Resumé

Over the past couple of decades, the level of turnout in Norwegian local elections has shown a steady decline. This decline is paralleled by a greater range of variation in turnout across Norwegian municipalities. Arguing from the perspective of rational utility-maximizing voters, the article examines to what extent such variations in the level of turnout may be accounted for by the policy performance of local authorities - in addition to certain structural features of the municipalities and local communities. Using aggregate data on turnout and measures of policy performance, the analysis testifies to the suggestion that policy exerts an influence in determining the level of turnout, indicating a turn in the direction of more rational voting behaviour as far as local elections are concerned. It is also evident from the analysis that the appearance of smaller single-issue parties does have a mobilizing effect in local elections, while the Downsian suggestion about the effect of party competition on turnout levels is not supported by the results of the analysis.

Introduction

Over the past thirty years the rate of turnout in Norwegian local elections has shown a steady decline, from a peak level of 81 percent in 1963 to 66 percent in 1991. This decrease is contrasted by a high and slightly increasing turnout in Norwegian general elections. Thus, while the rate of turnout in general elections was 79.1 percent in 1961, it increased to 83.2 percent in 1989. Even although the turnout in local elections still is of a considerable magnitude compared with that of other Western democracies, the downward trend in the turnout has raised some concern about the future viability of local democracy, particularly when considered in relation to the growth in local government activities that has taken place over this same period (Hansen & Sørensen 1988). Does this decline signify a weakened legitimacy and support of local government - or are we experiencing fundamental changes in the way in which local democracy works?

As far as the question of the legitimacy of local government is concerned, other data sources seem to point to the opposite development of that indicated by the turnout figures. In connection with the Norwegian Election Research Programme, voters have, over a number of years, been asked about their interest in local politics as compared to national and international

politics. According to these surveys the proportion of respondents who say that they are primarily interested in local political issues has been going up - from 38 percent in 1971 to 48 percent in 1987. Similarly, a broadly based survey of "the state of the arts" as regards Norwegian municipalities and local politics brings further evidence about the high level of popular support of and interest in local government affairs. According to this survey, of those respondents who wanted to change the distribution of tasks and functions between central, regional and local government (about half of the respondents wanted such changes), 54 percent wanted more tasks and functions to be transferred to the municipalities. (Baldersheim et al. 1990).

We are, in other words, confronted with two opposite trends with respect to the popular support of local government; on the one hand a decrease in the turnout and on the other hand an increased popular interest in and support of local political matters. This last trend is also more in tune with the actual development that has taken place in terms of municipal service provision, where the level of spending as a proportion of GNP has increased from 11 percent in 1960 to about 18 percent in 1990.

Parallel to this decline in electoral participation - if not a consequence thereof - is a frequent change in the composition of local councils. First, in every one of the past four municipal elections as many as 60 percent of the representatives of the councils either refused to stand for another period or were not re-elected (Larsen & Offerdal 1990). Second, this rather turbulent picture of local politics is further strengthened by more frequent turnovers in the political composition of the councils. After the local elections in 1991, as many as 40 percent of the municipalities experienced a change in the composition of the political majority in the councils. Furthermore, this change is - by and large - characterized by a weakening of the Labour and the Conservative Parties, to the advantage of the smaller bourgeois parties, particularly the Centre Party (the Agrarians).

The developments briefly described here raise several questions about the underlying causal factors as well as about the prospects for local democracy. In this paper I will, however, limit myself to an analysis of the level of turnout in local elections. In particular I will devote my attention to the relationship between voting and local government performance. Before entering on a further and more detailed discussion of this question, some theoretical considerations related to the logic of voting in political elections may be of relevance as a background and basis for the formulation of models of analysis to be employed in this study.

A Theoretical Framework

Basically, the general approach to be adopted in this analysis is a utilitarian

one, where the expected benefits to be derived from the outcome of an election are assumed to influence voting and the voting pattern. This perspective is rooted in the theoretical framework suggested by Anthony Downs (1957), who argues out from the basic axiom that citizens act rationally in politics in the sense that "... each citizen casts his vote for the party he believes will provide him with more benefits than any other" (Downs 1957, 36.)

On the basis of this axiom, Downs develops his reasoning on the logic of voting around the concept of expected party differential, which is calculated as the difference in utility to the voter between the parties (or candidates) running for election. The voter has to multiply this differential by the probability that his/her vote will make a difference for the outcome of the election - in the sense that the vote will tilt the balance in favour of his or her preferred party - in order to calculate the expected individual gross benefit from voting. Apart from depending on the number of eligible voters, this probability may also depend on how close the competing parties are in terms of electoral support - a point to which I will return later. Assuming that the voters have clear party preferences and that they are able to calculate their expected party differential, the cost of voting has to be included in this calculation. According to Downs, "time is the principal cost of voting" (Downs 1957, 265) and this includes time spent in becoming informed about the various party political platforms, time spent to deliberate - as well as time spent to go to the polls. By taking such costs into consideration, an individual's net utility from voting may algebraically be expressed by the following equation:

(1)

where U is the net utility accruing to the individual voter, B is the benefit - defined as party differential - that a voter receives from success of his/her party, P is the probability that the individual's vote will make a difference to the outcome of the election, C is the cost of voting.

Only where U > 0 will it be rational for an individual to cast his vote in the election. However, as implied by Downs's discussion, this calculus will normally lead to a negative value of U, which is to say that the gross utility to the individual voter (BP) in most circumstances will be less than the cost of voting (C). In that case, we are led to the conclusion that voting in elections is a typically irrational act - given this rather narrow utilitarian interpretation of voting and the individual motivation to vote.

Why this is so may be seen if we consider each component on the righthand side of the equation in some more detail. First, B is defined not as the total benefits that accrue to the voter if his party gains the majority, but as the difference in benefits between his preferred party and the

opposition. Apart from those voters who strongly identify with a particular party - and according to the Norwegian Election Research Programme this applies only to (or to as many as) some 30 percent of the electorate - to most voters this party differential will be of a rather moderate magnitude. In fact, according to the previously mentioned study by Baldersheim et al., as many as 40 percent of the respondents reported that they did not see any significant differences between the parties represented on "their" local councils; and they are thus indifferent to which party gains power. By definition, this yields a party differential of zero. Even if the "objective" differences between the parties, as e.g. measured by the size of budget proposals and the actual performance as to the level of service provision, may be significant, the voters may perceive such differences as being much less than they actually are. Conversely, hard-core party supporters will tend to magnify minor differences in policies between the competing parties. The point is that B has to be regarded as the voter's marginal utility from securing a majority for his preferred party rather than the opposition party.

