potential for third-party candidates drove him to finish the economic radicalization of the Democratic Party after 1934. Roosevelt viewed the principal threat to his reelection as being a leftist candidate such as Huey Long. In the Second New Deal, in the two years before his reelection in 1936, Roosevelt moved decisively leftward to forestall such a third-party attempt. In the process, he consolidated the economic orientation of the New Deal coalition, bringing the liberals from region D into alliance with the rural Southern Populists of Bryan. The New Deal pulled such old Bull Moosers as Harold Ickes out of the Republican Party and into the administration (Fine 1976, 392), where they were perfectly comfortable. In addition, they pulled black voters, urban ethnic minorities, and other Republican constituencies, making the Democrats the majority party for the middle third of the twentieth century.

The cleavage line between the two parties became, for the first time since before the Civil War, a vertical, class-defined boundary (as in Figure 1). Huckfeldt and Kohfeld (1989) document that during the New Deal, class was a good predictor of partisan voting in a way that it was not before or after the New Deal. The difference between the proportion of working-class whites voting Democratic and that voting Republican peaked in 1948 at about 44 percentage points. By maintaining the New Deal coalition, Democrats were able to win every presidential race from 1932 to 1964, with the exception of those won by Eisenhower, the World War II hero.

However, the price of being a majority coalition in a two-dimensional world is having to deal with the huge policy differences within different elements of the party. Just as the Republicans of 1912 had faced a split between the Progressive Republicans and the conservatives, the class solidarity of the Democrats did not eliminate the social policy differences between Southern segregationists and Northern liberals. The contradictions were already apparent in 1948, when Hubert Humphrey's riveting call for racial justice divided the Democratic Party. These contradictions presented the inevitable opportunity for Republicans to put together a countercoalition by further pivoting the cleavage line between the parties.

“The Decline of Class and the Rise of Race”: 1960 to 2000

The New Deal coalition put together by Franklin Roosevelt necessitated the suppression of the social policy differences between the racial conservatives in the Solid South and the racial liberals of the North. This coalition had faltered in 1948, when the Democratic Party split over a civil rights plank and the Dixiecrats ran as a third-party. But in 1960, the coalition had been patched up, and the parties were still primarily differentiated by economic ideology and class-based voting. As late as 1962, most of the public (55.9%) saw no difference between the two parties on civil rights, and the rest were evenly split (Petrocik 1981, 135–38) In April 1963, it was still only 4% of the population that felt that civil rights was the most important issue (Gallup 1972). It is no wonder that Kennedy felt he might be reelected the next year by the old New Deal coalition, including the Solid South.

However, after 1962, civil rights leaders succeeded in their policy of destabilization—breaking up the New Deal coalition by forcing the Kennedy administration to choose the side of federal law or state segregation in schools, interstate travel, and voter registration. After the success of the Birmingham protest in May 1963, Kennedy became the first Democratic president to ask Congress for a strong civil rights bill. By October of the same year, after the Birmingham bombings, Wallace's standing in the doorway at Tuscaloosa, and the March on Washington, the Gallup Poll revealed that 52% of the public felt that civil rights was the most important issue facing the country (Gallup 1972). Lyndon Johnson was convinced that he had to have a civil rights success to lead the Democrats to victory in 1964; by his actions in 1964 and 1965, he reached out to civil rights activists and brought them definitively into the Democratic Party. The New Deal coalition was shattered.

In Figure 2, Johnson's move is shown as an upward move from D_a to D_a′, not a move toward the center. Notice the logic of this flanking move. A move toward the center would have alienated the traditional New Deal Democratic activists—the economic liberals of labor and the consumer movement. However, an upward move gave no sign of alienating the party activists most concerned about economic policy. The Democrats' labor allies were solidly behind the Civil Rights Act. The upward move consequently succeeded in keeping economic liberals who were inclined toward social liberalism. At the same time, an upward shift earned the loyalty of the previously disaffected civil rights workers, who were to become a principal component of the post-1964 Democratic activist cadre, in a new, postrealigment activist equilibrium. In terms of the model, Johnson was able to earn the valence benefits of both economic and social liberal activists.

The shift in the public's perception of the Democratic Party created a huge landslide in 1964. Not only did Johnson win the White House by a margin of 16 million votes, but the Democrats added two seats to their majority in the Senate and 48 seats in the House. No member of the House who had voted for the 1964 Civil Rights Act was defeated in either party. Half the Northern members who had voted against the bill were defeated (Branch 1998, 522). Civil rights activists had succeeded in their goal of forcing the Democratic Party off its New Deal equilibrium.

