- Email: [email protected]
The microeconomic voter
Electoral Studies 18 (1999) 505–517 www.elsevier.com/locate/electstud
The microeconomic voter Simon Blount* University of New South Wales, Sydney, N.S.W. 2052, Australia
Abstract In the United States, aggregate and individual level studies of economic voting for the Congress have produced contradictory findings. The same is true for models of economic voting for the Australian Parliament. This paper presents data taken from a series of individual level studies which show that voters’ attitudes towards fiscal and microeconomic issues have been better predictors of the vote for the Australian House of Representatives over the last four elections than their attitudes towards macroeconomic issues. This finding suggests that the cause of the inconsistency between aggregate and individual level models of voting may be that aggregate models of economic voting which include only macroeconomic variables are inadequately specified, since they do not take broader aspects of the economy into account. 1999 Elsevier Science Ltd. All rights reserved. Keywords: Economic voting; Rationality; Micropolitics; Voter attitudes
1. Introduction There are two principal methods of empirically testing the theory of economic voting. The first involves the analysis of economic aggregates such as real income, inflation and unemployment. The second involves determining individual voters’ feelings about the state of the economy using questions in an attitudinal survey. However, aggregate and individual level tests of economic voting for the United States Congress have led to contradictory findings. The same is true for elections to the Australian Parliament. Further inquiry at the aggregate level is unlikely to resolve these inconsistencies (Lewis-Beck, 1988, p. 30). Therefore, this paper uses individual
* Tel.: ⫹ 61-02-9958-3094; fax: ⫹ 61-02-9967-4860 0261-3794/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 1 - 3 7 9 4 ( 9 9 ) 0 0 0 2 2 - 0
506
S. Blount / Electoral Studies 18 (1999) 505–517
level data to investigate an important assumption made in aggregate economic voting models of Australia: that voters at the national level are more likely to respond to macroeconomic issues than fiscal and microeconomic issues.1 I should note that studies of economic voting using individual level data have been severely criticised by Gerald H. Kramer (1983). Kramer argued that individual level survey questions do not distinguish between the economy generally and the government’s effect on it, and that economic voting should be modelled dynamically, rather than statically. Kramer concluded that evidence for economically rational voting at the individual level could be inferred only from aggregate data. Kramer’s first criticism, however, is not really true for welfare states. Where there is government intervention in the economy, it is reasonable for voters to associate changes in the economy with government actions (Lewis-Beck, 1988, pp. 62–63; Nannestad and Paldam, 1995, pp. 37–38). This criticism is also not relevant to the analysis in this paper, since Kramer was referring to a set of questions specifically designed to measure voters’ feelings about the economy (see Fiorina, 1978; Kinder and Kiewiet, 1979), whereas the analysis below focuses on voters’ feelings toward a number of individual election issues. In regard to Kramer’s second point, a solution to the problem may be to combine individual and aggregate level data (Kiewiet and Rivers, 1985), and this approach has been applied by Marcus (1988, 1992). However, in this paper, repeated cross sectional surveys of comparable respondents are used to give some indication of net change over time. Finally, in response to Kramer’s conclusion that economic voting at the individual level could only be inferred from aggregate level evidence, Robert S. Erikson (1990) found that aggregate models were themselves misspecified by demonstrating that once the control for electoral history was correctly specified, the effect of the economy on the congressional vote was reduced to statistical insignificance. This brief discussion, then, suggests that the individual level approach used in this paper is not demonstrably inferior to aggregate techniques. Using attitudinal data from the 1990, 1993, 1996 and 1998 Australian Election Studies (McAllister et al., 1990; Jones et al., 1993, 1996; Bean et al., 1998)2, this paper presents a multivariate analysis which suggests that individuals’ attitudes to fiscal and microeconomic issues are better predictors of the national vote than their attitudes to macroeconomic issues. I conclude that aggregate level models of econ1 Macroeconomic analysis is concerned with economic aggregates such as inflation, interest rates and unemployment. ‘Macroeconomic’ then, is clearly an appropriate label for variables quantifying voters’ attitudes towards these particular aspects of the economy. Microeconomic analysis is more concerned with the structure of particular markets, such as the labour and capital markets. It is appropriate then, to use this term to distinguish variables quantifying individuals’ attitudes towards industrial relations and privatisation. 2 The 1990, 1993, 1996 and 1998 Australian Election Studies were national systematic cross sectional surveys of voters using a mailback questionnaire. In 1993, the states of Queensland, Western Australia, South Australia and Tasmania were oversampled in order to ensure adequate responses for separate state analyses. Weighting this sample by state adjusts the sample to the equivalent of a random sample nationally. Collection of the data was carried out by McAllister et al. (1990); Jones et al. (1993, 1996); Bean et al. (1998). However, I am solely responsible for the analyses and conclusions presented above.
S. Blount / Electoral Studies 18 (1999) 505–517
507
omic voting which characterise the Australian economy solely as the sum of macroeconomic variables do not adequately operationalise the idea of the economy.
2. Aggregate models of economic voting Most aggregate models of economic voting at the national level in the United States have assumed that voters respond to changes in one or more of the macroeconomic conditions of real income, inflation and unemployment (Lewis-Beck, 1988, p. 29). An exception was the analysis by William A. Niskanen (1979), which tested for the influence of taxation as well as the macroeconomy on national elections. In the main, however, measures of aggregate changes in taxation tend to have been used to explain voting only at the subnational level (Eismeier, 1983, p. 369; Kone and Winters, 1993; Niemi et al., 1995; Pomper and Lederman, 1980; Turret, 1971). There are sound reasons for assuming that macroeconomic conditions in the United States should be more influential on the vote at the national level than at the subnational level. Whereas the federal executive may conceivably be held accountable for changes in the macroeconomy, state economies are influenced by external forces and governors have little real power to prevent labour and capital flight (Atkeson and Partin, 1995, p. 100). Fiscal conditions should be more important at the state level because even though the federal government levies the income tax, American state executives have discretion to levy a wide variety of other taxes (see Berry and Berry, 1992, 1994). Aggregate models of economic voting using Australian data have generally followed the procedure of the American literature by including only measures of the macroeconomy as independent variables.3 Christian Leithner (1993) carried out a series of multivariate analyses of data from the interwar years focusing attention on the relationship between real income and the vote. Anthony Mughan (1987) also confined himself to macroeconomic aggregates to carry out a comparative analysis of economic voting between the United Kingdom and Australia. Although Mughan produced significant results for the UK, he found that even after controlling for compulsory voting, lag times, the unusual circumstances of the 1975 federal election, and the effect of minor parties, neither inflation nor unemployment were statistically significant predictors of electoral outcomes in Australia. Roger Douglas (1975), however, included a variable for industrial unrest within an aggregate economic voting model that contained the conventional measures of unemployment, inflation and real wages. Although Douglas established a relationship between both unemployment and inflation and the vote, two flaws in his model suggest that these findings may be invalid. First, Douglas did not control for partisanship, which is by far the most important predictor of voting in Australia (Aitkin, 1982). Secondly, as Douglas himself acknowledged, because a conservative coalition remained in power for the entire
3 Australia has also been included in cross-national studies using aggregate data (e.g. Pacek and Radcliff, 1995).
508
S. Blount / Electoral Studies 18 (1999) 505–517
period of his analysis, it is possible that his results only measured the importance of these variables in determining support for the conservative side of politics. An individual level model of economic voting, however, did control for partisanship and did suggest that economic voting occurred in Australia. David John Gow (1990) investigated the probability of personal and collective economic grievances affecting the retrospective vote against the government. Gow found that the attitudinal data from the 1990 Australian Election Survey supported the proposition that sociotropic retrospective economic voting occurred in Australia.4 Gow suggested that his model succeeded where Mughan’s aggregate model failed because it estimated the probability of first preference voting against the government rather than the two party preferred vote (Gow, 1990, p. 70). Another reason for the failure of Mughan’s model may have been that the variables for inflation and unemployment were too closely correlated, and were therefore not sufficiently distinct measures of economic conditions (Leithner, 1993, p. 343). It is also possible that the longer time frame of Mughan’s aggregate model disadvantaged it compared to the cross-sectional procedure employed by Gow (Niemi et al., 1995, footnote 4). In this paper, I suggest that Mughan’s model failed because aggregate measures of the macroeconomy alone do not operationalise the idea of the Australian economy sufficiently broadly. There are two reasons why macroeconomic variables only partially represent the idea of the economy in Australia. First, in Australia, unlike the United States, the taxing power of the Commonwealth is quite strong relative to the states. Section 51 (ii) of the Australian constitution prohibits the states from collecting sales and excise taxes, and in 1942 the states gave the collection of income taxes over to the Commonwealth (Groenewegen, 1990, pp. 20, 94). Although there is some evidence that Australian voters do distinguish between the taxing responsibilities of the Commonwealth and the states (Compston, 1990, p. 260), the Commonwealth retains almost as much control over fiscal policy as it does over macroeconomic policy. Secondly, Australia has experienced attempts at substantial microeconomic reform in recent years, particularly in the area of industrial relations. Changes to the microeconomy tend to be permanent, whereas changes in the macroeconomy are cyclical. It is consistent with economic rationality that voters should have judged the government on policies which have led to permanent changes in the economy, rather than cyclical changes (Suzuki and Chappell, 1996, pp. 224–226). It follows that a model which does not include measures of fiscal and microeconomic policy does not fully test the theory of economically rational voting.
