Taxing Dividends in a Dual Income Tax System: The Nordic Experience with the Income Splitting Rules

Swedish tax law makes a distinction between “qualified” and “unqualified” shares in non-listed corporations. Dividends and capital gains from qualified shares are subject to the income splitting rules, whereas an unqualified shareholder may tax an unrestricted amount of dividends and capital gains as capital income. A share of a non-listed corporation is qualified if the following two conditions are met:

The corporation is a closely held corporation (CHC). In Sweden, a CHC is defined based on ownership concentration: A firm counts as a CHC if 4 or fewer owners control at least 50% of the shares. To prevent tax-motivated shifts of ownership shares across family members, the family network (defined in a broad sense to include relatives) count as one owner. Moreover, if several owners are active in the firm, these owners (and their relatives) also count as one owner (the extended CHC definition).

There is, however, an important exception to this rule. If passive owners, who are not taking an active part in profit generation, control more than 30% of the shares, the shares of the active owners always count as unqualified shares, and income splitting does not apply. Note that this holds true even when the two conditions specified above are met.

Following the terminology of Alstadsæter and Jacob (2012), I will use the term dividend allowance to denote the amount of dividends from corporation j that individual i is allowed to tax as capital income in fiscal year t. The dividend allowance can be written in the following stylized way for Sweden 1991–2005: 1DAijt=ρt×Eijt+(1+ρt)×𝟙DAijt−1>Dijt−1×[ DAijt−1−Dijt−1 ],{\rm{D}}{{\rm{A}}_{ijt}} = {\rho _t} \times {E_{ijt}} + \left( {1 + {\rho _t}} \right) \times {1_{{\rm{D}}{{\rm{A}}_{ijt - 1}} > {D_{ijt - 1}}}} \times \left[ {{\rm{D}}{{\rm{A}}_{ijt - 1}} - {D_{ijt - 1}}} \right], where E_ijt is the equity invested by owner i in corporation j (acquisition value of shares) at time t, ρ_t is the presumptive rate of return, and D_ijt–1 denotes dividends received from the corporation in the previous year. 𝟙_(.) is an indicator variable that takes on the value one if the saved dividend allowances are positive, and it is zero otherwise. Dividends in excess of the dividend allowance were added to labor income (progressive taxation).⁴

The intuition is straightforward: The allowance is given by a presumptive (imputed) return to the owner’s initial capital investment. This return is obtained by multiplying the acquisition value of shares with a presumptive rate of return, which is given by the risk-free rate of return (state lending rate) and a risk premium. If the CHC owner did not utilize the allowance a certain year, the unutilized allowance could be carried forward to the next year.

Capital gains from qualified shares, which were taxed upon realization, were also taxed as capital income up to the dividend allowance. Capital gains in excess of the dividend allowance were split equally into labor income and capital income.⁵ Hence, in practice the tax treatment of capital gains from CHCs was more lenient than for dividends.

The Swedish system underwent a profound reform in 2006. This reform has been covered empirically by Alstadsæter and Jacob (2016), Alstadsæter et al. (2017), and Jacob (2021). The reform strongly impacted on the way in which dividend allowances are calculated. It should be emphasized, however, that the abovementioned criteria for being covered by the splitting rules were retained. The government now introduced two options for calculating the year t dividend allowance. The CHC owner may either choose the simplification rule or the main rule. Using the above notation, we write 2DAijt={ DAijtSIMPLEorDAijtMAIND{A_{ijt}} = \left\{ {\matrix{ {DA_{ijt}^{SIMPLE}} \cr {or} \cr {DA_{ijt}^{MAIN}} \cr } } \right.

If the owner chooses the simplification rule the allowance is given by 3DAijtSIMPLE=sijt×qt+(1+ρt˜)×𝟙DAijt−1>Dijt−1×[ DAijt−1−Dijt−1 ],{\rm{DA}}_{ijt}^{SIMPLE} = {s_{ijt}} \times {q_t} + \left( {1 + \widetilde {{\rho _t}}} \right) \times {1_{{\rm{D}}{{\rm{A}}_{ijt - 1}} > {D_{ijt - 1}}}} \times \left[ {{\rm{D}}{{\rm{A}}_{ijt - 1}} - {D_{ijt - 1}}} \right], where sij_t is i’s ownership share in j, and q_t is a fixed amount that is constant across all corporations in year t regardless of the size of their equity. ρ˜t{{\tilde \rho }_t} is the interest rate used for accumulation of dividend allowances. Hence, when the CHC owner chooses this alternative, the allowance is simply a fixed amount (in proportion to the ownership share) plus accumulated unutilized allowances from earlier time periods.

