Our starting point in this analysis is found in two studies published in Political Studies in the early 1980s. First, using data from the city of Bath, Lewis attempted to clarify “whether attitudes to a list of public expenditures could be predicted from the political party preferences of the respondents [1]. Using a survey of two hundred respondents from the Bath register, he found significant differences between levels of support for a variety of public spending objectives of the supporters of the Labour and Conservative parties. While such an observation at the current time is not surprising given the evolution of British politics, until Lewis’ research, it was “not clear whether expenditures and taxations … [could] be used to maximize political support and whether they have a major impact on voting behaviour” [1]. More surprisingly perhaps, when Lewis’ analysis was replicated by Lowery and Sigelman [2]for the US case using General Social Survey data from 1973 to 1980, few such differences were evident. Even on the eve of Ronald Reagan’s election, few differences in spending preferences could be found between supporters of the Democrat and Republican parties. The contrasting UK and US patterns were not unexpected. Indeed, Lowery and Sigelman [2] noted the long-standing characterization of British parties as “more ideological and socially cohesive than their American counterparts.” Further, they suggested that, “party identification appears to play a rather minor role in the formation of policy opinion in the United States” (ibid.). Given the static snapshots of two electorates reported in these two studies, then, the differences between the two cases were essentially accounted for as a function of persistent differences between the natures of the two political party systems.

However, a lot of water has flown under the bridges of party politics in both countries since 1980. Indeed, they have followed a broadly similar pattern with the Thatcher and Reagan revolutions taking their respective parties sharply to the right in terms of negative assessments of the role of the state more generally and public spending more specifically. Then, in response to losing a series of elections, the New Democrats of Bill Clinton and the New Labour under Tony Blair adopted at least some of the anti-government rhetoric of Thatcher and Reagan in an attempt to modify their parties’ images as big spenders. But did these elite driven efforts to alter their parties’ identifies with respect to public spending muted the differences (or lack thereof) in preferences for public spending between their respective major parties as observed by Lewis [1] in the UK case and Lowery and Sigelman [2] in the US case? Answering this question may allow us to go beyond the conclusion drawn from earlier snapshots of the two party systems, where the differences or the lack thereof in these systems are persistent traits to consider when mass attitudes about the role of government are far more dynamic and, perhaps, elite-driven.

We address this question by constructing a Leader-Party Image model of the relationship between party images and preferences for spending. We test the model by matching UK and US data on aggregate public opinion on public spending at four points since 1980, treating the Clinton New Democrat and Blair New Labour transformations as interventions that might be expected to alter the preferences of partisans for public spending. Along the way, we control for both a Downs [5] view of the relationship between party positions and the distribution of public preferences for public spending, and the now dominant rival account of the relationship between partisan control and the preferences of the electorate for public spending found in the thermostatic model developed in the work of Wlezien [3] and Erikson, MacKuen, and Stimson [4]. We start by presenting a short overview of the electoral composition and political ideological differences between the respective major parties of both countries, developing theoretical expectations about the nature of changes we might expect to observe in preferences for public spending associated with the advent of New Democrats and New Labour and changes in partisan control. We then analyze aggregate change in partisan preferences at four points in time to test these hypotheses. We conclude by discussing the implications of these results for understanding the interplay of elite electoral strategies and mass attitudes about the role of government.

The histories of the two party systems with respect to rhetoric on public spending are readily summarized. In this regard, perhaps the best source is the parties themselves since their own words best summarize the image they wish to present to voters. To start with the UK, since becoming Parliamentary leader of the Conservative party in 1975 and then as Prime Minister from 1979 to 1990, Margaret Thatcher made opposition to public spending a centerpiece of her party’s image [6,7]. And indeed, public spending was reduced in a number of areas, most notably in education and housing. This emphasis on a more limited role for the state was continued under John Major’s governments (1990-1997) and became the party orthodoxy, as well as under the electorally less successful leadership of William Hague, Duncan Smith, and Michael Howard. It was not until the accession of David Cameron as Conservative leader that the party sought to significantly modify its rhetoric of full-throated opposition to public spending and excessive government. On the other hand, through the 1980s, Michael Foot, Neal Kinnock, and John Smith kept the ideology of the Labour Party rooted in its traditional support for greater public ownership, government intervention, wealth redistribution, and a strong welfare state [8]. The New Labour governments of Tony Blair (1997-2007) and Gordon Brown (2007-2010) rejected much of this traditional rhetoric and offered spending policies in numerous policy areas that reflected the Thatcherite emphasis on alternatives other than direct public spending.

