How unequal are opportunities in Korea, Japan, and Taiwan? Father's ...

[Very incomplete. Do not distribute without permission]

How unequal are opportunities in Korea, Japan, and Taiwan? Father’s Education and Son’s Achievement

March 11, 2012

Woojin Lee Department of Economics Korea University Seoul, Korea [email protected]

Atsuko Ueda Faculty of Political Science and Economics Waseda University Tokyo, Japan [email protected]

Jayee Ko Department of Economics Korea University Seoul, Korea [email protected]

Fengye Sun Graduate School of Economics Waseda University Tokyo, Japan [email protected]

Abstract According to a contemporary interpretation of equality of opportunity (Roemer, 1998), a society has achieved equality of opportunity if it is the case that what individuals accomplish, with respect to some desirable objective, is determined wholly by their choices and personal effort, rather than circumstances that are beyond their control. The present paper examines the extent of unequal opportunity for income acquisition in three East Asian countries: Korea, Japan, and Taiwan. In this paper, equality of opportunity for income acquisition is achieved in a country when the distributions of post-fisc income are the same for different types of citizens in that country, where a citizen’s type is defined by father’s education. We estimate the extent of unequal opportunity for income acquisition in these countries in two ways. First, given a country’s present tax policies, we calculate the extent to which existing income inequality of that country is due to differential circumstances, as opposed to differential effort. We examine, counterfactually, what would happen to the income inequality were all types endowed with the same circumstances, and compare it with the actual inequality. Second, given a country’s current inequality of circumstances across types, we calculate the optimal tax policy that would be required for equalizing post-fisc income distributions across types, and compare them with actual tax rates. In terms of the optimal tax rates required for equalizing opportunities for income acquisition among citizens, the level of unequal opportunity in Korea and Taiwan are found to be similar to those of Spain, Italy, and the United States reported in Roemer et al. (2003). The observed tax rates in Korea and Taiwan are, however, lower than those of the aforementioned three countries, and there are very large differences between the optimal and the observed tax rates. This implies that the taxbenefit policies of Korea and Taiwan have played very little role in correcting unequal opportunities for income acquisition among their citizens. Japan is different; the EOp tax rate is less than the observed tax rate or equal to zero. This is because the individuals classified as the least advantaged type in Japan are much better off than those in Korea and Taiwan, due to data availability. Keywords: Fiscal regimes, equality of opportunity, earnings distribution JEL Classification: D63, H00

1. Introduction

For the last four decades or so, the economic success stories of East Asian countries – such as Japan, Korea, Singapore, Taiwan, and recently China – have been well documented and widely discussed among scholars and policy makers. These countries have experienced not only fast growth; the fast growth was accompanied by relatively equal distribution of income and a high level of school enrollment. Very little has been known about the extent to which opportunities are equal in these countries among their citizens. The current paper aims at filling this gap by carrying out an empirical analysis for three East Asian countries. Using micro-survey data sets of Korea, Japan, and Taiwan, we examine how unequal opportunities are in these countries for income acquisition among their citizens, and calculate the extent to which their tax-benefit policies correct those unequal opportunities. To the best of our knowledge, our work is the first systematic empirical investigation of the opportunity inequality in East Asian countries. Inequality in income has many causes, and not all of them are equally objectionable. Indeed there is no a priori reason to believe that a country with a relatively low level of inequality of a final outcome will also have a relatively low level of inequality of opportunities for that outcome. Suppose a society’s reward scheme is such that it shrinks the variation in outcomes due to differential efforts while increasing the variation in outcomes due to differential circumstances, such as family backgrounds. That society is likely to have a high level of opportunity inequality and a low level of outcome inequality. The present paper examines the extent to which relatively equal distribution of income of the three East Asian

1

countries has been accompanied by equality of opportunities for income acquisition among their citizens. The theory of equality of opportunity poses two different, but closely interrelated issues: the first is the problem of measuring the degree of unequal opportunity in a society; the second is the design of a public policy intended to achieve equality of opportunity. Accordingly, our goals in the current paper are twofold. First, given a country’s present tax policies, we calculate the extent to which existing income inequality of that country is due to differential circumstances, rather than to differential effort. We examine, counterfactually, what would happen to the income inequality were all types endowed with the same circumstances, and then compare it with the actual inequality. Second, given a country’s current inequality of circumstances across types, we calculate the optimal tax policy that would be required for equalizing post-fisc income distributions across types, and compare them with actual tax rates. Although equality of opportunity is an ethical principle with almost universal appeal, there is indeed a spectrum of views regarding what defines opportunities and what is required for equalizing opportunities (Lefranc, Pistolesi, and Trannoy, 2006). The present paper adopts one contemporary interpretation (Arneson, 1989; Cohen, 1989, Roemer, 1993, 1998). According to that interpretation, a society has achieved equality of opportunity if it is the case that what individuals accomplish, with respect to

some desirable objective, is determined wholly by their choices and personal

effort, rather than circumstances that are beyond their control. The aim of equal opportunity is not to hold persons responsible for characteristics which are due to

2

their being in a disadvantaged type. A society equalizes opportunities if it makes the achievement of the objective of individuals be a function only of their efforts, not of their circumstances. According to Roemer (1998), equality of opportunity for a certain objective is achieved when the values of the objective are equal for all those who exercised a ‘comparable’ degree of effort, regardless of their circumstances. A tricky part in any theory of equal opportunity is to decide when two people in differential circumstances have exercised a ‘comparable’ degree of effort. Effort is a multi-dimensional set of responsible behaviors, including the acquisition of skill, intensity of job search, etc., which engender the final outcome of income acquisition. It should not merely be ‘labor hours’ or ‘intensity of labor,’ for two people, by virtue their different circumstances, may exercise very different amounts of labor. Roemer (1993, 1998) propose that we measure a person’s effort by the quantile at which he or she sits on the distribution of his or her type, for the quantile measure ranks a person’s effort by comparing him/her only to others of his/her own type. It turns out that effort in the Roemer’s theory is the residual that explains differential outcomes, once circumstances have been delineated.1 If the number of types is significantly small compared with the total number of individuals, there will usually be a large number of individuals in each type and thus there will ensue some distribution of the objective for each type. If some types have ‘better’ distributions of the objective than others, this must be due to their better circumstances. On the other hand, the differential outcomes of these individuals within a type are attributed to differential effort. 1

Thus random luck, another determinant of an outcome, appears as effort in their theory. But this may not be a bad aspect of their theory, for it may average out across individuals. 3

One advantage, among others, of the Roemer’s approach is that it is a computable concept of equality of opportunity. Indeed Roemer et al. (2003) empirically examine, for eleven Western countries, the extent to which fiscal policies equalize opportunity for income acquisition among citizens.2 Kim and Lee (2009) apply Roemer’s method to Korean data. The present paper employs the same method to the data sets of Korea, Japan, and Taiwan, and compares the main results with those reported in Roemer et al. (2003). In section 2, we briefly summarize our model and the strategy of counterfactual estimation. Section 3 describes data, explains our method of data analysis, and reports on our major results. Section 4 concludes.

2. The model

In this section, we present our basic model, which is largely based upon the model of Roemer et al. (2003). We assume that types are classified by the level of father’s education. Suppose L represents labor hours, w the wage rate, j(.) a function representing the government’s net tax (tax-benefit) policy, and x º wL pre-fisc income. We assume that the policy is affine; j(x ) = tx - T , where t is the marginal income tax rate and T is per-capita transfer payments. Thus the government policy is characterized by a pair (t,T ) , and post-fisc income is given by y = x - j(x ) = (1 - t )x + T . Let q be a variable capturing circumstances, and e an unobservable

2

See Aaberge et al. (2001), Dardanoni et al. (2005), and World Bank (2008) for related work. 4

variable capturing individual effort. We assume that the wage rate depends on both circumstances and individual effort3: w = w(q, e) .

(1)

We will denote the wage rates of the individuals in type q by w q and its distribution by F q (.) . Facing w , each individual chooses some optimal amount of labor hours: L* = L(w; t,T ) .4

(2)

Given equations (1) and (2), the optimal level of pre-fisc income is given by x * = wL* = w(q, e)L(w(q, e); t,T ) .

(3)

Equation (3) shows that circumstances affect the pre-fisc income through two channels: (a) Circumstances affect wages, and wages affect pre-fisc income directly, given L; and (b) Circumstances affect wages, and wages affect pre-fisc income indirectly, through their effect on labor supply. As in Roemer et al. (2003), we assume that individuals have a quasi-linear utility function: 1+

u(y, L) = y - aL

1 h

,

(4)

where y is post-fisc income. Then the optimal labor supply is h

h

æ (1 - t )w ö÷ æ (1 - t ) ö÷ h ÷÷ == çç ÷ w , L = çç çè a(1 + 1 / h) ÷÷ø çè a(1 + 1 / h) ÷ø *

(5)

In a general equilibrium analysis where w is determined by the supply and the demand, it will depend on the government policy as well. We take a partial equilibrium analysis, where w is considered the level of skill. 4 We are assuming that the optimal labor supply depends only upon w and j . In reality, however, 3

optimal labor supply will also be influenced by q and e as well as w and j . (Perhaps, one can think about the work ethic taught by parents.) 5

pre-fisc income is h

ö÷ 1+h æ 1-t ÷ w , x = wL = çç çè a(1 + 1 / h) ÷÷ø *

*

(6)

and post-fisc income is h

ö÷ 1+h æ 1-t ÷ w + T . (7) y * = (1 - t )x + T = (1 - t ) çç ÷ çè a(1 + 1 / h) ÷ø *

2.1 Estimation of the extent to which existing income inequality is due to differential circumstances

Here we examine, counterfactually, what would happen to the income inequality were all types endowed with the same circumstances. Suppose we were able to the wage equation (1) econometrically. Note that we cannot observe e ; we attribute it to the error term in the regression. For instance, suppose we estimate the following: wi = g(qi ) + ei ,

(8)

We now estimate counterfactual wages, labor hours, and pre-fisc incomes. Suppose g e (.) is the empirically estimated functions of equations (8), wie = g e (qi ) is the predicted value of the wage, and ei º wi - g e (qi ) is the residual from equations (8). We now assign the same circumstances ( q ) to all individuals, and compute the following counterfactual wages and labor hours.

wi = g e (q ) + ei ,

6

(9)

and Li = h(wi ; t,T ) .

(10)

Thus wi is the wage that would transpire were there no differences in people’s circumstances, and Li is the amount of labor hours that would transpire were there no differences in people’s circumstances. Note that we are attributing differential amount of individual effort to the estimated residuals. We now compute the following counterfactual level of pre-fisc income: xi = wi Li .

