Wage Incidence of Corporate Taxation: A Micro Level Approach Using Linked Employer-Employee Data

Transcription

1 Wage Incidence of Corporate Taxation: A Micro Level Approach Using Linked Employer-Employee Data Clemens Fuest, Andreas Peichl z, Sebastian Siegloch x, Work in progress, very rst version (as of October 28, 2011). Do not cite or circulate. { Abstract: The question of who bears the burden of corporate taxes is not a new, though still a very topical and controversial one. In this paper we provide empirical evidence on the wage incidence of the German business tax, which is raised at the community level. For our analysis, we use a very rich administrative linked employer-employee dataset and link it to data on the business tax rates of approximately 13,000 German communities. We set up a panel, covering 11 years and roughly 2 million employees per year. To the best of our knowledge, this paper is the rst to empirically study the business tax incidence on wages using national individual employee level data. The variation over time and communities allows us to identify causal e ects by applying a two-stage xed e ects estimator. We nd a wage elasticity of the business tax, which varies between -0.2 and JEL Codes: Keywords: H22, H25, J30, J38 Business Taxes, Wage incidence, Microdata, Germany University of Oxford, University of Cologne, CESifo and IZA, clemens.fuest@sbs.ox.ac.uk. z IZA, University of Cologne, ISER and CESifo, peichl@iza.org. x IZA and University of Cologne, siegloch@iza.org. { Andreas Peichl is grateful for the nancial support from the Deutsche Forschungsgemeinschaft DFG (PE1675). The usual disclaimer applies.

2 1 Introduction The questions of who bears the burden of corporate taxes was initially raised by Harberger (1962) and has been discussed ever since. Numerous theoretical studies have come up with answers stressing the importance of capital mobility and the substitutability between capital and labor. Despite the convergence of the theoretical predictions that labor bears a substantial share of the corporate tax burden in open economies, there is hardly any empirical evidence on the wage incidence of corporate taxes. The reason for this scarcity is certainly the absence of suitable data sources. In this paper we make an attempt to ll this research gap by analyzing the incidence of the German local business tax (Gewerbesteuer) on wages. The rate of the local business tax (LBT) is determined on the community level by the community speci c collection rate. In Germany, having approximately 13,000 communities, there is thus considerable variation as far as the corporate tax burden is concerned. Usually, the collection rate is between 200 and percent, yielding an eventual tax rate of between 7 and 14 percent. The tax revenue mostly accrues to the community and is its the most important source of revenue. In order to shed light on the wage incidence of the LBT and use the substantial variation across communities, we make use of a very rich administrative linked employer-employee dataset, covering about 2 million employees per year. We further extend the data by merging the business tax rates of German communities to the LIAB. Constructing a panel of eleven years (from to ) we make use of the huge variation over time and communities to identify causal e ects. As far as our empirical strategy is concerned we start with a basic static panel model, controlling for xed e ects on the individual, rm and community level. We then proceed to more sophisticated dynamic models à la Arellano and Bond (1991) and Blundell and Bond (), taking into account the potentially autoregressivity of the model and addressing potential endogeneity problems due to unobserved factors driving both community tax rate and wages. In addition, we construct a set of external instruments, such as indicators of municipal public good provision and information on community budgets. At a later stage, we will also use the exogenous variation in the German tax legislation and regional discontinuities to identify the wage incidence. To the best of our knowledge our study is the rst to address the e ects of corporate taxes on wages using a national employee level dataset. The richness of 1