In the case of the US, the differences in public spending rhetoric and policy proposals of the parties led by Gerald Ford and Jimmy Carter in the 1976 presidential election were, in retrospect, considerably muted. This changed remarkably with the successful campaign of the Republican candidate Ronald Reagan in 1980. [9] His administration from 1981 through 1988 reveled in identifying government as the problem, not the solution; a policy tone rigorously maintained by George H. W. Bush from 1989 through 1992, despite his occasionally reference to a “kinder, gentler nation,” and, with less electoral success, by Bob Dole in 1996. Still, the Republican Congress led by Newt Gingrich from 1995 through 1999 doubled down on opposition to public spending, even to the point of shutting down the government over a spending dispute with President Bill Clinton. And while the administration of George W. Bush (2001-2008) notably increased public spending in a number of areas, it maintained the Reagan rhetorical tradition of deep opposition to greater domestic public spending [10]. In contrast, and like Tony Blair, the New Democrats of Bill Clinton, abandoned what some perceived as the reflexively pro-spending positions of the unsuccessful campaigns of Walter Mondale and Michael Dukakis in favor of a more selective and cautious support for spending coupled with some significant reductions in public support, most notably for welfare spending. In sum, and in very broad strokes, the conservative parties of both countries undertook a strong march to the right with respect to public spending rhetoric in the late 1970s and early 1980s, with matching accommodations in their approaches to public spending on the part of Labour and the Democrats in the early 1990s.

How might these changes in the images/policies of the parties with respect to the role of the state have influenced aggregate preferences for public spending? The perspective that guides this research as well as the typical commentary on how Tony Blair and Bill Clinton led their respective parties out of electoral wilderness might be labeled the Leader-Party Image model. In this model, leadership changes first with new candidates at the top of the ticket signaling to voters a sharp change in what the party is and what it stands for. In response to this announcement, Labour and Democratic voters who already identify with their respective parties would be then expected to restructure their preferences to be in accord with the new image of the party. Thus, Bill Clinton and Tony Blair’s successful “New Labour” and “New Democrat” campaigns would be expected to lead to more moderation in the spending preferences of their party members. This model, then, assigns a critical role to the positions of leaders, especially so when leaders mark a sharp turn in party policy preferences. Voter partisan identities are then expected to lead to renewed conformity to these new policies so that voters’ preferences remain congruent with the new policy images of the parties. With respect to the changes announced by Blair and Clinton, this would mean policy images in regard to public spending.

We must note that our model makes the critical assumption that the images and policies of parties lead or influence voters’ preferences for public spending. It might alternatively be argued that public preferences, in fact, lead or determine party positions. Once we move beyond looking at the self-constructed images of parties and citizens’ preferences for public spending at a symbolic level, there is considerable room for thinking that public opinion and public policy are closely related at a substantive level, and that the former leads the latter. Indeed, Page and Shapiro [11] and Shapiro and Young [12] have reported considerable evidence in support of this traditional assumption of democratic theory. Even more generally, this view of the linkage between preferences and parties dominates the literature, whether that literature is founded, as we will see below, on a Downsian perspective (1957) or one informed by a macro politics perspective [3,4]. In such models, parties chase voters, but can only marginally influence them.

