Trading Offshore: Evidence on Banks Tax Avoidance

Transcription

1 Trading Offshore: Evidence on Banks Tax Avoidance Dominika Langenmayr KU Eichstätt-Ingolstadt and CESifo Franz Reiter University of Munich July 2016 Preliminary Version Abstract This paper proposes a novel way in which banks lower their tax burden: the strategic relocation of their proprietary trading activities to low-tax jurisdictions. Using regulatory data from the German central bank, we show that a 10 percentage point lower corporate tax rate increases a banks share of proprietary trading assets in total assets by 5 to 6 percentage points. Furthermore using Bureau van Dijk s Bankscope data set, we show that this increase mostly stems from the relocation of book profits, not of real activities. Keywords: Trading Profit Shifting, Multinational Banks, Corporate Taxation, Proprietary JEL Classification: H25, G21, F21 We thank Mihir Desai, Andreas Haufler, Martin Simmler, Hans-Werner Sinn, John Vella and participants of the 2nd MaTax meeting in Mannheim for helpful comments and suggestions. Dominika Langenmayr visited Oxford University Centre for Business Taxation while working on this paper; she is grateful for their hospitality. She also gratefully acknowledges financial support from the German Research Foundation (LA 3565/1-1). Franz Reiter gratefully acknowledges financial support from the Egon-Sohmen-Foundation. We also thank Deutsche Bundesbank for granting access to the External Position of Banks data base. 1

2 1 Introduction Do banks pay their fair share in taxes? In the last years, concerns arose that they do not: For example, five of the world s biggest investment banks (JP Morgan, Bank of America Merrill Lynch, Deutsche Bank AG, Nomura Holding and Morgan Stanley) paid no corporation tax in the United Kingdom in 2014, despite some of them reporting profits of several hundred million dollars there (Austin, 2015). Because of the immaterial nature of banking activities, it is likely that multinational banks can avoid taxation more easily than firms in other industries. However, very little research has systematically analyzed the ways in which banks lower their tax burden. Such specific research is necessary as the existing literature on corporate tax avoidance usually excludes the financial sector. It does so for good reason: Due to their specific business model, the methods by which banks lower their tax burden are likely different from those in other sectors. Therefore, this paper looks specifically at multinational banks and proposes a novel way of tax avoidance: The strategic relocation of proprietary trading activities. Proprietary trading accounts for a large share of banks profits: In the five years leading up to 2014, proprietary trading accounted on average for 32% of the aftertax profits of German banks (Deutsche Bundesbank, 2015a). Much of these trading activities take place in low-tax countries: Deutsche Bank, for example, has several subsidiaries in tax havens such as the Cayman Islands, Hong Kong, Ireland and Singapore (Deutsche Bank AG, 2014). Here we use a regulatory data set of all German banks foreign activities and find that a ten percentage points lower tax rate increases proprietary-trading-to-total-assetsratio held in foreign affiliates by 5 to 6 percentage points on average. In more detail, we analyze two international bank-level data sets to test the response of proprietary trading activities to tax differentials. In our main analysis we use the External Positions of Banks data base of Deutsche Bundesbank (2015b), a data set provided by the German central bank on all foreign subsidiaries and branches of German banks. While this data set is relatively small (with approximately 80 observations per month), it is of very high quality and provides a complete picture of the foreign activities of German banks. We then use within-bank-group-variation to identify the effects of tax rates on proprietary trading activities. We also confirm that our findings hold in a larger sample by using Bureau van Dijk s Bankscope data set, which has information on all major banks worldwide. 1

3 Using our preferred data set on German multinational banks, we find for banks and their foreign affiliates an average ratio of trading assets to total assets of 15.5%. We show that a 10 percentage points lower corporate tax rate increases this ratio by 5 to 6 percentage points. We obtain similar results when estimating the same regression with more international data from Bankscope. Moreover, we find evidence that the increase mainly stems from an artificial shifting of trading activities: only book profits are transferred to lower-taxed affiliates, but not the real activity of proprietary trading (measured by personnel expenses). These results have important implications for policymakers. Regulation on proprietary trading has increased over the last years, for example based on the Liikanen report from 2012 in the EU or on the Volcker rule from 2010 in the US. Such regulation has often forced banks to spin off their trading activities into separate corporations. However, this legal separation also facilitates the relocation of trading activities to offshore banking centers, which also offer low tax rates. At the same time, these offshore banking centers also regulate trading activities less. Therefore, as the combination of high taxes and high levels of regulation pushes trading activities abroad, governments may loose substantial tax revenues without decreasing the systemic risk from such activities. This paper contributes to the literature in several ways. A first related strand of literature considers the effects of taxation on the location of corporate activities and corporate profits (for a survey see Devereux, 2007). Most of this literature excludes the financial sector. An exception is Demirgüç-Kunt and Huizinga (2001), who provide indirect evidence for profit shifting by multinational banks. They show that the profitability of foreign banks rises relatively little with their domestic tax burden, indicating that foreign banks do not pass the tax on to their consumers, likely because the banks themselves can avoid the tax by shifting profits abroad. Heckemeyer and Overesch (2013) study the influence of taxation on leverage for both banks and nonbanks and show that on average, the marginal effect of taxation is similar in both groups. Gu et al. (2015) show that bank debt reacts to both corporate tax rates and intra-firm tax differentials, indicating profit shifting via internal debt. Most closely related to our paper is Merz and Overesch (2016), who analyze how various balance-sheet items of multinational banks respond to taxation. Their analysis also includes a regression on trading gains, where they find that these profits are particularly responsive to the corporate tax rates. In contrast to our paper, Merz and Overesch (2016) do not differentiate between profit shifting, the relocation of real activity, or a change in the 2