(11)

Then xi is the amount of pre-fisc income if there are no differences in people’s circumstances. Thus any difference in xi among individuals reflects only different degrees of individual effort. We now compute two inequality indices, such as the Gini coefficients: one from x i = wi Li (actual observation) and the other from xi = wi Li (counterfactual

observation). Now

G (wi Li ) - G (wi Li ) G (wi Li )

can be called the share of observed (pre-fisc)

income inequality due to unequal circumstances, while

G (wi Li ) G (wi Li )

may be called the

share of observed income inequality due to differential effort. One can further decompose the contribution of circumstances to the observed inequality into the two channels. In addition to equation (11), we also compute:

xi2 = wi Li ,

7

(12)

xi3 = wi Li .

(13)

Then the total effect of circumstances on inequality is decomposed into the three components as follows: G (wi Li ) - G (wi Li ) G (wi Li )

=

G (wi Li ) - G (wi Li ) G (wi Li ) - G (wi Li ) . + G (wi Li ) G (wi Li ) channel (a)

G (wi Li ) - G (wi Li ) G (wi Li )

=

channel (b)

G (wi Li ) - G (wi Li ) G (wi Li ) - G (wi Li ) . + G (wi Li ) G (wi Li ) channel (a)

(14)

(15)

channel (b)

Thus we decompose the total effect as follows: æ ö 1 çççG (wi Li ) - G (wi Li ) G(wi Li ) - G (wi Li )÷÷÷ The effect through channel (a): ç + ÷÷ ; ÷÷ 2 çç G (wi Li ) G (wi Li ) è ø÷ æ ö 1 çççG (wi Li ) - G (wi Li ) G(wi Li ) - G (wi Li )÷÷÷ The effect through channel (b): ç + ÷÷ . ÷÷ 2ç G (wi Li ) G (wi Li ) ÷ø çè So far we explained how we estimate the proportion of pre-fisc income inequality due to differential circumstances. We can also estimate the extent to which post-fisc income inequality is due to differential circumstances. This can be done by computing two inequality indices: one from yi = wi Li - Net Taxes (actual observation)

and

the

other

from

yi = wi Li - Net Taxes

(counterfactual

observation).

2.2 Calculation of optimal taxes that would be required to equalize opportunities

8

Suppose we denote the distribution of post-fisc income in type q by H q (.) . From equation (7), we know that y * (w; t,T ) < y if and only if 1

æ(y - T )(a(1 + 1 / h ))h ÷ö1+ h ç ÷÷ . w < çç (1 - t )(1 - t )h èç ø÷

Thus the distribution function of post-fisc

income in type q at policy (t,T ) is 1 ö æ ççæ (y - T )(a(1 + 1 / h))h ö1+h ÷÷ ÷÷ ÷÷ . H q (y ) = F q ççççç ÷÷ ÷÷ h çççè (1 - t )(1 - t ) ø ÷÷ çè ø

(16)

Now post-fisc income at the th quantile of H q (.) at the policy (t,T ) is given by v q (p; t,T ) such that: 1 ö æ çæ (v q (p; t,T ) - T )(a(1 + 1 / h))h ö1+h ÷÷ ç ÷÷ ÷÷ . p º F q ççççç ÷ h çççè ÷ø÷ ÷÷ (1 - t )(1 - t ) ÷ø çè

(17)

Roemer (1993, 1998) identify all those who sit at the th quantile of their type distributions as having expended effort in the comparable degree. As in Roemer et al. (2003), we impose the revenue neutrality condition, keeping constant the government revenue used for non-transfer-payment purposes. Suppose S is the value of government services (capturing non-transfer payments) per capita and F is the entire distribution of wages. Then the government budget constraint is h

ö÷ æ 1-t ÷ T (t; S ) = t çç çè a(1 + 1 / h) ÷ø÷

òw

1+ h

dF (w ) - S .

This means that for S and F given, the policy space is uni-dimensional.

9

(18)

Conceptually, equal opportunity is achieved when income levels are equal across types at any given degree of effort p . We may achieve such equalization in an efficient way by maximizing the minimum income level of the individuals, across all types, at the specific effort level in question. In other words, we may maximize: Min v q (p; t,T (t, S )) .

(19)

q

Unfortunately, there will, in general, be a continuum of solutions for such maximization problems, one for each p Î [0,1] . The first-best solution to the problem is achievable only when all these policies, t p , are identical for all ; we cannot expect this in general. Roemer (1993, 1998, 2003) propose that we instead maximize:

ò

1

Min v q (p; t,T (t, S ))d p .

(20)

q

0

An underlying assumption behind this formulation is that the objective function of the citizens in each quantile, namely Min v q (p; t,T (t, S )) , receives the same weight q

in an additive social objective function. The equal opportunity fiscal policy is the tax rate that maximizes equation (20). In general, the distribution function of pre-fisc income of the worst-off type will cross with the distribution functions of other types. In our application as well as Roemer et al. (2003)’s, however, the empirical distribution function of pre-fisc income of the worst-off type does not cross with the empirical distribution functions of the other types in almost all cases. (See Figure 2 of the present paper.) This means that the equal opportunity tax policy is equivalent to the tax policy that maximizes the average post-fisc income of the worst-off type at policy t : ¥

ò ((1 - t )x (w; t,T ) + T (t, S ))dF *

0

10

1

(21)

Thus the opportunity-equalizing fiscal policy is

t EOP = Max[1 where A =

òw

1+ h

dF 1 and B =

òw

hB , 0] , (1 + h )(B - A)

(22)

1+ h

dF .

Typically B will be significantly larger than A. In this case t EOP > 0 . But if the distribution of wages of the worst-off type is not very different from the distribution of wages of the whole society, then B - A will be small, and thus

t EOP = 0 . This means that there should be no redistributive taxation to equalize opportunities for income; there is so little inequality of opportunity, pre-fisc, and thus any taxation would be counter-productive, given the deadweight losses incurred. Roemer et al. (2003) compute one more tax rate, which they call a benchmark policy. A benchmark policy is the tax rate such that h

æ 1-t ÷÷ö T = t çç çè a(1 + 1 / h )÷÷ø

òw

1+ h

dF (w ) - S = 0 .

(23)

Thus the benchmark tax rate, t Bench , is one that would just suffice to raise government expenditures of S per capita and make no inter-citizen transfers. Roemer et al. (2003) also construct an index of measuring the extent to which fiscal regimes equalize opportunities for income acquisition as follows. Suppose

Y (t Obs ) , Y (t EOp ) , and Y (t Bench ) are the average post-fisc income of the worst-off type at the observed (actual) policy, the EOp policy, and the benchmark policy, respectively. They define a measure: Y (t Obs ) -Y (t Bench ) . n= Y (t EOp ) -Y (t Bench )

(24)

If n = 0 , then the observed fiscal policy is the benchmark policy, and if n = 1 , the

11

the observed policy is the EOp policy. Thus n can be thought of as the extent to which the observed policy achieves EOp, relative to the benchmark of no transfers.

3. Data analysis

3.1 Data sets Our empirical analysis uses three micro survey data sets: (i) the Korea Labor and Income Panel Study (KLIPS), compiled by the Korea Institute of Labor Studies; (ii) the Keio Household Panel Survey (KHPS), compiled by Keio University of Japan; and (iii) the Pane Study of Family Dynamics (PSFD), compiled by Academia Sinica of the Republic of China (Taiwan). The three data sets contain the information about pre-fisc incomes of individuals and households, personal and household characteristics (age, years of schooling, the number of household members, etc.), and family backgrounds (such as father’s education). The years studied are 2003-2004 for Korea, 2004 for Japan, and 2003 for Taiwan. Thus the data set of Korea covers two years, while those of Japan and Taiwan cover only one year. In the case of Korea, we just pooled samples of the two adjacent years. Pooling samples over two adjacent years would remove year-specific and transitory variation of incomes. Our samples consist of individuals who are male household heads and 30-55 years old. They are individuals born approximately between 1950 and 1975. The sample sizes are 1591 (per year) for Korea, 1193 for Japan, and 1038 for Taiwan. Tax payments are surveyed in the Keio Household Panel Survey, but are not

12

surveyed in the other two data sets. To maintain consistency, we thus simulate tax payments in all countries, using tax codes of individual countries.5 Social benefits received are, on the other hand, surveyed in all of the data sets. Table 1 summarizes our data sets.

[Table 1 about here]

3.2 Incomes As in Roemer et al. (2003), we use two definitions of income: standard income (ST-income) and equivalence income (EQ-income). Pre-fisc standard income is the sum of the individual’s labor income and his/her household capital income per adult. Capital income is the sum of interest income, dividend income, rental income, and other financial income. Capital gains/losses, fringe benefits, imputed rents, and incomes from home production are not included; information on them is not available. Most individuals in our sample have no capital income; the median capital income in Korea is, for instance, zero. The average number of adults is 2.335 in Korea (as of 2004), 1.93 in Japan, and 1.98 in Taiwan. Pre-fisc equivalence income is the sum of household labor and capital income adjusted by the equivalence scale (the square root of the household size). Thus the EQ-income takes account of differences in household needs. The average household 5

The reported taxes in surveys are often inaccurate. The reported tax payments are very likely to under-estimate the taxes paid by high income groups and those paid by the selfemployed, whose tax evasion is widespread in East Asian countries. If we used reported tax payments, tax payments would be under-estimated and thus the extent to which actual tax systems achieve equalization of opportunities for income acquisition would probably be more than what our estimates would indicate. 13

size is 3.20 (as of 2004) in Korea, 3.36 in Japan, and 3.80 in Taiwan.6 Post-fisc incomes are calculated by adding cash transfers and social insurance benefits to, and subtracting income taxes and social insurance contributions (such as pensions, health insurance premiums, unemployment insurance premiums, etc.) from, pre-fisc incomes. All incomes are expressed per annum terms in local currencies; they are expressed in ten thousands of Korean Won, thousands of Japanese Yen, and thousands of New Taiwanese dollar. As of 2004, one Japanese Yen is approximately 10 Korean Wons and one Taiwanese dollar is about 34 Korean Wons. One US dollar is approximately 1145 Korean Wons, 108 Japanese Yens, and 33.4 Taiwanese dollars. (See Table A-1 for some macroeconomic statistics for the three countries at 2004.) Korean incomes are expressed in real terms using the consumer price index (with 2004=100). Table 2 reports summary statistics of pre- and post-fisc ST- and EQ-incomes for the three countries.

[Table 2 about here]

The mean ST-income is 27,958,700 Wons (about $24,410) for Korea (2004/05), 5,832,710 Yens (about $53,911) for Japan, and 609,840 NT dollars (about 6

For Korea, we include all household members. For Japan and Taiwan, on the other hand,

household members are restricted to a couple (if married) and their children (if they have children). We exclude parents and other household members, for the surveys do not provide detailed income information of each household member. Adults also include only a male head and his wife in Japan and Taiwan.

