How to stop the race to the bottom

Abstract

Standard tax multipliers are a widespread feature of fiscal equalization systems. A simple theoretical model shows that actual tax multipliers respond positively to changes in standard tax multipliers. This theoretical prediction is tested empirically using data on municipalities in Germany. A quasi-experiment in the state of North Rhine-Westphalia is exploited to identify the incentive effect. The empirical results confirm that local business tax policy is shaped by standard tax multipliers. They provide a straightforward practical tool to avoid a race to the bottom in local business tax rates.

1 Introduction

It is a well-known normative principle among public economists that business taxation should not be decentralized to subnational levels of government. Otherwise, so the argument goes, local governments would engage in a harmful “race to the bottom” where they constantly try to undercut their neighbors’ business tax rates. Resulting tax rates would be inefficiently low (Oates 1972). A similarly widespread insight is that problems of overspending and reduced tax effort arise whenever budgeting involves a common pool of resources (see Raudla 2010 for a review on the use of the “budgetary commons” metaphor in existing literature).

Germany’s institutional setting involves business tax autonomy for local governments and a common pool of fiscal equalization transfers from the state to the local level. As a result, conventional wisdom points to overly low business tax rates as a likely outcome: Competition for mobile capital presumably pushes tax rates downwards. At the same time, one might suspect that the common pool of equalization transfers further reduces tax effort. Indeed, Köthenbürger (2002) shows that equalization schemes that rely on revenue equalization tend to reinforce tax competition. In contrast, fiscal equalization in the form of tax base or capacity equalization increases subnational tax rates and thus attenuates competitive forces, which may be efficiency-enhancing when competition effects are strong enough (Köthenbürger 2002; Bucovetsky and Smart 2006; Smart 2007). Municipal fiscal equalization in Germany adheres to the capacity equalization principle, which is also employed in the transfer systems of countries such as Canada and Australia. In such systems, jurisdictions’ tax bases are evaluated at a standard tax rate and compared to a benchmark level of spending or revenue to determine the size of the transfer.¹ The transfer to each jurisdiction decreases in its “fiscal capacity.”

In this paper, we argue that the so-called standard tax multipliers (fiktive Hebesätze or Nivellierungshebesätze) help to prevent both the race to the bottom and the raiding of the commons. Standard tax multipliers are employed in fiscal equalization schemes to calculate the fiscal capacity from taxes for which subnational governments enjoy tax autonomy. In the case of German municipalities, business tax is one of the most important components of fiscal capacity. The use of standard tax multipliers has the following effect: If the municipality’s actual business tax multiplier is smaller than the standard tax multiplier, the accounted standardized business tax revenue is greater than actual business tax revenue (and vice versa). While the effects of most mechanisms within the equalization system are unknown to the municipalities, the municipalities are well aware of the impact of the standard tax multiplier (i.e., the “overestimation” of tax revenue if the standard tax multiplier exceeds their business tax multiplier). Standard tax multipliers prevent municipalities from neglecting their own tax sources. They provide a signal for an “appropriate” tax rate level, which municipalities are incentivized to follow in order to maximize their revenue. Thus, practitioners at the state level have devised a clever mechanism to circumvent common pitfalls of local tax policy. Consistent with this argument, local business tax rates in Germany hardly appear to have been driven by a race to the bottom. Instead, they exhibited a steady upward trend over the past three decades.²

A growing body of empirical literature investigates the incentive effects of equalization systems on tax policy and demonstrates the positive impact of capacity equalization on local tax rates. Egger et al. (2010) exploit a change of the equalization formula in the state of Lower Saxony and show that this reform had a significant impact on municipalities’ business tax rates for four consecutive years. Büttner (2006) provides evidence that there is a positive relationship between the marginal contribution rate, defined as the rate at which an increase in the tax base reduces equalization transfers, and local business tax rates in the state of Baden-Württemberg. Smart (2007) investigates the effect of equalization among Canadian provinces, showing that an expansion of transfers leads to higher provincial tax rates.