4 amount or profitability of the proprietary trading activities. A second related line of literature considers the determinants of proprietary trading. Studying German equity trades, Hau (2001a) and Hau (2001b) show that foreign traders realize lower proprietary trading profits. Pinheiro and Ferreira (2011) study the characteristics of banks proprietary trading style. Fecht et al. (2013) analyze the interaction between proprietary trading and the returns obtained by the bank for retail investors. So far, this literature has not considered the impact of taxation on proprietary trading. The following section provides some background on proprietary trading and the taxation of banks. Section 3 presents our identification strategy, and section 4 describes the data sets. Section 5 discusses our results. Section 6 concludes. 2 Background: Proprietary Trading Within Banks Proprietary trading are all trades in stocks, bonds, derivatives or any other financial instrument that are carried out with the bank s own money (as opposed to the depositors money). Many banks derive a very large share of their profits from proprietary trading (see Figure 1). In our Bankscope sample, gains from proprietary trading account on average for 39% percent of banks pre-tax profits. Figure 1: Mean trading gains as share of pre-tax-profits Source: Bankscope Proprietary trading activities are highly mobile. As the trading strategy does not have to be developed in the same location as the trading is carried out, relocating 3

5 the trading activities requires the relocation of very few personnel. While some trading activities, especially high-frequency trading, profit from being close to stock exchanges, other trading activities can be carried out from almost anywhere in the world. Thus, there is large scope for relocation in response to taxation. In the following, we will attempt to differentiate between profit shifting and the real relocation of proprietary trading. We will call the relocation of trading activities profit shifting if very little personnel is relocated to the low-tax country, i.e. when the trading strategy is set in a high-tax country and traders in the low-tax country only carry out the exact instructions they receive from abroad. In contrast, if a significant number of employees are relocated, we will classify this as a relocation of real activities. Gains from proprietary trading are usually taxed at the same rate as profits from other banking activities. Note, however, that a few countries have specific corporate tax rates on banks that differ from the standard corporate tax rate (see Table 1 for an overview of the relevant tax rates of countries contained in External Positions of German Banks database). In some countries, capital gains of corporations are taxed at different rates than other profits. For instance, Hong Kong and Singapore have special zero tax rates for corporate capital gains, which also (but not only) apply to profits generated by the propriety trading activities of banks. Table 1: Corporate tax rates (CTR) affecting banks in percent Country German CTR CTR CTR CTR general banks general banks Argentina Yes Australia Yes Austria Belgium Yes Brazil Yes Bulgaria Canada Yes Cayman Islands Yes Chile Yes China Yes Curaçao Czech Republic Denmark Finland France Germany Yes banks active in trading? 4

6 Table 1: Corporate tax rates (CTR) affecting banks in percent Country German CTR CTR CTR CTR general banks general banks Greece Hong Kong Yes Hungary Yes India Yes Indonesia Yes Iran Ireland Yes Italy Yes Japan Yes Jersey Korea Yes Luxembourg Yes Malaysia Yes Malta Yes Mauritius Mexico Yes Netherlands New Zealand Yes Norway Pakistan Yes Peru Yes Philippines Yes Poland Yes Portugal Yes Qatar Russian Federation Yes Saudi Arabia Yes Singapore Yes Slovakia South Africa Yes Spain Sri Lanka Yes Sweden Switzerland Yes Taiwan Yes Thailand Yes Turkey Yes Ukraine Yes United Arab Emirates United Kingdom Yes United States Yes Vietnam Yes banks active in trading? CTR denotes statutory corporate tax rates. indicates special tax rates applying to corporate capital gains such as gains from proprietary trading, not only to banks. The last column indicates whether our preferred data base, the External Positions of Banks data base of Deutsche Bundesbank (2015b), contains bank affiliates holding trading assets in the country. Tax rate data from Ernst & Young (2011, 2014) and KPMG (2016). 5