Yet, there is little evidence that spending preferences lead party images on public spending in the manner suggested by such analyses of specific policy changes across a range of governments. This is especially evident in Lowery and Sigelman’s [2] finding, using 1980 data, that Republican and Democrat identifiers did not differ much in terms of spending preferences. Indeed, and with the important exception of welfare spending, preferences for public spending remained largely unstructured in the United States in 1984 following two years of the Reagan administration [13]. Further, our short history of party changes glosses over a great deal of detail in which conservative governments have greatly expanded spending and nominally leftish governments have cut spending. Rather, as Jacoby [13] has suggested, “the very issue of government spending seems to have a great deal of symbolic, rather than substantive, content.” This emphasis, of course, is not so very different from the macro politics view that specific policy preferences are less important than a general policy mood [3,4] for greater attention on specific policy preferences). Still, the emphasis or lack thereof by parties on spending as a symbol plausibly should be expected to overtime influence the distribution of preferences for spending at the substantive level. That is, at the level of symbol, we expect the images parties adopt to lead and then slowly shape the preferences of their supporters by redefining what it means to be a Labour supporter or a Democrat. Such changes over time may then influence substantive policy, if perhaps in a more delayed and punctuated manner [14] than traditional democratic theory would expect.

Perhaps even more importantly, the expectation that parties chase voters’ preferences but have no influence on the content of those preferences seems inconsistent with both the perspective of much of the larger literature on political parties. From that larger perspective, it is not uncommon to expect that parties shape or at least try to shape party images by framing and massaging issues so as to change the content of salience of preferences so as to win elections [15]. It also seems inconsistent with the actions of politicians such as Bill Clinton or Margaret Thatcher whose rhetoric often sought to educate citizens about the appropriate expectations about the role of government they should hold. Thus, it does not seem far fetched to expect that the arrow running from public preferences to party positions is complemented by another that runs from party positions to public preferences.

And there is some support for at least a conditional view of this causal direction. In studies of the evolution of public opinion on the culture war issues of abortion, gun control, and gay rights in the United States, research has demonstrated that the sharp cleavage between the opinions of party elites tended to lead the growing cleavage among party identifiers in the mass public [ 16,17]. Similarly, there is a long line of research–and contrary evidence as well–indicating that elite opinion on the European integration and the European Union leads mass opinion [18-21], for example, argued that greater attention should be given to “the role political elites play in teaching the public what to think about (European integration).” In short, there is evidence that party elites can at least some time and under some conditions move public opinion.

However, such research is typically conducted on new and emerging issues, not the persistent and established issue of public spending. Still, from this perspective on the causal relation between party elites and opinion, we would expect that the Thatcher/Reagan governments reshaped the responses of Conservative/Republican supporters over time so that they become more reflexively opposed to public spending as part of their party images. Conversely, and what will be the object of our empirical analysis, we would expect that the accession of New Labour under Tony Blair and the New Democrats under Bill Clinton led to reduced levels of support on the part of Labour and Democrat supporters, respectively, for more public spending. If this expectation and the theoretical assumptions underlying it are valid, two test implications might be expected. First, we would expect the differences in the preferences of left and right party supporters to be sharpened or muted in response to these changes in the symbolic images of parties with respect to public spending. In regard to the images of New Labour and the New Democrats, we would expect the differences between Democratic and Labour preferences on the one hand and Republic and Conservative preferences on the other should have declined following, respectively, the electoral successes of Bill Clinton and Tony Blair. Second, however, we would expect most of this muting to occur through a decline of preferences for public spending by Labour and Democratic supporters rather then stronger preferences of Conservative and Republican voters for greater spending. That is, it was the Democrats and Labour who attempted to change their images at the symbolic level and, perhaps to a lesser extent, at the substantive level. The rise of Tony Blair had little in the way of a modifying impact on the rhetorical opposition to spending of the party of Hague, and President Clinton did not give rise to a kinder, gentler Newt Gingrich. Indeed, the immediate responses of both conservative parties were to heighten their rhetoric against a perceived excessive role for the state and a deviation from the heritage of Thatcher/Reagan.

