• Tidak ada hasil yang ditemukan

Directory UMM :Data Elmu:jurnal:J-a:Journal of Economic Psychology:Vol21.Issue1.Feb2000:

N/A
N/A
Protected

Academic year: 2017

Membagikan "Directory UMM :Data Elmu:jurnal:J-a:Journal of Economic Psychology:Vol21.Issue1.Feb2000:"

Copied!
21
0
0

Teks penuh

(1)

A schema-theoretic approach to politicians

Õ

de®nitions of

tax issues

John Ashwortha, Bruno Heyndelsb,* a

Department of Economics, University of Durham, 23-6 Old Elvet, Durham DH1 3HY, UK

b

CEMS, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium

Received 5 March 1998; received in revised form 21 September 1998; accepted 10 September 1999

Abstract

This paper investigates how politicians de®ne the level of the tax burden in their jurisdiction in terms of a scale from low to high. This de®nition is driven by evaluative components in politiciansÕtax burden schema. These components are, in part, a€ective and, in part, a con-frontation of a€ect and cognition. Assuming that politicians have an objective function in which electoral and ideological objectives exist along with the maximisation of self-esteem, we present an empirical model of politiciansÕde®nitions. The model is tested, using ordered probit analysis, on a sample of survey responses of 682 Flemish politicians who were active at the local (municipal) level. The results indicate that all aspects of the politiciansÕ schema are represented by signi®cant variables. In addition to the measured tax burden, there are signi-®cant roles for the ideological position of the electorate, the proportion of taxes that can be exported, the income of the population and the tax burden in neighbouring municipalities together with the ideological position of the politician and whether (s)he has been directly involved in creating the burden. Ó 2000 Elsevier Science B.V. All rights reserved.

PsycINFO classi®cation:2260; 2340; 2960

www.elsevier.com/locate/joep

*Corresponding author.

E-mail addresses: John.Ashworth@durham.ac.uk (J. Ashworth), bheyndel@vnet3.vub.ac.be (B. Heyndels).

(2)

JEL classi®cation:D78; H71

Keywords:Tax choice; Schema theory

1. Introduction

The positive economic literature considers existing tax systems as the result of constrained maximising behaviour by self-interested politicians (Hettich & Winer, 1999). Government is seen as a black box modelled as if

it were an individual maximising expected votes, plurality, or as if it pur-sues ideological goals. At a lower level of aggregation, the literature starts from as if assumptions on rational behaviour of representatives and bu-reaucrats. Typically, in¯uences from decision-making processes at the

in-dividual level are assumed away. Psychological and experimental

economics, however, show that these processes have immediate relevance for economic behaviour, also in the political arena where tax choices are made (Conlisk, 1996).

A politicianÕs tax choices are based on whathe thinkshe is deciding on and on which options he thinks are available and relevant. Therefore, ``(b)efore we apply the methods of economic reasoning to political behavior we must characterise the political situation, not as it appearsÔobjectivelyÕ to the an-alyst, but as it appears subjectively to the actors'' (Simon, 1985, pp. 297, 298). The present paper looks at how politicians de®ne issues of taxation and analyses how this de®nition can be explained by current behavioural as-sumptions from the tax choice literature. To do so, we organise the current assumptions from (politico-)economics in a schema-theoretical framework from the psychological literature. As such, the hypothetical economic actor is linked to the ¯esh-and-blood actor from psychology.

The approach is used to analyse opinions on the level of the tax burden held by Flemish politicians who were active at the municipal level. A question on tax opinions was introduced in a large-scale survey which was sent during the summer of 1994 to local spokesmen of the six ``national'' parties 1in all

1These are: ``Agalev'' (green), ``SP'' (socialist), ``CVP'' (Christian democrat), ``VLD'' (liberal), ``VU''

(3)

308 Flemish municipalities. 2They were asked to rate the tax burden in their municipality onto a discrete opinion scale by answering the following ques-tion:

Do you feel the tax burden in your jurisdiction is low, rather low, average, rather high or high?

Completed questionnaires were returned by 682 local politicians. In Sec-tion 2 we discuss how politicianÕs de®nitions can be explained by their tax burden schemata. Section 3 presents the empirical analysis of the data. A summary and conclusion are in Section 4.

2. A politicians tax burden schema

In general, a politicianÕs opinion (O) corresponds with how he de®nes the level of the burden, that is with where he classi®es the burden on a contin-uous scale from low to high, so thatO 2 Š ÿ 1; ‡1‰. As a rule, we expect opinions to be positively related to the actual level of the tax burden…TAX†:

O ˆf…TAX† ‡k; f0>0; …1†

where f…TAX† is the ``true'' translation of tax burdens on the o€ered opinion scale. 3The k-term captures biases of individual responses.

Fig. 1 illustrates the transformation. The lower line gives the range of actual tax burdens. The upper line depicts the o€ered opinion scale. The

2

Our survey was organised ®ve months before the elections of October 1994. It is clear that this timing may have in¯uenced the relative importance of the respective determinants of politiciansÕopinions. More precisely it is possible that electoral determinants of opinions may be at their ``relative peak'' whereas ideological determinants are less important. Such a dynamic aspect corresponds with the political business cycle approach in Pommerehne and Schneider (1983) where behaviour ± instead of opinion ± is more driven by electoral considerations to the extent that the time before election is smaller. A set of panel data would be needed to test for potential political business cycle e€ects in politiciansÕopinions. Replies were received from 760 politicians. The ®nal sample of 682 was achieved by randomly ®ltering multiple respondents (so that we had at most one respondent of any political party in any municipality) and checking for a complete set of information (removing a further seven respondents). The full sample is 51% of the population.