7 A further potential issue is that controlled-foreign-corporation rules (CFC rules) in high-tax countries could affect bank affiliates in low-tax countries. CFC rules attribute certain income types (usually passive income) arising in foreign subsidiaries to the tax base of the parent company. However, in several countries, e.g. Germany, bank profits are exempt from CFC rules (Deloitte, 2014). As all banks in our preferred data set on the External Positions of German Banks are headquartered in Germany, we will not consider CFC rules. 1 Moreover, in 2013 the German legislator passed a law that requires the separation of commercial banking and proprietary trading in separate legal entities from July However, only very large banks in Germany are affected 2 and several kinds of business (e.g. market making) are not required to be outsourced to specialized subsidiaries. 3 Moreover, our preferred data set is not affected directly as the sample ends already in March 2015 before separation has become binding. However, as a consistency check, we also aggregated the data over all affiliates of a bank group in a country to account for a potential shifting of trading assets between entities in anticipation of the new law. Results with this aggregated data are very similar. 3 Research Design In the following, we will develop the hypotheses for the empirical test and discuss our empirical strategy. 1 German CFC rules completely exclude income from banking under the relatively loose condition of a commercially organized business operation in the foreign affiliate (see Förster and Schmidtmann, 2004; Ruf and Weichenrieder, 2012). As these rules are usually fulfilled with only very few employees in the tax haven, we assume that the CFC rules are not binding. Furthermore, according to a decision by the German Federal Fiscal Court, it is not even necessary that the affiliate has own employees or offices to fulfill the condition of a commercially organized business operation (BFH 13 Oct 2010, I R 61/09). In that case, a service contract with another affiliate was sufficient. 2 German law requires a bank in Germany to separate proprietary trading if its holds more than e100 billion trading assets in its balance sheet or it has total assets of more than e90 billion of which at least 20% are trading assets. 3 For a discussion of the German specialized banking law see Dombret et al. (2014). 6

8 3.1 Hypotheses Our paper aims two answer two questions. Do banks strategically relocate their proprietary trading in low-tax countries? And, if they do so, is this a profit shifting strategy or do they relocate real activities? An extensive literature has shown that firms relocate activities in response to tax rate differentials (for a survey see Devereux, 2007). However, most firms remain headquartered in high-tax countries, and face additional costs when they relocate activities away from their headquarter (Dischinger et al., 2014). Therefore, when deciding which activities to relocate to low tax countries, firms will take into account two factors: first, the cost of relocating this activity; and second, its profitability, which determines the potential tax savings. As explained above, proprietary trading meets these two criteria. It is highly mobile, as it is possible to separate the trading activities from the determination of the trading strategy. And it is highly profitable, as the example of German banks shows: In the five years up to 2014, proprietary trading accounted on average for 32% of their after-tax profits (Deutsche Bundesbank, 2015a). Thus, in the first part of the paper, we test the following hypothesis Hypothesis 1 Proprietary trading activities by banks are decreasing in the corporate tax rate. Banks can accomplish the relocation of proprietary trading activities in two ways: One possibility is that all activities related to proprietary trading (such as the formation of trading strategy, the decision on individual investments and the actual trading) take place in the low tax country. The other possibility is to only carry out the trades in the low-tax country, but decide on the investment strategy abroad. In this second case, the investment specialists that set the overall investment strategy and decide whether to invest in specific securities remain in the headquarter or in other, specialised affiliates. As these investment specialists are well-educated, costly personnel, the tax incentive is to deduct these cost in the high tax country. Thus, we propose the following second hypothesis: Hypothesis 2 An increase in trading activities in response to a tax rate decrease takes place without an accompanying increase in personnel expenses. 7

9 If this hypothesis holds, the relocation of proprietary trading would constitute a profit shifting strategy, similar to profit shifting via the relocation of patents in industrial firms. 4 It is important to separate profit shifting strategies from the relocation of real activities (which would be the case if all trading activities were relocated), as the welfare implications of the two activities may be different (Hong and Smart, 2010). We now outline our empirical approach before turning to the description of the data. 3.2 Econometric Approach Test of Hypothesis 1 Our first empirical specification tests whether more proprietary trading activities take place in low-tax affiliates. To do so, we look at the variation in tax rates that different affiliates of a multinational bank face. Accordingly, we estimate the following equation to test hypothesis 1: T rading ijkt T otalassets ijkt = β 0 + β 1 CT R jt + β 2 X ijt + δ k + Y ear t + u ijkt. (1) The dependent variable, T rading ijkt T otalassets ijkt is the ratio of trading assets to total assets held by affiliate i of bank-group k in country j as of year-month t. 5 We will use two data sources for this variable, the External Positions of Bank database by the Bundesbank and Bureau van Dijk s Bankscope, both of which are discussed in Section 4. The main explanatory variable of interest is CT R jt, the corporate tax rate of country j. We additionally use several control variables X ijkt, discussed below. δ k are dummies capturing fixed effects for bank-groups, and Y ear t are year dummies. If hypothesis 1 holds, we should observe β 1 < 0, as banks prefer low-tax countries to conduct their proprietary trading business. A threat to our identification strategy is that country characteristics other than the tax rate determine a country s attractiveness for proprietary trading. To address this concern, we would ideally add country fixed effects to our regression (and we will do so in robustness tests). However, the inclusion of country fixed effects also 4 For empirical evidence on the relocation of patents, see e.g. Karkinsky and Riedel (2012). 5 As an alternative dependent variable, we used the inverse hyperbolic sine of an affiliate s trading assets. Results were qualitatively and quantitatively similar. 8

