- Shopping Bag ( 0 items )
This book estimates the effects of economic conditions on the behavior of individual voters and on the outcomes of 42 elections in 15 countries. The conventional wisdom that poor economic conditions hurt governing parties is too simplistic. It does not hold for junior parties in coalition governments, who frequently gain at the expense of larger parties. It also does not hold in countries where responsibility for economic policy is unclear (more often than not). The extent to which the economy affects election outcomes depends strongly on the degree of electoral competition between parties.
Conventional wisdom asserts that economic conditions are closely linked to election outcomes. Bill Clinton turned this conventional wisdom into a cliché when, as a candidate in the 1992 American presidential election, he had his campaign staff put up a banner that hung across their campaign headquarters emblazoned with the words “It’s the economy, stupid!” This conventional wisdom is supported by much academic research. At least since the 1930s, voters in a variety of democratic countries have tended to hold governments accountable for bad economic times, reducing their support for parties holding government office in conditions of high unemployment or inflation or of low economic growth (Tufte 1978; Chrystal and Alt 1981; Hibbs 1977; Fair 1988; Lewis-Beck 1988; Markus 1988, 1992; Erikson 1989; Mackuen, Erikson, and Stimson 1992; Nadeau and Lewis-Beck 2001; Dorussen and Taylor 2002). These general findings hold whether the effects of economic conditions are modeled in terms of votes for government parties (generally referred to as “vote functions”) or in terms of government standing (generally referred to as “popularity functions”).
But it is clear that the economy does not always determine eithergovernment popularity or vote shares. In the American presidential election of 2000, Al Gore failed to win decisively as the incumbent party standard bearer despite a booming economy. In the Netherlands in 2002, all three members of the governing coalition lost votes in similarly excellent economic conditions. In Britain and Ireland in 1997, ruling parties also failed to win reelection despite booming economies. And in Britain in 1992, a ruling party succeeded in winning reelection despite an economy that was languishing. Turning to government popularity, it is clear that this often reflects voter concerns that go far beyond the economy. Support for George Herbert Walker Bush reached over 90 percent in the months immediately following the first Gulf War, despite a languishing economy.
These are exceptional cases, of course. The academic literature cited earlier has established that the conventional wisdom holds more often than not. However, even when governments as a whole gain or lose in accord with expectations, the individual political parties that are members of governing coalitions seldom find either their popularity or their vote shares moving in step. With coalition governments, it is often the case that some members of the coalition lose votes in an election, while others gain. The same goes for the major opposition parties in a multiparty system. Some of these may appear to benefit from a slow economy, but others do not.
Moreover, while government fortunes may appear more often than not to respond to economic conditions in general, there are numerous instances of countries with endemically high unemployment (for example, Spain, Greece, and Portugal) or persistently stagnant economies (for example, Japan) where these conditions did not lead citizens to vote ruling parties out of office. Indeed, from 1989 to 1999, in countries that are members of the European Union, the level of unemployment was repeatedly found in survey research to be viewed by voters as by far the most important problem facing their countries. Yet in all that time, few, if any, election outcomes were reported as having been determined by a government’s failure or success in tackling high levels of unemployment.
Why are some election outcomes apparently the result of economic conditions, while others are not? Why are some governing parties apparently hurt more than others by bad economic conditions, and why do some opposition parties appear to gain while others do not? Why do some sorts of economic conditions appear to determine election outcomes, while others (at least at certain times) do not? In this book we address these questions.
Parties compete for votes and in so doing provide the means for voters to hold governments accountable. Many voters see the management of the economy as one of the prime responsibilities of government. So an understanding of how voters react to parties in the light of economic conditions illuminates a central feature of democratic governance. Moreover, these questions have an importance to political science that goes beyond their obvious everyday relevance. Even though past research has established a general relationship between economic conditions and election outcomes, the evidence has been by no means conclusive and has given rise to a series of protracted debates in the literature, none of which shows much sign of convergence. This is the case for almost any topic that students of economic voting are concerned with. For example, we do not really know whether voters hold governments responsible for the general economic situation in their country or whether they are more immediately concerned about their own financial situation (whether economic voting is sociotropic or egocentric). Nor do we know with confidence whether voters respond retrospectively (holding governments responsible for past successes and failures) or prospectively (choosing parties for their economic expertise and policies) to economic conditions. Moreover, the modeled forms of vote and popularity functions differ considerably – within as well as between countries – as do the estimated effects of changes in these conditions.