3We assume thatf(TAX) captures the ``cognitive grounds'' of the translation of a tax burden onto an

(4)

transformation implied by (1), corresponds with an arrow from the lower to the upper line. Assuming the actual tax burden is TAX, the ``true'' or ``cognitive'' translation is depicted by the plain arrow. Actual responses of individualsi, j and k(dotted arrows), however, depart from this.

Schema theory suggests that politicians when ``choosing'' k activate an associative network (Fiske & Taylor, 1991). Here, the concept of taxation is related to what the decision maker considers relevant characteristics. The concept of tax burden is the superordinate node and relevant characteristics of the tax burden are contained in subordinate nodes of the schema. The schema concept is in the ®rst instance cognitive. Still, ``affective tags'' are attached to all nodes within the network (Lodge & McGraw, 1991, p. 1361). Thus, given an ``objective'' reality, evaluative components within the schema determine k and introduce subjectivity. This arises because the question re-quires active interpretation and processing of information. Beliefs about whether a tax burden is high or low correspond with what Boudon (1996) terms ``type III beliefs''; beliefs that cannot be validated by a confrontation with the real world because of their inherent subjectivity. The question under consideration is built around two concepts: theÔtax burdenÕand the (implicit) scale from low to high. It is the latter which involves active information processing and brings in subjectivity. No ``unique'' scale for tax burden ex-ists: it is not possible to de®ne a ``high'' or a ``low'' tax burden objectively. Therefore, each respondent must construct a scale for himself, which is by de®nition subjective.

In order to proceed, it is necessary to, ®rst, de®ne the nature of the nodes within politiciansÕtax burden schemata (i.e. the relevant characteristics of the concept ``tax burden'', where ``relevance'' indicates that the politicianÕs ob-jectives are a€ected), and, second, identify the nature of the a€ective tags on both superordinate and subordinate nodes.

(5)

2.1. Nature of nodes

A set of subordinate nodes extend the meaning of the concept ``tax bur-den''. The precise contents of individualsÕtax burden schemata are unknown. However, when he thinks about the tax burden, the individual politician inevitably thinks about these characteristics and the subordinate nodes are mentally inseparable from the object tax burden.

We hypothesise that the schema nodes can be deduced from his objec-tives. The way a politician thinks about the tax burden depends upon the expected in¯uence from this tax burden on the goals he wants to attain. A basic tax burden schema is given in Fig. 2. We assume that the politician has a triple objective function with electoral and ideological objectives supple-mented by a psychological objective, corresponding to the maximisation of self-esteem.

The ®rst two objectives correspond to the current politico-economic as-sumptions of vote-maximisation and maximisation of ideological objectives. Attainment of the electoral objective is a€ected to the extent that higher tax burdens lead to more political opposition (loss in votes). Thus, political oppositionassociated with the tax burdenis a relevant characteristic. From an ideological perspective, a relevant characteristic of the tax burden is the implied government intervention. The ®nal objective, self-esteem, is an ob-vious one in a context where politicians evaluate their own work. Such esteem is based on self-reference ``through which humans judge their own lives and relationships'' (Aaron, 1994, p. 15).

(6)

2.2. Nature of a€ective tags

As taxes are bad, 4Oin fact measures ``how bad'' respondents feel a given tax burden is. This evaluation may be purely a€ective or it may imply a confrontation of a€ect and cognition. The former re¯ects a respondentÕs

attitudes (A) towards the symbolic content of the schema nodes. The latter

relates to hisappreciation (g) of the actual contents of the nodes. Thus, we can write the individual speci®c biases in expression (1) as

kˆk‰A…ST†; g…RT†Š; …2†

whereST and RT are the symbolic and real contents of the schema nodes.

2.3. Attitudes

A politicianÕs evaluation depends on his predisposition to be more or less in favour of given concepts (nodes). Given the nature of taxes, negative (positive) attitudes towards a speci®c node translate into positive (negative) values ofk. From the political opposition node perspective, politicians expect the electorate to dislike taxes, and therefore they associate higher taxes with losses in votes. 5Taxation thus has negative symbolic value from an electoral perspective. As a result, politicians classify a given tax burden as higher than would have been the case if they would have had no electoral motivations.

From an ideological perspective, taxation has clear symbolic content as it is immediately linked to the (abstract) notion of government intervention. Unlike electoral objectives, ideological objectives have an individual speci®c impact. Whereas left-wing politicians are expected to consider government intervention a ``good'', right-wing politicians will consider it a ``bad''. Therefore we expect a negative relationship between politiciansÕ ideological position (IDEO) along a left±right scale and their attitude, and a positive relationship between ideology andk: left-wing politicians more likely classify a given tax burden as low than their right-wing colleagues.