14

$18,242) for Taiwan. The ratio of median ST-income to mean ST-income is 0.88 for Korea (2004/05), 0.86 for Japan, and 0.82 for Taiwan. Income inequality, measured by the Gini coefficient, is the lowest in Japan (0.2793), followed by Korea (0.3728). Taiwan has the highest level of income inequality (0.4471) in terms of the Gini coefficient. Our data sets show that Taiwanese income ($18,242) is much lower than Korean income ($24,410), although actual GDP per capita of Korea and Taiwan is very close each other. According to Table A-1, for instance, real GDP per capita of Korea is $21,330.22 and that of Taiwan is $23,693; Taiwan’s real GDP per capita is slightly bigger than that of Korea. There are two reasons for this discrepancy. First, the GDPs in Table A-1 are adjusted by the Penn World Table’s international price index, whereas the numbers in Table 2 are not adjusted. Indeed, unadjusted GDP per capita (calculated from the World Development Indicators) is $15,243 for Korea and $14,985 for Taiwan in 2004; thus Taiwan’s unadjusted GDP per capita is slightly lower than that of Korea. Second, incomes in our data sets are those for male household heads. Female labor participation rate and female wages are higher in Taiwan than in Korea.7

3.3 Types Roemer et al. (2003) partition the entire sample into types using two typologies: one according to the level of education of the more highly educated parent and the other according to the parents’ occupation (farmers, unskilled manual 7

Lf Note that GDP per capita is equal to w Lm + w , where wi is gender i’s average m f Pop

Pop

wages and Li / Pop is gender i’s labor force participation rate. 15

workers, skilled manual, and professionals and self-employed). In this paper, we use only the former typology. We partition the sample into three types (Edu_Pa1 to Edu_Pa3) according to the level of education of father. Mother’s schooling years are surveyed, but using them reduces sample sizes significantly in Japan and Taiwan. In East Asian countries the more highly educated parent is usually a father. The probability that the level of father’s education is greater than or equal to the level of mother’s is 90.8% in Korea, 88.3% in Japan, and 84.6% in Taiwan. Information on whether one’s parent is a school dropout is available in Korea, while this information is not available in Japan and Taiwan. We use the school dropout information only in Korea. In contrast with many Western countries, the level of father’s education in these countries is very low for the individuals born between 1950 and 1975, reflecting the low level of economic development of the three countries in those years. Indeed many individuals in Korea and Taiwan have fathers who have no formal education at all. (Most of their fathers are born before 1945.) In Korea, Edu_Pa1 consists of those individuals whose father has schooling years less than 6 (no formal education or incomplete primary education), Edu_Pa2 consists of those whose father has primary and junior-high school education (6-9 years of schooling), and Edu_Pa3 consists of those whose father’s schooling years are greater than or equal to 10 (at least one year of high school education and above). In Taiwan, the three types are classified as follows: Edu_Pa1 consists of individuals whose father’s schooling years are less than 6 (no formal education or incomplete primary education), Edu_pa2 consists of individuals whose father’s schooling years are between 6 and 12 (primary school, junior high school, or high school education),

16

and Edu_Pa3 consists of individuals whose father’s schooling years are greater than 12 (more than high school education). In Japan, Edu_Pa1 consists of those individuals whose father has schooling years less than or equal to 11 (no formal education, primary school education, junior high school education, or high school dropouts), Edu_Pa2 consists of those individuals whose father has 12-15 years of schooling (high school education, 2-year junior and technical college education, and 4-year college dropouts), and Edu_Pa3 consists of those whose father’s schooling years are greater than or equal to 16 (college education or higher). The cutoff schooling year between Edu_Pa1 and Edu_Pa2 in Japan is very high, compared with those in Korea and Taiwan. This is mainly due to data availability. The KHPS does not provide detailed information about father’s education; it surveys the level of father’s education starting only from junior high school graduation. Even if we take into account the fact that Japan has always been more advanced than Korea and Taiwan at least for the last 150 years, we suspect that this cutoff level is somewhat high. Because of this high cutoff level, the individuals in the least advantaged type in Japan (Edu_Pa1) are, we conjecture, much better off than those in the least advantaged types in Korea and Taiwan. Our results in Japan must be interpreted with some caution. Descriptive statistics by type are reported in Table 3.

[Table 3 about here]

Table 3 clearly shows that mean pre-fisc incomes are quite different across types in all of the three countries. In other words, circumstances significantly affect

17

individual and household incomes in Korea, Japan, and Taiwan. The ratio of the average type-1 ST-income to the average type-3 ST-income is 0.73 in Korea (for 2004/05), 0.79 in Japan, and 0.52 in Taiwan. The same ratio for EQ-income is 0.67 in Korea, 0.87 in Japan, and 0.47 in Taiwan. One might wonder whether this difference reflects different labor market experience of the individuals across types. This is not the case. In our samples, individuals in Edu_Pa3 are somewhat younger than those in Edu_Pa1; the higher average income of the former group cannot be due to more labor market experience. Rather, because of this age discrepancy, our sample is more likely to underestimate the degree of unequal opportunities among the three types. That is, were our samples to contain Edu_Pa3 type individuals of the same age as Edu_Pa1 type individuals, observed income differences would be greater. Figure 1 plots empirical distribution functions of pre-fisc incomes for each of the three types. In figure 1, top panels correspond to ST-incomes and bottom panels correspond to EQ-incomes.

[Figures 1 about here]

In all of the three countries, distribution functions are distinctively different across types. Also the distribution function of type 1 is first-order stochastically dominated by the distribution function of the other two types. The importance of circumstances on individual achievement is shown also in Table 4, which reports on the conditional probabilities of education.

18

[Table 4 about here]

The probability that a type 1 individual will have the highest level of education is very low (18% in Korea, 24% in Japan, and 11% in Taiwan), while the probability that a type 3 individual will have the highest level of education is very high (71% in Korea, 75% in Japan, and 62% in Taiwan). Conversely, the probability that a type 1 individual will have low levels of education is very high and the probability that a type 3 individual will have low levels of education is very low. We also examined the relationship between parents’ education and respondent’s income. In Korea, the probability that a type 1 individual will belong to the first quartile of ST-income distribution is 32.8%, whereas the probability that the same individual will belong to the fourth quartile of the ST-income distribution is 18.22%. On the other hand, the probability that a type 3 individual will belong to the first quartile of ST-income distribution is 18.22%, while the probability that the same individual will belong to the fourth quartile of the same income distribution is 33.01%. Similar patterns are observed in Japan and Taiwan. (See Table A-3 for details.) Summarizing, parents’ education has a great impact on individual incomes in all of the three East Asian countries.

3.4 Observed tax functions In order to obtain estimates of the actual mapping of pre-fisc incomes into net taxes, we regress individuals’ net taxes on their pre-fisc incomes:

Net taxes = -a + b * (Pre-fisc income) .

19

(25)

The estimated value of a is the observed value of T and that of b is the observed value of the marginal income tax rate (t). In Roemer et al. (2003), t is the marginal ‘income’ tax rate. One might argue that indirect taxes are as important as income taxes in Korea, Japan, and Taiwan. We can easily incorporate indirect taxes into the model so that t measures the ‘effective’ marginal income tax rate. Suppose the total amount of transfer payments, T , is divided into ‘cash transfers’, T1 = gT , and ‘non-cash transfers,’ T2 = (1 - g )T , where g Î [0,1] is the proportion of cash transfers in total transfers.

Then given the marginal income tax rate tW , the disposable income is (1 - tW )x + T1 , and the post-fisc income is obtained by subtracting indirect taxes

paid from and adding non-cash transfers to the disposable income. Thus if d is the proportion of indirect taxes paid in the total disposable income, then we have: y = (1 - tW )x + T1 - d((1 - tW )x + T1 ) + T2 = (1 - tW - d + tW d )x + (1 - gd )T .

(26)

Thus the effective marginal tax rate is tW + d - tW d and the effective amount of transfer payments is (1 - gd )T . Alternatively speaking, each individual pays taxes of (tW + d - tW d )x + gdT and receives transfer payment of T . Because tW d is usually small, the effective marginal tax rate is close to t = tW + d . Table A-4 reports values of d and g as well as those of S, estimated using the National Income and Product Accounts and the Government Revenue Statistics. Table 5 reports observed affine tax functions, estimated according to

20

equations (25) and (26). The first row reports the tax rate in (25) and the second in (26). [Table 6 about here]

We find that fiscal policies in the three East Asian countries are somewhat progressive; Figure 3 clearly shows that estimated tax functions are convex. Coefficients for quadratic regression equations are also all statistically significant. Nonetheless the affine fit is very good. The R2 is greater than 0.8 in all countries, and the regression with the quadratic or cubic terms does not add much explanatory power. (See Table A-4.) If we eliminate some excessively high incomers, then the affine fit, the quadratic fit, and a non-parametric fit would almost coincide. The estimated marginal income tax rates are 22-23% (ST-income) and 20-23% (EQ-income) in Korea, 28% (ST-income) and 31% (EQ-income) in Japan, and 28% (ST-income) and 25% (EQ-income) in Taiwan. Recall that in the case of Korea, where several years are covered, we pooled samples of every two adjacent years; the time variation in estimated tax rates is due neither to varying sample sizes nor to sample units. It is well known that OLS regressions are not robust; OLS regression results are highly sensitive to a small number of outliers. Figure 3 shows that there are a small number of observations with very high income. To see whether our estimated tax rates are influenced by these observations, we ran median regressions as well. (See Table A-5.) Median regressions somewhat reduce the estimated tax rates. The estimated tax rates are 15.3-16.3% (ST-income) and 15.1-15.2% (EQ-income) in Korea, 24.1% (ST-income) and 25.9% (EQ-income) in Japan, and 20.0% (ST-

21

income) and 19.2% (EQ-income) in Taiwan. We conjecture that the actual tax rates lie somewhere between the OLS and the median regression estimates. For the sake of comparison with the results reported in Roemer et al. (2009), we choose the OLS regression estimates.

3.5. Estimation wages and labor supply from post-fisc income Assuming that the individual with median income work 1 unit of time, we estimate aobs as follows: hobs

æ (1 - t obs )x ö÷ çç med ÷ çç obs obs ÷ ÷÷ø + a h (1 1 / ) è

= 1  aobs =

hobs (1 - t obs )x med . obs 1+ h

Then, using equation (6), we estimate wages as: 1

hobs

æ ö÷ æ 1 ö1+hobs 1 - t obs ÷ wi = ççç obs x i ÷÷÷ , where K obs º ççç obs ÷ø çè a (1 + 1 / hobs )÷÷ø èK

.

Labor supply is estimated as follows: Li = Kobswi h = x i / wi .

3.6 Estimation of education equation Because older people are more likely to have a lower level of education, we assume:

Edui = g 0 + g1 agei + g 2 Edu _ Pai + ni . (-)

(+ )

To examine the effect of circumstances on education inequalities, we need to set qi = Edu _ Pai at a certain counterfactual level. One natural choice for the counterfactual level of qi = Edu _ Pai might be the mean level of Edu _ Pai of the entire sample (Bourguignon et al., 2007).

22

This choice would make the distribution of incomes independent of circumstances. But this choice is not free from problems.

