Discussion Paper No. 80

INCOME DISTRIBUTION AND IN JAPAN

by

Fumihira NISHIZAKI, Yutaka YAMADA and Eisuke ANDO Economic Planning Agency

October 1997

Economic Research Institute Economic Planning Agency Tokyo, Japan

The views expressed here are the author’s and do not

represent those of Economic Planning Agency

INCOME DISTRIBUTION AND POVERTY IN JAPAN

by

Fumihira NISHIZAKI, Yutaka YAMADA and Eisuke ANDO

Economic Research Institute, Economic Planning Agency

October 1997

Abstract

This paper presents some preliminary results of measuring income distribution in Japan, based on the micro data sets of the National Survey of Family Income and Expenditure. The focus is mainly on the changes from 1984 to 1994. The analytical framework of this paper closely follows that of the OECD literature using the micro data base known as the Luxembourg Income Study.

Inequality in equivalent disposable income increased and the incidence of relative poverty also rose from 1984 to 1994. Earnings were the largest contributor to the increased overall inequality in disposable income. The increased share of earnings in disposable income played an important role, in addition to the effects of the high level of their share. Work-attachment and the development of earnings were undoubtedly important factors determining relative income and poverty.

Taxes and transfer systems effectively redistributed income. The weakness of transfers in redistribution of income is conspicuous in Japan. Transfers were not well “targeted” to the poor; they rather contributed positively to the overall inequality. Taxes, including social security contributions, are quite redistributive. Although the positive contribution of transfers to the overall inequality somewhat increased, taxes became more redistributive, resulting in a partial offsetting of the widening of market incomes.

The elderly benefited the most from the tax-transfer system across demographic groups. This tendency was strengthened over the recent decade as the public pension system matured. Households with a young head experienced a significant decline in relative income, reflecting the poor development of their earnings. However, this was probably due to an increase in the labor participation rate for young females.

i

Table of Contents

1. Introduction

2. Distribution across all population

3. Factors affecting income distribution: an aggregate analysis

4. Changes in well-being for sub-group of the population

5. Poverty and low income

Appendix: The National Survey of Family Income and Expenditure

Table 1. Trends in inequality indices Table 2. Equivalent disposable income: gains and losses by quintile Table 3. Sensitivity of the results (Gini coefficients) Table 4. Gini coefficients from Income Redistribution Survey Table 5. Gini coefficients from household expenditure surveys Table 6. Allocation of income components across decile groups Table 7. Aggregate inequality indicators before and after taxes and transfers Table 8. Contribution of income components to total inequality: Shorrocks decomposition Table 9. Decomposition of changes in total inequality: SCV Table 10a. Relative disposable income, by degree of work attachment of households Table 10b. Contribution of income components to changes in relative disposable income, by work attachment of households Table 11a. Relative disposable income, by age of households head Table 11b. Contribution of income components to the changes in relative disposable income, by age of households head Table 12a. Relative disposable income, by family type Table 12b. Contribution of income components to changes in relative disposable income, by family type Table 13. Contribution of earnings and employment to changes in earnings: young and older-worker households Table 14. Contribution of earnings and employment to changes in earnings: households with children Table 15. Relative disposable income of non-working households Table 16. Relative disposable income of children, the elderly and single earners with children Table 17. Population structure by work attachment and household type Table 18. Changes in inequality: between-group and within-group effects Table 19. Decomposition of the changes in MLD by work attachment Table 20. Trends in poverty using a relative threshold Table 21. Trends in poverty using a constant threshold Table 22. Poverty rates and poverty structure after taxes and transfers

iii

INCOME DISTRIBUTION AND POVERTY IN JAPAN1

This paper presents some preliminary results of measuring income distribution in Japan, focusing mainly on the changes over the recent decade. The analytical framework is provided by OECD Secretariat (1997), which closely follows the method used in Atkinson et al. (1995). This allows us to roughly compare our results with those for countries covered by the Luxembourg Income Study (LIS), although deriving a decisive conclusion from such a comparison is dangerous given the differences in definitions across countries.

1. Introduction

1.1 Data and methodology

The analyses in this paper are based on the micro data sets of the National Survey of Family Income and Expenditure (NSFIE), for the years of 1974, 1984, 1989 and 1994. However, there are some problems in using the data for 1974, including discontinuity of coverage (see Appendix). Given the purpose of this project toward understanding long-term trends in income distribution, we therefore focus mainly on the comparison between 1984 and 1994, while medium-term movements are occasionally referred. These two years are suited for comparison, since their cyclical positions were similar in the sense that both belonged to the early periods after recovering from long-lasting recessions (1980-1983 and 1991-1993).

Since the NSFIE lacks annual data on taxes, we have calculated those data by a micro-simulation method as is often done in other countries. This inevitably results in errors because the simulation formulae were to some extent simplified given the limitation of time. The definitions and methodology then applied for aggregation of data are basically in line with those provided by the terms of reference provided by the OECD Secretariat.

The unit of analysis is the household. Since households differ in size, equivalence scales are necessary to adjust income to account for differences in need. The specific equivalence measure used in this paper is mainly the square root of the family size (equivalence elasticity = 0.5). The results adjusted by the family size itself (equivalence elasticity = 1) are sometimes reported as a reference. Household income is re-weighted by the number of people in each unit (“person weight”).