4 McAllister (1992) (p. 204) also carried out an analysis of the economic voting questions in the 1990 AES, finding that sociotropic evaluations of the economy had a greater influence on the vote than evaluations based on the fortunes of respondents’ own households. He also found evidence that Australians did not really think that government actions had much effect on the economy, supporting the view put forward by Kramer (1983). However, in the absence of a control for partisanship, McAllister’s analysis cannot be considered a test of the economic voting hypothesis itself.
S. Blount / Electoral Studies 18 (1999) 505–517
509
3. Election issues My analysis begins with a brief summary of the issues in the 1990, 1993, 1996 and 1998 elections. The 1990 election tended to highlight environmental and other ‘new politics’ issues, and there is some evidence that these issues had an impact on the vote for the Australian Democrats, particularly in the Senate (McAllister and Bean, 1990; Papadakis, 1990; Blount, 1998). The 1993 election, however, was fought on the merits of a comprehensive platform for radical economic reform put forward by the Coalition. Many of the issues in this election, particularly those of tax reform, industrial relations, unemployment, and healthcare, exerted an unusually important influence on electors’ vote decisions (Bean, 1994). The 1996 election, in contrast, saw a return to a more familiar pattern of Australian electoral politics. The major parties played down the policy divisions between them, preferring to concentrate on questions of leadership and management. The election campaign was also tightly controlled and stage managed, with neither of the two major parties allowing much opportunity for their leaders to discuss policy. Although a few strong issues did emerge, such as taxation, industrial relations and healthcare, there was generally much less public interest in the 1996 election campaign than there had been in the 1993 campaign (Bean and McAllister, 1997, pp. 190–195). Finally, the most recent election was fought on a platform of tax reform, again put forward by the Coalition parties. The election was also notable for the emergence of a new party, ‘One Nation’, which capitalised on the disaffection of rural voters, many of whom were concerned about the decline of rural infrastructure. An issue which encapsulated this concern was the government’s proposal to divest Australia’s main telecommunications provider, Telstra, from public ownership. Despite the different overall policy emphases in each election, the attitudes of electors to a number of core issues, including inflation, interest rates, unemployment, taxation, health, education and the environment were canvassed for all four elections. Clearly, aggregate models of economic voting test the proposition that one or more of the cyclical issues of inflation, interest rates and unemployment have a significant influence on the outcome of national elections. This paper investigates the proposition that it is consistent with the assumption of economic rationality that issues of permanent change, such as taxation, industrial relations and privatisation, should also have an influence on the vote outcome. Three criteria must be fulfilled if an issue is to have an influence upon an individual’s vote decision. The issue must be perceptible, it must be accompanied by some perception that one party represents the person’s own position better than another, and it must arouse some minimal intensity of feeling (Campbell et al., 1960, p. 169–170). Two questions were asked in each of the attitudinal surveys concerning individuals’ attitudes towards election issues. The first question addressed the second condition outlined above, that one party, either the Australian Labor Party or the Liberal–National Coalition, should represent the respondent’s view better than the other. The second question addressed the third condition set out above, that an issue must arouse some intensity of feeling. Combining these two questions creates a variable that directly measures the con-
510
S. Blount / Electoral Studies 18 (1999) 505–517
ditions for issue salience (see Appendix A). The new variable indicates which of the two major parties most nearly represented the respondent’s own position on the issue and indicates the respondent’s intensity of feeling about it. Operationalising these conditions, one would expect the attitudinal frequencies of an issue that might affect an individual’s vote choice to be U formed. On a scale of 0–1, a U shape would indicate large numbers of Labour and Coalition supporters with strong intensity of feeling at points 0 and 1, and few individuals in between. U shaped frequencies are indicated by relatively large, negatively signed kurtosis values. Large standard deviations also suggest that opinion is not centred around the mean, but rather is spread towards opposite poles. Table 1 below gives the mean, standard deviation and kurtosis values for all of the issues surveyed in each election study. The issues are ordered into five groups: macroeconomic, fiscal and microeconTable 1 Salience of issues 1990
1993
k
1996
k
1998
k
k
Macro Unemp Inflat Int Rt
0.49 0.52 0.55
0.32 ⫺1.01 0.53 0.35 ⫺1.15 0.45 0.35 ⫺1.11 0.45
0.37 ⫺1.27 0.60 0.33 ⫺1.01 0.54 0.33 ⫺0.94 0.54
0.36 ⫺1.10 0.44 0.30 ⫺0.58 0.55 0.30 ⫺0.57 0.57
0.36 ⫺1.25 0.29 ⫺0.37 0.29 ⫺0.45
Fiscal/ Taxes Micro GST I. R. Wages Gvt Sp Privat
0.54
0.36 ⫺1.21 0.53 0.47 0.48 0.33 ⫺0.98 0.31 ⫺0.81
0.25 0.34 0.59 0.40 ⫺1.57 0.32 ⫺0.85 0.52
0.31 ⫺0.73 0.51 0.50 0.36 ⫺1.26 0.48
0.34 ⫺1.02 0.42 ⫺1.64 0.31 ⫺0.88
0.52
0.28 ⫺0.40 0.40
0.28 ⫺0.62
Health Health
0.41
0.36 ⫺1.19 0.33
0.40 ⫺1.23 0.49
0.40 ⫺1.54 0.43
0.37 ⫺1.27
Social Educ Envir S. Sec C.Care Immig
0.44 0.32
0.31 ⫺0.75 0.42 0.30 ⫺0.68 0.41 0.42 0.42
0.34 0.28 0.33 0.25
⫺1.08 0.50 ⫺0.49 0.49 ⫺1.07 ⫺0.06 0.60
0.33 ⫺0.92 0.38 0.31 ⫺0.79 0.45
0.43 ⫺1.57 0.27 ⫺0.25
0.28 ⫺0.48 0.49
0.25
0.15
0.47 0.53 0.52
0.26 ⫺0.26 0.50 0.23 0.46 0.19 1.90
0.24
0.44
Politic. Asia Def. State/T
0.45 0.57
Sources: 1990, 1993, 1996, 1998 AES. The measure for taxation in 1993 is the sum of two variables investigating attitudes towards business taxes and tariffs (see Appendix A). The main taxation issue in 1993 was the GST.
S. Blount / Electoral Studies 18 (1999) 505–517
511
omic, health, social, and political issues. Wages and government expenditure were included as microeconomic issues since centralised wage bargaining, which was a part of a prices and incomes accord operating in Australia in 1990, had the effect of delaying reform of the labour market (Kelly, 1992, p. 499), and government expenditure has an impact on the extent and shape of the private economy (see footnote 1). Keeping wages and government expenditure apart from the macroeconomic variables of inflation, interest rates and unemployment also allows for a test of the same macroeconomic variable for all four elections in the multivariate analysis below. Immigration was grouped as a social, rather than a political or economic issue, because the impact of immigration is more likely to be social in the short run, although the political and economic effects of immigration may become more apparent in the long run. Reading the row for unemployment, the mean, which increases from 0.49 to 0.60 from 1990 to 1996, and then falls to 0.44 in 1998, indicates that respondents’ attitudes initially tended to favour the Coalition’s view on the issue of unemployment, but tended to favour the Labour Party’s view on this issue in the last election. The standard deviations are relatively large, suggesting that opinion was moderately spread on this issue. The kurtosis values are also quite large, suggesting that opinion was polarised between the two major parties. The relatively large standard deviation and negatively signed kurtosis values suggest that this issue is quite U shaped, and is likely to have had some influence on the outcome of all four elections. The issue of the environment, on the other hand, consistently favours Labour’s position. However, the standard deviation and kurtosis values are smaller than those for unemployment. This suggests that the issue was less likely to have had an impact in the elections. An extreme example of a weak issue is Commonwealth–state relations in the 1996 election. The very small standard deviation and large positively signed kurtosis value indicate that this issue has the shape of an inverted U, with a large number of responses clustered around the centre.5 Such an issue is extremely unlikely to have had an impact on the outcome of the election under the conditions set out by Campbell et al. (1960). In general, the small standard deviations and small or positively signed kurtosis values indicate that political issues were unlikely to have had a significant effect on the outcome of the 1996 and 1998 elections. With the possible exception of education, particularly in the 1998 election, social issues were also unlikely to have affected the outcomes of any of the elections. The issue of health, on the other hand, is likely to have had a significant effect on all four elections. As far as the economic issues are concerned, voters’ attitudes towards fiscal and microeconomic issues appear likely to have had a greater impact on the last four election outcomes in Australia than their attitudes towards macroeconomic issues. In each year, either a taxation or an industrial relations issue had a larger standard deviation and kurtosis value than any of the corresponding macroeconomic issues.