This paper presents further evidence of the upward pressure that fiscal equalization exerts on local tax rates. We add to the literature by focusing the analysis on standard tax multipliers and using an innovative identification strategy with a new dataset. Standard tax multipliers are an institutional feature of any equalization scheme that relies on so-called representative tax systems, as well as being present in municipal fiscal equalization in all thirteen German territorial states.³ The hypothesis of standard tax multipliers as a driver of local tax policy has long been discussed in the applied literature on the evaluation of and reform options for fiscal equalization systems (e.g., Büttner et al. 2008; Parsche and Steinherr 1995; Goerl et al. 2013). Baskaran (2014) even takes this hypothesis as a given in his analysis of local tax mimicking by municipalities in Germany. He views a reform of standard tax multipliers in the state of North Rhine-Westphalia in 2003 as the cause of observable adjustments in actual tax multipliers. This is despite the fact that, to the best of our knowledge, an explicit test of this hypothesis is absent from the academic literature to date. We adapt the theoretical models used in Smart (2007) and Egger et al. (2010) to illustrate the interaction of local taxation and fiscal equalization. This allows us to derive the optimal business tax multiplier as well as the incentive effect of a change in the standard business tax multiplier. Beyond the mechanics exposed in the model, we believe that standard tax multipliers provide an easy-to-read signal to local policymakers. They view standard tax multipliers as a reference for an appropriate and politically feasible tax multiplier. In contrast to changes in eligibility criteria, adjustment levels, or marginal contributions rates, which may also influence local tax multipliers as shown in the previous literature, standard tax multipliers have the same magnitude as actual multipliers. As a result, changes in standard tax multipliers are easily translated into perceived necessary adjustments of local multipliers. As stated by Baskaran (2015) , hikes in standard tax multipliers also provide a window of opportunity for local officials to raise tax multipliers while deflecting the blame to the state level. Thus, we argue that changes in standard tax multipliers are more obvious and potentially more powerful trigger of local tax responses than previously analyzed fiscal equalization parameters.

Standard tax multipliers are often equal, or at least related, to the average of actual tax multipliers, creating an endogeneity problem in empirical analysis. A quasi-experiment in North Rhine-Westphalia allows us to solve this problem. Until 1995, North Rhine-Westphalia’s equalization scheme featured standard tax multipliers that were differentiated according to municipal population size. In 1993, the state constitutional court ruled that this arbitrary differentiation was not permissible. As a result, North Rhine-Westphalia’s state legislature had to adjust its municipal fiscal equalization scheme. The court ruling thus led to exogenous variation in the standard tax multiplier for small municipalities. The strict exogeneity of this reform is in contrast to other reforms where standard tax multipliers are adjusted to better reflect actual average tax multipliers.

Our empirical analysis is based on a balanced panel dataset of annual administrative data for all 396 municipalities of North Rhine-Westphalia. The dataset covers the time period from 1987 to 2002, thereby containing information on the pre-reform, reform and post-reform periods. It draws on a variety of official statistics data sources. The rich and unique dataset includes a number of municipal- and county-level control variables.

Our research design combines a municipal-level fixed effects model with a difference-in-differences approach, where local business tax multipliers are regressed on interaction terms between treatment groups and treatment points. Our identification strategy exploits the exogenous (quasi-experimental) variation of the standard tax multiplier for “small” municipalities induced by the reform of the North Rhine-Westphalian municipal fiscal equalization system in the mid-1990s. We find a positive effect of the standard business tax multiplier on local business tax multipliers, as predicted by theoretical considerations. The findings are robust to a number of alternative specifications.

Section 2 clarifies the institutional features of the German business tax and municipal fiscal equalization. Section 3 introduces the theoretical model. Section 4 explains our empirical approach and data. Section 5 presents the results of the empirical analysis. Section 6 concludes.

2 Institutional background

Germany’s federal structure is a key determining factor of the country’s fiscal landscape. The federal level, the three city-states and 13 territorial states, and the more than 11,000 municipalities each have differing degrees of tax autonomy over different taxes. For German municipalities, the business tax (Gewerbesteuer) and the equalization transfers (Schlüsselzuweisungen) provided to them by their federal state are two of the most important income sources.⁴ In 2013, municipal net revenue from the business tax and fiscal equalization transfers accounted for 15.8 and 14.3 % of aggregate municipal income, respectively (Federal Statistical Office 2014b).

2.1 Business taxation

It is a particularity of the German tax system that municipalities enjoy business tax autonomy. Each municipality sets its own local business tax multiplier (Gewerbesteuerhebesatz). In contrast, the business tax base and the basic tax rate (Steuermesszahl)⁵ are defined at the federal level. The resulting tax rate is determined by multiplying the local business tax multiplier with the basic federal tax rate. The business tax is charged on operating profits of corporate and non-corporate firms. In 2013, gross business tax revenue amounted to 43 bn EUR, making it Germany’s third most revenue-generating tax (Federal Statistical Office 2014a).