1.2 Main findings

Several important findings are worth listing at the outset.

1 This paper was originally presented at the Working Party No.1 of the Economic Policy Committee, OECD, October 9-10, 1997, as a room document. Preparation of the original tables from which the tables presented here were derived was conducted by the Working Group on Income Distribution, whose members were Fumio Funaoka (Shinshu University), Masato Okamoto (Management and Coordination Agency), Mariko Murata (Statistical Information Institute for Consulting and Analysis) and Fumihira Nishizaki (Economic Planning Agency).

-1-

Inequality in equivalent disposable income increased and the incidence of relative poverty also rose from 1984 to 1994. Earnings were the largest contributor to the increased overall inequality in disposable income. The increased share of earnings in disposable income played an important role, in addition to the effects of the high level of their share. Work-attachment and the development of earnings were undoubtedly important factors determining relative income and poverty.

Taxes and transfer systems effectively redistributed income. The weakness of transfers in redistribution of income is conspicuous in Japan. Transfers were not well “targeted” to the poor; they rather contributed positively to the overall inequality. Taxes, including social security contributions, are quite redistributive. Although the positive contribution of transfers to the overall inequality somewhat increased, taxes became more redistributive, resulting in a partial offsetting of the widening of market incomes.

The elderly benefited the most from the tax-transfer system across demographic groups. This tendency was strengthened over the recent decade as the public pension system matured. Households with a young head experienced a significant decline in relative income, reflecting the poor development of their earnings. However, this was probably due to an increase in the labor participation rate for young females.

2. Distribution across all population

2.1 Aggregate trends

The distribution of equivalent disposable income widened from 1984 to 1994 (Table 1). The Lorenz curve of 1994 was uniformly below that of 1984 and all aggregate inequality indicators (the , the squared coefficient of variation (SCV), the mean log deviation (MLD) and the Atkinson index with ε=0.5) showed an increase from 1984 to 1994. Changes in distribution largely occurred at the extremes, with progressively smaller movements around the mean (Table 2). A “hollowing out” of the middle was not conspicuous.

Changes in aggregate indicators from 1974 and from 1989 were also shown in Table 1. Note that figures for 1974 are not directly comparable with those for other years. Comparison of the distributions of disposable income between 1989 and 1994 is not straightforward since the SCV registered a decrease while other indicators showed an increase.

2.2 International comparison

We might be able to make a tentative comparison in terms of the Lorenz dominance with the distribution of other OECD countries documented in Atkinson et al. (1995). Of course, deriving a decisive conclusion from such a comparison is dangerous, since definitions of income and households are different across countries. Furthermore, measurement biases especially in capital and self-employed income, calibration errors in taxes and transfers, omission of in-kind benefits and other imputation, and even the lack of dynamic perspective will all require caution in international comparison.

With these caveats in mind, comparison of the 1994 distribution of disposable income (equivalence elasticity = 0.5) in Japan with the Lorenz curves in Table 4.3 of Atkinson et al. (1995) reveals that the former is “Lorenz superior” to Australia (1985), Canada (1987), France (1984), Ireland (1987), Italy (1986), Switzerland (1982), the United Kingdom (1986), and the United States (1986) and that it is

-2-

“inferior” to Belgium (1988), Germany (1984), Luxembourg (1985), Norway (1986), Finland (1987), and the Netherlands (1987).

The relative position of Japanese distribution in 1984 compared with the above-mentioned countries/years was the same as in 1994 except superiority to New Zealand (1988) and ambiguity with Germany (1984) and the Netherlands (1987). Japan was able to be classified as a modestly egalitarian country if these data are suited for international comparison.

2.3 Robustness

The results depend on the choice of equivalence elasticity, but the direction of changes in the Gini coefficients with the elasticity of 1, for example, is the same as in the case with the elasticity of 0.5. Like other OECD countries, income distribution is most equal when elasticity is 0.5 (Table 3).

In Japan, all the published Gini coefficients in various surveys are with household weight and zero elasticity. The Income Redistribution Survey (IRS)2, for example, which is conducted every three years by the Ministry of Health and , reports that the Gini coefficient of post-redistribution income (a similar concept to disposable income) increased during the recent decade and was 36.5 in 1992 (Table 4; for other results from published surveys, see Table 5). The IRS allegedly focuses on poor households and results in a relatively heavy weight of such households in its samples compared with the NSFIE. This coincides with the smaller Gini coefficient in 1994 in the latter when household weight and zero elasticity are used.

3. Factors affecting income distribution: an aggregate analysis

3.1 Composition of market income

The share of self-employment income significantly declined, reflecting a decrease in households engaged in agriculture and in small- and medium-sized establishments for manufacturing and retail, wholesale, and restaurants. Earnings of the spouse outweighed those of other household members during the recent decade.

One technical remark must be made about capital income, which accounted for only several per cent of market income. In the NSFIE, more than 70 per cent of the sampled households responded with zero capital income. This suggests significant underestimation of capital income since few households are reported to possess no interest-bearing financial assets in the same survey.

