Empirica 31: 27–42, 2004. © 2004 Kluwer Academic Publishers. Printed in the Netherlands.
27
Exploring Voter Preferences in Cultural Policy: A Case Study for Austria MICHAEL GETZNER Department of Economics, University of Klagenfurt, Universitaetsstrasse 65-67, A-9020 Klagenfurt, Austria E-mail:
[email protected]
Abstract. The aim of the paper is to explore the demographic, social, economic and political determinants of voting behavior in a recently held referendum on the construction of a new theater (“Musiktheater”) in the city of Linz (Upper Austria) in 2000. It was the first referendum on cultural policy of its kind in Austria, and it led to a rejection of the proposal by a majority of the voters (59.70% of “No” votes). Exploring the determinants of approval or disapproval of the proposition by using for group data in communities is thus an interesting question from an economic as well as a political point of view. We find econometric evidence for the influence of the size of the population in the community, the distance of the community from the city of Linz, income-related variables, variables denoting the economic structure of the community, and political variables such as the ratio of the communities’ debt to revenues and voters’ preferences in the last national election. Key words: Public referendum, cultural policy, theaters, performing arts, models of voting behavior. JEL codes: Z1, H4, H0
I. Introduction Cultural policy decisions are often at the center of public debate. In Austria, decisions made by the directorate of the Viennese “Burgtheater” or the Salzburg Festival are hotly discussed issues, even in newspapers and on television (e.g., regarding the novelty of the program or the engagement of certain actors or opera singers). However, many of the decisions affecting cultural policy are based on the recommendations of committees of experts and/or performers. The public debate can be considered mere “background noise” and only in very few cases are decisions dropped or significantly altered due to public debate (such as the design of the new “Museumsquartier” (museum district) in Vienna which was radically changed following public criticism). The project of a new “Linzer Musiktheater” in Linz (Upper Austria) has also been accompanied by a long-running debate on the necessity of building a new theater and the costs involved. In recent years, there had always been a political consensus among the parties represented in the provincial government that the old
28
MICHAEL GETZNER
“Musiktheater” (theater and opera house for the province of Upper Austria) had to be rebuilt due to poor acoustics and a lack of appropriate space for up-to-date productions. The first debates and planning steps directed towards modernizing or rebuilding the theater began in 1983. In 1991, a concrete planning process was initiated, including commissioning a comprehensive feasibility study for the building of a new theater. In 1996, a decision was made by the provincial government to build a new theater on the banks of the River Danube. In 1997, an international architectural competition was announced and in 1999, the first contracts with the architects and building companies were signed. The new building would have cost EUR 87.2 million, and would have constituted a new landmark on the banks of the River Danube. In 1999, the right wing populistic Austrian Freedom Party (FPÖ) left the political consensus of the provincial government (consisting additionally of the conservative Austrian People’s Party and the Social Democrats) and called for a referendum to be held on the project of a new “Musiktheater”. The FPÖ argued that the theater would be too expensive and that the funds devoted to the new building could be used more efficiently; however, the main argument was, according to political commentators, a short-sighted maximization of votes as no constructive arguments were brought into the debate by the FPÖ – a conclusion which can also be derived from the simple slogan “The man in the street against the big opera” (“Kleiner Mann gegen große Oper"). The other two governing parties, the conservative Austrian People’s Party (ÖVP) which had the majority in the provincial government and the regional parliament, and the Social Democrats (SPÖ), argued strongly in favor of a new “Musiktheater” being constructed. The Green Party, although not in the provincial government and in opposition in the regional parliament, also favored the realization of the project. As renovating the old building would be too costly, a new landmark building was thought to provide new possibilities for opera and theater productions. The ensuing political debate led to a political novelty in November 2000: a (non-binding) referendum with the single question, “Should a new theater be built in Linz?”, was held. This was the first referendum of its kind in Upper Austria, and the first one on cultural policy in Austria. At provincial level (in contrast to the national level), referenda are non-binding and their main objective is to “explore the electorate’s will”. Thus, the provincial government does not have to follow the electorate’s decision in a (non-binding) referendum. However, the decisions of voters in such a referendum certainly put strong pressure on decision makers. Unfortunately – from the viewpoint of opera and theater lovers – the referendum resulted in a surprisingly high share of “No” votes (59.70%) with a turnout of 50.05% of the electorate (source for all figures: Oberösterreichische Landesregierung, 2000), thus making it unfeasible for the provincial government to pursue its path of building a new music hall of the proposed kind.1 In the meantime, more cultural, architectural and financial concepts have been worked out, and a new “Musiktheater” will eventually be built on a different site and with an adapted
EXPLORING VOTER PREFERENCES IN CULTURAL POLICY
29