The success of the civil rights activists in forcing the Democratic Party to take a stronger civil rights position had the kind of disequilibrating effect that Aldrich (1995) hypothesized on the New Deal party activist equilibrium: As more civil rights activists became involved in politics, it encouraged more social liberals to become activists and drove social conservatives out of the Democratic Party. It moved the center of gravity of the Democratic activists upward. These social activists had other social concerns as well as civil rights, which became an agenda for the Democratic Party: women's rights, civil liberties, consumerism, enviro- nmentalism.

As Aldrich hypothesized, this also had implications for Republican activists. Let us consider the kind of person who had been a New Deal-era Republican activist. Presumably, her ideal point is near R_a (in Figure 2). In addition, however, she is probably more intensely concerned with economic policy than social policy, since for decades it had been only the economic dimension that differentiated the two parties. The party differential on social policy had been zero. This is indicated in Figure 2 by the ellipsoidal indifference curve for the Republican activists.

From Goldwater through Reagan, the strategy of Republican candidates was to build a coalition along the contract curve between economic and social conservatives. Barry Goldwater clearly served to destabilize the party activist equilibrium in 1964. In July of 1964, Goldwater was one of only eight non-Southern senators to vote against the Civil Rights Act. A few weeks later, the Republican National Convention refused to seat the traditionally black Southern delegates to the the Republican convention and nominated Goldwater to be the Republican candidate. It was a disruptive shock to the GOP and to the public's perception of what the GOP stood for (Branch 1998, 403; Carmines and Stimson 1989).

Goldwater won the electoral votes of five states of the Deep South in 1964, four of them states that had voted Democratic for 84 years (Califano 1991, 55). He forged a new identification of the Republican party with racial conservatism, reversing a century-long association of the GOP with racial liberalism. This in turn opened the door for Nixon's “Southern strategy” and the Reagan victories of the eighties.

A small initial success with social conservatives could lead to a positive response by some social conservative activists; their arrival in the party could alienate some of the social liberals within the Republican Party, who would become less active; this in turn would lead to a closer identification with social conservativism, which would encourage more still social conservatives to become involved (Aldrich 1995, 184–85). At the same time, social liberals might be increasingly tempted to become activists for the Democratic Party. As Carmines and Stimson (1989, 188) argue, the “Goldwater gamble” worked for the Republican party: “It did break the tie of southern whites to the Democratic party.”

The success of Goldwater in 1964 induced Wallace to run as an independent in 1968. Wallace attempted to mobilize voters on the newly salient social policy dimension. Wallace's campaign earned votes not only from Southern whites but from Northern social conservatives who were concerned about riots, court-ordered busing, the sexual freedom of the sixties, protests against the Vietnam War, and the breakdown of traditional values. Many of the same Milwaukee Polish Catholics who had voted for Kennedy over Humphrey in the 1960 Wisconsin primary voted for Wallace over Humphrey in the 1968 presidential election.

In a biography of George Wallace, Carter (1995) argues that Wallace and the voters he represented were a continuing obsession with Nixon during his first term, 1968–72. During this time, Nixon engineered various rapprochements with Wallace individually while appealing to potential Wallace voters with a strong “law and order” position and, especially, an anti-busing policy stance. The result was a smashing triumph of Nixon and the “Southern Strategy” over McGovern in 1972.

By the decade of the seventies, class had been largely displaced as the organizing principle of two-party competition in the United States, and “race conflict [was] the major new element in the party-system agenda …. [T]he most visible difference between the party coalitions that entered the turbulent 1960s and those that exited in the middle 1970s is to be found in their new-found distinctiveness on race-related policy issues” (Petrocik 1981, 148–49).

While Reagan was successful in keeping the coalition of economic and social conservative activists in the Republican camp, the tensions between the two were increasingly apparent during the eighties. Hostility between Southern populists and Eastern business interests is an ancient tradition in the United States. In a GOP that attempted to keep these two warring camps in the same party coalition, economic issues are as divisive as racial issues were for the New Deal Democrats. The old Rockefeller Republicans, especially the social liberals, felt increasingly alienated within their own party. Fortune magazine, the vehicle for big business, ran a front cover suggesting what they heard as the message from the Republican party: “GOP to Business: Drop Dead?” The article inside said, “In a political arena dominated by small-business populists, anti-government conservatives, and the religious right, corporate America's the odd man out—mistrusted, resented, impotent” (Kirkland 1995, 50). In this article, big business announced that they increasingly felt themselves the “disaffected” voter in a party alignment based more and more on social policy.

