Dual Income Taxation, Deductions and Income-Shifting. Olli Ropponen 1

Transcription

1 Dual Income Taxation, Deductions and Income-Shifting Olli Ropponen 1 Abstract: This paper employs a simple framework to illustrate the sharp incomeshifting incentives observed in Finnish dual income tax system. Deductions are shown to steepen these incentives and are therefore likely to increase tax compliance costs and may cause unnecessary fluctuations between the state tax base and the tax bases of municipalities. The deductions are also shown to exacerbate those cases where poorer entrepreneurs have to pay higher taxes than better off entrepreneurs. The observed negative features arising from inclusion of deductions into dual income tax system could be reduced by a suitable tax reform. (JEL H2, H25, H3, H32) Keywords: Dual income tax, Income-shifting, Tax Deductions 1 Introduction Globalization has pushed forward numerous changes around the world. A combination of an increase in both the international capital mobility and the role of multinational companies (MNCs) in the global economy has placed countries into more competitive environment than earlier. The use of active tax planning strategies of MNCs has pushed also countries into harmful tax competition in order to protect their tax bases (Fuest, Huber and Mintz 2005; Devereux and 1 VATT Institute for Economic Research, P.O. Box 1279 (Arkadiankatu 7), FI Helsinki, Finland. Tel: Fax: address: olli.ropponen@vatt.fi. 1

2 Loretz 2013). To become better equipped against other countries tax changes several countries have started to search for fundamental tax reforms that make them capable to tax different income components separately. The desire to choose more targeted tax treatments has supported a tendency for countries to replace their global income tax (GIT) systems, where all the income are taxed together, by the dual income tax (DIT) systems, where capital income is taxed separately from other sources of income (Sørensen 1994; Nielsen and Sørensen 1997). The discussion about the relative merits and the drawbacks of the DIT systems compared to GIT systems has been already going on for a while. Numerous equity and efficiency arguments favoring the DIT has been put forward. However, also some negative features are observed. Especially the income-shifting between labor income and capital income among self-employed has made their taxation difficult in the DIT systems (see Sørensen 1994; Boadway 2004; Sørensen 2005a,b; Pirttilä and Selin 2011). This paper provides new aspects for the ongoing discussion about the dual income taxation. It first shows in a simple framework that a DIT system combined with a single (EITC-like) deduction provides an entrepreneur with sharp and complicated income-shifting incentives, which are likely to increase the tax compliance costs and to make the tax bases of the state and the municipalities to fluctuate. Second, it shows that within this framework poorer entrepreneurs may end up paying higher taxes than better off entrepreneurs. In a DIT system an entrepreneur chooses between two types of income and his tax-bill is determined according to both the capital income tax rate and the tax rate for other income. His wage is considered as a remuneration for the labor 2

3 input and taxed according to earned income tax schedule, whereas the dividend is considered as a remuneration for capital input and taxed according to capital income tax rate. These two types of income provide some clear incentives for an entrepreneur. Very high earned income tax rates are likely to take the choice between wage and dividends towards dividends and the other way around for low earned income tax rates. Because of the multiple remuneration types, both income tax schedules matter for an entrepreneur reporting behavior and his tax-bill in a DIT system. Therefore, compared to a pure wage-earner or an investor, an entrepreneur has also an additional opportunity to affect his taxbill by choosing the relative levels of different remuneration types. Despite some of its drawbacks and motivated by its relative merits, several countries have replaced their GIT systems by the DIT systems. While Nordic countries (except Iceland) were among the first ones to implement this reform, it has been considered as an interesting option and worthwhile to being implemented also in several other countries. (Schratzenstaller 2004; Spenger and Wiegard 2004; Eggert and Genser 2005; Keuschnigg and Dietz 2007; Kleinbard 2010) After the first implementations also some improvements to the basic DIT system have been introduced. A good example of such improvement is a Norwegian reform in 2006, which resulted in multiple new neutrality properties (neutrality with respect to investment, financing and realization decisions) into their tax system (see Sørensen 2005a). Some further improvements for the tax system were suggested in Sørensen (2005b). The cornerstones of this proposal are the rate-of-return allowance (RRA) and the full carry forward for the unutilized RRA. This model taxes imputed normal return to shares 3

4 (determined via RRA) as capital income and above-normal returns as labor income. Compared to the current Finnish tax system it has multiple advantages. One major advantage is that the income-shifting incentive is reduced in this system. The paper proceeds as follows. The follow-up section reviews the related literature. Section 3 introduces a framework employed in the paper and derives its properties by providing both analytical and numerical results for it. Section 4 discusses the similarities between the framework and the Finnish tax system. Section 5 employs Finnish administrative data to provide empirical evidence on entrepreneur behavior. Finally, section 6 concludes. 4

