Estimates of average marginal tax rates on factor incomes in Japan

J. Japanese Int. Economies 25 (2011) 81–106

Contents lists available at ScienceDirect

Journal of The Japanese and International Economies journal homepage: www.elsevier.com/locate/jjie

Estimates of average marginal tax rates on factor incomes in Japan Hiroshi Gunji a, Kenji Miyazaki b,⇑ a b

Faculty of Economics, Daito Bunka University, Japan Faculty of Economics, Hosei University, Japan

a r t i c l e

i n f o

Article history: Received 10 November 2009 Revised 8 December 2010 Available online 18 February 2011 JEL classification: E20 E62 H20 O5

a b s t r a c t Gunji, Hiroshi, and Miyazaki, Kenji—Estimates of average marginal tax rates on factor incomes in Japan In this paper, we estimate average marginal tax rates on factor incomes in Japan from 1963 to 2007. We adapt the method of D.H. Joines [Estimates of effective marginal tax rates on factor incomes. J. Bus. 54 (2), 191–226.] to the Japanese tax and social security system. Average marginal tax rates on labor incomes without social security premiums range from 14% to 21%, whereas the rates on incomes with social security have increased from 21% to 33%. Tax rates on capital incomes have fluctuated between 35% and 58%. We also compare our estimates with average tax rates and the wedges from business cycle accounting. J. Japanese Int. Economies 25 (2) (2011) 81–106. Faculty of Economics, Daito Bunka University, Japan; Faculty of Economics, Hosei University, Japan. Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction Tax is often introduced into economic models to increase their realism, and sometimes to evaluate quantitatively welfare levels and policy effects. For accurate evaluation, estimating effective tax rates in the macroeconomy is crucial. The average marginal tax rate, which is a weighted average of the marginal tax rates of economic agents with different incomes, is more appropriate as an effective

⇑ Corresponding author. Address: Faculty of Economics, Hosei University, 4342 Aihara, Machida, Tokyo, Japan. Fax: +81 42 783 2611. E-mail address: [email protected] (K. Miyazaki). 0889-1583/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.jjie.2011.02.003

82

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

tax rate for macroeconomic analysis than is the average tax rate, which is simply the ratio of total tax revenues to national income. In this paper, we estimate Japanese average marginal tax rates. Many researchers have estimated average marginal tax rates for the US. Joines (1981) and McGrattan et al. (1997), updating the Joines series, used the amounts of income and tax revenue for each income bracket to estimate a series of tax rates on labor and capital incomes. Making fewer assumptions, Seater (1985) and Stephenson (1998), updating the Seater series, adopted the same method to calculate a series of tax rates on total income. Barro and Sahasakul (1983, 1986) used the statutory rate to compute a series of tax rates on total income. Akhand and Liu (2002) used a nonparametric approach to estimate a series of average marginal rates on total income. Following Joines (1981), most research into average marginal taxes has attempted to relax these assumptions, but has computed tax rates only on total income. In many studies in which dynamic macroeconomic models have been calibrated, the Joines (or its updated) series has been used so that the effects of taxes on each factor income can be evaluated separately.1 To similarly assist macroeconomists investigating the Japanese economy, in this paper we compute average marginal tax rates on capital and labor incomes by using the methodology of Joines. To determine the factors that were detrimental to the Japanese economy after 1990, many researchers have simulated Japanese business cycles in the 1990s by using neoclassical macroeconomic models that incorporate income tax. Hayashi and Prescott (2002) used a constant capital income tax rate of 0.48. Braun et al. (2006) set the labor income tax rate to 0.24, and Esteban-Pretel et al. (2010) set the labor and capital tax rates to 0.28 and 0.44, respectively. In these studies, average tax rates were used as marginal tax rates; it is important to estimate Japanese marginal tax rates accurately. To our knowledge, the only estimated average marginal tax rates on factor incomes for Japan are those obtained by McKee et al. (1986).2 The paucity of studies may be a product of the Japanese tax system. Many OECD countries adopt a self-assessment income tax system under which, even though wages and salaries are taxed at source, employees usually file a final tax return to make a year-end tax adjustment. In Japan, however, most employees have no such incentive because employers are obliged to make year-end tax adjustments for their employees. This makes it difficult to determine average marginal tax rates for all taxpayers. To be specific, we divided Japanese workers into three categories: employees not filing a final tax return; employees filing a final tax return; and self-employed workers filing a final tax return. The first two are withholding income taxpayers, and the last two are self-assessment income taxpayers.3 For the US, where most are self-assessment income taxpayers, the Statistics of Income published by the Internal Revenue Service can provide data on almost all taxpayers’ tax revenues for each income bracket. In Japan, the Sample Survey for Self-assessment Income (Shinkoku Shotoku Zei Hyohon Chosa) and the Statistical Survey of Actual Status for Salary in the Private Sector (Minkan Kyuyo Jittai Tokei Chosa), published by the National Tax Agency, present tax data for each income bracket. Thus, with some assumptions, one can construct series of average marginal tax rates for both self-assessment income taxpayers and withholding income taxpayers. The challenge is to estimate average marginal tax rates for all taxpayers by combining these series. In Japan, workers filing a final tax return are both self-assessment and withholding income taxpayers, and they are included in both surveys. To overcome this difficulty in estimating average marginal tax rates, we use a weight to estimate total average marginal tax rates. The weight is chosen so that the average tax rate on total income is equal to a weighted sum of average tax rates on self-assessment income and average tax rates on withholding income. By using this weight, we treat the weighted sum of the average marginal tax rates for self-assessment taxpayers and those for withholding income as the total average marginal tax rates. Furthermore, one can broadly consider social security premiums as a component of taxes on labor income. We calculate average marginal social security premium rates and add them to the average marginal tax rates on labor incomes. 1 For example, McGrattan (1994), McGrattan et al. (1997), Cole and Ohanian (1999), Chari et al. (2000), Siu (2008), and McGrattan and Ohanian (2010) used the Joines series. 2 McKee et al. calculated tax rates for 1979, 1981, and 1983 only, without using time series data. 3 Retirees, whose main incomes are no longer salaries and wages, are included in both categories of taxpayer.

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

83

Our results can be summarized as follows. Over the sample period, average marginal tax rates without social security premiums ranged from 14% to 21% and have decreased since the early 1990s. Average marginal tax rates including social security payments ranged from 21% to 33%. Average marginal tax rates for social security have increased from 5% to 15%. Average marginal tax rates on capital incomes ranged from 35% to 58%; they rose by 7% and 10% points in 1973 and 1974, respectively, and fell by around 5% in both 2002 and 2003. In addition, we make two comparisons. We first compare our results with the series of Japanese average tax rates obtained by Mendoza et al. (1994). Their average rates on labor incomes are similar to, and since the late 1980s, slightly higher than our average marginal tax rates; the tax rates on capital incomes obtained by Mendoza et al. are slightly lower than ours. Second, we compare our results with the Japanese labor and capital wedges from business cycle accounting (BCA) obtained by Kobayashi and Inaba (2006). Although our average marginal taxes on labor incomes account for 70% of labor wedges in terms of their level, our estimated tax rates on capital incomes cannot explain fluctuations in the capital wedge. This paper is organized as follows. In Section 2 (resp. 3), we compute average marginal tax rates on labor and capital incomes excluding (resp. including) social security premium rates. In Section 4, we compare our estimated tax rates with the average tax rates computed by Mendoza et al. (1994). In Section 5, we compare our rates with the labor and capital wedges from BCA obtained by Kobayashi and Inaba (2006). In Section 6, we offer concluding remarks. 2. Average marginal tax rates excluding social security premium rates In this section, we calculate average marginal tax rates without including social security premium rates. The average marginal tax rates on labor and capital incomes are denoted by MTRL and MTRK, respectively. Economic agents comprise two types of taxpayers: self-assessment taxpayers and withholding taxpayers. We calculate the average marginal tax rates for both types of taxpayers. By combining these figures with an appropriate weight, we estimate the total average marginal tax rates. In Section 2.1, we explain the calculation of average marginal tax rates for self-assessment income taxpayers. In Section 2.2, we explain the calculation of the average marginal tax rates for withholding income taxpayers. In Section 2.3, we report the average marginal tax rates at the macroeconomic level. Before describing our procedure, we comment on our sample period. Our estimated average marginal tax rates on factor incomes in Japan cover the period from 1963 to 2007. The main reason for using this sample period is that the Sample Survey for Self-assessment Income and the Statistical Survey of Actual Status for Salary in the Private Sector are available from 1963. Note also that we use two types of Systems of National Accounts (SNA). One is based on the 1968 System of National Accounts (68SNA), which spans from 1955 to 1998; the other, based on 93SNA, which has a base year of 2002, allows us to use data from 1980 to the present. Our series from 1963 to 1979 are based on 68SNA, while the series from 1980 to 2007 are based on 93SNA.4 2.1. Average marginal tax rates for self-assessment income taxpayers In this subsection, we consider the average marginal tax rates for self-assessment income taxpayers. We calculate these by using the methodology of Joines (1981).5 For computing average marginal tax rates for self-assessment taxpayers, our main data source is the Sample Survey for Self-assessment Income Tax produced by the National Tax Agency. This survey provides data on several types of income for each income bracket. We classify these incomes into labor and capital incomes, and then estimate the average tax rates for each income bracket. For other taxes, only total revenues are available. Each tax item is classified as a proportional tax on either capital income or total income. Adding the proportional 4

The 68SNA-based series from 1980 to 1998 are available from the authors on request. Having relaxed assumptions, Akhand and Liu (2002) proposed a nonparametric estimation. We do not use their method for two reasons. First, whereas Akhand and Liu estimated tax rates on total income, we estimate tax rates on individual factor incomes. Second, we do not have a large number of observations. Although it is possible theoretically to use their nonparametric method for our sample, it is not feasible in practice. In order to utilize their method, we need a rich dataset for individual taxpayers. 5

84

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

taxes to the average marginal tax rates for the self-assessment incomes yields the average marginal tax rates for self-assessment income taxpayers. We assume that taxpayers are homogeneous for each income bracket. The total income of group i (i = 1, . . . , N) is denoted yi = yli + yki, where yli and yki represent labor and capital incomes, respectively. Following Joines (1981), the amount of tax revenues of group i, ts(yi) is:

