the electoral cost of tax policy in flemish municipalities · 2021. 4. 10. · evidence of an...

The electoral cost of tax policy in Flemish municipalities

Jan Vermeir1

& Bruno Heyndels

Abstract. Tax policy is one of the important policy domains of government and therefore an important issue for voters. The literature shows that taxation can have an electoral cost. However, several factors influence this electoral impact. The nature of the tax itself, tax policy in neighbouring jurisdictions and the political characteristics of the government (the ideology and the political fragmentation) can have an influence. We test if tax policy has an influence on Flemish municipal election results, using panel data for the period 1982-2000. We find that the local property tax rate has a significant effect on the vote. Moreover, the average property tax rate in the neighbouring municipalities has a significantly positive effect, yielding evidence on yardstick voting. Contrary to the clarity of responsibility hypothesis, more fragmented governments are more punished for higher property tax rates and more rewarded for higher property tax rates in the neighbouring municipalities. We find no strong evidence for an impact of ideology.

1 Department of Economics, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussel. e-mail: [email protected], [email protected].

2

Introduction The power to tax is a major characteristic of government. In a democratic setting the use of this power is constrained by the electoral process. Introducing new taxes or increasing existing ones are unpopular policy measures that can be expected to translate into electoral losses for the incumbent government. Over the past decades, several authors have estimated the electoral cost of taxation. An important finding from this literature is that this cost of raising taxes depends on a multitude of intermediary factors. Characteristics of the tax itself as well as of the political context in which the decision is taken affect the electorate�s responsiveness. In the present paper, we analyse the electoral cost of tax policy in Flemish municipalities. On the basis of the existing literature, we identify four factors that may affect the electoral cost. A given change in taxes may be perceived and evaluated differently by the electorate depending on the visibility of the change or on the presence of similar measures taken in other (neighbouring) jurisdictions. Further, the political characteristics of the incumbent government may play a role. Electoral costs of taxation may depend on the fragmentation and/or ideological position of the government. In section 1, we briefly review the existing literature on the electoral cost of taxation. This literature forms part of the (large) literature on Vote- and Popularity functions (VP-functions). We identify the major intermediary factors and discuss the results found in earlier work. Section 2 gives a brief discussion of the institutional context and the role of local taxation. Section 3 summarizes the results from our empirical analysis into the electoral costs of tax policy in Flemish municipalities. Using a panel of data for the period 1982-2000, we examine if tax policy has an impact on election results. Thereby we concentrate on the two major local taxes: the local income tax and the local property tax rate. Section 4 discusses some extensions and section 5 concludes. 1. The electoral cost of taxation: a review of the literature Following Downs (1957), rational individuals vote instrumentally. Voters vote for the party or politician from which they expect the highest utility gain. Their expectations on how parties will perform after the elections are based on party platforms and � for incumbent parties � also on past performance. A positive evaluation of the incumbent�s past performance will translate into electoral gains. This logic underlies the responsibility hypothesis that has been the starting point of the empirical literature on vote and popularity-functions. The central point of attention in that literature has been the electoral effect of macro-economic variables. High economic growth, low inflation and/or unemployment have been found to positively affect the

3

incumbents� election results or approval ratings (Mueller, 2003). In a way, the voter rewards the incumbent for making appropriate use of the policy instruments at its disposal. Still, these instruments � notably tax and expenditure policies � are not only instruments in macro-economic policy. They, first of all, allow responding to the voters� demand for public goods and redistribution. Further they determine how the cost of these goods is distributed over the population. As a consequence, taxes have political relevance over and above their relevance as (macro-economic) policy instruments. This has been explicitly recognised in the literature on political budget cycles where it is demonstrated how governments manipulate taxes (and expenditures) in order to increase their re-election odds. Such manipulation is observed in addition to the manipulation of taxes and expenditures for macro-economic reasons (as implicit in the political business cycle literature). Today, governments in industrialised countries raise considerable amounts of tax revenue. These taxes directly constrain the individual voter�s capacity to spend on private goods and services. As such, it is natural to expect increased taxation to lead to electoral losses for the incumbent. Of course, the additional revenues may be used to increase public expenditures (which may create electoral gains). It should be noted, however, that electoral gains from public expenditures and electoral losses from taxation are to a large extent independent. Taxes are not really a price for public goods. From the individual voter�s perspective, it is always possible to provide the same set of public goods while taxing other taxpayers. Especially given the high complexity of most tax systems (based on a wide variety of taxes) the incumbent�s choice of how to raise additional revenue is highly relevant for the individual taxpayer. Ferejohn (1986), Rogoff (1990), Banks and Sundaram (1993), Besley and Case (1995), Persson and Tabellini (2000) and Revelli (2002) view elections in a principal-agent framework. The electorate is seen as the principal, and the incumbent government is the agent. Voters use a retrospective voting rule. They either punish or reward the incumbent for past performance or they choose the most competent government, using the information about past performance of the incumbent. Early empirical research on the link between taxes and elections or incumbent popularity focuses on U.S. state data. Pomper (1968) and Turett (1971), in early papers on the electoral cost of taxation, do not find a consistent significant correlation between tax variables and the election results of U.S governors. Later studies do find an impact on gubernatorial elections. Eismeier (1979, 1983), Kone and Winters (1993), Niemi et al. (1995) and MacDonald and Sigelman (1999) look at the effects of specific tax policy decisions such as tax introductions or increases. They find a significant effect. An electoral impact is also found in Besley and Case (1997) and Lowry, Alt and Ferree (1998). Hansen (1999), however, finds no influence of the rate of change in tax revenue on the approval rating of governors. Peltzman (1992)