architectural concept (it is not yet clear whether the new project will, in fact, reduce costs). However, it is interesting from an economic as well as from a political viewpoint to explore the determinants of voters’ behavior as, on the one hand, such an analysis can test for the explanatory power of economic and public choice models and on the other hand, such a venture provides an understanding of the political processes in cultural policy. The paper thus tries to answer the question as to which geographic, social, economic and political circumstances influenced the outcome of the referendum. The paper does not concentrate on theoretical modeling but empirically explores voters’ behavior regarding cultural policies. In the current context, three basic questions could be covered: (i) participation in elections and referenda in general, (ii) participation in the specific referendum on the theater in Linz, and (iii) the decision on how to vote (i.e., approval or rejection of the proposition). As data on question (i) are not readily available (this would require a different approach than the one pursued in the current paper), the paper tries to answer questions (ii) and (iii). The determinants of participation in the referendum might differ from the decision on how to vote. Thus, the methodology applied in the paper rests on a two-step procedure: First, a participation function is estimated. Second, the decision on how to vote is estimated separately. The structure of the paper is as follows: Section II gives a brief overview of the literature and discusses the (international) empirical evidence on the politicoeconomic determinants of voters’ preferences in referenda on cultural issues. Section III provides a brief sketch of the theoretical modeling background regarding the demand for cultural amenities such as the “Linzer Musiktheater”. As theoretical and empirical evidence on the influence of potential explanatory variables is ambiguous, there are a number of variables for which no a priori hypotheses regarding their influence on the participation and decision of voters can be built. Section IV builds on the empirical information from the referendum and discusses the importance of a number of politico-economic determinants of voting behavior. Section V summarizes the results and presents conclusions for cultural policy.
II. Voter Preferences and Cultural Policy Cultural policies are often said to involve non-transparent decision-making processes influenced by rent-seeking and interest-group interventions (Grampp, 1989; Schulze and Rose, 1998; Krebs and Pommerehne, 1995; Khakee, 1982; Lingle, 1992). In a few cases, cultural policies are subject to public referenda which elicit voters’ preferences directly. While public referenda on regional and local issues, including public expenditures on cultural amenities, take place quite often in Switzerland, they are rare in other European countries like, for instance, Germany or Austria. Starting with Pommerehne (1982), a few papers have dealt with the determinants of voting behavior in connection with cultural policy. Pommerehne (1982) explored the determinants of the demand for additional public subsidies
30
MICHAEL GETZNER
for two theaters in Basel in order to enhance the quality of their productions. Citizens were asked in two referenda whether they agreed with the proposition of increasing subsidies for the theaters. The main determinants regarding the approval rate were income, the increased tax rate faced by every voter, travel costs when going to a theater performance, and education. Regarding conclusions for policy advice, Pommerehne was cautious, basically because not all citizens took part in the referenda, thus the results of referenda might not have been “representative”. Following Pommerehne’s approach, Schulze and Ursprung (2000) discussed the determinants of voting behavior more recently using data from a Swiss referendum on public support for the Zurich Opera House. Their econometric estimations exhibit a rather similar pattern of socio-economic determinants of voting behavior (approval rate to public support). Determinants of support for cultural subsidies include, among others, income, education, distance from the opera house, and the share of voters who were self-employed (as a measure of the opportunity costs of attendance). Other papers on the determinants of voting behavior in cultural policy include Schneider and Pommerehne (1983) on the purchase of an important painting by the City of Basel and Withers (1979) on public support for the arts in Australia. The conclusions of these papers are basically comparable with the ones cited above, i.e., that support for the arts increases ceteris paribus with higher income and education, and with the availability of the cultural good for personal enjoyment (e.g., the distance of the residential area from the museum). However, no paper has dealt with the determinants of Austrian cultural policies so far. We now turn to the determinants of voting behavior regarding cultural policies in Austria and begin by sketching a model of voting behavior in public referenda, including a description of the empirical approach applied in this paper. III. Modeling Participation in the Referendum and the Demand for a “Musiktheater” The demand for a new “Musiktheater” in Linz is estimated on the basis of a twostep procedure. First, participation in the public referendum on the construction of the theater is explored; second, voting behavior in the referendum is modeled and empirically estimated. The first step is to assume an individual i’s utility function which includes cultural goods as an argument in the function: Ui = f (qi , gi , ci ),
(1)
where Ui is individual i’s utility derived from the consumption of private goods qi , public goods gi and cultural goods ci which may be a theater or opera house (this model is, of course, only a partial model). Microeconomic theory suggests that for the case ∂Ui /∂ci > 0, an increase in the quality or quantity of cultural goods ci results in an increase in the overall utility Ui , thus individual i has a positive willingness-to-pay for the production of these goods and might therefore accept
EXPLORING VOTER PREFERENCES IN CULTURAL POLICY
31
the proposition to build a new theater.2 The construction of a theater is a cultural good which may increase utility (depending on individual i’s preferences). In a corresponding public referendum, individuals agree to or reject the construction of a theater based on their individual assessment of the marginal utility they gain (or lose). However, even in the case of a positive marginal utility derived from the consumption of additional units of the cultural good, individuals might vote against the proposition if the tax price they have to pay is above the value of the marginal utility derived from the cultural good. In order to model an individual’s behavior in a referendum, binary choice models are commonly used. As we cannot observe or measure the utility deriving from cultural goods according to Equation (1), such models relate the probability of an individual, first, participating in the referendum, and second, agreeing to the proposition regarding the theater (a “Yes” vote in the referendum) to a set of socioeconomic variables. The reason for using a probability model is that we only can estimate partial models, i.e., we can only observe certain attributes of the individual. A number of other possible variables influencing voting behavior remain unknown. The probability of a “Yes” vote in the referendum (Pi ) is assumed to be influenced by a number of socio-economic variables such as income, age, education etc.: Pi = f (Income, Age, Education . . . .)