~i Þ; t s ðyi Þ ¼ syi þ sk yki þ f ðy where s denotes the proportional tax rate on total income, sk is the proportional tax rate on capital income, f() represents the progressive tax function, y~i ¼ ci yi denotes the income that is progressively taxed, and ci is the fraction of taxable income of group i. In this subsection, there are no proportional taxes on labor incomes. Joines (1981) considered two progressive tax functions for labor and capital incomes and assumed the fraction of taxable income to be constant for each income. By contrast, we consider one progressive tax function, but assume that the fraction of taxable income depends on i. The marginal tax rates of group i on labor and capital incomes are: s

dt ðyi Þ=dyli ¼ s þ ci fi0 ; s

dt ðyi Þ=dyki ¼ s þ sk þ ci fi0 ; where fi0 represents the progressive tax rates schedule. Each marginal rate of income tax is divided into proportional and nonproportional rates. We aggregate the marginal rates across groups to calculate the average marginal tax rates on labor and capital incomes (that is, MTRLs and MTRKs, respectively). Letting total tax revenues be P P T s ¼ Ni¼1 t s ðyi Þ, letting total labor income be Y l ¼ Ni¼1 yli , and letting total capital income be PN Y k ¼ i¼1 yki , we obtain: s

MTRLs ¼

s

s

MTRK s ¼

s

N N X dT dt ðyi Þ X dt ðyi Þ dyli ¼ ¼ ; dY l dyli dY l dY l i¼1 i¼1 s

s

N N X dT dt ðyi Þ X dt ðyi Þ dyki ¼ ¼ : dY k dyki dY k dY k i¼1 i¼1

Following Joines (1981), we assume that dyli/dYl = yli/Yl and dyki/dYk = yki/Yk. The assumption simplifies the above equations to:

MTRLs ¼ s þ

N X

ci wli fi0 ;

ð1Þ

i¼1

MTRK s ¼ s þ sk þ

N X

ci wki fi0 ;

ð2Þ

i¼1

where wli = yli/Yl and wki = yki/Yk. MTRLs and MTRKs are weighted averages of the marginal tax rates on labor and capital incomes for group i, with the weights wli and wki, respectively, representing the shares of labor and capital incomes that are subject to nonproportional taxes. We now investigate how the available Japanese data can be used to calculate s, sk, ci, fi0 , wli, and wki in (1) and (2). We then report our results for the average marginal tax rates of self-assessment taxpayers. 2.1.1. Estimation of s The proportional tax rate on total income s is:

s¼

amounts of proportional tax on total incomes : amounts of total incomes

To compute the denominator, one can use nominal national product (NNP) or national income at market prices from the Annual Reports on the National Accounts (Kokumin Keizai Keisan Nenpo), which is produced by the Economic and Social Research Institute, the Cabinet Office. An alternative to NNP is

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

85

national income (NI) at factor cost. However, as Joines (1981) explains, NI excludes net indirect tax in taxable income that, in theory, allows tax rates to exceed 100%. Hence, we follow Joines and use NNP for total income. The numerator, the total amount of proportional tax, is:

total national tax revenues  income tax  corporation tax  land tax  securities transaction tax þ local proportional tax: Except for data on the local proportional tax in the above equation, we can obtain all of the required data from Chapter 1 (Overview) of the National Tax Agency Annual Statistics Reports (Kokuzeicho Tokei Nenposho). We subtract self-assessment and withholding income taxes and national proportional tax on capital income from the total amount of national tax revenue, and add local proportional tax to obtain the total amount of proportional tax. Land taxes, introduced in 1996, have been suspended since 1998. Securities transaction taxes were abolished in 1999. Local proportional taxes are computed by subtracting local proportional taxes on capital income from total local taxes (that is, prefectural and municipal taxes). The local proportional tax on capital incomes is:

ðbusiness tax þ prefectural fixed asset tax þ real property acquisition taxÞ þ ðmunicipal fixed asset tax þ mine production tax þ special landholding taxÞ þ ðenterprise tax þ central planning tax þ water utility and land profit taxÞ: The first and second lines, respectively, represent ordinary prefectural and municipal taxes, and the last line represents special purpose municipal taxes. Data on these items can be obtained from the Annual Statistical Reports on Local Government Finance (Chiho Zaisei Tokei Nenpo) and the White Paper on Local Government Finance (Chiho Zaisei Hakusho) produced by the Ministry of Internal Affairs and Communications. Special landholding tax and enterprise tax were introduced in 1973 and 1975, respectively. 2.1.2. Estimation of sk Similarly to s, the proportional tax rates on capital incomes sk are computed from the following:

sk ¼

amounts of proportional tax on capital incomes : amounts of capital incomes

We interpret the denominator as hNNP, where 1  h is labor’s share in NNP, and is given by:

ð1  hÞNI ¼ compensation of employees þ ð1  hÞunincorporated enterprises income: The incomes of unincorporated enterprises are the sum of their operating surpluses and net receivable incomes. Net indirect taxes (=NNP  NI) and the total income of unincorporated enterprises are intermingled and difficult to divide into distinct capital and labor incomes. Therefore, we allocate them to labor and capital incomes based on the shares of labor and capital in total income. The numerator, the amount of proportional tax on capital incomes, is the amount subtracted from total tax revenues excluding income tax when calculating s. That is:

corporation tax þ land tax þ securities transaction tax þ local proportional tax on capital income: The source for all data (except for data on the local proportional tax on capital income) is Chapter 1 (Overview) of the National Tax Agency Annual Statistics Reports. The components of the local proportional tax on capital income are described in Section 2.1.1. 2.1.3. Estimation of ci The ratio of taxable income to total income for each income bracket i, ci, is estimated for each income group i from Section 2-5 (Results of Sample Survey for Self-assessment Income Tax, excerpt) of the National Tax Agency Annual Statistics Reports:

86

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

ci ¼

amounts of taxable income : total income

Note that, because of data availability, Joines (1981) assumes c to take the same value for all i. 2.1.4. Estimation of fi0 The derivative of the progressive tax function for each income bracket i, fi0 , is calculated from the following formula: f10 ¼ r1 =y1 and

fi0 ¼

ðr i =ni Þ  ðr i1 =ni1 Þ ~i =ni Þ  ðy ~i1 =ni1 Þ ðy

ði ¼ 2; . . . ; NÞ;

~i is the amount of taxable where ri is the amount of income tax, ni is the number of taxpayers, and y income for each income bracket i. For i = 1, we use r0 = y0 = 0, so f10 ¼ r 1 =y1 . These figures are taken from Section 2-5 of the National Tax Agency Annual Statistics Reports. To compute ri, we use the sum of withholding and self-assessment income taxes in Table 1 of Section 2-5.6 2.1.5. Estimation of wli and wki To estimate the distribution of labor and capital incomes subject to the nonproportional taxes, wli and wki, we must determine whether each type of assessment income is either labor or capital income. Having divided all incomes into three (labor income yli, capital income yki, and miscellaneous income ymi), Joines (1981) considered two cases: one in which ymi belongs to labor income and the other in which it belongs to capital income.7 Our allocation of ymi based on labor and capital income shares is novel. Specifically, the reported income items in Table (3) of Section 2-5 (Results of the Sample Survey for Self-assessment Income Tax, excerpt) of the National Tax Agency Annual Statistics Reports are classified into the following three types of income: 1. Labor income (yli): employment income and retirement income; 2. Capital income (yki):8 interest income, dividend income, real estate income, comprehensive capital gains, short-term separate capital gains, long-term separate capital gains, capital gains from stocks and other assets; 3. Miscellaneous income (ymi): business income, farm income, miscellaneous income, timber income, and occasional income. Capital gains on stocks, etc. have been taxed since 1989. Our classification closely follows that of Joines (1981). Having discussed the inclusion of capital gains in capital incomes, Joines (1981) computed marginal tax rates based on their inclusion and exclusion. However, in most studies based on the Joines data, tax rates are based on the inclusion of capital gains. We consider only the case in which capital gains are included in capital incomes.9 We further divide the case in which capital gains are included into three subcases: in the first, miscellaneous incomes are allocated to labor income; in the second, they are allocated to capital income; in the third case, they are allocated to both incomes based on factor shares. Joines (1981) used only the first two subcases. Because the first two are extreme cases, we next explain the third case. 6 The amount of tax minus the amount of tax credit should be theoretically equal to the amount of withholding income tax plus the amount of self-assessment income taxes in the table, although there exists a subtle statistical discrepancy. 7 In most studies based on the Joines data, miscellaneous income is classified as labor income. 8 The Japanese government adopts the source separation taxation system, under which self-assessment taxpayers can choose to include some of the listed capital incomes in their total income or have them taxed separately and proportionally. Most taxpayers are supposed to choose the latter; therefore, it seems appropriate to treat taxation on such capital incomes as proportional. However, some taxpayers include their capital incomes in their total income. Because the Sample Survey for Self-assessment Income does not provide data based on total income excluding capital incomes and because our formula can incorporate proportional taxation when fi0 is constant, we treat such capital incomes as income components that are progressively taxed. 9 Our estimated tax rates based on excluding capital gains are available from the authors on request.

