A Demographic Analysis of Income Distribution in Estonia(*).

Wilder, Lisa ; Benedict, Mary Ellen ; Viies, Marie 等

Journal of Economic Literature Classification Numbers: L2, P0

I. Introduction.

Above all, the transition [to market economies] has had, and will continue to have, a profound impact on peoples lives. In some of the countries undergoing transition, there has been a short-term drop in living standards; in others, human welfare has improved dramatically. Everywhere, it has changed the basic rules of the game as has irreversibly altered the relationship between people and their political, social, not to mention economic, institutions. (World Bank President James D. Wolfensohn, 1996).

The transition of many centrally planned economies to free markets provides an unsurpassed opportunity for those studying comparative economic systems and the impact of the transition on the citizens of a nation. This paper examines the impact of the new rules of the game on the Estonian public. By examining the demographic composition of the Estonian income distribution, we hope to identify those demographic groups most likely to be prospering and those groups who are not after four years of economic reform. It is hoped that this information will provide a vital first step in understanding the income distribution in transforming economies and how we may aid those in Estonia negatively affected by the new market mechanism.

This research will add two contributions to the literature on transition economies. First, it will enhance our understanding of income distribution by looking at an emerging market nation. Using a data set comparable to that used by other nations, we can examine whether Estonia's income distribution is what one expects, given what is found in other market economies. Numerous studies (Osberg, 1991; Levy and Murname, 1992; Gottshalk and Smeeding, 1997) indicate a growing income inequality for many market economies. These studies suggest that the increase in income and wage inequality, is in part due to the increased returns to schooling and technological skills. It would be expected that a market in transition would especially benefit those households whose heads possess skills and education that enable them to adapt to structural economic change.

Second, we provide information regarding the Estonian income distribution at a vital time in its history by analyzing the income distribution of households based upon gender, age, and education. This information is of interest to the economic community as well as to policy makers interested in aiding those struggling with structural economic change.

Estonia provides an interesting environment due to recent changes in the nation, its commitment to market mechanisms, and the availability of primary data on income by demographic characteristics. Estonia was the first former Soviet state to introduce its own currency in 1992 (Venesaar and Hachey, 1995). It is committed to rapidly developing trade ties both within the former Soviet environment and with Western Europe. Like many former Soviet economies, Estonia hopes to gain acceptance into the EU and has proceeded quickly with market reforms aimed at restoring competition, private ownership, and trade relationships with Europe and the rest of the world.

II. Recent Legal and Economic Changes in Estonia

The Baltic states gained independence from the Soviet regime in 1991 (Venesaar and Hachey 1995). Estonia, as did many other former Soviet states, implemented political, social, and economic reforms to move the country to a democratic political structure supported by a capitalistic economic system. Reform in ownership laws, social policy, and taxes, have moved Estonia toward a market economy.

Initially, changes in the ownership and enterprise laws began the establishment of the property rights essential for a market-based economy. State-controlled industries are slowly but surely being dismantled, with many operations moving to private ownership (Venesaar and Hachey, 1995). Second, the growing pains of economic transition led to several years of economic instability and high inflation. The negative effects of transition, as well as the private control of wealth, created the need for social safety nets. Unfortunately, the government had limited funds to provide a broad range of social protection. Thus, those who struggled the most with the economic transition because of a lack of skills or the inability to work had little government support. In order to provide for this social protection, Estonia implemented a third area of reform. The government implemented a personal income, enterprise, and value-added taxes, and a social tax on employers, to support existing government programs and to create new ones.

Despite these efforts to protect poorer citizens through redistributive programs, the move to a market economy has created growing income inequality. Using aggregated information on household income, the Estonian Institute of Economics (Lugas and Vartia, 1993; Rasulu, 1994; Venesaar and Hachey, 1995) estimated annual Gini coefficients before and after the political independence of the country. The studies contain several important results regarding the standard of living for Estonians in general and for income inequality:

-- Real average monthly household income (whether using gross or disposable income measures) decreased between 1992 and 1994 by as much as 70%. This decrease in purchasing power occurred while income taxes rose over tenfold.

