Taxes, Agglomeration Rents and Location Decisions of Firms

Abstract

The goal of this paper is to analyze the individual impact of tax rates and agglomeration rents as well as their interaction on location decisions of manufacturing firms within Belgium. Theoretically, both location determinants may weaken each other’s impact. Using a unique 10-year dataset concerning the number of newly setup manufacturing firms at the sector level for 43 Belgian districts, we show that local effective tax rates have a negative impact on location decisions. Moreover, location-specific supply-side agglomeration rents attract new firms and their impact appears to be even stronger for more spatially concentrated sectors. Finally, we show that a higher effective tax rate in a district does not necessarily deter new firms in more agglomerated districts, pointing to the existence of taxable location-specific agglomeration rents.

Appendices

Appendix 1: Belgian Regions and Districts

Belgium has an area of 30,528 square kilometer and a population of 10.4 million. It consists of 589 municipalities which have representative democracies taking care of their own expenditures and revenues. Municipalities levy on average 20 different taxes that account for more than 40 % of local revenues, of which the surcharges on the federal income and property tax are the most important (Heyndels and Vuchelen 1998). These municipalities are grouped into 43 districts for administrative and election purposes. Again these districts are grouped into ten provinces with their own provincial governor and the provinces themselves are divided into three regions with their own regional parliament: Flanders, Brussels and Wallonia. The analysis in this article focuses on the 42 districts in Flanders and Wallonia as illustrated in the following table as well as the district of Brussels.

Region	District
Flanders	Aalst
	Antwerpen
	Brugge
	Dendermonde
	Diksmuide
	Eeklo
	Gent
	Halle-Vilvoorde
	Hasselt
	Ieper
	Kortrijk
	Leuven
	Maaseik
	Mechelen
	Oostende
	Oudenaarde
	Roeselare
	Sint-Niklaas
	Tielt
	Tongeren
	Turnhout
	Veurne
Wallonia	Aarlen
	Aat
	Bastenaken
	Bergen
	Borgworm
	Charleroi
	Dinant
	Doornik
	Hoei
	Luik
	Marche-en-Famenne
	Moeskroen
	Namen
	Neufchateau
	Nijvel
	Philippeville
	Thuin
	Verviers
	Virton
	Zinnik

Appendix 2: Regional Surcharges on Property Tax in Belgium, 2003–2005

The ETR or real tax burden of a firm differs across districts because of several reasons such as more tax evasion in districts with a less efficient local tax administration (Moesen and Persoon 2002), different tax rulings and differences in deductible local taxes.

In this appendix we will give an example on how the surcharges on property tax can influence the ETR. The general property tax in Flanders is 2.5 %, while the property tax in Wallonia and Brussels is 1.25 %. Both the province and the municipality can ask a surcharge on this percentage. Our example is for the Flemish and Walloon municipalities with the highest and lowest surcharge in 2006. For example, we assume the cadastral income of a building (2,600 m$^{3})$ to be 7000 EUR and calculate the difference in taxes paid. Note that the taxes that are paid can be deducted from the firm’s taxable profit.

Flanders

Regional property tax: $7000\times 2.5\,\%=175$ EUR

Lowest surcharge in for example $\hbox {Diksmuide} = 1250 \rightarrow 175\times 12.50=2188$ EUR

Highest surcharge in for example $\hbox {Mol} = 2250\rightarrow 175\times 22.5=3938$ EUR

Wallonia

Regional property tax: $7000\times 1.25\,\%=87.5$ EUR

Lowest surcharge in for example $\hbox {Lasne} = 1200 \rightarrow 87.5\times 12= 1050$ EUR

Highest surcharge in for example $\hbox {Huy} = 3100 \rightarrow 87.5\times 31=2713$ EUR

These amounts can be subtracted from the taxable base for the corporate tax calculation and therefore lower the ETR.

The average variation in surcharges across districts is large according to Fig. 8. Note that on average the surcharges in Flanders are lower than in Wallonia. Because of the higher property tax in Flanders, however, this does not automatically imply a lower total property tax cost in Flanders.

Fig. 8

Surcharges on property tax in 2006 in Belgium at district level

Appendix 3: $t$ Tests ETR/Roads per $\hbox {km}^{2}$ and ETR/Regional Property Tax for 2000 and 2005

Table 4 $t$ Tests ETR/roads per $\hbox {km}^{2}$—year 2000

Table 5 $t$ Tests ETR/regional property tax—year 2000

Table 6 $t$ Tests ETR/roads per $\hbox {km}^{2}$; year 2005

Table 7 $t$ Tests ETR/regional property tax—year 2005

Appendix 4: Calculation of Incidence Risk Ratios

We consider ETR to be our categorical variable and look at three categories: low, intermediate and high ETR. The intermediate ETR category is the reference and we therefore construct only two dummy variables. Variable ETR2 gets a value of 1 if we consider a low ETR district while variable ETR3 obtains a value of 1 if we consider a high ETR district. If both variables have a value of 0, we are therefore automatically considering an intermediate ETR district.

We re-estimate equation (2) as in column (4) of Table 2 including now two ETR-terms and two $\hbox {ETR/Aggl(S)}_\mathrm{d}$-interaction terms:

$$\begin{aligned} E(n_{d,s,t} )\!&= \!\hbox {exp}(\beta _1 {\textit{ETR}}2_{d,t-1} \!+\!\beta _2 {\textit{ETR}}3_{d,t-1} \!+\!\beta _3 {\textit{Aggl}}\left( D \right) _{d,t-1} \!+\!\beta _4 {\textit{Aggl}}(S)_{d,t-1}\\&\quad +\beta _5 \left( {{\textit{ETR}}*{\textit{Aggl}}(D)} \right) _{d,t-1} +\beta _6 \left( {{\textit{ETR}}2*{\textit{Aggl}}(S)} \right) _{d,t-1}\\&\quad +\beta _7 \left( {{\textit{ETR}}3*{\textit{Aggl}}(S)} \right) _{d,t-1} +\beta _8 {\textit{Aggl}}(S)_{s,t-1}\\&\quad +\beta _9 \left( {{\textit{ETR}}*{\textit{Aggl}}(S)_{s,t-1} } \right) +\beta _{10} ({\textit{Aggl}}(S)_{d,t-1} *{\textit{Aggl}}(S)_{s,t-1} )\\&\quad +\beta 'x_{d,t-1} +\gamma 'd_s) \end{aligned}$$

The estimation results for the coefficients we need in order to calculate the IRR are the following:

$\beta _1$	-0.1615
$\beta _2$	-1.1987
$\beta _6$	0.0181
$\beta _7$	0.0175

Finally, the IRR comparing high versus intermediate ETR districts is calculated using the following expression: $\hbox {IRR}=\exp (\beta _2 +\beta _7$*value $\hbox {Aggl(S)}_\mathrm{d})$ where we allow $\hbox {Aggl(S)}_\mathrm{d}$ to take on different values. The results of these calculations are to be found in Table 3 and Fig. 7 in the main text.

Appendix 5: Results with Location-Specific Agglomeration Rents Measured at District-Sector-Level

In Table 8, the coefficients of the variables that we are most interested in are indicated in bold.

Table 8 Poisson estimation results with bootstrap and mean-centering; location-specific agglomeration rents measured at district-sector-level