10 poses a challenge to identification, as we then use only the information on affiliates in countries were the tax rate changes. While there are 35 changes in statutory tax rates in our sample, none of the tax havens in the sample changes its tax rate. Thus, the specification with fixed effects excludes exactly the countries to which we think that banks strategically relocate their proprietary trading business. Therefore, in our main regressions we do not use country fixed effects and instead employ several country-level control variables. In particular, we control for the log of the population as a proxy for country size, as larger countries also provide a larger market for raising funds that banks can use for proprietary trading. We also include inflation rates, as higher inflation can on the one hand discourage trading activities in a country because of higher risk premiums, and on the other hand make alternative capital investments at fixed nominal interest rates less attractive (lowering opportunity costs of proprietary trading). We control for GDP growth as countries that grow at higher rates offer more attractive markets for banks. We also include the share of country j s financial sector (financial and insurance activities) in the gross value added to account for the attractiveness of financial centers as the location of proprietary trading business. 6 In additional regressions, we will also control for regulation using an index on regulation of banks securities activities. To allow for a more precise estimation, we also include the natural log of total assets as a bank-level control variable to account for an affiliate s size. In the Bundesbank data set, we observe all German bank groups with foreign affiliates. Thus, all headquarters in the sample are in Germany. As Dischinger et al. (2013) show that firms are often reluctant to shift profits away from headquarters, we also rerun the regressions that use the Bundesbank data while excluding the German parents from the sample. We would expect to see stronger effects in this subsample, if the results from Dischinger et al. (2013) also hold for banks Test of Hypothesis 2 In the second part of the empirical analysis, we test whether the relocation of proprietary trading is mostly a shifting of book profits or the result of the relocation 6 The share of financial and insurance activities in total gross value added reflects the role of important financial centers: for instance in 2014 in the United Kingdom it is 8% and in Singapore 13%, compared to 4% in Germany and 4% in France. 9

11 of real activities. As an indicator for real activity we use variables on employment in the affiliates. Our second hypothesis predicts that an increase in trading activities in response to a tax rate decrease takes place without additional personnel expenses. To test this hypothesis, we use the following model: ln(p ersonnelexp ijt ) = β 1 CT R jt + β 2 ln(t rading ijt ) + β 3 CT R jt ln(t rading ijt ) + β 4 X ijt + Y ear t + u ijt. (2) The dependent variable is now ln(p ersonnelexp ijt ), the natural logarithm of personnel expenses in country j by bank i. T rading ijt are total trading assets held in bank affiliate i, and the other variables are as defined above. If a higher volume of trading in an affiliate is also associated with more personnel conducting these activities, we expect a positive estimate for β 2. However, given hypothesis 2 is true, we rather expect a positive β 3. This indicates that an increase in trading assets is associated with an increase in personnel expenses in high-taxed but not in low-taxed affiliates, implying that rather than shifting real traders, only the process of buying and selling is shifted to tax haven affiliates. The External Positions of Banks data base of the German central bank does not contain any information on employment, and therefore we can only use the Bankscope data set for this analysis. As we will discuss in the following section, this dataset has several disadvantages, in particular it only has information on subsidiaries, not on branches. The Bundesbank data shows that branches are heavily used in low-tax countries. The missing data should work against us finding the results predicted by hypothesis 2. In addition to the control variables discussed above, we also include customer deposits in X ijt as the customer business is very personnel-intensive. Besides employment, total personnel expenses obviously also depend on wage levels in the host countries. Therefore we additionally control for the annual average wage in country j. We use log transformations for personnel expenses, trading assets and all level control variables (total assets, customer deposits, population, wages), which allow us to interpret the estimated coefficients as elasticities. As a robustness check, in a further regression we use dummies for whether a country is classified as a tax haven by Dharmapala and Hines (2009) or Johannesen and Zucman (2014) instead of corporate tax rates in regression 2. 10

12 Because we identify differences in the effects of trading assets on personnel expenses depending on differences in absolute corporate tax rates, we cannot use country fixed effects. The main country-specific factors influencing personnel expenses are captured by the control variables we use, in particular we now control additionally for a country s average annual wage. Nevertheless, as a further robustness check we also split our sample into tax havens and non-haven countries and estimate the effect of trading assets on personnel expenses in these two subsamples separately, including countryfixed effects. Given hypothesis 2 is true, we expect a significantly positive coefficient for trading assets for non-haven countries and a smaller, less significant coefficient for tax havens. 4 Data and Descriptive Statistics The main data set we use is the External Positions of Banks data base of the German central bank (Deutsche Bundesbank, 2015b). German multinational banks are obliged to report detailed information on financial assets and loans of all their foreign subsidiaries and branches, and of the German parent companies. The data set includes information on fixed-interest assets that are held for proprietary trading and that we use as our main dependent variable. 7 As reporting to the Bundesbank is mandatory for all German banks with foreign activities, we have high-quality data covering a complete sample of German banks trading activities. Data is available on a monthly basis from June 2010 to March Figure 2 illustrates the yearly average total amount of trading assets held by German banks. In 2011, German banks held e139 billion in trading assets (foreign affiliates held 57 % thereof). In the two following years the total volume of trading assets declined to e110 billion. In 2014 it increased again to e116 billion, 53 % of which were held abroad. In which countries do German banks hold their trading assets? In Table 2 we list the top 10 countries in which German bank groups had proprietary trading assets in 2011, in the left panel ranked by the total trading assets that are held by German banks in a country and in the right panel by the average amount of trading assets in 7 This variable reflects the actual current value of trading assets held in an affiliate. The lowest value principle (which is the mandatory accounting principle for assets in Germany) does not apply to bank assets held for trading. In line with international financial reporting standards trading assets have to be assigned their fair value. 11