If the owner chooses the main rule the allowance is given by 4DAijtMAIN=ρt×Eijt+𝟙Wijt−1≥Wjt−1×𝟙sijt≥s_t×sijt×G(Wjt−1)+(1+ρ˜t)×𝟙DAijt−1>Dijt−1×[ DAijt−1−Dijt−1 ]\matrix{ {{\rm{DA}}_{ijt}^{MAIN}} \hfill & { = {\rho _t} \times {E_{ijt}} + {_{{{\rm{W}}_{ijt - 1}} \ge {{\rm{W}}_{jt - 1}}}}} \hfill \cr {} \hfill & { \times {_{{{\rm{s}}_{ijt}} \ge {{}_t}}} \times {{\rm{s}}_{ijt}} \times {\rm{G}}\left( {{{\rm{W}}_{jt - 1}}} \right)} \hfill \cr {} \hfill & { + \left( {1 + {{\tilde \rho }_t}} \right) \times {_{D{A_{ijt - 1}} > {D_{ijt - 1}}}}} \hfill \cr {} \hfill & { \times \left[ {D{A_{ijt - 1}} - {D_{ijt - 1}}} \right]} \hfill \cr } where W_ijt–1 is the active owner’s own wage received from corporation j in the previous year, and W_jt–1 is the total wage bill (including the owner’s own wage) in the corporation in the previous year. Wages in subsidiaries are also included. G(W_jt–1) is the wage-based allowance explained below. 𝟙_{Wijt–1≥Wjt–1} is an indicator function that takes the value of 1 if the active owner’s wage exceeds a certain threshold, and it is zero otherwise. Likewise, 𝟙_sijt≥st takes on the value of 1 if the ownership share exceeds a certain threshold s_t.⁶ The owner’s wage requirement for eligibility for the wage-based allowance can be written 5W_jt=min[ α+βWjt,W_tmax ].{{}_{jt}} = \min \left[ {\alpha + \beta {W_{jt}},_t^{\max }} \right].

To better understand the nature of the current Swedish splitting rules, I will now be more specific on the parameter values. In Swedish tax legislation, these parameters are often formulated in income base amounts (IBB) in the previous year. In 2023, the IBB amounted to SEK 74,300. The fixed amount when opting for the simplification rule is 2.75 IBB or q = SEK 204,300 in 2024. This amount must be thought of as large given that the minimum capital requirement for corporations in Sweden is currently SEK 25,000.

The wage-based allowance is simply 50% of the firm’s wage bill, i.e. G(W_jt–1) = 0.5 × W_jt–1, up to a very high ceiling of 50 × W_ijt–1. The wage requirement is given by α = 6 IBB, β = 0.05, and W_tmax=9.6_t^{\max } = 9.6 IBB. Hence, the owner’s wage requirement for eligibility to the wage based allowance may not exceed SEK 713,000 in 2024. If the wage requirement is not met, the individual owner receives no wage based allowance at all. Even though the old capital-based allowance remains in Equation (4) it nowadays plays a modest role – the wage based allowance is considerably more important for those using the main rule.⁷ Admittedly, the Swedish “main rule” is complex. In June 2024, a new government committee on income splitting rules presented a proposal with the stated purpose of simplifying the rules (SOU 2024:36). The committee suggests, for example, that both the wage requirement and the capital requirement should be abolished. It remains to be seen whether these proposals will be enacted into law.

Before closing this section, I encourage the reader to reflect upon the similarities between the Swedish post-2006 system and a normal wage model. In the introductory section, we made a distinction between the normal wage and the normal return models, and we remarked that Sweden, Norway, and Finland all adopted a normal return model. However, the current Swedish system apparently shares crucial features with a normal wage model, because if the firm owner withdraws a pre-specified wage income, she becomes eligible for a considerably larger dividend allowance. If the allowance is sufficiently large, the owner may tax all residual income as capital income.

The Norwegian income splitting rules that were in place in the years following the 1992 dual income tax reform applied to active corporate owners. An owner was considered as being active if she worked more than 300 hours in the firm annually. However, the splitting rules did not apply if passive owners held at least two thirds of the shares. This rule closely resembles the Swedish 30% (“utomståenderegeln”). However, unlike Sweden, there were no further restrictions with respect to ownership concentration in Norway 1992–2005.