More important for the magnitude of the first term on the right-hand size of this equation is the probability that the vote will make any difference to the outcome of the election (P). If the number of eligible voters in the electorate is V, P is defined as

(2)

which even in local elections and in smaller municipalities will be quite low. However, this probability may be - as suggested before - conditioned by how close the competing parties are as regards levels of electoral support. The probability of having a say in the election outcome increases, the closer the parties are in terms of electoral support. But also in that case the expected benefits of voting - defined by the term PB - will almost always and to most voters (if we are to believe the reported responses to the surveys) be infinitesimal, even at the local level where the probability that your vote will make a difference in any case is quite high compared to national elections.

The crucial factor in equation (1) is the cost of voting - the way it is defined and operationalized. An example may illustrate how even very minor costs can lead to a negative utility from voting. Let us for the sake of simplicity assume that the party differential can be measured in monetary terms, and as the difference in per capita spending between the voter's preferred party and the opposition parties in a particular municipality. In an average Norwegian municipality this figure could be estimated to about NOK 300 on an annual basis, corresponding to a total of NOK 1200 over a four-year election period. Let us furthermore assume that the

municipal electorate counts 4000 voters. We then arrive at the following
expected gross benefit to the voter:

(3)

If we now assume that the cost of voting is limited to the bus fare to and from the polling station - say NOK 30 altogether - we see that the costs are 100 times higher than the expected benefit, which leaves us with a negative net utility from voting. Even if we assume that party competition in this particular municipality is such that the probability of having a decisive vote increases ten times (which is to say that the "last 10 percent" of the voters are decisive in determining which party is getting the majority of the votes), we are still left with a negative net utility from voting. In this connection it should also be noted that a high degree of interparty competition, according to Downs, tends to decrease the distance between the competing parties in terms of policies, thereby contributing towards reducing the party differential for most voters. On the individual level we are, however, still left with the problem of how any individual voter can assume that his vote carries a high probability of being decisive in determining the outcome of the election. Or as B. Barry asks:

Can we really imagine anyone for whom the product of party differential and probability
of swinging the election could be other than infinitesimal? (1970, 18).

We are thus confronted with the paradox that even if the act of voting has a negative expected utility to most voters, the fact remains that the level of turnout in general and local elections is very high. In fact, those cases which will yield a positive utility are so extreme compared to most real situations that they are hardly worth considering empirically. To get a positive U in our example under the condition that the probability and costs are the same, B would have to exceed 120,000, which is far beyond any imagination within the framework of Norwegian local politics. The question is therefore whether this approach to the study of voting behaviour is worth considering at all. In their discussion of this calculus, Riker and Ordeshook state that:

We find this conflict between purpose and conclusion bizarre but not nearly so bizarre as a non-explanatory theory: The function of theory is to explain behaviour and it is certainly no explanation to assign a sizeable part of politics to the mysterious and inexplicable world of the irrational (1968, 25).

To get around the problems posed by Downs's formulation, Riker and Ordeshook embark on a reinterpretation of the individual variables of the model - especially the notion of costs incurred by the individual when voting. In particular, they point to several positive points of satisfaction to

the individual that follow from voting, such as compliance with the ethic of voting, satisfaction from affirming allegiance to the political system and from affirming a partisan preference, the social satisfaction from attending the polling station as well as the satisfaction of affirming one's efficacy in the political system. By taking these factors into consideration, they suggest that equation (1) has to be reformulated into

(4)

where D represents some aggregate of the various satisfactions from voting
(Riker & Ordeshook 1968, 28).

This formulation has been questioned particularly by Ferejohn & Fiorina (1974), who argue that while the original Downsian formulation treated the voting decision as an investment decision, the addition of the D term implies a change of the model to one of consumption decision. In the latter perspective one should investigate the structure of individual preferences. While Ferejohn and Fiorina's interpretation of the D term is still kept within a utilitarian perspective, I will argue that the problem of including this term - which in fact may be labelled "civic duty" - into the model, is that we in a sense are leaving a utilitarian model of explanation in favour of a deontological perspective based either on the norm of fairness or everyday Kantianism (Elster 1989). According to Jon Elster, neither of these two motivations - civic duty and fairness - is outcome-oriented, and for those subscribing to such norms the notion of a party differential upon which the Downsian perspective rests is irrelevant to the act of voting.

So far the implicit assumption has been that all voters act as utilitymaximizers in deciding whether to vote or not. We may, however - as Elster suggests - differentiate between three different motivations to vote: a utilitarian norm, a Kantian norm and a norm of fairness. While the everyday Kantians belong to the hard-core voters - but not necessarily the hard-core party identifiers - the decision to vote among those motivated by the norm of fairness depends on how other potential voters behave. According to Elster, "the norm of fairness enjoins us to follow the majority, whatever it is doing" (1989, 188). This implies that in a situation of a declining turnout, those subscribing to this norm may be more inclined to abstain from voting. Over time this bandwagon effect is self-reinforcing, and may well account for part of the long-term decrease in the turnout that we have experienced in Norwegian local elections. But considering that a majority of the voters do in fact vote, those inspired by the notion of fairness should be inclined to take part in the elections. The final group consists of utilitarians motivated by the Downsian logic. As B. Barry has pointed out:

. . . a Downsian type of analysis, reasonably enough one might think, applies much better
when a low sense of "citizen duty" gives it a chance to come through (1970, 18).

This implies that, for these voters, we are back to the initial formulation of the decision model, as suggested by equation (1). However, to overcome the problems related to a cost-benefit ratio below unity, I will - as suggested by Riker and Ordeshook - assume that the expected net utility from voting is positive, i.e. U > 0, for those who actually voted in the election. This means that we may simplify- perhaps to the point of over-simplification - our analytical problem by concentrating our attention on the effects of outcomes or policies on the pattern of voting. Before embarking on a more precise formulation of the analytical model to be employed in this study, some methodological issues and problems need to be discussed - if not resolved.

A Note on Methodology

The data analysed in this article are of the aggregate level variety. The constituency - in this case the municipality - and not the individual voter serves as our unit of analysis. This presents us with the problem of whether the approach discussed in the previous section, applying to individual voters, may be given a formulation and an interpretation which applies to an aggregate of voters. Will I, in fact, make a contextual fallacy - as opposed to an ecological fallacy - by assuming that factors working in explaining voting behaviour on the individual level may be transformed into a model applying to the level of a whole constituency? And vice versa, will the results of an aggregate analysis enable us to make any inferences as to the individual motivations for taking part in elections? If we are not going to make substantial modifications of the assumptions underlying the Downsian logic of voting, this model may not be directly applicable to the level of whole constituencies - or, more generally, to groups of voters. To me, therefore, the importance and strength of this perspective does not lie primarily in the logic of calculating individual utilities as a basis for making decisions on voting behaviour, but rather in its focus on the party policy as the basis for the voter's decision to vote or not to vote for a particular party. But this is, of course, the same as modifying some of the assumptions of the model.

