Voters’ Preferences Regarding Municipal Consolidation: Evidence from the Quebec De-Merger Referenda

ABSTRACT:

We use the results of the 2004 Quebec referenda on the mergers of municipalities to analyze the determinants of citizen preferences regarding municipal consolidation and fragmentation. The core hypotheses of our empirical model are generated from the economic theory of optimal jurisdictional size. Holding constant the influences of language and of a unit’s share of the merged population, we find voters are more likely to support de-merger when they expect that the merged unit will display a different public expenditure level than that of the municipality in which they reside. We also find support for de-merger is less when voters expect de-merger to increase the tax-price of local public services.

Notes

1 The merger targets, though large in number, nonetheless constituted a relatively small proportion of the total of general purpose local governments in the province. As of March 26, 2003, there were 1,113 municipal units functioning as general purpose local governments. In addition to numerous smaller units, the mergers affected the province’s major cities, including Montréal, Québec, Longueuil, Gatineau, Saguenay, Sherbrooke, Trois-Rivières, and Lévis.

2 Throughout this article, the phrase “local government structure” or variants thereof will refer to the degree of municipal government fragmentation or concentration as measured by the number of local units within a given geographic region.

3 See also Barzel (1969); King (1984); Fisher (2007), and Alesina, Baqir, and Hoxby (2004).

4 Additional potential benefits from consolidation might include broadening the range of public services available locally (through economies of scope) and affording the region greater influence in state or provincial political decision-making processes. Other potential advantages of greater municipal population size and geographic scope have been cited in support of regionalization, including better coordination of regional economic development and land use regulation policies and a redistribution of tax burden to tap upper income suburban residents for the support of central city services (some of which they use) and income support programs.

5 The OJS analysis described here represents an intermediate approach to an ongoing, long-term discourse between advocates of regional government like Downs (1994), Rusk (1995), and Orfield (2002) and those, such as Ostrom, Tiebout, and Warren (1961), and more recently Feiock (2004), for example, who prefer to focus on the decentralized governance processes that generate negotiated cooperative arrangements. OJS does not advocate either centralization or decentralization, but instead provides an analytical framework for striking an appropriate balance between the two.

6 Giertz’s findings with respect to the influence of median family income on decentralization proved to be somewhat controversial. Mullen (1980) found significant negative correlations between median family income and the ratio of state to state-plus-local expenditures. He argues that local autonomy is a good that is desirable in itself, is costly (due to foregone scale economies), and is likely to be a normal good. Giertz (1983), in a rebuttal to Mullen, argues that hypotheses with respect to the role of income differences in explaining fiscal concentration cannot be developed a priori, given that the degree to which functions are centralized at the state level depends on the interplay between the relative income elasticities of spending at the state and local levels and on the degree to which state and local spending are substitutes one for the other.

7 As noted, Filer and Kenny (1980) also look at voting outcomes; however, they examine the influence of disparities in income—a determinant of spending—while we look at disparities in spending levels directly.

8 Note that in order to obtain single-peaked preference curves, it is sufficient that (as we assume) the marginal dollar value of the public service is decreasing in the level of service provided and that the marginal tax payment is constant (or at least passes through the marginal value function from below). These assumptions are consistent with those used by Greene and Parliament, who refer to welfare losses attributable to consolidation as “political externalities” (Greene and Parliament, 1980). Greene and Parliament assume as well that equilibrium levels of public service output are Pareto-optimal. We instead initiate the voting process from status quo levels of output that may or not be optimal but which have been arrived at by majority vote.

9 Of course A is indifferent if .

10 The original data set from which the observations used in this study were derived included all 213 municipal units for which the Director General of Elections reported “results of the opening of the registers” (Directeur Général des Élections du Québec, 2004B) and which appear in the list of merged units current as of March 5, 2004 (Ministère des Affaires municipales, du Sport et du Loisir, 2004a). These 213 units were those which were authorized to choose, through the registration process described above, whether or not to participate in referenda on deconsolidation. In all, twenty units were deleted from the database for various reasons. One unit was deleted because demographic information did not appear in the Statistics Canada online Community Profiles database. The merged unit of Mont-Tremblant did not appear in the MAMSL data set, so its four component units were omitted. Four additional units, while appearing in the Statistics Canada database, had missing values for one or more of the variables of interest, and thus were excluded. Two units were dropped because they were the sole remaining unit in a merged entity some components of which had missing data. Three units were omitted because tax impact estimates were not available. Two units were dropped because expenditure data were not available. Finally, the three component units of one entity were excluded from the data set due to incompatibility in population figures between the Statistics Canada data and that appearing in the MAMSL data set. The final data set consisted of the remaining 193 observations. Among these 193, 80 units held referenda.