3 our data, allows us to address most of the relevant questions raised by the theoretical literature at the same time. We are able to control for openness - on the rm, not the country level -, which has been identi ed as an important factor. Moreover, skill di erence might be crucial, too. The relative burden of high- and low-skill is most likely to be di erent as the complementarity/substitutability between these labor inputs with regard to capital is di erent (see Griliches (1969)). The data allows us to this test hypothesis. Recent studies (Felix and Hines Jr. (2009), Arulampalam et al. (2010)) stress the importance of unions for the tax incidence. They argue that an increase in corporate taxes might reduce the rent to be shared between rm and workers in a bargaining framework. Using information on union coverage of the rm and whether a union contract is in place, we are able to investigate this questions more thoroughly. We are also able to see if the legal type of a rm (such as limited liability rms or stock corporations) has an e ect on the wage incidence. 2 Literature review One strand of the literaure views the incidence of corporate taxation from a general equlibirum perspective. Initiating by the seminal paper of Harberger (1962), showing the capital bears the whole burden in a close economy, the literature quickly evolved either adopting the model to an open economy case (Diamond and Mirrlees (1971), Bradford (1978), Kotliko and Summers (1987), Harberger (1995)), or extending the model by incorporating more sectors (Shoven (1976)) or introducing uncertainty (Ratti (1977)). 1 There are some computional general equilibirum (CGE) model estimates indicating that labor bears roughly 40 percent of the corporate tax burden when assuming a large open economy with reasonable portfolio, product and factor substitution elasticities (see Mutti (1985), Gravelle and Smetters (), Randolph (), Harbger () and Gravelle (2010) for a survey). As far as empirical evidence is concerned, there are a couple of recent studies trying to identify the incidence of corporate taxation on wages. While Hassett and Mahur () nd extremely high estimates of the wage incidence, most other studies nd a wage incidence of between 40 and 80 percent (Desai et al. (), Felix (), Felix and Hines Jr. (2009), Arulampalam et al. (2010),aus dem Moore et al. (2010),Dwenger et al. (2011) and Liu (2011)). 1 Recent survey on the theoretical literature of the general equilibrium are provided by Auerbach () or Harberger (). 2

4 As far as identi cation of the wage incidence is concerned, three clusters evolve: First, cross-country macro regressions where identi cation comes from the variation in tax rates across countries (or US States) over time (Hassett and Mahur (), Felix (). Second, studies on the rm level which either exploit variation in the tax rates in di erent countries (Desai et al. (), Arulampalam et al. (2010), aus dem Moore et al. (2010), Liu (2011), Liu (2011))or exploit di erent treatment of rm types within a country. Third, only Felix and Hines Jr. (2009) analyze the corporate tax incidence on individual wages, relying on a cross section and exploiting the variation in corporate tax rates between US States. In the latter, study the variation is not very also, because relying on a cross-section, cannot control for indivdual xed e ect. It is the main advantage of our study, that we have micro data panel with a large number of observations. Moreover, our data is national, which spares us to deal with cross-country comparisions. For instance, Desai et al. () stress the importance of national information on the micro level to assess the impact of a country s system of taxing corporate income, critizing cross-national macro-regressions that reduce "extent of tax variation that can be appropriately used to identify the distribution of tax burdens, since tax burdens between rms and between industries may have broader e ects that would be overlooked in standard partial equilibrium estimation." 3 German Taxation of Companies Taxation of rms in Germany consists of three pillars: the personal income tax (PIT), the corporate tax (CT) and the local business tax (LBT). Figure 1 shows the revenues of the three di erent taxes on rms for Germany. The local business tax is the most important tax with a revenue of about 40 billion euros which corresponds to approximately 6 percent of Germany s total tax revenue. There have been several changes of the German corporate tax legilation a ecting all three pillars in the recent years (see Table 5 in the Appendix for a synapsis). The local business tax applies to both non-corporate (Personengesellschaften) and corporate rms (Kapitalgesellschaften), certain free professions such as journalists, physiscians, lawyers or farmers are exempt. Since the LBT base, Y LBT, consists essentially of operating pro ts. 2 Until the business tax liabilities could be 2 From to half of the long-term debt service were added to the Y LBT. This changed with the the tax reform of. Instead of long-term debt services 25% of all interest payments 3

5 Figure 1: Revenues of company taxes Company tax revenues in bn. euros and as percent of total tax revenue annual revenue (bn. euros) share of total tax revenues 2010 year personal income tax: in euros (solid), as share (dashed) local business tax: in euros (solid), as share (dashed) corporate tax: in euros (solid), as share (dashed) deducted from their own tax base. Moreover, there is an allowance 24,500 euros up for non-corporate rms. The tax rate of the local business tax, LBT ; consists of two components, the basic federal rate (Steuermesszahl), fr LBT, and the a collection rate (Hebesatz), cr LBT : LBT = fr LBT cr LBT. The basic federal rate which was set at 5:0% from to and decreased to 3:5% in. 3. The collection rate, which works as mutlipier, is set by the communities and varied in 2009 between 200% and 500%. It is most important instrument of the German communities to raise tax revenue. Non-corporate rms (Personengesellschaften) are subject to the personal income tax. The tax base is the share of the operating pro ts assigned to the proprietor. Importantly, the a share of the tax business tax liabilities can be decuted from personal income tax base. 4 The taxable income from running a business is added to taxable (including a lump sum interest portion of rents, leasing rates and royalties) in as much as they exceed an allowance of e100,000 are added to Y LBT. 3 Note that prior to there was a reduced fed for non-corporate rms. For every 12,000 euros exceeding the allowance of 24,500 euros, fed was raised by one percentage point, so that the full basic federal rate of 5:0% had to be paid with a taxable income starting from 72,500 euros. 4 From to 1:8 fed Y LBT could be deducted. From onwards 3:8 fed Y LBT could be credited. 4

