Abstract
This article explores the role of fiscal institutions for local land use. It argues that tax-base mobility results in an incentive to expand commercial and residential land use, which is mitigated by fiscal redistribution. These predictions are investigated empirically using a data set of German municipalities. To identify differences in the exposure to fiscal redistribution, I exploit institutional characteristics of fiscal equalization grants using a regression-discontinuity analysis. The results support the role of fiscal incentives for local land use regulation, as commercial and residential land use is expanded much faster, and agricultural land use declines more rapidly in municipalities exempted from fiscal redistribution.
1. Introduction
Expanded land use for settlements, including commercial and residential land use, is often criticized because it reduces open space and exerts adverse effects on the environment (Nechyba and Walsh 2004; Nilsson et al. 2013). The economic literature has emphasized that land use is only one dimension of local economic development, and efficient land use will have to trade off environmental and other concerns against the economic benefits of urban expansion (e.g., Brueckner 2000). Because land use is typically regulated by local governments, this suggests that these local governments need to find a balance between the conflicting interests in the electorate. On the one hand, local governments face demands to generate more revenues to expand public services and to attract firms to enhance employment opportunities. On the other hand, residents seek to maintain natural amenities, and there is pressure by environmental groups to restrict land use expansion. Finding a balance between these different demands might prove difficult in practice, and political economy considerations may give rise to concerns that local land use might be inefficient. Yet such concerns apply at all levels of government. At the local level, one could argue that competition between governments might actually support the efficiency of public policy, as has been noted by many writers in the tradition of Tiebout (1956).
The literature in public finance has emphasized that a precondition for efficient local policies is a proper set of policy instruments. This includes various tax instruments that allow local governments to provide an efficient level of public services without distorting locational efficiency (Wildasin 1986). In practice, local governments face limitations in the set of available instruments and, in particular, often rely on mobile tax bases. Policy decisions exert fiscal externalities, and the equilibrium is generally inefficient (for a survey, see Wilson 1999). This creates a role for a system of fiscal redistribution that mitigates or even cancels fiscal externalities (e.g., Köthenbürger 2002; Bucovetsky and Smart 2006). Against this backdrop, this article explores the implications of fiscal competition and redistribution for local land use policy.
The starting point for the analysis is the standard model of fiscal competition, where local governments attempt to attract mobile capital. The model is augmented by land, which serves as an amenity or, if assigned to a commercial land use, is an input to production. In this model, I discuss the trade-off a local government faces when deciding on the fraction of land devoted to commercial land use. The analysis shows that under fiscal competition, the local jurisdiction is subject to a fiscal incentive to expand commercial land use. An extension shows that a similar incentive may exist with regard to residential land use. However, when fiscal competition is mitigated by fiscal equalization grants, the fiscal incentive to expand land use is reduced or absent.
The empirical testing ground is land use among municipalities in Germany. These municipalities offer a promising case for studying fiscal incentives, since their main revenue source is a tax on a mobile base: the local business tax is levied on profits of local firms and establishments. Moreover, municipalities decide on land use patterns, including residential and commercial land use. To identify differences in fiscal redistribution among municipalities, I exploit institutional details of a system of fiscal equalization grants, which strongly redistribute revenues of receiving municipalities while others are exempted. The results show that the amount of land dedicated to commercial and residential land use expands faster and agricultural land use declines more rapidly if municipalities are exempted from fiscal equalization.
The contribution of this article is threefold. The first contribution is to study the role of local public finances and in particular of interjurisdictional fiscal competition as a driver of expanding commercial and residential land use. Although it is well established that this competition need not be confined to taxes, the literature clearly focuses on fiscal instruments (Agrawal, Hoyt, and Wilson 2022). The urban economics literature has noted that decentral regulation may encourage excessive land use (for a survey, see Blöchliger et al. 2017). The main criticism is that local property taxes provide an incentive for reducing the capital intensity of housing and thus foster spatial expansion of cities (e.g., Brueckner and Kim 2003; Song and Zenou 2006). The role of interjurisdictional competition for land use has mainly been discussed in the context of growth-control regulations, where the population externality contributes to a proliferation of these restrictions (e.g., Brueckner 1995, 1998). In contrast, this article focuses on the role of fiscal competition for expanding and restricting commercial and residential land use. Employing a regression-discontinuity (RD) analysis, the analysis adds to the recent empirical literature that uses quasi-experimental methods to identify fiscal competition (e.g., Isen 2014; Agrawal 2015, Eugster and Parchet 2019). Although the core of this article is concerned with the effects of fiscal competition on regulation, there is a close parallel to the literature that views local regulation as the outcome of locational competition more generally (e.g., Agrawal and Trandel 2019; Hansen, Miller, and Weber 2020).
The second contribution is to provide empirical evidence on the incentive effects of fiscal redistribution on local government policies. While those effects are discussed in the theoretical literature in local public finance and fiscal federalism (e.g., Bucovetsky and Smart 2006), empirical evidence has focused on fiscal instruments such as tax rates (e.g., Dahlby and Warren 2003; Buettner 2006; Egger, Koethenbuerger, and Smart 2010; Dahlby and Ferede 2016; Ferede 2017) or public expenditures (e.g., Matheson 2005; Hindriks, Peralta, and Weber 2008). An exception is Han and Kung (2015), who find that an increase of revenue sharing has induced Chinese prefectures to devote less land to commercial use, which is in accordance with the findings in this article.
The third contribution is to ascertain the role of fiscal competition and equalization as drivers of land use in Germany. The expansion of residential and commercial land use has been identified as a source of environmental damage, and the federal government’s sustainable development strategy is committed to coordinate attempts to curtail its expansion (Federal Government [Bundesregierung] 2016). Although it has been argued that the desire to raise tax revenues might fuel the expansion of commercial land use because of a “fiscalization” of land use (Wassmer 2002; Langer and Korzhenevych 2018), the role of fiscal redistribution has not been explored. The results of this article confirm that this redistribution tends to mitigate the expansion of commercial and residential land use in Germany.