5 This indicates that many respondents either did not know which party was closest to their own view of the issue, or thought there was no difference between the parties (see Appendix A).
512
S. Blount / Electoral Studies 18 (1999) 505–517
Table 2 The importance of issues in the 1990, 1993, 1996, and 1998 Federal elections (betas) 1990 PID Lab. Leader Lib. Leader Macro Fisc/Micro Health Social Political R
1993
0.49** ⫺ 0.18** 0.07** 0.10** 0.10** 0.04** 0.00 n.a. 0.75
0.42** ⫺ 0.19** 0.12** 0.06** 0.14** 0.04** ⫺ 0.00 n.a. 0.77
1996 0.44** ⫺ 0.17** 0.11** 0.03 0.14** 0.06** 0.04 ⫺ 0.03 0.73
1998 0.50** ⫺ 0.07** 0.03 0.01 0.29** 0.05 0.00 0.00a 0.73
a The political variable here is the single issue of Australia’s relationship with Asia. Sources: 1990, 1993, 1996, 1996, 1998 AES. **P ⬍ 0.01.
In 1996 and 1998 in particular, it is unlikely that the issues of inflation and interest rates had a significant impact on individuals’ vote choices. A multivariate analysis tests these findings in a more rigorous way. The issues for each year were summed into macroeconomic, fiscal and microeconomic, social, and political scales, each confirmed by an alpha test of reliability (see Appendix A). These scales were then included, along with the variable for health, in a multivariate analysis to determine the relative weight of their impact upon the vote for each year. To reflect the reality of Australian elections, in which party allegiance plays an important role (Aitkin, 1982), the issues were analysed within a party identification model. The variables for party identification and party leaders which appear in the analysis below act as controls to prevent the overestimation of the influence of the issues on the vote. The model below tests the proposition that the attitudes of electors on microeconomic and fiscal issues has a greater influence on their vote than their attitudes towards macroeconomic issues. Comparing the relative influence of issues on the vote suggests that the standardised regression coefficient (beta) should be the result of interest. For each year, the regression equation is in the form: Hr ⫽ f(PID, Leaders, Issues) where Hr is the vote for the House of Representatives, PID is party identification, Leaders is the evaluation of the two main party leaders, and Issues represents macroeconomic, fiscal and microeconomic, social, political and health issues. The standardised regression coefficients suggest that in the past four federal elections in Australia, taxation and microeconomic issues have accounted for as much, or more, of the variation in the vote as macroeconomic issues. As the party identification model predicts, both party identification and attitudes towards party leaders
S. Blount / Electoral Studies 18 (1999) 505–517
513
account for much of the variation in the vote.6 The insignificant beta weights for social and political issues are also consistent with the findings of the summary analysis in Table 1. The three sets of issues which are highly significant are macroeconomic, fiscal and microeconomic, and health issues. However, macroeconomic issues appear to have had an insignificant influence on the vote in both the 1996 and 1998 elections. The beta weights suggest that microeconomic and fiscal issues were at least twice as influential as any other set of issues in the 1993 and 1996 elections, and in 1998, fiscal and microeconomic issues were a dominant influence upon the vote. These results are consistent with the suggestion that individuals do take economic conditions into account when voting for the Australian Parliament, but that the cyclical issues of inflation, interest rates and unemployment alone are inadequate measures of economic voting.
4. Conclusion Aggregate models of economic voting test the theory that voters make political decisions based on rational assessments of the economy. At the national level, these models have conventionally tested for a significant relationship between one or more of the macroeconomic variables of unemployment, interest rates, inflation, and the vote. However, an individual level analysis of the last four federal elections in Australia suggests that individuals’ attitudes towards these cyclical economic issues have had increasingly less influence on the way they voted than their attitudes towards issues of permanent economic change, such as taxation, industrial relations and privatisation. These findings reflect the particular circumstances of recent Australian elections, which have taken place within the context of significant attempts to introduce changes in the areas of taxation, conditions of employment and publicly owned utilities. But these findings also have implications beyond Australia, since they suggest that the theory of economic voting cannot be adequately tested by models which operationalise the idea of the economy solely as the sum of cyclical change economic variables. I conclude that aggregate models which include measures of fiscal and microeconomic change, as well as macroeconomic change, would offer a more powerful test of the theory of economic voting because they operationalise the idea of the economy in a more complete and meaningful way.
6 The insignificant result for the Prime Minister in the 1998 analysis may be accounted for by the the high profile of the leader of the emerging ‘One Nation’ party who tended to appeal to disaffected Coalition supporters.
514
S. Blount / Electoral Studies 18 (1999) 505–517
Appendix A Variable questions and values All questions are as they appear in the 1990 Australian Election Study. Substantial variations in subsequent surveys appear in brackets ⬍ > .
Issues Here is a list of important issues that were discussed during the election campaign. Whose policies—the Labor Party’s or the Liberal–National Coalition’s would you say came closest to your own views on each of these issues? For each issue, respondents were asked to reply: 1 ⫽ Labour; 2 ⫽ Liberal– National; 3 ⫽ No Difference; 4 ⫽ Don’t Know. Still thinking about these same issues, when you were deciding about how to vote, how important was each of these issues to you personally? For each issue, respondents were asked to reply: 1 ⫽ Extremely Important; 2 ⫽ Quite Important; 3 ⫽ Not Very Important. The combined variables in Table 1 were scored: 0 ⫽ Labour, Extremely Important; 2 ⫽ Labour, Quite Important; 4 ⫽ Labour, Not Very Important; 5 ⫽ No Difference/Don’t Know; 6 ⫽ Coalition, Not Very Important; 8 ⫽ Coalition, Quite important; 1 ⫽ Coalition, Extremely Important. The issues given in each survey were: 1990: Unemployment, Inflation, Interest Rates, Education, Environment, Health, Taxation, Goverment Spending, Wages. 1993: Unemployment, Inflation, Interest Rates, Education, Environment, Health, Medicare, GST, Business Taxes, Tariffs, Employment Contracts, Enterprise Bargaining, Social Security, Childcare. 1996: Unemployment, Inflation, Interest Rates, Education, Environment, Health and Medicare, Taxation, Industrial Relations, Immigration, Privatisation, Commonwealth–state relations, Asia, Defence. 1998: Unemployment, Inflation, Interest Rates, Education, Environment, Health, Taxation, GST, Industrial Relations, Immigration, Privatisation, Telstra, Asia. To make the issues in Table 1 more directly comparable, three sets of 1993 issues were combined: Health ⫹ Medicare ⫽ Health, Employment Contracts ⫹ Enterprise Bargaining ⫽ Industrial Relations, Business Taxes ⫹ Tariffs ⫽ Taxes. Two issues in 1998 were also combined in this table: Privatisation ⫹ Telstra ⫽ Privatisation. The scales in Table 2 were created as follows: 1990: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.77), Taxation ⫹ Goverment Spending ⫹ Wages ⫽ Micro (␣ ⫽ 0.79), Education ⫹ Environment ⫽ Social (␣ ⫽ 0.63), Health ⫽ Health. 1993: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.83), GST ⫹ Business Taxes ⫹ Tariffs ⫹ Employment Contracts ⫹ Enterprise Bargaining ⫽
S. Blount / Electoral Studies 18 (1999) 505–517
515
Micro (␣ ⫽ 0.85), Education ⫹ Environment ⫹ Social Security ⫹ Childcare ⫽ Social (␣ ⫽ 0.81), Health ⫹ Medicare ⫽ Health (␣ ⫽ 0.90). 1996: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.80), Taxation ⫹ Industrial Relations ⫹ Privatisation ⫽ Micro (␣ ⫽ 0.77), Education ⫹ Environment ⫹ Immigration ⫽ Social (␣ ⫽ 0.71), Health and Medicare ⫽ Health, Commonwealth–state relations ⫹ Asia ⫹ Defence ⫽ Politic (␣ ⫽ 0.74). 1998: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.79), Taxation ⫹ GST ⫹ Industrial Relations ⫹ Privatisation ⫹ Telstra ⫽ Micro (␣ ⫽ 0.85), Education ⫹ Environment ⫹ Immigration ⫽ Social (␣ ⫽ 0.76), Health ⫽ Health, Asia ⫽ Politic. Vote for the House of Representatives and Senate Which party did you vote for first in the House of Representatives? For each question, respondents were asked to reply: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Other/Independents; 6 ⫽ Green Parties/NDP; 7 ⫽ Call to Australia; 8 ⫽ Grey Power; 9 ⫽ Informal/Didn’t Vote. ⬍ 1993: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Other. > ⬍ 1996: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ Other; 9 ⫽ Informal/Didn’t Vote. > ⬍ 1998: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ One Nation; 7 ⫽ Other Party; 9 ⫽ Informal/Didn’t Vote. > In order to model the vote as variation along a left–right axis for the 1990, 1993 and 1996 elections, the scoring was: Labour Party ⫽ 0, Coalition ⫽ 1, Other ⫽ 5. In 1998, the scoring was: Labour Party ⫽ 0, Coalition and One Nation ⫽ 1, Other ⫽ 5. Scoring the One Nation Party the same as the Coalition parties reflects the fact that One Nation drew a substantial portion of its votes from disaffected Coalition supporters.