In a setting like this, there are salient incentives to invite passive owners, because by doing so the active owners circumvent the splitting rules and may tax all distributed corporate income as capital income. As mentioned above, in Sweden the rules for transferring shares to passive owners are restrictive: a large extended family network counts as one owner. On a similar note, in Norway spouses and nonadult children of active owners were not treated as separate passive owners. However, Thoresen and Alstadsæter (2010) report that a couple years after the 1992 reform, adult children were treated as passive owners. It thus seems like the Norwegian rules, at least in this respect, were more liberal than the Swedish ones.

An important feature of the Norwegian rules 1992–2005 was that corporate profits were divided into labor and capital income and taxed independently of the dividend payout decision. The government first calculated the presumptive return to capital as ρ_t ×A_jt, where A is total (gross) assets of the corporation. Residual (imputed) labor income of owner i is given by 6Zijt=sijt×[ Yjt−ω×(1+τP)Wjt−ρt×Ajt ],{Z_{ijt}} = {s_{ijt}} \times \left[ {{Y_{jt}} - \omega \times \left( {1 + {\tau _P}} \right){W_{jt}} - {\rho _t} \times {A_{jt}}} \right], where Y_jt is gross profits, ω is the share of wage costs that can be deducted (human capital deduction), and τ_p is the payroll tax.⁸ When Z_ijt < 0, it is carried forward against future positive imputed return to labor in corporation j. Capital gains (adjusted to account for retained profits in the firm) were taxed as capital income.

The Norwegian “shareholder income tax”, introduced in 2006, applies to shares from all corporations, regardless of ownership structure and activity status. It does not even differentiate between listed and non-listed shares. The theoretical foundations of the system have been laid out by Sørensen (2005), and the 2006 reform has been studied empirically by Alstadsæter and Fjærli (2009) and Alstadsæter et al. (2023). It should be emphasized that the Norwegian shareholder income tax is not literally an income splitting system, because no dividends or capital gains are taxed progressively with labor income. The basic principle behind the system is that the normal return to the share investment is exempted from owner-level taxes. However, as the “excess return” is taxed at a rate close to the top marginal tax rate on labor income, it is nevertheless related to an income splitting system.

The expression for the “dividend allowance” (rate of return allowance) corresponds to (1) for Sweden 1991–2005: 7DAijt=ρt×Eijt+(1+ρt)×𝟙DAijt−1>Dijt−1×[ DAijt−1−Dijt−1 ],{\rm{D}}{{\rm{A}}_{ijt}} = {\rho _t} \times {E_{ijt}} + \left( {1 + {\rho _t}} \right)\; \times \;{_{{\rm{D}}{{\rm{A}}_{ijt - 1}} > {D_{ijt - 1}}}}\; \times \;\left[ {{\rm{D}}{{\rm{A}}_{ijt - 1}} - {D_{ijt - 1}}} \right], where ρ_t now corresponds the risk-free rate of return (no risk premium).⁹ Similar to the old Swedish system, the allowance is given by the invested equity, and unutilized allowances can be carried forward.

While both Sweden and Norway made profound changes to their income splitting systems, the basic structure of the Finnish splitting system has been surprisingly stable since the advent of the dual tax system in Finland in 1993, an event that was studied by Pirttilä and Selin (2011). The Finnish income splitting system is very general: It applies to all non-listed corporations, regardless of ownership structure or activity status. However, in contrast to the current Norwegian system, shares in listed corporations are not part of the system.

The dividend allowance can be written 8DAijt=sijt×ρt×(Ajt−Bjt),{\rm{D}}{{\rm{A}}_{ijt}} = {s_{ijt}} \times {\rho _t} \times \left( {{A_{jt}} - {B_{jt}}} \right), where (A_jt – B_jt) refers to the book value of the corporation’s net assets. Note that, in Finland, unutilized allowances cannot be carried forward. However, the firm owner may of course expand future dividend allowances by retaining profits in the firm, thereby increasing equity. Capital gains are taxed as ordinary capital income and are not directly affected by the dividend allowance.

Behavioral responses to tax incentives within the Finnish system have been illuminated by e.g. Harju and Matikka (2016a), Harju and Matikka (2016b), and, most recently, Koivisto (2024). The incentives for entrepreneurship have been analyzed by Kanniainen et al. (2007).

Past and current splitting rules regimes in Sweden, Norway, and Finland are summarized in Table 1. The criteria for being covered by the rules were quite similar in Sweden and Norway up to 2006. After 2006, Sweden is the only country in which taxes on distributed profits are differentiated with respect to activity status and ownership concentration.