3.2 Evolution of the distribution of different income sources

Ranked by equivalent disposable income, total disposable income was more evenly distributed than market income, although the middle-income and higher-income group received a large share of transfers (Table 6). Earnings of the household head were allocated heavily to the middle-income group. Other types of earnings were concentrated in the higher-income group. The distributions of capital income and self-employment income were relatively unfavorable for the middle-income group.

2 The sample size of the IRS is about 9,000. The definition of post-redistribution income in the IRS is slightly different (including in-kind transfers through medical insurance) from that in the NSFIE.

-3-

If we regroup market income into two major components, “earnings” and “capital and self- employment income,” the share of earnings going to the lower-income group rose and that of capital and self-employment income fell over the recent decade. The latter types of income became more heavily allocated to the higher-income group. Individuals belonging to the lower-income group became more dependent on earnings, as the labor force moved from the self-employed to wage-earners. The share of transfers going to the lower-income group fell somewhat, implying a weakening role of transfers in redistribution. A larger share of taxes was paid by the higher-income group in 1994 than in 1984. 3 The effects of redistribution by taxes and transfers can be also seen by comparison of indicators of pre- and post-redistribution (Table 7). For example, the ratio of the Gini coefficient of disposable income and the pseudo-Gini coefficient of market income is such a measure (“redistribution rate”). The redistribution rate increased over the recent decade.

3.3 The Shorrocks decomposition

Decomposition of the SCV by the method proposed in Shorrocks (1982) shows that earnings accounted for nearly 100 percent of the total SCV, partly reflecting the largest share of earnings in equivalent disposable income (Table 8). Self-employment income was another important contributor to overall inequality. Capital income played only a minor role despite its own uneven distribution. The contribution of transfers was positive but small, implying its flat distribution across income groups. Taxes had an important negative contribution to overall inequality enough to offset that of self- employment income.

Over the recent decade, earnings were the major factor of the increase in overall inequality measured by the SCV (Table 9). This was due to the increased share of earnings and, to a lesser extent, a widening of their own distribution. The distribution of transfers became somewhat wider. This weakening of equalizing effect may have reflected maturity of the public pension system which contains earnings-related plans as an important ingredient. Contrary to this, taxes became more effective for equalization of disposable income over the recent decade, although a reverse movement was observed from 1984 to 1989 due to tax reforms.

3.4 International comparison of taxes and transfers

The distribution of taxes and transfers can be compared with Table 7.3 and 7.5 of Atkinson et al. (1995). Two characteristics of the Japanese tax distribution are a relatively large share of the bottom and top quintiles and a relatively small share of the third and fourth quintile. This might have reflected the government’s effort in supporting the need for expenditures related to education and housing by prime-aged, middle-class workers.

The distribution of transfers in Japan was even and considered not “targeted” in comparison with other OECD countries. Each of the lower three quintiles obtained less than 20 percent of the total transfers and the share of the top quintile far exceeds 20 percent. In Japan, the major part of transfers is public pension, which may have been received by a retired person living with his/her reasonably earning children or spouse. Transfers to working-age population such as unemployment insurance therefore had

3 However, an opposite movement was observed from 1984 to 1989. Taxes shifted their relative weight to the lower-income group, which weakened the redistribution effects of taxes and transfers. This reflects the tax reforms of September 1987 and December 1988, in which progressivity of income tax schedule was mitigated. For example, the maximum income tax rate was reduced from 70 to 50 percent due to the two consecutive reforms.

-4-

small shares. These facts can explain the above-mentioned results.

4. Changes in well-being for sub-groups of the population

4.1 Levels and changes in relative incomes of sub-groups of the population

According to grouping by age of the head and work-attachment, individuals in households with nobody at work had the lowest average equivalent disposable income in all age categories with equivalence elasticity 0.5 (Table 10a). This confirms the importance of work-attachment in determining the relative position of individuals. The change in relative incomes over time was not large compared with the differences across groups at a point of time. Note that the weights used are those of 1984, even for relative incomes in 1994, to avoid the impact of changing shares of the groups.

Equivalent disposable incomes rise with age of the head until retirement approaches, and decline thereafter (Table 11a). It is worth noting that the position of households with the retired head is well comparable to households with the prime-aged head. Over the recent decade, the relative position of households with the young head significantly declined while that of the older working-age heads improved.

Single-adult households accounted for about 7 percent of the population in person weight in 1994. Among two-adult (or more) households, those with children and two workers were the largest in their share in person weight, followed by those with children and only one worker, then those without children and two workers. The shares of single-adult households and of two-adult households without children and two workers increased, while those of two-adults households with children and with more than one worker (regardless of their number) decreased during the recent decade.

Grouping by the number of adults and children reveals that single-adult households with children had the lowest disposable income (Table 12a). This is followed by single-adult households without children, two-adult households with children and those without children when equivalence elasticity is 0.5. The configuration differs when elasticity is unity, where two-adult households fell behind single- adult households without children. In the former case, the relative position of single-adult households with children showed a significant deterioration over the recent decade. In conjunction with the deterioration of the relative position of young households, this might have reflected a rapid rise in the labor market participation of young female population.

