Labor Taxation and FDI Decisions in the European Union

Transcription

1 Open Econ Rev (2014) 25: DOI /s RESEARCH ARTICLE Labor Taxation and FDI Decisions in the European Union Åsa Hansson & Karin Olofsdotter Published online: 14 June 2013 # Springer Science+Business Media New York 2013 Abstract This paper uses panel data on bilateral FDI stocks in the European Union to empirically analyze the impact of labor and corporate taxations on FDI decisions. While the effect of corporate taxes on FDI is well documented, the impact of labor taxes on FDI has barely been explored. This is surprising since labor taxation may influence FDI as well; the taxation of labor affects the production cost and the ability to attract and retain productive labor, and thereby it also, ultimately, impacts the return to the investment. By employing a Heckman two-step estimation model, which controls for possible sample selection bias due to many zero bilateral observations, we find that labor taxes do influence FDI decisions. Keywords Labor taxation. Foreign direct investment JEL Classification F12. F15. F21. H24. H73 1 Introduction Researchers have paid much attention to the role corporate taxation plays in attracting foreign direct investment (FDI). As a result, there is a vast literature estimating the effect corporate taxes have on FDI, and there is now, it is fair to say, a consensus that corporate taxation has a statistically significant impact on FDI decisions. The exact magnitude, however, is uncertain. Common estimates of the semi-elasticity typically fall between 5 and 0, with a median of 2.9 (see e.g., de Mooij and Ederveen 2006 and Feld and Heckemeyer 2011), implying that a one percentage point increase in the corporate tax rate lowers FDI by 2.9 %. Other taxes, apart from the corporate tax rate, may impact a firm s net return and are, hence, likely to matter for FDI decisions as well. Knowledge of the impact of Å. Hansson (*) : K. Olofsdotter Department of Economics, Lund University Sweden, Box 7082, SE Lund, Sweden asa.hansson@nek.lu.se Å. Hansson Research Institute of Industrial Economics, Box 55665, SE Stockholm, Sweden

2 264 Å. Hansson, K. Olofsdotter other taxes on FDI decisions is limited, however. One important tax whose effect has largely been neglected is labor taxation. Taxation of labor, and that of high-skilled labor in particular, may affect multinationals activities and FDI decisions as taxation of labor is likely to directly influence the net return on investments by increasing firms costs. In addition, labor taxation may reduce work effort and firms ability to attract and retain productive workers and thus indirectly lower the investment return. This paper studies how taxation of labor influences FDI decisions. It closely follows Hansson and Olofsdotter (2010) and makes use of a two-step Heckman estimation model that controls for possible sample selection problems and, we believe, better reflects the nature of FDI decisions; that is, whether to invest or not and, given that investment takes place, the amount to invest. In addition, the Heckman estimation model controls for selection bias in data with many zero observations, like in bilateral FDI data. Using recent data on marginal and average labor and corporate tax rates for all the 27 EU member countries covering the period , we find that labor taxes have a negative impact on FDI. More specifically, within the EU27 we find a semielasticity of around 2, implying that a one percentage point increase in the difference in labor tax rates between a country pair lowers FDI stocks by about two percent. The rest of the paper is organized as follow. Section 2 provides some background as to why labor taxation may matter for FDI decisions, and discusses some previous studies. Section 3 reviews trends in FDI stocks within the EU and in corporate and labor tax rates, while Section 4 presents the method and data used. Section 5 reports the results and Section 6 provides discussion and some further analysis. Section 7 concludes the paper. 2 Why Labor Taxes may Matter for FDI Decisions Why is there extensive research on corporate taxation and FDI decisions, but very little research on labor taxation and FDI decisions? It is natural to assume that taxation of capital, rather than labor, affects capital flows and stocks such as FDI. Moreover, the extensive focus on corporate taxation as a determinant of FDI typically rests on the assumption that capital is mobile and responds to cross-country differences in corporate tax rates. Countries can thus attract FDI by lowering their corporate tax rate. The downward trend in corporate tax rates in the last decades has been attributed, at least partly, to this so-called tax competition. On the other hand, the link between FDI and labor taxation is less obvious and traditionally labor taxation has not been regarded as a factor influencing FDI decisions. One reason for this is that labor is less mobile than capital. Another reason is that labor taxation is not typically thought of as influencing firms costs and investments. This reasoning can be questioned, however. Even if labor is less mobile than capital, or should even be totally immobile, labor taxation could matter for FDI decisions as labor taxation affects multinational firms net returns and, consequently, their location and investment decisions. Whether labor taxation impacts firms net return depends on how labor taxes affect firms costs; that is, the incidence of labor taxation. Traditionally, the burden of labor taxation has been thought to fall entirely on employees and therefore not impose a cost on employers. This may not necessarily