The impact of policies on electoral behaviour has developed into a key question for electoral research with its traditional emphasis on issue- versus class-voting. By and large, this research may be separated into two major categories, (1) research using aggregate voting data/statistics, relating changes in voting patterns to economic and social conditions and changes in society and (2) survey data focusing on individual behaviour, and how this behaviour relates to various political issues (Budge & Farlie 1983). As a rule, those relying on aggregate data have focused on the way in which

various economic changes in society make an impact on the pattern of voting, while the survey data analysts have been more concerned with discrete policy issues and the way in which people's attitudes to such issues influence their party preferences. Furthermore, the aggregate level analyses are normally based on so-called objective statistical indicators of actual developments along certain dimensions of public policies or economic conditions in society - such as unemployment and inflation - whereas the measurement of issues in the survey studies is based on the repondents' attitudes to various specific issues or problems. If these two approaches are considered in relation to the Downsian model, it is the survey method which comes closest to that way of modelling voting behaviour. The differences between these two approaches - even if they are clearly related - illustrate some of the dilemmas of assuming that a model which is formulated to explain individual behaviour may also be adapted to the aggregate level. An illustration may serve to exemplify this dilemma. Let us assume that increased unemployment has contributed towards weakening the electoral support of the incumbent party, while we - on the individual level - in faet find a positive relationship between unemployment and voting for the incumbent party. The point is that there is nothing logically wrong in arriving at two such disparate conclusions. The first result may simply be produced by the faet that those who are employed - perhaps out of fear of becoming unemployed - realign and vote for the opposition party, while the unemployed may still have more confidence in the incumbent party. Thus, even if the two approaches to issue-voting are related, this relationship applies only to the emphasis given to issues or policies in explaining voting, and not to the logic of voting as such.

A major modification of the model to be employed here as compared to the individually based Downsian model is that voting in this study is to be regarded as a characteristic of the constituency as such. This approach is quite similar to the one adopted by Alford & Lee (1968) in their study of voting turnout in American cities, where the city and not the individual voter is the unit of analysis. However, while Alford and Lee use various structural features of the cities - such as political structure, social structure, other community characteristics as well as regional location - to explain turnout, I will mainly concentrate on the effects local policies, as measured by various output indicators, may have on the pattern of voting. Before presenting the model of analysis to be employed, some comments on Norwegian local policies and performance may be useful.

Local Policy Performance

Over the past three decades, the Norwegian municipalities have experienced

a substantial growth in their service provision and their budgets. Apart from an overall financial expansion, a major equalization of financial resources across the municipalities has taken place over this period, a development from which the smaller and most peripheral municipalities in particular have benefited. This is clearly demonstrated in Table 1, which displays some major indicators of the distribution of financial capabilities across variously sized municipalities.

Given the fact that the smallest municipalities suffered from a rather slim resource situation, and consequently a poor service provision, during the first two decades after World War 11, the data (see Table 1) clearly testify to the equalization of financial capabilities which has taken place over recent years. The fact that this development also has been part of a deliberate and explicit national policy of redistribution is readily apparent from the data on central government transfers to the municipalities. The smallest municipalities receive more than twice as much per capita than the larger ones, despite the fact that the per capita level of local taxes (income taxes) does not differ that much between the five groups of municipalities. It might be added that central government grants or transfers on the average amount to almost 40 percent of total local government revenues - and of course relatively more for the smallest municipalities, which accounts for the significantly higher level of expenditures in these as compared to larger municipalities. The fact that the smaller municipalities maintain a higher level of spending than the other municipalities should not, however, be taken as evidence of a higher level of actual service

provision. For most local government activities scale economies operate on the costs of service provision, such that the smallest municipalities suffer from disadvantages of scale, implying relatively high average cost per unit of the services produced and provided (Hansen & Sørensen 1985). This contributes towards reducing the actual differences in the level of service provision between smaller and larger municipalities.

The local authorities operate within relatively strict limits as to the total level of spending. This is largely a byproduct of the revenue system. The most important revenue source is the local income tax system, which gives the municipal authorities the freedom to choose the tax rate, but within limits fixed by the state. The maximum tax rate for Norwegian primary municipalities is 13.5 percent of taxable income, and this limit has for some time been reached by all Norwegian municipalities. Thus, in practice there are no possibilities for local authorities to increase their tax revenues. Furthermore, the level of government grants to be transferred to the individual municipalities is also determined by the central authorities - and according to a formula based on so-called objective criteria. Since taxes and grants together constitute more than 80 percent of total local government revenues, the local authorities themselves enjoy restricted autonomy in regulating the total level of activity as measured by financial resources. It is only with respect to the levy of fees and charges that local councils may exert some discretion, and this is also an area which increasingly has become a matter of local political conflicts, conflicts which normally follow a traditional left-right dimension. At the core of this conflict is the question to what extent the local authorities should replace a redistributional tax system in favour of a pricing system which does not take account of the individual consumers' purchasing power. Since the political visibility of fees and charges is rather high, I will expect such issues to become of increasing importance in connection with the voters' choice of what party to support in the elections. In fact, the high visibility and high political sensitivity of fees and charges suggest that the general information problem on the part of the voters inherent in the rational model of voting is of less importance at the local than at the national level. This is also suggested by G. Tullock who, in commenting on the information problems, argues that

Most of the models are simple, direct democratic systems, in which the tax cost of each measure is weighted against the benefit. Clearly, such situations do exist in the real world, particularly at the local level. Since it is very easy to understand the tax cost in such situations, and the benefit of each proposed measure is also easily understood by its beneficiaries, the perfect information model is not a violent distortion of the real world (1967, 106).