This would assign good

circumstances to the people in the bad types, but give poor circumstances to the people in the good types. Giving poor circumstances to the people in the good types is not the aim of the equal opportunity policy. One way of improving circumstances of the people in the bad types without hurting those in good types is to give better circumstances to the people in bad types while not changing the circumstances of the people in good types. As an example, suppose q3 is the minimum level of parent’s education among the people in type 3 (in Korea, this is 10 years of schooling). Suppose we define

ìïq qi = ïí 3 ïïqi î

for individuals in types 1 and 2 for individuals in type 3

Thus we are asking: what would be the wage distribution if individuals in types 2 and 3 were born to parents whose schooling years are equal to the minimum level of the parent’s schooling years in type 3? We calculate with two different choices of q :

ìïq 3 Case 1: qi = ï í ïïqi î

for individuals in types 1 and 2 , where q3 is the for individuals in type 3

minimum level of parent’s education among the people in type 3 (in Korea, this is 10 years of schooling). Case 2: qi = qmax for all i , where qmax = 18 in Korea. In the actual estimation of equation (29), we use the logistic transformation of the dependent variable. This is because in a linear regression like equation (29),

23

 can be negative. Thus, instead of equation (29), we run the following logistic Edu regression: log(

Edui - Edumin Edumax - Edui

) = g0 + g1agei + g2Edu _Pai + ni .

(27)

We now deduce: w min + w max exp( d0 +  d 1agei +  d 2 (agei )2 +  d 3 qi + ei ) , wi = 1 + exp( d + d age +  d (age )2 +  d q + e ) 0

1

2

i

3 i

i

i

where bˆi is the estimated coefficient of equation (30), and ei is its residual. To prevent some samples from being dropped when wi = w min

or

wi = w max , we use w min - a and w max + a , where a is a small number, instead

of w min and w max in the logistic transformation. We set w min = 0 (not the minimum in the data) and w max as the largest wage in the data. The effect of circumstances on education inequalities is computed as

i ) G(Edui ) - G(Edu G(Edui )

,

 Edu i

where

is

calculated

from

Edui = g 0 + g1 agei + g 2 Edu _ Pai + ni .

3.6 Estimation of wage equations

Consider the following wage equations: wi = a0 + a1 agei + a2 (agei )2 + a3 Edui + a4 Edu _ Pai + ui . (+)

(-)

(+)

24

(+)

(28)

We expect a positive sign for a1 and a negative sign for a2

because there would

be the effect of experience on wages. We expect a positive sign for a3 because the effect of education on wages should be positive (due to improved human capital). Finally, the effect of father’s education on wages must be positive: a4 > 0 . Because older people are more likely to have a lower level of education, we assume Edui = g 0 + g1 agei + g 2 Edu _ Pai + ni . Thus the following reduced form (-)

(+)

equations might suffice for equation (8):

wi = (a0 + a3g0) +(a1+ a3 g1)agei + a2(agei )2 +(a4 + a3 g2 )Edu _Pai +(ui + a3ni ) (+)

(-) (+)

(-)

(+)

(+) (+)

= b0 + b1agei + b2(agei )2 + b3Edu _Pai + ei

(29)

In the actual estimation of equation (29), we use the logistic transformation of wages. This is because in a linear regression like equation (29), w can be negative, in which case Li would be undefined. Thus, instead of equation (29), we run the following logistic regression: w - wmin log( i ) = d0 + d1agei + d2(agei )2 + d3qi + ei . wmax - wi

(30)

We now deduce: wi =

w min + w max exp( d0 +  d 1agei +  d 2 (agei )2 +  d 3 qi + ei ) , 1 + exp( d + d age +  d (age )2 +  d q + e ) 0

1

i

2

i

3 i

i

where bˆi is the estimated coefficient of equation (30), and ei is its residual. To prevent some samples from being dropped when wi = w min

or

wi = w max , we use w min - a and w max + a , where a is a small number, instead

25

of w min and w max in the logistic transformation. We set w min = 0 (not the minimum in the data) and w max as the largest wage in the data. We calculate with two different choices of q :

ìïq 3 Case 1: qi = ï í ïïqi î

for individuals in types 1 and 2 , where q3 is the for individuals in type 3

minimum level of parent’s education among the people in type 3 (in Korea, this is 10 years of schooling). Case 2: qi = qmax for all i , where qmax = 18 in Korea. One issue is the age effect. Older people may have lower levels of schooling, due to underdevelopment of East Asian countries at the ages of schooling. To control for this effect, we compute: w min + w max exp( d0 +  d 1age i +  d 2 (age i )2 +  d 3qi + ei ) , w = 1 + exp( d + d age +  d (age )2 +  d q + e ) a i

0

wia =

1

2

i

i

3

i

i

w min + w max exp( d0 +  d 1age i +  d 2(age i )2 +  d 3 qi + ei ) , 1 + exp( d + d age +  d (age )2 +  d q + e ) 0

1

i

obs

Lai = K h

2

i

i

hobs

( ) wia

3 i

,

and hobs

( )

obs Lai = K h wia

Then

G(wiaLai ) - G(wiaLai ) G(wiaLai )

.

can be called the share of observed income

inequality due to unequal circumstances after controlling for the age affect, while

26

G(wiaLai ) G(wiaLai )

may be called the share of observed income inequality due to differential

effort after controlling for the age effect. Decomposition of

G(wiaLai ) - G(wiaLai ) G(wiaLai )

into the two channels can be done in like manner.

The effect of circumstances on wage inequalities is computed as G (wi ) - G (wi ) . This must be close in number to the size of channel (a). (But it does G (wi )

not have to be identical.)

3.5 EOp tax rates We now estimate the EOp tax rates using equation (22). For that, there appear to remain two parameters to be estimated: h and a . Note that parameter h is the elasticity of labor supply with respect to the wage. (Recall equation 오류! 참조 원본을 찾을 수 없습니다..) Accurate estimation of the elasticity of labor supply is extremely difficult; estimated elasticities greatly vary depending upon the methods and the data sets employed. As in Roemer et al. (2003), we thus choose, rather than estimate, the parameter values of

h between 0.03 and 0.09. The benchmark value of h is 0.06. Roemer et al. (2003) estimate the parameter value of a by assuming that the hobs

æ (1 - t obs )x ö÷ med ÷÷ individual with median income work 1 unit of time: ççç obs çè a (1 + 1 / hobs )÷ø÷

Using the estimated values of aobs ,

= 1.

and t obs , they then deduce the distribution of

27

wages from the distribution of pre-fisc incomes. We argue that this step is unnecessary for the calculation of the EOp tax rate. Although A and B in equation (22) depend upon the estimated values of aobs and t obs , the optimal EOp tax rate is independent of the estimated values of these parameters. hobs

To see this, define K obs

æ ö÷ 1 - t obs ÷ º ççç obs obs ÷ èç a (1 + 1 / h )÷ø

. Roemer et al. (2003)

1 x . Thus estimated wages K obs

calculate wages from the following equation: w 1+h =

depend on the estimated values of aobs and t obs . But if we denote the distribution function of x by F(.) , then

A=

ò

¥ 0

w 1+hdF 1(w ) =

1 K obs

w 1+hdF (w ) =

1 K obs

ò

¥

0

xd F1(x ) ,

(31)

xd F(x ) .

(32)

and

B=

ò

0

¥

ò

0

¥

Thus, although the estimated values of A and B depend on the estimated values of aobs , and t obs , their ratio, A/B, is independent of them. Because the optimal tax rate depends only upon A/B and h , the optimal tax rate is also independent of the estimated values of aobs and t obs . In like manner, one can easily show that the estimated value of t Bench does not depend upon the estimated value of aobs , although it depends upon the estimated values of t obs and hobs . Tables 7 and 8 report the calculated EOp tax rates, as well as other statistics, for three chosen values of h : h = 0.03 , 0.06, and 0.09. For the sake of comparison, we also present the estimates for a few Western countries, reported in Roemer et al. (2003).

28

[Tables 7 and 8 about here]

The EOp tax rates are greater than observed tax rates in Korea and Taiwan. With the assumption of h = 0.06 , the EOp ST-income tax rates are 61.3-64.0% in Korea and 80.9% in Taiwan. The EOp EQ-income tax rates are, on the other hand, 67.7-69.2% in Korea and 83.1% in Taiwan. The levels of the EOp tax rates in Korea are similar to those of Spain (60.5% for ST-incomes and 55.6% for EQ-incomes), Italy (81.9% for ST-incomes and 82.9% for EQ-incomes), and the United States (64.7% for ST-incomes and not available for EQ-incomes). Out of the eleven countries studied in Roemer et al. (2003), they are the three countries that have the highest EOp tax rates and the lowest level of equal opportunity. Our analysis thus indicates that opportunities in Korea and Taiwan are as unequal as those in the three countries. The observed marginal tax rates in Korea and Taiwan are, however, about the same as or smaller than those in the three Western countries. In the case of STincome, Korea has the lowest marginal tax rate (22.2-22.7%), followed by Italy (23.2%), the United States (24.3%), Taiwan (26.7%), and Spain (37.6%). This implies that tax-benefit policies in Korea and Taiwan have played very little role in correcting unequal opportunities for income acquisition among their citizens. Japan is somewhat different from the other two East Asian countries. In terms of ST- and EQ-incomes, Japan’s EOp tax rates are lower than its observed tax rates for both h = 0.06 and h = 0.09 . Japan is certainly an egalitarian country, but this result is largely an artifact of an unusually high level of the cutoff schooling years

29

that defines the least advantaged type in Japan. Our assertion that Japan’s cutoff schooling year between Edu_Pa1 and Edu_Pa2 is high can be confirmed from two sources. First, the ratio of the average type 1 income to the average income is very high in Japan. The ratio of the average type 1 ST-income to the average ST-income is 0.84 in Korea (for 2004/05) and 0.71 in Taiwan, but is 0.93 in Japan. The same ratio for EQ-income is 0.82 in Korea and 0.68 in Taiwan, and 0.96 in Japan. Indeed, type 1 and type 2 distributions in Japan are almost identical according to Figure 2. Second, the fraction of individuals in Edu_Pa1 is 0.22 in Korea and 0.28 in Taiwan, but is is close to 0.5 (0.49) in Japan. Thus the results obtained from Japan are somewhat an artifact of the data availability, rather than the reflection of real equal opportunity in Japan.