While perhaps plausible, there are a number of other theories about the role of mass and elite opinion on public spending preferences that we need to account for, all of which more or less focus on actual electoral fortunes and patterns of party control. The first of these is a naive view that left parties should produce both greater preferences for and substantively more spending and right parties the opposite. But despite some evidence from municipal settings in support of this expectation [22], other studies have found nearly the opposite at the national level [23] or only mixed and weak results at the local level [24].

A second rival model meriting more serious attention is the traditional Downsian notion that parties seek to capture the centre of the political distribution where the bulk of votes lie. Leftist parties that move too far to the left and subsequently lose office need to move back to the centre if they are to succeed electorally. Conversely, rightist parties that move too far to the right and lose an election need to move back to the centre if they are to win again. There seems little doubt that Down’s idea lie behind Tony Blair and Bill Clinton’s efforts to reshape their parties in response to the Thatcher and Reagan revolutions. Yet, with respect to expectations about how patterns of party support are related to public preferences for spending over time, this model offers expectations that are quite different from those noted here. This is because the distribution of aggregate public opinion is generally assumed to remain stable in the Downsian model. It is party positions with respect to that distribution that change. As a result it is not that voters change their preferences, but that they move from one party to another that is now closer to their preferred spending preferences.

In such a model, the aggregate preferences within a party may change, but only as a result, for example, of a right-wing party losing centrist voters because of its perceived extremism to a left-wing party viewed as closer to the median voter. Conservative voters would, as a result, be more highly concentrated within the right-wing party. The average level of support for spending would therefore move rightward. The now more successful left-wing party would now attract support from voters representing a broader distribution of opinion about public spending. As a result, the average support for public spending of its supporters (new and old) would also become more conservative on average. This would lead to sharply different test implications from those noted earlier for the Leader-Party Image model. First, there should be no net difference between the average opinions found in the two parties even after a major shift in party fortunes because the resulting shift in ideological direction by one party should be mirrored by a symmetric shift in the same direction by supporters of the other party. Second, and as a result, the change in the average opinions found in the two parties should be positively related. When supporters of a rightist party become more conservative in terms of public spending, the supporters of leftist party should, on average, also become less supportive of public spending. Both test implications result from the fact that individual preferences remain stable. What changes is the party affiliation of centrist voters.

A third and very serious rival explanation can be found in the thermostatic model of Wlezien and Erikson [3,4] , a model that has now largely superseded the naïve model of party control earlier. In the thermostatic model, voters are assumed to be only weakly attentive to politics and public policy. Still, they are expected to assume that, as suggested by democratic theory, right-wing governments provide right-wing policies, including less public spending, and that leftwing governments provide the opposite. If true, then periods of party control should operate as a thermostat with sustained periods of left party control leading to increased preferences for lower levels of public spending and sustained periods of rightist governments leading to greater support for public spending. The longer governments remain in office, the more strongly this countervailing impulse should influence public preferences, leading in turn to an enhanced likelihood that the out-party will replace the governing party in a coming election.

If true, this model offers a set of test implications that differ from both the Leader-Party Image model developed here and the Downsian model just discussed. That is, as was the case with the Downsian model, there should be no net change in the difference in the distribution of opinions between the parties since it is expected that both left and right voters are attentive to the policy temperature and move their spending preference thermostats in the same direction. After a period of Labour or Democratic control, both Labour and Conservative voters or Democratic and Republic voters should favour less government spending. Accordingly, the change in a party’s image prior to a successful election will, in contrast to the party image model developed here, influence the distribution of opinion in both parties. However, that influence will not be a function of changes in party image per se, but will be a result of winning and providing (or rather, assumed to be providing) actual policy from the left or right of the political spectrum. Accordingly, and in contrast to the Leader- Party Image model, the movements of average opinion within the two parties should be related to each other. As with the Downsian model, that relationship should be positive as voters in both parties move right/left in response to patterns of party control.