4

finds some evidence of an electoral cost of revenue growth in senate and governor elections. The same author also considers U.S. presidential elections but fails to find any significant influence from tax policy. Still, in earlier work on U.S. presidential elections, Niskanen (1975, 1979) shows that an increase in federal tax revenues has a significant negative effect on the vote for the American presidential candidate of the incumbent party. Outside the US an impact of tax variables on election results has been found in the UK (Pissarides, 1980; Gibson and Stewart, 1992; and Gibson, 1996; Revelli 2002), in Denmark (Paldam and Schneider, 1980), in Sweden (Hibbs and Madsen, 1981) and in Canada (Happy, 1992; and Landon and Ryan, 1997). We can identify four important factors that can affect the electoral cost of tax policy. A first finding in the literature is that some taxes are electorally more costly. The visibility of a tax is put forward as a major reason for a tax to have a higher political cost. Kone and Winters (1993) find that changes in the sales tax have a higher impact on the vote for the incumbent party in American gubernatorial elections than changes in the income tax. The political cost of a whole range of taxes is studied by Landon and Ryan (1997).2 They consider the impact of tax policy on both the vote percentage of the opposition party and the probability of electoral defeat by the incumbent party in Canadian provinces. They find that an increase in sales tax has a significantly positive influence on both variables. The opposite is true for an increase in revenues from licences, permits and other fees: an increase results in an electoral gain for the incumbent party. Increases in direct personal taxes and fuel taxes have a significantly positive effect on the probability of electoral defeat by the incumbent party, but not on the vote for the opposition party. So the nature of the tax will affect its electoral cost and more conspicuous taxes are expected to have a larger cost. A recent addition to the literature has been the concept of yardstick competition. If cost shocks are correlated with cost shocks in neighbouring jurisdictions, tax policy in those jurisdictions can yield information on the quality of the incumbent government and reduce asymmetric information between voters and the incumbent. (Besley and Case, 1997; Revelli, 2002). Accordingly, voters will use tax policy in neighbouring jurisdictions as a yardstick to overcome political agency problems, analogous to relative performance evaluation and tournaments in the literature on incentives in firms (Holmstrom, 1982; Nalebuff and Stiglitz, 1983). Case (1994) and Besley and Case (1995) look at the impact of income tax liability in American gubernatorial elections. They find that an increase in the liability for high-income earners significantly increases the probability of a governor not being

2 They consider direct taxes on persons; corporate taxes; gasoline tax; natural resource taxes; sales taxes; miscellaneous indirect taxes; licenses, permits and other fees; and the provincial property tax.

5

re-elected.3 Importantly, re-election chances are also affected by changes in tax liabilities in neighbouring states. In a European context, Revelli (2002) who studies changes in property tax rates in English districts finds no significant effect of neighbour�s tax rate on the vote. However he finds evidence of an electoral impact of rate changes in the district itself. A third finding is that the electoral cost of tax policy can depend on the ideology of the incumbent party. Lowry, Alt and Ferree (1998) find evidence of an electoral cost of tax revenue in state elections, but that cost depends on the party of the governor. They analyse the influence of changes in tax revenues on governor and house elections in American states and find that Republicans are punished (rewarded) for increasing (lowering) tax revenues, while the opposite is true for Democrats.4 Peltzman (1992), however, finds that Democratic governors bear a higher electoral cost for expenditure growth than Republican governors. In the literature on economic voting there are two conflicting theories on the link between parties and voting. One view is that voters hold parties more responsible for their most salient goals. Powell and Whitten (1993) and Veiga and Veiga (2004) find that left-wing governments are more penalized for unemployment and that right-wing governments are more penalized for inflation. The Partisan Voter Model (Swank, 1993) and the �clientele hypothesis� (Rattinger; 1981,1991) suggest the contrary: left-wing governments benefit from unemployment, right-wing parties benefit from inflation In as much as lower taxes are a salient goal of right-wing governments, the findings of Lowry, Alt and Ferree (1998) that Republicans are punished for unexpected tax increases, are in line with the salient goal hypothesis. Another finding by Lowry, Alt and Ferree (1998) is that unified governments will face a higher electoral cost of tax policy than divided governments. When revenues or expenditures rise Republican governors lose votes when they control both branches. When only the governor is Republican and the legislative branch is controlled by the Democrats, there is no significant loss in votes for the Republicans (they win votes but the coefficient is not significant). This supports the �clarity of responsibility-hypothesis� which states that the impact of economic variables on voting is larger for governments with a greater clarity of responsibility (Powell and Whitten, 1993). Accordingly politically fragmented or divided governments are held less responsible for policy and economic developments than one-party governments or unified governments.5 3 They look at the income tax liability for joint filers with no dependents earning $40,000 and earning $100,000. The significant effect is only found for the latter category 4 The significance of these effects depends crucially on whether the changes are anticipated or not and on the presence of unified or divided government. 5 Other empirical evidence that the clarity of responsibility affects the electoral impact of economic variables is given by Anderson (1995, 2000) and Nadeau et al. (2002).

6

2. Elections and tax policy in Flemish municipalities 2.1. Elections in Flemish municipalities The Flemish region consists of 308 municipalities.6 Flemish municipalities have a parliamentary political system using mandatory voting. Municipal elections take place every 6 years (the most recent election year being 2000). In each of the 308 municipalities councillors are elected using a system of proportional representation (�Highest Averages Imperiali�). Depending on the size of the municipality, between 7 and 55 councillors are elected. A local government (mayor and aldermen) is formed by the party or by a coalition of parties that possesses a majority of the seats. The system of proportional representation leads to a highly diverse political landscape. In the most recent municipal elections, on average 5.16 parties presented themselves to the local voter. Whereas in some municipalities only two parties contested, the highest number of parties in a municipal contest was as high as 13. Many of these parties have a local character with no or only a modest link to one of the six main �national� parties. For most of the parties, however, such a link was present and made explicit by using the �national� party�s name in the campaign. The six �national� parties were: the ecologist Agalev, the christian democratic CD&V, the social democratic SP.a, the liberal democratic VLD, the nationalist VU and the extreme-right Vlaams Blok. For these parties, we put the word national between quotation marks as their political action is limited to only the Flemish Community, i.e. the Flemish speaking subset of the Belgian population. In federal elections, French speaking ecologists, Christian democrats, � participate too. These parties are separate entities without formal links with their Flemish speaking sister parties.7 Figure 1 gives for each �national� party as well as for the local parties taken together how often they were part of the local council or government (in percentage of municipalities.) The political dominance of the christian democrats is clear in figure 1. The CD&V governed in more than 60 % of all municipalities during the period 1982-2000. The two other major parties were the social democrats and the liberal democrats. The green party was part of government in only a small fraction of municipalities. The nationalist VU has seen its importance decline over the years. Another thing of note is the complete absence of the Vlaams Blok in the municipal governments. This

6 The other Belgian regions, Brussels and Wallonia consist of 19 and 262 municipalities respectively. 7 At the beginning of the seventies Belgian political families split up in Dutch speaking party and a French speaking party. Until then parties were truly national. In elections for the Council of the German Speaking Community German speaking parties participate.

7

reflects the so-called �cordon sanitaire� by which all major parties have tacitly agreed not to form a coalition with the �undemocratic Vlaams Blok�.8 Lastly, we observe that local parties entered the municipal coalition (or had a majority of the seats) in about a third of the municipalities.