2. Land Use Policy and Fiscal Redistribution
The theoretical discussion of the choice of local land use restrictions augments a simple model of fiscal competition with land. In a first step, I characterize the spatial equilibrium allocation treating policies as given, including the amount of land assigned to commercial use. This enables me to determine the amount of capital employed by local firms. In a second step, I discuss the effects of local policy. Specifically, I consider the effect of an increase of commercial land in a single jurisdiction on the capital-market equilibrium. Equipped with the understanding of how local governments’ choices affect equilibrium outcomes, I discuss the local policy choice. The goal of this analysis is to make empirical predictions about how the decision of a jurisdiction might change when it is subject to fiscal redistribution. An analysis of the competitive equilibrium strategies of all jurisdictions, which would be required to make welfare statements, is beyond the scope of this article, however.
Locational Choice
Commercial land use generates a profit (rent) to the owner. This profit varies with the input of labor and capital. Denoting land and capital used in jurisdiction i with Li and Ki and the number of immobile workers in this jurisdiction, each supplying one unit of labor, with Ni, the output is determined by a constant returns-to-scale production function F(Ki,Li,Ni). We set the output price to unity, and the profit from commercial land use in jurisdiction i is determined by the excess of output over cost of capital and compensation of workers (expressed in units of output):
1
where ri is the (before-tax) return on capital and wi is the wage rate. Profit maximization requires
2
and
3
Assuming for simplicity that labor supply is equal in all jurisdictions Ni = N, such that the setting is symmetric, equation [2] can be rearranged as a capital demand function for jurisdiction i:
4
The properties of this function are governed by the production function. It decreases in ri, ∂φi /∂ri = 1 / FiKK < 0, and, given input complementarity,1 increases in Li, ∂φi /∂Li = –FiKL / FiKK > 0.
I assume that capital is mobile such that the (after-tax) rate of return on capital is equal to the common rate ρ. Hence,
5
where τ is the tax rate. The capital-market equilibrium condition requires that input of capital in all M jurisdictions obeys
6
with K being the fixed capital supply in the economy. Equations [3]–[6] define a set of 3M + 1 conditions that determine equilibrium levels of factor prices wi, ri, the input of capital Ki, and the rate of return on capital τ + ρ for given Lj in all jurisdictions. Using equation [1], the profit of the landowner is determined. While the effects of a change in the local tax rate have been discussed at length in the literature (e.g., Wilson 1999), the next section focuses on a change in commercial land.
Effects of Increasing Commercial Land
Holding constant the choices of other governments, a local government’s decision regarding the amount of land assigned to commercial use affects the local demand for capital and the local wage rate. To determine the general-equilibrium effect of an increase in commercial land in jurisdiction i, I follow Wilson (1991) and consider the equilibrium net rate of return on capital in jurisdiction i as a function of the policy parameters in all jurisdictions. The effect of an increase of land Li in jurisdiction i on the net rate of return is positive, ∂ρ / ∂Li > 0.2 Hence, by taking a total differential of equation [4] and using equation [5], I obtain
From equation [3], the effect of an increase of commercial land on the wage rate is
Note that the effect on the wage rate has two components. The first reflects the direct productivity gain from using more commercial land. The second is an indirect effect that results from the increased input of capital.
Local Policy with Respect to Commercial Labor
Suppose public policy aims at maximizing the utility of an immobile worker residing and working in jurisdiction i. The utility of this worker is defined by a quasi-linear function:
7
where wi is the wage rate, u is utility from the consumption of a (pure) public good Xi, and v captures the benefit from land as an amenity with total area A, which is assumed to be the same across jurisdictions in line with the symmetry assumption. This benefit increases in the volume of land not assigned to commercial use, ν′ > 0. Provision of the public good is financed by the tax on capital, and the public budget constraint is
8
Optimization with regard to the amount of land assigned to commercial use requires 
Using the decomposition of the wage effect, the first-order condition becomes
9
This expression states that land use is optimal if the marginal benefit from land as an amenity (on the left side) is equal to its opportunity cost (on the right). The latter consist of three terms: positive direct and indirect effects of commercial land on the marginal product of labor and the benefit from higher tax revenues as the local tax base increases. The latter effect is absent if the tax rate is zero, and it increases in the tax rate. This is intuitive because an increase in the tax base exerts stronger revenue effects the higher the tax rate. The first term on the right side reflects the basic opportunity cost of using land as an amenity. The other terms on the right side are caused by relocation of capital. Although these terms add to the opportunity cost of using land as an amenity, they reflect a decline of the capital input in other jurisdictions and, hence, constitute negative externalities.
For each jurisdiction, the first-order condition defines a best response of the land use restriction given the choices of all other jurisdictions. The Nash equilibrium is characterized by the intersection of these response functions. Assuming that this equilibrium exists and is unique,3 it is socially inefficient because expanding commercial land by any local jurisdiction creates negative external effects on other jurisdictions (see Appendix A.1).
The notion of externalities of local land use restrictions is useful for interpreting the optimality condition [9]. The direct productivity effect of assigning more land to commercial use reflects the social opportunity cost of using land as an amenity. The indirect effects associated with mobility indicate that the private opportunity cost from the perspective of the local jurisdiction is higher than the social opportunity cost. Hence, from a social perspective, jurisdictions assign too much land to commercial use. Note that the inefficiency of the land use restriction is similar to the distortion of environmental policies discussed in Oates and Schwab (1988), although in their analysis, inefficiencies only arise from a fiscal externality.