In testing these competing expectations, we follow the general strategy suggested by Lewis and Lowery and Sigelman [1,2], but focus on change within and between parties in aggregate support for public spending in the UK and the US over the last three decades rather than sources of individual support for spending within a single cross-section of time. To do so, we rely on comparable measures for public spending found in the British Social Attitude Survey and, in the American case, the General Social Survey, the survey used by Lowery and Sigelman [2], in combination with a modest restructuring to simplify their party identification measures to provide a sharper direct contrast between the spending preferences of supporters of the two major political parties of each country. The two nationwide surveys provide us with the best opportunity to make a generalized comparison between the two countries. Both are respected, continuous surveys based on large random samples that are often used by scholars to track changes in public preferences over a number of policy areas. Indeed, they are one of the few survey instruments that allow us to track such changes over time, and far superior to the limited survey of Bath respondents used by Lewis [1]. Unfortunately, however, neither data set is without problems in terms of consistency of questions over time. The British data set was especially problematic in asking spending preference questions only episodically. In order to compile comparable data for both settings, we employ data from the 1985, 1990, 1994 and 2006 British Social Attitude Survey and 1985, 1990, 1996 and 2006 General Social Survey. Thus, we use a total of eight surveys, four for the US case and four for the British case. As we will see, even with the real limits of the consistency of questions across time, they still afford us sufficient empirical leverage to extract discernible results with respect to most of our expectations.

Starting with the dependent variables, in all eight years/surveys, respondents were asked their opinions about spending in eight specific policy areas. The following areas were listed in the British Social Attitude Survey: environment, health, police and law enforcement, education, military and defense, old age pensions, unemployment benefits, and culture and arts. And the following were listed in the General Social Survey: education, health, law enforcement, defense, environment, retirement benefits, unemployment benefits, and culture and arts. More specifically, respondents were asked if they supported much more, more, less, much less or about the same level of spending in each policy area. In initially analyzing these observations on spending preferences, we merged the intermediate categories to develop a three-point index of more, less, or about the same spending. We examine these spending preference data in aggregate, focusing on the proportion of the total within each party favoring more spending, calculating these proportions within each policy area and within and between the two parties at each of the four periods of observation. This produces in both national cases 32 observations of aggregate spending preferences.

The US results, in sharp contrast to Lowery and Sigelman’s findings for the 1980s, highlight strong differences between supporters of the two American parties, with a near majority or more supporting greater spending within both parties across most spending categories. More broadly, the mean support for increased spending across the year/policies was 40.12 percent for self-identified Republicans and 51.34 percent for self-identified Democrats. This is now quite similar to and even slightly greater than the results for the UK where, across the year/policies, 43.09 percent of Tories favored greater spending compared to 52.14 percent among Labour supporters. Indeed, the latter American difference (9.04 percent) is now slightly smaller than the former British difference (11.22 percent). [25-28]

Turning to the measure of party identification, the British survey identified respondents as supporters of Labour, the Conservative Party, and smaller parties. To better focus on the spending preferences of the two major parties, we examined only Labour and Conservative supporters. To make the American data comparable, we merged strong and weak Democrat/Republican respondents and removed the Independents from the analysis. Thus, we have 32 year/policy observations in both cases on spending preferences for the supporters of both major parties in each country. Given our test implications, we analyze these aggregate results both within party in each case as well as the average preference difference between the respective parties.

The key independent variables concern changes in party images and patterns of sustained party control. In the party image model discussed here, the Clinton New Democrats and Blair’s New Labour represent distinctive changes in elite projections of party images with respect to public spending falling within our four periods of observation. Thus, for the US, the 1985 and 1990 observations were coded zero on Clinton and the 1994 and 2006 observations were coded one. For the UK case, the 1985, 1990, and 1994 observations were coded zero on Blair and the 2006 observations were coded one.