Figure 1: Party landscape in de Flemish municipal councils and governments

Councils

0

20

40

60

80

100

Agalev

CD&VSP.a

VLD VU

Vl. Blok

Loca

l

Agalev

CD&VSP.a

VLD VULoc

al

Governments

Percentage

1982 1988 1994 2000

Source: Geys, 2003

2.2. Taxes in Flemish municipalities A major responsibility for the Flemish municipal governments is to set taxes. Tax revenues account for more than 40 percent of local revenues, the rest coming mainly from grants. Local governments have considerable freedom in choosing tax policy. This has led to a quite unique situation at the local government level in Flanders: municipalities use on average about 20 different taxes. Taken together, well over 100 different local taxes are in use. Still, while this remarkable diversification is indeed a prominent characteristic of the Flemish municipal tax system, it should be noted that revenues are highly concentrated in only two taxes. Table 1 shows the most important municipal taxes in terms of revenue. The local income tax (LIT) and the local property tax (LPT) generate about 80 percent of all tax revenues. Both taxes are of equal importance. They both are single rate surcharges on the federal

8 Note that the Vlaams Blok is indeed quite successful in some municipalities. It obtained on average 11 % of the valid votes and 9 % of the seats in the municipalities where it participated at the elections in 2000. In Antwerp, the Vlaams Blok is the largest party with around one third of the votes.

8

income tax and the regional property tax respectively. Their tax base is defined uniformly for the 308 Flemish municipalities. To test whether and how tax policy influences election results we can therefore focus on the impact of those two tax rates. Table 1: Five most important municipal taxes in terms of revenue in 2000 Tax Percentage of total tax revenue Local Income Tax 45.8 % Local Property Tax 36.7 % Tax on Motor-vehicles 2.2 % Environmental Tax 1.3 % Tax on domestic waste 1.1 % Source: Administratie Binnenlandse Aangelegenheden, 2002 The average local income tax rate in 2000 was 6.59 %, meaning that the LIT was 6.59 % of the federal income tax bill. Local income tax rates range from 0 tot 8.5 %. The average rate has remained nearly constant between 1988 and 2000. The average local property tax rate was 1073 % in 2000. This means that the local property tax was around ten times the amount of the regional property tax on which it was a surcharge. The regional property tax itself corresponds to 2.5 % of the assessed net rental value of the properties. So, the average local tax corresponds with (10.73*2.5=) 28,8 % of the net rental value. Given that � for budgetary reasons � no re-assessment took place since 1979 (when property values were expressed in 1975 prices) the tax base does not �really� reflect the net rental values. Since 1991, property values are adjusted for inflation (but not re-assessed). As a result, the link between the actual market (rental) value of properties and the tax has weakened. Just as for the local income tax rate, important variation exist for the local property tax rate which ranges from 438 % to 2000 %. 3. Empirical Analysis As evidenced in the previous section, local taxation is crucial to the Flemish municipalities. More precisely, and despite the highly diverse tax structures, two taxes play a central role: the local income tax and the local property tax. Given their budgetary importance in virtually all municipalities (the local property tax is levied in all 308 municipalities, the local income tax is in use in 305 municipalities), we concentrate on these two taxes in our empirical analysis of the electoral cost of taxation. The review of the literature (section 1) made clear that the electoral cost of taxation depends on a set of intermediary factors. These have to be accounted for in our empirical specification. Therefore, we start � in section 3.1 - with a discussion of each of the intermediary factors within the Flemish municipal

9

context. This discussion allows us to present an empirical specification (section 3.3) that allows estimating the electoral cost of taxation using panel data on three successive elections (1988, 1994 and 2000). As will be discussed in section 3.2, our panel is unbalanced due to data availability. While not being complete, we do cover 690 elections. The results are summarised in section 3.4. 3.1 Intermediary factors As mentioned earlier, several factors influence the electoral cost of tax policy. We discussed four possible factors. The first is the nature of the tax itself. A more conspicuous tax is expected to have a larger electoral cost. Second, tax policy in neighbouring jurisdictions can yield information for voters to evaluate tax policy of the incumbent government. A third factor that can possibly influence the electoral cost is the presence of fragmented or divided government. Finally the ideology of the incumbent government may be relevant. We will discuss these four factors in more detail in the context of Flemish local governments.

• Nature of the tax

The visibility of a tax has been identified as a major source of fiscal illusion, i.e. as a source of systematic misperception of the cost of public goods. Several sources of fiscal illusion have been identified in the literature: the complexity of the tax system, specific characteristics of the taxes and the tax collection process, the presence of grants as a source of revenue, � are well-known sources of possible misperception. Concentrating on the specific characteristics of taxes, we see three reasons why the local income tax is more likely to be underestimated than the local property tax. This implies three reasons why the latter tax may have higher political costs. At the same time, two arguments can be found that suggest the property tax has lower political costs. First, in contrast to the local income tax, the local property tax is more likely to be perceived as a truly local tax. Both taxes are surcharges. This means that � in practical terms - they are collected together with the federal (regional) tax on which they are based (see section 2). As a consequence, taxpayers may fail to identify the �municipal component�. The local component of the income tax is only a marginal fraction (around 7 %) of the total tax liability. In contrast, the local property tax is much larger than the regional tax on which it is based (the local component is about 90 % of the tax liability). As a result, taxpayers may not distinguish between the respective components of the tax and consider �the income tax� as a federal tax and �the

10

property tax� as a local one.9 To the extent that this is the case, the property tax is expected to have higher electoral costs for the incumbent government. A second crucial difference is the collection process of both taxes. The local income tax is collected through a Pay As You Earn-system whereby employers withhold taxes. In the case of the property tax the whole amount is paid when the tax notice arrives, which is once a year. As the explicit payment of the whole amount is more visible, the property tax may have a larger political cost. Third, the tax base elasticity differs between both taxes. The income tax grows automatically with inflation and growth. Constant income tax rates therefore allow for growing revenues. Given the progressive nature of the federal income taxes, the base elasticity is higher than 1. However the property tax base was not adjusted for inflation until 1991. Until then statutory rates had to be raised in order to increase revenues. These rate increases may have led to more media attention and a greater awareness of the property tax among voters. Again the property tax can be expected to have a higher electoral cost. On the other hand, the local property tax is a tax on (mainly) residential property. For rented properties, it is formally paid by the owner and (partly) shifted to the renter. To the extent that renter illusion occurs renters may not perceive (or underestimate) their tax burden. As a result, the local property tax may have a lower electoral cost among renters. 10 In addition to these perception-related differences between both taxes, it should be noted that both differ also in terms of the definition of the tax base. While the income tax is residence based, the property tax is source based. This opens - in the latter case � the possibility for tax exportation, possibly lowering the expected political cost associated with the property tax.