3 Labor Taxation and FDI Decisions in the European Union 265 be the case, however, especially for high-skilled workers who face an increasing number of opportunities and are unlikely to supply their labor perfectly inelastically. Several recent studies call the traditional view that labor supply is perfectly inelastic and, hence, taxes on labor are fully borne by employees into question (see e.g., Forslund et al. 2006, Daveri and Tabellinni 2000, and Bingley and Lanot 1999). There are many reasons for this and several of them stem from weakening power of the labor unions. Wage formation, for example, is more decentralized today than earlier (Kiander et al. 2004) due to, among other things, increased competition and an increasing share of workers working in foreign-owned companies. In addition, firms have a wider set of options compared to previously when it comes to production location and, thanks to technological progress means of production making the demand for labor more elastic. This weakens workers negotiation power and makes it less likely that 100 % of a tax cut will result in higher wages for workers and therefore not affect firms labor costs. Consequently, it is unrealistic to assume that the incidence of a tax fully falls on workers and has no effect on employers labor costs. Moreover, even if the burden of the labor tax were to fall entirely on the employees, labor taxation would likely affect firms ability to attract and retain productive labor and key personnel and, in addition, workers effort. An extensive literature has found that incentives and compensation policies matter for individuals effort (see e.g., Ehrenberg 1990, and Prendergast 1996, for reviews). This literature predicts a positive relationship between wage and work effort, something empirical work also has found support for (see e.g., Lazear 2000; Asch 1990; Bognanno and Ehrenberg 1990; Kahn and Sherer 1990). One explanation for this draws on the efficiency wage theory and suggests that workers compare their current pay with opportunities outside the firm, and that a higher net wage reduces the attractiveness of alternative opportunities and encourages greater work effort at their current jobs (Akerlof and Yellen 1986). Another explanation, stemming from the equity theory, suggests that workers compare their compensation to that of others in the same firm also implying that higher compensation increases work effort (Lawler et al. 1968). While taxes are set nationally, international differences may make work effort vary across countries and affect where firms choose to locate production, 1 in addition, high tax rates and a progressive structure make leisure less costly and may reduce effort rather than working hours. Results from the tax response literature support this view. Several studies reveal that especially high-income earners respond to a decrease in net return by reducing their effort rather than reducing their work hours (see e.g., Gruber and Saez 2002). Thus even if we assume that the burden of labor taxation falls entirely on workers, labor taxation can still affect production costs since lower net compensation for workers reduces their effort, which in turn increases production costs and lowers efficiency. The importance of a well-functioning labor market with a strong link between workers incentives and effort has also been pointed out as one of the twelve pillars of competiveness according to the World Competitiveness Report. To ensure competitiveness it is important to allocate workers to their most efficient use and to provide 1 Several high-tax countries impose favorable tax treatment on foreign experts in order to attract and keep key personal.

4 266 Å. Hansson, K. Olofsdotter incentives that motivate effort (World Economic Forum 2009). Large labor tax wedges and a progressive tax structure can weaken the link between incentives and effort and reduce competitiveness and ultimately the return on investments. Few empirical studies have estimated the effect of labor taxes on FDI, despite survey results suggesting that labor taxes do influence firms localization and investment decisions. For example, a survey by Braunerhjelm and Lindquist (1999), based on interviews with the 50 largest corporations in Sweden, shows that individual income taxation is one of the driving factors in Swedish firms decisions on where to locate their headquarters. Specifically, the survey discloses that favorable individual income taxation, followed by well-developed transportation, proximity to customers, and the general attractiveness of the region are the most important determinants of where firms locate headquarters. Although corporate taxation is found to be important, it ranks lower than individual income taxation. A few studies have investigated the effects of labor taxes on FDI. Egger and Radulescu (2011) model how corporate and labor taxation influence FDI decisions. They distinguish between labor taxes levied on employers and employees, and argue that both affect firms profits but for different reasons. Labor taxes levied on employers increase firms costs and thereby reduce profits, while taxes levied on employees reduce managers effort and thereby firms profits. Somewhat surprising Egger & Radulescu, hence, assume that the statutory incidence equals the economic incidence. In the empirical part of their paper the hypothesis is tested on FDI stocks from 49 countries for The result suggests that labor taxes matter for FDI decisions, and, interestingly, that the employee-borne part of the labor tax influences FDI stocks negatively. The progressivity in the personal income tax also has a negative and significant impact on FDI stocks. However, they do not find that the employer-borne part of the tax affects FDI stocks significantly, which may seem surprising, but could indicate that the statutory tax incidence is a poor proxy for economic incidence. Corporate tax rates are found to be more important for FDI decisions, however, both when it comes to significance and magnitude. Because labor taxation is strongly correlated with labor cost, the literature on labor and wage cost as determinants of FDI is of relevance and worth briefly mentioning. Several papers discuss the impact of labor or wage costs on FDI decisions. Since labor cost is an important factor cost, it is thought to mainly influence vertical FDI; that is, FDI motivated by lower production costs. There is mounting evidence that labor and wage costs influence localization decisions. Braconier et al. (2005) study, for instance, shows that wage costs have a strong negative effect on FDI flows from the US and Sweden. Countries with relatively cheap low-skilled labor attract more FDI than countries where low-skilled labor is more expensive, which supports the case for vertical FDI. Moreover, Becker et al. (2005) find that higher labor costs deter FDI flows to Germany and Sweden. This is also in line with results from Braunerhjelm and Thulin (2009) on FDI flows from Sweden. Specifically, they find that FDI flows have become increasingly sensitive to wage costs, but that the presence of agglomeration economies allows for higher wages. Additionally, Cheng and Kwan (2000) find wage costs to have a negative effect on FDI in 29 Chinese regions in the period 1985 to Contrary to this are studies by Chen (1996) and Head and Ries (1996) who attain no support for labor cost differences being important for location of FDI in China. Instead, they find agglomeration economies to be