6 income from other sources and the PIT schedule is applied. In most cases, pro ts of the rms are so high that companies are the highest tax bracket at face to top marginal tax rate, top P IT, which changed several times during the period under consideration. In addition a solidary surchage for the uni cation (Solidaritätszuschlag), soli, of 5:5% of the tax rate is added. Corporate rms are subject to the corporate tax. The tax base for the corporate tax is de ned similarly to the tax base of the personal income tax. The corporate tax rate has undergone several changes in recent years. Until a corporate tax imputation system existed in Germany, where retained pro ts where subject to a corporate tax rate, CT, of 45% in and 40%. 5 With the tax reform of, retained pro ts and dividends were equally taxed at 25%. 6 In CT was lowered to 15%. As for the PIT, a solidary surchage, soli, of 5:5% of the tax rate is added. In order to calcualte the e ective corporate tax burden for corporate (noncorporate) rms, it is necessary to rst take into account the local business tax bruden as well as the corporate tax (personal income tax) burden. Second, the deductions of the local business tax rate from the PIT base and in some years from its own base have to be takein into account. The e ective marginal tax rate for a corporate rm, corp EMT R, from to is thus 7 : EMT R = CT (1 + soli ) + fr LBT cr 1 + fr LBT cr LBT corp The e ective marginal tax rate for a non-corporate rm with retained pro ts exceeding 250,000 euros, non corp EMT R, is 8 : non corp EMT R LBT = top P IT (1 + soli ) + fr LBT cr LBT 1 + fr LBT ( cr LBT + 1:8) 5 Dividends were taxed at a di erent rate. 6 In there this rate was raised by 1.5 percentage points to nance the costs of a major ood in Germany. 7 Since the denominator is of the fraction is set to 1, since the local business tax cannot be dudected from its own tax base anymore. 8 Since the denominator of the fraction is set to 1 + fr LBT 3:8, since the local business tax cannot be dudected from its own tax base anymore, but parts of the LBT liablities can still be deducted from the PIT, where the multiplier of the basic federal rate changed from 1.8 to

7 4 Data For our analysis we combine two distinct data sources: rst community level data on the communities scal and budgetary situation and second detailed administrative. linked employer employee data. As far as the community data is concerned, we make use of statistics provided by the o cial statistical authorities of the 16 German federal states (Statistische Landesämter). The states collect information on the communities scal and budgetary situation. We combined and harmonized the annual state speci c datasets using speci c community identi cation numbers and constructed a community panel from to covering roughly 140,000 data points, i.e. community-years. Most importantly, the dataset contains information on the local collection rate, but also information on the population size and scal information on the communities expenses and revenues. Moreover, we added regional unemployment rates on a more aggregated level to control for local labor market conditions. The main data source for our analysis is the linked employer-employee dataset (LIAB) provided by the Institute of Employment Research (IAB) in Nuremberg, Germany (see Alda et al. () for more information on the dataset). The employee data are a sample of the administrative employment statistics of the German Federal Employment Agency (Bundesagentur für Arbeit), called the German employment register, which covers all employees paying social security contributions or receiving unemployment bene ts (see Bender et al. ()). The public sector is excluded, as civil servants are rarely observed in the social security data. Employee information recorded in the data include wages, age, tenure, quali cation, occupation and employment type (full-time, part-time or irregular employment). We observe between 1:6 and 2:0 million workers per year The rm component of the LIAB is the IAB Establishment Panel (cf. Kölling ()). The term establishment refers to the fact that the observation unit is the individual plant, not the rm; there can be several plants per company. The Establishment Panel is a representative, strati ed, random sample containing annual information on establishment structure and personnel decisions from 1993 onwards. It includes establishments with at least one worker for whom social contributions were paid, covering 16 industries and establishments from both the former West and East Germany. Among others, we extract the following information on establishments: value added, the export share as an indicator for openness, investment, 6