Fiscal Equalization
The foregoing discussion of the fiscal incentive to expand commercial land use assumes that the local budget is exclusively financed by a tax on a mobile base. In practice, local jurisdictions often receive a substantial amount of intergovernmental revenue by grants from upper-level governments. Depending on how grants are determined, fiscal incentives change. In particular, Bucovetsky and Smart (2006) note that equalization grants would limit externalities from tax-base mobility.
Fiscal equalization can be introduced into the model by assuming that municipality i receives a grant equal to the difference between “fiscal need” and “tax capacity.” Formally, the equalization grant amounts to Zi = Gi – Ci, where Gi represents the fiscal need (i.e., a politically determined figure of what is considered a reasonable level of public expenditure per resident). Fiscal capacity is defined by Ci = ϑKi and determines the extent to which revenues from the capital tax are taken into account. In a way, ϑ is similar to the representative tax rate known from provincial fiscal equalization in Canada (e.g., Smart 2007) and determines the implicit marginal contribution to the equalization scheme. Taking the grant into account, the local public budget constraint is
10
Depending on the magnitudes of Gi and ϑ, at a given level of the tax base Ki, total revenues need not be different under equalization grants. However, changes in the local tax base Ki exert weaker effects on local revenue. This mitigates the fiscal incentive to expand commercial land use; with full equalization, the fiscal incentive is absent.
To derive the effect of fiscal equalization on the incentive to expand commercial land use, I assume that optimal policy maximizes the quasi-linear utility function [7] subject to the budget constraint including grants in equation [10]. The condition for the optimal amount of commercial land is
11
Comparison with equation [9] reveals the effect of fiscal redistribution on commercial land use. More specifically, suppose the fiscal equalization grant is revenue neutral in the sense that the same amount of public services Xi could be provided if the jurisdiction would choose the same amount of commercial land as in the optimum without fiscal redistribution. Because part of the tax revenue generated from expanding commercial land is compensated for by lower grants, the opportunity cost of using land as an amenity decreases and the optimal amount of commercial land is reduced.4 Note that fiscal redistribution only reduces the fiscal incentive to expand commercial land. The incentive to attract mobile capital to increase the local wage rate is unaffected. Hence, if this effect is strong, even full equalization is unlikely to result in socially efficient policies. Of course, if the fiscal equalization grant were not revenue neutral, the number of public services that can be provided at the same amount of commercial land use would change. In particular, if the equalization grant exerts a positive revenue effect in the local public budget, the expansion of public services would further contribute to a decline in the opportunity cost of land as an amenity. However, note that the empirical analysis is concerned with a local treatment effect at levels of tax capacity, where fiscal equalization does not exert net-revenue or income effects.
The discussion has focused on the regulation of commercial land use, but there are many other types of land use restrictions,5 including on the total amount of land available for residential developments, partly motivated by environmental objectives (Anas and Rhee 2006). In this case, similar trade-offs may emerge as in commercial land use policies. In particular, revenues from taxing capital would benefit if local labor supply increases with new residents. Under capital tax competition, jurisdictions may have an incentive to devote land for residential developments to attract mobile residents and, indirectly, capital. A formal derivation of this incentive is discussed in an extension with mobile residents in the Appendix A.2.
3. Empirical Methodology
The empirical analysis examines determinants of land use regulation using data for local jurisdictions. Based on the theoretical discussion, it aims to test whether the intensity of land use is affected by fiscal redistribution. To answer this question, I explore local land use policies in an institutional setting, where local jurisdictions are subject to a system of fiscal equalization and compare the outcome with land use policies if revenues are determined under conditions of local fiscal autonomy. This distinction is characteristic of municipalities in Germany, which are also responsible for land use regulation and therefore serve as this article’s empirical testing ground for the relationship between fiscal redistribution and land use regulation.
While the details vary, the basic setup of fiscal redistribution is the same in all major German states.6 Jurisdictions with tax capacity above a certain level of fiscal need are considered “abundant” and do not receive equalization grants. Jurisdictions with low tax capacity relative to fiscal need receive equalization grants that are inversely related to tax capacity. For municipalities receiving those grants, the degree of fiscal redistribution is usually very high: an additional euro of tax revenues results in a decline of fiscal equalization grants by 80 cents on average (Buettner 2006).7 Hence, the effect of an expansion of commercial and residential land use on net revenues is substantially reduced.
Because the grants are not paid to abundant municipalities, the fiscal equalization formula is indeed nonlinear. Actually, employing the foregoing notation, the equalization grant is determined by a function,8 Zi = max(Gi – Ci,0). This function has two properties that are important for the empirical analysis. First, as tax capacity is a linear function of the tax base Ci = ϑKi, grants are a function of the tax base as well.9 Importantly, the function determining the grants is continuous. Although this is obvious for Gi > Ci or Gi < Ci, it also holds at the threshold level of the tax capacity (Γi). Formally, this level is defined by a tax base Ki = Γi, where 0 = Gi – ϑΓi. Hence, at the threshold level of the tax base, the left and the right limits of the function are identical:
Second, the first-order derivative of the grant function with respect to the tax base is discontinuous at the threshold level, and the left and the right limits of the function differ:
and
As jurisdictions with tax capacity close to the threshold level have about the same amount of net revenues, the discontinuity is not associated with an income effect. However, the effect of a change in the tax base on net revenues differs strongly to the left and the right of the threshold (i.e., it depends on whether or not tax capacity exceeds fiscal need).
Thus, given the institutional setting, the log of the relative tax capacity ci = logCi / Gi can be used as an assignment variable in a sharp RD analysis. Hence, I use regressions fitted to each side of the cutoff point ci = 0 to provide estimates of the treatment effect at the cutoff point. Denoting the outcome variable with y, I estimate functions yi = αL + βLci + εi, ∀i with –h ≤ ci < 0 and yi = αR + βRci + εi, ∀i with 0 ≤ ci ≤ +h, where h denotes the bandwidth around the cutoff point. The resulting estimate of the treatment effect is the difference in the intercept αR – αL.