5 Labor Taxation and FDI Decisions in the European Union 267 significant. Further support for a link between labor costs and FDI is provided in Bellak et al. (2008) who, after studying and surveying the pre-existing literature on labor costs and FDI to the Central and Eastern European Countries (CEEC), conclude that labor cost is an important negative determinant of FDI flows into CEEC. 3 FDI and Taxes in the EU15 and the New Member Countries We choose to study how labor taxation affects investment decisions within the EU27. Because we believe that FDI decisions differ between investments in the old EU (EU15) and in the new member countries (NMCs), we analyze the determinants of the two member groups separately. 2 Figures 1 and 2 display total FDI stocks within the EU15 and FDI stocks from the EU15 to the new member countries, respectively. Within the EU15, investment stocks have steady increased over the period apart from a dip in the early 2000s. The growth rate at which FDI increases is slightly lower after the dip than before, however. Stocks from the EU15 to the NMCs have increased sharply from 1995 to the peak in FDI stocks from the EU15 to the NMCs increased with more than 340 % during this period. Since the peak year, FDI stocks have declined by 60 %. Despite the sharp increase, the share of FDI stocks to the NMCs, as a total of FDI stocks within the EU27, only increased from 6.5 % in 1995 to 8.7 % by 2005, which implies that increased FDI stocks to the NMCs did not crowd out investments within the EU15. Figures 3 and 4 present the development of the corporate tax rates in the EU15 and the NMCs. Figure 3 shows the development of the statutory corporate tax rate since 1995, while Fig. 4 shows the development of the effective marginal and average corporate tax rate since Both figures display negative trends, and the decline is more distinct for the statutory rates than for the effective rates. 3 The decline in corporate tax rates is more pronounced in the NMCs, and the gap between the statutory corporate tax rates in the old and new member states has widened from 4 % in 1995 to 10 % in For the effective average and marginal tax rates the gap has doubled since 1998, which has led some to fear intensified tax competition from the enlargement. Hansson and Olofsdotter (2010) provide support for the notion that the NMCs lower corporate tax rates have indeed led to increased FDI flows from the EU15 to the NMCs. Figures 5 and 6 show the development of the labor tax rates in the EU15 and the NMCs since Figure 5 displays the top marginal statutory tax rates and Fig. 6 the implicit average tax rates (calculated as the ratio of total tax revenues to a proxy for potential labor tax base (Eurostat 2010)) on personal income. Labor tax rates, especially top marginal rates, have declined as well. The NMCs employ substantially lower top marginal tax rates (many of them apply fairly low and flat rates) and the gap between the two member groups has widened from almost 10 percentage points in the mid 1990s to over 19 percentage points in The gap between the implicit 2 Austria, Belgium, Denmark, Germany, Greece, Finland, France, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden, and the UK are counted as the old EU members or the EU15. Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, and Slovenia are counted as the new member countries (NMCs). 3 Tax reforms across countries have lowered statutory rates and broadened the base, resulting in substantially lower statutory rates and somewhat lower effective rates.

6 268 Å. Hansson, K. Olofsdotter Fig. 1 FDI stocks within EU15, millions of Euros. Source: Eurostat (2010) average tax rates for labor income has also increased: from 1 percentage point in 1995 to 4 percentage points in Increasing FDI stocks have thus coincided with decreases in both corporate and labor tax rates. Whether the increase in FDI can be attributed to lower tax rates, and if so which ones, needs to be further investigated, however. 4 Method and Data In order to estimate how labor taxation affects FDI decisions, we employ the two-step Heckman estimation technique. This is appropriate when FDI stock data are characterized by many zero observations in order to mitigate problems of sample selection bias Fig. 2 FDI stocks from the EU15 to the NMCs, millions of Euros. Source: Eurostat (2010)

7 Labor Taxation and FDI Decisions in the European Union 269 Fig. 3 Statutory corporate tax rates. Source: European Commission (Davis and Kristjánsdóttir 2010). 4 In addition, we believe that this estimation method fits nicely with the underlying FDI decision process. According to Razin et al. (2004) and Razin and Sadka (2006), foreign investment decisions can be characterized by a two-step procedure; first, a location decision on whether to invest or not in a particular country, and second, an amount decision on how much to invest. Razin et al. (2004) model this by assuming fixed setup costs for new investments, making bilateral FDI flows between an investing and a host country lumpy. 5 The fixed costs play no role in the decision regarding the amount of investment, but they do affect the location decision as well as the decision whether to engage in FDI at all. This two-fold nature of the investment decision suggests that the empirical estimation should make use of a sample selection procedure. The Heckman selection estimation model is, therefore, a natural candidate. Also, using this model allows the independent variables to have different effects on the decision to invest and the amount to invest, respectively. In the Heckman s selection model, selection from the sample is first predicted and then used to adjust the OLS estimates to account for the selection bias. In other words, the model assumes an underlying selection equation determining whether the dependent variable is observed or not (whether FDI takes place or not). More specifically, we estimate the following model: FDI * ijt ¼ X 1ijtβ 1 þ ε 1ijt s * ijt ¼ X 2ijtβ 2 þ ε 2ijt ð1þ ð2þ FDI ijt ¼ FDI * ijt ; s ijt ¼ 1 if s * ijt > 0 ð3þ FDI ijt ¼ 0; s ijt ¼ 0 if s * ijt 0 : ð4þ 4 Zero values can be due to no FDI stocks appearing, either because no FDI takes place or because it is too small to be reported. 5 The assumption of fixed set-up costs distinguishes FDI from purely financial transactions.