8 pro tability as measured on a one-to- ve scale, vacancies, layo s, total wages paid, legal form of the company and whether a collective agreement is in place. Per year we observe roughly 12; 000 establishments. The literature on the German local business tax is also relatively scarce. Büttner (), Büttner () set up a theoretical model to determine local business tax rates. He tests the model using data on German business tax collection rates of 327 districts (Kreise) from 1980 to Among others he nds that population is an important determinant of the business tax rate, corroborating the evidence established by Hoyt (1992). Larger communities have higher tax rates. In a subsequent paper, Büttner () analyzes the determinants of the business tax on the community level for 1111 communities of a German states from 1980 to He nds that budgetary variables causing a reduction cause an increase in the local tax rate. He also nds that communities with an older population have a business taxes. With the same data Büttner () analyzes the e ects of the local business tax rates on the tax base. He shows that the own tax rate is much more important for the tax base than the average neighboring rates. 9 9 In addition there are a couple of microsimulation studies, assessing the potentials of various reform proposal of the German business tax (see, e.g., Maiterth and Zwick () or Fossen and Bach ()). 7

9 Variable Obs Mean Std.Dev. Min Max Personal wage male west age tenure D_high-skilled 120, % D_medium-skilled 828, % D_low-skilled 178, % Firm output p. emp investment p. emp employees export share Community (district) collection rate population (in 1000) D_city_0-2K 34, % D_city_2K-5K 92, % D_city_5K-20K 162, % D_city_20K-50K 151, % D_city_50K-100K 109, % D_city_100K-500K 277, % D_city_500K+ 299, % 5 Empirical model Table 1: Descriptive statistics In order to assess the e ects of the business tax rates on wages, we estimate Mincerian type of wage equations on the individual, i.e. the employee s, level. The independent variables of interest is the local collection c;t, which we include in the regression equation with lags from zero to two The corresponding vector is called T. We further include three sets of controls, comprising standard variables which have been identi ed by previous research to have an impact on wages. At the employee level, the vector of covariates, X i;t, contains age, tenure, quali cation, race, gender, and occupation. At the rm level the control vector F j;t consists of output, investment, the export share, union coverage and the legal type of the rm. At the community level, we include population size and state dummies in the vector C c;t. 8

10 Hence, the wage equation for individual i in rm f and community c at time t is speci ed as: ln w ifc;t = 0 + a T c;t + X i;t + F j;t + C c;t + " i;t (1) We add time and individual xed e ects in our esimation. Nevertheless, it is likely that the collection rate is endogenous due to factors that drive both the wage and the collection rate. We thus will estimate a two-stage least squares xed e ects model and will use several instruments Z, which will be discussed in Section Results 6.1 Descriptive statistics Figure 2 depicts Germany s roughly 12,000 communities and the cross-sectional variation in the collection rates as of the year. The darker colors indicate communities with higher collections rates. States such as North Rhine-Westphalia or Saxony have on average a higher rate, whereas in Hesse or Brandenburg the rates are lower. 10 Figure 3 shows the variation of the collection rate over several categories. The by-state-variation con rms the patter demonstrated by Figure 2: North Rhine- Westphalia and Saxony together with the city states Berlin, Bremen and Hamburg have on average higher collection rates. Moreover, the bigger the city, the bigger the collection rate on average. The biggest variation is in the medium-sized community (small cities with a population of 20,000 to 50,000). State. Figure 4 shows the variation of the collection rate over time separately for each Looking at the collection rate by year one sees an upward trend of the collection rate over time in almost every state. increases over time in most states. Moreover, the within variation 10 Note that the black areas in some States (mostly in Rhineland-Palatinate and Schleswig- Holstein) are due to small community sizes and many black lines to be drawn to indicated the borders. 9

11 Figure 2: Communities and collection rates in 6.2 Panel estimates Instruments As noted above, we suspect the collection rate to be endogenous. A possible source of endogneity could be that the communities want to attract capital with a low collection rate, which in turn will a ect the wages. We therefore are in need of instruments to identify the incidence of changes in the business tax on wages. Given the general skepticisim towards the exogeneity of instruments we suggest several di erent instruments previously used in other studies to check how robust our ndings are with respect to the instrument choice. Out instrument vector Z contains the following elements: Lagged debt service (z1): Büttner () shows that the debt service of a community is an important determinant of the collection rate and. He uses the lagged debt service to 10