13 Figure 2: Total trading assets held by German bank groups Data is on total amount of fixed interest securities that are held for trading by German multinational banks. Data source is the External Positions data base of Deutsche Bundesbank (2015b). a bank affiliate. Note that we can not list countries in which less than three German bank groups conduct proprietary trading in Table 2 due to confidentiality reasons. Nevertheless, among the countries we are allowed to report on, most trading assets are held in parent companies in the home market Germany. Almost equally important is the United Kingdom, which is an important financial center in Europe (but also offers a considerably lower corporate tax rate than Germany). Comparing the average trading assets per affiliate reveals the important role of the United Kingdom in banks proprietary trading business: an average affiliate in the UK holds e5.6 billion in trading assets. Also some well-known tax haven countries play an important role in the banks trading activities: Singapore with a total of trading assets of around e4 billion, the Cayman Islands with around e3.4 billion and Hong Kong with e1 billion in German banks trading assets in All of these countries offer a zero tax rate for capital gains. 8 8 In the United States, a substantial part of trading assets is likely to be placed in affiliates in Delaware where banks can also profit from various corporate tax benefits. For instance, seven of Deutsche Bank s eight securities trading firms in the US are based in Wilmington, Delaware (Deutsche Bank AG, 2014). Unfortunately we cannot observe the exact location of a bank affiliate within the US in our data set. 12

14 Table 2: Top 10 countries for trading activities in 2011 Total trading assets Mean trading assets Rank Country Total trading Rank Country Mean trading (in e1,000) (in e1,000) 1 Germany 56,301,897 1 United Kingdom 5,591,453 2 United Kingdom 44,055,216 2 Germany 1,598,885 3 United States 9,452,124 3 Singapore 1,252,735 4 Singapore 4,082,724 4 United States 1,168,929 5 Cayman Islands 3,411,113 5 Cayman Islands 975,398 6 Hong Kong 1,037,524 6 Hong Kong 251,182 7 Russia 705,469 7 Russia 236,549 8 Poland 520,571 8 Poland 173,524 9 Luxembourg 339,863 9 Brazil 94, Brazil 257, Hungary 82,007 Total 138,500,000 Total 845,230 Data from External Positions of Banks data base of Deutsche Bundesbank (2015b). Euro values are country totals and means of fixed interest securities that are held for trading by German multinational banks. Countries in which less than 3 banks are active are not shown here due to confidentiality reasons. 13

15 The External Positions of Banks data base also provides information on total assets held in an affiliate which we use to control for an affiliate s size. Moreover, we use a set of country level control variables: the log of population, inflation rates, GDP growth and the share of the financial sector in a country s gross value added. We collect these macroeconomic controls from the International Financial Statistics of the International Monetary Fund, from the World Bank s World Economic Indicators data base, from the OECD data base and complement this information with data from national statistical agencies if necessary. All variables except population, GDP growth and financial sector share are available on a monthly basis. Population is an annual variable which we transform to monthly frequency by cubic spline interpolation. The share of the financial sector in gross value added is available on a quarterly base; we also transform it to monthly frequency by cubic spline interpolation. Monthly GDP growth is calculated from cubic spline interpolated absolute real GDP values which are available on a quarterly basis. To account for international regulatory differences we use the regulatory index on banks securities activities that is provided by a survey on bank regulation by the World Bank (2011). It ranges from 1 (unrestricted) to 4 (prohibited). Unfortunately the latest available data for this index is from the survey in 2011, thus we take it as a time constant variable in our regression. Table 3 gives an overview over the variables that we use in our regressions. We also want to investigate in an extension whether banks actually move traders to lower-taxed affiliates. For this we need information on employment in foreign bank affiliates which the External positions of German banks data base does not provide. Therefore we use the Bankscope data base compiled by Bureau van Djik to conduct this extension. Moreover Bankscope provides data on banks worldwide so that we can also test our main hypothesis with a more international sample and for a longer time period ( ). The dependent variable we use is the share of total assets held for trading in total assets of a bank affiliate. Descriptive statistics for the Bankscope sample are shown in the lower part of Table 3. 14