8 270 Å. Hansson, K. Olofsdotter Fig. 4 Effective marginal and effective average corporate tax rates. Source: Devereux et al. (2008) Equation (1) estimates the determinants of the amount of FDI, while Eq. (2) estimates the underlying selection equation where s ijt is one if the FDI stock from country i to country j is positive, and zero otherwise. The error terms are assumed to be normally distributed with a covariance σ 12 and with a correlation coefficient ρ.ifρ is positive, OLS estimation of Eq. (1) will yield biased results while Heckman provides consistent and asymptotically efficient estimates. We estimate the effect of tax rates on FDI by using unbalanced panel data on the bilateral outgoing FDI stocks between all 27 member countries of the European Union for the period We follow previous studies on FDI and use a gravity model where FDI is determined by standard gravity variables as well as taxes, agglomeration economies, and additional control variables. 6 We follow Blonigen and Davies (2004) and use a log-linear specification to deal with the skewness common in FDI data. In more detail, the amount Eq. (1) is estimated according to: lnfdistock ijt ¼ γ 1 Taxdifflab ijt þ γ 2 Taxdiffcorp ijt þ γ 3 lngdp it þ γ 4 lngdp jt þ γ 5 lndist ij þγ 6 lnagglodiff ijt þ γ 7 Z 1 ijt þ γ n Z n ijt þ λ t þ ε ij where lnfdistock ijt, is the natural logarithm of the stock of FDI from investing country i to host country j in year t. X 1ijt in (1) includes a number of variables where Taxdifflab ijt is the difference in labor tax rates between the host and investing countries, and Taxdiffcorp ijt is the corresponding difference for the corporate tax rates. lngdp it and lngdp jt are the logarithms of the investing and host country s GDP, respectively. lndist ij is the logarithm of the bilateral distance between country i and j. Agglodiff ijt is an agglomeration variable based on the difference in market potential between the host and investing country. Z is a number of additional control variables assumed to affect the stock of FDI, λ t is a time dummy, and ε ijt the error term. ð5þ 6 For other studies of FDI that employ a gravity framework, see, for example, Wei (2000), Stein and Daude (2007), Blonigen and Davies (2004), Bénassy-Quéré et al. (2005), Lahrèche-Révil (2006), Wolff (2007), Bellak and Leibrecht (2009), and Egger et al. (2009).

9 Labor Taxation and FDI Decisions in the European Union Percent NMCs EU Year Fig. 5 Top statutory personal income tax rates. Source: IBFD (2010) We use data on FDI stocks going out from the investing to the host country, provided by Eurostat, as our dependent variable. The data set records more than 2000 observations of FDI stocks to the NMCs and nearly 2900 observations of FDI stocks to the old member countries. For these observations, a EU15 country is the investing country in approximately 60 percent of the cases. Regarding different components of FDI, equity capital constitutes the largest part of FDI for both new and old member countries. The main independent variable of interest in this paper is the labor tax rate. According to theory, we should expect average taxes to influence discrete decisions such as whether to invest or not, and marginal tax rates to influence decisions such as how much to invest. Consequently, we want to use average tax rates in the selection equation and marginal tax rates in the amount equation. Since we are interested in how tax differences affect investment and location, we wish to use forward-looking tax measures, as investment decisions primarily depend on current and expected future tax rules. Defining these tax rates is not uncomplicated. We use two different sets of average and marginal labor tax rates. First, we use the implicit average tax rate on labor income from Eurostat (Eurostat 2010), and the top marginal income tax rate (IBFD 2010). The implicit average tax rate is calculated as the ratio of total tax revenues to a proxy for the potential labor tax base. 7 The top marginal income tax rate is the statutory marginal tax rate applying to personal income in the top bracket. Second, we use the average and marginal tax rates derived by OECD for a single person with an income that is 167 % of the average production worker s (APW) (OECD 2010). Our preferred tax set is the top marginal tax rate as it is forward-looking, and we also believe the top marginal tax rate to be more relevant for firms trying to attract productive high-skilled labor than the rate of a person with an income 167 % of the 7 Note that the implicit average tax rate is not the rate that applies to the top-bracket income earners but rather the overall average.