-- Only about 10% of households fell below the official poverty line (about $22/month in US currency). More than half of the Estonian households earned less than the average wage. This indicates a skewed income distribution with a small middle class in Estonia. Households containing many children, those with pensioners, and those with unemployed individuals, were the most likely to fall below the national average household income level.

-- Income inequality increased between 1990 and 1994. By 1994, those in the highest deciles had an average per capita monthly income that was fourteen times higher than those in the lowest decile. Estimated Gini coefficients for gross household income per capita rose from .20 in 1990 to nearly .40 in 1994.

Of significant concern to the Estonia government is the plight of specific groups such as the elderly. Previous studies of income distribution such as that by Mookerjee and Shorrocks (1982) for UK data and Blinder (1980) for income in the United States indicate that income distribution within age groups tends to be fairly uniform except for the over 60 group. In the senior age group, these studies have noted increased income dispersion when lifetime values of income are estimated. In Estonia, many elderly lost their savings as a result of the conversion from the ruble to Estonian currency. It is believed such a loss of lifetime savings will have an impact on income distribution for older cohorts and therefore on overall income inequality.

In our current study, we intend to go beyond these aggregated descriptive statistics. Thanks to a detailed survey of Estonian households, we are better able to determine the influence of demographic characteristics of heads of household on the income distribution. We use Gini coefficients and regression analysis to estimate income differences by head of household age, education, gender, and nationality. We assume that most characteristics of the head reflect the demographic make up of the household (e.g., age and nationality of the head is likely to be highly correlated with that of the spouse). In addition, because the head of household earnings and total income are 66 percent and 56 percent of the family values, respectively, we are comfortable using the head's information with household income values.

III. Methodology

Measures of income inequality are well established in the literature. We have elected to use the Gini coefficient as our measure of income inequality, primarily because of its common usage across studies and countries and because meaningful comparisons can be made in regard to income inequality within or between demographic groups. The Gini coefficient is a cumulative value of income shares and grows larger as income inequality grows. Using the Gini coefficient estimates, we address questions regarding whether inequality arises because certain groups have wide household income disparity or because some groups are more likely to be in the lower end of the income distribution while other groups are in the high end. Regression analysis further enables us to examine the impact of particular demographic characteristics holding other important factors constant. We will use regression analysis of household income and its components based on head of household demographic information to further examine the income determination process.

The Household Income and Expenditure Survey Data. In order to examine the income distribution in Estonia, it is necessary to have current information regarding earnings and other income for each household. The primary data used in this study are collected by the Estonian government's Statistical Office with the Household Income and Expenditure Survey (HIES). The survey uses information from questions that are similar in nature to US surveys such as the Panel Study of Income Dynamics.

Since July of 1995, The HIES collects monthly information on 1,000 randomly selected households from Estonia. The resulting database contains detailed information on total household income and its components (earnings, social transfers, other assets), and each household member's birth date, gender, education level, ability to work, occupation and industry, if working, and region of residence. A database on head of household information was compiled from the more general database, and the head of household data, in conjunction with household characteristics, are used to analyze household income. Because head of household is not specifically defined in the HIES, we define head of household as the individual with the highest earnings within the household.

The current analysis uses the data for the months of October, November and December of 1995 (months 3-6 of survey data). In order to conduct the analysis, a subset of the sample was eliminated due to missing demographic data and to include only households headed by adults (age 18 or older). The final set of data consists of 2,681 households from the original 3,000 observations.

Two issues are of particular concern in this analysis. First, we consider a statistical issue concerning the random selection of households. In order to select households to be included in the monthly study, an initial draw of 1,000 individuals is made from a database of Estonian citizens. Any Estonian not already included in this demographic study is equally likely to be selected to participate. However, information is collected not only on the individual, but on the entire household. This results in a selection bias toward larger households. Specifically, a household with five members is five times more likely to appear in the sample than a single person household. To compensate for this selection bias, the Estonian Statistical Office formulated weights that indicate the number of Estonians likely to be represented by each individual or each household. Data checks indicate that the final weighted sample is representative of the Estonian population and the weighted sample size is provided in all subsequent tables.