87

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106 Table 1 Average marginal tax rates of self-assessment income taxpayers. Year

MTRLs0

MTRK s0

MTRLs1

MTRK s1

MTRLsh

MTRK sh

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

0.2363 0.2380 0.2401 0.2302 0.2321 0.2407 0.2051 0.2044 0.1848 0.2119 0.2124 0.2238 0.2140 0.2372 0.2373 0.2463 0.2594

0.5481 0.5530 0.5369 0.4903 0.5323 0.5019 0.4491 0.4668 0.4560 0.5036 0.5390 0.6880 0.5918 0.6081 0.6185 0.6591 0.6782

0.2876 0.2832 0.2843 0.2720 0.2755 0.2932 0.2361 0.2326 0.2057 0.2355 0.2341 0.2429 0.2325 0.2491 0.2496 0.2546 0.2670

0.4688 0.4785 0.4691 0.4236 0.4291 0.4268 0.4066 0.4272 0.4317 0.4647 0.5144 0.6328 0.5484 0.5578 0.5669 0.6026 0.6186

0.2438 0.2446 0.2458 0.2365 0.2388 0.2489 0.2103 0.2092 0.1878 0.2154 0.2155 0.2260 0.2159 0.2385 0.2387 0.2474 0.2603

0.5074 0.5161 0.5040 0.4552 0.4707 0.4574 0.4270 0.4474 0.4465 0.4877 0.5300 0.6669 0.5778 0.5886 0.5988 0.6359 0.6550

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0.2573 0.2633 0.2757 0.2658 0.2582 0.2559 0.2705 0.2589 0.2446 0.2521 0.2590 0.2578 0.2765 0.2777 0.2344 0.2363 0.2320 0.2471 0.2624 0.2357 0.2386 0.2377 0.2393 0.2342 0.2256 0.2322 0.2463 0.2572

0.6345 0.6599 0.6744 0.6748 0.6904 0.6761 0.7135 0.7310 0.7162 0.6935 0.6475 0.6430 0.6689 0.6787 0.6629 0.7040 0.6805 0.7104 0.7171 0.6890 0.6447 0.6661 0.6384 0.6011 0.5891 0.6225 0.6304 0.6282

0.2683 0.2792 0.2877 0.2835 0.2767 0.2752 0.2967 0.2782 0.2717 0.2827 0.2886 0.2897 0.3148 0.3110 0.2641 0.2642 0.2581 0.2741 0.2840 0.2608 0.2642 0.2596 0.2637 0.2574 0.2504 0.2581 0.2748 0.2890

0.5756 0.6003 0.6242 0.6127 0.6250 0.6100 0.6456 0.6750 0.6591 0.6496 0.6095 0.5964 0.6043 0.6216 0.6007 0.6480 0.6300 0.6528 0.6694 0.6416 0.5980 0.6205 0.5953 0.5608 0.5459 0.5838 0.5919 0.5865

0.2590 0.2656 0.2775 0.2685 0.2610 0.2590 0.2747 0.2621 0.2489 0.2570 0.2640 0.2631 0.2820 0.2821 0.2380 0.2395 0.2353 0.2502 0.2648 0.2383 0.2416 0.2402 0.2421 0.2371 0.2290 0.2355 0.2503 0.2619

0.6087 0.6359 0.6552 0.6501 0.6639 0.6487 0.6864 0.7096 0.6928 0.6767 0.6330 0.6258 0.6431 0.6584 0.6414 0.6856 0.6634 0.6914 0.7030 0.6750 0.6294 0.6520 0.6250 0.5878 0.5739 0.6094 0.6170 0.6133

Note: The series from 1963 to 1979 are based on 68SNA and the series from 1980 to 2007 are based on 93SNA.

P The total labor income of self-assessment income taxpayers is denoted by Y sl ¼ Ni¼1 yli . The total PN s amount of capital incomes is denoted by Y k ¼ i¼1 yki . The total amount of miscellaneous incomes P is denoted by Y sm ¼ Ni¼1 ymi . We suppose that the income shares of each income bracket are the same as the macroeconomic income shares. By using the labor share 1  h, calculated in Section 2.1.2, we assume that miscellaneous incomes of (1  h)ymi and hymi are attributed to labor and capital incomes, respectively. Then, the distributions of labor and capital incomes subject to nonproportional taxes are:

wli ¼

yli þ ð1  hÞymi Y sli þ ð1  hÞY smi

and wki ¼

yki þ hymi : Y ski þ hY smi

When h = 0, all miscellaneous incomes are treated as labor incomes, and when h = 1, all are treated as capital incomes.

88

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

2.1.6. Estimation of the average marginal tax rates of self-assessment income taxpayers The above computations can be used to calculate the average marginal tax rates of self-assessment income taxpayers on labor incomes (MTRLs) and capital incomes (MTRKs). As already explained, we consider three cases relating to the treatment of miscellaneous incomes. The average marginal tax rates computed are defined as follows: 1. MTRLs0 and MTRK s0 : based on miscellaneous incomes being allocated to labor incomes; 2. MTRLs1 and MTRK s1 : based on miscellaneous incomes being allocated to capital incomes; 3. MTRLsh and MTRK sh : based on miscellaneous incomes being allocated to both incomes based on factor shares. The above three types of average marginal tax rates are summarized in Table 1. The plots of MTRLsh and MTRK sh are presented in Fig. 1, in which tax rates for withholding income taxpayers and a weighted average of rates for both taxpayers are also plotted. The average marginal tax rate on labor incomes declined until 1971, then rose gradually until 1992, and has since fluctuated. The average marginal tax rate on capital income peaked at around 70% in 1987. Since then, it has declined gradually. We discuss these results further after calculating total average marginal tax rates in Section 2.3. 2.2. Average marginal tax rates for withholding income taxpayers In this subsection, we calculate the average marginal tax rates for withholding income taxpayers. Under the withholding income tax system, labor and capital incomes are taxed at source. As explained

Fig. 1. Average marginal tax rates without social security premiums. Note: (i) (a) MTRLh and (b) MTRKh are the total average marginal tax rates on labor and capital incomes, respectively. The series from 1963 to 1979 are based on 68SNA and the series from 1980 to 2007 are based on 93SNA. (ii) All series are based on miscellaneous incomes being allocated to both capital and labor incomes according to their factor shares.

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

89

in the introduction, because employers should make year-end tax adjustments for their employees, most employees working for only one institution do not have to file a final tax return. Taxpayers whose labor incomes are from several sources are supposed to collect their incomes and file a final tax declaration. Under the progressive tax system, some of these taxpayers must make additional tax payments, which they have incentives to avoid. Because the Japanese tax system does not give citizens ID numbers, it cannot perfectly monitor such tax avoidance and, hence, cannot guarantee collecting all taxes due. For the data on withholding income, we used the Statistical Survey of Actual Status for Salary in the Private Sector, published by the National Tax Agency. However, this survey provides less detailed information than does the Sample Survey for Self-assessment Income used for self-assessment income taxpayers. The survey for self-assessment income taxpayers stratifies the data on taxpayers into income ranges based on total income, and the associated tables report income from various sources and total tax, as well as the number of taxpayers. By contrast, the survey for withholding income taxpayers reports only employment incomes for each stratum. Furthermore, because this survey provides such detailed information only for employees working in the private sector throughout the year, one cannot obtain sufficient data either for public sector workers or for private sector workers who only work for part of the year. Because of such data limitations, we make the following three assumptions. 1. We treat withholding taxes on incomes except wages and salaries as proportional taxes. 2. The income distribution of employees working throughout the year represents that for withholding taxpayers. 3. We ignore withholding taxpayers who do not earn wages and salaries. Although these assumptions only arise because of data limitations, they are justified on the following grounds. For the first assumption, because of the source separation tax system, capital incomes such as dividends, interest, and capital gains are currently taxed proportionally.10 The second assumption indicates that the income distribution for public servants and temporary workers is the same as that for private sector wage earners who work throughout the year. The bias induced by making this assumption might be at least partially offset by the fact that, in Japan, public servants earn relatively high incomes because their salaries are based on those of large-scale private corporations, and temporary workers earn relatively low incomes.11 The final assumption implies that withholding taxpayers are almost invariably employees. Unfortunately, we do not know the number of withholding taxpayers who do not earn wages and salaries. However, it is reasonable to ignore such taxpayers. By construction, as shown later, the derivative of the progressive tax function for the first income bracket is the average tax, which is simply the ratio of the amount of taxes to the amount of income, and this is invariant to the number of employees with zero income (who pay no tax). Therefore, treating withholding taxpayers who do not earn wages and salaries as employees with zero income does not affect the average marginal tax rate. Note that the Statistical Survey of Actual Status for Salary in the Private Sector collects data not from employees but from companies. This indicates that the survey counts employees who earn their salaries from more than one company at least twice. However, this double counting does not affect our calculation of average marginal tax rates. This is because, under the Japanese withholding income tax system, the average marginal tax rates for two employees whose incomes (YA and YB) are from one 10 As discussed in footnote 8, the Sample Survey for Self-assessment Income uses income bracket data based on total income including capital incomes, whereas the Statistical Survey of Actual Status for Salary in the Private Sector uses income bracket data based only on employment incomes rather than on total income. 11 Furthermore, a justification for the second assumption could be based on the number of workers. According to a recent Labor Force Survey, the number of government workers is around 2.3 million, which amounts to less than 5% of the total labor force. On the other hand, a recent Statistical Survey of Actual Status for Salary in the Private Sector shows that there are about six times as many private sector employees working throughout the year as there are working for part of the year. Because the survey collects data from each company, such temporary workers include those who have other jobs, are switching careers, or are quitting a job. Therefore, the actual number of temporary workers is even smaller. These arguments suggest that the vast majority of private sector employees work in the sector throughout the year are thus representative of all employees.

90

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

company are observationally equivalent to the rates for one employee whose incomes (Y1 and Y2) are from two companies, as long as either YA = Y1 and YB = Y2 or YA = Y2 and YB = Y1. Although most employees do not file a final tax return under the Japanese tax system, employees who earn their employment incomes from more than one company or who receive additional incomes are supposed to file a final tax return by law.12 Thus, some withholding income taxpayers are also selfassessment income taxpayers. However, because taxpayers do not have ID numbers, some might decide not to file an income tax return to avoid additional tax demands. Therefore, we cannot know the exact number of withholding income taxpayers or the exact number who file a final tax return. This issue is discussed in Section 2.3, in which we calculate overall average marginal tax rates on labor and capital incomes. Below, we explain the computation of average marginal tax rates for withholding income taxpayers. As are self-assessment income taxpayers, withholding income taxpayers are assumed to be homogeneous for each income bracket. The tax revenue of group i, tw(yi), is: w t w ðyi Þ ¼ syi þ sk yki þ sw k yki þ sl yl1i þ gðyl2i Þ;

where s and sk are as defined in the previous subsection, sw k is the additional rate of proportional taxation on the capital incomes of withholding taxpayers, sw l is the rate of proportional taxation on labor incomes except employment incomes, yl1i is labor income except employment income, yl2i is employment income (that is, yli = yl1i + yl2i), and g() represents the progressive tax function for employment income. Unlike self-assessment income tax, withholding income tax is assumed to be w represented by: sw k yki þ sl yl1i þ gðyl2i Þ. As already explained, we assume withholding taxes on incomes except wages and salaries to be proportional. The difference in expressions between self-assessment and withholding taxes arises solely because of data availability. The marginal tax rates of group i are: w

dt ðyi Þ=dyli ¼ s þ sw l

dyl1i dy þ g 0i l2i ; dyli dyli

w

dt ðyi Þ=dyki ¼ s þ sk þ sw k: We assume that dyl1i/dyli = yl1i/yli and dyk1i/dyki = yk1i/yki. Therefore, similarly to the previous subP P section, letting total tax revenues be T w ¼ Ni¼1 t w ðyi Þ, letting total labor incomes be Y l ¼ Ni¼1 yli , and PN letting total capital be Y k ¼ i¼1 yki , we obtain:

~w MTRLw ¼ s þ s l þ

N X

~ li g 0i ; w

i¼1 w k;

MTRK w ¼ s þ sk þ s where:

s~wl ¼ swl

PN

i¼1 yl1i

Yl

~ li ¼ and w

yl2i : Yl

In what follows, we explain how to use the available data to calculate average marginal tax rates. Because s and sk are as defined in the previous subsection, we explain how to use the available data to w 0 ~w ~ construct s l , sk , g i , and wli . Then, we report our calculated average marginal tax rates for withholding income taxpayers. w ~w 2.2.1. Estimation of s l and sk w ~w To estimate s l and sk , we need data on the amount of taxes for each factor income. As in the previous subsection, miscellaneous incomes are difficult to classify as either capital or labor incomes, and we consider three such cases.