10 272 Å. Hansson, K. Olofsdotter Percent NMCs EU Year Fig. 6 Implicit average personal income tax rates. Source: Eurostat (2010) APW. Moreover, the OECD tax data lack information on many of the new EU members. In addition, the definition of tax rates was changed in We calculate the difference in the average and marginal taxes between host and investing country and expect this difference to negatively influence FDI stocks. Measuring the other tax rate variable, the corporate tax rate, is not straightforward either. By the same logic as above, we use average in the selection and marginal in the amount equation. We choose to use effective tax rates as these take into account the tax base, depreciation rules, government tax compensations, etc. More specifically, they are calculated as the net present value of tax payments as a share of the net present value of pre-tax income. It should be noted, however, that these tax rates are based on a hypothetical investment project that requires a number of assumptions and simplifications. Data on effective marginal and average taxes are provided by Devereux et al. (2008) starting in Again, we use the differences in tax rates between host and investing country and expect this difference to be negatively related to the outward stock of FDI. As for the gravity variables, GDPs for the host and investing countries represent the sizes of the economies and, in the standard setting, are expected to have a positive effect on bilateral FDI stocks. From a theoretical point of view, the effect of geographical distance is ambiguous, as it may reflect trade costs. 9 The results from most empirical studies, though, suggest that distance tends to have a negative impact on FDI. It should be emphasized that despite its goodness of fit, the gravity framework for FDI is theoretically unfounded. 10 The obvious reason is that the multifaceted nature 8 The definition of the average and marginal tax rate was slightly changed in 2000, so that the years and use different definitions of these tax rates. 9 For a theoretical discussion of trade costs and FDI, see Neary (2009). 10 A more theoretically founded framework for FDI, provided by Carr, Markusen & Maskus (2001), includes distance and trade costs.

11 Labor Taxation and FDI Decisions in the European Union 273 of FDI makes the impact of markets and geography complex. While the simple bilateral gravity framework should work well for horizontal FDI, vertical FDI and export-platform FDI decisions have a definite multilateral character where the decision to invest in a specific country is not isolated from alternative investment locations. 11 We include a market-potential variable measuring the difference in market size between host and investing country. Market size is measured as the GDP of all other countries in the sample weighted by distance. Thus, if FDI is mainly in the form of export-platform FDI, where the host-country serves as a platform for exports to third markets, this variable is expected to be positive. For vertical FDI, on the other hand, market potential is less clear. 12 The market potential variable is also our agglomeration variables together with GDP in the host country. Agglomeration is found to affect FDI decisions (see e.g., Hansson and Olofsdotter 2012). In addition, several other variables thought to influence FDI decisions are included, such as the differences in labor cost, inflation rate, and the share of government investment. In addition, dummies for whether the host and investor country are actually members of the EU in a particular year are included. These variables as well as further data description are presented in Table 8 in the appendix. In the selection equation, Eq. (2), X 2ijt includes all variables in X 1ijt and, in addition, differences in trade openness and GDP per capita for identification. The average tax rate differentials are used for the tax variables. An alternative estimation technique to deal with problems arising due to a large number of zero observations is the Tobit estimation. The zero observations are then treated as a result from a censored process. Alternatively, we use a Tobit estimation technique to estimate the impact of labor taxes on FDI decisions. The Tobit estimator, however, assumes that the effects of the independent variables are the same for both the probability of being selected and the observed amount. 5 Results Table 1 reports the results from two Heckman two-step estimations of FDI within the EU27. The first columns in each specification report the estimates from the amount equations while the second reports the estimates from the selection equations. In the first specification we use our preferred labor tax measures: the top marginal and implicit average labor tax rate differentials. In the second specification we use the effective marginal and effective average labor tax differentials for a single person with an income of 167 % of the average production worker. Starting with our main variables of interest, labor tax differentials, they seem to influence the decision whether to invest or not and as well as the amount invested. In our preferred specification, both the top marginal and the implicit average labor tax rate differentials have a negative and statistically significant impact on FDI. The 11 The multilateral character of FDI, in combination with the impact of market access, is also related to the difficulty in defining the scope for agglomeration economies. 12 As discussed in, e.g., Blonigen et al. (2004), market potential will have no effect on vertical FDI where the multinational enterprise (MNE) seeks the single lowest cost producer by evaluating all possible locations. On the other hand, in cases where several activities are to be outsourced by an MNE, the market potential of a specific location is likely to have a positive impact on the FDI decision.

12 274 Å. Hansson, K. Olofsdotter Table 1 Heckman estimates: Determinants of FDI decisions within the EU27 EU27 EU27 Amount Selection Amount Selection Top marginal labor tax rate differential (5.25) b Effective marginal labor tax differential (2.07) a Effective marginal corp. tax differential (1.42) (0.40) lngdp(investor) (15.16) b (18.05) b (7.18) b (16.80) b lngdp(host) (19.15) b (4.08) b (20.61) b (3.06) b lndistance (26.66) b (3.18) b (25.86) b (4.86) b Market potential differential (10.46) b (5.88) b (7.37) b (4.94) b Labor cost differential (8.70) b (1.56) (10.99) b (1.49) Gov investment differential (7.58) b (1.50) (4.82) b (4.59) Inflation differential (3.80) b (0.58) (0.61) (2.70) b EU membership (host) b (14.10) (3.41) b (13.59) b (0.45) EU membership (investor) (10.96) b (6.89) b (13.47) b (2.18) a Implicit average labor tax differential (4.45) b Effective average labor tax differential (3.46) b Effective average corp. tax differential (0.76) (2.16) a Trade differential (12.40) b (9.72) b GDP per capita differential (4.10) b (1.54) Constant (5.89) b (18.12) b (0.67) (16.09) b Observations 5,527 3,175 Zeros 2,751 1,228 χ p-value Absolute value of z statistics in parentheses significant at 5 % significant at 1 % Also includes year dummies