(2)
Alternatively, Pi can also denote the probability of individual i participating in the referendum. In order to model individual i’s voting behavior empirically, a logit model which assumes that the probability of a “Yes” vote can be described by a cumulative logistic probability function of the observable socio-economic variables (vector Xi ) is taken as a starting point: Pi =
1 1+
e−(α+βXi )
.
As Pi /(1 − Pi ) = e−(α+βXi ) , taking the natural logarithm results in Pi = α + βXi , ln 1 − Pi
(3)
(4)
with α and β denoting the coefficients to be estimated. The problem of empirically estimating Equation (4) is the lack of data on individual voting behavior (Weck-Hannemann, 1990; Pindyck and Rubinfeld, 1991). However, for each of the 445 Upper Austrian communities in which the referendum on the new theater took place, group data are available. For instance, there are data on the share of “Yes” votes in each community and average data on a number of socio-economic variables describing the electorate of each community. By assuming identical individuals with respect to Xi , the probability of a “Yes” vote being cast by the
32
MICHAEL GETZNER
representative voter can be approximated by the fraction of voters in each community that actually voted “Yes”, i.e., Pi ≈ P˜i = zi /wi . zi denotes the number of “Yes” votes in each community, and wi denotes the number of voters participating in the referendum. (Alternatively, for estimating participation in the referendum, Pi ≈ P˜i = xi /yi , with xi denoting the number of voters who participated in the referendum and yi the electorate.) Taking the approximation for Pi discussed so far, we replace Pi by P˜i to derive Equation (5) (explaining “Yes” votes; participation rates are estimated analogously): log
zi /wi 1 − zi /wi
= α˜ + β˜ X˜ i + u,
(5)
where α˜ and β˜ are the coefficients to be estimated based on group data for each of the communities where the referendum took place and u denotes the error term. Equation (5) represents the basic empirical model describing voting behavior (and, alternatively, participation in the referendum). As the explanation of voting behavior relies on aggregate (group) data, the logit model can be estimated using ordinary least squares. The estimated parameters α˜ and β˜ are consistent when the sample in each group (population and number of votes) becomes arbitrarily large. As the single groups (communities) vary considerably in size, the error term u in Equation (4) is heteroscedastic, and therefore weighted least squares are used, each observation being multiplied by the appropriate weights qi , given by Equation (6): qi =
1 . (wi (zi · (wi − zi )))1/2
(6)
As far as the operationalization of vector X˜ i in Equation (5) is concerned, Table I shows the two dependent variables and four groups of explanatory variables hypothesized to explain voters’ behavior and participation in the referendum. While for some of the variables, a priori hypotheses can be formed, there are a number of variables whose influence on the dependent variable is not defined from a theoretical viewpoint. It has to be stressed again that the current paper concentrates on exploring, first, the participation of voters in the referendum on the theater, and second, on the support elicited by the referendum (the “Yes” votes). Thus, for a number of variables, possible explanations for their explanatory power and direction of influence are discussed which might not be convincing in all cases as there might not be existing theoretical or empirical arguments for all variables. The two dependent variables are first, PARTICIPATION, denoting the share of the electorate participating in the referendum and second, YES, denoting the share of voters participating in the referendum who approved of the proposition, i.e., agreed to the proposed construction of a new theater in Linz. Both dependent variables were transformed in order to conform to the logit model described above.