Even if total revenues - and thereby the total level of spending - are more
or less fixed by central authorities, local councils enjoy some autonomy as
to the way in which their financial resources are being spent across various

services. This also applies to services circumscribed by national legislation, such as the regulation of primary schools, health services and certain other social welfare services. Even if such legislation tends to be rather detailed in defining the responsibilities of local government as to what services to provide, local authorities are still left with some discretion both in determining the quantities as well as the quality of the services to be provided. In this sense, room is left for local political parties to appropriate the financial resources according to their own priorities. Several studies of spending decisions in Norwegian municipalities have clearly demonstrated that "politics matter", although the way in which ideological differences or different party political platforms make an impact on spending decisions may be more subtle than is usually assumed (Hansen & Kjellberg 1976; Hansen et al. 1988; Hansen 1981). However, it is on the disaggregate level that such party political differences make an impact on the spending decisions. On the aggregate level of spending such political effects are absent. Thus, by looking at the relationship between total municipal expenditure per capita in 1990 and the political composition of the local councils - measured as a proportion of socialist representatives on the council in that election period - we find a correlation close to zero. This is a strong indication to the effect that local councils have to treat the total level of resources within which they may operate as exogenously determined. The discretion left to the local councils to set their own priorities applies only with respect to the relative distribution of expenditure across various areas of service provision. The fact that we indeed observe certain differences in the way political parties choose to give priority to various services in their appropriation of resources, should therefore not make us oblivious of the fact that such differences operate only marginally. The fact remains that Norwegian municipalities have become increasingly similar both as to the level of expenditure and as to the way in which resources are allocated across various types of services, and that the variations among them may be accounted for by differences in factors relating to needs and production costs. This is also a reflection of the role attributed to local government by central authorities in the development of the Norwegian welfare state, namely that of securing universal access - on a spatial basis - of basic welfare services (Hansen 1991). To a certain extent, this also implies a depoliticization of local government affairs - at least when seen in relation to traditional party political cleavages, a point to which I will return later.

Let me, finally - before turning to a presentation of the model of analysis - make some additional comments on the basic assumption implicit in the Downsian logic of voting, namely that the voters actually are informed about the policies performed by the local councils. Do people actually make broadly based policy assessments when judging the performance of a particular council or a party, or are they primarily oriented towards

particular issues or policy areas? In the survey on people's attitudes to local government, the respondents were asked how they would spend an extra NOK 1000 across a total of 30 specified local government services. Figure 1 reports the distribution of responses as regards number of policy areas chosen. As is evident from this figure, most respondents concentrate on a few issues; 83 percent mention five or less issues which ought be granted more money. Furthermore, as Brekke (1993) demonstrates in his analysis of these data, the areas chosen are clearly related to the respondents' own immediate needs. These observations indicate that altruistic and ideological considerations are more or less absent in the way in which the respondents approach the problem of setting up their policy priorities. Moreover, even if there is a concentration of the responses on certain types of services, the variation in responses is still so large as to preclude the assumption that any single issue dominates the voters' orientation towards local political matters. Altogether, Brekke's analyses seem to support the suggestion made previously about a de-politicization of local politics in favour of an outright consumers' approach to local politics.

Models of Analysis

The empirical analyses to follow will employ a modified public choice perspective, emphasizing local government performance, in order to explain - or rather to account for - variations in turnout across Norwegian municipalities. Before entering into a discussion on the operationalization of the general model of analysis, a presentation of the dependent variables is in order.

In measuring the dependent variables, two different formulations will be used. First, I am interested in the actual level of turnout - and how this level varies across the municipalities. Under this formulation, turnout is measured as a percentage of all eligible voters who actually did cast their vote in the election in 1991. The second formulation focuses on the changes between the elections in 1987 and in 1991. Two options as to the measurement of change have been considered. The first is simply to look at the difference in percentage turnout between the two elections. The second option is to calculate the 1991 election results as a proportion of the results in 1987. I have settled for the first option, not least because it has a straightforward interpretation in relation to the explanatory variables chosen. Conceptually, the choice of the first rather than the second option is of no consequence. Calculation of the correlation coefficients between these two alternative formulations gives us coefficients close to unity.

Let me now turn to the independent variables to be employed in the analyses. The emphasis will be on the impact of municipal policies - as measured by expenditure figures- on the pattern of voting. I will, however, also include some:; other factors characterizing the municipalities as to structural features as well as political features.

Four variables denoting basic aspects of local policy performance have been chosen for the analyses. While three of these variables measure various aspects of the budgetary situation in the individual municipality, the fourth variable may be characterized as a measure of policy outcomes or impact, giving a summary expression of what may be called local standard of living or level of welfare. The four variables are:

1. gross municipal expenditures per capita in 1990;
2. total bloc grants from the state per capita in 1990;
3. total fees and charges per capita in 1990;
4. index of local living conditions in 1989.

The first variable - gross expenditures - is a straightforward measure of the total level of local government activity. The variable tells us how important local service provision - on average - is to the inhabitants. Even if this figure - as suggested before - to a certain extent is determined by central government, local authorities still have considerable discretion in

deciding for what purposes the money should be used. I will also assume that the higher the level of gross expenditures, the more autonomous are local authorities in deciding on the priorities. The implication is that the higher the gross expenditures, the more there is at stake at local elections in the sense that the winning majority is more likely to have a real say about how financial resources are being spent on various competing purposes. Thus, I expect turnout to increase with increasing level of gross expenditures per capita.

An alternative hypothesis as to the relationship between the level of political participation and level of government activity has been suggested by R. Inglehart (1990) who argues on the basis of "the principle of diminishing marginal utility". The principle implies that as government budgets increase above a certain level, the marginal utility to the voter to be derived from additional service provision will diminish. This contributes towards weakening the incentives to vote in elections, particularly as long as elections are still kept within a traditional class voting framework. This hypothesis about diminishing marginal utility from public service provision may well be of relevance as long as we restrict our attention to traditional welfare services, as seems to be the underlying assumption in Inglehart's argument. This is explicitly formulated by Clark & Inglehart (1990), who in their discussion of the "new political culture" argue that "the more extensive and complete the adoption of welfare state-type programs, the less the pressure for continued growth of these programs and of government in general. Success is the welfare state's main problem" (p. 15). However, if we look at the development of local government spending in Norway, there is clear evidence that as the total level of spending has increased, an increased proportion of the budget has been appropriated to services such as environmental protection and culture - exactly those services which are supposed to be highly valued by voters subscribing to post-materialist values. This is to suggest that as the level of local government spending increases, the incentives for post-materialist voters to turn out in elections are also increasing. The probability that such post-materialist values may be catered for by local councils may also be enhanced by the fact that there is a very high turnover among local councillors, which weakens elitedirected politics in favour of an elite-directing politics at the local level, to use another of Inglehart's expressions. The Norwegian system of proportional local elections, which gives the individual voter an opportunity to choose between candidates on a party ticket - an opportunity which has been frequently used - contributes towards a further weakening of the oligarchic tendencies among the local party elites. This contrasts with the suggestion by Clark & Inglehart (1990, 25) that proportional representation according to parties increases party control. This may be so in parliamentary elections, but not in local elections in Norway (a fact that is also recognized

by Clark and Inglehart). Actually, the "gender revolution" which took place in Norwegian local politics during the 19705, bringing women into a majority position in several Norwegian local councils, was by and large caused by this opportunity contained in the election system to vote for particular candidates on the ticket.