Our second concern regards the value of income obtained from this survey. It is well known that much activity takes place in the underground economy in Estonia (Atkinson and Micklewright, 1992). Although organized crime is a possibility, we expect that an even more common occurrence is the under reporting of general productive activity. Individuals may engage in private production (e.g., growing vegetables) or may earn income in a form not usually counted as money (e.g., bartering services). In addition, the extent of hidden activities may vary over time or by nation and the participation in underground activities may be greatest among those with little income. In order to account for at least part of this unreported income, survey participants were asked to identify any payments they received from any form of home production. These payments (called productive activity in this report) were summed for households. Approximately 12 percent of the sample reported income under this category. The final income values therefore capture some income not normally reported under other government auspices. However, because it is impossible to know exactly the extent of the underground economy in Estonia, we caution the reader to view our results with the understanding that our measure of income may underestimate total income for families, especially those in the lower end of the income distribution.

IV. Our Results: 1995 Income Distribution Estimates for Estonia

Income is defined as monthly disposable personal income per equivalent adult in the home. Disposable income includes all earnings, income from property assets, social transfers, productive activity payments, scholarships and gifts, less taxes. We use equivalent adult rather than per capita because, as noted by Jenkins (1991), "equal per capita incomes do not necessarily reflect equal commands over economic resources." This is particularly true when household members include children who generally do not add income. Because of this problem, when data permits, researchers use some form of household income per equivalent adult household member as the appropriate unit of observation when analyzing income distribution (For example, see Jenkins (1991), Smeeding (1991, 1997) and Wolffe (1987).) We measure equivalent adult as N(5), where N is the number in the household. Smeeding (1997) indicates that this measure has been used in many recent cross-sectional studies.

We first estimate Gini coefficients for total household income; then, we use regression analysis to estimate the head of household demographic effects on total income and its components. Please note that we will use the term "income" and drop "per equivalent adult" for exposition purposes in the remainder of the paper.

Table 1 presents the average monthly income of Estonian households for the fourth quarter of 1995 and the Gini coefficient estimates. Income is presented by education, nationality, gender and age subgroups of the head of household. Weighted values of the sample are listed by each summary statistic.

[TABULAR DATA 1 NOT REPRODUCIBLE IN ASCII]

An Overall Description of Income Distribution. The overall average monthly is 2378.77 kroon (US$208.84) and the Gini coefficient for household income is .290. This measure is in line with an estimate using income per capita of .32 by the Eesti Pank (1997) for the same quarter in 1995. The estimate is very similar to Gini coefficient estimates of other transitional economies. Smeeding (1997) uses a similar measure of income for 1992 (per equivalent adult) and finds that the Gini coefficient ranges from. 189 to .290.

We turn next to the relationship of income inequality to head of household characteristics.

The Influence of Education. Four categories of head of household educational levels are used: basic (some schooling), secondary (the attainment of a degree from a secondary school), some training (either vocational or some college training), and college (at least a college degree). We find that average monthly household income increases with years of education of the head. Table 1 indicates that those heads of household with at least a college degree are associated with more than twice as much household income when compared to those with only basic training, and more than one quarter than those with some training beyond a high school degree.

Within education groups, we find that there is a greater dispersion of household income associated with household heads who have more education. Within group inequality is lowest for those households headed by someone with a basic education (.207), while the other three educational groups have Gini coefficient estimates ranging from .261 to .278. Because the head of household does not necessarily hold the most education, it is difficult to interpret such a finding. However, this result can in part be explained by a larger range of returns as education grows for the main breadwinner of the household.

In addition, the relationship between education and growing income inequality can be attributed to the number of household members who earn income. As shown in Table 2, for those households headed by an individual with a basic education, more than half have none or one earner. It is therefore likely that incomes for households headed by those with little education are similar, especially if those households subsist on social transfers. In contrast, those households headed by individuals with more education are much more likely to have multiple members of the household adding earned income to the household's resources. Approximately 42 percent of those households headed by someone with at least a college degree have two earners; 35 percent and 38 percent of those with secondary degrees or advanced training, respectively, have two earners. The addition of more household earners is likely to increase income disparity in a country such as Estonia, where the opportunity cost of leisure is high and adult household members are more likely to work.