EXPLORING VOTER PREFERENCES IN CULTURAL POLICY
33
Table I. Description of variables Dependent variables PARTICIPATION ln (share of participating electorate/(1 – share of participating electorate)) YES ln (share of YES votes/(1 – share of YES votes)) Explanatory variables Geographical variables (Gi ) POP Population of the community (in 1,000, natural logarithm) DISTANCE Distance of the community from the provincial capital Linz (in km; classes in 20 km) Social variables (Si ) EDUC Share of inhabitants holding a university or college degree (in %) YOUNGAGE Share of inhabitants aged 0 to 19 (in %) Economic variables (Ei ) JOB Share of the population currently employed (in %) TOURISM Share of the labor force working in the tourism sector (in %) INDUSTRY Share of the labor force working in the industrial sector (in %) DEBT Community debt-to-revenue ratio (in %) Political variables (Li ) LIBERAL Share of the electorate voting for the Greens or Liberals (LIF – Liberales Forum) in the last nationwide election (1999; in %, natural logarithm)
The first group of explanatory variables comprises the geography and demography of the community (geographical variables Gi ). The variable POP denotes the size of the population.3 It might be hypothesized that in small communities, participation in referenda is greater due to higher “social control”: in a village, where each community member is known, it is more likely that citizens fulfill their “democratic duty” to participate. Furthermore, approval rates might be higher in smaller communities (for a discussion see Section 4.2 below).4 In general, the size of the population can be considered to have some influence on the dependent variable as in larger populations, free-riding is more likely, and the individual tax prices are smaller (low-cost situations). Distance from the provincial capital of Linz is operationalized by the variable DISTANCE. Voters in communities that are further away from Linz might – ceteris paribus – be less interested in participating in and voting in favor of the proposition compared to communities near Linz because they face higher travel costs when going to the theater (thus their opportunity costs of attendance are higher which lowers their demand for the theater; see, e.g., Schulze and Ursprung (2000) for empirical indications of the importance of distance from a theater in explaining
34
MICHAEL GETZNER
voting behavior). Furthermore, the hypothesis behind the influence of DISTANCE on the decision to participate is that in referenda such as the ones discussed in this paper, the incentive to go to the polls is greater for No-voters; if the interest in the music theater in communities more distant to Linz is lower (and the disapproval of the referendum is higher), distance might play a crucial influence on the estimation of a participation function. However, on the other hand, a larger distance to Linz lowers the interest in the music theater due to higher opportunity costs of attending, therefore voter turnout might also be lower. The second group of explanatory variables accounts for the social characteristics of the electorate (social variables Si ). The variable EDUC denotes the share of the population holding a college or university degree. In previous studies on cultural policy (Pommerehne, 1982; Schulze and Ursprung, 2000), it turned out that higher education levels correlate positively with support for the arts. Thus, higher education levels can be hypothesized to increase participation in and support for the proposition. The age structure of the population is described by the variable YOUNGAGE (denoting the share of residents aged 0–19). The share of young inhabitants aged 0 to 19 might increase participation in and support for the proposition based on the bequest value of cultural goods (parents caring for their children). Furthermore, young people might be more open-minded regarding modern architecture5 and thus be more interested in participating in and approving the referendum. The third group of variables (economic variables Ei ) is included in order to explore the relation between the dependent variables and the attributes of the local economy. The general underlying hypothesis is that wealthy communities might have a lower participation rate as the opportunity costs of participation in the referendum are higher. In contrast, approval rates might be higher as cultural goods, particularly theaters which may be considered elitist by a significant share of the population, are favored by those with higher incomes (Getzner, 2002). However, in the current study, there is no consistent income variable available for the communities in Upper Austria (no regional GDP or income-related variable such as income and/or labor taxes is readily available at community level).6 