Party identification Generally speaking, do you usually think of yourself as Liberal, Labour, National or what? Respondents were asked to reply: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Green; 6 ⫽ Independent; 8 ⫽ Coalition; 9 ⫽ No political Id. ⬍ 1993: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Other party (please specify); 6 ⫽ No party. > ⬍ 1996: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ Other Party; 7 ⫽ No Party. >
516
S. Blount / Electoral Studies 18 (1999) 505–517
⬍ 1998: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ One Nation; 7 ⫽ Other Party; 8 ⫽ No Party. > These categories were recoded to agree with the dependent variable: 0 ⫽ Labour, 1 ⫽ Coalition and One Nation; 5 ⫽ Other. Attitudes to leaders ...using the 0 to 10 scale please show how favourable or unfavourable you feel about the party leaders. Respondents were asked to circle a number on a scale for each candidate. The scale ranged from 0 to 10, where 0 represented ‘very unfavourable’, 5 represented neutral, and 10 represented ‘very favourable’. The responses create a Likert scale for each leader. A positively signed coefficient indicates a response favouring the Coalition leader. A negatively signed coefficient indicates a response favouring the Labour leader.
References Aitkin, D., 1982. Stability and Change in Australian Politics, 2nd ed. Australian National University Press, Canberra. Atkeson, L.R., Partin, R.W., 1995. Economic and referendum voting: a comparison of gubertanorial and senatorial elections. American Political Science Review 89 (1), 99–107. Bean, C., 1994. Issues in the 1993 Election. Australian Journal of Political Science 29, 134–157. Bean, C., McAllister, I., 1997. Short term influences on voting behaviour in the 1996 election. In: Bean, C., Bennett, S., Simms, M., Wharhurst, J. (Eds.), The Politics of Retribution: The 1996 Australian Federal Election. Allen and Unwin, St Leonards, N.S.W., Australia. Bean, C., Gow, D., McAllister, I., 1998. Australian election study 1998 [computer file]. Social Sciences Data Archive, The Australian National University, Canberra. Blount, S., 1998. Postmaterialism and the vote for the Senate in Australia. Australian Journal of Political Science 33 (3), 441–449. Berry, F.S., Berry, W.O., 1992. Tax innovation in the States: capitalising on political opportunity. American Journal of Political Science 36, 715–742. Berry, F.S., Berry, O.B., 1994. The politics of tax increases in the States. American Journal of Political Science 38 (3), 855–859. Campbell, A., Converse, P.S., Miller, W.S., Stokes, D.E., 1960. The American Voter. Wiley, New York. Compston, H., 1990. The electoral impact of tax: a comparative study of Sweden and Australia. Unpublished PhD Thesis, Australian National University, Canberra. Douglas, R., 1975. Economy and polity in Australia: a quantification of commonsense. British Journal of Political Science 5, 341–361. Eismeier, T.J., 1983. Votes and taxes: the political economy of the American governorship. Polity 15, 368–379. Erikson, R.S., 1990. Economic conditions and the congressional vote: a review of the macrolevel evidence. American Journal of Political Science 34 (2), 373–399. Fiorina, M.P., 1978. Economic retrospective voting in American national elections: a micro analysis. American Journal of Political Science 22 (2), 426–443. Gow, D.J., 1990. Economic voting and postmaterialist values. In: Bean, C., McAllister, I., Warhurst, J. (Eds.), The Greening of Australian Politics. Longman Cheshire, Melbourne. Groenewegen, P., 1990. Public Finance in Australia: Theory and Practice, 3rd ed. Prentice Hall, Sydney.
S. Blount / Electoral Studies 18 (1999) 505–517
517
Jones, R., McAllister, I., Denemark, D., Gow, D., 1993. Australian election study [computer file]. Social Science Data Archives, The Australian National University, Canberra. Jones, R., McAllister, I., Gow, D., 1996. Australian election study [computer file]. Social Science Data Archives, The Australian National University, Canberra. Kelly, P., 1992. The End of Certainty: The Story of the 1980s. Allen and Unwin, Sydney. Kiewiet, D.R., Rivers, D., 1985. A retrospective on retrospective voting. In: Eulau, H., Lewis Beck, M. (Eds.), Economic Conditions and Electoral Outcomes: The United States and Western Europe. Agathon Press, NY. Kinder, D.R., Kiewiet, R.D., 1979. Economic discontent and political behaviour: the role of personal grievances and collective economic judgements in congressional voting. American Journal of Political Science 23 (3), 495–527. Kone, S.L., Winters, R.F., 1993. Taxes and voting: electoral retribution in the American states. The Journal of Politics 55 (1), 22–40. Kramer, G.H., 1983. The ecological fallacy revisited: aggregate- versus individual-level findings on economics and elections, and sociotropic voting. American Political Science Review 77 (1), 92–111. Leithner, C., 1993. Economic conditions and the vote: a contingent rather than categorical influence. British Journal of Political Science 23, 339–372. Lewis-Beck, M.S., 1988. Economics and Elections: The Major Western Democracies. University of Michigan Press, Ann Arbor. Marcus, G.B., 1988. The impact of personal and national economic conditions on the presidential vote: a pooled cross sectional analysis. American Journal of Political Science 32 (1), 137–154. Marcus, G.B., 1992. The impact of personal and national economic conditions on the presidential vote, 1956–1988. American Journal of Political Science 36 (3), 829–834. McAllister, I., 1992. Political Behaviour: Citizens, Parties and Elites in Australia. Longman Cheshire, Melbourne. McAllister, I., Bean, C., 1990. Explaining Labor’s victory. In: Bean, C., McAllister, I., Warhurst, J. (Eds.), The Greening of Australian Politics. Longman Cheshire, Melbourne. McAllister, I., Jones, R., Papadakis, E., Gow, D., 1990. Australian election study [computer file]. Social Science Data Archives, The Australian National University, Canberra. Mughan, A., 1987. The ‘hip pocket nerve’ and electoral volatility in Australia and Great Britain. Politics 22 (2), 66–75. Nannestad, P., Paldam, M., 1995. It’s the Goverment’s fault! A cross-section study of economic voting in Denmark, 1990/93. European Journal of Political Research 28, 33–62. Niemi, R.G., Stanley, H.W., Vogel, R.J., 1995. State economies and state taxes: do voters hold governors accountable? American Journal of Political Science 39 (4), 936–957. Niskanen, W.A., 1979. Economic and fiscal effects on the popular vote for the President. In: Rae, D.W., Eismeier, T.J. (Eds.), Public Policy and Public Choice. Sage Publications, Beverly Hills. Pacek, A.C., Radcliff, B., 1995. Economic voting and the welfare state. The Journal of Politics 57 (1), 44–61. Papadakis, E., 1990. Minor parties, the environment and the new politics. In: Bean, C., McAllister, I., Warhurst, J. (Eds.), The Greening of Australian Politics. Longman Cheshire, Melbourne. Pomper, G.M., Lederman, S.S., 1980. Elections in America: Control and Influence in Democratic Politics. Longman, New York. Suzuki, M., Chappell, H.W., 1996. The rationality of economic voting revisited. Journal of Politics 58 (1), 224–236. Turret, S., 1971. The vulnerability of American governors 1900–1969. Midwest Journal of Political Science 15, 108–132.