5 2 Literature A DIT system combines a flat rate (or just slightly progressive) capital income tax with a progressive income tax schedule for other sources of income. Unlike in a GIT system, the tax rate for capital income may thus differ from the tax rate for other sources of income in a DIT system. For a pure wage-earner the level of capital income tax does not matter much in a DIT system 2 as his income is automatically considered as earned income and therefore his taxes are determined according to the earned income tax schedule (τ W ). A pure investor is not directly affected by earned income taxation as his compensation is considered as capital income and is thus taxed according to the capital income tax rate (τ D ). Thus the relative levels of the two types of income taxes do neither matter much for a wage-earner nor an investor tax-bill. For an entrepreneur the situation is very different. The discussion on the DIT systems has been very active since the 1990s. In his pioneering paper Sørensen (1994) provides a profound discussion on the most relevant issues related to DIT systems. It also discusses about the relative merits of these systems compared to GIT systems, where both capital income and earned income are first aggregated and then taxed with a single progressive tax schedule. Both theoretical (equity and efficiency) arguments and practical reasons are provided. The equity arguments for the dual income tax systems include reduction of the horizontal discrimination (according to life-cycle differences), reduction of tax favoritism towards human capital in- 2 The level of capital income tax does not matter for wage-earner directly, but might matter indirectly due to substitutability of different taxes. With a given total tax revenue, the level of earned income tax is likely to be higher the lower the level of capital income tax. 5

6 vestment compared to other investments and reduction in problem of taxing nominal rather than real income (inflation). The efficiency arguments for the DIT systems include the notions that the capital is likely to be more flexible than labor supply and that the DIT system reduces the disincentives to save. However, taxing of self-employed is considered as the Achilles heel of the DIT system, because compared to GIT systems, the DIT systems provide new room for tax avoidance via income-shifting from high-tax labor income (wage) to low-tax capital income (dividends). Nielsen and Sørensen (1997) study the DIT and contrast it with the GIT. They show formally that the DIT can be given rationality on pure efficiency grounds. Especially, via reducing the distortionary effects of progressive capital income taxation on the GIT systems. They also show that the tax favoritism towards investment in human capital, compared to other investment in the GIT systems, is offset by a DIT system. In addition to the theoretical studies also empirical studies on the DIT systems exist. The choice between company owner s wage and dividends is studied especially in Nordic countries. Fjærli and Lund (2001) study empirically this issue in the Norwegian DIT system. They employ micro data and show that in line with the incentives provided by the tax system the choice of a type of payout from corporations to owners is strongly motivated by taxes. They also find that social security benefits drive the division between wage and dividends. The wage may be paid in order to receive social security benefits even if it may not be tax-optimal. Alstadsæter and Jacob (2013) study this question with the Swedish data and find that closely held corporations are in many cases utilized in income-shifting to reduce tax burden. They also emphasize that the 6

7 conditions for the tax avoidance seems to require in addition to the incentive to avoid taxes, also the access to income-shifting channel and the awareness of the tax rules. Le Maire and Schjerning (2013) employ the Danish data and find evidence of income-shifting among the self-employed. Pirttilä and Selin (2011) study the effects of Finnish tax reform, where a GIT was replaced by a DIT. They find that the reform, which decreased the capital income tax rates, resulted in a small positive impact on the overall taxable income. They also show that self-employed substantially increased their capital income. This suggests that the positive response was probably partly offset by Achilles heel of dual income tax, income-shifting among self-employed. Harju and Matikka (2014) employ a dividend tax change to study income-shifting among the privately held businesses in Finland. They find active income-shifting behavior between tax bases and a positive relation between the tax incentives and the size of response. Motivated by the problems of present tax systems also multiple other than Nordic countries have put DIT systems in top of their lists of candidates for the tax reform. Spengel and Wiegard (2004) discuss problems observed in German corporate tax system. They conclude that their tax system should be simplified and made more neutral (w.r.t. investment decisions, financial decisions and choice of legal form) and its tax rates should be reduced. They conclude that the tax rate reductions are considered to be of most importance among mobile tax bases, which are prone to international tax competition. To tackle the observed problems they consider DIT and flat tax as leading candidates for the German tax reform. Keuschnigg and Dietz (2007) propose a tax reform (allowance for corporate equity combined with broad flat tax on 7

8 personal capital income), with neutrality properties with respect to investment, firm financial decision and organizational form. By suitably chosen tax rates it also prevents income-shifting from labor to capital income. Their simulation results suggest that the reform could add to GNP by 4-5 percent in the longrun in Switzerland. Eggert and Gensen (2005) study the implemented dual income tax structures. They conclude that it is not a simple task to implement separate taxation on capital and labor, yet there is (not much or) little pressure towards reintroducing a comprehensive income tax system. They consider an easy implementation of corporate income taxation (CIT) and personal income taxation (PIT) as one major advantage of DIT. They also conclude that the separate tax base for capital income allows countries to coordinate without them losing the sovereignty of choosing the labor income tax rates. The Nordic DIT system is largely motivated by addressing (both administrative and economic) problems in comprehensive income taxation as there seem to be little justification for taxing capital income at the same rate as labor income (Boadway 2004). Capital income taxation could also be coordinated (without a necessity to coordinate labor income simultaneously) if all nations (or e.g. all EU countries) adopted this system. Schratzenstaller (2004) surveys income tax designs of 15 EU member states, 10 central and east European accessions, US, Japan and Switzerland and finds a clear tendency away from comprehensive income tax systems and towards systems like DIT. She, however, argues that it is debatable whether this evolution is good, that is whether the pros of these changes are bigger than their cons. As a positive feature of these systems she concludes that the aim to reduce international capital flight seems to be well-founded among European countries. Sandmo 8