To tap the expectations of the thermostatic model, we created simple counter variables indicating for each of the four observation periods the number of elections the sitting party government had won. The counter, given the countervailing expectations of the thermostatic model, is scored positively in the Republican or Conservative cases and negatively in the Democrat/Labour cases. Thus, in the US case, the 1985 observations are coded 2 for the second Reagan administration, while the 1990 observations are coded three for the third term of consecutive GOP control. The 1996 observations are coded negative one for the first Clinton administration, and the 2006 observations are coded two for the second administration of George W. Bush. In the UK case, the 1985 observations are coded two for the second Thatcher government, while the 1990 observations are coded three for the third term of consecutive Conservative control. The 1994 observations are coded five for the fifth consecutive Tory government–in this case, John Major’s second. And the 2006 observations are coded negative three for third government headed by Tony Blair and the Labour party.

Given the data, we employ simple figures and then pooled regression analysis with dummies included for all of the policy areas but one–defense–as our reference category. Inclusion of time period dummies to create a saturated least squares dummy variable model [29] generated such extraordinarily high levels of collinearity that many estimates could not be generated for the more substantively interesting thermostat variables associated with Clinton/Blair. This led us to exclude time dummy variables from the analysis. Other analyses not reported here indicate that the use of robust standard errors clustered on the time period had little impact on the substantive results. Two tailed tests are used given our mix of competing expectations. Given the small n and the remaining presence of high levels of collinearity, statistical significance is reported at the 0.10, 0.05, and 0.01 levels.

We first examine the trends in the data on preferences graphically. As seen in Figures 1a and 1b, the trends in support for spending are generally quite stable over time for supporters of the Labour and Conservative parties. There is increased support for police and law enforcement over the first three periods and something of a decline in support for environmental spending in the last. But the most notable characteristic evident in these two figures is how consistent they are with each other. The preferences for more spending among Labour supporters move fairly consistently over time with those of supporters of the Conservative party. This does not bode well for our party image model, which would lead us to expect that the changes would be largely felt by the party–in this case the Labour party, following its reformulation as New Labour under Tony Blair–undergoing some effort by party elites to redefine the party’s stance vis-à-vis an expansive state. Nor, however, does it bode well for the Downsian alternative we have outlined.

Figure 1a: Proportion of Labour Supporting More Spending By Policy Area By Year.

Figure 1b: Proportion of Conservatives Supporting More Spending By Policy Area By Year.

The same observation applies to the American case, as seen in Figures 2a and 2b. There are some differences between the parties evident in the two figures. For example, support for retirement benefits among Democrats rises constantly over the period, while it falls sharply among Republicans between 1990 and 1996. And this same pattern, if to a lesser degree, applies equally to health policy. These reflect, no doubt, the battles between the two American parties together for supporters of both parties. Again, this seems to support neither the Downsian nor the party image model.

Figure 2a: Proportion of Democrats Supporting More Spending by Policy Area by Year.

Figure 2b: Proportion of Republicans Supporting More Spending by Policy Area by Year.

As a result of these party trends, the differences between supporters of the two parties in terms of their support for more spending, as seen in Figures 3a and 3b, are not especially sharp. Still, the British case, as seen in Figure 3a, offers perhaps some initial support for the party image model not evident from a simple comparison of Figures 1a and 1b. That is, the difference between the levels of support among supporters of the two parties is narrower across the range of policy areas in the last observation from 2006 than in the three prior periods. Still, this pattern is not at all evident in the American case, where the narrowest band of differences occurred in the second observation from 1990, fully two years before the Clinton administration, and increases considerably in the following two observations, 1996 and 2006. Still, to better sort out these impressions, and to control for our rival explanations, we must turn to our more complete tests using the pooled model of preferences for spending.

Figure 3a: Difference in Proportion of Conservatives and Labour Supporting More Spending by Policy Area by Year.

Figure 3b: Difference in Proportion of Democrats and Republicans Supporting More Spending by Policy Area by Year.