• Neighbouring jurisdictions Neighbouring municipalities� tax policy can be used as a yardstick to evaluate the tax policy of the incumbent government. The context of Flemish municipalities should be favourable to yardstick voting. As the two principal taxes are surcharges on a regional and a federal tax, the tax base of those taxes is defined in the same way for all municipalities. So voters can compare their tax policy basically by comparing the two rates. 9 However, a study by Heyndels (1989) shows both the local income tax and the local property tax are identified as a local tax by around half of the voters. Voters were asked to name the most important taxes in their municipality. Around 61 % of the voters could name a local tax. Of this group respectively 49.5% and 48.5 % mentioned respectively the local property tax and the local income tax as a local tax. 10 Heyndels and Smolders (1994) analyse the presence of different types of fiscal illusion in Flemish municipalities. They find no evidence for renter illusion.

11

Heyndels (1991) shows that voters� opinions on whether the level of taxation is high or low in their municipality compared to neighbouring municipalities, are generally consistent with reality. Neighbours� tax rates are also used as a point of reference by politicians (Ashworth and Heyndels, 1997, 2000; Heyndels and Ashworth, 2003). When voters evaluate relative performance, incumbent governments will be inclined to mimic the behaviour of neighbouring governments. Heyndels and Vuchelen (1998) offer evidence on tax mimicking among Belgian municipalities. They show that both tax rates are influenced by the tax rate in neighbouring municipalities.

• Partisan and ideological effects

The proportional system of representation in Flemish municipalities leads to a diverse political landscape where parties take distinct ideological positions. This is documented in table 2 which reproduces data from Deschouwer (1996). These data give the positioning of the main parties on a left-right scale. In a survey, local party presidents and spokesmen were asked to position their party on a left-right axis, with 0 representing extreme left and 10 representing extreme right. The CD&V, VLD and VU are at the right of the centre while Agalev and SP.a are at the left. Local parties are in the middle. The anti-immigrant party Vlaams Blok is at the extreme right of the spectrum.

Table 2: Ideological position of Flemish parties 1988 1994 Agalev 2.9 2.6 CD&V 5.1 5.3 SP.a 2.8 2.7 VLD 5.6 6.1 VU 5.1 4.7 Vlaams Blok n.a. 8.9 Local Parties 5.0 n.a.

Source: Deschouwer, 1996

As discussed in section 1, the electoral cost of tax policy can depend on the ideology of the incumbent party or coalition. Taxation is � generally speaking � a more salient issue for right-wing parties. Also in Flemish municipalities this is the case. A survey among municipal party representatives (Deschouwer, 1996) demonstrates this. To be more precise, representatives of the liberal VLD consider tax policy to be the single most important election theme. For the left-wing ecologists from Agalev tax policy is considerably less important. Likewise, a study of voter behaviour in Belgian national elections shows that economic issues � including tax policy-

12

are especially a voting motive for VLD-voters (Swyngedouw, 2000). This indicates that tax policy is a more salient theme for right-wing governments than it is for left-wing governments. According to the salient goal hypothesis we expect that right-wing governments are more penalized for high taxes. According to the partisan voter model and the clientele hypothesis however, high levels of taxation could benefit right-wing governments. Table 3: Ideological complexion of municipal governments

1983-1988 1989-1994 1995-2000 Average 4.8 4.8 5.0

Minimum 2.8 2.8 2.7 Maximum 5.6 5.6 6.1

Source: own calculations based on Deschouwer (1996) Many municipalities have coalition governments (see below). As a result ideological diversity is somehow moderated. Still, considerable ideological diversity exists at the government level. Table 4 gives the average ideological complexion of local governments for the period 1983-2000. This indicator is a weighted average of the ideological scores, where the weight is the relative number of mayor or aldermen of a certain party in the government (Kontopoulos and Perotti, 1997).11 As such the ideological complexion of Flemish local governments in the period 1982-2000 ranged from 2.7 to 6.1, with an average of 4.9.

• Fragmented government

While in some Flemish municipalities single party majorities dominate local politics, in most municipalities coalitions are formed and politics is highly fragmented. This is documented in table 4. This table gives � for each period � the composition of governments in terms of the number of parties. Around 60 % were coalition governments, with the average number of parties constituting a government being 1.7. Political fragmentation of the government may reduce the impact of tax policy on election results. The clarity of responsibility hypothesis allows coalition members to shift the blame of unpopular measures (like raising taxes) to one another. Ashworth, Geys and Heyndels (2004) find that Flemish municipal two-party coalitions are more likely to introduce a new tax than one party majorities. On the other hand they also observe that the likelihood of a new tax being introduced is lower in three and four-party local governments. They argue that this may be due to the fact that �gridlock� sets 11 To create the indicator data from the 1988 survey were used for governments in the period 1983-1988 and 1989-1994. Data from the 1994 survey were used for the governments in the period 1994-2000. Local parties were given an ideological complexion of 5.05 in 1994 (the unweighted average of the other parties).

13

in, i.e. that taking any policy measure (which introducing a new tax certainly is) becomes more difficult. Table 4: Political fragmentation in Flemish municipal governments

Source: (Geys, 2003) 3.2. Data We study the elections of 1988, 1994 and 2000 in the 308 Flemish municipalities.12 Our dependent variable is the percentage of the vote for the government parties. We introduce the vote of the government party/parties at the preceding election as a explanatory variable to control for the influence for past events (Peltzman, 1992). However the vote percentage of the government parties at the elections or the previous elections is not always available. Sometimes government parties do not participate at the elections following their government term. They may split up, or merge with another party � sometimes an opposition party � or simply disappear. It is not always easy to determine which party participating at an election corresponds to a certain government party. A party may participate at an election with a different name. However in that case, it is difficult to determine which party is the �successor� of the original party. Therefore we only consider these cases where government parties participate at the elections with the same name.13 Similarly, it is not always possible to calculate the vote percentage at the previous elections, because government parties did not always participate at these elections in the same form. A government party may be a breakaway faction that formed a new party or a part of a cartel list, and it is impossible to know which percentage of the vote this party would have received if it participated at the previous elections in its current form.

12 The electoral data come from the Elections Database of the Political Science Department of the Vrije Universiteit Brussel. Data on municipal governments come from the municipal data collection �Gemeentelijk Zakboekje�(1985, 1994, 1999, 2002). 13 In this we follow Buelens and Deschouwer (1997).