4. Conclusion

Employing the method of Roemer et al. (2003), we estimated the extent to which tax-benefit policies of Korea, Japan, and Taiwan equalize opportunities among citizens for the acquisition of income. As in Roemer et al. (2003), we proceeded by singling out one obvious circumstance (i.e., parents’ education), and attributing all remaining variation in incomes to differential effort. We find that opportunities in Korea and Taiwan are as unequal as those in Spain, Italy, and the United States, but the current fiscal policies in these East Asian countries have much smaller opportunity equalizing effects than the three Western countries. We make two final remarks

30

First, it is well known that wage schemes in East Asian countries are largely determined by seniority, and wages across different occupations and effort levels are highly compressed. Such a highly compressed wage structure is one of the main reasons for why inequality is relatively low in East Asian countries. On the other hand, our analysis indicates that opportunities in these countries are not so equal. Thus our analysis implies that relatively equal distribution of outcomes in East Asian countries is not due to small variation in outcomes across types, but perhaps due to small variation in outcomes across effort levels in each type. Second, one might argue that our exercise defines the feasible set of policies as affine taxation which is revenue neutral, with respect to the funding of nontransfer payment government spending. Much of that spending will also have an equal-opportunity effect, such as money spent on education and health, and we have not attempted to estimate that effect. We can, however, observe the relative magnitudes of this spending by considering the benchmark situation, in which there are no cash transfers. The benchmark tax rate is in the range of 6.5%-6.7% in Korea, 6.2% in Japan, and 6.1% in Taiwan, and these benchmark tax rates are again no higher than those in the three countries. Thus potential equal-opportunity effects from general government services would not be large in East Asian countries. Due to data availability, our calculation in this paper is limited to only three East Asian countries. Studying the subject with a more comprehensive set of East Asian countries and comparing the results with Latin American countries is left for future research.

31

References

Aaberge, R., U. Colombino, and J. Roemer. 2001. “Equality of opportunity versus equality of outcome in evaluating income redistribution policies: Empirical evidence based on Italian data.” mimeo. Arneson, R. 1989. “Equality and equality of opportunity for welfare.” Philosophical Studies 56, 77-93. Björklund, A., M. Jäntti, and J. Roemer, 2010, “Equality of opportunity and the distribution of long-run income in Sweden,” mimeo. Bourguignon, F., F. H. G. Ferreira, M. Menéndez. 2007. “Inequality of opportunity in Brazil.” Review of Income and Wealth 53 (4), 585-618. Cohen, G. 1989. “On the currency of egalitarian justice.” Ethics 99, 906-944. Dardanoni, V., G. Fields., J. Roemer, and M. Puerta. 2005. “How demanding should equality of opportunity be, and how much have we achieved?” in S. Morgan, D. Grusky, and G. Fields eds., Mobility and Inequality: Frontiers of Research in Sociology and Economics, Stanford University Press: Stanford, CA Dworkin, R., 1981a. “What is equality? Part I: Equality of welfare.” Philosophy and Public Affairs 10, 185-246. Dworkin, R., 1981b. “What is equality? Part II: Equality of resources.” Philosophy and Public Affairs 10, 283-345. Kim, W. and W. Lee, 2009, “Roemer's conception of equal opportunity and Korea's income tax-benefit policy,” Quarterly Economic Analysis 15 (3), 129-168 [written in Korean]. Jusot, F., S. Tubeuf, and A. Trannoy, “Effort or circumstances: Which one

32

matters in health inequality?” mimeo. Lefranc, Arnaud, Nicolas Pistolesi, and Alain Trannoy, 2008. “Inequality of Opportunities vs. Inequality of Outcomes: Are Western Societies All Alike?” Review of Income and Wealth 54 (4), 513-546. Lefranc, A., N. Pistolesi, and A. Trannoy. 2009. “Equality of opportunity: Definitions and testable conditions, with an application to income in France.” Journal of Public Economics 93, 1189-2007. Peragine, V., Roemer, J. 1993. “A pragmatic theory of responsibility for the egalitarian planner.” Philosophy and Public Affairs 10, 146-166. Roemer, J. 1998. Equality of Opportunity. Cambridge, MA: Harvard University Press. Roemer, J. et al. 2003. “To what extent do fiscal regimes equalize opportunities for income acquisition among citizens?” Journal of Public Economics 87, 539-565. World Bank (Latin America and the Caribbean Region). 2008. Measuring Inequality of Opportunity in Latin America and Caribbean, Volume 1: Main Report, Washington, DC: World Bank

33

Table 1: Summary of household surveys used

Survey name

Years studied Selected samples Sample size per year Taxes

Korea Korea Labor and Income Panel Study (KLIPS) 2003/2004 Male household heads aged between 30-55 1501 Simulated

1

Japan Keio Household Panel Survey (KHPS)

Taiwan Panel Study of Family Dynamics (PSFD)

2004 Male household heads aged between 30-55 1193 Simulated

2003 Male household heads aged between 30-55 1038 Simulated

Table 2: Summary statistics

Pre-fisc ST income Pre-fisc EQ income Post-fisc ST income Post-fisc EQ income Obs.

Median Mean Gini Median Mean Gini Median Mean Gini Median Mean Gini

Korea

Japan

Taiwan

2466.09 2795.87 0.3728 1908.54 2229.88 0.3615 2301.60 2471.75 0.3489 1761.77 1991.38 0.3381 3002

5020.00 5832.71 0.2739 3464.10 3914.55 0.2793 4395.04 4766.86 0.2520 2801.09 3086.14 0.2557 1193

498.00 609.84 0.44707 346.41 446.28 0.44711 453.4562 530.0337 0.42054 304.3423 379.3054 0.43456 1038

Source: Korea Labor and Income Panel Study (Korea), Keio Household Panel Survey (Japan), and Panel Study of Family Dynamics (Taiwan) Notes: (1) Pre-fisc ST income is the sum of the individual’s labor income and his household capital income per adult. Post-fisc ST income is pre-fisc ST income plus social benefits from the government, minus tax payments. (2) Pre-fisc EQ income is the sum of household labor and capital income, with the sum adjusted by the equivalence scale (the square root of household size). Post-fisc EQ income is pre-fisc household income plus social benefits minus tax payments, with the total being divided by the equivalence scale. (3) Korean income is expressed in real terms (2004=100) and in ten thousands of Korean Won. (4) Japanese income is expressed in thousands of Japanese Yen. (5) Taiwanese income is expressed in thousands of NT$.

2

Table 3: Summary statistics, by type

Country Korea

Japan

Taiwan

type Edu_Pa1 Edu_Pa2 Edu_Pa3 Edu_Pa1 Edu_Pa2 Edu_Pa3 Edu_Pa1 Edu_Pa2 Edu_Pa3

Pre-fisc ST income (mean)

Pre-fisc EQ income (mean)

2394 2815 3183 5419 6030 6857 429 600 830

1897 2201 2647 3747 4002 4306 297 436 632

Post-fisc ST income (mean) 2138 2489 2791 4466 4904 5530 384 526 701

Post-fisc EQ income (mean) 1718 1968 2333 2990 3130 3332 256 372 531

Age (mean) 44.4 40.7 39.1 45.6 42.3 40.8 47.5 44.3 43.1

School year (mean)

Obs.

11.5 13.1 14.7 12.7 13.8 15.4 9.1 11.7 13.8

758 1534 710 585 460 148 289 490 259

Source: Korea Labor and Income Panel Study (Korea), Keio Household Panel Survey (Japan), and Panel Study of Family Dynamics (Taiwan) Notes: (1) Pre-fisc ST income is the sum of the individual’s labor income and his household capital income per adult. Post-fisc ST income is pre-fisc ST income plus social benefits from the government, minus tax payments. (2) Pre-fisc EQ income is the sum of household labor and capital income, with the sum adjusted by the equivalence scale (the square root of household size). Post-fisc EQ income is pre-fisc household income plus social benefits minus tax payments, with the total being divided by the equivalence scale. (3) Korean income is expressed in real terms (2004=100) and in ten thousands of Korean Won. (4) Japanese income is expressed in thousands of Japanese Yen. (5) Taiwanese income is expressed in thousands of NT$. (6) Using the level of father’s education, we classify the three types as follows: Edu_Pa1 Edu_Pa2 Edu_Pa3 Korea 0~5 6~9 ³ 10 Japan 0~11 12~15 ³ 16 Taiwan 0~5 6~12 ³ 13

3

Table 4: Conditional probabilities of education

Korea Father’s schooling years All

0-5 yrs. 6-9 yrs. 10 & higher

Japan Father’s schooling years All

0-11 yrs. 12-15 yrs. 16 & higher

Taiwan Father’s schooling years All

0-5 yrs. 6-12 yrs. 13 and higher

Respondent’s schooling years 0-9 yrs. 10-12 yrs. 0.314 0.464 0.108 0.479 0.017 0.270 0.139 0.426

13 and higher 0.222 0.413 0.713 0.436

sum 1.000 1.000 1.000 1.000

Respondent’s schooling years 0-9 yrs. 10-15 yrs 0.121 0.641 0.068 0.503 0.007 0.241 0.086 0.538

16 & higher 0.238 0.429 0.752 0.376

sum 1.000 1.000 1.000 1.000

Respondent’s schooling years 0-9 yrs. 10-12 yrs. 0.664 0.221 0.284 0.418 0.081 0.297 0.339 0.333

13 and higher 0.114 0.298 0.622 0.328

sum 1.000 1.000 1.000 1.000

Source: Korea Labor and Income Panel Study (Korea), Keio Household Panel Survey (Japan), and Panel Study of Family Dynamics (Taiwan)

4

Table 5: Estimated affine income tax functions

ST income Country Korea Japan Taiwan

t 0.210 0.246 0.279 0.311 0.267 0.321

T 262.33 252.63 601.79 584.35 78.27 75.65

t 0.218 0.253 0.313 0.344 0.246 0.301

T 246.52 237.40 444.10 431.23 46.91 45.34

R2 0.863 0.872 0.834

EQ income Country Korea Japan Taiwan

R2 0.819 0.843 0.835

Notes: (1) We compute the observed income tax function by running the following regression: Net taxes = t * x -T , where x is the pre-fisc income. In calculating net taxes, we add payroll taxes and subtract social benefits. Such estimates are reported in the first low. The second row estimates are by equation (18). (2) Standard income is the sum of the individual’s labor income and his household capital income per adult. (3) Equivalence income is the sum of household labor and capital income adjusted by the equivalence scale (the square root of household size). (4) Korean income and net taxes are expressed in ten thousands of Korean Won. It is expressed in real terms using the consumer price index (2004=100) . (5) Japanese income and net taxes are expressed in thousands of Japanese Yen. (6) Taiwan income and net taxes are expressed in thousands of NT$.

5

Table 6: Regression results: Korea Dependent variables are not transformed. Edu Cons Age

15.41*** (48.17) -0.0924*** (-13.21)

Age^2 Edu Edu_dad Obs.

0.219*** (20.63) 3002

Wage from ST incomes -6082.3*** (-5.22) 251.3*** (4.40) -2.461*** (-3.56) 220.6*** (17.57) 19.27* (2.51) 2747

-4123.0*** (-3.37) 303.6*** (5.05) -3.339*** (-4.59)

66.45*** (8.74) 2747

Wage from EQ incomes -2989.2** (-3.02) 130.6** (2.69) -1.403* (-2.39) 170.3*** (15.97) 18.61** (2.84) 2853

-1486.6 (-1.45) 171.6*** (3.40) -2.092*** (-3.43)

55.21*** (8.62) 2853

Dependent variables are logistically transformed. 　Edu Cons Age

15.41*** (48.17) -0.0924*** (-13.21)

Age^2 Edu Edu_dad Obs.