4Some (left-wing) politicians may treat the

abstractconcept ``taxation'' as a good to the extent that it symbolises government intervention (see below). Even then, they are assumed to consider theactualtax burden as a bad.

5Note that this attitude exists irrespective of the nature ofactualopposition: the respondent ``just''

(7)

Like the intervention node, the ``self '' node generates individual speci®c attitudes. More precisely, a politicianÕs attitude is expected to depend on his involvement (INVOL). A respondent maximising his self-esteem de®nes situations in which he is a part as favourably as possible. We expect a positive relationship between involvement and attitude, and thus a negative rela-tionship between involvement andk. A respondentÕs involvement depends on his political position: a given tax burden is more likely classi®ed as low by a member of a party holding an absolute majority (SINVOL), than by a member of the opposition. Members of parties within a coalition (CINVOL) hold intermediate positions.

2.4. Appreciation

The appreciation of the relevant characteristics of the tax burden relates directly to the ``real world'' contents of the schema nodes (as opposed to the symbolic contents which underlie attitudes). For example, the political op-position node contains information on actual opposition. Politicians will classify a given level of the tax burden as higher to the extent that they believe that it induces more political opposition.

Relevant characteristics within the government intervention node are the

actual size of government intervention and the actual redistribution among

taxpayers or progression of the tax system. As with attitudes, appreciation is individual speci®c here. A larger intervention or redistribution will be wel-comed by one respondent, while another will disapprove. In general, we expect reactions to depend on a respondentÕs ideological position (left-wing politicians welcome positive changes in the degree of actual intervention or redistribution). Formally, this means that opinions on the level of the tax burden depend on the interaction of IDEO and the actual size and redis-tributive aspects of government intervention; the latter two empirically de-®ned in the next section.

3. Flemish politicians' opinions on the level of the tax burden

(8)

politiciansÕ``true'' opinions…O†on the one hand and a vector of explanatory variables (X) on the other:

O ˆb0X ‡e;

whereb0 is a vector of coecients and e is the error term. A number of ex-ogenous variables can readily be deduced from the hypotheses in Section 2. However, the analysis has in effect an unobserved dependent variable and crucial independent variables are also unobserved (more precisely politiciansÕ beliefs about the real content of the schema nodes).

Dependent variable: the ``true'' opinion of politicians…O†is unknown. What

we do have are the politiciansÕopinions as they are translated on the offered taxation opinion scale. Given the question posed in the survey, ®ve possible answers are observed and so the actual responseOis derived as follows:

OˆLowˆ0 if O60;

OˆRather lowˆ1 if 0<O6l1;

OˆAverageˆ2 if l1 <O6l2;

OˆRather highˆ3 if l2 <O6l3;

OˆHighˆ4 if l3 <O;

where the l's are unknown parameters which delineate the boundaries for the responses. As Greene (1993) states ``this is a form of censoring'', and the unknown parameters must be estimated with the parameters of the basic model.

Independent variables: PoliticiansÕ opinions depend on the actual level of

the per capita tax burden (TAX) but also on their attitudes and appreciation (which determinek). Their attitudes can be proxied by their ideological po-sition and their involvement with tax policy and so IDEO and INVOL readily enterXalong with TAX. Appreciation, however, is based on politi-ciansÕunobserved beliefsabout real world events. We assume that politicians make estimates based on information which is readily available. They can use the size of the tax burden as a measure of the actual size of government intervention and some measure of the progression inherent in the tax code (PROG) as a proxy for the distribution among taxpayers.

(9)

electoral outcomes. The tax choice literature identi®essourcesof opposition and shows that their impact depends on the form and organisation of op-position (Ashworth & Heyndels, 1997).

The ultimate source of political opposition is public opinion (PO) on taxation and its burden. This contains cognitive as well as a€ective elements. In general, public opinion re¯ects the publicÕs attitudes (PA) towards taxa-tion and its appreciataxa-tion ± h ± of the quantity (Q) and quality (QUAL) of government intervention. Public opinion determines ``latent'' opposition. This is translated into actual opposition with the help of interest group ac-tivity (IG). Thus, politiciansÕbeliefs on political opposition will be positively related to public discontent and to the activity of interest groups. Actual political opposition (OPPOS) can thus be written as

OPPOSOPPOSfPO‰PA;h…Q;QUAL†Š;IGg: …3†

Note that the information problem is not solved by expression (3). The publicÕs attitudes as well as its appreciation of actual government interven-tion cannot be observed. Neither can be the activity of interest groups. One more step is needed to identify observable determinants. Interest group ac-tivity is proxied by the number of active interest groups (NIG). The existing tax choice literature provides a large number of (observable) determinants of public opinion:

· PA: Public attitude is positively related to the publicÕs ideological position

on a left±right scale (ELEC). A more right-wing electorate will have a more negative attitude and will therefore more strongly oppose taxation. Thus, the model predicts that the electorateÕs ideological position a€ects politiciansÕ opinions: irrespective of their own ideological position, politi-cians will de®ne a given tax burden as higher to the extent that the elec-torate is more right-wing. Second, the publicÕs attitude towards taxation depends on past experience. As suggested by Peacock and Wiseman (1961), citizens use an historical perspective when evaluating their govern-ment. Higher taxes in the past increase political acceptance of higher bur-dens today. We expect opposition to be positively related to the recent evolution of the burden (TAXCH).