The important point to be made here is that public budgets are multifunctional, with differing marginal utilities across the various functions. If the service provision in certain areas approaches an optimal level according to some welfare function, this may enable the public authorities to reallocate financial resources to service areas with high marginal utility to the voters, such as environmental policies. The fact that the voters also have the possibility to exert a direct influence on the election of individual candidates - although still within the framework of party tickets - should therefore produce a positive incentive to vote in a situation of "abundant" budgetary resources.

As far as the second policy variable is concerned, it has been included in the model in order to account for the distributional aspects of central government policies in relation to local government. As can be observed in Table 1, the amount granted by central government to the municipalities is clearly and negatively related to population size. As Madsen noted in his study of electoral outcome and macro-economic policies in Scandinavia: "Norwegian voters in particular have been concerned about the distributional effect of economic growth" (1980, 27). This also applies to the distribution of budgetary resources across the municipalities. Even if central government grants in a strict sense are not a measure of local government performance, the size of the grant indicates the extent of the financial burdens needed to keep a high level of local government activity which has to be carried by the local tax-payers. High central grants contribute towards reducing the price to be paid by the local recipients of the services. They also imply that the local political authorities are in a better and more autonomous position to decide on the level of services to be provided, without having to charge their own voters with the full costs of keeping up a "comfortable" level of service provision. The political effect of this is - I will suggest - to maintain a high level of political support, and thereby a high level of turnout in the local elections.

Conversely, a high level of fees and charges indicates a rather tight financial situation in the municipality, and where the local tax-payers carry a disproportionate burden for securing the local service provision. As already suggested, while taxes and state grants may be regarded as more or less invisible revenue sources by the voters, fees and charges are not. Furthermore, as a taxing system they are regressive and in this sense quite similar to a poll tax system. In their study of the 1990 London local elections, Gibson & Stewart (1992) observed a significant - although not

very strong - effect of the level of poll tax on the pattern of voting, such that the incumbent party loses from a high poll tax level. This demonstrates the high political salience of such revenue sources, but it is not clear from their analysis whether the level of poll tax had any effect on the turnout in these elections.

Two alternative and competing hypotheses may be suggested as to the effects of fees and charges on the level of turnout. First, a high level of fees and charges may be regarded as a sign of problems of governance, in the sense that the local councils are unable to adjust their spending priorities to their available revenue resources. The popular interpretation of this situation is that the financial situation is running out of political control. In my interpretation it is rather the councillors who are running out of control, by acting as some kind of freeriders on the budgetary balance. This situation is likely to occur in municipalities with weak, politically fragmented and therefore unstable majorities, where councillors belonging to this majority do not feel any obligations to secure a budgetary balance within the framework of ordinary revenues. The easy way out of the resulting overspending is to increase fees and charges. Since no one particular political party can be blamed for this, the political consequence will be a general loss of electoral support for the local council as such, and with a consequently low level of turnout in local elections.

Second, a high level of fees and charges may be a result of clearly stated political objectives - as a step in the direction of privatization of local service provision. Such political objectives have been stated by the rightwing political parties. If such ideological motives are present in explaining high fees and charges, this -1 will suggest - contributes towards mobilizing the voters, thus resulting in high turnout. I will keep both interpretations as open options.

The fourth policy variable to be included is an index of local living conditions. It is constructed as an additive index based on 15 indicators of living conditions, including various measures of local government service provision. Among the factors included in this index are: the level of unemployment, the coverage of various health services (such as number of medical doctors per 1000 inhabitants), the coverage of kindergartens, average personal income, measures of mortality, the level of education as well as the proportion of the population living on disability pensions and social welfare benefits. The indicators are presented in more detail in the appendix. As to the effect of this variable on the level of turnout, the considerations spelled out in connection with the level of gross expenditures apply here as well. However, in contrast to gross expenditure - which to most voters is a rather abstract measure of government performance - this index gives us a more direct and politically more visible measure of the performance of local government activity. I hypothesize this variable to be positively related to the level of turnout.

As far as other social characteristics of the municipalities are concerned,
three further variables have been included in the model:

5. population size, measured as the natural logarithm of size;
6. proportion of population above 18 years of age with university education;
7. proportion of eligible voters aged between 18 and 24 years.

The first of these three variables, size of population, is an obvious choice, since population size is inversely related to the probability that an individual's vote will nicike a difference for the outcome of the election. Because of the substantial range in differences in size of Norwegian municipalities, I have chosen to use the natural logarithm of size in the analysis. The hypothesis is thai turnout decreases as population size increases.

As for the other variable, we know from studies of electoral behaviour that the better educated are more likely to be politically active than the less educated. Voting is, of course, no exception from that rule. But, as Inglehart (1990) suggests, the better educated may be more inclined to prefer other and more unconventional modes of political participation. According to Inglehart, the net result of rising levels of education and political skills

. . . has been a stagnation of electoral turnout and other forms of elite-directed participation,
on one hand, together with the growth of elite-directing forms of participation, on the other
hand. (1990, 336).

As I suggested before, the way in which Norwegian local elections have developed over the past two decades - in terms of high turnover of councillors - the elections provide an opportunity to elite-directing forms of participation. Altogether, I therefore hypothesize turnout to be positively related to a high level of education.

In several studies of political participation it has been observed that younger voters are less likely to participate than are older voters. Furthermore, as Inglehart has suggested, there is a cohort difference as to political attitudes and values between younger and older voters, in the sense that the younger generations are more likely to subscribe to postmaterialist values than are older generations. Also this generational value shift, following the suggestions made by Inglehart, would imply a lower turnout among younger voters. If the proportion of young voters differs across the municipalities, this could account for lower turnout in municipalities with a high proportion of younger voters. Thus, I hypothesize a negative relationship between this variable and turnout.

Finally, three variables capturing political features of the municipalities
have been included in the model of analyses:

8. turnout in the local election in 1983;

9. percentage of socialist votes in the 1987 election;
10. number of political parties contesting the election in 1991.

The first variable attempts to operationalize the bandwagon effect suggested by Elster under the norm of fairness. Ideally, this bandwagon effect ought to be defined as a long-term trend, including several previous elections. At the present stage of research, I have decided to use only the 1983 election, on the assumption that this election gives a fair representation of the trend in turnout in the individual municipalities over the past couple of decades. It may be argued that by including the level of turnout in a previous local election to account for the turnout level in a later election, we come close to a tautological formulation of the model. It is, indeed, tautological, in the sense that any self-reinforcing trend of this type is tautological. In order to avoid some of the problems related to auto-correlation, this variable has been lagged by yet another election period, in the sense that I use turnout in 1983 rather than in 1987 to depict this bandwagon effect. Following the suggestions made by Elster, I hypothesize that a positive deviation in 1983 leads to a high turnout in 1991. In other words, I assume this variable to be positively related to the level of turnout in 1991.