Thus, the economic situation of a community is described by three variables commonly correlated to (average) income. The variable JOB denotes the share of the population currently employed and thus receiving income. Communities with an above-average unemployment rate may thus have higher participation and a lower approval rate in the referendum. The first hypothesis can be supported by arguing that the opportunity costs of participating in the referendum are higher for people currently in employment than for unemployed citizens. The second hypothesis can be derived from the positive income elasticity of demand for cultural goods (Netzer, 1992; Frey, 2000; Schulze and Rose, 1998). The variable TOURISM denotes the share of the population employed in the tourism sector. Citizens working in that sector are often seasonal workers and thus might be less interested in participating in the referendum and approving the
EXPLORING VOTER PREFERENCES IN CULTURAL POLICY
35
proposition because they have fewer opportunities to go to the theater (and also have below-average incomes). Furthermore, many employees in the tourism sector have to work during the week and often on weekends and might therefore not have the opportunity to take part in the referendum. Thus a lower participation rate can be expected in communities with a comparably high share of employees in the tourism sector. The variable INDUSTRY (share of workers in the industrial sector) also accounts for the classification of a community as having – ceteris paribus – a lower interest in the public production of and support for cultural goods as industrial workers have a below-average income. Their demand for cultural goods might – ceteris paribus – be lower. Consequently, a positive income elasticity of demand for cultural goods corresponds to a negative sign of the coefficient for both variables. However, for the variable TOURISM, a different expectation might also be formulated: workers in the tourism sector might prefer a new theater as it increases the attractiveness of Upper Austria as a tourist destination, therefore contributes to the stability of their jobs. On the other hand, if they perceive the new theater as being a competitive advantage of the city of Linz over their own community, they might disagree with the proposition. The last group of variables (political variables Li ) describes the political circumstances in each community. The possibly most important variable, preferences for political parties, can be described by the share of the electorate which voted for a particular party in the national elections that took place only one year prior to the referendum (Schulze and Rose, 1998). We take the variable LIBERAL as the share of the electorate which voted for the Greens or Liberal Party which are both widely known for their liberal and positive attitudes towards cultural goods. Finally, the variable DEBT accounts for the community’s debt in relation to the Community’s total revenue. When citizens already perceive their tax burden as being above-average, and fear that they might have to pay higher taxes due to high community debt, their participation in and lack of support for the referendum might be comparably high. (However, this might only be a subjective perception as the tax burden resulting from constructing the new theater was only EUR 63.00 per person as a one-off payment.)
IV. Empirical Results: Estimating the Demand for a New “Musiktheater” 1. EXPLAINING PARTICIPATION IN THE REFERENDUM ON CULTURAL POLICIES
Having clarified the theoretical background and the hypothesized direction of influence of the explanatory variables, we now turn to explaining voters’ decisions empirically. As mentioned before, we apply a two-stage procedure: first, a function exploring participation in the referendum is estimated (voters going to the polls irrespective of whether they accept or reject the proposition; variable PARTICIPATION). Second, the determinants of the outcome of the referendum (variable YES)
36
MICHAEL GETZNER
Table II. Estimating voters’ participation in the referendum
Dependent variable: Variable
Constant POP DISTANCE EDUC YOUNGAGE JOB TOURISM DEBT
Participation equation PARTICIPATION Est. 1 Coefficient (t-statistic)
Yes-vote equation YES Est. 2 Coefficient (t-statistic)
1.8125 (6.1871∗∗∗ ) −0.2092 (−14.6025∗∗∗ ) −0.0078 (−15.4908∗∗∗ ) 0.0099 (3.0836∗∗∗ ) 0.0235 (5.6573∗∗∗ ) −0.0065 (−1.8671∗ ) −0.0236 (−5.1071∗∗∗ ) 0.0003 (1.949∗ )
0.3830 (4.4161∗∗∗ ) −0.0972 (−7.8143∗∗∗ ) −0.003 (−6.8248∗∗∗ ) 0.0217 (7.3344∗∗∗ )
INDUSTRY LIBERAL Adj. R2 S.E. of regr. Log likelihood F-statistic N
0.6223 0.2476 −6.2131 105.2713∗∗∗ 445
−0.0052 (−3.3921∗∗∗ ) 0.0412 (3.2701∗∗∗ ) 0.5762 0.2175 50.5191 121.7248∗∗∗ 445
WLS estimation; ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.