· h(.,QUAL): VotersÕ appreciation of the quality of governmentÕs

(10)

re-Table 1

Ordered probit estimation of politicians' opinions on tax burdena

Expected sign Linear Logarithmic

Intercept + ÿ4.207 ÿ0.001 4.314 8.113

(6.471) (1.460) (9.537) (4.723)

TAX + 0.643 0.567 10.605 7.262

(0.188) (0.108) (2.265) (1.658)

IDEO + ÿ0.129 0.047 0.232 0.047

(0.204) (0.022) (0.187) (0.022)

ELEC + 1.973 0.825 3.511 2.197

(1.268) (0.296) (1.793) (0.879)

EXPO ÿ ÿ0.064 ÿ0.067 ÿ0.062 ÿ0.067

(0.010) (0.008) (0.010) (0.008)

INC ÿ ÿ0.133 ÿ0.011 ÿ5.524 ÿ3.819

(0.017) (0.001) (0.630) (0.391)

FRAG + 1.798 0.973

(1.211) (1.091)

NEIGH ÿ ÿ0.032 ÿ0.026 ÿ0.209 ÿ0.270

(11)

lated to the share of debt expenditures (DEBT). In fact, debt expenditures could be seen as payments for services with zero present bene®t. Further, the ``value'' a citizen gets for his tax money is positively related to the ef-®ciency of public non-debt expenditures (EFFIC).

The ``money'' he pays for a given amount of public expenditures may be lowered by exporting taxes, that is by having non-residents pay part of the local tax burden. The larger the share of taxes that can be exported (EXPO) the lower is the e€ective tax price. The presence of grant ®nancing (GRAN) may have a similar e€ect. Under tax exportation or grant ®nancing, non-residents pay part of the ``money''. Thus, for any given set of government services, the perceived ``value for money'' will go up. Finally, the ``value for money'' a citizen gets is a€ected positively by the size of the local population (POP). This will be the case if the publicly provided services display some publicness: in the limiting case of pure public goods, the ``value'' of public services to any given citizen is not a€ected by the number of people using the service. So, if additional inhabitants contribute (by paying taxes) the value for money for initial citizens increases.

Table 1 (Continued)

Expected sign Linear Logarithmic

l1 0.690 0.671 0.722 0.689

(0.077) (0.073) (0.079) (0.075)

l2 2.239 2.190 2.344 2.245

(0.099) (0.097) (0.105) (0.099)

l3 3.324 3.248 3.460 3.322

(0.114) (0.111) (0.117) (0.113)

LL ÿ829.017 ÿ833.642 ÿ812.889 ÿ820.783

LL0 ÿ920.905 ÿ920.905 ÿ920.905 ÿ920.905

v2 183.774 174.524 216.032 200.242

Pseudo-R2 0.290 0.279 0.330 0.311

v2(R) 9.250 15.788

(13) (13)

q2 0.498 0.490 0.506 0.506

j 0.155 0.144 0.198 0.194

aNotes: Estimated standard errors are in parentheses; LL is the maximised value of the log-likelihood

function; LL0 the log-likelihood computed with only the constant term;v2the test of overall signi®cance

of the equation;v2…R†the test of the omission of variables from the general model withRthe number of

omitted variables; andq2the proportion of predictions that is correct; Pseudo-R2is the measure suggested

(12)

· h(Q,á): The level of government intervention can be proxied by the tax

burden. The publicÕs appreciation of this intervention depends ®rst of all on the publicÕs preferences. These can be proxied by the average income (INC) and by the ideological position of the population (which adds to the ``attitude effect'' of ELEC already discussed). At the same time, evaluation of government intervention is affected by the underlying cog-nitive process. The complexity of public intervention typically creates in-formational problems which are likely to persist as the individual has little incentives to inform himself (Downs, 1957). Systematic mispercep-tion of ®scal variables can result. The literature on ®scal illusion gives a number of sources of such misperception. 6 Most important are the complexity of the tax system and the role of grant ®nancing. The com-plexity hypothesis of ®scal illusion says that citizens underestimate the actual tax burden to the extent that the tax system is fragmented, that is: composed of many small taxes (FRAG). The illusionary role of grants is associated with the so-called ``¯ypaper e€ect'': the idea is that voters underestimate theirmarginaltax burden to the extent that expendi-tures are ®nanced more byunconditional grants (these lower theaverage

tax burden; the ®scal illusion explanation is that voters confuse average and marginal tax burdens). As we already showed that the mere presence of grants lowers political opposition, the ®scal illusion hypothesis suggests that this effect will be even stronger to the extent that the grants are uncon-ditional in nature.

Finally, even given correct information on the tax burden, voters may ®nd it hard to evaluate this. Therefore, the mimic literature suggests that the indi-vidual may use information on the tax burden in neighbouring jurisdictions (NEIGH) as a yardstick to evaluate the burden in their own jurisdiction.7

6

Oates (1988) and Dollery and Worthington (1996) provide reviews of the literature on ®scal illusion. The tax complexity hypothesis is typically associated with Wagner (1976). Heyndels and Smolders (1995) give a detailed discussion. The ¯ypaper e€ect refers generally to the empirical observation that unconditional grants coincide with larger levels of spending. Many theoretical explanations can be found in the literature. The ®scal illusion explanation goes back to Courant and Rubinfeld (1979) and Oates (1979).