The microeconomic voter Simon Blount* University of New South Wales, Sydney, N.S.W. 2052, Australia
Abstract In the United States, aggregate and individual level studies of economic voting for the Congress have produced contradictory findings. The same is true for models of economic voting for the Australian Parliament. This paper presents data taken from a series of individual level studies which show that voters’ attitudes towards fiscal and microeconomic issues have been better predictors of the vote for the Australian House of Representatives over the last four elections than their attitudes towards macroeconomic issues. This finding suggests that the cause of the inconsistency between aggregate and individual level models of voting may be that aggregate models of economic voting which include only macroeconomic variables are inadequately specified, since they do not take broader aspects of the economy into account. 1999 Elsevier Science Ltd. All rights reserved. Keywords: Economic voting; Rationality; Micropolitics; Voter attitudes
1. Introduction There are two principal methods of empirically testing the theory of economic voting. The first involves the analysis of economic aggregates such as real income, inflation and unemployment. The second involves determining individual voters’ feelings about the state of the economy using questions in an attitudinal survey. However, aggregate and individual level tests of economic voting for the United States Congress have led to contradictory findings. The same is true for elections to the Australian Parliament. Further inquiry at the aggregate level is unlikely to resolve these inconsistencies (Lewis-Beck, 1988, p. 30). Therefore, this paper uses individual
* Tel.: ⫹ 61-02-9958-3094; fax: ⫹ 61-02-9967-4860 0261-3794/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 1 - 3 7 9 4 ( 9 9 ) 0 0 0 2 2 - 0
506
S. Blount / Electoral Studies 18 (1999) 505–517
level data to investigate an important assumption made in aggregate economic voting models of Australia: that voters at the national level are more likely to respond to macroeconomic issues than fiscal and microeconomic issues.1 I should note that studies of economic voting using individual level data have been severely criticised by Gerald H. Kramer (1983). Kramer argued that individual level survey questions do not distinguish between the economy generally and the government’s effect on it, and that economic voting should be modelled dynamically, rather than statically. Kramer concluded that evidence for economically rational voting at the individual level could be inferred only from aggregate data. Kramer’s first criticism, however, is not really true for welfare states. Where there is government intervention in the economy, it is reasonable for voters to associate changes in the economy with government actions (Lewis-Beck, 1988, pp. 62–63; Nannestad and Paldam, 1995, pp. 37–38). This criticism is also not relevant to the analysis in this paper, since Kramer was referring to a set of questions specifically designed to measure voters’ feelings about the economy (see Fiorina, 1978; Kinder and Kiewiet, 1979), whereas the analysis below focuses on voters’ feelings toward a number of individual election issues. In regard to Kramer’s second point, a solution to the problem may be to combine individual and aggregate level data (Kiewiet and Rivers, 1985), and this approach has been applied by Marcus (1988, 1992). However, in this paper, repeated cross sectional surveys of comparable respondents are used to give some indication of net change over time. Finally, in response to Kramer’s conclusion that economic voting at the individual level could only be inferred from aggregate level evidence, Robert S. Erikson (1990) found that aggregate models were themselves misspecified by demonstrating that once the control for electoral history was correctly specified, the effect of the economy on the congressional vote was reduced to statistical insignificance. This brief discussion, then, suggests that the individual level approach used in this paper is not demonstrably inferior to aggregate techniques. Using attitudinal data from the 1990, 1993, 1996 and 1998 Australian Election Studies (McAllister et al., 1990; Jones et al., 1993, 1996; Bean et al., 1998)2, this paper presents a multivariate analysis which suggests that individuals’ attitudes to fiscal and microeconomic issues are better predictors of the national vote than their attitudes to macroeconomic issues. I conclude that aggregate level models of econ1 Macroeconomic analysis is concerned with economic aggregates such as inflation, interest rates and unemployment. ‘Macroeconomic’ then, is clearly an appropriate label for variables quantifying voters’ attitudes towards these particular aspects of the economy. Microeconomic analysis is more concerned with the structure of particular markets, such as the labour and capital markets. It is appropriate then, to use this term to distinguish variables quantifying individuals’ attitudes towards industrial relations and privatisation. 2 The 1990, 1993, 1996 and 1998 Australian Election Studies were national systematic cross sectional surveys of voters using a mailback questionnaire. In 1993, the states of Queensland, Western Australia, South Australia and Tasmania were oversampled in order to ensure adequate responses for separate state analyses. Weighting this sample by state adjusts the sample to the equivalent of a random sample nationally. Collection of the data was carried out by McAllister et al. (1990); Jones et al. (1993, 1996); Bean et al. (1998). However, I am solely responsible for the analyses and conclusions presented above.
S. Blount / Electoral Studies 18 (1999) 505–517
507
omic voting which characterise the Australian economy solely as the sum of macroeconomic variables do not adequately operationalise the idea of the economy.
2. Aggregate models of economic voting Most aggregate models of economic voting at the national level in the United States have assumed that voters respond to changes in one or more of the macroeconomic conditions of real income, inflation and unemployment (Lewis-Beck, 1988, p. 29). An exception was the analysis by William A. Niskanen (1979), which tested for the influence of taxation as well as the macroeconomy on national elections. In the main, however, measures of aggregate changes in taxation tend to have been used to explain voting only at the subnational level (Eismeier, 1983, p. 369; Kone and Winters, 1993; Niemi et al., 1995; Pomper and Lederman, 1980; Turret, 1971). There are sound reasons for assuming that macroeconomic conditions in the United States should be more influential on the vote at the national level than at the subnational level. Whereas the federal executive may conceivably be held accountable for changes in the macroeconomy, state economies are influenced by external forces and governors have little real power to prevent labour and capital flight (Atkeson and Partin, 1995, p. 100). Fiscal conditions should be more important at the state level because even though the federal government levies the income tax, American state executives have discretion to levy a wide variety of other taxes (see Berry and Berry, 1992, 1994). Aggregate models of economic voting using Australian data have generally followed the procedure of the American literature by including only measures of the macroeconomy as independent variables.3 Christian Leithner (1993) carried out a series of multivariate analyses of data from the interwar years focusing attention on the relationship between real income and the vote. Anthony Mughan (1987) also confined himself to macroeconomic aggregates to carry out a comparative analysis of economic voting between the United Kingdom and Australia. Although Mughan produced significant results for the UK, he found that even after controlling for compulsory voting, lag times, the unusual circumstances of the 1975 federal election, and the effect of minor parties, neither inflation nor unemployment were statistically significant predictors of electoral outcomes in Australia. Roger Douglas (1975), however, included a variable for industrial unrest within an aggregate economic voting model that contained the conventional measures of unemployment, inflation and real wages. Although Douglas established a relationship between both unemployment and inflation and the vote, two flaws in his model suggest that these findings may be invalid. First, Douglas did not control for partisanship, which is by far the most important predictor of voting in Australia (Aitkin, 1982). Secondly, as Douglas himself acknowledged, because a conservative coalition remained in power for the entire
3 Australia has also been included in cross-national studies using aggregate data (e.g. Pacek and Radcliff, 1995).
508
S. Blount / Electoral Studies 18 (1999) 505–517
period of his analysis, it is possible that his results only measured the importance of these variables in determining support for the conservative side of politics. An individual level model of economic voting, however, did control for partisanship and did suggest that economic voting occurred in Australia. David John Gow (1990) investigated the probability of personal and collective economic grievances affecting the retrospective vote against the government. Gow found that the attitudinal data from the 1990 Australian Election Survey supported the proposition that sociotropic retrospective economic voting occurred in Australia.4 Gow suggested that his model succeeded where Mughan’s aggregate model failed because it estimated the probability of first preference voting against the government rather than the two party preferred vote (Gow, 1990, p. 70). Another reason for the failure of Mughan’s model may have been that the variables for inflation and unemployment were too closely correlated, and were therefore not sufficiently distinct measures of economic conditions (Leithner, 1993, p. 343). It is also possible that the longer time frame of Mughan’s aggregate model disadvantaged it compared to the cross-sectional procedure employed by Gow (Niemi et al., 1995, footnote 4). In this paper, I suggest that Mughan’s model failed because aggregate measures of the macroeconomy alone do not operationalise the idea of the Australian economy sufficiently broadly. There are two reasons why macroeconomic variables only partially represent the idea of the economy in Australia. First, in Australia, unlike the United States, the taxing power of the Commonwealth is quite strong relative to the states. Section 51 (ii) of the Australian constitution prohibits the states from collecting sales and excise taxes, and in 1942 the states gave the collection of income taxes over to the Commonwealth (Groenewegen, 1990, pp. 20, 94). Although there is some evidence that Australian voters do distinguish between the taxing responsibilities of the Commonwealth and the states (Compston, 1990, p. 260), the Commonwealth retains almost as much control over fiscal policy as it does over macroeconomic policy. Secondly, Australia has experienced attempts at substantial microeconomic reform in recent years, particularly in the area of industrial relations. Changes to the microeconomy tend to be permanent, whereas changes in the macroeconomy are cyclical. It is consistent with economic rationality that voters should have judged the government on policies which have led to permanent changes in the economy, rather than cyclical changes (Suzuki and Chappell, 1996, pp. 224–226). It follows that a model which does not include measures of fiscal and microeconomic policy does not fully test the theory of economically rational voting.