12

To be precise, separate withholding tax systems allow taxpayers earning capital incomes to make separate tax payments.

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

91

For withholding income taxpayers, all types of income, except for employment income ðY w l2 Þ, can be classified into one of the following three categories:   1. Labor income, except for employment income Y w l1 : retirement income;  w 2. Capital income Y k : interest income, dividend income, and capital gains on listed stocks;   3. Miscellaneous income Y w m : remuneration, fees, and the incomes of nonresidents. w w Let the corresponding total tax revenues be denoted by T w l1 ; T k , and T m . These figures, as well as those for the total amount of employment income, Y w , are available from Section 3-1 (Statistics of Taxl2 ation) of the National Tax Agency Annual Statistics Reports. Whereas capital incomes are taxed proportionally, retirement incomes are not. Although taxable retirement incomes after deductions are taxed progressively under the Japanese taxation system, because of limited data availability, we assume that retirement incomes are taxed proportionally.13 w ~w For calculating s l and sk , we consider three cases: the case in which miscellaneous incomes are treated as labor incomes; the case in which miscellaneous incomes are treated as capital incomes; and the case in which these incomes are treated as either labor or capital incomes in proportion to the corresponding factor shares. Because the first two are extreme cases, we explain the computation w ~w of s l and sk in the third case. By using labor’s share 1  h, we divide miscellaneous incomes into labor w w ~w incomes ð1  hÞY w m and capital incomes hY m , respectively. Therefore, sl and sk can be estimated from:

s~wl ¼

w Tw l1 þ ð1  hÞT m w w Yw þ Y þ ð1  hÞY l1 l2 m

and

swk ¼

w Tw k þ hT m w : Yw þ hY k m

When h = 0, all miscellaneous incomes are treated as labor incomes, and when h = 1, all miscellaneous incomes are treated as capital incomes. 2.2.2. Estimation of g 0i The derivative of the progressive tax function for each income bracket i, g 0i , is calculated from g 01 ¼ r 1 =yl21 and

g 0i ¼

ðri =ni Þ  ðr i1 =ni1 Þ ðyl2i =ni Þ  ðyl2i1 =ni1 Þ

ði ¼ 2; . . . ; NÞ;

where ri is the amount of withholding tax on employment income, ni is the number of taxpayers, and yl2i is the level of employment income for income group i. These figures are available from Table 6 (Breakdown of the number of employment income earners, total amount of salary and amount of tax by range of salary, employment income earners who worked through a year) of Section 3-2 (The Results of the Statistical Survey of Actual Status for Salary in the Private Sector, excerpt) in the National Tax Agency Annual Statistics Reports. Given the method of constructing g 0i , the inclusion of employees with zero income does not affect the average marginal rate. To show this, we create a new income group i = 0 in which r0 = yl20 = 0 and g 00 ¼ 0. For any n0 > 0, g 01 ¼ ðr 1 =n1  r 0 =n0 Þ=ðyl21 =n1  yl20 =n0 Þ ¼ g 01 ¼ r 1 =yl21 , which implies that the derivative of the progressive tax function for the first income bracket is the average tax rate. Because this is simply the ratio of the amount of taxes to the amount of income, it is independent of the num~ l0 (described later) is always zero, ber of zero taxpayers. Furthermore, because yl20 = 0, the weight w and thus has no effect on the average marginal rate. This provides a justification for our third assumption (described previously) that withholding taxpayers who do not earn wages and salaries can safely be ignored. ~ li 2.2.3. Estimation of w ~ li , we must make To compute the distribution of labor incomes subject to nonproportional taxes, w some adjustments. The total amount of employment income from Table 6 of Section 3-2 in the 13 Because retirement incomes account for no more than 2–5% of total labor incomes, this assumption only affects our results in a minor way.

92

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

P National Tax Agency Annual Statistics Reports, Ni¼1 yl2i , is not the same as that from Table 1 of Section w 3-1 (Statistics of Taxation), Y l2 . The former is based on the incomes of those who work throughout the year in the private sector, whereas the latter includes public servants and temporary workers. In this context (as already explained), we assume that public servants and temporary workers have the same income distribution as do those working throughout the year in the private sector. Thus, when miscellaneous incomes are allocated to both labor and capital incomes in proportion to the corresponding factor shares, the weight becomes:

~ li ¼ w

Yw l1

þ

Yw l2

yl2i Yw l2 : w  PN þ ð1  hÞY m i¼1 y2li

w w Note that Y w l1 ; Y l2 , and Y m are as defined previously. Data on these items can be obtained from Table 1 of Section 3-1 in the National Tax Agency Annual Statistics Reports. Data on yl2i comes from Table 6 of Section 3-2. When h = 0 (resp. h = 1), all miscellaneous incomes are allocated to labor incomes (resp. capital incomes).

2.2.4. Estimation of the average marginal tax rates of withholding income taxpayers Using the methods described above, we calculate the average marginal tax rates on labor incomes (MTRLw) and capital incomes (MTRKw) of withholding income taxpayers. As in the previous section, in calculating average marginal tax rates, we consider the following three treatments of miscellaneous incomes: w 1. MTRLw 0 and MTRK 0 : miscellaneous incomes are treated as labor incomes; w 2. MTRL1 and MTRK w 1 : miscellaneous incomes are treated as capital incomes; w 3. MTRLw h and MTRK h : miscellaneous incomes are allocated to labor and capital incomes in proportion to the corresponding factor shares.

The above three types of average marginal tax rates are presented in Table 2. Compared with those reported in Table 1, the average marginal tax rates on both factor incomes for withholding income taxpayers are always below those of self-assessment income taxpayers. Like MTRLsh and MTRK sh ; MTRLw h and MTRK w h are plotted in Fig. 1. Fig. 1 indicates that average marginal tax rates for withholding income taxpayers are much lower than those for self-assessment income taxpayers. As is shown later, w because MTRLw h and MTRK h account for a major part of total average marginal tax rates, the behavior of withholding income tax rates is similar to that of total average marginal tax rates. We explain the latter series in the next subsection. 2.3. Estimation of total average marginal tax rates We divided taxpayers into two types, and computed the marginal tax rates on both factor incomes. Based on these computations, we estimate total average marginal tax rates at the macroeconomic level. As already discussed, some withholding taxpayers make year-end tax adjustments. Thus, the withholding income taxpayers who make declarations are included in both the Sample Survey for Self-assessment Income Tax and the Statistical Survey of Actual Status for Salary in the Private Sector. Ideally, one would divide taxpayers into those paying only self-assessment income tax, those paying only withholding income tax, and those paying both types of tax. In addition, one should take account of the numbers of the three types of taxpayers and of the income levels and taxes in all income brackets. Then, to compute average marginal tax rates, one should use as weights the proportions of the three types of taxpayers. However, because data limitations prevent application of these procedures, all we can do is to compute the average marginal tax rates of self-assessment income taxpayers and withholding income taxpayers and then combine them into a weighted average of the tax rates. To do so, we use an alternative weight, denoted by a. When miscellaneous incomes are allocated to labor and capital incomes according to factor shares, the total average marginal tax rates on factor incomes can be computed from:

93

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106 Table 2 Average marginal tax rates of withholding income taxpayers. Year

MTRLw 0

MTRK w 0

MTRLw 1

MTRK w 1

MTRLw h

MTRK w h

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

0.1623 0.1680 0.1599 0.1495 0.1457 0.1497 0.1486 0.1468 0.1466 0.1544 0.1628 0.1433 0.1312 0.1409 0.1418 0.1499 0.1612

0.3542 0.3606 0.3724 0.3362 0.3411 0.3548 0.3558 0.3708 0.3987 0.3960 0.4588 0.5628 0.4697 0.4576 0.4593 0.4898 0.4883

0.1628 0.1691 0.1605 0.1500 0.1457 0.1501 0.1488 0.1469 0.1464 0.1546 0.1636 0.1433 0.1310 0.1412 0.1419 0.1503 0.1620

0.3632 0.3698 0.3763 0.3370 0.3400 0.3498 0.3509 0.3668 0.3940 0.3913 0.4513 0.5541 0.4628 0.4517 0.4558 0.4842 0.4838

0.1624 0.1683 0.1600 0.1496 0.1457 0.1499 0.1487 0.1468 0.1465 0.1545 0.1630 0.1433 0.1312 0.1410 0.1418 0.1500 0.1613

0.3571 0.3635 0.3735 0.3365 0.3407 0.3530 0.3539 0.3693 0.3971 0.3944 0.4564 0.5605 0.4680 0.4561 0.4584 0.4882 0.4871

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0.1648 0.1727 0.1774 0.1790 0.1794 0.1816 0.1887 0.1912 0.1807 0.1786 0.1866 0.1925 0.1961 0.1946 0.1852 0.1725 0.1715 0.1806 0.1775 0.1715 0.1710 0.1702 0.1650 0.1617 0.1638 0.1713 0.1821 0.1745