Num ber of governm ent parties

1983-88 1989-1994 1995-2000

1 party 126 40.9%

140 45.5%

120 39.0%

2 parties 132 42.9%

136 44.8%

149 48.4%

3 parties 43

14.0% 27

8.8% 31

10.1% > 3 parties 7

2.3% 5

1.6% 8

2.6%

14

The consequence of this, is that our panel is unbalanced. Moreover, we also dropped these observations where a mayor or alderman (any member of the municipal government) was a candidate on a new list or the list of an opposition party in the next election. In those cases it is difficult to determine which parties are to be considered as the real government parties and accordingly which parties are to be held responsible for tax policy. As a result our dataset includes 690 observations � of the possible 924 observations. 3.3. Empirical model To investigate the electoral cost of taxation, we follow a two-step procedure. First, we estimate a vote function relating the votes obtained by the incumbent party (parties) - Ve - to the main economic, political and budgetary variables (Paldam and Schneider, 1980; Nannestad and Paldam, 1994).14 The vote function can be written as: Ve = β0 + β1Ve-1+ β2 INC + β3 UN + β4NPG + β5 EXP + β6 LIT + β7 LPT + β8 LITN + β9 LPTN + u Second, and to gain better insight into the mechanisms underlying the taxation-electoral result nexus, we estimate a more encompassing model that integrates a number of interactions: Ve = β0 + β1Ve-1+ β2 INC + β3UN + β4NPG + β5 EXP + β6 LIT + β7 LPT + β8 LITN + β9 LPTN + β10 IDEO + β11 LIT*NPG + β12 LPT*NPG + β13 LIT*IDEO + β14 LPT*IDEO + β15 LITN*NPG + β16 LPTN*NPG + β17 LITN*IDEO + β18 LPTN*IDEO + u Where Ve is the vote percentage of the government party (or parties) and Ve-1 is the vote percentage of the same party (parties) in the previous elections, the elections that brought them into power. The literature on economic voting states that governments are held accountable for economic developments. While (macro-) economic policy is � of course � mainly a federal and regional responsibility, local governments may have a (marginal) influence or be held accountable by the electorate despite their objective inability to interfere. Therefore, we introduce real per capita income (INC) and the unemployment rate (UN) in our regression. We expect income to have a positive effect � prosperity translating into electoral support � and unemployment to negatively affect the incumbent�s election result. 14 Municipality and year indices are left out for convenience.

15

In our first regression we investigate the impact of tax rates in the municipality and the neighbouring municipalities without looking at the possible interaction with clarity of responsibility. However, we already include the number of government parties (NPG) as a measure of clarity of responsibility in the regression. When a government is made up of more coalition partners, the responsibility of each coalition partner for policy and economic developments is less clear. Also, voters can shift their votes between different government parties (Powell and Whitten, 1993; Whitten and Palmer, 1999). Accordingly, the number of government parties is expected to have a positive effect on the vote. As said earlier, we use the rates of the local income tax and the local property tax as an indicator of local tax policy (LIT and LPT). We use the level of tax rates in the election year. We expect these to have a negative effect on the election result. The theory on yardstick competition suggests that voters can use information on tax rates in neighbouring municipalities (Besley and Case, 1995; Revelli, 2002). We use the unweighted average of LIT and LPT rates in border-sharing municipalities as our variables (LITN and LPTN).15 The level of tax rates in neighbouring municipalities is expected to have a positive impact on the election results. Finally, we include the per capita expenditures in the municipality. To better understand the precise nature of the electoral impact of tax policy, we introduce a set of interaction variables. First, we allow for the possibility of an effect of government fragmentation, measured by the number of parties in government. The clarity of responsibility hypothesis suggests that the electoral cost of taxation is lower if the number of coalition members is larger. Therefore we expect β11 and β12 to be positive (thus offsetting the effects captured by β6 and β7). The same reasoning applies to the electorate�s evaluation of neighbours� tax policy. While higher taxes in neighbouring municipalities are expected to be politically rewarding for the incumbent as such higher taxes are seen as proof of his policy, such reward may be moderated if � because the government is a coalition of parties � voters do not �know� who to reward. As a result, we expect β15 , β16 < 0. A second interaction is possible with the ideological position of the incumbent (IDEO). The salient goal hypothesis suggests that � to the extent that lower taxes are a salient goal for right wing governments � right wing governments are more punished for higher levels of taxation. Therefore, we expect β13 , β14 < 0 and β17 , β18 > 0.

15 We also include neighbouring municipalities in the Brussels and Walloon region, as the context is very similar. The local income tax and the local property tax are also the major local taxes and the tax bases of both taxes are defined on the federal level (although the property tax is a regional tax).

16

We �standardize� our variables. Apart from facilitating comparisons among independent variables, �standardizing� also reduces the multicollinearity due to interaction variables (as the variables are �centered�). The interaction terms are the product of the �standardized� variables, but are not �standardized� themselves, according to the prescriptions in the literature on interactions. (Jaccard et al.; 90,33; Aiken and West; 91,11). A final thing to note is that local elections do not take place in a (political) vacuum. The impact of national or regional politics on local elections is well established (Heath et al., 1999; Jérôme and Lewis-Beck, 2000 and Revelli, 2002). In extreme situations, local elections can be considered as referenda for national politics and, consequently, the local election results may show little relationship with local developments and policies. Such a multi-level dependency may, in the Belgian situation, reveal itself by the fact that changes in the popularity of national parties translate into changes for the corresponding municipal parties. To capture this possibility, we included a set of party-year dummies.16 In what follows, we will present the results with and without these dummies. 3.3. Results We estimate pooled OLS regressions. 17 The regressions without party-year dummies are shown in Table 6.18 Column 1 shows the regressions without interactions. The vote percentage at the previous election has as significant positive impact, as expected. Of the economic variables the unemployment percentage has no effect on the vote. This is not surprising as most policy instruments to fight unemployment are in the hands of the federal and regional government. Per capita income however has a significant negative impact on the incumbent parties. A possible explanation is that at lower levels of per capita income, voters tend to stay with or go back to traditional government parties, reminding of the �clientele hypothesis� by Rattinger (Rattinger, 1981, 1991). Separate regressions (not shown here) show us that 16 We include party-year dummies for the five �national� parties that participated in municipal government. A dummy is one if a certain party was part of municipal government in that election. 17 We also estimated fixed effects regressions. The results are in appendix. We show F statistics to test for the joint significance of the fixed effects. The test rejects for most base regressions that the fixed effects are jointly significant at a 5 % level, however they are significant at a 10 % level. We also estimated random effects regressions (not shown). However a Hausman test showed for each regression that fixed effects was preferred to random effects. The fixed effects approach introduces municipality specific constants in the specification. The presence of such constants would reflect that in some municipalities incumbents consistently have a better or worse election results, regardless of which parties are in power. There is no logical argument for such systematic municipality-specific differences. 18 Note that the interaction regressions may not always produce a zero intercept � as �standardized� regressions do � because the interaction variables are not �standardized�.