Using an RD design enables me to identify the local average treatment effect of being subject to fiscal equalization. The estimate indicates a change in the outcome variable that results when the municipality is subject to strong fiscal redistribution. Although the degree of fiscal redistribution is low for the municipalities not receiving equalization grants, it is not zero. At the same time, the degree of redistribution among municipalities included in fiscal equalization is very high but not complete. The degree of redistribution is roughly halved for the municipalities that are exempt from fiscal equalization.10
When specifying the RD estimator, an appropriate choice of bandwidth must be made. The estimate can become more precise if the bandwidth is large. However, depending on the true underlying relationship between yi and ci, the difference in the intercept terms delivers a biased estimate of the local treatment effect (Lee and Lemieux 2010) and the bias increases with the bandwidth. To find the right balance between precision and bias, I implement the bandwidth selection procedure by Calonico, Cattaneo, and Titiunik (2014). To check for robustness, I use squared and higher-order terms of the right-hand variable ci, include total jurisdiction size as a control variable, and explore the treatment effect estimates associated with placebo values of the cutoff point. Estimation is carried out using different specification of kernel weights K(ci / h) based on the relative tax capacity.
If local governments seek to avoid or even purposefully induce inclusion in fiscal redistribution, the number of observations would be systematically higher on one side of the threshold. In this case, a key identification assumption of the RD estimator would be violated (McCrary 2008). To check this, I follow Cattaneo, Jansson, and Ma (2020) and provide tests of whether the density of the assignment variable is continuous at the threshold.
For outcome variables, I use changes in the share of land assigned to commercial, residential, or agricultural land use between periods t + 1 and t. Focusing on the change of land use is useful, because a change of the assignment of land into commercial or residential uses is not easily revertible. Once space is devoted to commercial or residential uses, the landowner has a protected right to use it in this way, and hence, a reversal is only possible with the consent of both landowner and municipality and may require removing existing structures.11
The decision on land use is made by the respective municipal or city council. In contrast to the decision on the budget, land use decisions can be made during the year. However, the data on land use are collected only annually, which is why I use annual changes. The data monitors changes in land use regulation from decisions on individual parcels of land and from modifications of underlying development plans. Major regulatory changes, such as setting up a new development plan for the municipality as a whole, may have a considerable lead time. As I use annual data, this may lead to serial correlation in the residuals, suggesting that I use clustering at the municipality level.
The research approach uses the ability to precisely determine whether a jurisdiction is subject to fiscal equalization grants. However, assignment into the institutional treatment is based on annual data. This raises the question of how to deal with cases where the assignment varies from year to year. In particular, the tax capacity changes with the evolution of tax revenues. If the year used to determine the fiscal status shows relatively high revenues, a jurisdiction may be temporarily exempt from equalization even if it is included in all other years. Conversely, a jurisdiction might be subject to an adverse revenue shock. Consequently, it may change status and receive fiscal equalization grants temporarily, even if is exempt in the other years. In these cases, a forward-looking jurisdiction may not change its land use policy. Therefore, in the empirical analysis, I use data for several years (2008–2013) and focus on municipalities that are either subject to redistribution in all years considered or exempt from equalization in all years. For robustness testing purposes only, I include all municipalities.
4. Data
Municipalities form the lowest tier of governments in the German federation. They have a constitutionally guaranteed fiscal autonomy. This includes their own budget, which is financed by tax revenues from the local business tax, a property tax, and shares in federal taxes. Since the municipalities decide about the local rate of the business tax, they are traditionally engaged in local tax competition (e.g., Buettner 2001).12 Besides fiscal autonomy, municipalities are responsible for building permits and setting up zoning plans (e.g., Gunlicks 2003). This raises the question of the relationship between fiscal autonomy with mobile tax bases and land use policy (Götze and Hartmann 2021). However, with the exception of three urban states, the German states run systems of fiscal equalization that strongly redistribute revenues among municipalities.
Preconditions for using the sharp RD design are that a sufficient number of municipalities are exempt from fiscal equalization and it is possible to measure the assignment into treatment precisely. To meet these requirements, the empirical analysis focuses on the state of Bavaria, where all municipalities operate under the same equalization system, a substantial fraction of municipalities are exempt from fiscal equalization (abundant), and relative tax capacity is precisely reported in the official statistics.
Details about the data sources and the definitions are provided in Appendix B. The data refer to all 2,056 municipalities of Bavaria in the six years between 2008 and 2013. Data on land use refers to the entry in the official cadastral land register. This implies that the data directly reflect the regulation of the land use. For example, if the municipality decides that a plot of land that was previously assigned to agricultural use is devoted to commercial or residential purposes, this is reflected in the data accordingly as a commercial or residential land use. This holds regardless of whether the buildings required for a corresponding use have been erected and whether new businesses or households have moved in. All changes of land use in the data therefore reflect policy decisions. This facilitates the empirical analysis of land use policy.
Land use is reported for each parcel and aggregated at the municipal level.13 About half of all land is used for agriculture. About a third is covered by forests. The settlement area captures only about 12% of total land. While the largest fraction of this area (about 42%) is used for transport, the fraction of residential use is second, amounting to roughly a quarter of all settlement areas (about 24%). Residential land use includes buildings and areas predominantly used for housing. Commercially used land amounts to about 5% of the settlement area. This includes buildings and areas used for industrial production and other commercial purposes. A relatively large fraction (about 22%) is characterized as a residual category (other settlement area). A more detailed breakdown is not available at the municipal level.14
Descriptive Statistics
Next, I focus on commercial and residential land use. Because it is the biggest category of nonsettlement land use, I also explore effects on agricultural land use for comparison.15 Since the size of municipalities differs, I consider the different categories of land use and their change over time relative to the total area of each municipality.