4 McAllister (1992) (p. 204) also carried out an analysis of the economic voting questions in the 1990 AES, finding that sociotropic evaluations of the economy had a greater influence on the vote than evaluations based on the fortunes of respondents’ own households. He also found evidence that Australians did not really think that government actions had much effect on the economy, supporting the view put forward by Kramer (1983). However, in the absence of a control for partisanship, McAllister’s analysis cannot be considered a test of the economic voting hypothesis itself.
S. Blount / Electoral Studies 18 (1999) 505–517
509
3. Election issues My analysis begins with a brief summary of the issues in the 1990, 1993, 1996 and 1998 elections. The 1990 election tended to highlight environmental and other ‘new politics’ issues, and there is some evidence that these issues had an impact on the vote for the Australian Democrats, particularly in the Senate (McAllister and Bean, 1990; Papadakis, 1990; Blount, 1998). The 1993 election, however, was fought on the merits of a comprehensive platform for radical economic reform put forward by the Coalition. Many of the issues in this election, particularly those of tax reform, industrial relations, unemployment, and healthcare, exerted an unusually important influence on electors’ vote decisions (Bean, 1994). The 1996 election, in contrast, saw a return to a more familiar pattern of Australian electoral politics. The major parties played down the policy divisions between them, preferring to concentrate on questions of leadership and management. The election campaign was also tightly controlled and stage managed, with neither of the two major parties allowing much opportunity for their leaders to discuss policy. Although a few strong issues did emerge, such as taxation, industrial relations and healthcare, there was generally much less public interest in the 1996 election campaign than there had been in the 1993 campaign (Bean and McAllister, 1997, pp. 190–195). Finally, the most recent election was fought on a platform of tax reform, again put forward by the Coalition parties. The election was also notable for the emergence of a new party, ‘One Nation’, which capitalised on the disaffection of rural voters, many of whom were concerned about the decline of rural infrastructure. An issue which encapsulated this concern was the government’s proposal to divest Australia’s main telecommunications provider, Telstra, from public ownership. Despite the different overall policy emphases in each election, the attitudes of electors to a number of core issues, including inflation, interest rates, unemployment, taxation, health, education and the environment were canvassed for all four elections. Clearly, aggregate models of economic voting test the proposition that one or more of the cyclical issues of inflation, interest rates and unemployment have a significant influence on the outcome of national elections. This paper investigates the proposition that it is consistent with the assumption of economic rationality that issues of permanent change, such as taxation, industrial relations and privatisation, should also have an influence on the vote outcome. Three criteria must be fulfilled if an issue is to have an influence upon an individual’s vote decision. The issue must be perceptible, it must be accompanied by some perception that one party represents the person’s own position better than another, and it must arouse some minimal intensity of feeling (Campbell et al., 1960, p. 169–170). Two questions were asked in each of the attitudinal surveys concerning individuals’ attitudes towards election issues. The first question addressed the second condition outlined above, that one party, either the Australian Labor Party or the Liberal–National Coalition, should represent the respondent’s view better than the other. The second question addressed the third condition set out above, that an issue must arouse some intensity of feeling. Combining these two questions creates a variable that directly measures the con-
510
S. Blount / Electoral Studies 18 (1999) 505–517
ditions for issue salience (see Appendix A). The new variable indicates which of the two major parties most nearly represented the respondent’s own position on the issue and indicates the respondent’s intensity of feeling about it. Operationalising these conditions, one would expect the attitudinal frequencies of an issue that might affect an individual’s vote choice to be U formed. On a scale of 0–1, a U shape would indicate large numbers of Labour and Coalition supporters with strong intensity of feeling at points 0 and 1, and few individuals in between. U shaped frequencies are indicated by relatively large, negatively signed kurtosis values. Large standard deviations also suggest that opinion is not centred around the mean, but rather is spread towards opposite poles. Table 1 below gives the mean, standard deviation and kurtosis values for all of the issues surveyed in each election study. The issues are ordered into five groups: macroeconomic, fiscal and microeconTable 1 Salience of issues 1990
1993
k
1996
k
1998
k
k
Macro Unemp Inflat Int Rt
0.49 0.52 0.55
0.32 ⫺1.01 0.53 0.35 ⫺1.15 0.45 0.35 ⫺1.11 0.45
0.37 ⫺1.27 0.60 0.33 ⫺1.01 0.54 0.33 ⫺0.94 0.54
0.36 ⫺1.10 0.44 0.30 ⫺0.58 0.55 0.30 ⫺0.57 0.57
0.36 ⫺1.25 0.29 ⫺0.37 0.29 ⫺0.45
Fiscal/ Taxes Micro GST I. R. Wages Gvt Sp Privat
0.54
0.36 ⫺1.21 0.53 0.47 0.48 0.33 ⫺0.98 0.31 ⫺0.81
0.25 0.34 0.59 0.40 ⫺1.57 0.32 ⫺0.85 0.52
0.31 ⫺0.73 0.51 0.50 0.36 ⫺1.26 0.48
0.34 ⫺1.02 0.42 ⫺1.64 0.31 ⫺0.88
0.52
0.28 ⫺0.40 0.40
0.28 ⫺0.62
Health Health
0.41
0.36 ⫺1.19 0.33
0.40 ⫺1.23 0.49
0.40 ⫺1.54 0.43
0.37 ⫺1.27
Social Educ Envir S. Sec C.Care Immig
0.44 0.32
0.31 ⫺0.75 0.42 0.30 ⫺0.68 0.41 0.42 0.42
0.34 0.28 0.33 0.25
⫺1.08 0.50 ⫺0.49 0.49 ⫺1.07 ⫺0.06 0.60
0.33 ⫺0.92 0.38 0.31 ⫺0.79 0.45
0.43 ⫺1.57 0.27 ⫺0.25
0.28 ⫺0.48 0.49
0.25
0.15
0.47 0.53 0.52
0.26 ⫺0.26 0.50 0.23 0.46 0.19 1.90
0.24
0.44
Politic. Asia Def. State/T
0.45 0.57
Sources: 1990, 1993, 1996, 1998 AES. The measure for taxation in 1993 is the sum of two variables investigating attitudes towards business taxes and tariffs (see Appendix A). The main taxation issue in 1993 was the GST.
S. Blount / Electoral Studies 18 (1999) 505–517
511
omic, health, social, and political issues. Wages and government expenditure were included as microeconomic issues since centralised wage bargaining, which was a part of a prices and incomes accord operating in Australia in 1990, had the effect of delaying reform of the labour market (Kelly, 1992, p. 499), and government expenditure has an impact on the extent and shape of the private economy (see footnote 1). Keeping wages and government expenditure apart from the macroeconomic variables of inflation, interest rates and unemployment also allows for a test of the same macroeconomic variable for all four elections in the multivariate analysis below. Immigration was grouped as a social, rather than a political or economic issue, because the impact of immigration is more likely to be social in the short run, although the political and economic effects of immigration may become more apparent in the long run. Reading the row for unemployment, the mean, which increases from 0.49 to 0.60 from 1990 to 1996, and then falls to 0.44 in 1998, indicates that respondents’ attitudes initially tended to favour the Coalition’s view on the issue of unemployment, but tended to favour the Labour Party’s view on this issue in the last election. The standard deviations are relatively large, suggesting that opinion was moderately spread on this issue. The kurtosis values are also quite large, suggesting that opinion was polarised between the two major parties. The relatively large standard deviation and negatively signed kurtosis values suggest that this issue is quite U shaped, and is likely to have had some influence on the outcome of all four elections. The issue of the environment, on the other hand, consistently favours Labour’s position. However, the standard deviation and kurtosis values are smaller than those for unemployment. This suggests that the issue was less likely to have had an impact in the elections. An extreme example of a weak issue is Commonwealth–state relations in the 1996 election. The very small standard deviation and large positively signed kurtosis value indicate that this issue has the shape of an inverted U, with a large number of responses clustered around the centre.5 Such an issue is extremely unlikely to have had an impact on the outcome of the election under the conditions set out by Campbell et al. (1960). In general, the small standard deviations and small or positively signed kurtosis values indicate that political issues were unlikely to have had a significant effect on the outcome of the 1996 and 1998 elections. With the possible exception of education, particularly in the 1998 election, social issues were also unlikely to have affected the outcomes of any of the elections. The issue of health, on the other hand, is likely to have had a significant effect on all four elections. As far as the economic issues are concerned, voters’ attitudes towards fiscal and microeconomic issues appear likely to have had a greater impact on the last four election outcomes in Australia than their attitudes towards macroeconomic issues. In each year, either a taxation or an industrial relations issue had a larger standard deviation and kurtosis value than any of the corresponding macroeconomic issues.