4.2 Accounting for changes in relative income of population sub-groups

Across work-attachment groups, it is natural that changes in relative disposable income of households with two workers are largely explained by the contribution of earnings (Table 10b). For households with only one worker, the negative contribution of capital and self-employment income was not sufficiently offset by the effects of earnings. A large contribution of earnings to relative income of non-working households looks peculiar. This by definition should have been related to the behavior of working retirement-age individuals.

Earnings growth also played an important role in explaining the improved relative position of households with an older working-age head (Table 11b). Households with the retirement-age head benefited most from an increase in transfers. On the other hand, it was not the case for households with the older working-age head. This shows that an improvement in the public pension system in Japan did

-5-

not significantly affect the retirement decisions by workers in this group4.

It is noteworthy that the decline in the relative position of single-adult households without children was particularly related to earnings (Table 12b). Earnings growth contributed positively to changes in the relative positions of other groups by the number of adults and children.

4.3 Factors affecting earnings across household types

Earnings contributed to a relative decline in disposable income of households with the young head on average. However, the shares of the three work-attachment groups were essentially unchanged (Table 13, a nearly zero “employment effect”). It seems that higher youth unemployment and increased female labor participation offset each other to some extent. Households with the older working-age head were characterized by a strong positive effect of earnings. The “employment effect” was negative but relatively small. In fact, the participation rate in the labor market of the population aged 60 to 64 increased from 1984 to 1994, although a decreasing trend was observed until 1988.

Comparison between single-adult and two-adult households both with children reveals that there were no significant differences in terms of the direction of effects of earnings growth or work-attachment (Table 14). For both types of households, an increase in earnings and a fall in the share of non-working households had non-negative effects to the changes in relative income position over the recent decade.

4.4 Developments for selected population groups of special policy interest

We focus our attention on special categories of population in this section: non-working households with the working-age head, children and the elderly captured as individuals, households with a retired head and single-earner households with children.

Each non-working group with a working age head classified by the age of the head had relative income well below 100 percent (Table 15). Among them, relative incomes tended to rise along with the age of the head. This phenomenon can be explained by increasing transfers to the head as well as self- employment income of other household members.

Children had equivalent disposable income below average by 11 percent (Table 16). Older individuals had also income below average but to a lesser extent. Specifically, individuals over 75 years old were enjoying nearly comparable income to the average population, helped by transfers and sometimes by cohabitation with their children. It is interesting to know about the distribution of income among different cohorts. Children are characterized by a relatively even distribution of income. The distribution tends to be polarized as the age exceeds 50. The polarization is most conspicuous for individuals over 75. Retired households had disposable income below average by 17 percent. However, this category of households registered an increase in relative income over the decade due to enhanced transfers.

Low relative income also occurs for single-earner households with children, especially those with a single parent. These types of households experienced a significant fall in relative income over the recent decade. In both cases, however, earnings had a positive effect while self-employment income worked negatively to a considerable extent.

4 This does not imply the possibility that the public pension system was not entirely discouraging incentives to work. In fact, an adverse feature of the past pension scheme in this respect was reformed in 1994.

-6-

4.5 Decomposition of the changes in MLD by work-attachment

In Japan, changes in the population structure by work-attachment were surprisingly small as far as households with the working-age head are concerned (Table 17). The share of non-worker households increased, but only slightly over the recent decade. The increase occurred mainly for households without children, possibly reflecting population aging. Changes in the structure of total population confirm the progress in aging. At the same time, two-adult households without children showed an increase in its share.

The decomposition of the changes in the overall MLD from 1984 to 1994 by work-attachment reveals that the within-group component (a component due to inequality within each group) was far larger than the between-group component (a component due to differences in average incomes between groups) although both contributed to the widening of the overall distribution (Table 18).

A further decomposition into the “pure” within- and between-group effects (those with the population shares held constant) and the structural effects (those due to changes in the share of each group) shows that the structural effects were dominant in explaining the changes in the overall MLD (Table 19). The “pure” within effects also contributed to a widening of the overall distribution but to a lesser extent. The “pure” between effects were negative, implying changes in the relative incomes between groups alleviated a widening of the overall distribution. Among the structural effects, the increase in the share of non-working households was dominant. Those households also played an important role in the “pure” between effects.

5. Poverty and low income

5.1 Aggregate trends in poverty

Relative poverty increased from 1984 to 1994 in terms of the head count number of the poor regardless of the (Table 20). The income gap and the Gini coefficient among the poor in relative terms rose. As a result, the Sen index (the head count ratio multiplied by the sum of the income gap ratio and the Gini coefficient of the poor weighted by 1 minus the income gap ratio) increased.

The level of a head count ratio can be compared with the low income rates in Table 7.2 of Atkinson et al. (1995). The low income rate is defined as that with equivalent incomes of less than half of the median. In Japan, low income rates were between 7 - 9 percent, which are below those in Australia, Canada, Ireland, Italy, the United Kingdom, and the United States, and above those in Belgium, Germany, Luxembourg, and the Netherlands.

Despite aggravation of relative poverty, absolute poverty decreased from 1984 to 1994 in terms of the head count number of the poor regardless of the poverty threshold, reflecting increased average real income (Table 21). Growing average income limits the absolute poor within the extreme lower tail of distribution. Therefore, income gap and the Gini coefficient among the poor significantly increased in absolute terms.

