Spatial fiscal interactions among French municipalities within inter-municipal groups

ABSTRACT

Based on a unique data set of French municipalities and a large number of budgetary variables, we estimate the extent of spatial fiscal interaction among the 33,484 French municipalities in 2008 by accounting for inter-municipal cooperation. Using a spatial autoregressive model with inter-municipal group fixed effects, we show that spatial interactions among French municipalities are inflated by correlated effects that affect similarly municipalities that cooperate together. Removing these confounding effects leads to considerably smaller positive spatial interactions for tax decisions and even negative ones for capital expenditures. In addition, we observe a clear distinction between complementary choices on current budget items and substitutable choices on capital budget items.

KEYWORDS:

• Inter-municipal cooperation
• spatial autocorrelation
• fiscal decisions interaction

JEL CLASSIFICATION:

• C31
• H71
• H72

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed here.

Notes

¹ The theoretical implications of these spillovers (see Williams 1966; Pauly 1970; Oates 1972; Boskin 1973; for seminal papers) have been tested recently by, among others, Case, Rosen, and Hines (1993), Baicker (2005), Lundberg (2006) or Solé-Ollé (2006).

² The competition to attract the mobile tax base can occur though taxes, resulting in a ‘race to the bottom’ of tax rates (Wilson 1986; Zodrow and Mieszkowski 1986) or through expenditures (Keen and Marchand 1997). Numerous empirical papers examine the evidence for fiscal competition (see, for instance, Brett and Pinkse (2000) for Canada; Ladd (1992) and Brueckner and Saavedra (2001) for US; Revelli (2001) and Buettner (2001) for Germany; Heyndels and Vuchelen (1998) for Belgium; Feld and Kirchg¨assner (2001) for Switzerland; Feld, Josselin, and Rocaboy (2002), Jayet, Paty, and Pentel (2002), Leprince, Paty, and Reulier (2005) and Leprince, Madiès, and Paty (2007) for France). The paper by Chirinko and Wilson (2013) is one of the few papers that find a negative slope of the reaction function of home tax policy to neighbouring tax policies. It shows how the reaction function changes from positive to negative as one controls for aggregate shocks and allows for delayed responses to neighbouring tax changes.

³ Originally formulated by Salmon (1987), and first modelled theoretically and estimated empirically by Besley and Case (1995), yardstick competition describes the competition among politicians that originates from the voters’ comparative performance evaluations.

⁴ See also Bramoull´e et al. (2009) for a discussion on identification.

⁵ The local business tax was replaced in 2010 by a new tax called the Territorial Economic Contribution (CET).

⁶ The 6 ‘villages died for France’ are Beaumont-en-Verdunois, Bezonvaux, Cumières-le-Mort-Homme, Fleury-devant-Douaumont, Haumont-près-Samogneux and Louvemont-Côte-du-Poivre.

⁷ M14 is the budgetary instruction applicable to all municipalities and inter-municipal groups.

⁸ The property tax rate on undeveloped land can exceed 100%, allowed for the municipality.

⁹ For instance, the standard deviation inside inter-municipal groups is on average 0.14 for the dummy variable ‘Mountain’ and 0.07 for the share of tourist accommodation.

¹⁰ We acknowledge a tax base effect which influences the tax receipt. However, our results are robust (see Appendix) when we only focus on tax rates.

¹¹ See Brueckner (2003) for a survey in the context of spatial fiscal interdependence.

¹² These identification conditions are also met for all the other budget items considered in this article, except in two instances where the restriction $\rho {\beta }_1+{\beta }_2=0$ could not be rejected with a p-value of 0.15. For these two cases, results should be considered with care.

¹³ We thank an anonymous Referee for suggesting this example.

¹⁴ We also estimated all the models by removing the outliers in the database. Specifically, for each dependent variable, we removed the observations above 97.5% and below 2.5% of the distribution. It appears that removing outliers has very little effect on the previous results, in either qualitative or quantitative terms.

¹⁵ The results for all the other dependent variables are available upon request from the authors.

¹⁶ The maps displaying the ‘départements’ and the ‘zones d’emploi’ are in the appendix (See Figure A1 and Figure A2 in the Appendix).