[TABULAR DATA 2 NOT REPRODUCIBLE IN ASCII]

The Influence of Nationality. Nationality of the head of household also seems to play a role in the distribution of average monthly income. We observe that the average monthly household income associated with Estonian heads of household exceeds income for non-Estonian headed households by 252 kroons (US$22.12). In addition, we find that the income of Estonian headed households tends to be less equalized than that of Non-Estonians. The Gini coefficient for Estonian headed households is .306, while for Non-Estonian households it is .248, suggesting a smaller degree of inequality within the foreign cohort.

The influence of nationality becomes more pronounced when we look at the educational attainment of Estonians compared to Non-Estonians in our sample. Almost 48 percent of Non-Estonian heads of household had greater than a secondary school education. In contrast, 43 percent of Estonian heads had some training or college. However, Estonians still have significantly more household income on average.

The Influence of Gender. There are significant differences in household income by the gender of the head of household. In families with a male head of household, average monthly income is 2869.66 kroons (US$251.95); the estimate is 2028 kroons (US$178.05) for households headed by a female. The variation of household income was significantly larger when associated with male headed households, as indicated by the Gini coefficients of .328 and .257 for males and females, respectively.

The gender differences in income are likely due to three reasons First, we find that female headed households tend to have fewer earners than households headed by males. More than 51 percent of households headed by females have zero or one earner, as compared to 21 percent of male-headed households (See Table 2). Thus, households headed by males were more likely to have multiple earners than those headed by females. Thus, at least for household income, part of the difference of income inequality between genders is due to the differences in the economic resources brought in by additional household members.

Second, the distribution of age categories by gender also affects the pattern of income inequality by gender. For both males and females, households associated with the oldest age group for head of household (those over age 59) exhibit the lowest income inequality and lowest monthly household income. The Gini coefficients for households with female and male heads are estimated as. 190 and .259, respectively. However, the size of the effect is much greater for older females due to the relative percentage of household heads who comprise the oldest age group. More than 35 percent of the female heads of household are over age 59; only 23 percent of males fall into this age category. This may be a result of Estonia's migration patterns after World War II when many males relocated (Kahk, 1991). Thus, the distribution of income of households headed by older females lowers income inequality for all female headed households substantially.

Third, although males and females are similarly distributed by educational levels, the household incomes associated with female heads with a basic education are much more equalizing for the income distribution than are the household incomes headed by similarly educated males. Female headed households in this education category have an estimated Gini coefficient of approximately. 16, but the male estimate is about 60 percent higher (.262).

The Influence of Age. Head of household age is also associated with income inequality. We see that average monthly household income is highest for those households headed by an individual in 25-44 age group. This outcome may possibly be the result of transformation processes. In other words, those in the 25-44 age category may be better able to take advantage of the move to a free market system and as the main breadwinner, they bring the household forward economically.

In addition, we see a general trend with increased income inequality within the two youngest cohorts and a drastically more equalized distribution of income among the elderly. As Table 1 indicates, the Gini coefficient for households headed by the youngest group is about 42 percent higher when compared to the oldest group (.299 versus .211). This result is not surprising since the elderly are much more likely to rely on pensions for income which exhibit little variance.

When we review the income distribution of households by the age and gender of the head, we find that households associated with the oldest male cohort have the most equalizing effect on the household income distribution associated with male heads because within group income disparity for the oldest category is relatively low (.259) compared to other age groups. In contrast, the within group income distribution of the youngest females increases overall income inequality for female headed households, although the effect is small due to the small proportion of households in this category.

Household Size and Income. Table 2 presents household income based upon number of household members and earners. The table also shows the breakdown of the three main components of average monthly income: earnings from work, productive activity income (income obtained from entrepreneurial activities, as described earlier) and social transfers (child care, unemployment benefits, pensions, and welfare payments).

Our estimates suggest that households are more likely to rely on earnings than any other component of household income. Small and very large household (more than 6 people) have 19 percent or more of total income from social transfers. Large households also rely on increased productive activity. This is probably due to the smaller average earnings brought in by each individual in these larger households. As Table 2 indicates, the average monthly earnings (per equivalent adult) for households with 6 or more workers is lower than that of 2 to 5 earner households. On the other hand, social transfers are larger for households with many earners, suggesting some equalization through the social safety net.