16 Table 3: Descriptive Statistics Variable Obs. Mean Std. Dev. p1 p50 p99 Frequ. Source External Positions of German Banks data base (2010/ /03) Trading assets / Total assets 4, M External Positions data base Corporate tax rate 4, M Ernst & Young (2011, 2014) Total assets (in million e) 4,255 23,043 60, , ,900 M External Positions data base Population (in thousands) 4, , , ,334 1,338,773 A M World Bank, IMF Inflation rate (%) 4, M IMF GDP growth (%) 4, Q M IMF Regulation 4, Survey of World Bank (2011) Financial sector share 4, Q M OECD, complemented by data from national statistical agencies Bankscope data base ( ) Trading assets / Total assets 2, A Bankscope Corporate tax rate 2, A Ernst & Young (2011, 2014) Total assets (in million USD) 2,953 18,300 92, , ,000 A Bankscope Population (in thousands) 2, , , ,652 1,344,668 A World Bank, IMF Inflation rate (%) 2, A IMF GDP growth (%) 2, A IMF Regulation 2, Survey of World Bank (2011) Financial sector share 2, A OECD, complemented by data from national statistical agencies Average annual wage (in USD) 2,448 45,792 21,463 2,273 51,238 94,812 A OECD, complemented by data from national statistical agencies Personnel expenses (in million USD) 2, , ,420 A Bankscope Customer deposits (in million USD) 2,448 10,300 60, , ,000 A Bankscope Trading assets in the External Position of German Banks data base is fixed-interest assets that are held for trading; in the Bankscope data base it is the total amount of assets held for trading. M/Q/A indicate monthly, quarterly and annual frequency, respectively. In regressions using German bank data, population is transformed to monthly frequency by cubic spline interpolation. Monthly GDP growth is calculated from cubic spline interpolated absolute real GDP values. Financial sector share is the share of finance and insurance activities in total value added in a country. Monthly frequency is calculated by cubic spline interpolation. Regulation is an index for the extend to which banks may engage in securities activities in a country, ranging from 1 (unrestricted) to 4 (prohibited). 15

17 Despite these advantages of Bankscope, there is a substantial drawback with regard to the extent of affiliates covered in the data set. Firstly, Bankscope only has information on subsidiaries but no information on branches. This is a major disadvantage as the comprehensive Bundesbank data set shows that 57 % of all affiliates of German banks that engage in trading are branches rather than subsidiaries. Secondly, the coverage of the data is unclear. There are many missing values for total trading assets, and we do not observe all subsidiaries of multinational bank groups. For example, the Bundesbank data base reports seven subsidiaries of German banks that are active in trading in Singapore. 9 But in Bankscope there is only one German-owned bank active in Singapore, and there is no information on its trading assets. Overall we therefore prefer the Bundesbank data due to its comprehensive sample coverage and its excellent quality. Nevertheless we also use Bankscope as a consistency check for our results and to investigate further issues that we cannot analyze with Bundesbank data. 5 Estimation Results In this section we present the regression results. Table 4 reports results for estimations with External Positions of Banks data of Deutsche Bundesbank (2015b). In Table 5 we test our hypotheses with Bankscope data. 5.1 Trading Activities in German Banks Regressions in Table 4 analyze the effect of statutory corporate tax rates on trading assets held by German banks and their foreign affiliates. The dependent variable is the share of fixed-interest trading assets in the bank affiliate s total assets. We bootstrap standard errors and cluster them by bank and country-year. In columns (1) to (3) we use the complete sample of German bank affiliates including German headquarters. In regression (1) the significantly negative coefficient of indicates that a ten percentage point lower corporate tax rate implies on average a 5.2 percentage points higher ratio of trading assets to total assets. Compared to the mean trading-to-totalassets-ratio of 15.5% in our sample, this implies on average an increase by one third. 9 In the External Positions of Banks data base of Deutsche Bundesbank (2015b) we observe seven subsidiaries that hold either fixed-interest trading assets or trading derivatives (or both). As data on derivatives is only available from a relatively small time period beginning in December 2013, in our analysis we focus on fixed-interest trading assets as dependent variable (available from June 2010). 16