As for socialist strength, this variable will be used to calculate interparty competition in the municipalities. Socialist strength is measured as percentage of votes for the Labour Party, the Socialist Left Party, the Communist Party and the Marxist-Leninist party (AKP). I assume that the closer to 50 percent of the votes that these four parties come, the higher the level of inter-party competition. Following Downs, I hypothesise that high inter-party competition will increase the level of turnout, which is to suggest that the closer to 50 percent of the votes the socialist parties are, the higher the level of turnout. To capture this effect, the variable has to be transformed into the following polynomial formulation of the relationship between turnout (T) and socialist strength (S):

This formulation has a maximum value when

and when the second derivative of the equation is negative:

Thus, the hypothesized relationship implies that a > 0 and b < 0.

Two quite opposite hypotheses may be made as to the effect of the number of parties on the level of turnout. First, the number of parties running for election may signify a rather anarchic political situation in a municipality, with considerable problems in forming stable majority coalitions. Such political fragmentation may lead to diminishing legitimacy and support of the local political regime, thus implying a reduction in the level of turnout. Alternatively, a large number of parties may signify a move from a class-based party system towards a less ideological and issuebased party system. This provides the voters with a richer choice as to what political platforms to support. If we look closer at the more recently formed parties contesting local elections, we find that they are clearly issueoriented, but at the same time even more ideologically oriented than traditional class-based political parties, such as the Labour and the Conservative Parties. Several of these parties are to be found at both extremes on a left-right dimension - such as the neo-liberalists (Progressive Party) and the Anti-Immigration Party on the very right wing of this axis, and the Marxist-Leninists at the other extreme of the axis. We do, however, also find parties subscribing to post-materialist values, such as the Environmentalists. In the sense that many political parties provide the electorate with a better choice, I will assume that this has a positive effect on turnout. In choosing between these two interpretations, I am inclined to select the latter one, on the assumption that a large number of parties tends to increase the level of turnout.

As regards the analysis of change in turnout between the 1987 and the 1991 election, I will employ the same set of independent variables as I do for variations in the level of turnout. It may be argued that also the independent variables should be measured in terms of changes over the four-year election period. This may be particularly relevant as far as the three budgetary measures are concerned. This will also be done in the analysis. However, to most voters such measures of change are merely abstract calculations, to which it may be difficult to relate. It is the reality as perceived on the day of election which is of most importance for most voters. Still, most people will have at least a vague notion of whether things are improving or becoming worse as far as local policy performance is concerned. I have therefore chosen to use change in gross expenditure per capita between 1986 and 1990 to measure such changes. This variable will also be used in the model of variations in turnout, and I hypothesize this variable to be positively related to both the level of turnout and changes in the level of turnout.

Let me now, before turning to a presentation and discussion of the results
of the analyses, present a summary of the model of analysis with the
suggested hypotheses.

Results of the Analysis

Ordinary linear regression analysis has been employed to estimate the effects of the independent variables on the two dependent variables. Inspection of the correlation matrix between the independent variables demonstrates no problems related to possible multicollinearity - with one exception, namely the relationship between the two terms denoting interparty competition. Since the inclusion of this polynomial formulation is justified on theoretical grounds, there is no reason to modify the model in order to avoid this high correlation. As we also will see, the high intercorrelation does not produce any significant disturbance to the statistical analysis - either in terms of abnormal beta-values or with respect to any noticeable differences between the ordinary and the adjusted explanatory power of the models (R squared).

Before presenting the estimated regressions, it may be worthwhile to consider the variation in and the frequency distribution of the level of turnout in 1991 across the Norwegian municipalities. This is displayed in Figure 2, which also compares the actual distribution with a normal distribution. Among the 439 municipalities included in the analysis, the level of turnout varies between 53 and 85 percent, and with an average of 67.5, which is clos eto the total national figure of 66 percent. We also observe

that the actual distribution by and large follows a normal distribution, thus
justifying the conduct of ordinary tests of significance of the regression
coefficients.

Let me now turn to an inspection of the estimated regression equations.
Table 3 displays the estimated regression coefficients and their /-values for
the equation with turnout level in 1991 as the dependent variable.

The first thing to be noted is the high explanatory power of the model. Altogether 69 percent of the variation in the dependent variable is accounted for by our independent variables. Given the large number of observations upon which the estimation of the model has been performed, this is by any standard a noticeable explanatory power. However, in contrast to this very high fit of the model as such, I have been somewhat less successful in my selection of independent variables. Only five out of the twelve selected independent variables display significant effects upon the dependent variable (using a one-tailed test with a level of significance of 0.05). However, of the remaining variables, we find three that measure local government performance, namely gross expenditures, change in gross expenditures between 1986 and 1990 and fees and charges per capita. In the sense that local government performance constitutes the core element of the explanatory model employed here, this result may seem rather discouraging. This should not, however, lead us to reject our basic hypothesis that "policy does matter". As is evident from the table, the second most important explanatory variable is the index of local living conditions,

a variable which probably is a better and more immediate measure of the impact on the citizens of local (and central) government performance than the level of gross expenditures. As suggested before, variations in costs of producing local government services across variously sized municipalities, may represent major distortions in equalizing expenditures with the actual level of service provision. This is not to suggest that we may substitute one of the two variables for the other. As is obvious from the figures in the table, the two variables display opposite effects on the level of turnout - the index of living having a positive effect, while the effect of gross expenditure being negative. In fact, if the index of living conditions is removed from the model, the impact of all remaining policy variables is reduced, while the effect of population size increases substantially. This suggests that these policy variables tap different dimensions of policy performance. If we inspect the zero-order correlation between these two variables, we find it to be positive and significant (0.29 with p<0.001). The fact that they have opposite effects on the level of turnout seems to suggest that while the index of living conditions gives a direct measure of the service level, the level of gross spending may - when controlling for the index - indicate the efficiency by which the financial resources are being spent in the municipalities. Under this interpretation, a high level of spending may simply be interpreted as low cost-effectiveness on the part of the local council, thus contributing to decreased popular support for the council as such, and thereby lower turnout in the election.

As far as the other policy measures are concerned, we do observe that state grants are positively and significantly related to the level of turnout, while the effect of fees and charges is positive, but not significant. It should be noted that these two explanatory variables display a fairly strong and negative zero-order correlation (0.69). The fact that both variables are positively related to the level of turnout indicates that we have in fact been able to separate the two dimensions along which the variables were hypothesized to effect the level of turnout.