5.2 The structure of poverty

This section uses another definition of poverty, the share of individuals in specific groups which

-7-

fall into the bottom quintile of the distribution of equivalent disposable income. The use of such a definition is aimed at examining demographic patterns at risk of poverty. It should be noted that 20 percent of the total population is always classified as suffering from poverty according to this definition.

Non-working households were exposed to a relatively high risk of poverty in general (Table 22). Among them, households with the young head had a very high rate of poverty, followed by those with children. Households with the retirement-age head had relatively small risk. In Japan, the share of those households is dominant in the group of non-working households because of its low unemployment rate. The risk of one-worker households is not significantly different from the population average (20 per cent). However, single-adult households with children and with only one worker were exposed to higher risk.

The risk of poverty was reduced for non-working households as a whole over the recent decade. An improvement in the situation for households with the retirement-age head mainly contributed to this change, reflecting the maturing of the public pension system. Instead, non-working households with a young head increased their risk to some extent. Furthermore, the risk increased for one-worker households, especially for those with a single adult and children.

One noteworthy feature of the structure of poverty in Japan was a very low share of non-working households. This was a natural consequence of low unemployment and high labor participation of the elderly from an international point of view. Nearly half of the total poor (20 per cent of the population) were one-worker households. However, the share of non-working households among the poor increased over the recent decade, probably due to population aging.

-8-

References

Atkinson, A. B., L. Rainwater and T. M. Smeeding (1995), Income Distribution in OECD Countries, Social Policy Studies No. 18, OECD.

OECD Secretariat (1997), “Income Distribution and Poverty in Selected Countries,” a background document for the Working Party No.1 of the Economic Policy Committee, unpublished.

Shorrocks, A. F. (1982), “Inequality Decomposition by Factor Components,” Econometrica, Vol. 50, No. 1.

-9-

Appendix: The National Survey of Family Income and Expenditure

This appendix briefly describes the National Survey of Family Income and Expenditure (NSFIE), from which our analyses were based. The NSFIE has been conducted every five years since 1959 by the Management and Coordination Agency. The latest survey, the 8th in the series, was undertaken between September and November 1994. It should be noted that the main purpose of the NSFIE is not to examine income distribution per se, but to grasp overall family budget structure from the perspective of income, consumption, and assets. However, the present survey has a large sample of about 60,000 households (including 4,700 one-person households), so that it also enables comprehensive analyses on income distribution if necessary adjustment is implemented.

The final sampling unit of the NSFIE is the household. The survey is undertaken separately for two-or-more person households and one-person households. Two-or-more person households are defined as a group of two or more persons sharing dwelling and living expenses. One person households refer to a person living alone, a person residing with an ordinary household but keeping a separate budget, or an individual person residing in a dormitory, boarding house, or lodging house. It covers all households in the whole country except certain types of households. The exceptions are: households running restaurants or inns on the same premise; those running boarding houses; those with boarders; foreigners’ households; two-or-more person households with four or more live-in employees; one-person households under 15 years of age; those residing with live-in employees; students; inmates of social and correctional institutions; and inpatients in hospitals and sanatoriums. These exceptions do not seem to be serious for our purposes since they cover less than 5 percent of households in Japan.

There is an important discontinuity about coverage of the survey. The surveys until 1979 did not cover households which were mainly or exclusively engaged in agriculture. Although this paper presents aggregate inequality indicators for 1974 as a reference, they cannot be compared directly with data after 1984 given the discontinuity and other technical problems found in the micro data set.

In this survey, “yearly income” is defined as the total pre-tax cash income of all members of the household in the 12 months preceding the survey date. The yearly income and its items are asked for the household head, his/her spouse, and other household members. It differs from the total household income concept, as defined in the United Nations guidelines, in which the latter concept also includes income in kind, imputed rents from owner-occupied dwellings, and employer contributions to social security and other similar schemes.

A major problem when using this survey for LIS-like international comparison is the lack of yearly disposable income data, which should therefore be estimated. For employee households, we could have exploited the fact that their disposable income and tax payment were reported for some months (September, October and November for two-or-more person households, for example). These monthly data may be readily transformed into annual data by seasonal indices obtained from other surveys such as the Family Income Expenditure Survey. However, such a method is not applicable to other types of households which lack even monthly data on taxes. Therefore, we chose a micro-simulation method for all types of households in order to obtain consistent estimates.

-10-

Table 1 Trends in inequality indices 1

Equivalence scale elasticity=0.5

1974-1994 1984-1994 1989-1994 Gini levels Initial level 25.2 26.0 (26.6)2 (24.9) (25.9) Final level (26.3) 26.5 26.5 Gini % change (-1.2) 4.9 1.7 Absolute change (-0.3) 1.2 0.4 SCV % change (-1.6) 21.7 -10.6 Absolute change (-0.5) 5.3 -3.5 MLD % change (4.1) 13.5 5.9 Absolute change (0.5) 1.5 0.7 Atkinson % change (0.0) 10.9 3.3 Absolute change (0.0) 0.6 0.2 1. Absolute change is the difference in the value of the index. 2. Figures in brackets are those excluding households with a head mainly or exclusively engaged in agriculture, forestry or fishery.