17

income has indeed a significant positive impact on the vote of the Green Party, a party that was in power in only a few municipalities, and on the vote of the extreme right Vlaams Blok, a party that has never been in power (see section 2.1.). This suggests that less traditional parties fare better at higher levels of income and that traditional government parties fare better at lower levels of income. The number of government parties has a significant positive effect on the vote for the government parties, as expected. This confirms that fragmented governments lose less votes as it is less clear which party voters should hold responsible for policy and economic developments. The main objective of our analysis was to find out if tax policy has an impact on election results. We see that the level of the property tax rate has a significant negative influence on the vote for the government parties. This shows that tax policy has indeed an effect on elections: incumbent parties are punished for higher tax rates. The income tax rate has also a negative impact on the vote, although smaller and not significant. The property tax rate seems to have a higher electoral cost. As mentioned earlier, this could be due to the fact that the local income tax is less perceived as a local tax. Moreover income taxes are witheld by the employer, whereas the collection of the property tax is more explicit. The average local property tax rate in the neighbouring municipalities has a significant positive impact on the election result. So, voters indeed use tax rates in neighbouring municipalities as a yardstick to evaluate tax policy in their own municipalities. In column 2 we show a regression where we dropped the insignificant variables. The results are similar. In a second step, we control for the interaction of clarity of responsibility � number of government parties- and ideology on the electoral cost of the tax rates. The results are shown in the third column of Table 2. Note that here the coefficient of the LPT rate gives us the electoral impact of the rate for an average number of government parties and for an average ideological complexion (Aiken and West, 1991). The interaction variables give us the change in this impact for a larger number of government parties or for more right wing governments. Again the LPT rate has a significant negative impact and the neighbours� LPT rate has a significant positive impact � at the 10 % level - on the electoral result of the government parties. The interaction of the neighbours� LPT rate with the number of government parties has a significant impact. Whereas fragmented governments � who have a less clear responsibility � would be expected to be less rewarded for higher neighbours� tax rates, we find evidence of the contrary. Governments with an average ideological complexion and number of parties are not significantly punsihed for higher LIT rates. However more right-wing governments are punished for higher LIT rates. This is in line with the salient goal hypothesis. Considering that lower taxes are a salient goal for right-wing governments, they are expected to be more severly punished for not

18

reaching that goal. In the last column we repeated the regression where we deleted the insignificant variables � one by one. The interaction of the LIT rate however becomes insignificant and is not included. The interaction of the LPT rate with the number of government parties becomes significant. Contrary to the clarity of responibility-hypothesis fragmented governments are more punished for higher LPT rates and more rewarded for higher rates in neighbouring municipalities. Lastly, the ideological position of the government has a significant positive effect, meaning that right wing governments obtained better election results. In the next regressions we take into account the effect of possible national effects on municipal elections by including party-year dummies for the five �national� parties who participated in municipal governments. The results are shown in Table 7. We do not show the party-year dummies themselves. The results are similar to the regression without party-year effects. Again, contrary to the clarity of responsibility hypothesis more fragmented governments are more punished for a higher LPT rate (significant at a 10 % level) and are more rewarded for a higher rate in the neighbouring municipalities. There is no significant interaction with ideology. Note that the effect of the ideological position of the government itself has lost its significance, now that we control for the electoral evolution of the �national� parties.

19

Table 6. Pooled OLS regressions (without party-year dummies) Dependent Variable Vote share

of incumbent party/ies

Vote share of incumbent party/ies



Method

Pooled OLS Pooled OLS Pooled OLS Pooled OLS

C .159E-02 (.06)

.498E-02 (.03)

Vote preceding election .638 (22.44)

.641 (22.63)

.638 (22.41)

.639 (22.64)

Real per capita income -.160 (-4.37)

-.137 (--4.79)

-.176 (-4.74)

-.146 (-5.07)

Unemployment rate -.042 (-1.18)

-.046 (-1.28)

Number of government parties

.142 (4.76)

.134 (4.62)

.159 (5.02)

.154 (5.02)

Ideology .056 (1.78)

.056 (1.89)

Per capita expenditures -.225E-02 (-.07)

.989E-02 (.30)

LIT rate -.039 (-1.29)

-.021 (-.70)

LPT rate -.102 (-2.54)

-.117 (-3.25)

-.103 (-2.55)

-.105 (-2.91)

LIT rate neighbours .028 (.81)

.031 (.89)

LPT rate neighbours .084 (1.99)

.111 (3.03)

.071 (1.66)

.104 (2.84)

LIT rate * number of government parties

-.020 (-.70)

LPT rate * number of government parties

-.060 (-1.45)

-.073 (-2.06)

LIT rate * ideology -.067 (-1.79)

LPT rate * ideology .025 (.59)

LIT rate neighbours * number of government parties

-.013 (-.37)

LPT rate neighbours * number of government parties

.089 (2.01)

.0783 (2.19)

LIT rate neighbours * ideology

-.281E-02 (-.08)

LPT rate neighbours * ideology

.027 (.59)

Adjusted R2 .480 .480 .483 .485

Values between brackets are t-values

20

Table 7. Pooled OLS regressions (with party-year dummies)

Dependent Variable Vote share of incumbent party/ies




Method

Pooled OLS Pooled OLS Pooled OLS Pooled OLS

C .397E-02 (.15)

.631E-02 (.24)


.649 (23.44)

.640 (22.28)

.648 (23.42)

Real per capita income -.088 (-1.95)

-.071 (-1.88)

-.088 (-1.91)

-.083 (-2.35)

Unemployment rate -.854E-02 (-.22)

-.019 (-.49)


.215 (3.52)

.127 (3.53)

.214 (3.44)

.135 (3.74)

Ideology -.046 (-.78)

Per capita expenditures .935E-02 (.29)

.017 (.50)

LIT rate -.013 (-.43)

-.204E-02 (-.06)

LPT rate -.097 (-2.43)

-.099 (-2.72)

-.103 (-2.54)

-.096 (-2.68)


.023 (.66)

LPT rate neighbours .118 (2.75)

.134 (3.55)

.117 (2.66)

.123 (3.33)


-.012 (-.44)


-.074 (-1.80)

-.081 (-2.35)

LIT rate * ideology -0.43 (-1.13)

LPT rate * ideology .020 (.46)


-.019 (-.50)


.098 (2.14)

.096 (2.60)


-.013 (-.38)


.017 (.36)

Adjusted R2 .501 .504 .501 .509

Values between brackets are t-values.