12 Figure 3: Cross-Sectional variation in collection rates by State by Population (in 1000) SH HH NI HB NW HE RP BW BA SL BE BB MV SN ST TH instrument for the collection rate. Lag de cits (z2): In the same study, Büttner () also uses the lagged de cit is an instrument for the collection rate. Lagged collection of neighboring communities (z3): There are many studies showing that the tax competition leads to tax mimicking due, e.g., to yardstick competition (see, e.g.,ladd (1992), Besley and Case (1995), Kirchgässner and Pommerehne (1996) or Bordignon et al. ()). For the German local business tax the relevant reference is Büttner (). We build an population and distances weighted average of the collection rate of the ten nearest neighboring communities. 11

13 Figure 4: Time variation in collection rates SH HH NI HB NW HE RP BW BA SL BE BB MV SN ST TH Border dummy (z4): There is considerable variation in the collection rates across states (see Figure 3). We therefore suggest a dummy variable as an instrument which is equal to one if the community is close to an inter-state border. 11 A map of the border region de nition can be found in the Appendix. Election year dummy (z5): A recent, unpublished study by Foremny & Riedel (2011) shows that local business taxes are a ected by the electoral cycle. 11 Being close is de ned as having a neighboring community in a di erent state that is less than 10 kilometers away. For the city states Berlin, Hamburg and Bremen, which constitute a single community each, the distance criterion is relaxed due to the size of these communities. 12

14 Note that all Z vary only at the community level (or in case of the election year dummy even on the more aggregate State level). When instrumenting for the local collection rate we therefore have to restrict the set of exogenous regressors in the rst stage regression to those which are determined at the community level as well 12 The top panel of Figure 5 is a graphical representation of the instrument relevance, plotting the collection rathe against the respective instrument. 13 It can be seen that the releationship between lagged de cit debt service and neighbors collection rate is positive, following intuition, whereas the correlation between a communities collection and the average rate of its neighbors appears to be by far the strongest. The correlation between tin the lower two graphs of the gure, we show the change in the distribution of the collection rate when the dummy instrument is switched on and o. The interpretation of the density plots is not as straightforward. Yet, it becomes apparant that the density mass is shifted to the left in election years and for communities in border regions, which is in line with the priors about the instruments e ects on the endogenous variable. First Stage First stage results are presented in Table 2 and corroborate the graphical illustration. All instruments have the expected signs and are sign cant. With the exception of the election year dummy, F-statistics of the excluded instruments are high enough to pass the standard threshold and rule out weak instruments. Yet, the partial r-squared are quite low, with the exception of the neighbors average collection rate, which is mostly likely due to the high general t of the model as indiciated by the r squared statistic. As far as the covariates are concerned, we reassuringly nd that large cities tend to set relatively high tax rates, a stylized fact which has been identi ed both for the US (see, e.g., Hoyt (1992)) and for Germany (see, e.g., Büttner ()) and is explained by the higher market power of bigger cities. Also grunda and grundb, two other tax rates set by the communities, are positively correlated with the collection rate. For the instruments to be valid, they additionally must not be correlated with the idiosyncratic part of the wage equation s error term. In our view, it is very plausible to assume that both the presence of an election and being located in a border region is exogeneous. Moreover, we circumvent the danger of reverse causality by using the 12 In other words it would make sense to include age of individual i as a regressor in a rst-stage regressions of the collection rate in community c. We therefore run the 2SLS estimation by hand and correct the error terms by using the bootstrap. 13 Note that debtservice and de cit is measured in log per capita. 13

15 Figure 5: Instrument relevance rst lag of the neighboring collection rate, the community de cit and debt service. In addition, the presence of time and state xed e ects (and the interaction of those 14 ) should guarantee that no other unobserved local event should have an impact on both the instrument and the individual wages. Second stage DISCLAIMER: The following results are preliminary due to four reasons: (1) They are estimated on the district level not on the community level. The community level data is not available, yet. Community level information are average within districts using the population size as weights. (2) There are still some instruments for some State-years missing. (3) We estimate the e ects for corporate rms with one establishment only. (4) Community controls have not been included yet in the second stage. In the future, we will be able to estimate the e ects also for non-corporate rms as well as for multi-establishments rms for both legal types of companies. The rst stage results are not a ected by these restrictions. Table 3 shows results from the second stage. In column 1, we show do not instru- 14 With the exception of speci cation 5 where the interaction of state and year xed e ects would destroy the instrument, since community elections vary across states. 14