9 (2005) takes a different point of view into discussion about the DIT systems as he reviews the reasons for inequality growth in the OECD area. Regarding taxation he concludes that the expenditure tax may be better than DIT, both on efficiency point of view and redistribution point of view. Kleinbard (2010) considers the Nordic implementations as large scale experiments that provide evidence supporting the implementability of the DIT system also in US. He considers that especially due to the pressure to reduce the most important component of the capital income taxation, the corporate tax rate, the US has little choice but to embrace a DIT system. Norwegian DIT reform in 2006, which moved the tax burden away from normal returns and towards economic rents and risky returns to capital, enhanced a birth of a new branch for the DIT literature. Sørensen (2005a) describes in detail a design of Norwegian tax reform proposed to shareholder income tax. He shows that the proposed new system is neutral with respect to investment and financing decisions and decisions to realize capital gains. Also income-shifting problem of the dual income tax system can be solved with suitable tax rate choices in the new system. Sørensen continues to study the issue in another paper (Sørensen 2005b), where he provides an overview of the DIT and discusses alternative methods for taxing business income under the DIT. He concludes that for non-listed companies the imputed normal return should be taxed as capital income and above-normal returns should be taxed by labor-income tax. Compared to Norwegian shareholder income tax (in Sørensen 2005a), the proposed tax system requires less administrative capacity (as it only relates to non-listed companies) and offers greater flexibility in choosing tax rate structure. Lindhe, Södersten (2012) add to Sørensen 9

10 (2005a,b) by studying the proposed tax system in an open economy. They show that in an open economy some distortions still appear also in a new system. This paper contributes to the DIT literature in two ways. First, it shows with a simple framework that a DIT system combined with deductions provides non-trivial and sharp income-shifting incentives, which are likely to increase the costs of tax compliance and cause fluctuations between the state tax base and the tax bases of municipalities. Second, this paper shows that the DIT system makes poorer entrepreneurs to pay higher taxes than better off entrepreneurs in some cases. The deductions are shown to increase this difference. This paper also contributes to an increasing bunching literature (see Saez 2010) by providing evidence that Finnish entrepreneurs do bunch according to tax incentives provided by the tax system. 10

11 3 Entrepreneur Optimization Problem 3.1 Framework In this paper we concentrate on a simple framework, which is illustrated in figure 1. 3 An entrepreneur is assumed to own a company with a pre-determined pre-tax income of amount of Y. All of this company income is then distributed to an entrepreneur as wage and dividends. 4 An entrepreneur has one choice variable: the amount his company pays to him as wage income, w (0 w Y ). The rest of the pre-determined income, Y w, will be paid as dividends. The taxable wage income, which is taxed at rate τ W, is derived by subtracting a (single) deduction from the wage income. We assume there are no deductions from the dividend and no other capital income from other sources. Thus the paid dividend, Y w, is also the taxable capital (dividend) income, which is taxed at rate τ D. The objective of an entrepreneur is to maximize his net-of-tax income. 5 The result of an optimization of an entrepreneur in our framework depends 3 This framework captures some of the features in the Finnish DIT system and might therefore be considered as a very stylized version of it, which abstracts from most of its details. A more detailed structure of the taxation of an entrepreneur in a Finnish DIT system is illustrated in figure 1 in Ropponen (2012). It also reveals the complicatedness of the entrepreneur taxation. In this paper, the only margin an entrepreneur can react is the choice between the wage and dividend (the reporting of the choice of split of the income). In reality it is possible that the pre-tax income would depend on the whole tax system, but it is abstracted here. Also the corporate income tax is abstracted in the framework. 4 Thus there are no retained earnings in our framework. 5 With fixed Y the maximization of net-of-tax income is equivalent to minimizing the sum of taxes. 11

12 Y(income) w(wage) Y w(dividend) D(deduction) taxable wage taxable dividend t W t D tax on wage tax on dividend Figure 1: Framework for stylized model crucially on the tax rates (τ W and τ D ) as well as the exact shape of the deduction in the earned income tax schedule. We consider throughout the paper a single (EITC-like) deduction with its shape illustrated in figure 2. 6 The amount of the deduction in the stylized model is a function of wage income, D = D(w). First, with low wage incomes there is a phase-in region, where the deduction increases with a constant rate from zero to its maximum value D max as wage 6 Earned Income Tax Credit (EITC), which is in use in U.S. and includes phase-in, plateau and phase-out regions, serves as a known example of a credit of the similar type than the deduction employed in our framework. Similar deduction schemes are also in use in the Finnish tax system (see e.g. figure 10 about a deduction of income from work [työtulovähennys] in Ropponen 2012). 12