0.4508 0.4802 0.4952 0.4851 0.4984 0.4812 0.5092 0.5369 0.5392 0.5439 0.5010 0.4807 0.4988 0.5080 0.5178 0.5724 0.5733 0.5694 0.5677 0.5750 0.5557 0.5721 0.5367 0.4932 0.4827 0.5364 0.5334 0.5245

0.1659 0.1741 0.1787 0.1803 0.1806 0.1821 0.1893 0.1919 0.1898 0.1872 0.1959 0.2018 0.2061 0.2051 0.1955 0.1808 0.1795 0.1888 0.1858 0.1777 0.1769 0.1763 0.1717 0.1707 0.1739 0.1795 0.1923 0.1839

0.4464 0.4757 0.4910 0.4832 0.4985 0.4847 0.5127 0.5409 0.5270 0.5275 0.4874 0.4689 0.4854 0.4934 0.4984 0.5508 0.5358 0.5428 0.5451 0.5521 0.5245 0.5452 0.5055 0.4630 0.4520 0.5036 0.5023 0.4863

0.1650 0.1730 0.1776 0.1793 0.1797 0.1817 0.1888 0.1914 0.1829 0.1806 0.1888 0.1947 0.1981 0.1965 0.1870 0.1739 0.1729 0.1820 0.1787 0.1724 0.1720 0.1711 0.1660 0.1632 0.1656 0.1727 0.1839 0.1763

0.4494 0.4790 0.4941 0.4846 0.4984 0.4822 0.5102 0.5381 0.5345 0.5378 0.4962 0.4768 0.4946 0.5037 0.5120 0.5659 0.5589 0.5597 0.5600 0.5676 0.5461 0.5646 0.5256 0.4803 0.4689 0.5237 0.5205 0.5078

Note: The series from 1963 to 1979 are based on 68SNA and the series from 1980 to 2007 are based on 93SNA.

MTRLh ¼ aMTRLsh þ ð1  aÞMTRLw h; MTRK h ¼ aMTRK sh þ ð1  aÞMTRK w h: When h = 0 (h = 1), all miscellaneous incomes are treated as labor (capital) incomes. The weight a is chosen so that the average tax rate on total income is equal to a weighted average of the average tax rate on self-assessment income and that on withholding income. That is:

total income tax self-assessment income tax withholding income tax ¼a þ ð1  aÞ : total income self-assessment income withholding income The denominator of the first term on the right-hand side is computed from the total amount of income reported in Table 1 of Section 2-5 in the National Tax Agency Annual Statistics Reports. The

94

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

numerator of the first term is obtained from the sum of self-assessment and withholding income taxes in the same table. The denominator of the second term is calculated from the sum of the payment amounts for each income type in Tables 4 to 9 of Section 3-1. The numerator is taken from the total amount of withholding income taxes in Table 1 of the same section. To compute the left-hand side, we assume that the average tax rate for all income taxpayers is the same as that for the withholding income taxpayers who file a final tax return. Given this assumption, the ratio of the sum of the denominators on the right-hand side to the sum of the numerators is the average tax rate for all taxpayers.14 The weight a for each fiscal year is presented in the second column of Table 3. The weight has declined from 20% to reach its current level of around 10%. Recent values are 2–3% above the share of the self-employed in the total labor force.15 Based on the weight a for each fiscal year, we compute weighted averages of the average marginal tax rates of self-assessment and withholding income taxpayers to obtain the total average marginal tax rates on labor and capital incomes. The average marginal tax rates are summarized in Table 3. Fig. 1 illustrates MTRLh and MTRKh for self-assessment income taxpayers, withholding income taxpayers, and a weighted average of both taxpayers. More weight is assigned to withholding taxpayers than to self-assessment income taxpayers. Therefore, the total average marginal tax rates are similar to, and slightly above, those for withholding income taxpayers. The average marginal tax rate on labor income increased until the late 1980s and then fluctuated around 20%. In more detail, from its peak in 1992, the rate generally fell to reach its lowest level in 2003. This decline is arguably the result of Japan’s lengthy economic slump (the so-called ‘‘lost decade’’) and of tax-cutting reforms designed to stimulate the Japanese economy. Since 2004, along with a rebound in stock prices, economic recovery has started, which has increased the average marginal tax rate in the progressive taxation system. Note that these tax rates do not incorporate social security premiums. As discussed in the introduction, social security contributions can be broadly considered as taxes on labor incomes. In the next section, we calculate average marginal tax rates that incorporate social security premium rates. The average marginal tax rate on capital incomes is at least 30% higher than the average marginal tax rate on labor incomes, mainly because of corporation tax. Tax rates on capital incomes increased from 36% to 58% in 1987 and then fluctuated around 55%. The rates rapidly increased by seven percentage points in 1973 and by 10 percentage points in 1974 because of the 1973 oil crisis,16 the 1974 revision of the statutory rate of corporation taxation, and the Act on Special Measures Concerning Taxation (introducing gasoline tax, motor vehicle tax, and local road tax). In the following year, the rates dropped, possibly because of economic depression and as firms adjusted to new tax systems. In 2002 and 2003, the rates fell by five percentage points, perhaps because of the reformed tax system.

3. Average marginal tax rates including social security premium rates As noted in Section 1, social security premiums can be broadly interpreted as taxes. In this section, we estimate the average marginal tax rates including social security premiums, denoted by MMTRL. The five types of social security payments considered are pension insurance, health insurance, employment insurance, accident compensation insurance, and long-term care insurance. In the following subsections, we calculate the average marginal premium rates for each security payment. In the final subsection, we add up those rates to obtain total average marginal rates of social security 14

In fact,

a þ b þ c ða þ cÞ þ ðb þ cÞ ¼ dþeþf ðd þ f Þ þ ðe þ f Þ ,when c/f = (a + b + c)/(d + e + f), where a, b, and c (resp. d, e, and f) are the amounts of taxes (resp. incomes) for those paying only self-assessment income tax, those paying only withholding income tax, and those paying both types of tax. 15 The figures on the number of self-employed workers and the total labor force obtained from labor force data are fiscal year averages from the Labor Force Survey (Rodoryoku Chosa) published by the Ministry of Internal Affairs and Communications. 16 The 1973 oil crisis triggered unexpected inflation, a rise in nominal interests, and thus an increase in nominal capital income and associated taxes.

95

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106 Table 3 Total average marginal tax rates. Year

a

MTRL0

MTRK0

MTRL1

MTRK1

MTRLh

MTRKh

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

0.1611 0.1614 0.1511 0.1526 0.1541 0.1554 0.1742 0.1773 0.1846 0.1819 0.2085 0.1309 0.1269 0.1120 0.1150 0.1283 0.1329

0.1742 0.1793 0.1720 0.1618 0.1590 0.1639 0.1585 0.1570 0.1536 0.1649 0.1731 0.1538 0.1417 0.1517 0.1528 0.1623 0.1742

0.3854 0.3916 0.3972 0.3597 0.3706 0.3777 0.3721 0.3878 0.4093 0.4155 0.4755 0.5792 0.4852 0.4744 0.4776 0.5115 0.5136

0.1829 0.1875 0.1792 0.1686 0.1657 0.1724 0.1640 0.1621 0.1573 0.1693 0.1783 0.1563 0.1439 0.1533 0.1543 0.1637 0.1759

0.3802 0.3873 0.3903 0.3502 0.3537 0.3618 0.3606 0.3775 0.4010 0.4047 0.4645 0.5644 0.4736 0.4636 0.4686 0.4994 0.5017

0.1755 0.1806 0.1730 0.1629 0.1600 0.1652 0.1594 0.1579 0.1541 0.1656 0.1739 0.1541 0.1419 0.1519 0.1529 0.1625 0.1745

0.3813 0.3881 0.3932 0.3546 0.3607 0.3692 0.3666 0.3831 0.4062 0.4114 0.4717 0.5744 0.4819 0.4709 0.4746 0.5072 0.5094

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0.1294 0.1227 0.1252 0.1255 0.1233 0.1236 0.1269 0.1358 0.1365 0.1429 0.1440 0.1388 0.1173 0.1164 0.1083 0.1111 0.1208 0.1107 0.1078 0.1064 0.1039 0.0977 0.1012 0.1043 0.1081 0.1131 0.1111 0.1054

0.1767 0.1838 0.1897 0.1899 0.1891 0.1908 0.1991 0.2004 0.1895 0.1891 0.1970 0.2016 0.2055 0.2043 0.1905 0.1796 0.1788 0.1880 0.1866 0.1783 0.1780 0.1768 0.1725 0.1693 0.1705 0.1782 0.1892 0.1832

0.4746 0.5023 0.5177 0.5089 0.5221 0.5053 0.5351 0.5633 0.5633 0.5653 0.5221 0.5032 0.5188 0.5279 0.5335 0.5870 0.5863 0.5850 0.5838 0.5871 0.5649 0.5813 0.5470 0.5045 0.4942 0.5462 0.5442 0.5354

0.1791 0.1870 0.1924 0.1933 0.1924 0.1936 0.2030 0.2036 0.2010 0.2009 0.2093 0.2140 0.2188 0.2174 0.2029 0.1900 0.1890 0.1983 0.1964 0.1865 0.1859 0.1844 0.1810 0.1797 0.1822 0.1884 0.2015 0.1950

0.4631 0.4910 0.5077 0.4995 0.5141 0.5002 0.5295 0.5591 0.5451 0.5449 0.5049 0.4866 0.4994 0.5084 0.5095 0.5616 0.5472 0.5550 0.5585 0.5616 0.5321 0.5526 0.5146 0.4733 0.4621 0.5127 0.5123 0.4969

0.1772 0.1844 0.1901 0.1905 0.1897 0.1913 0.1997 0.2010 0.1919 0.1915 0.1997 0.2042 0.2080 0.2065 0.1925 0.1812 0.1805 0.1895 0.1880 0.1794 0.1792 0.1779 0.1737 0.1709 0.1725 0.1799 0.1913 0.1854

0.4700 0.4982 0.5142 0.5053 0.5189 0.5028 0.5325 0.5614 0.5561 0.5577 0.5159 0.4974 0.5120 0.5217 0.5260 0.5792 0.5715 0.5743 0.5754 0.5791 0.5547 0.5731 0.5356 0.4916 0.4803 0.5334 0.5312 0.5189