16 Table 2: First stage results Model (1) (2) (3) (4) (5) (6) population 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) property tax (agric.) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) property tax (general) 0.315*** 0.306*** 0.302*** 0.316*** 0.316*** 0.293*** (0.018) (0.019) (0.018) (0.018) (0.018) (0.019) local UR 0.437*** 0.459*** 0.350** 0.462*** 0.451*** 0.345** (0.111) (0.111) (0.108) (0.111) (0.111) (0.108) lag de cit 0.001*** 0.001*** (0.000) (0.000) lag debt service 0.006*** 0.005*** (0.001) (0.001) lag neighbors rate 0.002*** 0.002*** (0.000) (0.000) border region *** (0.005) election year *** (0.001) constant 2.185*** 2.158*** 1.533*** 2.184*** 2.180*** 1.518*** (0.047) (0.045) (0.050) (0.047) (0.047) (0.049) year FE yes yes yes yes yes state FE yes yes yes yes yes Adjusted R Observations Partial R 2 (x10) F-Test of excl. IV Note: Standard errors (in parentheses) clustered at the community level. Signi cance levels are 0.1 (*), 0.05 (**), and 0.01 (***) ment for the collection rate. In the following columns we use the predicted values from the di erent rst stages presented in Table 2. Note that we do not correct standard errors due to the large number of observations, which yields asymptotically correct standard errors even without correcting for the rst stage estimation. The table shows that the collection rate has the expected negative e ect on wages and that the instrumental variable approach increases both the signi cance of the e ect and its absolute magnitude. The 2SLS estimates for the instruments z1 to z3 for are almost the same. Correspondingly, the coe cient in speci cation (7) is also somewhat in line. Using the boder region indicator as an instrument decreases the negative e ect of the collection rate on wages, whereas the election year instument yields a higher e ect in absolute terms. Note that all other covariate yield 15

17 expected signs with positive but decreasing returns to tenure 15. The skill dummies reveal positive returns to education, although there is not much variation of the skill variable within individuals over time. The number of employees in the rm as well as ebit and investment per employee yield a positive impact on wages. Plants with under a sector speci c union contract pay higher wages than plants under a rm or no union contract. Moreover, the community controls such as the population size of the communities or the local unemployment rate show the expected signs. Table 3: Fixed E ects 2SLS results Model (1) (2) (3) (4) (5) (6) (7) Instrument - z1 z2 z3 z4 z5 z1-z4 lag1 collection rate tenure tenure squared high-skilled omitted medium-skilled low-skilled no. of employees ebit p.c invest p.c export share sector union contract omitted rm union contract no union contract community controls no no no no no no no constant yes yes yes yes yes yes yes state dummies yes yes yes yes yes yes yes year*state dummies yes yes yes yes yes no yes year xed e ects yes yes yes yes yes yes yes person xed e ect yes yes yes yes yes yes yes N r Notes: z1 = lagged de cit, z2 = lagged debt service, z3 = lagged neighbors collection rate, z4 = border region, z5 = election year. Bold gures indicate signi cance of at least 5%. In order to be able to interpret the coe cents on the collection rate, we have to make two transformations. First, we have to translate the change in the collection rate of 0.01 points as indicated by the coe cents into a percentage change of the e ective marginal tax rate a rm has to face, taking into account other taxes as well (see also Section 3). This transformation will give us an elasticity. Second, we will express the incidence in money terms to see whether the estimated elasticities are reasonable. 16 Table 4 provides the results. The elasticies of the average wage with 15 Age has been omitted due to the inclusion of year xed e ects. 16 A recent study by Dwenger et al. (2011) questions the exactly this reasonability of many 16