are investigated.7 Table II shows the empirical results of estimating a “participation function” (dependent variable PARTICIPATION). The first estimation of Table II (Est. 1) includes, first, the geographical variables POP and DISTANCE. The estimation shows that the variable POP exerts a significantly negative influence on the dependent variable: other things being equal, participation rates are higher in communities with a smaller population. Thus, the hypothesis of “social control” presented above might be quite realistic. The variable DISTANCE also has the expected (significant) influence; the participation
EXPLORING VOTER PREFERENCES IN CULTURAL POLICY
37
in the referendum is increasing the closer a community is to Linz, the provincial capital of Upper Austria. Second, accounting for social variables such as the age structure of the population increases the explanatory power of the model. The variable EDUC (denoting the share of the population holding a college or university degree) exhibits a significantly positive influence on the dependent variable, indicating that participation in the referendum increases with higher education. 8 The coefficient for the variable YOUNGAGE is also significantly positive. The higher the share of young people aged up to 19 in the community, the higher the participation rate in that community. Third, the economic variables Ei also contribute to an improved understanding of voter turn-out. The coefficient for the variable JOB is significantly negative, indicating that voters currently employed might, in fact, face higher opportunity costs of participating in the referendum and thus participate to a lower extent. However, a joint estimation including the variables JOB and TOURISM further increases the statistical quality of the model. The variable TOURISM exhibits a significantly negative influence on the dependent variable. The estimation indicates that in communities with a higher share of the labor force employed in the tourism sector, participation in the referendum is – ceteris paribus – lower. However, variables hypothesized to be significantly influential do not prove to contribute to an explanation of the participation rate. The sign of the coefficients of other economic variables, such as the variable INDUSTRY, is negative as expected but they fall far from being significant, and are not reported in detail here. Fourth, while the variable LIBERAL does not exhibit a significant influence on participation in the referendum, thus indicating that political preferences do not seem to influence voters’ decisions to participate, the variable DEBT contributes significantly to explaining participation rates which increase with the community’s higher debt-to-revenue ratio. Voters are apparently sensitive to the fiscal problems of their communities and are therefore more interested in influencing the use of scarce public funds. Est. 1 exhibits an adj. R2 of slightly over 0.62.9 To sum up, most of the hypotheses discussed in Section 3 seem to be supported empirically. Communities with a larger population (lower “social pressure” to participate) and further away from the city of Linz (lower interest in the music theater) exhibit a lower voter turnout while a higher share of young people in the community increases participation in the referendum (due to the bequest value of cultural goods, and their potentially higher support for modern architecture). Communities with higher labor force participation rates, higher shares of workers in the tourism sector (higher opportunity costs of participating) and lower debtto-revenue ratios (lower “fiscal sensitivity” in low-debt communities) have lower participation rates.
38
MICHAEL GETZNER
2. EXPLAINING SUPPORT FOR THE PROPOSAL TO BUILD A NEW “ MUSIKTHEATER ” Based on the empirical exploration of the determinants of participation in the referendum on the “Musiktheater”, we now turn to the second step of our analysis, which is explaining the outcome of the referendum, i.e., the approval rate (dependent variable YES). First, the estimation again relates to the demography and location of the community. The patterns already observed when explaining the participation rate can also be found in the model for the variable YES. Est. 2 in Table II shows that the approval rate significantly depends on the size of the community and its location. The variable POP exhibits a significantly negative influence, indicating that the approval rate is higher in small communities than in larger ones. This might be considered a surprise as day-to-day experience would suggest that voters in larger communities with more opportunities to enjoy cultural events might be more “used” to consuming cultural goods. The idea behind this reasoning is that in order to derive increased utility from cultural activities, the individual has to build up some sort of “consumption capital”: enjoying art involves education in order to fully understand and benefit from cultural activities. However, the empirical evidence presented in the estimations in Table II indicates a counter-intuitive influence of the size of the community, although this result can also be interpreted in the same way as the discussion of the determinants of participation in the referendum. If there is indeed “social pressure” to participate in the referendum, the share of “Yes” votes may be larger in smaller communities because voters disagreeing with the proposition may generally face a higher incentive to participate and vote “No” (hence, the share of “No” votes is higher in larger communities where “social pressure” to participate is lower). Thus, increasing participation in general might also lead to a relatively higher participation of voters voting “Yes”. The variable DISTANCE again exhibits a significantly negative influence on support for the proposition. The approval rate is lower in communities that are further from Linz compared to communities closer to the city. Such statistical correlation can be reasoned by the aforementioned influence of travel costs and the demand for attendance at the theater. Voters in nearby communities face a greater incentive to go to the theater due to their lower opportunity costs. Second, the variable EDUC exhibits a significantly positive influence on the approval rate (variable YES) indicating that communities with a higher than average education amongst voters – ceteris paribus – exhibit greater support for the construction of a new theater. The share of young people in the community (YOUNGAGE) again significantly influences approval rates which might be reasoned by the larger share of young inhabitants and the perceived bequest values of cultural goods. Third, while the coefficient for the variable JOB has the expected sign indicating a higher approval rate with lower unemployment (cultural goods as “normal
EXPLORING VOTER PREFERENCES IN CULTURAL POLICY
39
goods”), it is not significant at a satisfactory level and thus left out from the estimation. A closer look at the economic structure of the labor force in individual communities indicates a significant influence for the variable INDUSTRY. As expected, this variable has a significantly negative coefficient and contributes to an improvement of the explanatory power. (The variable TOURISM does not add as much explanatory power to the model and is thus left out.) Finally, we again account for a model including political variables. While the variable DEBT does not play a significant role in explaining the dependent variable’s variance, the political preferences of the electorate do. The variable LIBERAL, denoting the share of voters who supported the Greens or Liberal Party in the last national election exhibits a significantly positive sign of the coefficient as theoretically expected. As both parties pursue liberal and supportive policies towards the arts, voters from both parties also supported the construction of a new theater. The adj. R2 of the model amounts to 0.58. However, the coefficient is not stable and the result is thus only a broad indication as the variable YOUNGAGE has to be left out due to possible multi-collinearity.