13 Deduction 0 D MAX plateau phase in phase out I II III IV a b c d Wage Figure 2: Deductions and marginal tax rates for wage income in stylized model changes from a 7 to b. Second, from wage income b to c, there is a plateau, where the deduction stays at its maximum value D max. Third, in the phaseout region from wage income c to d the deduction decreases from its maximum value to zero and stays zero thereafter. 3.2 Analysis This subsection provides a theoretical analysis of the optimal choice between wage and dividends in the stylized model described above. The following four subsections provide reasoning for the optimal wage schedule as a function of company income level, Y. The sections are divided according to income levels that correspond to those in figure 2 (I-IV). The analysis is conducted under 7 a is chosen to be 0. 13

14 the following assumption: Assumption 1 t D > t W Let us first consider the plausibility of assumption 1. Earned income, including wage income, is typically taxed more heavily than dividends paid out of a company. However, taking into account the corporate taxes that have already been paid from the profits might make dividends more expensive than wage for an entrepreneur owning a non-listed company of limited liability. As our stylized model includes only a single tax rate (τ D ) for dividends, it is considered to include both the corporate income taxation and the capital income taxation. In Finland an entrepreneur, who owns a non-listed company with limited liability, first pays corporate income tax (t corporate = 26% in year 2011; t corporate = 24.5% in year 2012) from the company profits. After the corporate tax, 70 percent 8 of (those) dividends that (do not exceed a certain net wealth related amount and) are paid out of the company are taxed according to capital income tax rate (a flat rate of t capital = 28% in 2011; a progressive capital income taxation in 2012 with rate t capital = 30% up until e and 32% for the capital income that exceeds 50000e). The corporate income tax and the capital income tax together sum up to single tax rates of t D = 40.50% (= t corporate + (1 t corporate ) 0.70 t capital = *0.7*0.28) in 2011 and t D = 40.36% ( *0.7*0.3) in 2012 (i.e. net-of-tax rates 8 As long as dividends were below both a net wealth related amount and a given euro amount, only 70 percent of dividends were taxable dividends in Finnish tax system. In general, dividends are split into earned income and capital income according to a net wealth of a company. In addition, capital income types are further split (and taxed differently) according to whether they exceed a certain euro level or not. 14

15 are and 59.64). 9 These average rates correspond to wage income of 6 500e/month, whereas the average wage income in Finland was in 2010 just over 3 000e/month. Therefore, the assumption that t D > t W seems plausible for many entrepreneurs in Finland Phase-in Region (I) Let us first consider a case, where a company income, Y, is between a and b. Thus also the wage income remains between a and b (a w b) and therefore in the phase-in region of the deduction (see figure 2). If all the income were paid as dividends, the taxes would be t D Y. The slope of the deduction in the phase-in region is D max. Therefore, a one unit b a increase of the wage income would both increase the wage tax by t W (1 D max b a ) and decrease dividend tax by t D. Because t D > t W, this increase in the wage income decreases the total tax-bill by t D t W (1 Dmax ) units. Because the b a phase-in-region is linear, each additional one unit wage increase provides the same decrease in the taxes as long as a Y b. Therefore, in the phase-in region it is optimal for the entrepreneur to pay all its income as wage income. If some units would have been paid as dividends, those units would be subject to tax rate t D instead of t W (with t W < t D ) and in addition part of the deduction would be lost, compared to case where no dividends are paid Plateau Region (II) Next we consider income levels, Y, which lie in between b and c. As shown above none of the wage levels less than b can be optimal (because then one 9 Here we have used 30% as capital income rate for year

16 would be able to reduce taxes by increasing the wage income up until b). If the wage income is increased from b by one unit, the deduction remains the same as long as the new wage level is at most c. A one unit increase would therefore imply a t W unit increase in taxes on wage income and a t D unit decrease in taxes on dividend income, resulting in a t D t W unit reduction (t D > t W ) in total taxes. Therefore, also in the plateau region it is optimal to pay all the income as wage (w OP T = Y, for all Y c) Phase-out Region (III) Let us next consider income levels c Y d. From the above we know that wages less than c cannot be optimal. Let us thus consider what happens to the total tax-bill when the wage is increased from c by one unit. Because the slope of the phase-out region is D max, increasing wage income from c by one d c unit increases taxable income by 1 + Dmax d c and therefore wage income tax by t W (1 + Dmax d c ). The decrease in the capital tax is t D. Now let us make the following assumption: Assumption 2 t D < t W (1 + D max d c ). Under this assumption 11 a wage income increase would increase the total taxes making wages above c but at most d, suboptimal. Therefore, under assumptions 1 and 2, w OP T = c for all c Y d This will also be observed in figure 3 in section 3.3 (and in figures 5 and 7 in section 4), where up until a company income level of c=35 000e all the income is paid as wage. 11 Combining assumptions 1 and 2 tells us that our results are valid for such dividend (plus corporate) tax rates, t D, that are greater than that for wage income, t W, but with some upper limit (which is determined via the slope of the phase-out region). 12 In figures 3, 5 and 7 this shows up as a flat part (no increase in the optimal wage 16