Altogether, and judged from our basic perspective, these observed effects do indicate that policy does matter in explaining the level of turnout. What the empirical results indicate is that the theoretical interpretation of the chosen policy measures may be more subtle than initially assumed, which should urge us to be more careful in the operationalization of such variables in future studies of turnout. If we judge the overall result of the effects of the policy measures, we do not find any support for Inglehart's suggestion about the diminishing marginal utility from government service provision and the assumed effect on level of turnout. The results presented here suggest the reverse relationship, namely that turnout increases as level of local service provision increases. This suggestion will, however, be considered more closely in the next part of this analysis. Let me therefore now turn to the other hypothesized relationships.

First, the results testify to the effect that turnout increases with diminishing population size of the municipality. Population size comes out as the third most important explanatory factor in this model. In their discussion of political participation and the size of local units, Dahl & Tufte (1974) suggested - on the basis of voting data from various countries - that "within countries, among local units of the same legal type, there is no general relationship between turnout and unit size" (Dahl & Tufte 1974, 61). This is obviously not the case as far as Norway is concerned; the effect of population size on turnout is strong and negative. It might have been argued that the smaller the municipality is, the better are the opportunities for other forms of political participation, such as those subsumed in Inglehart's concept of elite-directing participation. In the survey on citizens' attitudes to local government, the respondents were asked whether they had taken part in various forms of political activity, such as participating in campaigns on specific issues, taken part in political demonstrations, written to newspapers about specific political issues, submitted complaints or proposals to the local authorities or having contacted members of local councils on specific issues. On the basis of these questions an index of political activity was constructed. If this variable is cross-tabulated with size of municipality (actually a composite variable of size and urbanization), we find political activity to be positively related to size, such that the larger and more urbanized the municipality is, the more likely are citizens to be politically active (the gamma is 0.206, which is a significant correlation). While between 60 and 70 percent of the respondents in the largest cities reported that they had taken part in one or more of such activities, the corresponding figure for respondents from the smallest and least urbanized municipalities is slightly less than 40 percent. Thus, contrary to our commonsense assumptions, it may seem that there is a substitution effect between voting in elections and other and more direct forms of political participation, such that the larger the municipality, the less likely are people to vote, but at the same time more likely to participate on other political arenas - while voting in elections is the dominating mode of political participation for citizens in smaller municipalities. In other words, Inglehart's suggestion is clearly supported by our data, but within an institutional context which - at least to me - is somewhat surprising. One possible explanation may be that this relationship is produced by the fact that political competence - as a result of more widespread educational opportunities - is higher in the larger towns and cities. As we can see from Table 3, the percentage of voters with university education is negatively related to the level of turnout, which gives conditional support to the above suggestion, although the coefficient is not significant. It is conditional in the sense that we cannot make any inference from this result to the level of individual voters; this negative relationship is not the same as saying that it is those with a university education who are most likely to abstain from voting.

Let me now turn to the political factors included in the model. The variable displaying the strongest effect on the level of turnout is the lagged turnout variable, measuring turnout at the local election in 1983. This variable was interpreted as a self-reinforcing bandwagon effect, which works on a long-term perspective. The positive effect of this variable supports this interpretation. Despite this observation, I am not completely satisfied with this operationalization of the bandwagon effect. First, the variable lacks a long-term dimension. Second, the effects of other variables may be contained in the measurement of this variable. This is clearly demonstrated if this variable is removed from the regression model, which has the consequence that the effect of proportion aged 18-24 years becomes the second most important explanatory variable (the T-value increases from 0.2072 to -4.7088), and where also the sign of this effect is reversed from a weak positive to a strong negative effect - the latter effect being in accordance with my hypothesis.

As far as party competition is concerned, we do not find significant effects of the two terms included in the formulation, and the signs of these terms indicate a reverse relationship of that hypothesized, such that a high level of party competition contributes to lower turnout. In other words, the estimation gives us; a U-formed relationship rather than the hypothesized opposite. One possible explanation of this may be that the presence of hard-core party loyalists - and thereby voters - is higher in municipalities with one dominant party than in politically more "open" municipalities.

Finally, number of parties contesting the election has a rather strong, significant and positive effect on the level of turnout. Although this effect is in accordance with our hypothesis, it seemingly contradicts the suggestion made in the previous paragraph on the effect of party competition. However, the presence of many parties does not have to be an indication of a high degree of party competition. Most parties contesting the local elections are minor ones - at least at the local level. In fact, the presence of several minor parties may be taken as evidence of frustration and protest against one large dominant party. This suggestion will not be further pursued here, but an inspection of the zero-order correlations between the number of parties on the one hand and the share of votes for the major parties testifies to a fairly strong negative relationship - with the exception of the Conservative Party, where I find a strong positive correlation. This may be due to the fact that the Conservatives have been rather weak in most rural municipalities, with their strongholds in the larger and most urbanized municipalities. This is, however, a problem which in itself calls for a more thorough investigation.

Let me now turn to a presentation of the results for the second model,
with changes in turnout between 1987 and 1991 as the dependent variable.

The estimated regression coefficients and their T-values are presented in
Table 4 above.

First we notice that the explanatory power of this model is much lower than the previous one, but is still of some magnitude, explaining about 31 percent of the variation in the dependent variable. Second, we observe that the order of relative importance of the various independent variables is somewhat changed compared with the former model, but that some of the effects - with one important exception - are in the same direction as that observed in Table 3. The only effect having changed the sign compared to the previous analysis, is the effect from turnout 1983. Here we observe a fairly strong negative effect, indicating that increases in turnout between 1987 and 1991 were highest in municipalities having the lowest level of turnout in 1983. In other words, as far as changes in turnout are concerned, the bandwagon effect suggested by Elster is not supported by our data. As for the variables measuring local government performance, four out of five display a positive relationship with changes in the level of turnout.

Once again, we do not find any empirical support for Inglehart's suggestion about diminishing marginal utility from voting. Our measure of local living conditions is in fact the variable displaying the strongest positive effect on changes in the level of turnout which suggests that the level of well-being in a local community is strongly conducive to the act of voting in local elections. A reservation to this conclusion should, however, be

added, namely the faet that the level of gross expenditure displays a
significant and negative effect on changes in the level of turnout.

As for the political variables, we do not find any support as to the suggestion that a high level of party competition will increase the level of turnout. The significant U-shaped relationship between percentage of socialist votes and changes in the level of turnout suggests that turnout is at its highest in municipalities with one dominant party coalition. We do, however, also note that number of parties contesting the election exerts a positive effect on changes in the level of turnout. As suggested before, this variable should not in any sense be equated with the level of party competition, since many of these parties are very minor ones - even in a local setting. The positive and significant effect from this variable does indicate that the appearance of new - and frequently single-issue - parties does have a mobilizing effect in local elections. The results observed here do, however, not tell us whether this mobilization is to the benefit of these new and smaller parties or the traditional political parties.