It was this fault line in the Republican Party that Clinton was able to take advantage of in the nineties. Many cosmopolitans, while historically loyal to the Republican Party, were frightened of a GOP that was hostile to affirmative action and abortion rights and aggressively pushed traditional social values over economic conservatism. Clinton's actions in the midnineties—especially welfare reform, a major crime bill, support for NAFTA, a balanced budget—upset many of his liberal supporters but made many upper-income social liberals comfortable voting for him in 1996. Clinton was acknowledged to have moved right on the economic dimension, while preserving the Democrat's position of social liberalism (see Figure 3). Clinton made it legitimate for a professional suburban homeowner with a six-figure income, concerned about taxes, crime, and welfare fraud, to vote Democratic.

The special prosecutor's investigations into the Clinton sex scandals had the perverse (for the Republicans) effect of further helping to drive group C cosmopolitans into the Democratic fold. While much of the new Republican party was convinced that the public would be as outraged as they were by the Clinton scandal, it became apparent that the Moral Majority was in fact a minority. Social liberal Republicans and independents did not want to be a part of Starr's sexual inquisition. The impeachment issue became a defining moment for the Republican party in much the same way that the Civil Rights Act of 1964 was a defining moment for the Democrats: It identified clearly which voters were now vulnerable to the opposition party.

The extent of the realignment is shown by the shift in voting behavior on the part of cosmopolitans and populists. As shown in Table 3, cosmopolitans voting for Democratic congressional candidates rose from 46% in the seventies to 65% in the decade of the nineties. The percentage of low-income moral traditionalists, whom we would identify as populists, voting for Democratic congressional candidates dropped from 63% in the seventies to 29% in the nineties. The fact that this statistic is for congressional voting rather than presidential voting, and over five elections in each decade, suggests that the trend is broader, deeper, and more lasting than simply a feeling based on Clinton or other particular presidential candidates. Low-income social conservatives have left the New Deal coalition for the Republican Party, and high-income social liberals are increasingly comfortable voting in the Democratic Party (Smith 2002).

A number of political scientists have documented how partisan voting has evolved since 1960. According to Huckfeldt and Kohfeld (1982, 2), party voting has been characterized since the New Deal by “the decline of class and the rise of race.” They point out that the difference between the proportion of blacks and that of whites voting Democratic rose sharply, from less than 25% in 1960 to more than 50% in 1968 (3). This trend has continued: Democratic presidential candidates have never won more than 47% of the white vote, or less than 83% of the black vote, since 1976 (Connelly 2000, 4). Even a winning candidate like Clinton in 1996 won only 43% of the white vote but 84% of the black vote. Social liberals are now the core of any winning Democratic coalition, and in 2000, social conservatives were the core of the winning Republican coalition.

Choices, Credible Commitment, and Path Dependence

In considering Figure 2, we might wonder whether a Republican candidate has a choice about appealing to disaffected social activists of either the liberal or the conservative persuasion. Could Nixon, for example, have appealed to civil rights activists rather than to Wallace supporters in 1972? Was there an equally viable “Northern Strategy” that would have substituted for his “Southern Strategy” in that year? After all, in 1957–58, Nixon had hoped to earn support from black voters by positioning the Republican Party as a backer of a civil rights bill (Mann 1996).

We argue that candidates are constrained by recent historical events that introduce an asymmetry in the calculations of party candidates. In 1972, for example, Nixon could not credibly make a claim for pro-civil rights activists. That option had already been ruled out by the public's awareness of the early changes in the cadre of Republican activists in the 1964 national convention and their selection of Goldwater as the candidate in that year. As recently as 1962, the public felt that Republicans were equally as likely to see that African Americans got fair treatment in jobs and housing (21.3% for Republicans, 22.7% for Democrats, 55.9% seeing no difference in the two parties). But by late 1964, the Democrats had already earned the civil rights advocacy reputation by 56% to 7%, with 37% seeing no difference (Petrocik 1981, 135–38). A determined Republican could at best have neutralized the civil rights issue, without constructing a coalition. But by 1972, Nixon was able to show the way for two decades of Republican victories by giving the civil rights advocacy honors to the Democrats, while earning the support of social conservative activists.

Similarly, in the midnineties, Clinton was under a great deal of pressure from liberals in his party to restore the Democratic Party to the economic liberalism of the New Deal. However, in the midnineties, the Democratic Party had irrevocably lost a vital component of the successful New Deal coalitions: the South. As a result, Clinton had only one choice; he could keep the Democratic Party as a minority party, isolated in the upper left-hand quadrant in Figure 3, or he could reach out to the cosmopolitans in the upper right-hand corner. Appealing to Southern and other social conservatives was no longer credible for the party that had supported civil rights and affirmative action for 30 years—but reaching out to cosmopolitans via a more moderate economic policy was viable. Figure 3 suggests the location of the various presidential candidates over the last 25 years.