17 3.2.4 Beyond Phase-out Region (IV) Suppose next that Y > d. An entrepreneur can now choose wage level from phase-in, plateau and phase-out regions as well as beyond these regions. From the above we know that w opt = c for all 0 Y d. With wage levels d w Y increasing wage by one unit decreases tax-bill by t D t W units. 13 Thus an entrepreneur chooses w = Y rather than any other wage above d (d w < Y ). We also know from the above that w = c makes one better off than any w d. Thus the final comparison for income levels larger than d is between wages w = c and w = Y. The choice w = Y is better than w = c when total taxes from w = Y, T (Y ), is less than those from w = c, T (c), i.e. T (Y ) < T (c) t w Y [t w (c D max ) + t D (Y c)] < 0 t W D max + (t W t D )(Y c) < 0. Intuitively, the equation states that increasing wage from c to Y decreases dividend taxes by t D per unit yet increases wage taxes by t W per unit and in addition one loses the whole deduction. Solving for Y gives us the following condition for w = Y being optimal (i.e. better than w = c): Y > c + t W t D t W D max (1) Therefore, under our assumptions, with high enough total income Y, choosing w = Y becomes optimal (like in figures 3 and 5). With smaller Y, w = c is optimal. Thus for high enough income levels there occurs a discrete jump in the optimal wage schedule. It is worth noticing that without deduction the jump does not occur. schedule) from e to e. 13 For all wages w d the deduction is 0. 17

18 3.3 Entrepreneur Optimization Problem - Example Let us next illustrate the solution of the derived entrepreneur optimization problem with a numeric example. Let us choose the tax rates to be τ W = 25% and τ D = 29%. 14 Let the phase-in region of the deduction be from zero to e (b = 15000), the plateau from e to e (c = 35000) and the phase-out region from e to e (d = 50000). Wage PHASE IN a=0 PLATEAU b=15000 Stylized Model PHASE OUT w=c=35000 c=35000 JUMP (> 96000) Income Figure 3: Optimal wage schedule in a stylized framework 14 Note that as our model abstracts from corporate income tax, the capital income tax can be thought to illustrate both corporate and capital income tax. The exact numbers are, however, not that important in our stylized model, but rather the fact that there are parameter values, which imply some features into the tax system and that these features are also observed in the true Finnish DIT system (see section 4). 18

19 The optimal wage schedule for an entrepreneur with the chosen parameter values is given in figure 3. First note that with the above parameter choices both assumption 1 and 2 hold. Thus, according to sections and all the income is paid as wage both in the phase-in region and in the plateau region (up until income level of c = e in figure 3). In the phase-out region (section above) all the income is paid as dividends and thus the optimal wage schedule stays flat at this region. Then finally, with high enough income level (beyond the phase-out region, see section 3.2.4), it becomes optimal to pay all the income as wage. With the chosen parameters the income level is Y > c + t W t D t W D max = = The most dramatic income-shifting incentive in this illustration occurs at the income level of e. Here, a one unit increase in the income level changes the optimal wage level from e to close to e. This is a very large jump in the optimal wage schedule and is likely to have some drawbacks. We will come back to these drawbacks in the next section. 19

20 4 Optimization in Finnish Tax System The two following subsections provide illustrations of entrepreneur optimization problem in Finnish tax systems in years 2011 and The similarities of the optimal wage schedules between our stylized model and the true tax systems are discussed. 4.1 Entrepreneur optimization problem in 2011 Let us next compare the Finnish income tax system faced by an entrepreneur together with our stylized model. Finland has a DIT system where the earned income is taxed with the progressive tax schedule and the capital income is taxed with (about) flat rate. Figure 4 provides the marginal (solid line) and average (dashed line) tax rates for earned income and capital income tax in 2011 and 2012 and for our stylized framework. The progressiveness of the earned income tax schedule in the true tax system is easily observed in upper and middle left hand graphs. Except for two regions (around e and e) the marginal tax rates increase. 15 Our stylized framework 16 follows similar pattern for small wages, yet differs 15 All the decreases in the marginal tax rate are consequences of the end of some deduction. The decreases around e (upper left graph and middle left graph) follow from the end of the phase-out region of the basic allowance (perusvähennys). The decreases around e follow from the end of the phase-out regions in the deduction of earned income in municipality taxation (kunnallisveron ansiotulovähennys) and in the deduction of income from work (työtulovähennys). 16 First, in the phase-in region (from a to b) the deduction starting from zero increases with the same pace as wage. Thus all of wage income is deducted and the marginal tax rate for this region is zero. Second, in plateau region (from b to c) each additional unit of 20