0.219*** (20.63) 3002

(Wage) from ST incomes

-5.621*** (-9.16) 0.110*** (3.64) -0.00119** (-3.27) 0.0856*** (12.94) 0.00405 (1.00) 2747

-4.861*** (-7.73) 0.130*** (4.19) -0.00153*** (-4.10)

0.0224*** (5.72) 2747

6

(Wage) from EQ incomes

-4.971*** (-9.05) 0.0594* (2.21) -0.000683* (-2.10) 0.0726*** (12.27) 0.0128*** (3.52) 2853

-4.331*** (-7.72) 0.0769** (2.79) -0.000977** (-2.93)

0.0284*** (8.13) 2853

Table 7: Education inequality due to differences in father’s education Case 1: Country

Korea

Japan

Taiwan

Index

G(Edu)

) G(Edu

Total

GINI

0.1192

0.0999

0.1622

GE0

0.0479

0.0369

0.2308

GE1

0.0267

0.0178

0.3345

GINI

0.0920

0.0347

0.6225

GE0

0.0155

0.0027

0.8273

GE1

0.0151

0.0025

0.8327

GINI

0.1688

0.1042

0.3828

GE0

0.1808

0.1364

0.2455

GE1

0.0533

0.0250

0.5318

Index

G(Edu)

) G(Edu

Total

GINI

0.1192

0.0759

0.3630

GE0

0.0479

0.0289

0.3960

GE1

0.0267

0.0109

0.5904

GINI

0.0920

0.0209

0.7728

GE0

0.0155

0.0010

0.9358

GE1

0.0151

0.0010

0.9367

GINI

0.1688

0.0722

0.5723

GE0

0.1808

0.1229

0.3201

GE1

0.0533

0.0162

0.6967

Case 2: Country

Korea

Japan

Taiwan

7

Table 8: Education inequality due to differences in father’s education: Age fixed Case1: Country

Korea

Japan

Taiwan

Index

G(Edu

a

)

a ) G(Edu

Total

GINI

0.1180

0.0978

0.1716

GE0

0.0452

0.0355

0.2135

GE1

0.0245

0.0167

0.3184

GINI

0.0929

0.0351

0.6226

GE0

0.0155

0.0027

0.8245

GE1

0.0151

0.0026

0.8301

GINI

0.1672

0.1072

0.3589

GE0

0.1687

0.1328

0.2128

GE1

0.0485

0.0249

0.4855

a ) G(Edu

Total

Case2: Country

Korea

Japan

Taiwan

Index

G(Edu

a

)

GINI

0.1180

0.0745

0.3688

GE0

0.0452

0.0282

0.3764

GE1

0.0245

0.0103

0.5806

GINI

0.0929

0.0211

0.7728

GE0

0.0155

0.0010

0.9347

GE1

0.0151

0.0010

0.9356

GINI

0.1672

0.0729

0.5638

GE0

0.1687

0.1192

0.2931

GE1

0.0485

0.0157

0.6754

8

Table 9: Pre-fisc income inequality due to differences in father’s education: Case 1 Wages are calculated from ST incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(wL)

 ) G(wL

G (wL)

 ) G (wL

GINI

0.3145

0.3080

0.3140

0.3077

0.0217 0.0205 (94.38) 0.0012 (5.62)

GE0

0.2035

0.1956

0.2029

0.1953

0.0401 0.0378 (94.35) 0.0023 (5.65)

GE1

0.1748

0.1663

0.1742

0.1659

0.0509 0.0481 (94.36) 0.0029 (5.64)

GINI

0.2714

0.2576

0.2703

0.2571

0.0528 0.0498 (94.43) 0.0029 (5.57)

GE0

0.1548

0.1418

0.1538

0.1414

0.0870 0.0821 (94.32) 0.0049 (5.68)

GE1

0.1264

0.1123

0.1253

0.1117

0.1161

GINI

0.4077

0.3737

0.4046

0.3731

0.0849 0.0804 (94.64) 0.0045 (5.36)

GE0

0.3412

0.2864

0.3362

0.2851

0.1645 0.1552 (94.35) 0.0093 (5.65)

GE1

0.3241

0.2606

0.3188

0.2586

0.2022 0.1909 (94.39)

Total

Channel A

Channel B

0.1095 (94.30) 0.0066 (5.70)

0.0114

(5.61)

Wages are calculated from EQ incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(wL)

 ) G(wL

G (wL)

 ) G (wL

GINI

0.3281

0.3205

0.3275

0.3203

0.0238 0.0225 (94.44) 0.0013 (5.56)

GE0

0.1999

0.1906

0.1991

0.1904

0.0479 0.0452 (94.37) 0.0027 (5.63)

GE1

0.1938

0.1833

0.1930

0.1829

0.0564 0.0533 (94.39) 0.0032 (5.61)

GINI

0.2793

0.2724

0.2788

0.2721

0.0257 0.0243 (94.34) 0.0015 (5.66)

GE0

0.1503

0.1439

0.1498

0.1437

0.0438 0.0413 (94.32) 0.0025 (5.68)

GE1

0.1377

0.1290

0.1370

0.1287

0.0654 0.0617 (94.32) 0.0037 (5.68)

GINI

0.4180

0.3808

0.4144

0.3803

0.0900 0.0851 (94.57) 0.0049 (5.43)

GE0

0.3307

0.2706

0.3248

0.2696

0.1850 0.1745 (94.35) 0.0105 (5.65)

GE1

0.3307

0.2622

0.3246

0.2604

0.2125 0.2006 (94.38)

9

Total

Channel A

Channel B

0.0119

(5.62)

Table 10: Pre-fisc income inequality due to differences in father’s education:Case 2 Wages are calculated from ST incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(wL)

 ) G(wL

G (wL)

 ) G (wL

GINI

0.3145

0.2997

0.3134

0.2990

0.0492 0.0465 (94.38) 0.0028 (5.62)

GE0

0.2035

0.1867

0.2022

0.1859

0.0862 0.0813 (94.34) 0.0049 (5.66)

GE1

0.1748

0.1559

0.1734

0.1550

0.1130

GINI

0.2714

0.2526

0.2700

0.2518

0.0722 0.0682 (94.42) 0.0040 (5.58)

GE0

0.1548

0.1375

0.1535

0.1367

0.1168

GE1

0.1264

0.1077

0.1249

0.1069

0.1542 0.1455 (94.31) 0.0088 (5.69)

GINI

0.4077

0.3494

0.4032

0.3473

0.1482 0.1401 (94.54) 0.0081 (5.46)

GE0

0.3412

0.2566

0.3342

0.2536

0.2568 0.2422 (94.33) 0.0146 (5.67)

GE1

0.3241

0.2205

0.3157

0.2167

0.3314 0.3126 (94.32) 0.0188 (5.68)

Total

Channel A

Channel B

0.1066 (94.34) 0.0064 (5.66)

0.1101

(94.32) 0.0066 (5.68)

Wages are calculated from EQ incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(wL)

 ) G(wL

G (wL)

 ) G (wL

GINI

0.3281

0.3127

0.3269

0.3122

0.0485 0.0458 (94.41) 0.0027 (5.59)

GE0

0.1999

0.1818

0.1984

0.1812

0.0937 0.0884 (94.34) 0.0053 (5.66)

GE1

0.1938

0.1717

0.1921

0.1709

0.1184

GINI

0.2793

0.2696

0.2786

0.2692

0.0361 0.0341 (94.35) 0.0020 (5.65)

GE0

0.1503

0.1414

0.1496

0.1410

0.0618 0.0582 (94.33) 0.0035 (5.67)

GE1

0.1377

0.1258

0.1368

0.1252

0.0903 0.0851 (94.32) 0.0051 (5.68)

GINI

0.4180

0.3582

0.4131

0.3563

0.1475 0.1394 (94.53) 0.0081 (5.47)

GE0

0.3307

0.2423

0.3231

0.2395

0.2757 0.2601 (94.32) 0.0157 (5.68)

GE1

0.3307

0.2250

0.3218

0.2215

0.3302

10

Total

Channel A

0.1117

0.3114

Channel B

(94.34) 0.0067 (5.66)

(94.31) 0.0188 (5.69)

Table 11: Pre-fisc income inequality due to differences in father’s education: Age fixed & Case 1 Wages are calculated from ST incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G (w L )

a a

G (w L )

a a

a a G (w L )

a a G (w L )

GINI

0.3088

0.3017

0.3082

0.3014

0.0239 0.0226 (94.37) 0.0013 (5.63)

GE0

0.1976

0.1894

0.1970

0.1891

0.0432 0.0408 (94.35) 0.0024 (5.65)

GE1

0.1677

0.1592

0.1671

0.1588

0.0530 0.0500 (94.36) 0.0030 (5.64)

GINI

0.2562

0.2407

0.2550

0.2401

0.0627 0.0592 (94.38) 0.0035 (5.62)

GE0

0.1431

0.1292

0.1420

0.1286

0.1011

GE1

0.1132

0.0991

0.1121

0.0986

0.1295 0.1221 (94.31) 0.0074 (5.69)

GINI

0.4010

0.3681

0.3979

0.3675

0.0835 0.0789 (94.50) 0.0046 (5.50)

GE0

0.3271

0.2743

0.3222

0.2731

0.1650 0.1556 (94.35) 0.0093 (5.65)

GE1

0.3077

0.2476

0.3026

0.2457

0.2015 0.1901 (94.39)

Total

Channel A

Channel B

0.0953 (94.32) 0.0057 (5.68)

0.0113

(5.61)

Wages are calculated from EQ incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

a a

G (w L )

a a

G (w L )

a a G (w L )

a a G (w L )

Total

Channel A

Channel B

GINI

0.3257

0.3186

0.3251

0.3184 0.0225 0.0212 (94.43) 0.0013 (5.57)

GE0

0.1965

0.1878

0.1957

0.1875 0.0454 0.0429 (94.37) 0.0026 (5.63)

GE1

0.1901

0.1802

0.1893

0.1799 0.0535 0.0505 (94.39) 0.0030 (5.61)

GINI

0.2648

0.2558

0.2641

0.2554

0.0356 0.0336 (94.34) 0.0020 (5.66)

GE0

0.1387

0.1306

0.1381

0.1303

0.0608 0.0573 (94.33) 0.0034 (5.67)

GE1

0.1251

0.1150

0.1244

0.1145

0.0846 0.0798 (94.32) 0.0048 (5.68)

GINI

0.4159

0.3785

0.4123

0.3780

0.0911

GE0

0.3278

0.2675

0.3219

0.2664

0.1873 0.1767 (94.35) 0.0106 (5.65)

GE1

0.3256

0.2579

0.3195

0.2562

0.2130 0.2010 (94.38) 0.0120 (5.62)