7More generally, it is suggested that voters will compare with ``jurisdictions of reference'' which may or

(13)

4. Results

As all the above gives no indication of the appropriate functional form to be used, Table 1 gives results for two di€erent functional forms; a linear and a logarithmic representation of the independent variables.8 In each case an estimation of the complete model is provided, followed by results which apply various linear restrictions and remove insigni®cant variables. For ease of interpretation, the anticipated sign indicated in the previous sections is provided in column 2. As can be seen from the v2 statistics provided, the removal of the insigni®cant variables and the imposed restrictions are sup-ported by the data.

Overall, the explanatory power is in line with that anticipated from studies of this type as can be seen by the pseudo-R2 in the 0.28±0.33 region. A summary indicator for the predictive power of the model is provided byj. The value shows that there is an improvement in the prediction beyond that of using the most selected answer (``average'') for all respondents. However, this improvement, whilst signi®cant, is only moderate.9 The weight of evi-dence, certainly in terms of the likelihood function, favours the logarithmic speci®cation indicating that proportionate changes in the variables so de-lineated is the more likely manner in which the politicians view these variables when forming their opinions. This is consistent with the Weber± Fechner Law which suggests that people consider relative instead of absolute (price) di€erences (Thaler, 1980). 10

The internal predictions are more explicitly illustrated in Table 2. Here, the rows and columns give the actual and predicted responses respectively. For

8A description of the variables and the data sources is given in Appendix A. Only those variables where

taking a log is appropriate are transformed.

9

The statistic is a version of CohenÕs kappa, see Veall and Zimmermann (1996). Following Bishop, Feinberg and Holland (1975), the null hypothesis that the true value ofjis zero is rejected with a test statistic of 8.00. Further, the 95% con®dence interval forjis [0.132, 0.255] which re-inforces the view that, whilst signi®cant,jis not large.

10

(14)

example, of the 48 politicians responding that their tax burden was low (Oˆ0), the model predicts only six correctly. What is important, however, is that predictions which are incorrect fall in one of the neighbouring categories. This is illustrated by the zero-values in the distant off-diagonal cells.11 We can therefore conclude that the equations help explain the for-mation of opinions about tax burden.

Before turning to a discussion of the signi®cant variables in the restricted model, it is worth noting that only the growth in burden and population have unexpected signs. As regards population (POP), the e€ect is positive: the greater the population, the greater is the perceived burden. This is at odds with the theoretical prediction in the situation of scale economies. Our result suggests the existence of diseconomies of scale (crowding e€ects) but these are insigni®cant. The growth in the tax burden (TAXCH) a€ects opinions neg-atively, thus rejecting the idea of the presence of an historical benchmark for the evaluation of the present tax burden. All other variables a€ect opinions on tax burden in the anticipated manner. For grants (CGRAN and UG-RAN), interest groups (NIG), debt (DEBT), the degree of fragmentation (FRAG) and the interaction terms involving ideology the e€ect is, however, not signi®cant.12

Table 2

Internal predictions from the logarithmic model

Predicted Low Rather low Average Rather high High Total

Actual

Low 6 14 28 0 0 48

Rather low 8 13 55 0 0 76

Average 2 27 266 23 6 324

Rather high 0 0 119 39 14 172

High 0 0 22 19 21 62

Total 16 54 490 81 41 682

11

As the internal predictions for each speci®cation are broadly the same with a small improvement from the logarithmic model, Table 2 provides only the predictions from this model. Following Siegel and Castellan (1989), it is possible to calculate ac statistic which measures the relationship between two ordinally scaled variables. In this case,cˆ0.775, which is signi®cant at 1% with a test-statistic of 13.31. Again, this suggest that the model o€ers some prediction without being in any sense a good ®t.

12The linear functional form ®nds a more signi®cant role for grants which is not present in the

(15)

As expected, the tax burden (TAX) has a strongly signi®cant and positive e€ect. Higher tax burdens are in general classi®ed as higher, indicating the ``cognitive'' benchmark around which opinions vary (as in Fig. 1). Ac-counting for the interaction with the political colour of the electorate mod-i®es, but does not change, this interpretation. This can be seen in Table 3, which contains the marginal e€ects for the signi®cant variables.13

The politicianÕs ideological position (IDEO) is a ®rst signi®cant ``evalua-tion'' variable. We ®nd the expected positive e€ect: right-wing politicians view the burden to be higher ceteris paribus, re¯ecting a more negative predisposition towards government intervention.14Comparatively, the e€ect of a one unit change in ideological position to that of a 1000 BEF change in the tax burden suggests that the change in ideological position has around a quarter of the e€ect of the tax change.