Equivalence scale elasticity=1.0

1974-1994 1984-1994 1989-1994 Gini levels Initial level 28.8 29.4 (29.6)2 (28.3) (29.1) Final level (29.4) 29.7 29.7 Gini % change (-0.7) 3.2 1.0 Absolute change (-0.2) 0.9 0.3 SCV % change (3.0) 14.0 -5.9 Absolute change (1.1) 4.8 -2.5 MLD % change (4.1) 9.3 4.1 Absolute change (0.6) 1.3 0.6 Atkinson % change (2.7) 8.7 2.7 Absolute change (0.2) 0.6 0.2 1. Absolute change is the difference in the value of the index. 2. Figures in brackets are those excluding households with a head mainly or exclusively engaged in agriculture, forestry or fishery.

-11-

Table 2 Equivalent disposable income: gains and losses by quintile

Equivalence scale elasticity=0.5

Quintile 1 2 3 4 5

1984-1994 -0.4 -0.3 -0.1 0.1 0.7

-12-

Table 3 Sensitivity of the results (Gini coefficients)

Weight Equivalence 1984 1989 1994 scale elasticity

Person 1.0 28.8 29.4 29.7 0.5 25.2 26.0 26.5 0.0 27.1 28.0 28.6 Household 0.0 30.3 31.5 32.2

-13-

Table 4 Gini coefficients from Income Redistribution Survey

Pre-redistribution Post-redistribution Post-redistribution Post-redistribution Year 1 income income through taxes2 through social security3 Gini (a) Gini (b) 1 -b/a (%) Gini (c) 1-c/a (%) Gini (d) 1-d/a (%)

1961 39.0 34.4 11.8 - - - - 1966 37.5 32.8 12.6 36.1 3.7 34.2 8.7 1971 35.4 31.4 11.4 33.8 4.4 33.4 5.7

- 1974 37.5 34.6 7.8 36.4 2.9 35.8 4.5

14 1977 36.5 33.8 7.4 35.2 3.7 36.1 1.2 -

1980 34.9 31.4 10.0 33.0 5.4 33.2 5.0 1983 39.8 34.3 13.8 38.2 3.8 35.8 9.8 1986 40.5 33.8 16.5 38.8 4.2 35.6 12.0 1989 43.3 36.4 15.9 42.1 2.9 37.9 12.5 1992 43.9 36.5 17.0 42.6 3.2 38.1 13.2

1. The year when data were collected. Note that the publication of the survey is titled with next year. 2. Pre-redistribution income minus tax payment excluding social security contribution. 3. Pre-redistribution income plus social security benefit including imputed medical benefit minus social security contribution. Source : Ministry of Health and Welfare, Income Redistribution Survey.

Table 5 Gini coefficients from household expenditure surveys

Two-or-more person household, pre-redistribution income1

Year NSFIE 2 FIES3

1979 27.1 - 1980 - 27.2 1981 - 27.3 1982 - 27.9 1983 - 27.5 1984 28.0 27.2 1985 - 28.5 1986 - 29.0 1987 - 28.7 1988 - 28.0 1989 29.3 28.6 1990 - 29.2 1991 - 29.6 1992 - 29.3 1993 - 29.2 1994 29.7 29.2 1995 - 29.6

1. Calculated from decile data. Sources : 2. Management and Coordination Agency, National Survey of Family Income and Expenditure. 3. Management and Coordination Agency, Family Income and expenditure Survey.

-15-

Table 6 Allocation of income components across decile groups

Equivalence scale elasticity=0.5

Percent, and changes in percentage points

1994 1984 Changes, 1984-1994 Earnings three bottom deciles 13.0 11.8 1.1 four middle deciles 36.5 37.6 -1.1 three top deciles 50.5 50.6 -0.1

Capital and self-employment income three bottom deciles 17.8 23.3 -5.5 four middle deciles 27.5 31.3 -3.8 three top deciles 54.7 45.4 9.4

Market income three bottom deciles 13.7 15.0 -1.2 four middle deciles 35.1 35.9 -0.8 three top deciles 51.2 49.2 2.0

General government transfers three bottom deciles 27.5 28.0 -0.5 four middle deciles 37.5 32.7 4.8 three top deciles 35.0 39.3 -4.2

Taxes three bottom deciles 11.3 12.6 -1.3 four middle deciles 29.7 30.9 -1.2 three top deciles 59.0 56.6 2.4

Total disposable income three bottom deciles 15.7 16.3 -0.6 four middle deciles 36.5 36.7 -0.2 three top deciles 47.8 47.0 0.8

-16-

Table 7 Aggregate inequality indicators before and after taxes and transfers

Equivalence scale elasticity=0.5

1994 1984 % changes, 1984-1994 SCV Before taxes and transfers (1) 53.6 40.1 33.7 After taxes and transfers (2) 29.6 24.3 21.7 % changes due to taxes and transfers (2)/(1)-1 -44.9 -39.4

Gini Before taxes and transfers (1) 34.0 29.8 14.0 After taxes and transfers (2) 26.5 25.2 4.9 % changes due to taxes and transfers (2)/(1)-1 -22.0 -15.2