13 Labor Taxation and FDI Decisions in the European Union 275 resulting semi-elasticity is 2.5, implying that a one percentage point increase in the marginal tax rate differential reduces FDI stocks by 2.5 %. The other tax rate differentials, based on the tax difference of a single person with an income of 167 % of the APW, also have negative impact on the amount invested albeit smaller in magnitude. In the selection equation the effective average labor tax rate differential now has a positive and statistically significant impact. Interestingly, the effects of corporate tax rates on FDI decisions are weaker and only statistically significant in the selection equation in the second specification. Turning to the traditional gravity variables, the results provide support for the importance of the role they play for FDI amounts and in determining whether FDI takes place or not. More specifically, the GDPs of the investors and host countries have a positive and statistically significant impact on FDI amounts and in the first specification in the selection. In the second specification, GDP in the host country has a negative and statistically significant impact on FDI taking place. As expected, distance has a negative and statistically significant impact on FDI amounts. The effect of distance in the selection is more ambiguous with a negative coefficient in our preferred specification while positive in the second specification. The agglomeration variable market potential affects the amount invested positively while the effect on whether FDI tales place or not is negatively influenced by market potential. Of the other variables, it is noteworthy that labor cost differences and differences in government investment between host and investing country both impact the amounts invested negatively. When the investing country is an EU member both the probability that investment takes place and the amount invested increases. However, whether the host country is an EU member or not seems to impact the amount invested positively but have, if any, a negative impact on whether investment takes place or not. As pointed out by Hansson and Olofsdotter (2010), corporate tax rate differentials impact FDI decisions within the EU15 and from the EU15 to new member countries differently. There is reason to expect the same to hold for labor taxation. Hence, we now distinguish between the impacts of the different determinants of FDI decisions within EU15 and from the EU15 to the new member states (NMCs). The regression results for investment decisions within EU15 and from the EU15 to the NMCs are reported in Tables 2 and 3, respectively. The results reveal some interesting differences. Starting with Table 2, labor tax rate differentials have a negative and statistically significant impact on the amount invested within the EU15 in both specifications. The semi-elasticity ranges from 4.6 to 1.4, numbers larger in magnitude (in absolute terms) than those presented in Table 1. In the selection the labor tax differential has a positive and statistically significant impact however. This suggests that the old EU members are more likely to invest in high labor-tax countries but that the amount invested is negatively affected by the labor tax rate. The corporate tax rate differential seems to be less of a deterrent for the amount invested within the EU15; the differential even has a positive impact in the preferred specification. Whether FDI investment takes place or not is negatively influenced by the corporate tax rate, while only statistically significantly in the second specification. Among the other control variables, higher labor costs in the host country is a deterrent on the amount invested while market potential has a positive impact on FDI amounts.

14 276 Å. Hansson, K. Olofsdotter Table 2 Determinants of FDI decisions within the EU15 Within EU15 Within EU15 Amount Selection Amount Selection Top marginal labor tax differential (9.84) b Effective marginal labor tax differential (3.99) b Effective marginal corp tax differential (1.96) a (0.88) lngdp(investor) (10.47) b (11.52) b (9.00) b (11.83) b lngdp(host) (22.81) b (2.28) a (20.61) b (2.65) b lndistance (28.03) b (9.80) b (27.48) b (9.79) b Market potential differential (6.43) b (3.50) b (6.53) b (3.21) b Labor cost differential (2.70) b (0.08) (6.14) b (0.71) Inflation differential (1.74) (1.11) (1.97) a (1.23) Gov investment differential (0.63) (2.40) a (1.19) (2.82) b Implicit average labor tax differential (2.57) b Effective average labor tax differential (3.74) b Effective average corp tax differential (0.94) (2.04) a Trade differential (9.59) b (8.73) b GDP per capita differential (5.68) b (3.70) b Constant (4.17) b (13.63) b (5.51) b (13.64) b Observations 1,967 1,967 Zeros χ p-value Absolute value of z statistics in parentheses significant at 5 % significant at 1 % Also includes year dummies

15 Labor Taxation and FDI Decisions in the European Union 277 Table 3 Determinants of FDI from the EU15 to the NMCs EU15 to NMCs EU15 to NMCs Amount Selection Amount Selection Top marginal labor tax differential (0.31) Effective marginal labor tax differential (5.02) b Effective marginal corp tax differential (4.90) b (2.27) a lngdp(investor) (22.30) b (6.86) b (10.13) b (9.68) b lngdp(host) (6.74) b (1.91) (12.12) b (4.67) b lndistance (17.15) b (7.98) b (10.17) b (1.32) Market potential differential (2.46) b (0.75) (4.96) b (4.37) b Labor cost differential (5.26) b (2.14) a (5.15) b (2.29) a Gov investment differential (2.88) b (1.52) (0.42) (0.76) Inflation (host) differential (3.19) b (1.77) (1.97) a (0.04) EU membership (host) (0.01) (4.66) b (1.43) (6.10) b Implicit average labor tax differential (1.84) Effective average labor tax differential (0.32) Effective average corp tax differential (0.07) (2.60) b Trade differential (8.53) (8.26) GDP per capita differential (1.32) (1.67) Constant (2.94) b (0.51) (7.20) b (6.58) b Observations 1, Zeros χ p-value Absolute value of z statistics in parentheses significant at 5 % significant at 1 % Also includes year dummies