Descriptive statistics for the size of municipalities and their land use are provided in Table 1. The share of land assigned to commercial uses is rather small. In 2013, on average about 53.1% of total area of municipalities is assigned to agricultural land use. In the years under consideration, from 2008 to 2013, the average share of land used commercially is about 0.6% of total area. The average annual change of this share of land is positive: on average, 0.01% of total land is added to commercial land use every year. Declines of commercial land use are relatively infrequent—less than 20% of observations report a decline. Putting the change of land use in relation to the fraction of land already devoted to commercial land, commercial land use increases on average by 1.8% every year.
On average, residential land use also displays increases every year. The mean annual increase of residential land is around 1% every year. Declines are even less likely than in the case of commercial land—less than 8% of observations report a decline. The table shows that the increases in commercial and residential land use are reflected in declining agricultural land use.
Average Change in Commercial Land Use, 2008–2013
Figure 1 is a map of the changes in commercial land use between 2008 and 2013. The map classifies the change in four groups with declining share, no or small increase, modest increase, and strong increase. The increase in commercial land use is a bit more intense in the southern part of Bavaria, but stronger increases are found in other regions. Figure 2 depicts the average change in residential land use. This map classifies the change in four groups with declining or constant shares and small, modest, and sharp increases. Residential land use shows stronger increases in the southern parts of Bavaria, in particular, close to the agglomeration around the city of Munich.
Table 1 also presents descriptive statistics on other characteristics of municipalities. The average population size slightly exceeds 6,000 residents, the largest city (Munich) has about 1.4 million residents. The average equalization grant in the sample is about €167 per capita with a maximum of about €1,015. The table reports a binary indicator for “abundant” jurisdictions receiving zero equalization grants. In the data set, about 14% of municipality-year observations fall in this category.
Average Change in Residential Land Use, 2008–2013
The tax capacity per capita is €586 on average. This includes revenues from the local business tax, the municipal shares of income and value-added taxes, and the local property tax.16 Because there are single data points with negative or very high revenues, typically reflecting legal disputes over the business tax filings of large firms, municipalities with relative tax capacity below the 1% percentile and those with relative tax capacity above 99% are excluded. Nevertheless, the descriptive statistics point at relatively strong cross-sectional variation in tax capacity. For some jurisdictions, tax capacity only amounts to €223 per capita, whereas others report €1,759 per capita.
Relative Tax Capacity of the Estimation Sample
The fiscal need amounts to €808 per capita, on average. It shows much lower dispersion than tax capacity, which reflects the redistributive nature of the fiscal equalization system. Note that fiscal need includes a basic allowance, featured in the theoretical analysis, and some additional allowances reflecting municipal spending obligations related to welfare aid and local unemployment.
The distribution of relative tax capacity is shown in Figure 3, which provides a histogram along with a kernel estimate of the density. While the plot documents that most jurisdictions display relative tax capacity below unity, the cutoff point is still well within the distribution.
Because the data refer to the six years between 2008 and 2013, the relative tax capacity of a municipality displays some variation over time. While most of the municipalities display a relative tax capacity below or above unity in all years, some municipalities change their fiscal position and receive grants in some but not all years. Table 1 reports that 1,555 municipalities (about three-quarters of all municipalities) have relative tax capacity below unity and receive grants in all years. A total of 146 municipalities (about 7.1% of all municipalities) always report relative tax capacity above unity and receive no equalization grants. Figure 4 is a map displaying the frequency of relative tax capacity at or above unity in the period under consideration.
Number of Years Relative Tax Capacity at or above Unity for a Municipality, 2008–2013
5. Empirical Results
The discussion of the empirical results begins with three sets of plots showing the distribution of outcome variables around the cutoff point. Since the forcing variable is right-skewed, all plots depict the forcing variable (relative tax capacity) in logs. If the relative tax capacity exceeds unity, or the log of the relative tax capacity exceeds zero, jurisdictions do not receive fiscal equalization grants. Hence, to the left of the threshold, jurisdictions are subject to fiscal equalization. Therefore, fiscal incentives to expand commercial and residential land use should be weak. Municipalities to the right of the cutoff point are exempt from fiscal equalization and face a stronger fiscal incentive to expand land use.
Figure 5a reports the means of annual changes in commercial land as a fraction of total land. Observations depict the means for an equal number of 20 bins on each side of the cutoff point.17 Whereas the change of commercial land use is close to zero to the left of the cutoff point, the means point at increases of commercial land use to the right. Figure 5b is based on the same data and reports the means for a smaller number of bins, which minimizes the integrated mean squared error of the prediction as suggested by Calonico, Cattaneo, and Titiunik (2014). Figure 5b also depicts the 95% confidence interval. The results are qualitatively similar. Note that the confidence intervals are larger to the right of the cutoff point, which reflects the smaller number of observations.
Change in Commercial Land Use with or without Fiscal Redistribution
Change in Residential Land Use with or without Fiscal Redistribution
Figure 6 reports RD plots for the annual changes in residential land as a fraction of total land. Again, the left plot (Figure 6a) is based on 20 bins left and right of the cutoff point. Figure 6b reports the means for the number of bins, which minimizes the integrated mean squared error of the prediction, along with the confidence interval. Both show that residential land increases stronger to the right of the cutoff point. Compared with commercial land use, the treatment seems slightly more local to the discontinuity.