21 for bigger wages. 17 The capital tax rate was 28 percent in year In 2012 it was 30 percent up until e capital income and 32 percent for the excess capital income. For our stylized model we choose a tax rate for earned income to be 25 percent and 29 percent for capital income. 18 The phase-in region for the deduction is chosen to be from zero to e (b = 15000), plateau from e to e (c = 35000) and phase-out region from e to e (d = 50000). Let us next turn to optimal wage schedule in the Finnish tax system. Figure 5 shows the optimal wage schedule of an entrepreneur in Finland in year 2011 (left graph) and in our stylized model (right graph). In more detail the left graph describes the optimal wage schedule of a single owner of a non-listed company with a given net wealth share of e and estimated entrepreneur income of e 19 in year 2011 in Finland (for details, also for the regimes wage income is taxed with a given tax rate τ W, which is thus also the marginal tax rate in this region. Third, in the phase-out region (from c to d) an additional unit of wage income increases taxable wage income by two units, because also the deduction is changed by one unit. Thus the marginal tax rate is equal to two times the earned income tax rate (2τ W ). From earned income level d on there are no deductions and thus the marginal tax rate is τ W from this income level on. 17 As will be seen later, despite the simplicity of our framework, it is still capable for providing similar features in optimal wage schedule than what is observed in the true tax system. 18 Note that as our model abstracts from corporate income tax, the capital income tax can be thought to illustrate both corporate and capital income tax. The exact numbers are, however, not that important in our stylized model, but rather the fact that there are parameter values, which imply problematic features into the tax system and that these features are also observed in the true Finnish DIT system. 19 The pension payments and social security payments depend on the estimated en- 21

22 Earned Income (euro) Capital Income (euro) Tax rate MTR 2011 ATR Tax rate Earned Income (euro) Capital Income (euro) Tax rate MTR 2012 ATR Tax rate MTR ATR Earned Income (euro) Capital Income (euro) Tax rate Stylized Model MTR ATR Tax rate Stylized Model Earned Income (euro) Capital Income (euro) Figure 4: Marginal and average tax rates for earned income and capital income in Finland in years 2011 and 2012 and in a stylized framework 22

23 A - E, see Ropponen 2012). The income is the gross income of a company that is a sum of pre-tax profits and gross wages. 20 The owner is assumed to distribute all the company gross income between wage and dividend. We show here only the optimal wage schedule as the corresponding optimal dividends appear as residuals. Wage Finnish Tax System 2011 E JUMP upper bound for wage C D optimal wage A B Wage PHASE IN PLATEAU PHASE OUT Stylized Model JUMP (> 96000) w=c=35000 a=0 b=15000 c= Income Income Figure 5: Optimal wage in Finnish 2011 tax system and in a stylized framework The most stunning thing in the optimal wage schedule in figure 5 appears at the income level of around e. Here we observe just like in our simple stylized model, a discrete jump where a one euro increase in the entrepreneur income implies a change in optimal wage of tens of thousands of euros. 21 Even trepreneur income. 20 Gross income is the amount an owner would receive if there were no taxes at all. 21 In the true tax system not all of the compensation is paid as wage after the jump, because there it is optimal to pay dividends up until 9 percent of the net wealth share of the company as these dividends are tax-free according to Finnish tax law. The entrepreneurs also must pay pension payments, which are determined according to the estimated yearly earnings. 23

24 if the stylized model is dramatically simpler than the true tax system, it still provides this same feature. The other common feature between the true tax system and the stylized model is that with small income levels it is optimal to pay only wage. This follows in the true system from both the deductions and the low tax rates for low wages. The observed sharp and complicated income-shifting incentives (a jump) that arise due to deductions (added into DIT system) 22 are likely to cause tax compliance costs. An entrepreneur either has to find out about tax rules by himself or pay to an accounting company to do the optimization. Otherwise he ends up paying higher taxes than optimal. Next we provide an example of the magnitude of the possible additional costs. In addition to compliance costs another potential cost of jump is that it makes an entrepreneur tax-bill more vulnerable for uncertainty. Suppose that in the beginning of the year an entrepreneur believes that his business will yield e that year, that is e per month. Suppose that (according to the optimal wage schedule in figure 3) he pays himself wage of e each month. Next suppose that his business faces a downturn starting from July and lasting until the end of the year. During this downturn the company does not make any profit and thus the business yields only e for the whole year. An entrepreneur has paid (until the end of June) all the e as wage. Because the deduction is zero for this income level, he pays e earned income tax (25%) from his wage income. If he would have known that his company income will be e, he would have optimally chosen to pay (see figure 3) e as wage income (15 000e tax-free and e with 22 Note that without deduction all the income would be paid as wage with every income level as τ W < τ D. 24

25 25 percent rate; i.e e) and the rest, e as dividends (29 percent of e is e). The total tax payments would in this case be e, which is 23.3% of e. In this example an entrepreneur total tax-bill is increased by about 7% (from 23.3% to 25%) due to uncertainty. Wage PHASE IN PLATEAU PHASE OUT Stylized Model JUMP (> 96000) MINIMUM w=c=35000 a=0 b=15000 c= Wage No Deduction MINIMUM Income Income Figure 6: Optimal wage schedules with and without deduction. The third drawback resulting from the jump arises if one cannot wait for the next year to receive dividends 23, but has to pay himself such a minimum amount of wage, which exceeds the optimal wage. Figure 6 shows the optimal wage schedules in the presence and in the absence of deduction in the stylized model. Let us consider a case where the income is units and one has to pay at least units of wage. With the deduction he cannot pay the optimal amount of wage (35 000e), whereas without the deduction ( e) he can. Thus the deduction makes this type of problem more pronounced and actually promotes a redistribution from worse-off entrepreneurs (who are more 23 Of those companies that paid dividends at all from their 2011 income, 80% of the payments were done on the following year, when the company income is known. 25