2.2 Municipal fiscal equalization

In 2013, municipal fiscal equalization transfers in Germany totaled 29.4 bn EUR (Federal Statistical Office 2014b). These transfers serve a double purpose. First, most German municipalities lack sufficient own revenue sources to fund their tasks. The transfers they receive from their respective federal state via its municipal fiscal equalization system thus serve a fundamental financing function. Second, the transfers are designed to reduce differences in municipalities’ capacities to provide public goods.

Municipal fiscal equalization systems function similarly in all German states. All of them employ the same basic mechanism of comparing a fictitious measure of “fiscal need” with a standardized measure of “fiscal capacity.” Total fiscal equalization transfers $\sum \nolimits _{i=1}^{I}T_{i}$ (Schlüsselmasse, i.e., the sum of all equalization transfers paid out in one year in the state in question) are predetermined. The fiscal equalization transfer $T_{i}$ of municipality i equals

$$\begin{aligned} T_{i}=\alpha (\beta N_{i}-C_{i})\quad \forall i\quad \text {with }\, \beta N_{i}>C_{i}. \end{aligned}$$

(1)

$T_{i}$ depends on the combined effect of the following factors: adjustment level $\alpha$, i.e., the degree to which the difference between fiscal need and fiscal capacity is equalized; fictitious measure of fiscal need, which is calculated by multiplying a fiscal need number $N_{i}$ by the basic amount $\beta$; standardized measure of fiscal capacity $C_{i}$.⁶

Municipalities whose fiscal capacity exceeds their fiscal need are called “abundant” and do not benefit from fiscal equalization transfers. The basic amount is determined via an iterative process and equals⁷

$$\begin{aligned} \beta =\frac{\sum \nolimits _{i=1}^{I}T_{i}+\alpha \sum \nolimits _{i=1}^{I}C_{i}}{\alpha \sum \nolimits _{i=1}^{I}N_{i}} \,\forall \,i\quad \text {with }~\beta N_{i}>C_{i}. \end{aligned}$$

(2)

While the derivation of the fictitious measure of fiscal need is negligible with respect to the focus of this paper, the derivation of the standardized measure of fiscal capacity is not. Fiscal capacity is the sum of standardized business and property tax revenue and the (unstandardized) municipal share of VAT and income tax revenue. To assure local tax multiplier autonomy, municipal fiscal equalization systems employ so-called standard tax multipliers to evaluate tax revenue from taxes for which the municipalities set tax multipliers (business and property tax). Standard tax multipliers are set by the respective federal states. Standardized business tax revenue $R_{i}^\mathrm{std}$ equals

$$\begin{aligned} R_{i}^\mathrm{std}=s\times \frac{R_{i}}{m_{i}} \end{aligned}$$

(3)

with $R_{i}:=$ business tax revenue, $m_{i}:=$ business tax multiplier and $s:=$ standard tax multiplier.⁸

If the actual business tax multiplier is smaller than the standard tax multiplier, the accounted standardized tax revenue is greater than the actual business tax revenue (and vice versa). While the effects of most mechanisms within the equalization system are unknown to the municipalities, they are well aware of the impact of the standard tax multiplier (i.e., the “overestimation” of tax revenue if the standard tax multiplier exceeds their business tax multiplier).

3 A simple theoretical model

To understand the incentive effect of standard tax multipliers, we develop a simple theoretical model of local taxation and fiscal equalization with two revenue-maximizing local jurisdictions. It is a version of the models employed by Egger et al. (2010) and Smart (2007), which we extend to include the standard tax multiplier as well as the basic amount. It allows us to derive the optimal business tax multiplier and the incentive effect of a change in the standard business tax multiplier. Suppose there are two municipalities i and j whose sole income sources are business taxation and fiscal equalization transfers. The business tax base $B_{i}$ of municipality i does depend not only on its own business tax rate $m_{i}$, but also on the one of municipality j, $m_{j}$:

$$\begin{aligned} B_{i}=B_{i}^{0}+\gamma m_{j}-\delta m_{i} \end{aligned}$$

(4)

where $B_{i}^{0}\ge 0$ and $\delta >\gamma \ge 0$. Tax revenue $R_{i}$ thus becomes

$$\begin{aligned} R_{i}=m_{i}\left( B_{i}^{0}+\gamma m_{j}-\delta m_{i}\right) . \end{aligned}$$

(5)

Fiscal capacity $C_{i}$ is

$$\begin{aligned} C_{i}=\frac{sR_{i}}{m_{i}}=s\left( B_{i}^{0}+\gamma m_{j}-\delta m_{i}\right) \end{aligned}$$

(6)

where s again denotes the standard tax multiplier.