Figure 7 turns to agricultural land. Because there is no fiscal incentive to increase this type of land use, I do not expect to observe increases to the right of the cutoff point. Instead, as the expansion of residential and commercial land use comes at the expense of other types of land use, a decline of agricultural land seems likely. Again, the left plot (Figure 7a) is based on 20 bins left and right of the cutoff point. Figure 7b reports the means for the number of bins, which minimizes the integrated mean squared error of the prediction, along with the confidence interval. Both show a stronger decline of agricultural land to the right of the cutoff point, at least close to the cutoff point.
Change in Agricultural Land Use with or without Fiscal Redistribution
All plots support the view that jurisdictions exempt from fiscal equalization tend to expand the use of land used for commercial and residential purposes more than the others. To obtain estimates of the actual magnitude of these effects and to provide specification tests, I turn to the regression analysis.
Table 2 reports point estimates for the treatment effect on the change in the share of land assigned to commercial use and related statistics. The table reports 10 different estimates, all obtained by local RD analysis but using different specifications. If local governments seek to avoid or purposefully induce inclusion in fiscal redistribution, many observations would systematically be found on one side of the threshold. For each specification, therefore, the table reports a test statistic for the continuity of the density of the assignment variable at the threshold.
The first column reports results from a basic setting with bandwidth set to 25 log points. This restricts the analysis to a subsample comprising only around 18% of all observations. More observations are left of the cutoff: 1,172 observations have fiscal need exceeding tax capacity and hence are subject to redistribution, 260 observations are “abundant” (i.e., have tax capacity exceeding fiscal need and thus exempt from fiscal redistribution). However, the manipulation test statistic proves insignificant, indicating that the continuity of the density of the assignment cannot be rejected. The local regressions use linear polynomials and a triangular kernel. The resulting point estimate of the treatment effect indicates that the expansion of commercial land use is faster in the abundant municipalities: annual expansion is found to exceed the control group by 0.0275% of total area. Given that the average expansion of commercial land use in the data set amounts to 0.01% of total area, this is a sizable effect, indicating that the speed of expansion of commercial land use is two to three times higher. Based on the cluster-robust standard error, the effect is significantly different from zero. I also report a bias-corrected standardized test statistic following Calonico, Cattaneo, and Titiunik (2014), which shows a p-value below 5%.
Columns (2) and (3) report alternative specifications, which use the same bandwidth and degree of polynomial but different kernels. Based on the rectangular kernel, as reported in column (2), the point estimate is slightly reduced. If a parabolic kernel is used, the point estimate is more similar. This suggests that the treatment effect is local to the discontinuity.
Regression Discontinuity Estimates for Commercial Land
Column (4) reports results of specifications where the bandwidth is not fixed arbitrarily but chosen to strike an optimal balance in the bias-variance trade-off that emerges under a first-order polynomial following Calonico, Cattaneo, and Farrell (2020). The procedure uses a slightly larger bandwidth and yields quite similar results as in column (1). Because land use decisions of neighboring municipalities may be correlated, column (5) reports results of a specification where the standard errors are clustered by county rather than municipality. This robustness check delivers smaller standard errors suggesting that spatial correlation, if present, does not create a risk of over-rejecting the null.
When higher-order polynomials are fitted, the optimal bandwidth is dramatically enlarged. When quadratic functions are fitted to the left and right of the cutoff point, as shown in column (6), the optimal bandwidth is set to 43 log points and the number of observations increases to almost 42% of the sample. When cubic or quartic polynomials are fitted, as shown in columns (7) and (8), respectively, the majority of municipalities are included. While qualitative results are similar, due to the higher data demands, the results obtained with higher-order polynomials should be interpreted with caution; Gelman and Imbens (2019) suggest relying in general on local linear or quadratic polynomials. Column (9) reports results based on first-order polynomials, which include the (log) size of the municipality as a covariate. The treatment effect estimate proves robust, indicating that the information about the size of the municipality can safely be omitted.
Specifications (1)–(9) exclude all municipalities switching their status of abundancy over time. As discussed, if switching is temporary, given the difficulty to revert an expansion of land use, this seems reasonable. However, if the change in status is expected to be permanent, switching the status might well exert effects on land use. As a robustness check, therefore, column (10) reports results where municipalities that change their status are also included in the sample. While the number of observations increases, the treatment effect turns out to be smaller. This supports the view that at least some of the switching is expected to be temporary.
Note that regardless of whether switching municipalities are excluded or included, the data does not point at self-selection into or out of the treatment. Defining bins of +/− 0.5 log points around the cutoff, in the sample excluding switching municipalities, 31 observations are at the left, 29 are right of the cutoff. If switching municipalities are included, the observation numbers are 38 and 34, respectively. The absence of manipulation is also indicated by the formal tests of whether the density of the assignment is continuous at the threshold.
For a robustness check, I explore the estimation of treatment effects using placebo values of the cutoff point. As expected, for cutoff points with relative tax capacity of −20 log points or −40 log points counterfactual treatments show no effects on commercial land use.18
These results suggest that if municipalities are exempt from fiscal redistribution, they intensify commercial land use. This confirms the theoretical view that the expansion of commercial land use by municipalities is advantageous in a setting with fiscal competition because it attracts mobile capital. The expansion of commercial land use is only one means of improving locational attractiveness. Another means, emphasized in the fiscal competition literature, would be to reduce the tax rate on mobile capital. Indeed, the municipalities under consideration here have discretion in adjusting the local business tax rate. The existing empirical literature for German municipalities has confirmed a negative effect of the exposure to fiscal competition on the local tax rate (e.g., Buettner 2006; Egger, Koethenbuerger, and Smart 2010). It might be interesting to see if the data used here confirms this finding. However, due to a tax reform that was implemented in 2008, right at the beginning of the period under study, this is not clear. This reform introduced a substantial income tax credit for local business tax payments.