26 likely to face this restriction) to better-off entrepreneurs. We come back to this issue below, where we consider year 2012 tax system and this problem becomes even more pronounced. 4.2 Entrepreneur optimization problem in 2012 Let us next consider year 2012 Finnish tax system. Even if multiple small changes occurred in the Finnish tax system in the structure of the tax system and the levels of tax rates did not change much from year 2011 to 2012 (see figure 4). However, two changes are worth to mention. First, the corporate income tax rate was decreased from 26% to 24.5%. Second, the capital income tax rate of 28% was replaced by a two-step capital income taxation with tax rate of 30% up until capital income of e and 32% for the capital income exceeding e. In addition some of the deductions were changed and the inflation adjustments were made to the state income tax schedule. Due to all these changes the marginal tax rates differ in 2012 from those in year 2011 tax system, yet not very much. Figure 7 shows the optimal wage schedule for an entrepreneur facing a year 2012 Finnish tax system (left graph) and in our modified stylized model (right graph). The stylized model is otherwise the same as that for the year 2011 tax system, but modified by adding progressiveness into earned income tax schedule. More specifically the earned income tax rate is chosen to be 30% from income level e on. The graphs share some similar features, yet differ in some parts from those for year First, like in 2011 tax system, up until a certain level it is 24 For a detailed discussion see Ropponen

27 Wage Finnish Tax System 2012 upper bound for wage C D optimal wage A B E Wage PHASE IN PLATEAU PHASE OUT Stylized Model w=c=35000 a=0 b=15000 c= Income Income Figure 7: Optimal wage in Finnish 2012 tax system and in a stylized framework with progression optimal to pay only wage, because of the deductions. Second, unlike in 2011 tax system, no jumps are observed, but in contrast with high levels of income it is optimal to pay just little wage and most of the income is paid as dividends. This difference between the tax systems of the consequent years has some implications. First, the sharp change in the incentives from year 2011 to year 2012 tax system is likely to make the tax bases to fluctuate between state and municipalities. An entrepreneur with company income of e would optimally pay according to stylized model for year 2011 all its income as wage (see figure 5) resulting in a tax-bill of e ( e*0.25). According to year 2012 stylized model (see figure 7) his tax-bill would be e ((35 000e e)* e*0.29), which does not differ much (less than 2.3%) from the year 2011 tax-bill. However, the difference of the collected tax revenues may make a bigger difference. If an entrepreneur pays all the income as 27

28 wage, his municipal tax will be in the Finnish 2011 tax system about 87% of the state tax. 25 Next suppose that an entrepreneur pays e as wage and the rest of the income as dividends (like in modified stylized model for year 2012). The state income tax is now e, the municipality tax e and the corporate income tax e. 26 An approximation of the Finnish tax system where roughly 30% of the corporate tax goes for the municipalities and 70% for the state, leads to municipality tax being about 73% of the state tax. 27 Thus, the sharp income-shifting incentives combined with yearly changes in a tax system are likely to affect in non-negligible way to the relative levels of taxes collected between the state and the municipalities. Second, as also seen already above, if an entrepreneur would like to optimize but must pay himself some minimum amount of wage without having an opportunity to wait for the next year to receive dividends, he may end up 25 Because an entrepreneur must pay pension payments and other payments, the maximum amount of wage that can be paid is about e in 2011 tax system. The taxes and payments sum up to 48.1% of the company income with this choice. The state tax is now e and the municipality tax is e. Because all the income is paid as wage, no corporate income tax is paid. 26 In this case the taxes and payments sum up to 47.3% of the company income and an entrepreneur pays himself dividends of e. 27 The corporate income tax is divided in Finnish tax system between the state, the municipalities and the congregations according to a given formulary defining the relative shares. The division of the earned income tax revenues depends on the earned income levels for each taxpayer and thus also differs from the division of the corporate income tax revenues. Thus the choice between wage and dividends affects also tax revenues. Proportional share for the municipalities is 31.99% in 2011 and 28.34% in 2012 (see e.g. Government proposal; HE 180/2014). The corresponding numbers for the congregation are 2.55% and 2.30%. The state thus collects in total e and the municipalities e in this case. 28

29 receiving much higher tax-bill than with optimal wage-dividend choice. Suppose an entrepreneur must pay from a e company income (at least) e as wage in the modified stylized model. 28 His tax-bill is then e (50 000* * *0.29), whereas with optimal wage schedule (35 000e as wage and e as dividends) it would be e ((35 000e e)* e*0.29). Thus the restriction, which is more often faced by poorer entrepreneurs than better-off entrepreneurs yields about 10% higher tax-bill in this case. 28 Marginal tax rate for additional wage income is now 30% and for dividends it is 29%. Thus it is optimal to pay e as wage in this case. 29