Turning to the base for the dividend allowance, Sweden 1991–2005 and Norway post-2006 both use injected equity, whereas imputed capital income depends on gross assets in Norway 1992–2005 and net assets in the entire period in Finland. In Sweden and Norway, but not in Finland, the presumptive rate of return is tied to the risk-free interest rate.

The incentives faced by those covered by an income splitting system do of course depend on the interaction between the parameters of the splitting system and the tax rates. Table 2 summarizes the relevant current tax rates in the three countries.

The differential between the combined dividend and corporate tax rate and the combined personal top marginal tax rate and the payroll tax rate is an important determinant of the incentive to shift labor income to capital income. In the current Swedish setting this differential is almost 30 percentage points (66-36.48=29.52) for dividends within the generously determined dividend allowance. In Finland, the same differential is sizeable as well – around 25 percentage points for dividends that are below EUR 150,000 and taxed within the dividend allowance. It is clear though that the level of the dividend tax within the allowance is significantly lower in Finland compared to Sweden.

In Norway, the tax rate differential is substantial within the dividend allowance. However, as the presumptive rate of return equals the risk-free rate, it is perhaps more relevant to compare the tax rate on excess dividends with the top marginal tax rate on labor incomes. Apparently, the top tax rate on labor income and the combined dividend and corporate tax rates are close, by construction.

A crucial feature of the Swedish tax rules, not captured by Table 2, is that dividends since 2006 are taxed differently depending on whether the shares are qualified, unqualified (see Section 2.1) or listed. As can be inferred from Table 2, dividends from qualified shares are taxed at 20% at the owner level. However, dividends and capital gains from unqualified shares, i.e. shares in non-listed firms not subject to income splitting, are taxed at a higher rate of 25%. Income from listed shares, finally, are taxed at 30%, which is the “standard” proportional tax rate on capital income in the Swedish dual income tax system. An interesting consequence of this tax differentiation is that it is not always beneficial to avoid the splitting rules.¹⁰ This is especially true for corporate owners who are entitled to large dividend allowances.

Sweden, Norway, and Finland introduced different normal return models in the early 1990s. Even more striking, however, is that the countries followed very different paths in the mid-2000s. Sweden introduced more generous rules in 2006, with more emphasis on the wage-based dividend allowance, while Norway in the same year moved to a general system that in some respects resembled the Swedish splitting rules of the 1990s. It is beyond the scope of this paper to assess the reasons for these reforms, but a detailed description of the Swedish policy context is found in Selin (2021).

The role of the splitting rules in the public debate seems to have been different in Norway and Sweden. In Sweden, the splitting rules or “the 3:12 rules” were very unpopular among business owners in the 1990s.¹¹ The rules were criticized for being too restrictive against successful entrepreneurship, see the discussion in Section 3.3 below. Another common complaint was that the rules were too complicated, imposing high compliance costs on small business owners. The latter line of criticism motivated the “simplification rule” discussed above. Sweden’s 2006 reform was implemented under a Social Democratic government, and the rules were subsequently made more lenient under a centerright government. The more generous Swedish wage sum rule in 2006 was motivated by the desire to promote entrepreneurship: The wage sum was seen as a proxy for risk-taking (Edin et al., 2005).

In Norway, on the other hand, the focus in the political discussion seems to have been on the taxplanning responses induced by the pre-2006 version of the splitting rules (Sørensen, 2005).

In short, the stated motive behind the Swedish reform was to promote entrepreneurship and reduce compliance costs, and the motive behind the Norwegian reform was to combat income shifting and to achieve a more neutral taxation of shareholder income (Sørensen, 2005).

Is income splitting needed in the first place? Or would it be better to tax capital income and labor income at the same rates? This is undoubtedly the most fundamental policy question related to income splitting. From a public finance perspective, the answer depends critically on the real elasticities of labor and capital incomes. If the two elasticities are equal, then from an efficiency point of view there is nothing to be gained from lower taxes on capital. It should be noted, however, that optimal tax theory considers taxes on real returns, and the inflation argument mentioned in the introduction may still motivate lower taxes on capital than on labor when nominal incomes are taxed. Optimal taxes also depend on redistributive concerns and how different sources of income are distributed in the population.