21

4. Conclusion We find that tax policy has a significant effect on the election results of Flemish municipal governments. The level of the local property tax rate has a significant negative effect on the vote. Moreover, the average property tax rate in the neighbouring municipalities has a significantly positive effect, yielding evidence on yardstick voting. We find that the number of government parties interacts with the electoral cost of the local property tax rate. However our findings are not in line with the clarity of responsibility hypothesis. More fragmented governments are more punished for higher property tax rates and more rewarded for higher property tax rates in the neighbouring municipalities. Lastly, we find no strong evidence for an impact of ideology on the electoral impact of tax policy.

22

Appendix Table 8. Fixed Effects regressions (without party-year dummies) (Values between brackets are t-values) Dependent Variable Vote share of

incumbent party/ies




Method

Fixed effects Fixed effects Fixed effects Fixed effects


.436 (10.04)

.427 (9.49)

.437 (10.10)

Real per capita income -.038 (-.40)

-.036 (-.37)

Unemployment rate .893E-02 (.16)

.019 (.32)


.484 (7.61)

.474 (7.63)

.519 (7.92)

.481 (7.76)

Ideology .098 (1.84)

Per capita expenditures -.151 (-.01)

.044 (.65)

LIT rate .030 (.45)

.044 (.65)

LPT rate -.197 (-2.28)

-.260 (-4.54)

-.201 (-2.32)

-.258 (-4.52)


.060 (.80)

LPT rate neighbours -.094 (-.78)

-.110 (-.88)


.055 (.80)


-.135 (-1.84)

LIT rate * ideology -.063 (-1.14)

-.085 (-1.99)

LPT rate * ideology -.082 (-1.19)


-.028 (-.46)


.148 (2.00)


-.040 (-.73)


.086 (1.38)

Adjusted R2 .519 .523 .525 .526

F test joint significance fixed effects

F(295, 385)=1.1857

p-value = [.0588]

F(295, 391)=1.3013

p-value = [.0076]

F(295, 376)=1.2013

p-value = [.0469]

F(295, 390)=1.3230

p-value = [.0049]

23

Table 9. Fixed Effects regressions (with party-year dummies) (Values between brackets are t-values) Dependent Variable Vote share of

incumbent party/ies




Method

Fixed effects Fixed effects Fixed effects Fixed effects


.443 (10.35)

.417 (8.73)

.432 (10.16)

Real per capita income -.036 (-.33)

-.020 (.17)

Unemployment rate .084 (1.12)

.067 (.88)


.520 (4.86)

.439 (7.10)

.535 (4.77)

.479 (7.67)

Ideology -.078 (-.74)

Per capita expenditures .027 (.23)

.015 (.12)

LIT rate .062 (.92)

.054 (.79)

LPT rate -.183 (2.14)

-.232 (-3.39)

-.189 (-2.18)

-.190 (-2.72)

LIT rate neighbours .087 (1.12)

.081 (1.02)

LPT rate neighbours -.068 (-.55)

-.080 (.62)


.042 (.61)


-.108 (-1.44)

LIT rate * ideology -.040 (-70)

-.088 (-2.07)

LPT rate * ideology -.080 (-1.14)


-.043 (-.65)


.096 (1.17)


-.055 (-.95)


.068 (1.00)

Adjusted R2 .541 .541 .540 .552

F test joint significance fixed effects

F(295, 383)=1.1849

p-value = [.0597]

F(295, 388)=1.2987

p-value = [.0080]

F(295, 362)=1.1887

p-value = [.0588]

F(295, 386)=1.3665

p-value = [.0020]

24

Bibliography Ackaert, J. 1994. De Gemeenteraadsverkiezingen. Leuven: Davidsfonds.

Aiken, L.S., S.G. West. 1991. Multiple Regression: Testing and Interpreting Interaction. Newbury Park, Ca: Sage Publications

Administratie Binnenlandse Aangelegenheden. 2002. De Gemeentefinanciën 2002: De financiële toestand van de Vlaamse gemeenten op basis van de rekeningen van 2000. Brussel: Ministerie van de Vlaamse Gemeenschap.

Anderson, C. 1995. Blaming the government: Citizens and the Economy in Five European Democracies. Armonk, N.Y.: M.E. Sharpe.

Anderson, C.J. 2000. « Economic voting and political context : a comparative perspective. » Electoral Studies 19: 151-170.

Ashworth, J. and B. Heyndels. 1997. �Politicians� Preferences on Local Tax rates: an Empirical Analysis.� European Journal of Political Economy. 13(3): 479-502.

Ashworth, J. and B. Heyndels. 2000. �A schema-theoretic approach to politicians� definitions on tax issues.� Journal of Economic Psychology. 21: 21-42.

Ashworth, J., Geys, B. and B. Heyndels. 2003. �Determinants of tax innovation: The Case of Environmental Taxes in Flemish Municipalities.� Paper presented at the EPCS-conference, Aarhus, Denmark, April, 2003.

Besley, T. and A. Case. 1995. �Incumbent behavior: Vote-seeking, tax-setting and yardstick competition.� American Economic Review 85: 25-45.

Buelens, J. and K. Deschouwer. 1997. �De gemeenten en de lokale politiek in het Brussels Hoofstedelijk Gewest.� Res Publica 39(1): 89-99.

Case, A. 1994. Taxes and the Electoral Cycle: How sensitive are governors to coming elections?� Business Review March/April 17-26.

Deschouwer, K. 1996. �Nationale partijen en gemeenteraadsverkiezingen.� In De dorpsstraat is de Wetstraat niet: Een onderzoek naar de opvattingen van plaatselijke woordvoerders van nationale partijen, Beulens, J. and K. Deschouwer. (red.). Brussel: VUBPRESS.

Downs, A. 1957. An Economic Theory of Democracy. New York: Harper & Row.

Eismeier, T.J. 1979. �Budgets and ballots: The political consequences of fiscal choice.� In Public Policy and Public Choice, (eds.) D. Rae and T. Eismeier. Beverly Hills: Sage.

Eismeier, T.J. �Votes & taxes: The political economy of the American governorship.� Polity 15: 368-379

Gemeentelijk Zakboekje ´85/´86. 1985. Antwerpen: Kluwer.

Gemeentelijk Zakboekje ´94/´85. 1994. Zaventem: Kluwer Editorial.

25

Gemeentelijk Zakboekje 1999: Databank. 1999. Diegem: Kluwer Editorial.

Gemeentelijk Zakboekje 2002: Databank. 2002. Diegem: Kluwer.

Geys, B. 2003. �Hoe verdeeld is de lokale politiek? Politieke fragmentatie in Vlaamse gemeenten.� te publiceren in Res Publica.