The involvement of politicians (INVOL), indicating whether they are member of the ruling party or coalition leads them to classify a given tax burden as lower. This corresponds to the prediction that respondentsÕ

in-Table 3

Marginal e€ects of signi®cant variables in the logarithmic model

Burdenˆ0 Burdenˆ1 Burdenˆ2 Burdenˆ3 Burdenˆ4

TAX ÿ0.018 ÿ0.028 ÿ0.027 0.049 0.024

IDEO ÿ0.004 ÿ0.007 ÿ0.006 0.011 0.006

INVOL 0.056 0.081 0.051 ÿ0.128 ÿ0.060

ELEC 0.019 0.032 0.029 ÿ0.053 ÿ0.026

EXPO 0.006 0.009 0.009 ÿ0.016 ÿ0.008

INC 0.001 0.002 0.002 ÿ0.003 ÿ0.002

NEIGH 0.002 0.004 0.003 ÿ0.006 ÿ0.003

13

The marginal e€ects in Table 3 re¯ect the in¯uence of a 1 unit change in TAX (where 1 unit corresponds to 1000 BEF ± see Appendix A) on the predicted probabilities of each answer occurring. For example, a 1000 BEF rise in the (average) tax burden would increase the probabilities of answers ``3'' and ``4'' while decreasing the probabilities of the other answers (the marginal e€ects all sum to zero from the requirement that the probabilities sum to unity). It should be remembered that, whilst the size of the e€ect is a matter of interest, the signi®cance, i.e. whether an e€ect exists is taken as being more important than the actual size of that e€ect. The aim of our analysis is to establish the salient features in the politiciansÕ

opinion formation. If a variable is signi®cant, it is important and cannot be omitted without bias, no matter how small the e€ect.

14There is a considerable degree of multicollineaity in the full equations, particularly relating to the role

(16)

volvement leads them to have more positive attitudes. Although the results of the full model give some evidence of greater involvement if sole power (SINVOL), as opposed to power in a coalition (CINVOL), exists, it is pos-sible to impose the restriction that the e€ect of power is the same no matter how exercised (INVOL). Thus, whether an incumbent politician has to share the political power or not with coalition members does not a€ect his opinions. In that sense, politicians within a coalition have homogenous opinions (which is especially interesting given that the survey was organised a few months prior to elections). It is also interesting to note in Table 3 that the e€ect of involvement (INVOL equal to 1 instead of 0) is approximately the same size as the full range of the ideological spectrum.

As for the politiciansÕ attitudes, we ®nd that the electorateÕs attitude, as proxied by its ideological position (ELEC) has the expected signi®cant e€ect. This is an important result as it indicates that what politicians think about the tax level depends also on what the electorate thinks about it. Irrespective of his own ideological position, a politician classi®es a given tax burden as higher to the extent that he is confronted with a more right-wing electorate. This is clear indication that politicians, when forming an opinion, are in-¯uenced by the position of the local voters. It reveals the presence of electoral considerations and motives. Whereas the ideological position of the elec-torate has the expected e€ect, the interaction of this with the burden has a negative sign despite the anticipation of a positive e€ect. The results indicate that there will be more ± not less ± political opposition towards a given tax burden to the extent that the electorate is more left-wing. It can be seen from the marginal e€ects that this is the dominant e€ect.

(17)

politicians consider a given tax burden as higher to the extent that it exceeds the level in neighbouring municipalities. This may lead to mimicking be-haviour (as is indeed found to exist among Belgian municipalities in Heyndels & Vuchelen, 1998). As can be seen from Table 3, the marginal effect of a unit change in neighboursÕ burden has about an eighth of the effect of the same change in the municipalityÕs own burden.

5. Summary and conclusions

PoliticiansÕ behaviour depends on how they perceive the world. In this paper, we illustrate how schema theory can link current behavioural as-sumptions from the tax choice literature to how politicians de®ne tax issues. The general idea is illustrated through the analysis of how politicians de®ne the level of the tax burden in their jurisdiction.

We hypothesise a triple objective function in which the politician cares about electoral, ideological and self-esteem motives. A politicianÕs tax burden schema summarises what the tax burden ``truly means'' for him and how he evaluates its di€erent aspects. An ordered probit analysis of a sample of survey responses of 682 Flemish politicians who were active at the municipal level of government, shows that di€erences in opinions can be attributed to evaluative components of the politicianÕs tax burden schema. This evaluation takes place at two distinct conceptual levels. First, at a symbolic level, pol-iticiansÕattitudes toward the tax burden a€ect how they de®ne its level. More precisely, it is observed that the politicianÕs ideological position and his in-volvement with tax policy a€ect his opinion. Second, at a ``real'' level, pol-iticiansÕappreciation of relevant characteristics plays a role. We ®nd evidence that characteristics of the tax burden relating to the respondentÕs ideological and electoral objectives a€ect his de®nition of the tax burden.