Atkinson Before taxes and transfers (1) 12.4 8.4 47.3 After taxes and transfers (2) 5.9 5.3 10.9 % changes due to taxes and transfers (2)/(1)-1 -52.3 -36.7

-17-

Table 8 Contribution of income components to total inequality: Shorrocks decomposition

Equivalence scale elasticity=0.5

Percent, and changes in percentage points

1984 1994 Changes, 1984-1994 Percentage contribution to total inequality of disposable income (1) Earnings 94.1 94.8 0.7 Capital and self-employment 41.2 41.8 0.6 Transfers 2.6 4.3 1.7 Taxes -37.9 -40.9 -3.0 Transfers and taxes -35.3 -36.6 -1.3

Percentage shares (2) Earnings 82.5 93.0 10.5 Capital and self-employment 30.8 16.8 -14.0 Transfers 6.7 10.8 4.1 Taxes -20.0 -20.6 -0.5

Relative inequality indicator (1)/(2) Earnings 1.14 1.02 -0.12 Capital and self-employment 1.34 2.49 1.15 Transfers 0.38 0.40 0.01 Taxes 1.89 1.99 0.10

-18-

Table 9 Decomposition of changes in total inequality: SCV

Equivalence scale elasticity=0.5

Absolute changes 1

1984-1994 Earnings Part due to change in: Shares 2 (1) 3.0 Component inequality 3 (2) 2.1 Total (1)+(2) 5.2 Capital and self-employment income Part due to change in: Shares 2 (3) -7.4 Component inequality 3 (4) 9.8 Total (3)+(4) 2.4 Transfers Part due to change in: Shares 2 (5) 0.4 Component inequality 3 (6) 0.2 Total (5)+(6) 0.6 Taxes Part due to change in: Shares 2 (7) -0.3 Component inequality 3 (8) -2.6 Total (7)+(8) -2.9 Total change in SCV Part due to change in: Shares 2 -4.3 Component inequality 3 9.5 Total 5.3 1. Absolute change is the difference in the value of the SCV index. 2. Change in SCV due to changes in the share of each component in total income. Where the sign is negative, inequality is reduced. 3. Change in SCV arising from widening or narrowing in the distribution within each component.

-19-

Table 10a Relative disposable income, by degree of work attachment of households

Equivalence scale elasticity=0.5 Population in households with a working-age head

No worker One worker Two workers

Level 1994 62.7 88.7 110.0 % change,1984-1994 9.2 -5.3 3.5 Percentage point change 5.3 -5.0 3.8

Table 10b Contribution of income components to changes in relative disposable income, by work attachment of households

Equivalence scale elasticity=0.5 Population in households with a working-age head

No worker One worker Two workers

Earnings 17.2 2.9 22.7 Capital and self-employment income -15.4 -10.3 -17.1 Transfers 3.8 1.2 0.6 Taxes 3.7 0.9 -2.7

Disposable income 9.2 -5.3 3.5

-20-

Table 11a Relative disposable income, by age of households head

Equivalence scale elasticity=0.5

Young Prime Age Older Age Retired Household Household Household Household (below 30) (30-49) (50-64) (above 64)

Level 1994 75.5 93.7 120.1 92.7 % change,1984-1994 -7.8 -1.4 2.6 -1.3 Percentage point change -6.4 -1.3 3.0 -1.3

Table 11b Contribution of income components to the changes in relative disposable income, by age of households head

Equivalence scale elasticity=0.5

Young Prime Age Older Age Retired

Household Household Household Household

Earnings -3.6 12.0 21.9 -2.1 Capital and self-employment income -4.2 -13.9 -17.9 -20.0 Transfers -0.3 1.0 0.3 16.4 Taxes 0.3 -0.5 -1.6 4.3

Disposable income -7.8 -1.4 2.6 -1.3

-21-

Table 12a Relative disposable income, by family type

Equivalence scale elasticity=0.5

Single adult, Single adult, no Two adults, with Two adults, no

with children children children children

Level 1994 56.9 81.6 93.4 120.4 % change,1984-1994 -2.9 -6.5 -1.4 1.3 Percentage point change -1.7 -5.7 -1.3 1.6

Table 12b Contribution of income components to changes in relative disposable income, by family type

Equivalence scale elasticity=0.5

Single adult, Single adult, no Two adults, with Two adults, no

with children children children children

Earnings 14.2 -23.8 13.9 12.2 Capital and self-employment income -15.3 1.1 -15.5 -14.9 Transfers -1.8 12.8 0.9 3.8 Taxes 0.0 3.3 -0.7 0.3

Disposable income -2.9 -6.5 -1.4 1.3

-22-

Table 13 Contribution of earnings and employment to changes in earnings: young and older-worker households

Equivalence scale elasticity=0.5 Households with a young and older working-age head

Change in earned income per capita Contribution to total change due to: for each group relative to the Earnings effect Employment effect population average (2) (3) (1)

Young adult heads, 1984-94 -15.6 -15.7 0.1 Older working-age heads, 1984-94 14.6 15.3 -0.7