Regression Estimation Results. To further examine the demographic characteristics of income distribution, a regression analysis was conducted on the natural log of income and on each of the components of income described previously. All analyses are based on household income and head of household characteristics.

Factors included in the regression include variables related to the head of household (gender, nationality, age, age squared, elderly, disabled, education), to the household (the number of children, urban or rural resident, and number of able household members not in the labor force) and to the employment of the head of household (public, self, part time, labor force participation, industry and occupation type). Definitions and descriptions of all variables are in Appendix 1.

The regression results are presented in Table 3. The analysis indicates that female headed households have about 21 percent less in average monthly household earnings than male headed households. Households associated with female heads receive about 20 percent more in social transfers, which does little to reduce total household income differences by the gender of the head. The age of the head of household is associated with higher total family income because earnings, social transfers, and productive income are positively associated with age. Further, elderly headed households receive over 100 percent more in social transfers compared to households headed by the nonelderly. Overall, there is a positive effect on overall household income due to the head of household age, and elderly headed households receive about 13 percent more income than households with heads younger than age 65, holding all else constant.

[TABULAR DATA 3 NOT REPRODUCIBLE IN ASCII]

The number of children under the age of 18 in the home slightly increases household earnings, but decreases household productive activity and social transfers, leading to lower average household income. Household earnings increase by only 1 percent for each child, but each additional child decreases household social transers by 13.7 percent and productive activity by 15.4 percent, leading to an almost 15 percent decline in total household income for each additional child. This result is surprising, given the Estonian government's efforts to aid families with children.

We also study the effect of adults members of the household who are not directly contributing to household income through their labor efforts. Each additional adult not in the labor force (working age, not disabled) decreases average household earnings by 4 percent; however, income from social transfers and productive activity grows as the number of household members not active in the labor force grows, so that the overall affect on total income is that each able adult not formally working leads to an increase of total household income of approximately 10 percent.

Turning to the education coefficients, we find that educated heads of household are associated with higher household earnings and productive activity income than those heads with a basic education. What is interesting is the smaller effect on household earnings for those heads who have received some advanced training relative to those with a secondary degree. Although households headed by individuals with advanced training have 12 percent less in earnings than those headed by individuals with a secondary education, these households make up for the difference with a higher return to productive activity and with slightly more in social transfers, thereby advancing overall household income. As to be expected, those households headed by an individual with at least a college degree have the highest earnings and total income.

In reviewing other head of household characteristics, we find that a disabled head of household leads to lower household earnings and productive activity when compared to the estimate for households headed by someone not disabled. However, the loss in earnings is somewhat covered by social assistance, and the overall income difference between households with disabled and nondisabled heads is approximately 12 percent. Unemployed heads of household are also associated with lower earnings when compared to those working. Even though social transfers and productive activity are equalizing, households with an unemployed head still have 51 percent less in total household income than those with an employed head.

Finally, we review industry and occupational differences, as well as city and urban differences. We find that heads of household who are in agricultural industries are associated with lower household earnings, social transfers, and productive activity income, when compared with other industry groups. This result is exacerbated when we consider that urban dwellers receive approximately 10 percent more in household income than rural dwellers. Mining and construction workers are associated with the largest returns to household earnings, which translates into higher household income. Regarding occupations (professional, skilled labor, service workers, and unskilled workers), we find that heads of household in the professional fields are associated with larger household earnings than the other three occupation categories; therefore, those households have relatively higher total income. Surprisingly, households headed by skilled workers have lower average earnings than those headed by the unskilled, but they have about a 26 percent return to productive activity as compared to the unskilled group, leading to a difference in household income between the two groups of more than 9 percent. This suggests that households headed by skilled workers use informal market methods to increase overall household income.

Overall, it appears that, on net, the programs designed to aid those households with individuals who cannot work, particularly for the elderly, have somewhat of an equalizing effect on household income. These programs, however, have not tackled the issues of a lack of education or regional differences, nor have they done enough for households with children or for households with a disabled head. As with many transitional post-Soviet economies, Estonia faces a difficult challenge as the country dismantles collective farms and as economic development centers on the main cities. The rural poor have low income, not only because of a lack of skills of the main breadwinner, but also because job opportunities are situated in urban sites.