5 This indicates that many respondents either did not know which party was closest to their own view of the issue, or thought there was no difference between the parties (see Appendix A).
512
S. Blount / Electoral Studies 18 (1999) 505–517
Table 2 The importance of issues in the 1990, 1993, 1996, and 1998 Federal elections (betas) 1990 PID Lab. Leader Lib. Leader Macro Fisc/Micro Health Social Political R
1993
0.49** ⫺ 0.18** 0.07** 0.10** 0.10** 0.04** 0.00 n.a. 0.75
0.42** ⫺ 0.19** 0.12** 0.06** 0.14** 0.04** ⫺ 0.00 n.a. 0.77
1996 0.44** ⫺ 0.17** 0.11** 0.03 0.14** 0.06** 0.04 ⫺ 0.03 0.73
1998 0.50** ⫺ 0.07** 0.03 0.01 0.29** 0.05 0.00 0.00a 0.73
a The political variable here is the single issue of Australia’s relationship with Asia. Sources: 1990, 1993, 1996, 1996, 1998 AES. **P ⬍ 0.01.
In 1996 and 1998 in particular, it is unlikely that the issues of inflation and interest rates had a significant impact on individuals’ vote choices. A multivariate analysis tests these findings in a more rigorous way. The issues for each year were summed into macroeconomic, fiscal and microeconomic, social, and political scales, each confirmed by an alpha test of reliability (see Appendix A). These scales were then included, along with the variable for health, in a multivariate analysis to determine the relative weight of their impact upon the vote for each year. To reflect the reality of Australian elections, in which party allegiance plays an important role (Aitkin, 1982), the issues were analysed within a party identification model. The variables for party identification and party leaders which appear in the analysis below act as controls to prevent the overestimation of the influence of the issues on the vote. The model below tests the proposition that the attitudes of electors on microeconomic and fiscal issues has a greater influence on their vote than their attitudes towards macroeconomic issues. Comparing the relative influence of issues on the vote suggests that the standardised regression coefficient (beta) should be the result of interest. For each year, the regression equation is in the form: Hr ⫽ f(PID, Leaders, Issues) where Hr is the vote for the House of Representatives, PID is party identification, Leaders is the evaluation of the two main party leaders, and Issues represents macroeconomic, fiscal and microeconomic, social, political and health issues. The standardised regression coefficients suggest that in the past four federal elections in Australia, taxation and microeconomic issues have accounted for as much, or more, of the variation in the vote as macroeconomic issues. As the party identification model predicts, both party identification and attitudes towards party leaders
S. Blount / Electoral Studies 18 (1999) 505–517
513
account for much of the variation in the vote.6 The insignificant beta weights for social and political issues are also consistent with the findings of the summary analysis in Table 1. The three sets of issues which are highly significant are macroeconomic, fiscal and microeconomic, and health issues. However, macroeconomic issues appear to have had an insignificant influence on the vote in both the 1996 and 1998 elections. The beta weights suggest that microeconomic and fiscal issues were at least twice as influential as any other set of issues in the 1993 and 1996 elections, and in 1998, fiscal and microeconomic issues were a dominant influence upon the vote. These results are consistent with the suggestion that individuals do take economic conditions into account when voting for the Australian Parliament, but that the cyclical issues of inflation, interest rates and unemployment alone are inadequate measures of economic voting.
4. Conclusion Aggregate models of economic voting test the theory that voters make political decisions based on rational assessments of the economy. At the national level, these models have conventionally tested for a significant relationship between one or more of the macroeconomic variables of unemployment, interest rates, inflation, and the vote. However, an individual level analysis of the last four federal elections in Australia suggests that individuals’ attitudes towards these cyclical economic issues have had increasingly less influence on the way they voted than their attitudes towards issues of permanent economic change, such as taxation, industrial relations and privatisation. These findings reflect the particular circumstances of recent Australian elections, which have taken place within the context of significant attempts to introduce changes in the areas of taxation, conditions of employment and publicly owned utilities. But these findings also have implications beyond Australia, since they suggest that the theory of economic voting cannot be adequately tested by models which operationalise the idea of the economy solely as the sum of cyclical change economic variables. I conclude that aggregate models which include measures of fiscal and microeconomic change, as well as macroeconomic change, would offer a more powerful test of the theory of economic voting because they operationalise the idea of the economy in a more complete and meaningful way.
6 The insignificant result for the Prime Minister in the 1998 analysis may be accounted for by the the high profile of the leader of the emerging ‘One Nation’ party who tended to appeal to disaffected Coalition supporters.
514
S. Blount / Electoral Studies 18 (1999) 505–517
Appendix A Variable questions and values All questions are as they appear in the 1990 Australian Election Study. Substantial variations in subsequent surveys appear in brackets ⬍ > .
Issues Here is a list of important issues that were discussed during the election campaign. Whose policies—the Labor Party’s or the Liberal–National Coalition’s would you say came closest to your own views on each of these issues? For each issue, respondents were asked to reply: 1 ⫽ Labour; 2 ⫽ Liberal– National; 3 ⫽ No Difference; 4 ⫽ Don’t Know. Still thinking about these same issues, when you were deciding about how to vote, how important was each of these issues to you personally? For each issue, respondents were asked to reply: 1 ⫽ Extremely Important; 2 ⫽ Quite Important; 3 ⫽ Not Very Important. The combined variables in Table 1 were scored: 0 ⫽ Labour, Extremely Important; 2 ⫽ Labour, Quite Important; 4 ⫽ Labour, Not Very Important; 5 ⫽ No Difference/Don’t Know; 6 ⫽ Coalition, Not Very Important; 8 ⫽ Coalition, Quite important; 1 ⫽ Coalition, Extremely Important. The issues given in each survey were: 1990: Unemployment, Inflation, Interest Rates, Education, Environment, Health, Taxation, Goverment Spending, Wages. 1993: Unemployment, Inflation, Interest Rates, Education, Environment, Health, Medicare, GST, Business Taxes, Tariffs, Employment Contracts, Enterprise Bargaining, Social Security, Childcare. 1996: Unemployment, Inflation, Interest Rates, Education, Environment, Health and Medicare, Taxation, Industrial Relations, Immigration, Privatisation, Commonwealth–state relations, Asia, Defence. 1998: Unemployment, Inflation, Interest Rates, Education, Environment, Health, Taxation, GST, Industrial Relations, Immigration, Privatisation, Telstra, Asia. To make the issues in Table 1 more directly comparable, three sets of 1993 issues were combined: Health ⫹ Medicare ⫽ Health, Employment Contracts ⫹ Enterprise Bargaining ⫽ Industrial Relations, Business Taxes ⫹ Tariffs ⫽ Taxes. Two issues in 1998 were also combined in this table: Privatisation ⫹ Telstra ⫽ Privatisation. The scales in Table 2 were created as follows: 1990: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.77), Taxation ⫹ Goverment Spending ⫹ Wages ⫽ Micro (␣ ⫽ 0.79), Education ⫹ Environment ⫽ Social (␣ ⫽ 0.63), Health ⫽ Health. 1993: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.83), GST ⫹ Business Taxes ⫹ Tariffs ⫹ Employment Contracts ⫹ Enterprise Bargaining ⫽
S. Blount / Electoral Studies 18 (1999) 505–517
515
Micro (␣ ⫽ 0.85), Education ⫹ Environment ⫹ Social Security ⫹ Childcare ⫽ Social (␣ ⫽ 0.81), Health ⫹ Medicare ⫽ Health (␣ ⫽ 0.90). 1996: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.80), Taxation ⫹ Industrial Relations ⫹ Privatisation ⫽ Micro (␣ ⫽ 0.77), Education ⫹ Environment ⫹ Immigration ⫽ Social (␣ ⫽ 0.71), Health and Medicare ⫽ Health, Commonwealth–state relations ⫹ Asia ⫹ Defence ⫽ Politic (␣ ⫽ 0.74). 1998: Unemployment ⫹ Inflation ⫹ Interest Rates ⫽ Macro (␣ ⫽ 0.79), Taxation ⫹ GST ⫹ Industrial Relations ⫹ Privatisation ⫹ Telstra ⫽ Micro (␣ ⫽ 0.85), Education ⫹ Environment ⫹ Immigration ⫽ Social (␣ ⫽ 0.76), Health ⫽ Health, Asia ⫽ Politic. Vote for the House of Representatives and Senate Which party did you vote for first in the House of Representatives? For each question, respondents were asked to reply: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Other/Independents; 6 ⫽ Green Parties/NDP; 7 ⫽ Call to Australia; 8 ⫽ Grey Power; 9 ⫽ Informal/Didn’t Vote. ⬍ 1993: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Other. > ⬍ 1996: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ Other; 9 ⫽ Informal/Didn’t Vote. > ⬍ 1998: 1 ⫽ Liberal Party; 2 ⫽ Labour Party; 3 ⫽ National Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ One Nation; 7 ⫽ Other Party; 9 ⫽ Informal/Didn’t Vote. > In order to model the vote as variation along a left–right axis for the 1990, 1993 and 1996 elections, the scoring was: Labour Party ⫽ 0, Coalition ⫽ 1, Other ⫽ 5. In 1998, the scoring was: Labour Party ⫽ 0, Coalition and One Nation ⫽ 1, Other ⫽ 5. Scoring the One Nation Party the same as the Coalition parties reflects the fact that One Nation drew a substantial portion of its votes from disaffected Coalition supporters.