Policy Matters

As demonstrated in the previous section, the empirical analyses testify clearly to our major hypothesis that policy matters in explaining the level of turnout at Norwegian local government elections. Does this observation signify a change in the direction of a decline in class-oriented voting towards a more pragmatic issue-oriented voting, as, e.g., observed by Franklin (1985) in his analysis of changes in electoral choice in Britain? The present analysis does not give any definite answer to this question. The results are indicative of a more rational voting behaviour on the part of the electorate, but in order to assess this problem, the analysis needs a temporal dimension. Furthermore, this analysis has not taken into consideration changes in the distribution of votes across different parties. In their study of the last general election, Valen et al. (1990) observed rather strong effects of various socio-economic variables (such as unemployment and immigration) on changes between 1985 and 1989 in the relative share of votes for the Labour, Conservative and Progressive (neo-liberalist) Parties, but not for the Socialist Left Party. In the sense that the Socialist Left Party to a considerable extent may be regarded as an exponent of post-materialist values, while the other three parties have a more traditional materialist left-right profile, these results indicate that issue-voting is constrained within a traditional class-structured framework. In an attempt to determine to what extent similar relationships are present in local elections, a modified version of the model used to examine the variations in turnout was used to analyse variations in the share of the votes for the major political parties

in the 1991 election. With the exception of the models for the Conservative, the Progressive and the Christian Democratic Parties, we find rather strong effects of the policy variables for the three remaining parties - that is, Labour, the Socialist Left Party and the Centre Party (Agrarians). Since these three parties in fact constitute the current majority coalition in the Norwegian parliament (if not a formalized coalition), these results suggest that there is an interaction effect between voting according to a traditional ideological framework and issue voting.

This problem will not be pursued further here. It is, however, likely that a consumer's approach to the study of the pattern of voting is of more relevance at the local than at the national level. But this does not imply that we are leaving a class-oriented perspective on local politics. Rather, following sociologists like Castells (1977 and 1978) and Saunders (1981), class is being reinterpreted according to location in relation to collective consumption and not merely to location in the labour market. That this may be a more satisfactory approach than basing the analysis on a traditional class concept has been clearly demonstrated by P. Dunleavy who points out that

Because consumption cleavages and their effects on political alignment are fundamentally
structured by state intervention, they cross-cut occupational class lines (1979, p. 443).

This is, however, a problem that calls for another analysis than the one
presented here.

---

NOTES

1. Data have been provided by the Norwegian Social Science Data Services (NSD) and from a survey on attitudes to local government, conducted by H. Baldersheim, P. A. Pettersen, L. Rose and myself. H. Valen and G. Vogt have kindly given me access to data from the Norwegian Election Research Programme. I am particularly indebted to Helene Roshauw (NSD) who has organized the data and with great patience supervised and encouraged my work on this paper. I also appreciate the comments and suggestion from an anonymous referee.

REFERENCES

Alford, R. R. & Lee, E. C. 1968. 'Voting Turnout in American Cities', American Political
Science Review 62.

Baldersheim, H., Hansen, T., Pettersen, P. A. and Rose, L. 1990. Publikums syn på
kommunepolitikk og kommunale tjenester. LOS-rapport 90/2. Oslo/Bergen.

Barry, B. 1970. Sociologists, Economists and Democracy. London: Collier-Macmillan.

Brekke, 0. 1993. Egeninteressens rolle i publikums prioriteringer av kommunale tjenester.
Oslo: Institute of Political Science.

Budge, I. & Farlie, D. J. 1983. Explaining and Predicting Elections. London: George Allen
& Unwin.

Castells, M. 1977. The Urban Question. London: Edward Arnold.

Clark, T. & Inglehart, R. 1990. The New Political Culture. Chicago and Ann Arbor: Draft
manuscript.

Dahl, R. A. & Tufte, E. R. 1974. Size and Democracy, Stanford: Stanford University Press.

Dunleavy, P. 1979. 'The Urban Basis of Political Alignment', British Journal of Political
Science 9.

Elster, J. 1989. The Cement of Society. Cambridge: Cambridge University Press.

Ferejohn, J. A. and Fiorina, M. P. 1974. 'The Paradox of Not Voting: A Decision Theoretic
Analysis', American Political Science Review 68.

Franklin, M. N. 1985. The Decline of Class Voting in Britain. Oxford: Clarendon Press.

Gibson, J. G. & Stewart, J. D. 1992. 'Poll Tax, Rates and Local Elections', Political Studies
40.

Hansen, T. 1981. 'Transforming Needs into Expenditure Decisions', in Newton, K. ed.,
Urban Political Economy. London: Frances Pinter.

Hansen, T. 1988. 'Finansiell utvikling i kommunene', in Kjellberg, F. ed., Nytt inntektssystemnye-
utfordringer for kommuner og stat. Oslo: Kommuneforlaget.

Hansen, T. 1991. 'Desentralisering som kollektivisering av tjenesteforbruk', in Nagel, A. H.
ed., Velferdskommunen. Bergen: Alma Mater.

Hansen, T. & Kjellberg, F. 1976. 'Municipal Expenditures in Norway: Autonomy and
Constraints in Local Government Activity', Policy and Politics 4.

Hansen, T. and Sørensen, R. 1985. Bykommuner under finansielt stress. Oslo: Universitetsforlaget.

Hansen, T. & Sørensen, R. (1988). 'The Growth of Local Government in Norway: The Sky
is the Limit?', in Kjellberg, F. & Taylor, J. eds., Local Finances in the Contemporary State:
Theory and Practice. Oslo: NORAS.

Hansen, T., Roshauw. H. & Sørensen, R. 1988. Analyse av kommunale finanser. Oslo:
Universitetsforlaget.

Inglehart, R. 1990. Culture Shift. Princeton NJ: Princeton University Press.

Madsen, H. J. (1980). 'Electoral Outcomes and Macro-Economic Policies: The Scandinavian
Cases', in Whiteley, P. ed., Models of Political Economy. London: Sage.

Riker, W. & Ordeshock, P. 1968. 'A Theory of the Calculus of Voting', American Political
Science Review, 62.

Saunders, P. 1981. Social Theory and the Urban Question. London: Hutchinson.

Tullock, G. 1967. Tovard a Mathematics of Politics. Ann Arbor: University of Michigan
Press.

Valen, H., Aardal, Ei. & Vogt, G. 1990. Endring og kontinuitet Stortingsvalget 1989. Oslo:
Statistisk Sentralbyrå.