Gibson, J.G. and J.D. Stewart. 1992. �Poll tax, rates and local elections.� Political Studies 40: 516-513.

Hansen, S.B. 1999. �Life is not fair: Governors� job performance ratings and state economies.� Political Research Quarterly 52: 167-188.

Happy, J.R. 1992. �The effect of economic and fiscal performance on incumbence voting: The Canadian case.� British Journal of Political Science 22: 117-130.

Heyndels, B. 1989. �Fiscale perceptie bij de gemeenteraadsverkiezingen.� De Gemeente 4: 159-63.

Heyndels, B. 1991. �De financiering van de Belgische gemeenten in de jaren tachtig.� Driemaandelijks Tijdschrift van het Gemeentekrediet 177: 3-18.

Heyndels, B. 1992. �De gemeentelijke fiscaliteit in het Vlaamse Gewest: een analyse van de begrotingen 1991.� Het Tijdschrift van het Gemeentekrediet 180: 5-14.

Heyndels, B. 1993. �Federale belastinghervormingen en hun invloed op de Brusselse gemeentefinanciën.� BRES Periodiek Bericht 5: 1-8.

Heyndels, B. and J. Ashworth. 2003. �Self-serving bias in Tax Perceptions: Federalism as a Source of Political Instability.� Kyklos. 56 (1): 47-68.

Hibbs, D.A., Jr. and H.J. Madsen. 1981. �The impact of economic performance on electoral support in Sweden, 1967-1978.� Scandinavian Political Studies 4: 3350.

Holmstrom, B.R. 1982. �Moral hazard in teams.� Bell Journal of Economics and Management Science 93(2): 284-64.

Jaccard, J., Turrisi, R. and C.K. Wan. 1990. Interaction effects in Multiple Regression. Newbury Park, Ca: Sage Publications.

Kone, S.L. and R.F. Winters. 1993. �Taxes and voting: Electoral retribution in the American states.� The Journal of Politics 55: 22-40.

Landon, S. and D.L. Ryan. 1997. �The political costs of taxes and government spending.� Canadian Journal of Economics 30: 85-111.

Lowry, R.C., Alt, J.E. and K.E. Ferree. 1998. �Fiscal policy outcomes and electoral accountability in American states.� American Political Science Review 92: 759-774.

26

MacDonald, J.A. and L. Sigelman. 1999. �Public assessments of gubernatorial performance. A comparative state analysis.� American Politics Quarterly 27: 201-215.

Mueller, D.C. 2003. Public Choice III. Cambridge: Cambridge University Press.

Nadeau, R., Niemi, R.G. and A. Yoshinaka. 2002. �A cross-national analysis of economic voting: taking account of the political context across time and nations.� Electoral Studies. 21: 403-23.

Nalebuff, B.J. and J.E. Dtiglitz. 1983. �Prizes and incentives: towards a general theory of compensation and competition.� Bell Journal of Economics 14: 21-43

Nannestad, P. and M. Paldam. 1994. �The vp-function: A survey of the literature on vote and popularity functions after 25 years.� Public Choice 79: 213-245.

Niemi, R.G., Stanley, H.W. and Vogel, R.J. 1995. �State economies and state taxes: Do voters hold governors accountable?� American Journal of Political Science 39: 936-957.

Niskanen, W.A. 1975. �Bureaucrats and Politicians.� 1975. Journal of Law and Economics 18: 617-643.

Niskanen, W.A. 1979. �Economic and fiscal effects on the popular vote for the president.� In Public Policy and Public Choice, (eds.) D. Rae and T. Eismeier. Beverly Hills: Sage.

Paldam, M. 1981. �A preliminary survey of the theories and findings on vote and popularity functions.� European Journal of Political Research 9: 181-199.

Paldam, M. 1991. �How robust is the Vote Function? A Study of Seventeen Nations over Four Decades.� In Economics and Politics: The Calculus of Support., (eds.) H. Norpoth, M.S. Lewis-Beck and J-D. Lafay. Ann Arbor: The University of Michigan Press.

Paldam, M. and F. Schneider. 1980. �The macro-economic aspects of government and opposition popularity in Denmark 1957-78.� Nationalokonomisk Tidsskrift 2: 149-170.

Powell, G.B. and G.D. Whitten. 1993. �A cross-national analysis of economic voting: Taking account of the political context.� American Journal of Political Science 37: 391-414.

Peltzman, S. 1992. �Voters as fiscal conservatives.� The Quarterly Journal of Economics 107: 327-361.

Pissarides, C.A. 1980. �British government popularity and economic performance.� The Economic Journal 90: 569-581.

Perotti, R. and Y. Kontopoulos. 2002. �Fragmented fiscal policy.� Journal of Public Economics 86: 191-222.

Pomper, G.M.. 1968. Elections in America: Control and Influence in American Politics. 1968. New York: Dodd, Mead & Company.

27

Rattinger, H. 1981. �Unemployment and the 1976 election in Germany: Some findings at the aggregate and the individual level of analysis� In Contemporary Political Econom., (eds.) D.A. Hibbs and H. Fassbender. Amsterdam: North-Holland Publishing Company.

Rattinger, H. 1991. �Unemployment and elections in West-Germany� In Economics and Politics: The Calculus of Support., (eds.) H. Norpoth, M.S. Lewis-Beck and J-D. Lafay. Ann Arbor: The University of Michigan Press.

Revelli, F. 2002. �Local taxes, national politics and spatial interactions in English district election results.� European Journal of Political Economy 18: 281-299.

Swank, O.H. 1993. �Popularity functions based on the partisan theory.� Public Choice 75: 339-56.

Swyngedouw, M. 2001. �The subjective cognitive and affective map of extreme right voters: using open-ended questions in exit polls.� Electoral Studies 20: 217-41.

Turett, S. 1971. �The Vulnerability of American Governors, 1900-1969.� Midwest Journal of Political Science 15: 108-132.

Verbeek, M. 2000. A Guide to Modern Econometrics. Chichester, England: John Wiley & Sons.

Vereniging van Vlaamse Steden en Gemeenten. 2002. Financiële uitdagingen voor gemeenten en OCMW�s. Brussel: VVSG.

Vereniging van Vlaamse Steden en Gemeenten. 2003. De lokale belastingverhogingen nader bekeken: analyse van de tariefstijgingen in de periode 2001-2003. Brussel: VVSG.

Whitten, G.D. and H.D. Palmer. 1999. �Cross-national analysises of economic voting.� Electoral Studies 18: 49-67.

the electoral cost of tax policy in flemish municipalities · 2021. 4. 10. · evidence of an...

Documents