1. Change in earned income per capita of each group is relative to earned income per capita for the whole population. 2. Change in relative earned income arising from a change in the relative position of earnings of each work-attachment group("Earnings effect"). 3. Change in relative earned income due to changes in the share of individuals in each work-attachment group("Employment effect"). - 23

Table 14 Contribution of earnings and employment to changes in earnings: households with children

Equivalence scale elasticity=0.5 Households with children

Change in earned income per capita Contribution to total change due to: for each group relative to the Earnings effect Employment effect population average (2) (3) (1)

Single-parent household, 1984-94 3.1 0.8 2.3 Two-parent household, 1984-94 1.1 1.0 0.1

1. Change in earned income per capita of each group is relative to earned income per capita for the whole population. 2. Change in relative earned income arising from a change in the relative position of earnings of each work-attachment group("Earnings effect"). 3. Change in relative earned income due to changes in the share of individuals in each work-attachment group("Employment effect").

Table 15 Relative disposable income of non-working households

Equivalence scale elasticity=0.5 Percent, and changes in percentage points

By age of household head Young head Prime-age Older working- Working-

head age head age head

Level, 1994 39.1 56.6 65.2 63.7 Changes,1984-1994 -9.2 12.0 4.4 6.0

-24-

Table 16 Relative disposable income of children, the elderly and single earners with children

Equivalence scale elasticity=0.5 Percent, and changes in percentage points

- By age of individuals By age of Single-earner households 25

- household head with children Age 0-17 Age 65-75 Age 75+ Over 65 Single Two not working parent parents

Level, 1994 89.4 93.7 99.3 82.7 54.4 85.5 Changes,1984-1994 -2.5 1.0 0.5 6.2 -9.8 -4.3

Table 17 Population structure by work attachment and household type

Percent, and changes in percentage points

1994 1984 Changes,1984-1994 Working-age population By work attachment Two workers 56.5 55.1 1.4 One workers 41.4 43.3 -1.8 -

26 No workers 2.1 1.6 0.4

- Total population

By age of head Young head 5.0 6.2 -1.2 Prime-age head 52.5 59.7 -7.2 Older working-age head 30.3 26.6 3.7 Retirement age head 12.2 7.5 4.7 By family type Single adult with childen 0.5 0.7 -0.2 Single adult, no childen 6.6 5.3 1.3 Two adults with childen 57.0 69.5 -12.5 Two adults, no childen 35.8 24.4 11.4

Table 18 Changes in inequality: between-group and within-group effects

Equivalence scale elasticity=0.5 Levels and absolute changes

Total MLD MLD Between group MLD Within group

Level in 1994 12.6 0.7 11.9 changes,1984-1994 1.5 0.3 1.2 share of total change 100.0% 19.8% 80.2%

-27-

Table 19 Decomposition of the changes in MLD by work attachment

Equivalence scale elasticity=0.5 Absolute changes

28 Total MLD Total MLD decomposition That part due to non-workers

- Structural "Pure" "Pure" within Structural "Pure" "Pure" within Sub-total

effect between group effect effect between group effect MLD group effect group effect

Changes,1984-1994 1.5 1.8 -1.0 0.7 2.3 -0.8 -0.1 1.5

Table 20 Trends in poverty using a relative threshold 1

Equivalence scale elasticity=0.5

Changes in percentage points, unless otherwise indicated

1984-1994 40% Head-count ratio 0.6 50% median income Head-count ratio 0.8 Income gap ratio 2.5 Change in Sen index 2 Per cent 23.1 Absolute 0.6 60% median income Head-count ratio 1.0 1. "Relative threshold" poverty lines are fixed in terms of real median income in each period. 2. Absolute change is the difference in the value of the index.

Table 21 Trends in poverty using a constant threshold 1

Equivalence scale elasticity=0.5

Changes in percentage points, unless otherwise indicated

1984-1994 40% median income Head-count ratio -1.4 50% median income Head-count ratio -2.9 Income gap ratio 4.2 Change in Sen index 2 Per cent -29.7 Absolute 0.8 60% median income Head-count ratio -5.7 1. "Constant threshold" poverty lines are fixed at the level of real median income in the earliest period. 2. Absolute change is the difference in the value of the index.

-29-

Table 22 Poverty rates and poverty structure after taxes and transfers1

Equivalence scale elasticity=0.5

Percent, and changes in percentage points

1994 1984 1984-1994 Households with no workers All non-working houeholds rate 42.5 55.9 -13.4 share 20.1 13.4 6.7 Young head rate 87.6 73.7 13.9 share 0.2 0.2 0.0 Retirement-age head2 rate 39.4 51.2 -11.8 share 15.0 8.4 6.6 Single adult children rate 70.6 94.4 -23.8 share 0.3 0.7 -0.4 Two adults children rate 70.3 63.6 6.7 share 0.7 0.9 -0.1 One-worker households All one-worker households rate 23.0 20.3 2.8 share 44.8 42.3 2.5 Single adult children rate 67.1 52.2 14.9 share 1.6 1.5 0.1 Young adult head rate 38.9 31.1 7.8 share 7.4 7.4 0.0 1. Data are presented for selected groups only and shares do not sum to 100. 2. Mainly retired households.

-30-