These more informal observations drawn from Figures 1 through 6 are confirmed by the pooled regression results reported in Table 1. The two substantively important sets of estimates are reported in the first two rows. Contrary to expectations by the model on party images developed here, the Clinton and Blair dummies generated positive and significant estimates for their parties’ respective supporters (13.96 and 24.24) despite some collinearity and relatively few observations. We had expected that the moderating impact of New Labour and New Democratic policies would weaken support for expansive spending among their parties’ supporters. But instead, the opposite seems to have occurred. Even more contrary to that model, positive and significant estimates of nearly equal magnitude (12.86 and 22.21) were similarly generated for the supporters of the Republican and Conservative parties. We had expected that the impacts of Clinton and Gore would be primarily felt by their own supporters and not the supporters of their opponents. Thus, public support for spending increased across the board following the rise of Prime Minister Blair and President Clinton, a result that does not support the first test implication of our party image model. As a result, and contrary to our second test implication, as seen in the third and sixth columns of the table, the Clinton and Blair variables had no great effect on the difference in the aggregate level of support found between the parties of the left and the right. The estimates of the two dummies (1.10 and -2.02) were smaller than their standard errors.

Table 1: Values under coefficients are standard errors.*=p<0.10; **=p<0.01, two tailed test.

At the same time, these results provide little support for the alternative Downsian model. We interpreted that model as implying that ideologically attracting centrist voters from the right to the left by moderating the spending rhetoric and policies of the left-wing party, would leave the difference of the mean positions of the two parties the same, and this proved the case as indicated by the nonsignificant Blair and Clinton estimates in models 3 and 6 of Table 1. But this lack of net difference should have resulted from the supporters of the right leaning parties becoming even more opposed to spending as more moderate supporters left for the more attractive alternative closer to the center of the preference spectrum, while that same movement would leave the left-wing party with a more diverse set of supporters and, thus, on average less supportive on average than they had in the past. Again, supporters of both parties became more supportive of spending following the Clinton and Blair transformations of their respective parties. This should not be interpreted as suggesting that the Downsian model does not matter for electoral outcomes. It almost certainly does as is fully illustrated by the election of Clinton and Gore. But this does not mean that such elections lead to simple or obvious impacts on public attitudes about public spending. Thus, neither the party image model nor the Downsian model was fully confirmed. But before addressing a more satisfactory explanation of the Clinton and Blair dummies, we need to consider the third account–the thermostat model.

We must first note that the positive Clinton and Blair dummies just discussed also pose problems for the thermostatic model by suggesting that far from trying to turn down the spending thermostat during the Clinton/Blair governments, supporters of both parties became even more supportive of spending. Thus, this first result is anomalous for all three sets of theoretical expectations we have generated, an issue we return to below.

Still, the most direct test of the thermostat model is provided by the estimates reported in the third row of Table 1. We expected, given the inverse coding of the thermostat measure, that positive estimates would be generated for supporters of both the Democratic and Labour parties on the left and those of the Republican and Conservative parties on the right. And this certainly seems to be the case. The positive and significant coefficients of models 1, 2, 4, and 5 in the third row of Table 1 provide confirmation of this explanation. Indeed, the magnitudes of values of the coefficients for the parties of the left and right are quite similar. As a result, and also as expected by the thermostat model, the thermostat coefficients (-0.90 and -1.22, respectively) reported in models 3 and 6 for the difference between the average levels of support for more spending between the two parties are comparatively tiny and not statistically discernible. In short, the estimates reported in the third row of the table provide strong support for the thermostat model.