The instabilities in the results of different studies made the editors of a recent symposium exclaim: “We all prefer to think that the instability is apparent only. That is, it is due to something we are missing or doing wrongly – if we could just find the ‘trick’, everything would be well” (Lewis-Beck and Paldam 2000: 114).
We do not presume to claim that we have found the “trick” that makes everything well, but we do believe that we have taken a major step in the right direction. In this book we argue that, indeed, past researchers have been doing something wrong. We assert that most existing studies in the field of economic voting mis-specify the dependent variable in their analyses and that many of the instabilities in the findings are a consequence of this mis-specification. Studies at both the aggregate and the individual levels have generally relied on a very crude distinction between the standing of (or votes for) governing parties and the standing of (or votes for) opposition parties. Obviously, this crude distinction does not adequately describe the choice process in multiparty systems. Anderson (1995) and Stevenson (2002) do distinguish between parties that are members of a coalition government, but neither of these authors distinguishes between opposition parties, thus still failing to fully specify the choice process. In Chapter 1, we will review the literature and explain how and why the customary specification of the dependent variable can be expected to yield invalid estimates of the effects of economic conditions.
Past studies of economic voting have looked in detail at variations across political systems and at variations across voters – see the outline of the edited volume by Dorussen and Taylor (2002), who organize their book around this distinction. Our study, by contrast, while conducting its analyses at the level of individual voters, explicitly focuses on competition between parties. We analyze the effects of economic conditions on electoral support for each of a country’s parties, treating government/opposition status as a variable that helps us to understand the different impact that the same economic conditions have on different parties. Government versus opposition status is, however, not the only characteristic of parties that we distinguish; we also look at their size, ideological complexion, and whether they control government ministries with responsibility for economic affairs. We will show – by estimating aggregate election outcomes on the basis of our individual level models – that particular opposition parties as well as particular government parties are affected very differently by improving or deteriorating economic conditions. We will even show instances of some governing parties gaining votes (at the expense of other governing parties) as a result of a worsening economy.
Because different countries have governments that are composed of very different types of parties (and parties that find themselves in very different competitive situations vis-à-vis each other), similar economic developments can have very different implications for individual voting decisions (and hence for election outcomes) in one country than in another. Such developments can even have different implications within one country at one election than at another. We will show that the character and competitive situation of individual parties make a big difference to the consequences of economic developments. It is no wonder, therefore, that highly unstable findings are reported in studies that take no account of distinctions between parties.
Another source of instability in the findings of past studies is that many of them were conducted in venues that contained a great deal of nonrandom “noise.” Studies of vote choice at the time of an election need to take account of all the factors that influence election outcomes, so that effects of the economy can be isolated after having controlled for everything else. The problem here is that we simply do not know what all the factors are that influence election outcomes. Although we have a fairly good understanding of what makes people vote the way they do, when estimates derived from this knowledge are aggregated to the level of the election outcome we do not, in general, reproduce the results, in terms of vote shares for different parties, that occurred empirically (Anker 1992; Dalton and Wattenberg 1993; Erikson 2002). We understand voting behavior in general, but the track record of political scientists in predicting the outcomes of specific elections is almost as bad as their track record in establishing consistent effects of economic conditions. Specific election outcomes appear to be determined not only by general forces but by all sorts of factors, such as campaign slogans, political scandals, and candidate traits that may be specific to particular elections. So, our knowledge is by no means sufficiently detailed to serve as a basis for controlling all the factors that could contaminate our findings – especially if the effects we are looking for are rather small.
