An empirical analysis of municipal choice in Portugal

Abstract

In assessing specific combinations of supply and demand influences on local public choice, four policy issues with differing ideological attributes were compared over a period of two electoral terms in Portuguese municipalities. Results confirmed the significance of ideology and the non-neutrality of the electoral agenda in shaping specific expenditure issues (culture and urban waste management). There was also some evidence demonstrating the political influence of interest groups, especially in reference to less visible issues. The findings draw attention to the possibility of opportunistic cycles and for political discrimination in central government investment allocation among municipalities.

Acknowledgements

I am grateful for the helpful comments of Rui Baleiras and of John Stirling Wilks.

Notes

1 The archipelagos of Madeira and of Azores constitute two regions and are ruled by regional governments. Given their special status these two regions are not included in the empirical study.

2 The most important local revenue of municipalities comes from taxes on urban real estate. Obviously in urban areas this revenue is much more important than in rural municipalities. In rural municipalities central government transfers are more than 50% of municipal revenues.

3 There are some issues (not the most important) in which the mayor has competence to decide alone.

4 A bureaucrat, according to Niskanen (1973: 11), is ‘the senior official of any bureau with a separate identifiable budget’.

5 See Samuelson (1954).

6 It is not possible to aggregate social preferences into optimal collective action.

7 In principle the expected sign for N's influence is (–), because of scale economies in consumption. However, in the case of strong congestion in consumption the expected sign might be (+).

8 In all issues analysed in this paper there is no specific tax to finance the expenditure on a specific issue.

9 Many median voter applications follow Borcherding and Deacon (1972) in considering that the population of each community equally shares local taxes that are not connected directly to a specific expenditure.

10 Departing from Consumption Theory, the median voter's (community) demand could be defined by a Cobb-Douglas demand function (Bergstrom and Goodman, 1973): e = K y_m ^aθ^b N^{(α−1) (1 + b)}. (a) is the demand income elasticity; (b) is the demand price elasticity; K is a constant; (α) is the degree of rivalry on consumption, that varies between 0 (non-rivalry) and 1 (private good). For a review of the median voter model see Bahl (1980); Cruz (1998); Holcombe (1989); Meltzer and Richard (1983); Pommerehne and Frey (1976); Pommerehne et al. (1978).

11 The effect of additional unconditional grants (given to local governments) on increasing local public expenditures is stronger than the same increment in the income of the population of the community.

12 Bailey and Connolly (1998) reviewed the literature on the ‘flypaper effect’.

13 There are other ways in the literature to define the fiscal illusion influence, for example the relationship between non-transferred local revenue and the total revenue of local government. Letting R be the total local revenue, the level of independence from unconditional transfers of the community is . We would expect that lower ‘per capita’ expenditure is associated with higher proportions of local own revenue, but it can also express a stronger ability of municipalities to collect local revenue and, in consequence, a greater ability on the part of the municipality to generate ‘per capita’ expenditure. Due to this mixed effect I think the first measure is preferable.

14 In the empirical literature on interest group political influence it is possible to find several measures of group power (number of members, level of resources, number of interventions in media, amount of contributions to candidates in elections); for a review of the empirical interest group literature see Potters and Sloof (1996). Ahmed and Greene (2000), Baumgardner (1993), Congleton and Bennett (1995), Congleton and Shughart II (1990), and Shapiro and Papadakis (1993) represent a sample of papers combining the median voter hypothesis with the interest group hypothesis.

15 The inclusion of this variable was not statistically significant (as expected) and increased very much the co-linearity between the variables.

16 In the Portuguese case it makes sense to include ideological preferences, because political parties support candidates in local elections.

17 However it is rather expanded in Scandinavian literature; several articles are collected in Ratts⊘ (1998).

18 The congestion parameter in median voter empirical estimation for several countries indicates that local government provision on the issues considered in this study are of ‘private good’ type, examples are: Bahl (1980); Bergstrom and Goodman (1973); Gemmell et al. (2002); Reiter and Weichenrieder (1997); Turnbull and Djoundourian (1994). It is assumed that the same occurs in Portugal.

19 The mayor may not assign them specific executive functions, but if the party that supported the mayor did not get a majority of seats in the municipal assembly, important executive decisions may be blocked.

20 See Frey (2000) and Throsby (2001) for a brief review on cultural economics.

21 Note that in Portugal the tax autonomy of municipalities is very limited, because all local revenue that is not directly linked to the provision of services (without a direct tax-price) is subordinated to central government administration (change of tax rates, delimitation of fiscal basis, settling, collection).

22 Municipal elections occurred in 1993, in 1997 and in 2001, always in December. Fifty-three municipalities were not included because: CDS did not run in the contest in at least one year of the sample and in the years it did run it won more than 5% of votes (32 municipalities); there is no data on at least one of the variables (13 municipalities); the dimension of the municipality changed with the creation of new municipalities (5 municipalities); the municipality was created in 1998 (3 municipalities).

23 I assume equal regression coefficients for different cross-sectional units (municipalities), with the effects of omitted individual specific variables being treated as fixed constants over time in the fixed-effects estimation, and as random variables in the random-effects estimation. The ignorance in respect to a better specification and the need for preserving the degrees of freedom conditioned the option.

24 Obtained from panel data that include all the years in the analysis (not separating the electoral stages), in order to guarantee more observations and more confidence in the estimated results: ∗ indicates that H₀ is statistically rejected for a level of significance of 5%; ∗∗ indicates that H₀ is statistically rejected for a level of significance of 1%, the null hypothesis is the random effects model.

25 As a matter of fact, if the random effects model is the ‘valid’ model, the fixed effects estimator will still produce consistent estimates of the identifiable parameters. So, the option taken was not too risky.

26 Data at 1995 constant prices. Regional differences on prices were considered.

27 Data for dependent variables, median voter variables (except income) and interest group variables is from the Regional Statistical Year-Book of 1996, 1997, 1998, 1999, 2000, 2001 and 2002, published by the National Statistical Institute (INE). Data on municipal income and on income tax was collected from the internet site of the Exchequer – DGCI – http://www.min-fin.pt. Political data is from the internet site of National Elections Committee (CNE) – http://www.cne.pt.

28 On the other hand, affecting the degree of congestion in the consumption of local goods is only imperfectly under the control of mayors and is normally not feasible in one electoral term. The degree of congestion in consumption depends on the ‘technical’ characteristics of the commodities. It is different from the level of congestion in consumption that directly depends on the quantity of provision. The former is not related to mayors' decisions, whereas the latter is related strongly. In addition, in mainland Portugal, municipalities do not have competence to make significant changes to the structure of taxes, so it is not to be expected that in any given electoral term mayors will change the level of fiscal illusion. In mainland Portugal central government defines the structure of the most important taxes whose revenue reverts to municipalities. The competence of municipalities in the fiscal issue is residual. Finally, the incumbency effect is independent of electoral stages, because the visibility of the mayor and his experience are functions of the former mandate. Thus the Chow test does not include the variables POPULATION, ILLUSION and INCUMBENT.

29 LM is equal to 1 for 56 observations in each regression.

30 In the fixed effects estimation the coefficient of RM was not statistically significant. However, according to the random effects estimation (Appendix 1) RM is statistically significant with the expected sign.

31 Social democrats from 1991 until 1995 and Socialists between 1995 and 2002.