11

0.0861 (94.57) 0.0049 (5.43)

Table 12: Pre-fisc income inequality by birth group: Case 1, Korea Wages are calculated from ST incomes with h =0.06. Year

1950-1955

1956-1960

1961-1965

1966-1970

1971-1975

Index

G(wL)

 ) G(wL

G (wL)

 ) G (wL

GINI

0.3874

0.3740

0.3865

0.3734 0.0361 0.0341 (94.36)

0.0020 (5.64)

GE0

0.3027

0.2829

0.3013

0.2820 0.0684 0.0645 (94.35)

0.0039 (5.65)

GE1

0.2611

0.2401

0.2597

0.2391 0.0845 0.0797 (94.35)

0.0048 (5.65)

GINI

0.3365

0.3220

0.3354

0.3213 0.0449 0.0424 (94.40)

0.0025 (5.60)

GE0

0.2661

0.2515

0.2650

0.2509 0.0572 0.0539 (94.36)

0.0032 (5.64)

GE1

0.2010

0.1836

0.1998

0.1828 0.0907 0.0856 (94.37)

0.0051 (5.63)

GINI

0.2912

0.2837

0.2907

0.2834 0.0271 0.0255 (94.38)

0.0015 (5.62)

GE0

0.1583

0.1513

0.1578

0.1510 0.0460 0.0434 (94.35)

0.0026 (5.65)

GE1

0.1435

0.1354

0.1429

0.1351 0.0586 0.0553 (94.36)

0.0033 (5.64)

GINI

0.2711

0.2671

0.2708

0.2669 0.0157 0.0148 (94.39)

0.0009 (5.61)

GE0

0.1440

0.1404

0.1438

0.1402 0.0265 0.0250 (94.35)

0.0015 (5.65)

GE1

0.1321

0.1277

0.1318

0.1275 0.0345 0.0325 (94.35)

0.0019 (5.65)

GINI

0.2332

0.2318

0.2331

0.2318 0.0062 0.0058 (94.38)

0.0003 (5.62)

GE0

0.1119

0.1105

0.1117

0.1104 0.0129 0.0121 (94.38)

0.0007 (5.62)

GE1

0.0940

0.0928

0.0938

0.0928 0.0120

0.0113 (94.43)

0.0007 (5.57)

Channel A

Channel B

Total

Channel A

Channel B

Wages are calculated from EQ incomes with h =0.06. Year

1950-1955

1956-1960

1961-1965

1966-1970

1971-1975

Index

G(wL)

 ) G(wL

G (wL)

 ) G (wL

GINI

0.3750

0.3600

0.3735

0.3598 0.0406 0.0383 (94.44) 0.0023 (5.56)

GE0

0.2642

0.2428

0.2622

0.2423 0.0829 0.0783 (94.35) 0.0047 (5.65)

GE1

0.2660

0.2378

0.2636

0.2369 0.1094 0.1032 (94.37) 0.0062 (5.63)

GINI

0.3323

0.3250

0.3317

0.3248 0.0228 0.0215 (94.43) 0.0013 (5.57)

GE0

0.2048

0.1968

0.2041

0.1966 0.0402 0.0379 (94.36) 0.0023 (5.64)

GE1

0.1893

0.1807

0.1885

0.1804 0.0468 0.0442 (94.39) 0.0026 (5.61)

GINI

0.3247

0.3140

0.3239

0.3136 0.0342 0.0323 (94.41) 0.0019 (5.59)

GE0

0.1895

0.1787

0.1887

0.1783 0.0592 0.0559 (94.36) 0.0033 (5.64)

GE1

0.1816

0.1686

0.1806

0.1681 0.0745 0.0703 (94.38) 0.0042 (5.62)

GINI

0.3008

0.2957

0.3004

0.2955 0.0177 0.0167 (94.39) 0.0010 (5.61)

GE0

0.1632

0.1577

0.1627

0.1575 0.0346 0.0327 (94.36) 0.0020 (5.64)

GE1

0.1703

0.1631

0.1698

0.1627 0.0444 0.0419 (94.36) 0.0025 (5.64)

GINI

0.2685

0.2677

0.2684

0.2678 0.0029 0.0027 (94.61) 0.0002 (5.39)

GE0

0.1402

0.1393

0.1401

0.1393 0.0059 0.0056 (94.41) 0.0003 (5.59)

GE1

0.1229

0.1223

0.1228

0.1223 0.0050 0.0047 (94.51) 0.0003 (5.49)

12

Total

Table 13: Post-fisc income inequality due to differences in father’s education: Case 1 Wages are calculated from post-fisc ST incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(y )

G(y)

Total

GINI

0.2889

0.2835

0.0187

GE0

0.1763

0.1703

0.0339

GE1

0.1454

0.1389

0.0446

GINI

0.2173

0.2056

0.0537

GE0

0.0973

0.0866

0.1106

GE1

0.0800

0.0709

0.1138

GINI

0.3401

0.3227

0.0513

GE0

0.2572

0.2246

0.1268

GE1

0.2040

0.1794

0.1208

Wages are calculated from post-fisc EQ incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(y )

G(y)

Total

GINI

0.3057

0.3001

0.0183

GE0

0.1736

0.1670

0.0377

GE1

0.1639

0.1564

0.0454

GINI

0.2321

0.2275

0.0200

GE0

0.0992

0.0937

0.0556

GE1

0.0944

0.0884

0.0635

GINI

0.3742

0.3249

0.1316

GE0

0.2960

0.2063

0.3030

GE1

0.2444

0.1760

0.2796

Note: y º wL - Net Taxes is the actual post-fisc income, and   - Net Taxes is the counterfactual post-fisc income. y º wL

13

Table 14: Post-fisc Income inequality due to differences in father’s education: Case 2 Wages are calculated from post-fisc ST incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(y )

G(y)

Total

GINI

0.2889

0.2776

0.0392

GE0

0.1763

0.1644

0.0677

GE1

0.1454

0.1319

0.0928

GINI

0.2173

0.1994

0.0824

GE0

0.0973

0.0820

0.1573

GE1

0.0800

0.0668

0.1649

GINI

0.3401

0.3227

0.0513

GE0

0.2572

0.2246

0.1268

GE1

0.2040

0.1794

0.1208

Wages are calculated from post-fisc EQ incomes with h =0.06. Country

Korea

Japan

Taiwan

Index

G(y )

G(y)

Total

GINI

0.3057

0.2953

0.0338

GE0

0.1736

0.1621

0.0659

GE1

0.1639

0.1494

0.0881

GINI

0.2321

0.2235

0.0371

GE0

0.0992

0.0902

0.0908

GE1

0.0944

0.0848

0.1010

GINI

0.3742

0.3249

0.1316

GE0

0.2960

0.2063

0.3030

GE1

0.2444

0.1760

0.2796

Note: y º wL - Net Taxes is the actual post-fisc income, and   - Net Taxes is the counterfactual post-fisc income. y º wL

14

Table 15: EOp policy ST income, h = 0.06 Country

year

t obs

t EOp

T EOp

t Bench

n

Korea

2003/2004

0.210

0.6062

1436.62

0.0669

0.3929

Japan

2004

0.279

0.2020

819.92

0.0617

0.6442

Taiwan

2003

0.267

0.8091

417.18

0.0614

0.3643

Spain

1991

0.376

0.605

663.9

0.080

0.748

Italy

1993

0.232

0.819

21.3

0.156

0.160

USA

1991

0.243

0.647

13578.0

0.182

0.200

Belgium

1992

0.531

0.535

158.0

0.316

0.999

Sweden

1991

0.524

0

-30207.0

0.203

overtax

EQ income, h = 0.06 Country

year

t obs

t EOp

T EOp

t Bench

n

Korea

2003/2004

0.218

0.6208

1136.63

0.0839

0.3702

Japan

2004

0.313

0

-365.30

0.0918

overtax

Taiwan

2003

0.246

0.8308

300.99

0.0841

0.2863

Spain

1991

0.400

0.556

823.7

0.100

0.840

Italy

1993

0.247

0.829

16.4

0.154

0.186

USA

1991

NA

NA

NA

NA

Belgium

1992

0.555

0.661

238.0

0.260

0.900

Sweden

1991

0.569

0

-24258.0

0.185

overtax

NA

Note: The results on non-East Asian countries are taken from Roemer et al. (2003)).

15

Figure 1: Empirical distribution functions of pre-fisc income, three EDU types, by type

Korea Korea, 2003/2004, ST income

Taiwan

.8

1

1 0

0

.2

.2

.4

.4

.6

.6

.8

1 .6 .4 .2 0 0

5000

10000

15000

0

5000

ST Income Edu_Dad1

10000

Edu_Dad2

Edu_Dad3

Edu_Pa1

Korea, 2003/2004, EQ income

Edu_Pa2

15000

Edu_Dad2

Edu_Pa3

.8 .6 .4 .2 5000

10000

15000

0

1000

EQ Income

Edu_Dad3

Edu_Pa2

4000

0 0

EQ Income Edu_Dad1

3000

1

1 .6 .4 .2 0

10000

2000 ST Income

Taiwan, 2003, EQ income

.8

1 .8 .6 .4

5000

1000 Edu_Pa1

Edu_Pa3

Japan, 2004, EQ income

.2 0

0

15000

ST Income

0

EQ income

Taiwan, 2003, ST income

Japan, 2004, ST income

.8

ST income

Japan

Edu_Pa1

16

Edu_Pa2

Edu_Pa3

Edu_Pa1

2000 EQ Income Edu_Pa2

3000

4000 Edu_Pa3

Appendix: Additional tables

Table A-1: Korea, Japan, and Taiwan in 2004 at a glance

Exchange rate Real GDP per capita ($, chain index) Real GDP per equivalent adult ($, chain index) Real GDP per worker ($, chain index) Secondary school enrollment ratio (net) Total tax revenue as % of GDP Taxes on income and profits as % of GDP Taxes on goods and services as % of GDP VAT rate Total social expenditure as percentage of GDP

Korea ₩1145.32/$1 21330.22 23655.94 43758.43 0.883 14.25 5.82 6.26 10% 3.35

Japan ¥108.19/$1 29203.49 31382.81 56141.31 0.999 16.50 8.37 5.52 5% 17.30

Taiwan NT$33.43/$1 23693.82 26239.60 52235.23 0.795 12.50 4.12 5.79 5% 2.33

Ratio(K/J) 10.59₩/¥ 0.730 0.754 0.779 0.884 0.864 0.695 1.134 2.000 0.194

Ratio(K/T) 34.26₩/NT$ 0.813 0.902 0.838 1.111 1.14 1.413 1.081 2.000 1.438

Source PWT 6.3 PWT 6.3 PWT 6.3 PWT 6.3 World Development Indicators See note * See note * See note * See note * See note *

* Source for Korea: World Development Indicators Source for Japan: Financial Statistics of Japan, Ministry of Finance, Government of Japan, Cabinet Office, Government of Japan, National Institute of Population and Social Security Research

Source for Taiwan: Ministry of Finance, Republic of China Directorate-General of Budget, Accounting and Statistics, Executive Yuan, Republic of China

17

Table A-2: Education and income Pre-fisc ST income Respondent’s schooling years

Korea

Japan

Taiwan

Respondent’s ST income (4 year average) 0-25%

25-50%

50-75%

75-100%

sum

0-9 yrs.