We believe that we have found a venue for studying economic effects on election outcomes that is not subject to nearly so much nonrandom noise as the conventional venue of voting choice at national elections. Specifically, we study voter behavior not at the time of national elections, but rather at the time of elections to the European Parliament. For reasons to be discussed more fully in Chapter 2, our findings nevertheless allow us to draw conclusions about voters’ actual choices in national elections. It has been established in past research (van der Eijk and Franklin 1996) that the low saliency of European Parliament elections provides scant stimulus that would divert voters from their baseline national party preferences. Elections to the European Parliament are thus not genuinely Europe-wide elections but rather what have been termed “second order national elections” (Reif and Schmitt 1980; Reif 1984). For that reason, what we see at these elections are the same effects that are relevant in national elections but uncontaminated by the idiosyncrasies of national election campaigns. Because we study voters resident in fifteen different countries at three points in time – 33,000 of them in all – our study has the power to evaluate effects that are expected to be quite small. We will enlarge on our research design in later Chapters.
Our design also enables us to address another problem arising from past research, which is that the main finding from aggregate-level studies (that good economic conditions benefit incumbent government parties) has never been unambiguously replicated at the individual level. Individual-level studies (Fiorina 1978; Kinder and Kiewiet 1979, 1981; Lewis-Beck 1988; Nadeau and Lewis-Beck 2001) generally assess the effect of respondents’ assessments of economic conditions on their support for a government or an opposition party. But these assessments of economic conditions themselves turn out to be strongly affected by preferences for the governing party or parties (Wlezien, Franklin, and Twiggs 1997; Bartels 2002; Duch and Palmer 2002). Only for the United States, looking at a sequence of up to ten presidential elections, has evidence been found for an effect of real economic conditions on individual voting decisions (Markus 1988, 1992; Nadau and Lewis-Beck 2001).1 A major objective of this book is to determine the effect of objective economic conditions on individual-level vote choice across a much larger number of electoral contexts (we study forty-two contexts in all).
Plan of the Book
In Chapter 1, we will review the literature on economic voting and develop our argument regarding why and how past studies mis-specify the choice process, pointing out how this mis-specification will in many cases have biased the results of such studies. We show how a focus on political parties as the primary objects of electoral contestation enables us to examine the process by which individuals make their electoral choices – a process that must be properly specified if the role of economic conditions is to be discovered – and how a failure to do so inevitably results in unstable and sometimes incoherent findings. In order to be able to focus on political parties, in this book we adopt a two-stage model of vote choice based on the work of Anthony Downs (1957) that will be introduced in detail in Chapter 2. In that chapter, we also explain how this model meets the conditions set out earlier so as to remove or at least ameliorate the various problems enumerated in Chapter 1. We believe that this model has a good chance of arriving at stable and consistent estimates for the effects of economic conditions. In Chapter 3, we will describe in greater detail the theoretical expectations that govern our choice of variables for the model, the hypotheses that we derive from this theoretical reasoning, and the data we employ to test those hypotheses. Most of the hypotheses pertain to individual-level behavior and relate either to expected effects of economic conditions (and alternative ways of specifying those effects) or to the influence of control variables that our model needs to take into account. These hypotheses are tested in Chapters 4 and 5, which deal with individual-level behavior. There is one hypothesis, however, that concerns the outcomes of elections and the way in which those outcomes are affected by economic conditions. That hypothesis is tested in Chapter 6, while Chapter 7 concludes. In an Epilogue, finally, we sketch out how further research into economic voting might be designed so as to avoid the endemic errors that have characterized the field up to now.
Studying Economic Voting
In this chapter, we develop our argument that previous investigations into economic voting have been hobbled by model specification problems. We believe that the main problem in previous work has been a failure to focus on political parties as actors that compete for votes. This failure means that previous research has been unable to take account of the extent of party competition (which can vary from election to election). We argue that this failure to take account of party competition leads to models that are intrinsically mis-specified and that this is one of the root causes for the unstable and sometimes incoherent findings in the literature.