What do we know about the real elasticities of labor and capital supply? Several studies using Nordic data have found that the taxable labor income elasticity is around 0.2, implying a revenue-maximizing top marginal tax rate on labor above 60% (including both direct and indirect taxes).^12,13 The elasticity of capital supply is much harder to determine empirically, but a study by Jakobsen et al. (2019) using Danish data suggests that capital supply is more elastic than labor supply.

Even if the real elasticities between labor and capital income were different, it is far from certain that a DIT is desirable. If the shifting elasticity is sufficiently large, i.e. if people shift a lot of income when the tax differential increases, standard optimal tax theory still recommends equalizing the two tax rates (Saez and Stantcheva, 2018). The reason is that people otherwise waste resources on tax planning, and the government does not gain much revenue from the higher tax rate on labor income.

A difficult but important task in empirical tax research is to disentangle real response from shifting responses. Harju and Matikka (2016b) examined how owners of privately held corporations subject to the Finnish splitting rules reacted to a 2005 dividend tax reform. Before the reform there was no double taxation of dividends in Finland, but afterwards dividends were only tax exempt up to a monetary threshold of EUR 90,000. Owners with large dividend distributions therefore experienced a tax increase. Harju and Matikka(2016b) studied how the reform affected reported wage and dividend income of firm owners, while decomposing the response into a real response and a shifting response. They found that the shifting response accounted for two-thirds of the total response.

It is important to recognize, however, that the shifting response is under influence of the government. If income splitting rules are designed in such a way that it is easy to convert labor income into capital income, the shifting elasticity will be large, and vice versa. If the government chooses to tax labor and capital income differently, a key challenge is to design splitting rules that prevent socially wasteful tax avoidance activities.¹⁴

One policy alternative that might be considered is to tax different types of capital income at different rates, depending on how easy it is to shift labor income into a particular type of capital income. As we saw in Section 2.7, the Norwegian system after 2006 combines a dual income tax, i.e. progressive taxes on labor income and a low (22%) proportional tax on capital income, with a much higher owner-level tax on excess dividends, i.e. dividends that exceed the dividend allowance. As shown in Section 2.7 above, the combined corporate and dividend tax rate is roughly equal to the top marginal tax rate on labor income. At the same time, the top tax rate on labor income is much lower than in Sweden. Clearly, the Norwegian example shows that income splitting rules may imply high marginal tax rates on dividend distributions (especially if the dividend allowance is low). The distortion of dividend taxes at the owner level on investment and economic activity is a hotly debated issue. According to the so-called new view of dividend taxation, dividend taxes are not distortionary when marginal investment is financed by retained earnings. Recently, this view has received some empirical support from Yagan (2015) on a sample of large US firms. In small open economies, such as the Nordic ones, residence-based taxes at the owner level are expected to have less impact on domestic investment than source-based corporate taxes (Boadway and Bruce, 1992).

High dividend taxes could nevertheless affect the allocation of capital in the economy. It has been pointed out that dividends transferred within groups of companies are tax exempt in Norway (and in the European Union), and sæter et al. (2023) have documented an increase in holding companies, where a lot of wealth is accumulated without being distributed to the owners. The incentives to accumulate wealth in corporations under the Norwegian shareholder income tax have also been highlighted by Bjerksund and Schjelderup (2021a, 2021b). They claim it is profitable in Norway to save in shares through companies and to defer dividends.¹⁵ There are strong reasons to monitor such lock-in effects carefully in the future.

In the survey in the first part of this article, I deliberately focused on the coverage and eligibility criteria of the income splitting rules, as the Nordic countries differ considerably in this respect. In Sweden in 1991 and in the old Norwegian system, the ambition was to target the income splitting rules at the owners of firms who have the ability to convert labor income into dividends or capital gains, i.e. the owners of firms with concentrated ownership who also work in their own firms. The reasoning would be that income splitting rules should not affect the dividend policy of firms in contexts where the problem of shifting does not exist. The government’s problem, however, is to decide where to draw the line and to prevent owners from manipulating the characteristics that determine eligibility. Ownership structure can be adjusted if different rules apply to different groups of firm owners, and it is inherently difficult for the government to observe the owner’s work effort in the firm.

Even if it would be possible for the government to perfectly observe activity status, the challenge to combat manipulation of the ownership structure remains. This was indeed a big issue in the old Norwegian system, where passive owners were invited to circumvent the splitting rules (Thoresen and Alstadsæter, 2010). According to Sørensen (2005), the share of corporations in Norway subject to income splitting fell from 55% in 1992 to 32% in the year 2000. Although such loopholes can be closed, the task certainly comes with an administrative burden. This speaks in favor of more general rules, like those found in the stable Finnish system, or in the current Norwegian system.