V. Discussion, Summary and Conclusions Studying actual voting behavior is very attractive to economists from the viewpoint of eliciting preferences for public goods such as improved environmental conditions or cultural goods like the theater which was subject to a referendum in Upper Austria. However, there are a number of limits to the analysis, suggesting a cautious interpretation of the empirical results. There are theoretical and empirical limits to eliciting “true” preferences for public goods by means of a referendum such as strategic behavior which might, however, pose no severe problem in the current context.10 The probably most significant empirical limit of the current analysis is the lack of data on individual voting behavior. Only group data are available, thus the empirical models estimated are able to account for only around 60% of the variance in voters’ behavior. Compared to the explanatory power of different approaches such as contingent valuation surveys, however, the explanatory power of the estimations presented in this paper is much higher, while a number of recent papers on explanations of voting behavior particularly regarding cultural policy result in comparatively higher statistical quality. This shortcoming in the estimations presented here might also be reasoned by the lack of a consistent variable more closely correlating to the voters’ income. As only data on the employment of the labor force are available, conclusions are weak regarding the influence of average income figures or regional GDP per capita. Nevertheless, those variables which are employed in the analysis and are weakly related to income indicate the theoretically expected influence on participation and approval rates. Beside the main limits of the analysis, the estimations broadly confirm the hypotheses regarding explanations of voting behavior. Higher education and employment increase support for a new theater while the increasing opportunity costs
40
MICHAEL GETZNER
of going to the theater (e.g., in communities further from the capital city of Linz) lead to lower approval rates. In addition to socio-economic determinants of voting behavior, we find empirical evidence for political determinants: already high debtto-revenue ratios in the communities apparently lead to higher participation rates, and the political preferences of the electorate also manifest themselves in voters’ preferences. The share of votes for the Green and Liberal Party increases the share of voters agreeing to the proposition to build a new theater. The general conclusion for cultural policy lies in the fact that the electorate’s preferences for cultural goods seem to be explainable by a few socio-economic variables and political attributes, at least to a certain extent. While the concern is often voiced that voters can be influenced by political marketing or short-sighted arguments, particularly in cultural policy, it seems on the one hand as though political debates and advertising only play a minor role in forming preferences in referenda on cultural policies. On the other hand, empirical results such as the current study may help to understand the political processes, particularly in cultural policies. Furthermore, the winners and losers of cultural policies become transparent as the costs and benefits are distributed unequally across the jurisdiction. Acknowledgements The author would like to thank G. Schulze, M. Schuster, R. Zanola, D.C. Mueller and an anonymous referee for critical and very helpful comments on a draft of the paper, and the participants of the European Public Choice Society’s Conference (April 2002, Belgirate, Italy) and the Association of Cultural Economics’ International Conference (June 2002, Rotterdam, The Netherlands) for discussion and comments. The author would also like to thank the Oberösterreichische Landesregierung (Upper Austrian Provincial Government) for supplying the underlying data on voting behavior in the referendum. All errors are, of course, the responsibility of the author. Notes 1. It is, however, interesting that there is political consensus that the old theater is far too small. Rebuilding and extending the existing theater might, however, be as costly as building a completely new one. 2. Of course, there are a number of pitfalls such as strategic voting behavior depending on the tax price each individual has to pay. However, as Seaman (1981) pointed out, the tax price each individual has to pay in the concrete situation is very small. Thus there might only be a modest incentive to behave strategically. However, due to space restrictions, this paper does not deal with the voting paradox or strategic behavior in detail. 3. Data for all variables except DISTANCE were taken from the database of the Austrian Central Statistical Office (Statistik Österreich, 2001). Detailed data and estimations (also those reported here) are available on request from the author. 4. The size of the population might also influence participation in and approval of the referendum as the “tax price” each individual has to pay for public funding of the arts influences individual
EXPLORING VOTER PREFERENCES IN CULTURAL POLICY
5. 6.
7.
8.
9.
10.