Note: The series from 1963 to 1979 are based on 68SNA and the series from 1980 to 2007 are based on 93SNA.

premiums, MSST. We then estimate MMTRL by adding MSST to the average marginal rates excluding social security premiums, MTRL, estimated in the previous section. Before estimating the premium rates on labor incomes for each type of social security, we address two issues. First, MSST and MTRL are based on different weights. We use the income ratio as the weight for MTRL. However, for the MSST weight, because of limited data availability, we use the ratio of the number of insured persons to the total number of workers (the labor force). Insured persons paying lump-sum contributions are generally low-income earners. Therefore, total average marginal rates based on the numbers of insured persons are below those based on income levels. Second, social security payments for employees are paid by both employees and employers. Thus, the effective rates for social security premiums should be estimated as described below. Let wB and wA

96

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

denote the before-tax and after-tax wage rates, respectively. This means that wB = (1 + se)wA, where se is a tax on employers’ payments of wages. The before-tax wage rate for employees is wA = wB/(1 + se), so the after-tax wage rate is as follows:

ð1  sl ÞwA ¼

1  sl B w ¼ 1 þ se

  sl þ se B w ; 1 1 þ se

where sl is a tax on employees’ wages. Therefore, the effective tax rate paid by both employers and employees is (sl + se)/(1 + se). 3.1. Average marginal premium rates for pension insurance In this subsection, we estimate the average marginal rates of pension insurance. Pension insurance is classified into: (i) the National Pension; (ii) Employees’ Pension Insurance; and (iii) the Mutual Aid Associations’ Pension.17 Every Japanese person above the age of 20 is required by law to join the National Pension, and every regular employee is supposed to belong to either of the other Employees’ Pension Insurance scheme or to the Mutual Aid Associations’ Pension.18 We calculate the weighted average of these premium rates using the statutory rates and the number of members covered by each type of insurance. The premium for the National Pension is constant and independent of the level of income and, thus, the marginal premium rate should be set to zero. Because the premium for the remaining pensions is proportional to the income level, the marginal premium rate is, with adjustments, the corresponding statutory rate. The weight is considered as the ratio of the number of persons covered by each type of insurance to the total labor force.19 We use the marginal rates and the weights for the other pensions to estimate the weighted average of the premium rates. Data on the numbers of insured persons and the statutory insurance premium rates are from the Social Security Year Book (Shakai Hosho Nenkan) published by the National Federation of Health Insurance Societies. Labor force data are fiscal year averages from the Labor Force Survey (Rodoryoku Chosa) published by the Ministry of Internal Affairs and Communications. To calculate the marginal premium rate for either Employees’ Pension Insurance or the Mutual Aid Associations’ Pension, the corresponding statutory rate on employment incomes must be adjusted. Employment incomes are divided into regular earnings and special earnings, and the latter, paid twice a year, typically account for a large share of overall labor incomes in Japan. Before 2002, the statutory premium rate was imposed only on regular earnings,20 and another, much lower, rate is imposed on special earnings. Since the introduction of the total remuneration system in 2003, both types of earnings are subject to the same rate of pension insurance. Thus, we must recalculate the premium rate on total employment incomes. Unfortunately, we cannot obtain data on the amounts of special earnings or on the premium rate on special earnings for each pension fund from the Social Security Year Book. To recalculate the premium rates, we assume that the premiums on special earnings before 2002 were zero. We then construct the ratio of employment earnings to total earnings from a different data source. We estimate the rates of pension insurance (except for the National Pension) on labor incomes by using the product of the statutory rates and the adjustment coefficient q, defined as:

q¼

annual cash earnings : annual cash earnings þ annual special earnings

17 The Mutual Aid Associations consist of (iii-1) the Mutual Aid Association of National Government Employees; (iii-2) the Mutual Aid Association of Local Government Employees; (iii-3) the Mutual Aid Corporation of Private School Personnel; (iii-4) the Agriculture, Forestry, and Fisheries Organization Employees Mutual Aid Association; (iii-5) the Mutual Aid Associations of Public Corporation Employees; (iii-6) the Seamen’s Insurance; and (iii-7) the Farmer Pension Fund. The last three associations (iii-5)–(iii7) were liquidated in 1983, 1985, and 1985, respectively. 18 This system has been in operation since 1961, and covers our entire sample period, 1963–2007. 19 Contributors only to the National Pension are not necessarily part of the labor force. However, the marginal premium rate for the National Pension is zero, and therefore this weight does not affect our estimation. 20 To be precise, the monthly amount of the premium is calculated from the product of the statutory rates and the index of monthly regular earnings, which rounds off fractions of actual monthly payments.

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

97

Annual earnings are computed from the sum of the monthly contractual cash earnings in each year. The data on cash earnings and special earnings are from the Basic Survey on the Wage Structure (Chingin Kozo Kihon Tokei Chosa Houkoku) published by the Ministry of Health, Labor, and Welfare. 3.2. Average marginal premium rates for health insurance In this subsection, we estimate the average marginal rates of health insurance. The six types of health insurance are: the health insurance managed by government and by associations; National Health Insurance; Employees’ Insurance; Day-Laborers’ Health Insurance; Seamen’s Insurance; and Mutual Aid Association Insurance.21 Every Japanese person aged over 20 is required by law to join one of these associations. As in the previous subsection, we calculate the weighted average of these premium rates using the statutory rates and the number of members covered by each type of insurance. On the one hand, because the premium for National Health Insurance is effectively constant, we set its premium rate to zero. On the other hand, because the premium for the remaining insurance is proportional to the income level, the marginal premium rate is, after adjustment, the corresponding statutory rate. The weight is considered as the ratio of the number of persons covered by each type of insurance to the total labor force.22 Because the marginal premium rate for National Health Insurance is zero, we use the marginal rates and the weights for the other pensions to estimate the weighted average of the premium rates. The data used for our calculations are available from the Social Security Year Book, the Labor Force Survey, and the Basic Survey on the Wage Structure. As in the previous subsection, adjustments are needed when using statutory rates on employment incomes. The total remuneration system was introduced for health insurance in 2003, and different rates were imposed on regular earnings and special earnings before 2002. Fortunately, data on insurance premium rates on special earnings for health insurance are available. Therefore, our estimates of the average marginal premium rates before 2002 are:

q  ðrate on regular earningsÞ þ ð1  qÞ  ðrate on special earningsÞ; where q is as defined in the previous subsection. According to this equation, the premium rate is a weighted sum of the insurance rate on cash earnings and that on special earnings. 3.3. Average marginal premium rates for employment insurance In this subsection, we estimate the average marginal rates for employment insurance. The three types of employment insurance are: Employment Insurance for General Persons; Day-Laborers’ Insurance; and Seamen’s Insurance. We use the statutory rates and the numbers of insured members in each scheme to calculate the weighted average of these premium rates. The marginal premium rate is the corresponding statutory rate when the premium is proportional to the income level, and the weights are the ratios of the numbers of insured members to the labor force. These data are taken from the same sources as those used in the previous subsections. Not every worker is insured for unemployment. Therefore, unlike for pension and health insurance, we need information on the numbers of uninsured workers. We specify this as the difference between the total labor force and the total number of insured persons. Furthermore, the insurance premium for Day-Laborers’ Insurance is in the form of a lump sum and is independent of the level of income. For Day-Laborers’ Insurance contributors and for uninsured workers, the marginal premium rate is zero. 3.4. Average marginal premium rates for accident compensation insurance In this subsection, we estimate the average marginal rates of accident compensation insurance. We classify accident compensation insurance into Workmen’s Accident Compensation Insurance, Accident 21 As described in the previous subsection, the Mutual Aid Association consists of seven associations: (iii-1)–(iii-7). However, because of limited data availability, we use the first four associations to calculate the average marginal premium rates. 22 Note that contributors to National Health Insurance are not necessarily part of the labor force. However, the marginal premium rate for National Health Insurance is zero, and therefore this weight does not affect our estimation.

98

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

Compensation for National Government Employees, and Accident Compensation for Local Government Employees. The premium rate for Workmen’s Accident Compensation Insurance varies between occupations. Rosai Hoken Tekiyo Jigyo Saimoku No Kaisetsu (Details on Enterprises covered by Accident Compensation Insurances) published by Japan Rodo Press provides the premium rates of 94 occupations. To estimate the average marginal premium rates, we calculate the median of those rates in each year.23 For Accident Compensation for National Government Employees and Accident Compensation for Local Government Employees, we can obtain the premium rates from the Social Security Year Book. Each premium rate is levied on employers, but not on employees (sl = 0). 3.5. Average marginal premium rates for long-term care insurance Long-term care insurance was introduced in 2000. Persons aged 40–65 (termed class 2 persons) must pay this insurance premium. Data on the number of insured persons and the insurance premium rate are taken from the Social Security Year Book. For uninsured persons, the marginal premium is zero. The number of uninsured persons is the difference between the labor force and the number of insured persons. We compute the weighted average of the rates for the insured and uninsured by using the shares of the corresponding groups as weights. 3.6. Average marginal tax rates including all social security premium rates The average marginal premium rates for all types of social security are shown in columns 2–6 of Table 4. Although the pension insurance rate increases gradually, there is little change in the other four rates. Note that the rates of employment insurance, accident compensation insurance, and long-term care insurance are no more than 1%. Now, we can estimate the average marginal tax rates including social security premiums, MMTRL. Let MSST be the sum of all of the social security premium rates, which are reported in column 7 of Table 4. When miscellaneous incomes multiplied by the capital share h are allocated to capital incomes, we obtain:

MMTRLh ¼ MTRLh þ MSST: When h = 0 (h = 1), miscellaneous incomes are treated as labor (capital) incomes. The results are presented in Table 5. The solid line in Fig. 2 represents the time series for MMTRLh. The rates excluding social security premiums (represented by the broken line, MTRLh) range from 14% to 21% for the sample period. Social security premiums increase the rate by around five percentage points up to 1975 and by around 14 percentage points from 2000. Premiums for the National Pension and National Health Insurance, which account for a large share of social security, are independent of income; that is, the marginal premium rate is zero. This might reduce the progressiveness of the marginal tax rate on labor income. In the next section, we compare our average marginal taxes on labor incomes with average taxes. 4. Comparing our tax rates with existing average tax rates In this section, we compare our series of average marginal tax rates with the average tax rates calculated by Mendoza et al. (1994). They used data on tax revenues from the OECD’s Revenue Statistics and data on income and expenditures from the OECD’s National Accounts of OECD Countries. They present a series of average tax rates on consumption, labor, and capital for seven OECD countries for the period 1965–1996.24 In this section, we compare our series with their Japanese series. In Fig. 3, we 23 We use the median instead of the mean for the following two reasons. (1) Whereas many occupations have rates of below 1%, a few have extremely high rates of over 10%. (2) We cannot obtain data on the number of workers in each occupation. 24 They originally estimated a series for 1965–1986; a series that extends to 1996 is available from E.G. Mendoza’s Web site

ðhttp : ==econweb:umd:edu=mendoza=pp=newtaxdata:pdf Þ: We cannot updated their series because some definitions in the Revenues Statistics changed in 1997.