Assuming that both municipalities are non-abundant, the respective fiscal equalization transfers $T_{i}$ are derived by inserting Eq. (2)⁹ into Eq. (1):

$$\begin{aligned} T_{i}= & {} \alpha \left[ \frac{\sum T_{i,j}+\alpha s\left( B_{i}^{0}+\gamma m_{j}-\delta m_{i}+B_{j}^{0}+\gamma m_{i}-\delta m_{j}\right) }{\alpha \left( N_{i}+N_{j}\right) }N_{i}\right. \nonumber \\&\left. -\,s\left( B_{i}^{0}+\gamma m_{j}-\delta m_{i}\right) \right] \end{aligned}$$

(7)

As an auxiliary assumption, suppose that both municipalities seek to maximize their revenue from taxes and transfers:

$$\begin{aligned} \underset{m_{i}}{\max \,}R_{i}+T_{i} \end{aligned}$$

(8)

The reduced-form equation for the optimal tax rate of municipality i then becomes:

$$\begin{aligned} m_{i}*= & {} \frac{1}{4\delta ^{2}-\gamma ^{2}}\left[ 2\delta B_{i}^{0}+\gamma B_{j}^{0}+\alpha s\bigg \{ \gamma \delta +2\delta ^{2}\right. \nonumber \\&\left. \left. +\,\frac{1}{N_{i}+N_{j}}\left( \gamma N_{j}(\gamma -\delta )+2\delta N_{i}(\gamma -\delta )\right) \right\} \right] \end{aligned}$$

(9)

This leads to the following first derivative with respect to the standard tax multiplier s:

$$\begin{aligned} \frac{\partial m_{i}^{\divideontimes }}{\partial s}=\frac{1}{4\delta ^{2}-\gamma ^{2}}\left[ 3\delta \gamma N_{i}+N_{j}\left( 2\delta ^{2}+\gamma ^{2}\right) \right] >0 \end{aligned}$$

(10)

Proposition

An increase in the standard tax multiplier increases the optimal tax multiplier chosen by the municipalities.

Given this relationship, the use of standard tax multipliers prevents municipalities from neglecting their own revenue sources and provides a clever way to circumvent the common pool problem. What is more, many municipalities consider the standard tax multiplier as a signal for their own tax policy. Even if—as is likely the case—local policymakers do not fully understand the intricacies of fiscal equalization and the effect of the basic amount, they recognize intuitively that they should respond to changes in the standard tax multiplier to avoid transfer losses. In contrast to changes in the adjustment level $\alpha$, which also induces tax multiplier reactions, changes in s are easily translated into appropriate adjustments of the local tax multiplier, as both have the same magnitude. Therefore, a race to the bottom in local business tax rates does not occur when standard tax multipliers are used in equalization.

4 Empirical approach

4.1 Quasi-experiment

We exploit a quasi-experiment in the state of North Rhine-Westphalia for an empirical test of our proposition. North Rhine-Westphalia offers a promising case to study given that it is the most populous German state with over 17 million inhabitants. Moreover, it stands out as a state where municipalities’ business tax multipliers are high relative to those found elsewhere in Germany. The same applies to its standard tax multipliers.

Like those in the other 12 territorial states, the 396 North Rhine-Westphalian municipalities receive state transfers through a municipal fiscal equalization system. Each year, several billion euros (8 bn in 2014) are paid out as equalization transfers. North Rhine-Westphalia currently sets a single standard tax multiplier with respect to the business tax. Until 1995, the equalization scheme featured standard tax multipliers that were differentiated according to population size. The fiscal capacity of municipalities with up to 150,000 inhabitants (“small” municipalities) was calculated using a standard tax multiplier of 350. The fiscal capacity of municipalities whose population size exceeded this threshold (“big” municipalities) was evaluated with a standard tax multiplier of 380. In 1993, the state constitutional court ruled that this arbitrary differentiation was not permissible (VerfGH 9/92, 22/92).¹⁰As a result, North Rhine-Westphalia’s state legislature was required to adjust its municipal fiscal equalization scheme. Standard tax multipliers for municipalities with less than 150,000 inhabitants were increased in three equal steps between 1996 and 1998 to reach the larger cities’ multiplier. This 30-points change amounted to an increase of 8.5 % in the standardized tax multiplier. The court ruling thus led to sizeable exogenous variation in the standard tax multiplier for small municipalities. To the best of our knowledge, this quasi-experiment has not been used in the literature to date.