16 278 Å. Hansson, K. Olofsdotter As seen in Table 3 determinants of FDI decisions from the EU15 to the NMCs differ from those within the EU15. For example, labor tax differentials do not influence FDI negatively; on the contrary, the impact is even positive in the second specification. The corporate tax rate differentials, however, have a negative and statistically significant impact on the amount invested. The semielasticity for the effective marginal corporate tax rate differential varies between 3.8 and 2.4. Another interesting difference concerns the market potential variable. This now has a negative and statistically significant impact on the amount invested (and either an insignificant or negative influence on the discrete FDI decision), suggesting that agglomeration economies are not driving FDI to the NMCs or, alternatively, that FDI to NMCs are not driven by exportplatform motivation. Labor cost differences seem to play a more important role for FDI decisions to the NMCs than within the EU15. The coefficients of labor cost differentials are negative and statistically significant in both the selection and the amount equation, and the coefficients are larger in absolute magnitude compared to those presented in Table 2. This further supports the hypothesis that FDI decisions to the NMCs are driven by lower costs rather than by providing an export-platform. So far, we have looked at how the differences in labor taxation between host and investing country impact FDI decisions. The host and the investing countries tax rates may have different impacts, however. As Razin and Sadka (2006) suggest, only host country tax rates matter for investment amounts once FDI is present, while the tax rate in the investing country is more important for location decisions. We investigate whether this is the case in our sample, both for labor and corporate taxation, by incorporating the level of the marginal and average labor and corporate tax rates in the host and investing countries. Table 4 presents the results from the regressions for FDI decisions within EU15 and from the EU15 to the NMCs for our preferred specification. Within the EU15 our results suggest that both the labor tax rate in the host and in the investing country matter for the amount of FDI invested, and with expected signs and about equal magnitudes. Only the host labor tax rate is statistically significantly correlated with FDI taking place. Turning to the corporate tax rate, only the rate in the host country seems to matter. A high corporate tax rate in the host country has a negative and statistically significant impact on the investment choice but a positive impact on the amount invested within the EU15. Unlike Razin & Sadka we do not find the corporate tax rate in the investing country to matter for whether investment takes place or not. For FDI investments from the EU15 to the NMCs corporate tax rates seem to play a more important role than labor tax rates. Again host and investing countries tax rates have a similar impact on the amount invested. The host country s tax rate enters with a negative coefficient while the investing country s corporate tax rate increases the amount invested. In the selection equation only the host country s labor tax rate statistically significantly impact whether FDI takes place. In addition, we investigate whether the impact of labor taxation on FDI decisions differs over time by dividing the sample into two subsamples, one comprising the years up to and including 2002 and one comprising the years after The results

17 Labor Taxation and FDI Decisions in the European Union 279 Table 4 Determinants of FDI: levels of tax rates in host and investing country Within EU15 EU15 to NMC Amount Selection Amount Selection Marginal labor tax rate (host) (7.01) b (1.07) Marginal labor tax rate (investor) (7.48) b (1.63) Effective marginal corp tax rate (host) (3.99) b (2.74) b Effective marginal corp tax rate (investor) (1.25) (3.78) b lngdp(investor) (7.91) b (11.71) b (18.63) b (7.00) b lngdp(host) (18.27) b (1.54) (6.61) b (1.70) lndistance (27.46) b (9.58) b (15.38) b (8.11) b Market potential differential (6.13) b (3.52) b (3.05) b (0.57) Labor cost differential (2.58) b (0.10) (3.21) b (2.37) a Government investment differential (0.30) (2.37) b (3.03) b (1.78) Inflation differential (1.53) (1.13) (2.70) b (2.05) b EU membership (host) (0.46) (5.04)** Average labor tax rate (host) (2.50) a (2.15) a Average labor tax rate (investor) (1.72) (0.79) Effective average corp tax rate (host) (2.54) a (1.21) Effective average corp tax rate (investor) (0.98) (0.60) Trade differential (9.67) b (8.58) b GDP per capita differential (5.73) b (0.65) Constant (2.47) a (7.44) b (3.46) b (0.75) Observations 1,967 1,967 1,064 1,064 χ p-value Absolute value of z statistics in parentheses significant at 5 % significant at 1 % Also includes year dummies