For the vast majority of municipalities, this created an opportunity to raise the tax rate without increasing the burden on businesses, which are subject to the income tax (Buettner, Scheffler, and von Schwerin 2014).19 At any rate, estimates reported in the Appendix show that the preferred RD specification does support a smaller increase in tax rates of municipalities exempt from redistribution.20
Table 3 turns to the treatment effect of fiscal competition on residential land use. As before, the table shows point estimates using the RD method employing a set of different specifications. Column (1) shows the treatment effect obtained from a specification with bandwidth fixed to 25 log points using a triangular kernel and a first-order polynomial. The positive coefficient points to a faster expansion of residential land use in the absence of fiscal equalization. More specifically, the average increase of residential land exceeds the control group by 0.0288% of total area. Given the average expansion of residential land use among all municipalities by 0.0316% of total area, this indicates that the speed of expansion almost doubles if municipalities are exempted from fiscal equalization. Based on the cluster-robust standard error and the bias-corrected standardized test statistic following Calonico, Cattaneo, and Titiunik (2014), the effect is weakly significant with a p-level of about 10%. A rectangular kernel results in a smaller effect (see column (2)), while the parabolic kernel (see column (3)) yields a similar effect, indicating that the treatment effect is local to the discontinuity. If an optimal bandwidth is chosen with the method of Calonico, Cattaneo, and Farrell (2020), the treatment effect is found to be larger and the p-level declines (see column (4)). Based on the point estimate for the treatment effect of 0.0377% of total area, the speed of expansion more than doubles. As before, results are robust when taking account of possible spatial correlation. Allowing for higher-order polynomials is associated with much larger optimal bandwidth. Nevertheless, treatment effects are similar though estimated with lower precision. Column (9) reports an estimate including the total size of a municipality (in logs) as a control. Point estimate and inference are not much affected. The specification reported in column (10) includes municipalities that are only exempt from fiscal redistribution in some but not all years. The treatment effect is much smaller and no longer statistically significant. For residential land use, estimates of treatment effects using placebo values of the cutoff point also do not show any effects.21
Regression Discontinuity Estimates for Residential Land
Regression Discontinuity Estimates for Agricultural Land
Table 4 shows treatment effects on agricultural land use. The table uses the same specifications as in the foregoing analyses. Column (1) shows that a basic RD estimate based on a bandwidth of 0.25 log points, a local linear polynomial, and triangular kernel delivers a negative treatment effect, pointing to a decline of agricultural land use. Given the average decline in agricultural land use of −0.15% of total area, the point estimate suggests that the speed of decline doubles if a municipality is exempted from fiscal equalization. Across different specifications, the quantitative estimates of the treatment effect vary, but the qualitative results prove robust. The qualitative differences across specifications are similar to the analysis of commercial and residential land. The treatment effect is local to the discontinuity, is robust against potential spatial correlation, and is estimated with less precision when higher-order polynomials are used. The treatment effect is robust against inclusion of total size (in logs) as a control and disappears if the sample is extended to include municipalities exempt from fiscal redistribution in some but not all years. Moreover, robustness checks conducted using placebo values of the cutoff point do not show any effects on agricultural land use.22
6. Conclusions
The theoretical analysis in this article considers the land use policy of a local government relying on revenues from a tax on mobile capital. Optimal land use equates the marginal benefit from land that serves as an amenity with its opportunity cost. This cost is determined by the wage increase that results from an expansion of land use and by a fiscal incentive. The latter stems from the taxation of mobile capital. In particular, an expansion of commercial land use attracts mobile capital, tax revenue increases, and the supply of public services expands. The link between land use and tax revenues fuels concerns about an inefficient expansion of land use by jurisdictions competing for mobile capital. Under fiscal redistribution, however, the incentive for land expansion is reduced. An extension shows that there is also a fiscal incentive to expand residential land use to attract mobile residents.
As an empirical testing ground for analyzing the fiscal incentives to expand commercial and residential land use, the article considers the land use policy among municipalities in Germany. To identify differences in the environment faced by the municipalities, I exploit institutional characteristics of the fiscal equalization system. In particular, I make use of the discontinuity in the system of fiscal equalization, which involves strong fiscal redistribution among the majority of municipalities, but at the same time exempts municipalities with higher levels of tax capacity from redistribution. This causes strong institutional variation in the degree of fiscal redistribution, which enables me to base the empirical analysis on a sharp RD design.
The empirical analysis uses a data set covering the evolution of official land use between 2008 and 2013 in the 2,056 municipalities of Bavaria, a large German state. The results confirm a significant effect of fiscal equalization on the land assigned for commercial and residential use. Specifically, I find that land assigned to commercial use increases two to three times faster in municipalities exempted from fiscal equalization than in those receiving equalization grants. As for residential land, the results show that municipalities exempt from fiscal equalization also expand this form of land use about two times faster than other municipalities. At the same time, land allocated to agriculture decreases faster in municipalities exempt from fiscal equalization. These results suggest that municipal fiscal equalization severely curtails the fiscal incentives to expanding land use in Germany.
By relying on official land use data, the analysis reveals the effects of fiscal redistribution on municipal land policy. The effect on actual land use may be different, however, because once land is formally assigned to a specific use, private-sector action is usually required to enable the actual use. Whether such action actually occurs in a timely way or whether the policy decision to expand commercial or residential land is ineffective cannot be examined in the context of this analysis and is left to further research.
Because my empirical results on the effects of fiscal redistribution on land use are derived from German data, the question arises whether similar effects hold in other settings as well. Regarding commercial land use, the strong effects may reflect the heavy reliance of German municipalities on the local business tax. In more conventional settings, where the property tax is the main source of local tax revenue, the fiscal incentive to expand commercial land might be smaller, and perhaps a stronger incentive emerges with regard to residential land. If local sales taxes are important, the fiscal incentive to increase land for shopping opportunities may be strong in particular. At any rate, the analysis in this article supports the view that the “fiscalization” of land use is a relevant concern.