In what follows, we diagnose the deficiencies we see as endemic in existing studies of economic voting (many of these deficiencies apply to other studies of electoral behavior as well) and set out our ideas about what a viable approach should look like in general terms. This will set the scene for Chapter 2, in which we present our own approach to the study of economic voting and describe the ways in which it mitigates or eliminates the deficiencies of past studies.
Conceiving the Dependent Variable
Studies of economic voting have been conducted at the aggregate level, treating the country or election as the unit of analysis, and at the individual level, treating the survey respondent as the unit of analysis. In aggregate-level studies, the dependent variable has been the proportion or percentage voting for the government party (or parties in the case of coalition governments), while in individual-level studies the dependent variable has been a dummy variable indicating whether the respondent voted for a government party or not. Aggregate-level studies try to account for variations in government support on the basis of economic conditions (generally inflation, unemployment, and economic growth), sometimes with controls for systemic characteristics. Individual-level studies try to account for variations in government support by means of a “standard model” consisting of a variety of independent variables found empirically to be important in explaining support for particular parties, together with measures of economic perceptions – the assessments given by individuals regarding the state of the economy at the time of the survey (and perhaps over the recent past and/or near future).
The most important problem that is common to both approaches is that they fail to take account of competition between parties, as already mentioned. This root problem manifests itself in two different ways: a focus on support for the government and a focus on electoral choice.
In existing approaches to economic voting, the dependent variable is defined as a choice between government and opposition parties. The focus on this dichotomy seems to make sense because researchers assume that, if economic conditions have any influence on the voting act, this influence must involve an assessment of who is to be credited or blamed for the state of the economy (Anderson 1995). Here the supposed role of governments in economic management is generally taken for granted and, because of this, it seems natural to construe the dependent variable in terms of the distinction between government and opposition. But this supposes that all members of a governing coalition will benefit equally from good economic times and will suffer equally if economic times are bad. Empirically, however, it turns out that parties sharing government power often fare very differently at the same election. Some government parties lose votes while others gain, and similarly with opposition parties. Moreover, at the time of an election voters are not presented with the opportunity to cast their ballot for or against the government but rather to vote for a party or candidate. In a two-party system, such as exists in the United States, elections may be conceived as providing voters with the opportunity to support or reject the government by casting their ballot either for the party of the president or for the other major party, but even in that system the choice that voters face is rarely presented in such terms. In other countries, where multiple parties vie for voters’ favor, the government – opposition dichotomy does not adequately represent the choices on offer.1 Where several parties share power in a coalition government, it is abundantly clear that casting a ballot involves more than choosing between government and opposition. In such cases, each voter can choose to support only one of several government parties. Even when the government consists of a single party, there will generally be more than one opposition party; yet, each voter can choose to support only one of these parties.
If we only focus on whether the government gains or loses support, as most previous researchers have done, then we fail to examine the actual decision-making process that voters experience. Any model that focuses on the choice between government and opposition is bound to mis-specify the choice process. Yet that is what most existing approaches to the study of economic voting do. Exceptions (for example, Duch and Stevenson 2003) that attempt to focus on party choice by employing multinomial or conditional logit methods suffer from other problems that we will detail in Chapter 2.
The second way in which existing approaches fail to adequately represent political parties is caused by their exclusive focus on the actual choices that voters make (or the aggregated counterparts of these choices in terms of the share of the vote given to each party). This focus also seems natural, since voters generally have to choose a single recipient of their vote. But choice informs us only (and in a very crude way) about a voter’s orientation toward a single political party. It does not reveal anything about that voter’s orientation toward other parties, except for the fact that they were not chosen. Bingham Powell (2000) has eloquently argued that, by focusing only on the choices made by voters, analysts have assumed that these reveal all we need to know about voter preferences; but this is not so (2000: 160).
1. Studying economic voting; 2. Party choice as a two-stage process; 3. Hypotheses and data: the theoretical and empirical setting; 4. Effects of the economy on party support; 5. The economic voter; 6. From individual preferences to election outcomes; 7. The economy, party competition, and the vote.