18 respect to the e ective tax rate range between 27 and 98 percent depening on the speci cation. Translated into money terms, the elasticities imply that an increase of the tab liabilities by one euro leads to the reduction of the wage bill by between 36 cents (z4) and 1.30 euros (z5). 17 Table 4: Fixed E ects 2SLS results z1 z2 z3 z4 z5 z1-z4 Estimate Elasticity Incidence Notes: z1 = lagged de cit, z2 = lagged debt service, z3 = lagged neighbors collection rate, z4 = border region, z5 = election year. Incidence is de ned as the Euro response of the total wage bill to a one Euro increase of the tax bill. 7 Conclusions To be completed. elasticities estimates in the literature. 17 Note that the incidence is calculated at the sample means. The mean pro ts before taxes in our sample are 77.5 million euros. Given an average marginal tax rate of 41.9%, the average tax payments are 32.5 million euros. The average annual wage bill is 43.1 million euros. The average plant size in the sample is 1093 employees with an average monthly wage of 3290 euros. 17

19 References Alda, H., Bender, S. and Gartner, H. (), European Data Watch: The Linked Employer-Employee Dataset Created from the IAB Establishment Panel and the Process-Produced data of the IAB (LIAB), Schmollers Jahrbuch: Journal of Applied Social Science Studies 125(2), Arellano, M. and Bond, S. R. (1991), Some Test of Speci cation for Panel Data: Monte Carlo Evidence and an Application to Employment Equations, Review of Economic Studies 58(2), Arulampalam, W., Devereux, M. P. and Ma ni, G. (2010), The Direct Incidence of Corporate Income Tax on Wages, IZA Discussion Paper No Auerbach, A. J. (), Who Bears the Corporate Tax? A Review of What We Know, NBER Working Paper No aus dem Moore, N., Kasten, T. and Schmidt, C. M. (2010), Do wages rise when corporate tax rates fall? evidence form the german business tax reform, mimeo. Bender, S., Haas, A. and Klose, C. (), The IAB Employment Subsample , Schmollers Jahrbuch: Journal of Applied Social Science Studies 120(4), Besley, T. and Case, A. (1995), Incumbent Behavior: Vote-Seeking, Tax-Setting, and Yardstick Competition, Amercian Economic Review 85(1), Blundell, R. and Bond, S. (), Initial Conditions and Moment Restrictions in Dynamic Panel Fata Models, Journal of Econometrics 87(1), Bordignon, M., Cerniglia, F. and Revelli, F. (), In Search of Yardstick Competition: A Spatial Analysis of Italian Municipality Property Tax Setting, Journal of Urban Economics 54(2), Bradford, D. F. (1978), Factor prices may be constant but factor returns are not, Economics Letters 1(3),

20 Büttner, T. (), Determinants of Tax Rates in Local Capital Income Taxation: A Theoretical Model and Evidence from Germany, CESifo Working Paper No Büttner, T. (), Local Business Taxation and Competition for Capital: The Choice of the Tax Rate, CESifo Working Paper No Büttner, T. (), Tax Base E ects and Fiscal Externalities of Local Capital Taxation: Evidence from a Panel of German Jurisdictions, Journal of Urban Economics 54(1), Desai, M. A., Foley, C. F. and Hines, J. R. (), Labor and Capital Shares of the Corporate Tax Burden: International Evidence, mimeo, presented at the International Tax Policy Forum and Urban-Brookings Tax Policy Center conference on Who Pays the Corporate Tax in an Open Economy?, 18 December,. Diamond, P. A. and Mirrlees, J. (1971), Optimal Taxation and Public Production, American Economic Review 61, 8 27 and Dwenger, N., Rattenhuber, P. and Steiner, V. (2011), Sharing the burden: Empirical evidence of corporate tax incidence, mimeo. Felix, R. A. (), Passing the Burden: Corporate Tax Incidence in Open Economies, Federal Reserve Bank of Kansas City Regional Research Working Paper Felix, R. A. and Hines Jr., J. R. (2009), Corporate Taxes and Union Wages in the United States, NBER Working Paper No Fossen, F. M. and Bach, S. (), Reforming the German Local Business Tax - Lessons from an International Comparison and a Microsimulation Analysis, FinanzArchiv 64(2), Gravelle, J. C. (2010), Corporate Tax Incidence: Review of General Equilibirum Estimates and Analysis, Congressional Budget O ce Working Paper No