4.2 Data sources

Our empirical analysis is based on a balanced panel dataset of annual administrative data for all 396 municipalities of North Rhine-Westphalia. The dataset covers the time period from 1987 to 2002, thereby containing information on the pre-reform, reform, and post-reform periods. It draws on a variety of official statistics data sources, namely North Rhine-Westphalia’s statistical office (IT.NRW), the Regional Database Germany (Regionaldatenbank Deutschland) and the Federal Employment Agency (Bundesagentur für Arbeit). The rich and unique dataset includes municipal business tax multipliers, inhabitants, income tax and VAT shares, employees at place of employment, GDP (at county level), disposable income of private households (at county level), municipal debt, tax bases and revenues from property and business tax, commuters and municipal surface area. There are 375 “small” and 21 “big” municipalities up until 1999. From 2000 onwards, one additional city has more than 150,000 inhabitants.

Table 1 provides summary statistics for the most important variables, which are reported in per capita terms with the exception of business tax multipliers. Between 1987 and 2002, business tax multipliers in North Rhine-Westphalia varied between 250 and 490, with an unweighted average of 373.28. Table 1 also illustrates some data availability issues. None of the tabulated control variables are available for all years. Municipal income tax shares, GDP and surface area have only been reported since 1992.¹¹ There are no data on the disposable income of private households before 1995 or on the number of employees before 1993. More detailed summary statistics are provided in Tables 6, 8 and 7 in Appendix.

Table 1 Summary statistics, 1987–2002

Figure 1 depicts the difference between average business tax multipliers of “big” and “small” municipalities between 1992 and 2002.¹² The three dashed vertical lines mark the three reform years where standard business tax multipliers for small municipalities were raised. As shown, the average business tax multiplier of big municipalities was more than 70 % points higher than that of small municipalities at the outset. During the three reform years, the difference in averages dropped sharply, to a level of 60 % points and below.

Fig. 1

Difference in average business tax multipliers between “big” and “small” municipalities, 1992–2002. Source: IT.NRW, own calculations

4.3 Empirical model

Our research design combines a municipal-level fixed effects model with a difference-in-differences approach. The dependent variable is the business tax multiplier $m_{i,t}$ of municipality i in year t. Our independent variables of interest are the interaction terms $\hbox {TG}_{i}\times \hbox {TP}_{t},\,t=1996,\ldots ,1998$ between treatment groups ($\hbox {TG}_{i}=1$ if population $\le 150,000$ and 0 otherwise)¹³ and treatment points ($\hbox {TP}_{1996}=1$ if $t=1996$, $\hbox {TP}_{1997}=1$ if $t=1997$, $\hbox {TP}_{1998}=1$ if $t=1998$ and 0 otherwise).

We include two types of control variables to adjust for observable time-variant differences between municipalities: $\mathbf{X_{i,t}}$ and $\mathbf{Z_{c,t}}$ represent column vectors of municipal-level variables (debt p.c., share of income tax etc.) and county-level variables (GDP p.c. etc.), respectively. Furthermore, we control for municipal and year fixed effects $(\lambda _{i},\varPhi _{t})$. The municipal fixed effects account for unobserved but time-invariant omitted municipal-level factors that may influence business tax multipliers. By adding year fixed effects to the regression equation, we are able to control for common shocks affecting tax rates across all municipalities in a given year. We use the following regression model with $t=1995,\ldots ,1998$:

$$\begin{aligned} m_{i,t}= & {} \alpha \hbox {TG}_{i}+\delta _{1996\,}\hbox {TG}_{i}\times \hbox {TP}_{1996}+\delta _{1997}\hbox {TG}_{i}\times \hbox {TP}_{1997}+\delta _{1998}\hbox {TG}_{i}\times \hbox {TP}_{1998}\nonumber \\&+\,\beta \mathbf{X_{i,t}}+\theta \mathbf{Z_{c,t}}+\lambda _{i}+\varPhi _{t}+\varepsilon _{i,t} \end{aligned}$$

(11)

where the error term $\varepsilon _{i,t}$ is clustered at the county level.

Our coefficients of interest $\delta _{1996},\,\delta _{1997}$ and $\delta _{1998}$ measure how the business tax multiplier differential between “small” municipalities (treatment group) and “big” municipalities (control group) changed c.p. between the reference year 1995 and 1996, and 1997 and 1998, respectively.