11 For a representative example of these reports see Étude Économique Conseil (2004). For a typical summary see Québec (2004e).

12 The structural consequences of the referenda were not as clear-cut as might have been suggested by the wording of the referendum question itself. De-merger would not restore the full list of service responsibilities to the reconstituted units. Instead, a significant number of functions would be retained by an “agglomeration,” consisting of all of the component units of the merged municipality. All preexisting municipal functions would be performed by the agglomeration for those communities choosing to remain in the merged entity. Each reconstituted municipality would regain jurisdiction over only a limited subset of municipal functions. Reconstituted municipalities would be governed by a municipal council and would be represented on the agglomeration council based on their share of the agglomeration’s total population. Decisions of the agglomeration council would be subject to a veto on the part of the totality of representatives of those municipalities choosing not to be reconstituted. The summaries of projected impacts of de-merger provided to each community by the province explained the structural implications of de-merger. It is not clear how widely understood these consequences were among the voters.

13 Source: Québec, Ministère des Affaires Municipales, du Sport et du Loisir. Principales Conséquences Financières de la Réorganisation. http://www.mamsl.gouv.qc.ca/publications/organisation/cons_fina_defu.pdf. Accessed October 2004.

14 As noted above, we have no means of assessing either the accuracy of these estimates or the degree of credibility afforded the estimates by voters, nor do we know the extent to which voters were aware of the estimates.

15 Sources: Population 2001: 2001 Census of Population. Statistics Canada, 2001 Community Profiles. Released June 27, 2002. Last modified: 2005-11-30. http://www12.statcan.ca/english/Profil01/CP01/Index.cfm?Lang=E. Accessed various dates, 10/1/04–6/30/6. Total Expenditure: Service de l’information financière et de la vérification, Direction des finances municipales, Ministère des affaires municipales, Quebec. Spreadsheet provided to the authors.

16 We do not include a measure of the disparity in the degree of income inequality between unmerged and merged units. In this respect, we are consistent with the existing literature. Studies that include measures of income heterogeneity look at income disparities within the larger region and not at differences between income disparities within localities and average income disparities within the region. For example, Alesina, Baqir, and Hoxby’s (2004) measure of disparities in income distribution—the Gini coefficient—applies to counties and not to the individual localities within counties.

17 Source: 2001 Census of Population. Statistics Canada, 2001 Community Profiles. Released June 27, 2002. Last modified: 2005-11-30. http://www12.statcan.ca/english/Profil01/CP01/Index.cfm?Lang=E. Accessed various dates, 10/1/04–6/30/6.

18 Oates (1972) offers cultural (including language) indicators as demand determinants, the degree of variation in which may influence the desired degree of centralization within nation-states. As noted above, Oates finds little or no influence if these variables on revenue decentralization. He does, however, find a significant positive influence on decentralization exerted by an index of “sectionalism,”“a measure of the extent to which people in geographical subareas of a country identify self-consciously and distinctively with that area.” Oates’s findings with respect to racial diversity contrast with those of Alesina, Baqir, and Hoxby (2004) (for municipalities and school districts, within counties), noted above, and Martinez-Vazquez, Rider, and Walker (1997) (for school districts, within metropolitan areas and within states) both of whom find significant positive associations between measures of racial diversity and numbers of governmental units. While racial and language differences are qualitatively different, both represent dimensions of cultural diversity. On balance, especially in light of anecdotal portrayals of the high degree of sensitivity to linguistic differences that seems to be widespread within Quebec, it seems reasonable to control for language diversity in our model.

19 Source: Statistics Canada, 2001 Community Profiles. Released June 27, 2002. Last modified: 2005-11-30. http://www12.statcan.ca/english/Profil101/CP01/Index.cfm?Lang=E. Accessed various dates, 10/1/04–6/30/6.

20 Source: 2001 Census of Population. Statistics Canada, 2001 Community Profiles. Released June 27, 2002. Last modified: 2005-11-30. http://www12.statcan.ca/english/Profil01/CP01/Index.cfm?Lang=E. Accessed various dates, 10/1/04–6/30/6.

21 The only difference between Tables 2 and 3 is in regard to population size, with average population size greater at the register stage for communities opposing de-merger and smaller at the voting stage for communities opposing de-merger. We use population share rather than population size in our multivariate analysis since population share comports with our operating hypothesis that political influence in a merged unit will be proportionate to share of total population, not to absolute population size. A community may be large in population but relatively small compared to the total population of the merged entity. Indeed, we found no a priori reason to expect absolute population size to influence citizen preferences regarding de-merger.

22 As a matter of fact a value of rho equal to zero would mean that estimations using a Heckman’s selection model and OLS would lead to the same coefficients.