99

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106 Table 4 Average marginal tax rates for social security. Year

Pension insurance

Health insurance

Employment insurance

Accident insurance

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

0.0196 0.0211 0.0242 0.0241 0.0255 0.0251 0.0279 0.0295 0.0291 0.0290 0.0335 0.0342 0.0334 0.0388 0.0385 0.0381 0.0387

0.0269 0.0285 0.0233 0.0255 0.0268 0.0274 0.0279 0.0281 0.0285 0.0285 0.0298 0.0293 0.0296 0.0311 0.0314 0.0323 0.0347

0.0046 0.0049 0.0051 0.0051 0.0053 0.0053 0.0050 0.0052 0.0053 0.0054 0.0054 0.0056 0.0056 0.0056 0.0055 0.0057 0.0062

0.0042 0.0044 0.0043 0.0044 0.0043 0.0046 0.0050 0.0053 0.0066 0.0066 0.0068 0.0073 0.0090 0.0089 0.0086 0.0086 0.0087

Care insurance

0.0553 0.0588 0.0569 0.0591 0.0618 0.0624 0.0658 0.0681 0.0695 0.0695 0.0754 0.0765 0.0776 0.0843 0.0840 0.0847 0.0882

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0.0440 0.0444 0.0444 0.0430 0.0454 0.0515 0.0505 0.0495 0.0502 0.0539 0.0658 0.0599 0.0580 0.0612 0.0616 0.0684 0.0710 0.0700 0.0709 0.0708 0.0705 0.0703 0.0698 0.0709 0.0710 0.0732 0.0759 0.0790

0.0348 0.0362 0.0363 0.0363 0.0361 0.0362 0.0360 0.0363 0.0376 0.0382 0.0373 0.0395 0.0389 0.0389 0.0391 0.0394 0.0392 0.0400 0.0396 0.0397 0.0408 0.0404 0.0404 0.0432 0.0427 0.0430 0.0435 0.0441

0.0062 0.0063 0.0064 0.0064 0.0064 0.0065 0.0064 0.0064 0.0067 0.0068 0.0069 0.0070 0.0069 0.0057 0.0057 0.0057 0.0057 0.0057 0.0057 0.0057 0.0057 0.0086 0.0087 0.0088 0.0088 0.0090 0.0079 0.0081

0.0104 0.0112 0.0113 0.0111 0.0112 0.0112 0.0114 0.0114 0.0119 0.0119 0.0121 0.0125 0.0118 0.0124 0.0125 0.0105 0.0105 0.0105 0.0103 0.0104 0.0103 0.0104 0.0104 0.0086 0.0087 0.0088 0.0080 0.0081

0.0953 0.0980 0.0984 0.0969 0.0991 0.1054 0.1042 0.1036 0.1064 0.1108 0.1222 0.1188 0.1155 0.1182 0.1189 0.1241 0.1264 0.1262 0.1265 0.1265 0.1311 0.1366 0.1362 0.1372 0.1383 0.1420 0.1431 0.1470

0.0038 0.0069 0.0068 0.0057 0.0070 0.0080 0.0078 0.0078

MSST

compare our average marginal tax rates on labor income, MMTRL, with the corresponding average tax rates presented by Mendoza et al. (1994), ATRL.25 In the early 1980s, the average and marginal tax rates on labor income were similar, and since then, the average rate has slightly exceeded the marginal rate. Fig. 3 indicates that average marginal tax rates are similar to average tax rates. This is the product of the Japanese tax and social security system. Generally, average marginal tax rates are raised if income is taxed progressively but lowered if there are lump-sum taxes. As already noted, whereas the Japanese social security system levies social security contributions on most employers in proportion to their wages, it imposes lump-sum taxes on others, including the self-employed. 25 Because of how our variables are constructed, our average marginal tax rates on labor income should be compared with the sum of their average tax rates on consumption and labor income.

100

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106 Table 5 Average marginal tax rates on labor income with social security premiums. Year

MMTRL0

MMTRL1

MMTRLh

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

0.2295 0.2381 0.2289 0.2209 0.2208 0.2263 0.2242 0.2251 0.2231 0.2344 0.2485 0.2303 0.2193 0.2360 0.2368 0.2470 0.2625

0.2382 0.2463 0.2361 0.2277 0.2275 0.2348 0.2298 0.2302 0.2268 0.2389 0.2537 0.2328 0.2215 0.2376 0.2383 0.2484 0.2642

0.2308 0.2394 0.2299 0.2220 0.2218 0.2277 0.2252 0.2260 0.2236 0.2351 0.2493 0.2305 0.2195 0.2362 0.2370 0.2472 0.2627

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0.2721 0.2819 0.2880 0.2867 0.2883 0.2962 0.3033 0.3040 0.2958 0.2998 0.3192 0.3204 0.3210 0.3225 0.3094 0.3037 0.3053 0.3142 0.3132 0.3048 0.3091 0.3134 0.3087 0.3065 0.3088 0.3201 0.3323 0.3303

0.2745 0.2850 0.2907 0.2901 0.2916 0.2990 0.3072 0.3072 0.3073 0.3116 0.3314 0.3327 0.3343 0.3356 0.3218 0.3141 0.3154 0.3245 0.3229 0.3131 0.3170 0.3210 0.3172 0.3170 0.3205 0.3303 0.3446 0.3420

0.2725 0.2824 0.2885 0.2873 0.2888 0.2967 0.3039 0.3046 0.2983 0.3023 0.3218 0.3230 0.3234 0.3247 0.3114 0.3053 0.3069 0.3157 0.3145 0.3059 0.3103 0.3145 0.3099 0.3081 0.3108 0.3218 0.3344 0.3324

Note: The series from 1963 to 1979 are based on 68SNA and the series from 1980 to 2007 are based on 93SNA.

It is interesting that the progressiveness of income tax has a small effect. In referring to Feenberg and Coutts (1993) and Prescott (2004) assumed that the ratio of marginal tax rates excluding rates of social security contributions to the average tax rates of Mendoza et al. (1994) was 1.6 for the US. Our results indicate that, in Japan, the ratio is unity, which is lower than that referred to by Prescott. This difference arises because the minimum taxable level of personal income is higher in Japan than in the US. In Fig. 4, we compare our average marginal tax rates with average tax rates on capital income. The difference between average rates and average marginal rates was in decline until the late 1980s. After that, the gap increased again. This suggests that Japanese taxes on capital income are as progressive as are taxes on labor income without social security premiums.

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

101

Fig. 2. Average marginal tax rates on labor income with social security premiums. Note: Both series are based on miscellaneous incomes being allocated to both capital and labor incomes according to their factor shares.

5. Comparison with BCA In this section, we compare our results with wedges from BCA developed recently by Chari et al. (2007a). In standard calibration analysis, one chooses an appropriate dynamic macroeconomic model with a plausible set of parameters, estimates exogenous shocks from actual data, and conducts a simulation to evaluate the impact of each shock on the endogenous variables. In BCA, one uses a standard dynamic general equilibrium model to estimate the shock variables, called wedges, from actual endogenous variables, and conducts a simulation to investigate the extent to which each wedge contributes to actual business cycles. The BCA wedges are interpreted as taxes that prevent the economy from achieving its Pareto optimum allocations. The four wedges considered by Chari et al. (2007a) are efficiency, labor, government, and investment wedges. Kobayashi and Inaba (2006) and Otsu (2008) applied BCA to the Japanese economy and concluded that efficiency and labor wedges play an important role in business cycles. These findings are consistent with those of Hayashi and Prescott (2002), who found that technology shocks (known as efficiency wedges in BCA terminology) were the most significant contributors to the depression of the 1990s, Japan’s lost decade. Prescott (2004) compared labor wedges and average marginal tax rates on labor income for the G7 countries including Japan, and concluded that taxes on labor incomes can almost completely account for labor wedges. For other discussions on wedges, see Golosov et al. (2006) and Shimer (2009). Taking BCA analyses into account, we compare our estimated tax rates on labor and capital incomes with labor and capital wedges.26 Note that we use capital wedges although BCA often uses

26 We are grateful to Masaru Inaba for providing data on the labor and capital wedges analyzed in this section. Note that Kobayashi and Inaba (2006) used 93SNA from 1980 to 2003, but with a base year of 1997 rather than our base year of 2002.

102

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

Fig. 3. Marginal and average tax rates on labor income. Note: Our series is based on miscellaneous incomes being allocated to both capital and labor incomes according to their factor shares.

investment wedges. There is controversy about the relationship between capital and investment wedges. Christiano and Davis (2006) point out that, in models with investment adjustment costs, replacing investment wedges by capital wedges may affect the results. In response, Chari et al. (2007b) showed that both wedges are equivalent mathematically. Thus, for comparison with our series, we use capital wedges. Fig. 5 illustrates the average marginal tax rate on labor income (MMTRL) and the labor wedge. The labor wedge exceeds the marginal tax rate for the whole period. The difference rises from about five percentage points at the beginning of the period to about 15 percentage points at the end. The labor wedge increased from 1985, whereas the marginal tax rate remained unchanged at around 30%. Overall, the average marginal tax rate explains around 70% of the labor wedge on average. Kobayashi and Inaba (2006) and Otsu (2008) argued that the labor wedge is an important contributor to Japanese business cycles. Given their results, the average marginal tax rate is also a major factor. Increased social security burdens might cancel out the effect of tax cuts on the Japanese economy. A rise in the marginal tax rate on labor income raises the relative disutility of labor and, consequently, lowers labor supply. In addition, a negative technology shock lowers firms’ labor demands. Thus, in the lost decade, not only might slow technological progress have reduced output, but reduced labor demand and increased social security premiums might have lowered labor supply. Both would have had negative effects on the Japanese economy in the long run. The argument that tax plays an important role in explaining the economy is not new. As already mentioned, to simulate labor supply using a simple neoclassical general equilibrium model, Prescott (2004) estimated marginal tax rates for the G7 countries in the periods 1970–1974 and 1993–1996. On the basis that model predictions are consistent with actual values, Prescott claimed that labor wedges are completely accounted for by marginal tax rates.