0.522

0.320

0.113

0.046

1.000

10-12 yrs.

0.261

0.281

0.261

0.196

1.000

13 & higher

0.163

0.200

0.295

0.343

1.000

0-9 yrs.

0.444

0.354

0.131

0.071

1.000

10-15 yrs.

0.289

0.294

0.234

0.183

1.000

16 & higher

0.153

0.162

0.290

0.394

1.000

0-9 yrs.

0.463

0.315

0.165

0.057

1.000

10-12 yrs.

0.234

0.257

0.306

0.202

1.000

13 and higher

0.079

0.132

0.309

0.479

1.000

Pre-fisc EQ income Respondent’s schooling years

Korea

Japan

Taiwan

Respondent’s ST income (4 year average) 0-25%

25-50%

50-75%

75-100%

sum

0-9 yrs.

0.507

0.293

0.154

0.046

1.000

10-12 yrs.

0.270

0.290

0.248

0.192

1.000

13 & higher

0.151

0.194

0.284

0.371

1.000

0-9 yrs.

0.475

0.232

0.172

0.121

1.000

10-15 yrs.

0.287

0.268

0.239

0.211

1.000

16 & higher

0.142

0.218

0.297

0.343

1.000

0-9 yrs.

0.372

0.352

0.185

0.091

1.000

10-12 yrs.

0.185

0.228

0.315

0.272

1.000

13 and higher

0.044

0.097

0.224

0.635

1.000

18

Table A-3: Father’s education and respondent’s income Pre-fisc ST income Respondent’s schooling years

Korea

Japan

Taiwan

Respondent’s ST income (4 year average) 0-25%

25-50%

50-75%

75-100%

sum

0-9 yrs.

0.317

0.274

0.222

0.187

1.000

10-12 yrs.

0.254

0.250

0.258

0.239

1.000

13 & higher

0.190

0.228

0.287

0.294

1.000

0-9 yrs.

0.284

0.268

0.234

0.214

1.000

10-15 yrs.

0.233

0.252

0.252

0.263

1.000

16 & higher

0.169

0.189

0.291

0.351

1.000

0-9 yrs.

0.443

0.266

0.190

0.100

1.000

10-12 yrs.

0.239

0.241

0.263

0.257

1.000

13 and higher

0.100

0.193

0.328

0.378

1.000

Pre-fisc EQ income Respondent’s schooling years

Korea

Japan

Taiwan

Respondent’s ST income (4 year average) 0-25%

25-50%

50-75%

75-100%

sum

0-9 yrs.

0.326

0.273

0.211

0.190

1.000

10-12 yrs.

0.248

0.247

0.268

0.237

1.000

13 & higher

0.178

0.227

0.255

0.341

1.000

0-9 yrs.

0.280

0.250

0.229

0.241

1.000

10-15 yrs.

0.244

0.246

0.270

0.241

1.000

16 & higher

0.149

0.250

0.291

0.311

1.000

0-9 yrs.

0.339

0.315

0.187

0.159

1.000

10-12 yrs.

0.198

0.220

0.247

0.335

1.000

13 and higher

0.058

0.143

0.290

0.510

1.000

19

Table A-4: Estimated values of d , g , and S

Country

d

g

S

Korea

0.045

0.822

188.9 (10,000 Won)

Japan

0.045

0.644

365.3 (1000 Yen)

Taiwan

0.073

0.459

38.0 (1000 NT$)

Source: Various sources (such as the National Income and Product Accounts, the Government Revenue Statistics, etc.) Notes: (1) S is the value of government services (capturing non-transfer payments) per capita. (It is calculate from the identity, Total Revenue=Transfer Payments + S.) (2) d is the proportion of indirect taxes paid in the total disposable income, where the disposal income is the income after income taxes are paid and government transfers are received. (3) g is the proportion of cash transfers in total government transfers. (Total government transfers consist of cash transfers and non-cash transfers (in-kind benefits).)

20

Table A-5: Observed affine and quadratic income tax functions: OLS regressions Korea t ST income

0.210 0.100

EQ income

s

1.05*10

-5

0.218

T (10,000 Won)

R2

262.33

0.863

83.31

0.955

246.52

0.819

0.139

6.00*10-6

120.39

0.897

t

s

T (10,000 Won)

R2

601.79

0.872

159.41

0.934

444.10

0.843

Japan

ST income

0.279 0.153

EQ income

6.68*10

-6

0.313 0.232

4.29*10-6

214.82

0.875

t

s

T (10,000 Won)

R2

78.27

0.8339

18.13

0.9143

46.91

0.8352

12.94

0.9050

Taiwan

ST income

0.267 0.126

EQ income

31.5*10

-6

0.246 0.141

29.4*10-6

Source: Korea Labor and Income Panel Study (Korea), Keio Household Panel Survey (Japan), and Panel Study of Family Dynamics (Taiwan) Notes: (1) We compute the observed income tax function by running the following regression:

Net taxes = t * x + s * x 2 - T , where x is the pre-fisc income. In calculating net taxes, we add payroll taxes and subtract social security benefits. (2) Standard income is the sum of the individual’s labor income and his household capital income per adult. (3) Equivalence income is the sum of household labor and capital income adjusted by the equivalence scale (the square root of household size). (4) Korean income and net taxes are expressed in ten thousands of Korean Won. (One US dollar is approximately equal to eleven hundred Korean Won.) It is expressed in real terms (2004=100). (5) Japanese income and net taxes are expressed in thousands of Japanese Yen. (6) Taiwan income and net taxes are expressed in thousands of NT$.

21

Table A-6: Observed affine and quadratic income tax functions: Median regressions Korea t ST income

0.152 0.043

EQ income

s

1.66*10

-5

0.151

T (1,000 Yen)

Pseudo R2

122.53

0.599

4.90

0.730

109.37

0.537

0.094

9.38*10-6

47.73

0.599

t

s

T (1,000 Yen)

Pseudo R2

415.44

0.666

64.25

0.731

274.52

0.598

Japan

ST income

0.241 0.117

EQ income

9.17*10-6

0.259 0.215

5.07*10-6

193.17

0.618

t

s

T (1,000 Yen)

Pseudo R2

30.26

0.4102

2.51

0.4974

26.84

0.4780

13.71

0.5368

Taiwan

ST income

0.1996 0.0845

EQ income

3.94*10-5

0.1922 0.1419

3.03*10

-5

Source: Korea Labor and Income Panel Study (Korea), Keio Household Panel Survey (Japan), and Panel Study of Family Dynamics (Taiwan) Notes: (1) We compute the observed income tax function by running the following regression:

Net taxes = t * x + s * x 2 - T , where x is the pre-fisc income. In calculating net taxes, we add payroll taxes and subtract social security benefits. (2) Standard income is the sum of the individual’s labor income and his household capital income per adult. (3) Equivalence income is the sum of household labor and capital income adjusted by the equivalence scale (the square root of household size). (4) Korean income and net taxes are expressed in ten thousands of Korean Won. (One US dollar is approximately equal to eleven hundred Korean Won.) It is expressed in real terms (2004=100). (5) Japanese income and net taxes are expressed in thousands of Japanese Yen. (6) Taiwan income and net taxes are expressed in thousands of NT$. 22

Table A-7: Regression results in Japan and Taiwan: Dependent variables are not transformed. Japan Edu

Cons

Age

-15420.4***

-14788.3***

-3307.2

-2900

(15.76)

(-5.36)

(-4.99)

(-1.53)

(-1.31)

0.0193*

655.5***

756.2***

105.7

167.7

(2.26)

(4.84)

(5.44)

(1.04)

(1.62)

-6.705***

-7.800***

-0.362

-1.037

(-4.29)

(-4.86)

(-0.31)

(-0.87)

Edu

Obs.

Wage from EQ incomes

8.379***

Age^2

Edu_dad

Wage from ST incomes

312.0***

193.6***

(8.66)

(7.15)

0.378***

138.2***

255.7***

59.53*

132.1***

(14.13)

(3.85)

(7.46)

(2.21)

(5.19)

1193

1189

1189

1193

1193

Taiwan Edu

Cons

Age

-1818.5*

-1535.1*

-582.9

-345

-21.87

(-2.46)

(-1.98)

(-1.11)

(-0.62)

-0.104***

68.09*

87.88*

9.688

25.44

(-7.78)

-1.98

-2.44

-0.4

-0.98

-0.656

-0.946*

-0.00585

-0.239

(-1.66)

(-2.30)

(-0.02)

(-0.81)

Edu

Obs.

Wage from EQ incomes

14.12***

Age^2

Edu_dad

Wage from ST incomes

59.52***

48.28***

-9.8

-11.28

0.371***

9.321*

30.64***

9.035**

26.27***

-17.26

-2.01

-7.12

-2.76

-8.53

1038

969

969

986

986

23

Table A-8: Regression results in Japan and Taiwan: Dependent variables are logistically transformed. Japan (Edu)

Cons

Age

(Wage) from ST incomes

-1.548*

-7.870***

-7.724***

-6.400***

-6.264***

(-2.52)

(-7.07)

(-6.87)

(-5.97)

(-5.79)

-0.00256

0.216***

0.240***

0.130**

0.151**

(-0.26)

(4.13)

(4.55)

(2.58)

(2.98)

-0.00233***

-0.00258***

-0.00125*

-0.00148*

(-3.86)

(-4.24)

(-2.15)

(-2.53)

Age^2

Edu

Edu_dad

Obs.

(Wage) from EQ incomes

0.0721***

0.0647***

(5.18)

(4.82)

0.291***

0.0219

0.0490***

0.00834

0.0326**

(9.43)

(1.57)

(3.77)

(0.63)

(2.62)

1193

1189

1189

1193

1193

Taiwan (Edu)

Cons

Age

-8.005***

-7.557***

-5.084***

-4.511**

-3.13

(-5.01)

(-4.61)

(-3.48)

(-2.95)

-0.0393***

0.197**

0.229**

0.0224

0.0603

(-4.32)

-2.65

-3

-0.33

-0.85

-0.00227**

-0.00273**

-0.000111

-0.000672

(-2.66)

(-3.13)

(-0.14)

(-0.83)

Edu

Obs.

(Wage) from EQ incomes

1.373**

Age^2

Edu_dad

(Wage) from ST incomes

0.0941***

0.116***

-7.16

-9.74

0.102***

0.0280**

0.0617***

0.0267**

0.0682***

-6.96

-2.78

-6.77

-2.92

-8.05

1038

969

969

986

986

Note: (1) t statistics are in parentheses * p