21 Gravelle, J. G. and Smetters, K. A. (), Does the Open Economy Assumption Really Mean that Labor Bears the Burden of a Capital Income Tax, Advances in Economic Analysis & Policy 6(1), Article 3. Griliches, Z. (1969), Capital-skill complementarity, Review of Economics and Statistics 51(4), Harberger, A. C. (1962), The incidence of the corporation income tax, Journal of Political Economy 70, Harberger, A. C. (1995), The ABCs of Corporate Tax Incidence: Insights into the Open-Economy Case, in Tax Policy and Economic Growth: Proceedings of a Symposium sponsored by the American Council for Capital Formation, ACCF Center for Policy Research, Washington, pp Harberger, A. C. (), Corporate Tax Incidence: Re ections on What Is Known, Unknonw and Unknowable, in J. W. Diamond and G. R. Zodrow, eds, Fundamental Tax Reforms: Issues, Choices, and Implications, MIT Press, Cambridge, MA, pp Harbger, A. C. (), Corporation Tax Incidence: Re ections on What Is Known, Unknown and Unknowable, in G. R. Diamond, John W. & Zodrow, ed., Fundamental Tax Reform: Issues, Choices and Implications, MIT Press, Cambridge, MA, pp Hassett, K. A. and Mahur, A. (), Taxes and Wages, AEI Working Paper No Hoyt, W. H. (1992), Market Power of Large Cities and Policy Di erences in Metropolitan Areas, Regional Science and Urban Economics 22(4), Kirchgässner, G. and Pommerehne, W. W. (1996), Tax Harmonization and Tax Competition in the European Union: Lessons from Switzerland, Journal of Public Economics 60(3), Kölling, A. (), The IAB-Establishment Panel, Schmollers Jahrbuch: Journal of Applied Social Science Studies 120(2),

22 Kotliko, L. J. and Summers, L. H. (1987), Tax Incidence, in A. J. Auerbach and M. Feldstein, eds, Handbook of Public Economics, Vol. 2, Elsevier, Amsterdam, pp Ladd, H. F. (1992), Mimicking of Local Tax Burdens Among Neighboring Counties, Public Finance Review 20(4), Liu, Li & Altshuler, R. (2011), Measuring the burden of the corporate income tax under imperfect competition, mimeo. Maiterth, R. and Zwick, M. (), A Local Income and Corporation Tax as an Alternative to the German Local Business Tax: An Empirical Analysis for Selected Municipalities, Journal of Economics and Statistics 226(3), Mutti, John & Grubert, H. (1985), The Taxation of Capital Income in an Open Economy: The Importance of Resident-Nonresident Tax Treatment, Journal of Public Economics 27(3), Randolph, W. C. (), International Burdens of the Corporate Income Tax, CBO Working Paper -09. Ratti, Ronald A. & Shome, P. (1977), On a General Equilibrium Model of the Incidence of the Corporation Tax under Uncertainty, Journal of Public Economics 8(2), Shoven, J. B. (1976), The Incidence and E ciency E ects of Taxes on Income from Capital, Journal of Political Economy 84(6),

23 Figure 6: Border region 22

24 Table 5: Changes in corporate tax legislation Tax Year Tax Base Tax Rate Personal Income Tax (PIT) Operating Pro ts of non-corporate rm %; allowance=6,322e plus additions and deductions minus %; allowance=6,681e for non-corporate rms, incl. allowance %; allowance=6,902e free professions such as %; allowance=7,206e journalists, physiscians, %; allowance=7,235e lawyers or farmers %; allowance=7,235e %; allowance=7,664e %; allowance=7,664e %; allowance=7,664e %; allowance=7,664e %; allowance=7,664e Corporate Tax (CT) Operating pro ts 45.0% on retained pro ts; 30% on dividends 40.0% on retained pro ts; 30% on dividends for corporate rms 40.0% on retained pro ts; 30% on dividends Operating pro ts 25.0% 25.0% 26.5% 25.0% 25.0% 25.0% 25.0% 15.0% Local Business Tax (LBT) - Operating pro ts from PIT or CT plus 5% on tax base in case rm is corporate; long-term debt minus loss In case rm is non-coporate: for corporate and from preceding years and minus an 1% if tax base+allowance <36,500e and non-corporate rms; allowance of 24,500efor non-corporate 2% if tax base+allowance <48,500e free professions are exempt rms 3% if tax base+allowance <60,500e 4% if tax base+allowance <72,500e 5% if tax base+allowance >72,500e Operating pro ts from PIT or CT plus 3.5% on tax base for non-corporate & corporate 25% of all interest payments in case those exceed 100,000eminus loss from preceding years and minus an allowance of 24,500efor non-corporate rms Solidarity Surcharge - 5.5% of CT or LBT 23