18 In regression (2) we add an index for a country s strictness in the regulation of banks securities activities, with higher values indicating stricter control. This changes our results only slightly, the corporate tax rate coefficient of again indicates a preference for trading in low-taxed affiliates. In the regression in column (3) we use country fixed effects rather than bank group fixed effects. The estimate for the corporate tax rate is again in an equal range as in regressions (1) and (2), however it is not significant. Results for the subsample of foreign affiliates (completely excluding German headquarters from the analysis) are similar: regressions (4) and (5) indicate a 5.7 to 5.8 percentage points lower trading-to-total-assets-ratio in response to a ten percentage points higher corporate tax rate. With a mean of 18.1% in foreign affiliates this corresponds to an average decrease by 32%. The use of country fixed effects rather than bank group fixed effects leads to an even more negative but insignificant coefficient. The insignificant results when using country fixed effects may come from the fact that we have only 35 changes in corporate tax rates in our sample. Most of them were tax rate cuts in countries that initially had a relatively high corporate tax rate and thus are unattractive locations for proprietary trading. Countries that are most attractive for the profitable trading business from a tax perspective have mostly stable (and often zero) corporate tax rates over the sample period. Thus using cross-country variation adds important information. Overall, the regressions shown in Table 4 confirm our first hypothesis. Banks trading business is very mobile and multinational banks locate their trading activities mainly in affiliates in low tax countries. What conclusions can we draw from the control variables? In all our regressions we find a negative estimate for the log of total assets, with significant coefficients when using bank group fixed effects. This implies that proprietary trading plays a more important role in smaller affiliates. A larger population also leads to a higher tradingto-total-assets-ratio in an affiliate in the regressions without country fixed effects. The share of the financial sector in a country s total gross value added yields no significant coefficient in all of the regressions in Table 4. This implies that the share of trading assets in total assets is not particularly high in financial centers and that banks do not locate disproportionally many trading assets relative to other activities in these financial centers. 17

19 Table 4: Effect of Tax Rates on Proprietary Trading Sample: All entities Foreign affiliates only (1) (2) (3) (4) (5) (6) Corporate tax rate (-4.34) (-5.02) (-0.63) (-5.04) (-4.37) (-1.12) ln(total assets) (-2.01) (-1.68) (-1.24) (-1.89) (-1.68) (-1.02) ln(population) (3.47) (3.88) (0.20) (3.81) (3.29) (-0.04) Inflation rate (0.83) (0.80) (-1.83) (0.75) (0.78) (-2.01) GDP growth (0.25) (0.18) (0.51) (0.18) (0.08) (0.65) Financial sector share (-0.77) (-0.67) (-0.02) (-0.58) (-0.47) (-0.06) Regulation (0.28) (0.21) Year FE Yes Yes Yes Yes Yes Yes Bank group FE Yes Yes No Yes Yes No Country FE No No Yes No No Yes R Observations 4,255 4,255 4,255 3,006 3,006 3,006 The dependent variable is fixed interest securities that are held for trading divided by bank s total assets. Financial sector share captures a countries share of gross value added that is acquired in the banking and insurance sector. Regulation is an index for the extend to which banks may engage in securities activities in a country, ranging from 1 (unrestricted) to 4 (prohibited). Monthly bank data for 2010/ /03 from External Positions of Banks data base of Deutsche Bundesbank (2015b). t-statistics in parentheses, based on bootstrapped standard errors clustered by bank and by country-year. 18

20 5.2 Trading Activities in Banks Worldwide Whereas our preferred data set from the German central bank is very precise and comprehensive, it is restricted to German multinational bank groups and their foreign branches and subsidiaries. As a consistency check of our results in a more international context we also use Bankscope data provided by Bureau van Djik and repeat our regressions. This data set will also form the basis for the analysis in Section 5.3. Table 5 presents the results. Table 5: Effect of Tax Rates on Proprietary Trading, Bankscope Data Sample: All entities Foreign affiliates only (1) (2) (3) (4) (5) (6) Corporate tax rate (-2.11) (-2.12) (0.24) (-2.02) (-2.01) (0.63) ln(total assets) (-1.00) (-0.81) (-1.96) (-1.12) (-0.94) (-1.59) ln(population) (-0.04) (1.12) (-0.15) (-0.06) (1.09) (0.18) Inflation rate (0.54) (0.37) (-0.99) (0.54) (0.38) (-1.28) GDP growth (0.19) (0.84) (1.51) (0.15) (0.82) (1.30) Financial sector share (-2.21) (-2.38) (-0.37) (-2.11) (-2.30) (-0.06) Regulation (-1.86) (-1.84) Year FE Yes Yes Yes Yes Yes Yes Bank group FE Yes Yes No Yes Yes No Country FE No No Yes No No Yes R Observations 2,953 2,953 2,953 2,389 2,389 2,389 The dependent variable is total trading assets divided by bank s total assets. Financial sector share captures a countries share of gross value added that is acquired in the banking and insurance sector. Regulation is an index for the extend to which banks may engage in securities activities in a country, ranging from 1 (unrestricted) to 4 (prohibited). Yearly bank data for from Bankscope data base. t-statistics in parentheses, based on bootstrapped standard errors clustered by bank and by country-year. 19