41
voting decisions (Seaman, 1981). However, in the current case study, tax rates do not differ between the communities and so tax prices do not depend on the size of the population. The planned music theater would have been a modern “land-mark” building. A crude income measure such as personal taxable income per capita (mean of the financial district in which the community is located) as well as per capita communal tax (based on the wage sum) did not exhibit a significant influence and were thus left out. If individual data were available, more sophisticated approaches such as Heckman’s sample selection procedures might be useful. However, we can only account for group data in the current context. A model with a nonlinear (quadratic) function was also tested. The hypothesis behind such a functional form can be considered in the fact that voters with both a strong positive and a strong negative opinion of the new theater have an incentive – ceteris paribus – to exhibit a higher participation rate. As the participation rate depends on the education level, a quadratic specification seems reasonable. However, no such functional form proved to be significant in the econometric estimations. A number of other estimations were also carried out, including the development of the model in a number of separate estimations and variables not mentioned here (e.g., other political or public finance variables). The results reported here are reasonable and have the best statistical fit. Strategic behavior in eliciting preferences is an important concern in the theoretical discussion. However, strategic behavior in the current context requires a lot of thinking and computation regarding the whole incentive structure and the political processes which might not be feasible for average voters. Furthermore, voters are in a low-cost situation when participating in the referendum. The costs of gathering the large amounts of information might easily exceed the benefits of behaving strategically in the referendum.
References Frey, B.S. (1997) ‘The Evaluation of Cultural Heritage: Some Critical Issues’, in M. Hutter and I. Rizzo, eds., Economic Perspectives on Cultural Heritage. New York: St. Martin’s Press, pp. 31–49. Frey, B.S. (2000) Arts and Economics: Analysis & Cultural Policy. Berlin: Springer Verlag. Frey, B.S., and Pommerehne, W. (1993) ‘Economics, Democracy and Arts: Love’s Labor Lost?’, Discussion paper B 9303, December 1993, Department of Economics, Saarbrücken: University of Saarland. Getzner, M. (2002) ‘Determinants of Public Cultural Expenditures: An Exploratory Time Series Analysis for Austria’, Journal of Cultural Economics 26(4), 287–306. Grampp, W.D. (1989) ‘Rent-Seeking in Arts Policy’, Public Choice 60(1), 113–121. Khakee, A. (1988) ‘The Rationale for Urban Government Action for Arts Funding’, Journal of Cultural Economics 12(1), 1–18. Krebs, S., and Pommerehne, W. (1995). ‘Politico-Economic Interactions of German Public Performing Arts Institutions’, Journal of Cultural Economics 19(1), 17–32. Lingle, C. (1992) ‘Public Choice and Public Funding of the Arts’, in R Towse and A. Khakee, eds., Cultural Economics. Berlin: Springer-Verlag, pp. 21–30. Netzer, D. (1992) ‘Cultural Policy in an Era of Budgetary Stringency and Fiscal Decentralization: The US Experience’, in R. Towse and A. Khakee, eds., Cultural Economics. Berlin: Springer Verlag, pp. 237–245. Oberösterreichische Landesregierung (2000) Volksbefragung 2000 – Musiktheater Linz. Informationen der Abteilung Statistik des Amtes der Oberösterreichischen Landesregierung, Folge 41/2000, Linz.
42
MICHAEL GETZNER
Peacock, A. (1994) ‘The Design and Operation of Public Funding of the Arts: An Economist’s View’, in A. Peacock and I. Rizzo, eds., Cultural Economics and Cultural Policies. Dordrecht: Kluwer Academic Publishers, pp. 167–184. Pindyck, R.S., and Rubinfeld, D.L. (1991) Econometric Models & Economic Forecasts. New York: McGraw-Hill. Pommerehne, W. (1982) ‘Steuern, Staatsausgaben und Stimmbürgerverhalten’, Jahrbücher für Nationalökonomie und Statistik 197(5), 438–462. Schneider, F., and Pommerehne, W. (1983) ‘Private Demand for Public Subsidies to the Arts: A Study in Voting and Expenditure Theory’, in W.S. Hendon and J.L. Shanahan, eds., Economics of Cultural Decisions. Cambridge, MA: Abt Books, pp. 192–206. Schulze, G.G., and Rose, A. (1998) ‘Public Orchestra Funding in Germany – An Empirical Investigation’, Journal of Cultural Economics 22(3), 227–247. Schulze, G.G., and Ursprung, H.W. (2000) ‘La donna e mobile - or Is She? Voter Preferences and Public Support for the Performing Arts’, Public Choice 102(1), 131–149. Seaman, B.A. (1981) ‘Economic Theory and the Positive Economics of Arts Financing’, American Economic Review 71(2), 335–340. Statistik Österreich (2001) Data excerpt of the ISIS databank. Statistik Österreich, Vienna. Weck-Hannemann, H. (1990) ‘Protectionism in Direct Democracy’, Journal of Institutional and Theoretical Economics 146(3), 389–418. Withers, G.A. (1979) ‘Private Demand for Public Subsidies: An Econometric Study of Cultural Support in Australia’, Journal of Cultural Economics 3(1), 53–64.