The view that the strength of the fiscal incentives hinges on a country’s tax system is in accordance with the tax competition literature, which has pointed out that inefficiencies are primarily driven by the lack of suitable tax instruments. However, the theoretical analysis has shown that externalities from land use policies are not only the result of fiscal incentives. If local policy aims at increasing wages of immobile workers, expanding commercial land use might be a way to attract investments that result in higher wages or employment. Because the identification strategy here focused on a fiscal incentive, our analysis is silent on whether this incentive is empirically relevant. If this were the case, socially inefficient policies would result even in the absence of fiscal competition.
From a broader perspective, my findings strengthen the view that lack of harmonization of land use regulation can be an important driver of inefficient land use (Burchfield et al. 2006). In fact, it is tempting to relate the finding of a strong sensitivity of land use policy to local fiscal competition to the size of the municipalities under consideration. The vast majority of municipalities in the data are very small, with average population size of about 6,000 inhabitants. Hence, the empirical setting is characterized by a heavily decentralized land use policy, which is prone to socially inefficient outcomes because of externalities.
Acknowledgments
For data support, I thank Martin Bürner from the Bavarian Office for Statistics and Data Processing. For helpful comments and discussions, I thank David Agrawal, Calin Arcalean, Martin Altemeyer-Bartscher, Jan Brueckner, Wolfgang Buchholz, Bill Hoyt, Jay Wilson, two anonymous referees, and participants at various seminars and conferences, including the ZEW public finance conference in Mannheim, a conference on fiscal competition and fiscal federalism at the University of Kentucky's Martin School of Public Policy and Administration, the meeting of the Verein fuer Socialpolitik in Augsburg, the conference of the Canadian Economic Association, seminar at the University of Regensburg, and the IWH research seminar.
Footnotes
Supplementary materials are available online at: https://le.uwpress.org.
↵1 Formally, FiKL > 0, FiKN > 0, FiNL > 0.
↵2 To derive the effect of commercial land on the equilibrium return on capital, I consider the total differential of the capital-market equilibrium condition [6]:
and solve for .↵3 The literature on capital tax competition has discussed sufficient conditions under which the intersection of the response functions will gives rise to a unique Nash-equilibrium; for example, Taugourdeau and Ziad (2011).
↵4 If the local jurisdictions have discretion in setting the local tax rate, fiscal equalization grants may exert an effect on the local tax rate. In fact, they provide an incentive to increase the local tax rate (Bucovetsky and Smart 2006), which may partly offset the effect a higher marginal contribution rate.
↵5 For a taxonomy of land use restrictions in the United States, see Quigley and Rosenthal (2005).
↵6 The system differs in the three urban states.
↵7 In terms of the notation in Section 2, the degree of redistribution is determined by ϑ / τ. If ϑ = τ, there is full redistribution, implying that revenues are equalized except for differences in fiscal need.
↵8 In the context of German municipal fiscal equalization, for municipalities that receive fiscal equalization grants, the basic fiscal need Gi is equivalent to the what is referred to as main component (Hauptansatz) divided by the rate of equalization (Ausgleichssatz).
↵9 In German municipal equalization, the marginal contribution rate ϑi is equivalent to the standardized tax rate (Nivellierungshebesatz) for the business tax divided by the rate of equalization.
↵10 For a municipality with median tax rate receiving equalization grants, the degree of redistribution is about 80%. If the municipality is considered abundant and does not receive equalization grants, the degree of redistribution is about 40%. For details on the Bavarian equalization system, see Bavarian Ministry of Finance 2022.
↵11 Referring to the change of land use is also common practice by geographical indicators of “land consumption,” for example, Melchiorri et al. (2019).
↵12 Recently, however, federal policy has implemented measures to reduce tax competition. In 2004, a minimum tax rate of 7% was introduced. Since 2008, up to a tax rate of 13.3%, business tax payments can be credited against the income tax liability (Buettner, Scheffler, and Schwerin 2014).
↵13 Note that because of migration of the statistic from the digital land register toward a geographic information system, I exclude more recent years as well as observations for two counties in 2013, which first shifted to the new reporting standard.
↵14 The residual category refers to space used for buildings and open space, which are not primarily residential or commercial land. This includes space for public institutions, including public services, churches, and cemeteries. This heterogeneous category also includes land used for commercial activities such as fairgrounds, exhibition sites.
↵15 The results for nonsettlement land use are qualitatively and quantitatively similar if forestland is included. However, since the proportion of forested areas tends to be particularly high in peripheral regions, where the incentive to increase commercial or residential land may be particular small, the results focus on agricultural land use.
↵16 Appendix Table C.1 provides figures for the revenue shares of the main taxes. Note that the German property tax is relatively unimportant as a revenue source. This partly comes from the fact that assessment is held fixed since the 1960s.
↵17 Plots were constructed using the rdplot command of Stata’s RD package.
↵18 Results are provided in Appendix Table C.2, columns (1) and (2).
↵19 The tax credit applies to business taxes paid up to a tax rate of 13.3% (equivalent to a collection rate (Hebesatz) of 380%). This corresponds to the 95th percentile of the tax rate distribution in 2008, implying that this regulation applies to almost all Bavarian municipalities.
↵20 If an optimal bandwidth is chosen using the method of Calonico, Cattaneo, and Farrell (2020), with linear polynomials and a triangular kernel, the treatment effect points at a decline in the business tax rate by about 0.5 log points; see Appendix Table C.3.
↵21 Results are provided in Appendix Table C.2, columns (3) and (4).
↵22 Results are provided in Appendix Table C.2, columns (5) and (6).
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.