(18)

Table 4

data de®nitions and sources

Name of variable

De®nition Sources Average value

TAX Actual per capita tax burden in the municipality (in 1994) in thousands BEF Ministry Flemish Community

12.23

IDEO Respondents located themselves on a leftÿright scale with values from 0 for extreme left up to 10 for extreme righta

Own survey 4.17

INVOL Dummy that takes the value 1 (0) if the respondentÕs party is (not) a member of the ruling group. This variable is separated into dummies capturing sole or coalition power (SINVOL and CINVOL)

Ysebaert and Asselberghs (1994)

0.37

PROG Progression of the tax system calculated as the percentage share of income taxes in the combined revenue from income and property taxes (these taxes together account for 80% of local tax revenues on average, the former is far more progressive)

Ministry Flemish Community

54.31

NIG (potential) in¯uence of organised interest groups. Respondents were asked if various groups had contacted their party in order to in¯uence the political programme or the list of candidates for the 1994 local elections. NIG gives the number of interest groups which contacted the respondentÕs party

Own survey 1.41

ELEC ElectorateÕs ideological position, calculated as a weighted average of the votes at the (most recent) 1988 election with the weights being the average value party members gave to our ideology-questionb

Ysebaert and Asselberghs (1994) and own survey

4.85

TAXCH Change in tax burden between 1987 and 1994 in thousands BEF Ministry Flemish Community

2.55

DEBT Level of per capita debt in the municipality (budget data) in thousands BEF Ministry Flemish Community

6.39

EFFIC Eciency of government non-debt expenditures. Calculated as the complement of the measure of government ineciency as given for the Flemish

municipalities (based on FDH analysis)

De Borger and Kerstens (1993)

0.93

EXPO Indicator for the share of taxes paid by residents (i.e., ®rms and non-resident house-owners) (budget data).c Expressed as a percentage

Ministry Flemish Community

10.36

GRAN Share of grants in municipal revenues (budget data). Also sub-divided into conditional (CGRAN) and unconditional grants (UGRAN)

Ministry Flemish Community

0.53 (0.37 & 0.16)

(19)

Name of variable

De®nition Sources Average value

INC Average income within the municipality in thousands BEF National Institute of Statistics

305.41

FRAG Fragmentation of the tax system calculated using a standard Her®ndahl index (this index is given by FRAGˆPir…i†

2

, wherer…i†is the share of tax revenue coming from tax sourcei†, as is standard practice in the literature on ®scal illusion (Wagner, 1976). The greater the fragmentation the smaller the value

Ministry Flemish Community

0.36

NEIGH Average per capita tax burden in neighbouring municipalities in thousands BEF

Ministry Flemish Community

12.08

aThis approach has the advantage that it is directly aimed at the respondent

Õs actual position. However, it is likely that individualsÕhave di€erent conceptions about the scale itself. Di€erences in positions along the left±right axe may then re¯ect di€erences in these conceptions. Therefore, in a second stage we chose an alternative measure for the ideological position. This measure is based on the respondentÕs party aliation. Each party was located along the left±right scale on the basis of the average value its members gave to our ideology-question. The underlying assumption is that on average respondents from any given party have the correct conception of our left±right scale and that within-party di€erences in ideological position can be attributed totally to di€erences in conceptions of the scale itself. Use of this scale did not materially a€ect the results hence only one set of estimations is recorded.

b

As a substitute, the ideological position of the electorate was approximated by the share of socialist votes in the 1988 elections, which again did not have any signi®cant e€ect on the results.

c

The de®nition used in the text implies that local income and property taxes are assumed to be paid by residents. Especially with respect to the property tax, alternative assumptions are feasible. To account for the possibility that renters shift the tax to home-owners who live outside the municipality, we added to EXPO a proportion of the property tax corresponding with the proportion of renters in the population. This gave similar results to those in the text and all signi®cant variables remained unchanged. A still stronger assumption is that property taxes are fully capitalised into house values and are therefore ``sunk costs'' for all taxpayers. Under this very extreme assumption all property taxes could be considered to be ``exported'' in the long run. Use of this most extreme option increased the signi®cance of FRAG in the preferred logarithmic model and GRAN in the linear model with all other inferences remaining the same.

(20)

Acknowledgements

We thank Marc Deloof, Harold Hochman, Veronique Weets and two

anonymous referees for useful suggestions. Earlier versions of this paper were presented at seminars at the Institute for Empirical Research, Univer-sity of Zurich and at the UniverUniver-sity of Science and Technology, Dragvoll (Norway). We thank Iris Bohnet, Lars-Erik Borge, Isabelle Busenhart, Jurg

de Spindler, Reiner Eichenberger, Bruno Frey and Jùrn Rattsù for valuable and stimulating comments. The British Council and the Fonds voor Weten-schappelijk Onderzoek Vlaanderen are acknowledged for their ®nancial support.

Appendix A. Data de®nition and sources

See Table 4.

References

Aaron, H. J. (1994). Distinguished lecture on economics in government. Public policy, values and consciousness.Journal of Economic Perspectives,8, 3±21.

Aldrich, J. H., & Nelson, F. D. (1984).Linear probability,logit and probit models. Beverley Hills, CA: Sage. Alm, J., McKee, M., & Skidmore, M. (1993). Fiscal pressure, tax competition and the introduction of state

lotteries.National Tax Journal,46, 463±476.

Ashworth, J. S., & Heyndels, B. (1997). PoliticiansÕpreferences on local tax rates: An empirical analysis. European Journal of Political Economy,13, 479±502.

Besley, T., & Case, A. (1995). Incumbent behaviour, vote seeking, tax setting and yardstick competition. American Economic Review,85, 25±45.

Bishop, Y. M., Feinberg, S. E., & Holland, P. W. (1975). Discrete multivariate analysis: Theory and practice. Cambridge, MA: MIT Press.