30 5 Empirical Evidence The observed features of the DIT system might not be of great importance if the entrepreneurs did not follow the incentives provided by the system. To illustrate the reflection of the Finnish tax system on the behavior of the entrepreneurs let us consider an example of a particular dividend related incentive. 29 Finnish tax system makes dividend income in some cases very cheap compared to wage income. In years 2011 and 2012 the dividend income that is both below 9 percent of the net wealth of the company (share) and below a given euro amount (90 000e in 2011, e in 2012) is tax-free for an entrepreneur. 30 Let us illustrate this incentive by the administrative data. Figure 8 plots paid dividends as a function of entrepreneur share of the net wealth of the company in years 2011 and The concentration of dividends to this net wealth limit is clearly seen in the figure: Both graphs show a concentration of dividends on the amounts which equal to 9 percent of the net wealth share (solid line) of the company as well as on the euro limit, e in 2011 and e in Even if we observe a clear concentration 29 Here we do not take a stand on optimizing behavior, but concentrate on illustrating that entrepreneurs do take advantage of the cheap dividends provided by the tax system. 30 A corporate tax had to be paid from the profit before dividend payment, but no tax on paid dividends was levied on top of it. From 2014 on these dividends are no longer tax-free, but still face a lighter tax treatment than other dividends. 31 In more detail the figure shows the paid dividends for those main owners of a non-listed company who have paid dividends at all (those entrepreneurs who pay all the income as wage are not observed in the data). Määttänen and Ropponen (2014) show that with 3 percent accuracy about one half of the entrepreneurs choose their dividends to equal 9 percent of the net wealth in

31 Dividend Dividend Net Wealth Net Wealth Figure 8: Net wealth and dividend payments of non-listed companies to their main owners in 2011 and 2012 of the dividends, there are some entrepreneurs who choose their dividends differently. For them this choice may not be optimal or they just do not optimize (minimize their total tax-bill). Non-optimizing behavior might result in from multiple different reasons including not understanding the tax rules as well as choosing their compensation more in form of wage, because it provides access to social security, whereas dividend as a form of compensation does not. 32 Although not all entrepreneurs choose the dividends in the same way, figure 8 still provides evidence of entrepreneurs following the incentives provided by the tax system, especially when choosing between wage and dividends. The studies by Pirttilä and Selin (2011) and Harju and Matikka (2014) provide some further evidence on entrepreneurs following the tax incentives provided by the Finnish tax system. 32 The reasoning via access to social security was suggested in Fjærli and Lund (2001). 31

32 6 Conclusions The discussion about the benefits and drawbacks of the DIT systems, compared to the GIT systems has been going on already for a long time. I provide new aspects for this ongoing discussion. I show that deductions combined with a DIT system provides sharper and more complicated income-shifting incentives than when combined with a GIT system. This is likely to increase tax compliance costs. European Commission (2013) finds information as a key cost item in determining the tax compliance costs, suggesting that these costs are likely to fall especially on small businesses. It also finds that these costs are lower for simpler tax codes. Blaufus, Eichfelder and Hundsdorfer (2014) estimate the aggregate tax compliance costs of German income taxpayers and find their magnitudes to be of several percentages of the income tax revenue. For self-employed their estimates are even higher. Thus a movement towards simpler tax code might increase welfare by reducing the compliance costs especially for small business. In addition to compliance costs, the observed sharp income-shifting incentives between wage and dividend income, driven by deductions, may also have other negative effects. These may change dramatically the relative amount of taxes collected by state and municipalities. The abolishment of these incentives would reduce the fluctuations between tax bases. Another negative consequence of the sharp incentives is that even if deductions are typically designed to promote redistribution from better-off individuals to less well-off individuals, these may work the other way around for entrepreneurs in a DIT system. Compared to those entrepreneurs who have to pay wage for themselves each month, those who can choose wage-dividend combination only in the end 32

33 of the year may benefit from this choice more in the presence of deductions than when they are absent. Our study implies that the deductions not only erode the tax base and imply higher tax rates, but also make the incentives for income-shifting sharper than without any deductions in a DIT system. One way to reduce the observed negative consequences would be by making changes to deductions. One solution might be to remove the deductions. However, this would not as such solve the income-shifting problem completely. It might also turn out to be politically difficult. Another solution might be to remove the phase-out regions of the deductions, as these are the biggest reason for the observed income-shifting incentives in the first place. This would, however, probably be too costly. Third possible change for the deductions might be by making their phase-out region very gentle. This would also, however, increase costs, yet not as much as the removal of it. Tax-bill which was independent of the actions taken by an entrepreneur, would remove the incentive for income-shifting and would thus mitigate the observed weaknesses of the current Finnish DIT tax system. An attractive reform tackling the observed income-shifting incentives in the Finnish DIT system would be to follow steps towards a Norwegian DIT tax system. The new system might include an imputed normal return, which would be taxed as capital income. Above-normal returns would be then taxed by labor-income tax. As long as the total tax-bill of an entrepreneur is not be subject to his choice between different compensation forms, there is no incentive for incomeshifting. 33