Party identification Generally speaking, do you usually think of yourself as Liberal, Labour, National or what? Respondents were asked to reply: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Green; 6 ⫽ Independent; 8 ⫽ Coalition; 9 ⫽ No political Id. ⬍ 1993: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Other party (please specify); 6 ⫽ No party. > ⬍ 1996: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ Other Party; 7 ⫽ No Party. >
516
S. Blount / Electoral Studies 18 (1999) 505–517
⬍ 1998: 1 ⫽ Liberal; 2 ⫽ Labour; 3 ⫽ National (Country) Party; 4 ⫽ Australian Democrats; 5 ⫽ Greens; 6 ⫽ One Nation; 7 ⫽ Other Party; 8 ⫽ No Party. > These categories were recoded to agree with the dependent variable: 0 ⫽ Labour, 1 ⫽ Coalition and One Nation; 5 ⫽ Other. Attitudes to leaders ...using the 0 to 10 scale please show how favourable or unfavourable you feel about the party leaders. Respondents were asked to circle a number on a scale for each candidate. The scale ranged from 0 to 10, where 0 represented ‘very unfavourable’, 5 represented neutral, and 10 represented ‘very favourable’. The responses create a Likert scale for each leader. A positively signed coefficient indicates a response favouring the Coalition leader. A negatively signed coefficient indicates a response favouring the Labour leader.
References Aitkin, D., 1982. Stability and Change in Australian Politics, 2nd ed. Australian National University Press, Canberra. Atkeson, L.R., Partin, R.W., 1995. Economic and referendum voting: a comparison of gubertanorial and senatorial elections. American Political Science Review 89 (1), 99–107. Bean, C., 1994. Issues in the 1993 Election. Australian Journal of Political Science 29, 134–157. Bean, C., McAllister, I., 1997. Short term influences on voting behaviour in the 1996 election. In: Bean, C., Bennett, S., Simms, M., Wharhurst, J. (Eds.), The Politics of Retribution: The 1996 Australian Federal Election. Allen and Unwin, St Leonards, N.S.W., Australia. Bean, C., Gow, D., McAllister, I., 1998. Australian election study 1998 [computer file]. Social Sciences Data Archive, The Australian National University, Canberra. Blount, S., 1998. Postmaterialism and the vote for the Senate in Australia. Australian Journal of Political Science 33 (3), 441–449. Berry, F.S., Berry, W.O., 1992. Tax innovation in the States: capitalising on political opportunity. American Journal of Political Science 36, 715–742. Berry, F.S., Berry, O.B., 1994. The politics of tax increases in the States. American Journal of Political Science 38 (3), 855–859. Campbell, A., Converse, P.S., Miller, W.S., Stokes, D.E., 1960. The American Voter. Wiley, New York. Compston, H., 1990. The electoral impact of tax: a comparative study of Sweden and Australia. Unpublished PhD Thesis, Australian National University, Canberra. Douglas, R., 1975. Economy and polity in Australia: a quantification of commonsense. British Journal of Political Science 5, 341–361. Eismeier, T.J., 1983. Votes and taxes: the political economy of the American governorship. Polity 15, 368–379. Erikson, R.S., 1990. Economic conditions and the congressional vote: a review of the macrolevel evidence. American Journal of Political Science 34 (2), 373–399. Fiorina, M.P., 1978. Economic retrospective voting in American national elections: a micro analysis. American Journal of Political Science 22 (2), 426–443. Gow, D.J., 1990. Economic voting and postmaterialist values. In: Bean, C., McAllister, I., Warhurst, J. (Eds.), The Greening of Australian Politics. Longman Cheshire, Melbourne. Groenewegen, P., 1990. Public Finance in Australia: Theory and Practice, 3rd ed. Prentice Hall, Sydney.
S. Blount / Electoral Studies 18 (1999) 505–517
517
Jones, R., McAllister, I., Denemark, D., Gow, D., 1993. Australian election study [computer file]. Social Science Data Archives, The Australian National University, Canberra. Jones, R., McAllister, I., Gow, D., 1996. Australian election study [computer file]. Social Science Data Archives, The Australian National University, Canberra. Kelly, P., 1992. The End of Certainty: The Story of the 1980s. Allen and Unwin, Sydney. Kiewiet, D.R., Rivers, D., 1985. A retrospective on retrospective voting. In: Eulau, H., Lewis Beck, M. (Eds.), Economic Conditions and Electoral Outcomes: The United States and Western Europe. Agathon Press, NY. Kinder, D.R., Kiewiet, R.D., 1979. Economic discontent and political behaviour: the role of personal grievances and collective economic judgements in congressional voting. American Journal of Political Science 23 (3), 495–527. Kone, S.L., Winters, R.F., 1993. Taxes and voting: electoral retribution in the American states. The Journal of Politics 55 (1), 22–40. Kramer, G.H., 1983. The ecological fallacy revisited: aggregate- versus individual-level findings on economics and elections, and sociotropic voting. American Political Science Review 77 (1), 92–111. Leithner, C., 1993. Economic conditions and the vote: a contingent rather than categorical influence. British Journal of Political Science 23, 339–372. Lewis-Beck, M.S., 1988. Economics and Elections: The Major Western Democracies. University of Michigan Press, Ann Arbor. Marcus, G.B., 1988. The impact of personal and national economic conditions on the presidential vote: a pooled cross sectional analysis. American Journal of Political Science 32 (1), 137–154. Marcus, G.B., 1992. The impact of personal and national economic conditions on the presidential vote, 1956–1988. American Journal of Political Science 36 (3), 829–834. McAllister, I., 1992. Political Behaviour: Citizens, Parties and Elites in Australia. Longman Cheshire, Melbourne. McAllister, I., Bean, C., 1990. Explaining Labor’s victory. In: Bean, C., McAllister, I., Warhurst, J. (Eds.), The Greening of Australian Politics. Longman Cheshire, Melbourne. McAllister, I., Jones, R., Papadakis, E., Gow, D., 1990. Australian election study [computer file]. Social Science Data Archives, The Australian National University, Canberra. Mughan, A., 1987. The ‘hip pocket nerve’ and electoral volatility in Australia and Great Britain. Politics 22 (2), 66–75. Nannestad, P., Paldam, M., 1995. It’s the Goverment’s fault! A cross-section study of economic voting in Denmark, 1990/93. European Journal of Political Research 28, 33–62. Niemi, R.G., Stanley, H.W., Vogel, R.J., 1995. State economies and state taxes: do voters hold governors accountable? American Journal of Political Science 39 (4), 936–957. Niskanen, W.A., 1979. Economic and fiscal effects on the popular vote for the President. In: Rae, D.W., Eismeier, T.J. (Eds.), Public Policy and Public Choice. Sage Publications, Beverly Hills. Pacek, A.C., Radcliff, B., 1995. Economic voting and the welfare state. The Journal of Politics 57 (1), 44–61. Papadakis, E., 1990. Minor parties, the environment and the new politics. In: Bean, C., McAllister, I., Warhurst, J. (Eds.), The Greening of Australian Politics. Longman Cheshire, Melbourne. Pomper, G.M., Lederman, S.S., 1980. Elections in America: Control and Influence in Democratic Politics. Longman, New York. Suzuki, M., Chappell, H.W., 1996. The rationality of economic voting revisited. Journal of Politics 58 (1), 224–236. Turret, S., 1971. The vulnerability of American governors 1900–1969. Midwest Journal of Political Science 15, 108–132.