IV. Conclusion

We have investigated the Estonian income distribution for the fourth quarter of 1995. We find that many of the trends evident in other market-oriented countries exist in Estonia. Our results indicate that household income inequality in Estonia is relatively similar to that of other transitional economies. Households headed by males tend to increase the dispersion of the income distribution, while female headed household income tends to be equalizing. Part of this outcome can be explained by the equalizing effect of incomes of households with elderly female heads, the number of female-headed households that rely on income other than earnings, and the tighter income distribution of households headed by females with basic education as compared to other age/ gender groups. Income disparity is larger for households headed by native Estonians as compared to Non-Estonians and those households with educated heads. We find that the incomes of the elderly headed households decrease overall income inequality. Finally, our results indicate that the size of the household and the number of earners affects the average monthly income and its distribution.

These trends are to be expected. As Estonia moves from a centrally planned to a market economy, income inequality is expected to rise. Those individuals with the ability to adapt to structural economic change do so more quickly than whose who lack the resources and talents to move into the new economic system. Households headed by the elderly and, to a lesser degree, the lower-educated, reduce overall income inequality in Estonia; however, it is likely that this is due to a lack of adaptability to economic change by the main breadwinners in these homes. The fact that those households associated with low income inequality often have few earners in the household suggests that the need for outside income may be due to the inability to keep up with transitional processes. The regression analysis confirms this problem and suggests that the social programs have reduced income inequality mainly for the households headed by the elderly. However, we still find a significantly large difference in household earnings and income for those in the agricultural industries and for those who live in rural areas.

Notes

(1.) Even for households with more than three people, the head's earnings comprise more than one half of total family earnings and about 45 percent of total family income. Estimates are available upon request of the authors.

(2.) A Gini coefficient measures the distance between perfect income inequality (represented on a Lorenz diagram by a 45 degree line) and the true income distribution. A Gini coefficient of 0 would, therefore, indicate perfect income equality (i.e., 10 percent of the population holds 10 percent of total income and so on). If a single individual holds 100 percent of the society's total income, then the Gini value would be 1.0. See the discussion in Miller, Economics Today, pp. 709-710, for a detailed explanation.

(3.) The Statistical Office of Estonia is very careful in its collection of data However, as with all new survey instruments, it is possible that some problems exist either in data collection or compilation. The Statistical Office noted several problems, in particular with the weight assignments described in the next paragraph, and subsequently made changes after the first month of the data collection. Therefore, we used only the data from the fourth quarter of 1995. Ideally, a full year of data would be most appropriate for income distribution analysis and the authors received the raw 1996 data in the fall of 1997. However, after several months of data cleaning, we find that the 1996 data are still not ready for analytical purposes. Although precise income distribution estimates will be improved with annual data, we believe that the main thrust of this report (association of head of household characteristics to income distribution) is captured by the fourth quarter data analysis employed in this report.

(4.) Any measure of `equivalent adult' has some arbitrary assumptions underlying it, and we elected this measure because it is used by current researchers and because of its simplicity. We also performed the analysis using income per capita and per adult. These estimates did not change the general relationship of head of household characteristics to the income distribution, particularly when comparing per capita and per equivalent adult estimates.

(5.) The average exchange rate for US-Estonia currency in the fourth quarter of 1995 was 11.39EEK/$.

(6.) The estimates are: Poland, .290; Hungary, .289; Czech Republic, .207; and Slovak Republic, .189. See Smeeding (1997).

(7.) Of course, this group could also include the households most likely involved in productive activity not reported to the government. As we note earlier in the paper, we control for informal productive activity as reported in the HIES and we caution the reader to consider this element in interpreting our results.

(8.) The frequency distribution of education and nationality is available from the authors.

(9.) Studies of earnings and income often use a log linear model to account for heteroskedasticity. Regression coefficients can be interpreted as percentage changes due to a unit change in the independent variable.