21 Estimates for multinational data presented in Table 5 qualitatively confirm our results that we found for German bank groups. In regressions (1) and (2) the coefficient of the statutory corporate tax rate is significant and negative, indicating a higher trading-to-total-assets ratio in lower-taxed affiliates. The results suggest that a ten percentage points lower tax rate induces a 2.6 percentage points higher trading-tototal-assets-ratio in an affiliate. As the share of trading in total assets is on average 6.2% in the sample this corresponds to an decrease on average by 42%. Excluding headquarters from the sample leads to very similar results. Country fixed effects rather than bank group fixed effects again yield insignificant coefficients for the corporate tax rate. Note that comparability to the results in the previous section is limited for the following reasons: First, the dependent variable here captures the total amount of assets held in the trading book whereas in the regressions for German banks we employ the specific balance sheet position of fixed-interest assets held for trading. Second, as outlined in section 4 the trading data in Bankscope suffers from a substantial number of missing values and the fact that a major part of proprietary trading is conducted in foreign branches, whereas Bankscope only contains subsidiaries of banks. Nevertheless results on the corporate tax influence on trading assets held in subsidiaries are qualitatively comparable to those we found for our comprehensive Bundesbank dataset for German multinational banks. Thus also with data on bank groups worldwide we can confirm our hypothesis that banks tend to locate their proprietary trading assets in low-taxed affiliates. Regarding the control variables an interesting result arises for the regulatory index: as expected, affiliates facing a stricter regulation in securities activities hold a lower share of trading assets in total assets. The financial sector share has a negative effect on the trading-to-total-assets ratio in regressions using bank group fixed effects, although it is close to zero. This confirms the finding for German multinational banks that banks do not disproportionally locate trading activities compared to other activities in affiliates located in financial centers. 5.3 Profit Shifting or Shifting of Real Activity? Trading business is presumably very mobile, but traders do not necessarily have to reside in the country where trading takes place. A bank group that wants to relocate some of its trading business into a tax haven affiliate may choose a physical move of 20

22 trading personnel. It is also possible to endow the tax haven affiliate with a minimum of employees who simply execute orders they get from the traders in the high-taxed headquarter (or any other high-taxed affiliate). Highly developed trading technology nowadays ensures that there would be no organizational obstacles to this strategy. Then the bank group could benefit at the same time from both highly qualified trading personnel in developed (high-tax) countries, from tax-deductibility of personnel expenses for traders in the high-tax country and from the tax advantages tax havens offer. Unfortunately our preferred data set on German banks does not provide any information on employment in bank affiliates. Also Bankscope does not report the number of employees in an affiliate, but at least total personnel expenses of a bank subsidiary. Besides the number of employees, also the wage level in a country influences the total amount of personnel expenses. Therefore we use it as an additional country-level control variable. In our basic specification we regress a subsidiary s log of personnel expenses on the natural logarithm of total trading assets held in this subsidiary. To specifically test the impact of trading volume on employment in low-tax countries we then add an interaction term between the log of trading assets and the corporate tax rate in regression (2), and between trading assets and a tax haven dummy in regression (3). Table 6 shows the results of these specifications. In regression (1) without the interaction term between trading assets and the corporate tax rate a significantly positive coefficient for trading assets arises, indicating that banks that engage more in trading also need more personnel that conducts trading activities. But including the corporate tax rate and the interaction term between the corporate tax rate and trading assets in column (2) reveals that this relationship only holds for affiliates in high-tax countries: the coefficient for trading assets is no longer positive. Instead we get a significantly positive coefficient of for the interaction term in column (2). This suggests that a higher amount of trading assets leads to more employment primarily in high-taxed subsidiaries. The lower the corporate tax rate the weaker is this relationship. Similar results arise when including a tax haven dummy and an interaction term between this tax haven dummy and trading assets: the coefficient for trading assets in regression (3) is significantly positive, and it is almost completely offset if the country is a tax haven. These results speak against the shifting of real activities (or traders). In columns (4) and (5) we split the sample into tax havens and non-haven countries and use countryfixed effects. For tax havens we get a weakly significant and positive coefficient of for trading assets, for non-havens it is highly significant and more than twice as large 21

23 Table 6: Effects on real activity Sample: All countries Havens Non-Havens (1) (2) (3) (4) (5) ln(trading assets) 0.044*** *** 0.046*** 0.024* 0.053*** (6.70) (-2.78) (7.01) (1.84) (6.56) Corporate tax rate ** (-2.28) CTR * ln(trading assets) 0.344*** (4.44) Haven dummy 0.723*** (2.61) Haven * ln(trading assets) ** (-2.05) ln(customer deposits) 0.590*** 0.584*** 0.592*** 0.682*** 0.578*** (15.88) (16.02) (16.00) (12.60) (13.30) ln(wage) 0.352*** 0.261** 0.340*** *** (3.64) (2.41) (3.74) (-3.54) (0.82) Inflation rate 0.131*** 0.127*** 0.137*** * (3.86) (4.16) (4.20) (-1.88) (1.35) ln(population) 0.215*** 0.150*** 0.235*** (7.65) (4.26) (7.57) (-0.86) (0.95) GDP growth ** ** (-2.24) (-0.96) (-2.57) (0.71) (0.52) Financial sector share (-0.96) (-0.55) (-1.22) (-1.53) (-1.15) Year FE Yes Yes Yes Yes Yes Country FE No No No Yes Yes R Observations 2,448 2,448 2, ,101 Yearly bank data for from Bankscope. Dependent variable is log of personnel expenses. Wage is the average annual wage in the respective country. Financial sector share captures a countries share of gross value added that is acquired in the banking and insurance sector. t-statistics in parentheses, based on standard errors clustered by bank and by country-year. 22