And in the end, we would also interpret the pattern of significant and nonsignificant estimates for the Blair and Clinton dummy variables as supportive of the thermostat model. Indeed, it is most likely that the significant and positive estimates for Blair and Clinton in the within-party models reported in columns 1, 2, 4, and 5 are an artifact of our perhaps overly simple coding of the thermostat variable in which we simply counted the number of times a party in control had won consecutive elections. If, contrary to our coding scheme based on a simple count, support for public spending across the public as a whole rises at a nonlinear rate over the course of conservative governments, then we would expect the Clinton and Blair coefficients to have pickedup and represent some of this additional nonlinear and countervailing effect of the consequences for very long periods of party control on the distributions of preferences about public spending across both parties. And as would be expected by the thermostat model, the net difference of the Blair and Clinton dummies as accidental thermostat variables is small. The Clinton and Blair estimates in models 3 and 6 are, as noted earlier, not statistically discernible. Unfortunately, given the limited number of annual observations and the relatively few changes in party control over this period, collinearity precludes full testing this plausible, although ad hoc, interpretation of the results.

We started this analysis by focusing on a Leader-Party Image model of the relationship between spending preferences and the changing positions of parties that emphasized the ability of political parties’ leaders to, via their manipulation of party images, dynamically shape the spending preferences of their followers as they seek to win elections. That they shape or at least try to shape party images so as to win elections seems clear [30], leading to a plausible expectation that the causal arrow typically running in the relevant theoretical literature from public preferences to party positions might be complemented by another running in the opposite direction, from party positions to public preferences. As we noted earlier, there is some evidence that the second causal arrow has been important for new and emerging issues, such as culture war issues in the United States [16,17] and European integration within the European Union [18-21]. But contrary to our expectations, we have not found any evidence that even significant changes in party images with respect to public spending in turn shapes the preferences of their supporters for more or less public spending in a manner that would serve to either enhance or dampen the differences in levels of support for spending between their own followers in comparison to those of parties on the other end of the ideological spectrum. Rather, it seems that patterns of party control influence in a similar manner the level of support for spending found among supporters of both parties. Thus, the net effect of these changes on aggregate differences between parties in their level of support for public spending is nil. As a result, the thermostatic model still provides the best account of how parties, especially their patterns of party control, influence distributions of policy preferences for more or less government.

Still, this null finding leaves us with something important. That is, these results suggest that far from being the rescuers of their parties via their efforts to move their parties more toward the Downsian center of the political distribution of opinion, President Clinton and Prime Minister Blair were fortunate beneficiaries of the normal rhythms of party fortunes where long periods of preferences for conservative (leftist) government are eventually exhausted. It may have helped both men and their parties to move to the center. But doing so did not reshape the differences between the parties’ respective electorates in terms of their preferences for public spending. Still, it is sometimes nice to be in the right spot at the right time if you wish to win an election. Less pleasant, of course, is the position of Gordon Brown or Al Gore, candidates who had to battle uphill against an accumulating tendency for everyone to turn the thermostat toward less spending.

Finally, we are not yet fully convinced that efforts such as those of Tony Blair and Bill Clinton to reshape the perception of their parties on the part of the public are wholly wasted. After all, such moves toward the center are well founded in Downsian theory and common among parties seeking to regain control of government. But the effect of such image changes may be other than the direct effect we have outlined here. In light of these results, perhaps a more likely effect is that such changes modestly hasten the time before a majority of the public reaches for the thermostat and/or delays the time when the public begins to assume that the new party must be implementing policies of the left or right and, therefore, slows the countdown to the next change in party electoral fortunes. In other words, the party image model might interact with the thermostatic model to set the pace of thermostatic change. Testing such an interaction, however, will require more data covering more elections and a broader range of party/ideological configurations. Further, it may be the case that the ability of leaders to shape the policy preferences of party voters may be more important in some political systems than others, with the capacities of leaders enhanced in political systems with more authoritarian and semi-democratic structures. Thus, it might be useful to further analyze the Leader-Party Image model in other cases and at other times. Finally, it is also true that the relationship between voter preferences and party positions may change over longer periods of time. We have noted earlier, for example, that our results in terms of the mean preferences of Democratic and Republican voters differed sharply from those reported by Lowery and Sigelman [2] for an earlier period. Thus, while the Blair and Clinton cases are especially interesting in terms of leaders trying to change the images of their parties, further analysis of other periods may prove useful.