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

103

Fig. 4. Marginal and average tax rates on capital income. Note: Our series is based on miscellaneous incomes being allocated to both capital and labor incomes according to their factor shares.

However, we disagree with Prescott (2004) to some extent. Prescott used Mendoza et al.’s (1994) series of average tax rates for individual G7 countries, and multiplied each average rate by 1.6 to estimate the marginal rates. Prescott’s figure of 1.6 comes from empirical research on the US (Feenberg and Coutts, 1993), and is assumed to apply to all G7 countries including Japan. However, as we showed in the previous section, the ratio excluding social security premiums in Japan is about unity. Thus, we argue that the tax rate on labor income accounts for around 70% of the labor wedge in the Japanese economy. The difference between average marginal tax rates and labor wedges widened from the early 1990s. The 1988 revision of the Labor Standards Law, which reduced legal working hours, may be at least partially responsible for this difference.27 An alternative possible cause is the recent rapid accumulation of the fiscal deficit. To stimulate the economy, the Japanese government has continued not only cutting taxes but also issuing bonds. Japan’s fiscal deficit as a proportion of GDP is the highest among the OECD countries. According to the Ricardo–Barro effect, current fiscal deficits are essentially future taxes. Fig. 6 illustrates the average marginal tax rate and the capital wedge. Whereas the marginal tax rate stays around 50%, the capital wedge fluctuates greatly, ranging from 33% to over 70%. The average of the marginal tax rate (0.55) is slightly above that of the capital wedge (0.47). The first-order autocorrelation coefficient for the capital wedge of 0.12 suggests that fluctuations in the capital wedge could be explained by a short-run shock; an example is an unexpected monetary shock. According

27 Hayashi and Prescott (2002) argued that Japan’s lost decade is explained by the 1988 revision of the Labor Standards Law as well as low productivity rather than by financial collapses. In their model, labor input is decomposed into working hours and the number of workers; therefore, their framework differs from ours. By reducing legal working hours, the law revision acts as a wedge, distorting endogenous labor input in the BCA framework.

104

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

Fig. 5. Marginal tax rates on labor income and labor wedges. Note: Our series is based on miscellaneous incomes being allocated to both capital and labor incomes according to their factor shares.

to Kobayashi and Inaba (2006) and Otsu (2008), the investment wedge, which is equivalent to the capital wedge, has not made any significant contribution to Japan’s business cycles.

6. Concluding remarks In this paper, we have applied the method of Joines (1981) to estimate average marginal tax rates for Japan from 1963–2007. We considered both self-assessment taxpayers and withholding income taxpayers. We calculated their average marginal tax rates separately, and then combined these tax rates into a weighted average to obtain total average marginal tax rates. Moreover, we included social security tax rates, which consist of pension insurance, health insurance, employment insurance, accident compensation insurance, and long-term care insurance. We obtained the following results. The average marginal tax rates on labor incomes without social security premiums and on capital incomes both increased until the late 1980s and then fluctuated around 17% and 50%, respectively. Because of increased social security payments, the tax rates with social security premiums range from 22% to 33% for the entire sample period. Average marginal tax rates on capital incomes, ranging from 35% to 58%, changed dramatically following the 1973 oil crisis and because of tax reforms. In addition, we made two comparisons. First, we compared our estimates with Mendoza et al.’s (1994) average tax rates. The average tax rates on labor incomes are similar to, and from the late 1980s were above, the average marginal tax rates with average marginal premium rates of social security. This is because taxation is progressive but social security is regressive. The average tax rates on capital incomes were slightly below the marginal tax rates for most of the sample period. This suggests that Japan’s capital tax is relatively progressive.

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

105

Fig. 6. Marginal tax rates on capital income and capital wedges. Note: Our series is based on miscellaneous incomes being allocated to both capital and labor incomes according to their factor shares.

Second, we compared our estimates with the labor and capital wedges estimated by Kobayashi and Inaba (2006). We found that marginal tax rates on labor incomes account for about 70% of the labor wedge. This implies that Prescott (2004) exaggerates the importance of taxing labor incomes. Because the difference between labor tax rates and wedges has increased, other contributory factors, such as reduced legal working hours and a rapidly accumulating fiscal deficit, should be considered. Our marginal tax rates cannot explain fluctuations in capital wedges. A future research task is as follows. It would be worthwhile computing marginal taxes from a different data source. Microeconomic data have recently become more accessible in Japan. If we obtained an adequate number of individual observations, we could construct average marginal tax rates on total incomes using the nonparametric method proposed by Akhand and Liu (2002). Comparing such microeconomic-based estimates to our macroeconomic-based estimates would be intriguing. Acknowledgments We are grateful to Masaru Inaba, Keisuke Otsu, Garry Fleming, Yosuke Takeda, Masayoshi Tsurumi, and two anonymous referees for many helpful comments, and to Chie Hanaoka for excellent research assistance. Financial support from the Zengin Foundation for Studies on Economics and Finance is acknowledged with thanks. An earlier Japanese version of this paper won the Excellence Award in Zei Ni Kansuru Ronbun (Paper on Tax) from the Tax Payment Association. All the data are in Excel files downloadable from our website (http://sites.google.com/site/amtrjapan/home). References Akhand, H., Liu, H., 2002. Marginal income tax rates in the United States: a non-parametric approach. J. Monet. Econ. 49 (2), 383–404.

106

H. Gunji, K. Miyazaki / J. Japanese Int. Economies 25 (2011) 81–106

Barro, R.J., Sahasakul, C., 1983. Measuring the average marginal tax rate from the individual income tax. J. Bus. 56 (4), 419–452. Barro, R.J., Sahasakul, C., 1986. Average marginal tax rates from social security and the individual income tax. J. Bus. 59 (4), 555– 566. Braun, R.A., Esteban-Pretel, J., Okada, T., Sudou, N., 2006. A comparison of the Japanese and US business cycles. Japan World Economy 18, 441–463. Chari, V.V., Kehoe, P.J., McGrattan, E.R., 2000. Sticky price models of the business cycle: can the contract multiplier solve the persistence problem? Econometrica 68 (5), 1151–1179. Chari, V.V., Kehoe, P.J., McGrattan, E.R., 2007a. Business cycle accounting. Econometrica 75 (3), 781–836. Chari, V.V., Kehoe, P.J., McGrattan, E.R., 2007b. Comparing alternative representations and alternative methodologies in business cycle accounting. Staff Report 384, Federal Reserve Bank of Minneapolis. Christiano, L., Davis, J., 2006. Two flaws in business cycle accounting. NBER Working Paper 12647. Cole, H.L., Ohanian, L.E., 1999. The great depression in the United States from a neoclassical perspective. Quarterly Review, Federal Reserve Bank of Minneapolis, issue Win, 2–24. Esteban-Pretel, J., Nakajima, R., Tanaka, R., 2010. TFP growth slowdown and the Japanese labor market in the 1990s. J. Japanese Int. Economies 24 (1), 50–68. Feenberg, D.R., Coutts, E., 1993. An introduction to the TAXSIM model. J. Pol. Anal. Manage. 12 (Winter), 189–194. Golosov, M., Tsyvinski, A., Werning, I., 2006. New dynamic public finance: a user’s guide. NBER Macroecon. Annu. 2006, 317– 363. Hayashi, F., Prescott, E.C., 2002. The 1990s in Japan: a lost decade. Rev. Econ. Dynam. 5 (1), 206–235. Joines, D.H., 1981. Estimates of effective marginal tax rates on factor incomes. J. Bus. 54 (2), 191–226. Kobayashi, K., Inaba, M., 2006. Business cycle accounting for the Japanese economy. Japan World Econ. 18 (4), 418–440. McGrattan, E.R., 1994. The macroeconomic effects of distortionary taxation. J. Monet. Econ. 33 (3), 573–601. McGrattan, E.R., Ohanian, L.E., 2010. Does neoclassical theory account for the effects of big fiscal shocks? Evidence from World War II. Int. Econ. Rev. 51 (2), 509–553. McGrattan, E.R., Rogerson, R., Wright, R., 1997. An equilibrium model of the business cycle with household production and fiscal policy. Int. Econ. Rev. 38 (2), 267–290. McKee, M.J., Visser, J.J.C., Saunders, P.G., 1986. Marginal tax rates on the use of labour and capital in OECD countries. OECD Econ. Stud. 7, 45–101. Mendoza, E.G., Razin, A., Tesar, L.L., 1994. Effective tax rates in macroeconomics cross-country estimates of tax rates on factor incomes and consumption. J. Monet. Econ. 34 (3), 297–323. Otsu, K., 2008. Jitsubutsu keiki junkan riron to nihon keizai (real business cycle theory and the Japanese economy, in Japanese), Kinyu Kenkyu 27 (4), 45–86. Prescott, E.C., 2004. Why do Americans work so much more than Europeans? Fed. Reserve Bank Minneapolis Quart. Rev. 28 (1), 2–13. Seater, J.J., 1985. On the construction of marginal federal personal and social security tax rates in the US. J. Monet. Econ. 15 (1), 121–135. Shimer, R., 2009. Convergence in macroeconomics: the labor wedge. Am. Econ. J.: Macroecon. 1 (1), 280–297. Siu, H., 2008. The fiscal role of conscription in the US World War II effort. J. Monet. Econ. 55 (6), 1094–1112. Stephenson, F.E., 1998. Average marginal tax rates revisited. J. Monet. Econ. 41 (2), 389–409.