Boudon, R. (1996). TheÔcognitivist modelÕ± a generalizedÔrational-choice modelÕ.Rationality and Society, 8, 123±150.

Case, A. (1993). Interstate tax after TRA86.Journal of Policy Analysis and Management,12, 136±148. Conlisk, J. (1996). Why bounded rationality.Journal of Economic Literature,34, 669±700.

Courant, P. G., & Rubinfeld, D., (1979). The stimulative e€ects of intergovernmental grants: Or why money sticks where it hits. In: P. Mieszkowski, W. Oakland,Fiscal federalism and grants-in-aid(pp. 5±21). Washington, DC: The Urban Institute.

Davidson, R., & McKinnon, J. G. (1993).Estimation and inference in econometrics. Oxford: Oxford University Press.

De Borger, B., & Kerstens, K. (1993). De technische ecientie in de gemeentelijke voorzieningen. Economisch en Sociaal Tijdschrift,4, 665±690.

(21)

Fiske, S. T., & Taylor, S. E. (1991).Social cognition. New York: McGraw-Hill.

Genser, B., & Weck-Hannemann, H. (1993).Fuel taxation in EC countries: A politico-economy approach, preliminary version, University of Konstanz.

Greene, W. (1993).Econometric analysis(2nd ed.). New York: Macmillan.

Hettich, W., & Winer, S. (1999).Democratic choice and taxation ± a theoretical and empirical analysis. Cambridge: Cambridge University Press.

Heyndels, B., & Smolders, C. (1995). Tax complexity and ®scal illusion.Public Choice,65, 127±141. Heyndels, B., & Vuchelen, J. (1998). Tax mimicking among Belgian municipalities.National Tax Journal,

51, 89±102.

Ladd, H. F. (1992). Mimicking of local tax burdens among neighbouring countries.Public Finance Quarterly,20, 450±467.

Lewis, A. (1982).The psychology of taxation. Oxford: Martin Robertson.

Lodge, & McGraw (1991). Where is the Schema? Critiques.American Political Science Review 85, 1357± 1364.

McElvey, W., & Zavoina, R. (1975). A statistical model for the analysis of ordinal level dependent variables.Journal of Mathematical Sociology,4, 103±120.

Murphy, A. (1996). Simple LM tests of mis-speci®cation for ordered logit models.Economics Letters,52, 137±141.

Oates, W. (1988). On the nature and measurement of ®scal illusion: A survey. In G. Brennan et al., Taxation and ®scal federalism: Essays in honour of Russell Matthews(pp. 65±88). Sydney: Australian University Press.

Peacock, A., & Wiseman, J. (1961).The growth of public expenditure in the United Kingdom. Princeton, NJ: Princeton University Press.

Pommerehne, W. W., & Schneider, F. (1983). Does government in a representative democracy follow a majority of votersÕpreferences? An empirical examination. In H. Hanusch,Anatomy of government de®ciencies(pp. 61±88). Heidelberg: Springer.

Simon, H. A. (1985). Human nature in politics: The dialogue of psychology with political science. American Political Science Review,79, 293±304.

Siegel, S., & Castellan, J. (1989).Non-parametric statistics for the behavioural sciences. London: McGraw-Hill.

Thaler, R. (1980). Toward a positive theory of consumer choice. Journal of Economic Behavior and Organization,1, 39±60.

Veall, M. R., & Zimmermann, K. F. (1996). Pseudo-R2 measures for some common limited dependent variable models.Journal of Economic Surveys,10, 241±259.

Referensi

Dokumen terkait

Judul Tesis : HUBUNGAN ANTARA KOMBINASI HEMODIALISIS / HEMOPERFUSI DENGAN TOTAL BODY WATER (TBW), EKSTRA CELLULAR WATER (ECW) DAN INTRA CELLULAR WATER (ICW) PADA

Kedua: Membaca penting karena dapat memberikan pencerahan baru pada pemikiran

Pada Aspek Metode Resolusi Konflik, Gaya Manajemen Konflik Situasi Konflik dan Reaktualisasi Pendidikan Karakter .... 185 Tabel 4.3 Situasi Konflik yang Terjadi di dalam

“ HUBUNGAN ANTARA KOMBINASI HEMODIALISIS (HD) / HEMOPERFUSI (HP) DENGAN TOTAL BODY WATER (TBW), EXTRA CELLULAR WATER (ECW) DAN INTRAi. CELLULAR WATER (ICW) PASIEN HEMODIALISIS

Tujuan dari penelitian ini adalah untuk mengetahui hubungan antara kadar prealbumin dengan status nutrisi pada anak sakit kritis.. Mengetahui status nutrisi pada anak

UPAYA PENGASUH DALAM MEMBERIKAN LAYANAN PENGASUHAN PADA BALITA DI TEMPAT PENITIPAN ANAK (TPA) SE-KECAMATAN COBLONG KOTA BANDUNG.. Universitas Pendidikan Indonesia |

Respon stres yang terjadi pada anak sakit kritis dapat. menyebabkan malnutrisi karena proses katabolisme yang

Berdasarkan kondisi tersebut peneliti tertarik untuk menelaah secara lebih mendalam terhadap upaya yang dilakukan pengasuh di TPA yang ada di Kecamatan Coblong