4.4 Discussion of identification

Our identification strategy exploits the exogenous (quasi-experimental) variation of the standard tax multiplier for “small” municipalities induced by the reform of the North Rhine-Westphalian municipal fiscal equalization system in the mid-1990s. In contrast to later changes to standard business tax multipliers, this reform was prompted by a court ruling and is therefore truly exogenous. The chosen identification strategy thus circumvents typical endogeneity concerns.¹⁴ The validity of identification hinges on the assumption of a common trend between treatment and control groups.¹⁵ We assume that business tax multipliers would have evolved in parallel in the absence of treatment (conditional on other included independent variables). Without treatment, $\delta _{1996},\,\delta _{1997}$ and $\delta _{1998}$ would have to be zero.

Differences in administrative status between the treatment and the control groups might pose a potential concern regarding this identifying assumption. All 21 cities in the control groups are cities with county status. Of the 375 municipalities in the treatment group, only two have county status, while the remaining 373 belong to a county. If there had been systematic differences in or changes to the financing structure or spending responsibilities of municipalities with as opposed to without county status during our sample period, this might constitute a violation of the common trend assumption. We know of no such major shifts during the period of interest. Moreover, the revenue sources of cities with county status are equivalent to those of municipalities belonging to a county: Both rely on the same types of taxes, fees and charges, transfers, etc. In contrast, counties are financed solely through state transfers and the Kreisumlage, a financial contribution levied from municipalities within the county. Through this levy, municipalities belonging to a county share the responsibility for financing county-level spending. Given this administrative and fiscal setup, we believe that our treatment and control groups are sufficiently comparable.

Systematic differences in the degree to which both groups suffer from fiscal distress and find themselves under the supervision of regulatory authorities might also bias our estimation results. In recent years, regulatory authorities have been bound by official decrees to ensure that local tax multipliers of municipalities operating under budget consolidation plans are higher or at least equal to average state-wide tax multipliers of municipalities in their population size range. This might induce upward movements in tax multipliers which are unrelated to standard tax multipliers. There are unfortunately no official records on municipalities with budget consolidation plans in the mid-1990s. However, according to the Ministry of the Interior, the practice of actively influencing tax multipliers is a relatively new phenomenon. To the best of their knowledge, no official decrees existed during our sample period that would have mandated regulatory authorities to make higher tax rates a precondition for the approval of budget consolidation plans. What is more, budget consolidation plans were much less widespread during our sample period than they are today. Thus, we are confident in the validity of the common trend assumption.

We investigate the common trend by plotting the development of the average business tax multipliers of “small” municipalities (treatment group) and “big” municipalities (control group) between 1987 and 2010 (Fig. 2). The former is represented by the gray dashed and the latter by the black dotted line. The corresponding standard tax multipliers are shown in gray (“small” municipalities) and black (“big” municipalities/ all municipalities). Figure 2 supports the common trend assumption. Both groups have seen a gradual upward trend since 1987. The development of their business tax multipliers has been similar for most of the time period. Visible exceptions with some convergence of averages occurred during the reform years 1996–1998 and 2003 (see also Fig. 1).¹⁶

Fig. 2

Development of average business tax multipliers and standard business tax multipliers, 1987–2010. Source: IT.NRW, own calculations

5 Results

5.1 Main results

Table 2 shows results for two regressions where $t=1995,\ldots ,1998$. We restrict our main analysis to the reform period as we expect municipalities to react instantaneously to changes in applicable standard tax multipliers. Specification I contains baseline results for a regression without any control variables apart from the usual municipal and year fixed effects. The regression displayed in specification II includes income tax shares, GDP and disposable income of private households, surface area and the number of employees at place of work (each per capita) as additional controls. In line with our expectations, the interaction terms $\hbox {TG}_{i}\times \hbox {TP}_{t},\,t=1996,\ldots ,1998,$ between treatment group and treatment point dummies are highly significant with positive estimated coefficients in both regression specifications. According to the baseline specification, business tax multipliers of small municipalities were about 3.7 % points higher in 1996 than in 1995 ($\delta _{1996}=3.694$), c.p. They rose by another 5.7 percentage points in the following year ($\delta _{1997}-\delta _{1996}=5.712$). A smaller adjustment of about 2.6 % points took place in 1998, the final year of the reform ($\delta _{1998}-\delta _{1997}=2.641$). Given the annual increase of the standard tax multiplier of 10 % points, the degree of adjustment of small municipalities’ tax multipliers is remarkable.