18 280 Å. Hansson, K. Olofsdotter Table 5 Different time periods <=2002 >2002 Amount Selection Amount Selection Top marginal labor tax differential (2.80) b (5.70) b Effective marginal corp tax differential (4.54) b (0.68) lngdp(investor) (9.35) b (15.60) b (12.43) b (8.54) b lngdp(host) (13.79) b (2.70) b (12.61) b (4.69) b lndistance (24.81) b (5.15) b (12.69) b (0.67) Market potential differential (7.37) b (5.44) b (7.23) b (0.28) Labor cost differential (8.51) b (1.04) (3.16) b (1.24) Gov investment differential (5.74) b (2.82) b (5.22) b (0.42) Inflation differential (3.85) b (0.21) (1.81) (0.38) EU membership (host) (14.17) b (0.25) (3.89) b (0.56) EU membership (investor) (6.36) b (0.70) (4.70) b (1.41) Implicit average labor tax differential (3.24) b (3.15) b Effective average corp tax differential (2.54) b (2.21) a Trade differential (11.56) b (5.92) b GDP per capita (host) differential (4.82) b (2.10) a Constant (0.27) (12.82) b (6.69) b (12.07) b Observations 3,027 2,500 Zeros 1,700 1,051 χ p-value Absolute value of z statistics in parentheses significant at 5 % significant at 1 % Also includes year dummies

19 Labor Taxation and FDI Decisions in the European Union 281 from these regressions, based on our preferred tax measures (shown in Table 5), suggest that labor taxation has become a more important determinant of FDI since Labor taxation has a negative and statistically significant influence on FDI taking place in both periods and with similar magnitudes. But the impact of labor taxation on the amount invested increases both in absolute value and in level of significance in the later period. Specifically, the semi-elasticity is 1.7 in the earlier and 4.2 in the later period and the level of significance increases from 2.80 to The corporate tax rate changes from having a negative and statistically significant impact on FDI decisions in the earlier period to having a positive and statistically significant impact in the later period. Concerning the amount decision, the corporate tax rate changes from having a positive and significant influence to becoming insignificant. So far we have not included host and investor country dummies. As found by Hansson and Olofsdotter (2010) and Wolff (2007) results can be sensitive to the inclusion of country dummies. Wolff, for example, finds that including country and time controls makes the corporate tax rate coefficient insignificant. This is not surprising as country variation is lost when country dummies are included leaving only time variation in tax rates left for identification. 13 Nevertheless, we incorporate host and investor country dummies to see how sensitive our results are to the exclusion of country dummies. Results are presented in Table 6 were the first two columns report the results without dummies (for comparison) and the last two columns report results when country dummies are included. Inclusion of country dummies does change the results. The labor tax rate becomes insignificant in both the selection and in the amount specification. Including dummies makes the corporate tax rate statistically significant in the selection equation with the expected negative sign. Some of the other control variable becomes less statistically significant as well, for instance, labor cost differences. Finally, we re-run some of the regressions using a Tobit estimation technique. Table 7 reports results from four Tobit estimations for FDI stocks within the EU27 (with and without country dummies), within the EU15, and from the EU15 to the NMCs. The results from the Tobit estimations are similar to those from Heckman s two-step estimations. The top marginal labor tax differential is negatively correlated with FDI stocks within the EU27 and within the EU15. As before, the labor tax differential does not seem to be an important deterrent of FDI from the EU15 to the NMCs, while the corporate tax rate is negatively correlated with the amount of FDI. Again market potential has a positive impact on FDI within the EU27 and EU15 while an insignificant impact on FDI decisions from the EU15 to the NMCs. Strangely, labor cost differences seem to be of less importance for FDI to NMCs than within EU27 and EU15, however. Again, including country dummies, as done in the last column of Table 7, lessens the level of significance and magnitude of several variables including labor taxation and labor cost. Despite the loss in significance the impacts are still statistically significant at standard significance levels. Labor taxation still imposes a negative 13 Following Wolff (2007) we regress the host country dummy on host country tax rates for both labor and corporate taxation. The host labor tax rate is highly significant and results in a R 2 of The host country corporate tax rate is also highly significant but results in a smaller R 2 of 0.22.

20 282 Å. Hansson, K. Olofsdotter Table 6 Determinants of FDI decisions: including host and investor country dummies EU27 EU27 Amount Selection Amount Selection Top marginal labor tax rate differential (5.25) b (0.19) Effective marginal corp. tax differential (1.42) (1.32) lngdp(investor) (5.16) b (18.05) b (1.11) (2.24) a lngdp(host) (19.15) b (4.08) b (2.95) b (4.32) b lndistance (26.66) b (3.18) b (22.91) b (0.86) Market potential differential (10.46) b (5.88) b (3.61) b (5.25) b Labor cost differential (8.70) b (1.56) (1.87) (2.29) a Gov investment differential (0.009) b (1.50) (0.006) (0.32) Inflation differential (3.80) b (0.58) (1.62) (1.19) EU membership (host) (14.10) b (3.41) b (1.65) (1.96) a EU membership (investor) (10.96) b (6.89) b (3.56) b (6.18) b Implicit average labor tax differential (4.45) b (0.65) Effective average corp. tax differential (0.76) (2.52) a Trade differential (12.40) b (0.93) GDP per capita differential (4.10) b (2.37) a Constant (1.37) (5.68) b (2.62) b (6.89) b Country dummies No no yes yes Observations 5,527 5,527 5,527 5,527 χ p-value Absolute value of z statistics in parentheses significant at 5 % significant at 1 % Also includes year dummies