When tax treatment differs depending on ownership structure, which is currently the case in Sweden, firm owners have incentives to spend resources on switching tax status. These resources have more productive uses in the economy.

In Sweden’s popular debate around the turn of the millennium, a common concern was that the tax rules of closely held corporations were too restrictive against successful entrepreneurs. A widespread view was that owners of rapidly growing firms were prohibited from withdrawing a fair amount of dividends as capital income. This critique is worth serious attention, because in the old Swedish income splitting system, and in the current Norwegian system, the dividend allowance is a function of invested equity and the interest rate used to accumulate unutilized dividend allowances. This implies that the base for the allowance is insensitive to firm performance at a given level of equity investment. This problem is not present in the same way in Finland, because firm growth is likely to materialize in a higher book value of net assets. And in Sweden post-2006, firm performance is accounted for through the wage sum, as successful firms tend to have more employees.¹⁶

Connecting the base of the dividend allowance to firm performance is likely, however, to induce distortions. Lindhe et al. (2004) compares the cost of capital of different organizational forms and financing in the early incarnations of the splitting rules in Sweden, Norway, and Finland. They found that the cost of capital for investments financed by retained earnings in Finland is too low compared with equity finance. Finnish corporate owners have incentives to retain profits in the firm and invest the funds both in financial and physical capital.

Turning to the Swedish system and the wage based allowance, the concern is the opposite, namely that firms’ factor utilization is distorted in favor of labor, see the discussion in Sørensen (2010, Section 7.4). The more employees an entrepreneur hires, the more lenient the tax treatment of dividends and capital gains.

I emphasized in the introduction that the alternative to the normal return model is the normal wage model. Under the latter regime, the government imputes a wage income, while residual income distributed from the firm is taxed as capital income. In Section 2.2, we concluded that the current Swedish system is a hybrid between a normal return and normal wage model. A pure version of such a system was, however, implemented in Iceland in 1997 in order to prevent income shifting from progressively taxed labor income to proportionally taxed dividend income in the Icelandic version of the DIT (Matheson and Kollbeins, 2012). In Iceland, the earnings of the business owner are imputed by using industry specific minimum wages, with annual adjustments to account for inflation and productivity growth. Since 2010, there is an additional restriction that 50% of dividend payouts that exceed 20% of corporate net assets are taxed as labor income.

The normal wage model is also in place in the Netherlands. The Dutch income tax system differs from the Nordic DITs insofar that presumptive capital returns and not actual returns to a larger extent are taxed, see, e.g. Cnossen and Sørensen (2021) and Cnossen and Bovenberg (2001). However, similar to the Nordic countries, the separation of progressive labor income taxation and proportional capital income taxes motivate special income splitting rules for business owners. In the Netherlands, the owner’s wage requirement is the highest of the following amounts: (i) 75% of the wage from the most comparable employment, (ii) the wage of the company’s highest earning employee, and (iii) EUR 48,000 in 2022. There is also a share requirement (c.f. Sweden post-2006). The owner is required to own at least 5% of the shares in the corporation to be allowed to tax distributed income as capital income. Bettendorf et al. (2017) have studied these rules empirically.

A downside of the normal wage model is that productivities are genuinely difficult to impute. Moreover, the term “wage model” is slightly misleading, because the tax system is a function of annual earnings; therefore, the government also needs to have an idea on how to impute work hours. The government may either use very standardized amounts (like in the Netherlands or Sweden) to impute earnings, or industry-specific amounts (like in Iceland). In both cases, imputations may substantially deviate from the “true” earnings, and there are potentially large administrative costs involved in running such a system effectively. Moreover, there are reasons to believe that true earnings are considerably more heterogenous than the marginal rate of return requirement used to impute the presumptive rate of return in the Finnish system.

An interesting property of the normal wage model is that labor income in excess of the imputed wage on the margin is taxed as capital income. Consequently, the marginal incentive to supply labor is given by the lower capital income tax rate. Selin and Simula (2020) consider an optimal tax model in which agents differ with respect to productivity, labor supply elasticity, and fixed cost of switching from being a regular employee to become an “income shifter”. If people who are more elastic in their labor supply also have lower costs of income shifting, it may actually be optimal for the government to allow for some income shifting, even when abstracting from standard arguments in favor of taxing capital income more leniently than labor income.¹⁷