Table 2 Regression results

Adding time-variant controls slightly affects the coefficients of interest (specification II). Per capita income tax shares and GDP each turn out to be individually significant covariates. Per capita disposable income of private households, surface area and employees at place of employment are jointly significant with the remaining controls and further improve the goodness of fit as measured by the $R^{2}$. The results of specification II support the general magnitude and the direction of the reform effect. However, they also suggest that the development of business tax multipliers is affected by time-variant factors aside from the reform. We expect the common trend between treatment and control groups to hold conditional on these time-varying factors. Our results are stable across all tested model specifications.¹⁷

5.2 Robustness checks

To validate our results, we perform equivalent regressions using the full dataset where t runs from 1987 to 2002 and corresponding interactions terms $\hbox {TG}_{i}\times \hbox {TP}_{t},\,t=1987,\ldots ,1994,\,1996,\ldots ,2002$ and year fixed effects are added. Specification III of Table 3 shows the results of such a regression without any further control variables.

Table 3 Regressions results (extended time period)

The coefficients of interest, $\delta _{t}$, belonging to this regression are also illustrated in Fig. 3. Importantly, the interaction terms $\hbox {TG}_{i}\times \hbox {TP}_{t},\,t=1987,\ldots ,1994$ belonging to the pre-reform period are all individually and jointly statistically insignificant. In contrast, the interaction terms $\hbox {TG}_{i}\times \hbox {TP}_{t},\,t=1996,\ldots ,2002$ of the reform and post-reform period are all highly significant with positive coefficients, indicating an upward shift of business tax multipliers triggered by the reform. The estimated adjustment during the reform years 1996 to 1998 is exactly the same as in specification I of Table 2. In the years following the reform, estimated coefficients $\delta _{1999}$ to $\delta _{2002}$ remain fairly stable. This lends support to the notion of an immediate response to each annual change of the standard tax multiplier.¹⁸ Due to limitations in data availability (see Sect. 4.2), a regression using pre-reform data and a set of control variables is not possible. However, the analysis can be extended to post-reform years. This is done in specification IV of Table 3 where t runs from 1995 to 2002 and per capita income tax shares, GDP, disposable income, surface area and employees again have been included as control variables. Again, the reform effects are significant and their magnitude and direction are in line with our expectations.

As an additional robustness check, we rerun specifications I and II of Table 2, this time excluding municipalities that were abundant, i.e., did not receive any equalization transfers, at any point between 1995 and 1998. This reduces the number of municipalities in the treatment group to 316. There are 19 municipalities left in the control group. Municipalities that did not benefit from equalization transfers presumably faced weaker incentives to raise their tax multipliers following the increase in their standard tax multiplier. Some incentive effect remains as it is very hard, if not impossible, for most municipalities to predict whether their fiscal capacity might exceed their fiscal need in a given year. Nonetheless, we expect the estimated treatment effect to be stronger than in our baseline specification.

Fig. 3

Coefficients on interaction terms. Notes Dotted lines mark 95 % confidence intervals around point estimates.

Source: IT.NRW, own calculations

Table 4 shows the corresponding regression results. The estimated treatment effect is very similar and slightly more pronounced than in our baseline specifications, confirming our expectations.

Table 4 Regression results (excluding abundant municipalities)

5.3 Extension

Lastly, we adapt our model to test if “small” muncipalities’ reactions to the reform differed systematically depending on their pre-reform business tax multipliers. We expect to find a more pronounced effect for “small” municipalities with a “low” pre-reform business tax multiplier. We operationalize these considerations by distinguishing two groups within our original treatment group: Treatment group 1 consists of the 217 “small” municipalities whose business tax multiplier was smaller than 380 in 1995 ($\hbox {TG}1_{i}=1$ if population $\le 150{,}000$ and $m_{i,1995}<380$ and 0 otherwise). Treatment group 2 refers to the 158 “small” municipalities with business tax multipliers greater than or equal to 380 in 1995 ($\hbox {TG}2_{i}=1$ if population $\le 150{,}000$ and $m_{1995}\ge 380$ and 0 otherwise). The corresponding interaction terms are defined as $\hbox {TG}1_{i}\times \hbox {TP}_{t}$ and $\hbox {TG}2_{i}\times \hbox {TP}_{t},\,t=1996,\ldots ,1998$. We estimate the following regression equation with $t=1995,\ldots ,1998$:

$$\begin{aligned} m_{i,t}= & {} \alpha _{1}\hbox {TG}1{}_{i}+\alpha _{2}\hbox {TG}2_{i}+\sum _{t=1996}^{1998} \delta _{1,t}\,\hbox {TG}1_{i}\times \hbox {TP}_{t}+\sum _{t=1996}^{1998}\delta _{2,t}\,\hbox {TG}2_{i}\times \hbox {TP}_{t}\nonumber \\&+\,\beta \mathbf{X_{i,t}}+\theta \mathbf{Z_{c,t}}+ \lambda _{i}+\varPhi _{t}+\varepsilon _{i,t} \end{aligned}$$

(12)

Table 5 shows the results for differentiated treatment groups, with specification V displaying the regression without controls (except for the usual municipal and year fixed effects) and specification VI including the same time-variant controls as specifications II and IV.

Table 5 Regression results (two treatment groups)

In line with our expectations, we find a much stronger reform effect on the business tax multipliers of treatment group 1 than on those of treatment group 2. All interaction terms between treatment group 1 and treatment point dummies are highly statistically significant with positive estimated coefficients $\delta _{1,t}$. The magnitude of the total effect for treatment group 1 pinpoints the strength of the effect triggered by the reform (specification V: $\delta _{1,996}+\delta _{1,1997}+\delta _{1,1998}=47.943)$. Apparently, municipalities in treatment group 1 used the reform as an opportunity to raise business tax multipliers even beyond the 30 percentage point increase of the reform.

The size of the estimated coefficient on the interaction term between treatment group 1 and treatment point 1996 is rather low and close to the one of treatment group 2 (specification V: $\delta _{1,1996}-\delta _{2,1996}=0.951$). This is not the case in 1997 and 1998: According to specification V, the business tax multipliers of treatment group 1 were about 13.4 % points higher in 1997 than in 1995 ($\delta _{1,1997}=13.375$) and continued rising in 1998 ($\delta _{1,1998}=17.284)$.

In contrast, the estimated effects of the interaction terms between treatment group 2 and the treatment point dummies are rather stable ($\delta _{2,1996}=3.144,\delta _{2,1997}=3.956$ and $\delta _{2,1998}=4.855).$ Moreover, statistical significance of treatment group 2’s interaction terms is low compared with those of treatment group 1 and in case of $\delta _{2,1998}$ depends on the specification used.

In summary, we find that the rise of the standard business tax multiplier had an effect on the business tax multipliers of all “small” municipalities, but this effect was particularly strong for municipalities with a “low” pre-reform business tax multiplier (i.e., pre-reform business tax multiplier below post-reform standard tax multiplier).

6 Conclusion

Conventional economic wisdom suggests that decentralized business taxation and a common pool of equalization transfers among local jurisdictions should lead to a race to the bottom in local business tax rates. In practice, however, a simple institutional device, standard tax multipliers, is used to counteract downward pressure on municipal tax rates and tax effort. Standard tax multipliers are employed in fiscal equalization schemes in all German territorial states to assess a municipality’s fiscal capacity independently of its actual tax multiplier.

Using the case of North Rhine-Westphalia in the mid-1990s, this paper empirically analyzes the impact of standard business tax multipliers on municipal business tax policy. The results show that upward shifts in standard business tax multipliers lead to immediate upward adjustments in actual business tax multipliers. This is true for all affected municipalities. The reaction is more pronounced for municipalities whose business tax multipliers are below post-reform standard tax multipliers. The findings are robust to a number of alternative specifications. They also reflect the positive incentive effect derived from theoretical considerations.

Our results have important implications for the practical design of fiscal equalization schemes. They highlight the importance of the parameters of equalization systems for shaping local tax policy. Through its choice of the standard tax multiplier, a state can influence the level of municipal tax rates and the weight of competitive downward forces. Some states choose to set standard tax multipliers that are so low that they have virtually no signaling effect while others induce a race to the top in local taxation through regular adjustments of standard multipliers. This partially explains why there is far greater heterogeneity in business tax multipliers across federal states than within states in Germany.

By consequence, standard tax multipliers should be regarded as a tool for governments to shape lower-level tax policy, with important consequences for their own competitiveness.

Appendix

See Tables 6, 7 and 8.

Table 6 Summary statistics (means) by year, 1987–2002

Table 7 Summary statistics (means) for treatment group by year, 1987–2002

Table 8 Summary statistics (means) for control group by year, 1987–2002