(10.) Taking the partial derivative with respect to age suggests that the age squared term has no economic effect on income or its components.

References

Ahde, P. and Teet Rajasalu. 1992. "On the Economic Structure of Estonia and Finland Before the 1990s" Research Institute of the Finnish Economy Discussion Paper.

Atkinson, Anthony and John Micklewright. 1992. Economic Transformation in Eastern Europe and the Distribution of Income. Cambridge: Cambridge University Press.

Blinder, Alan S. 1980. "The Level and Distribution of Economic Well Being" in M. Feldstein, ed., The American Economy in Transition. Chicago: University of Chicago Press.

Eesti Pank (Bank of Estonia) Bulletin No. 2, 29, 1997.

Gottshalk, Peter and Timothy M. Smeeding. 1997. "Cross-National Comparisons of Earnings and Income Inequality." Journal of Economic Inequality, 35,2, pp. 633-681.

Jenkins, Stephen. 1991. "The Measurement of Income Inequality" In L. Osberg, ed. Economic Inequality and Poverty: International Perspectives. London: M.E. Sharpe, Inc.

Kahk, J. 1991. World War Two and the Soviet Occupation in Estonia: A Damages Report. Tallinn: Perioodika Publishers.

Levy, A. and R.J. Murname. 1992. "US Earnings Levels and Earnings Inequality: A Review of Recent Trends and Proposed Explanations." Journal of Economic Literature, 30, 3, pp. 1333-81.

Lugus, Olev and P. Vartai. 1993. Estonia and Finland - a Retrospective Socioeconomic Comparison. Helsinki. The Research Institute of the Finnish Economy.

Miller, Roger. 1994. Economics Today. Eighth edition. Harper Collins College Publishers: NY NY.

Mookherjee, D. and A. Shorrocks. 1982. "An Income Decomposition Analysis of the Trend in UK." Income Inequality Economic Journal, 92, pp. 886-902.

Osberg, Lars, ed. 1991. Economic Inequality and Poverty: International Perspectives. London: M.E. Sharpe, Inc.

Rajasalu, Teet. 1994. Estonian Economy: Summer 1994. Tallinn: Estonian Academy of Sciences, Institute of Economics.

Smeeding, Timothy M. 1991. "Cross-National Comparisons of Inequality and Poverty." In L. Osberg, ed. Economic Inequality and Poverty: International Perspectives. London: M.E. Sharpe, Inc.

--. 1997. "The International Evidence on Income Distribution in Modern Economics: Where Do We Stand?" In Jon Neill, ed., Poverty and Inequality, The Political Economy of Redistribution. Kalamazoo, MI: W.E. UpJohn Institute for Employment Research.

Venesaar, Urve and George A. Hachey. 1995. Economic and Social Changes in the Baltic States in 1992-1994. Tallinn: Estonian Academy of Sciences, Institute of Economics. World Bank, Policy Research Department.

Welfare During the Process of Transition Research Paper Forthcoming. World Bank, World Development Report Team. 1996 World Development Report.

Wolffe, Edwards N. 1987. ed. International Comparisons of the Distribution of Household Wealth. New York: Oxford University Press.

Appendix 1
Descriptive Statistics of Regression Variables

Variable Mean

Dependent Variables:
Household After Tax Earnings 1234.18
Household Social Transfers 428.58
Household Productive Activity 111.70
Household Disposable Income 2376.46
LnEarnings 4.84
Ln Social Transfers 4.50
Ln Productive Activity 0.69
Ln Income 7.52
Demographic Characteristics
Gender (1 if head is female) 0.58
Nationality (1 if head is Estonian) 0.68
Age (age of head) 49.50
Age Squared 2717.94
Elderly (1 if head is over age 65) 0.27
Disabled (1 if head is disabled) 0.02
Secondary Education 0.27
Some Training Beyond Secondary 0.30
College+ 0.15
Employment Characteristics
Unemployed (1 if head is unemployed) 0.36
Public (1 if head works in a publicly owned facility) 0.30
Self(1 if family has self-employment income) 0.04
Part Time (1 if head has part-time earnings) 0.05
Industry classification of head:
Agriculture, Forestry or Fishing 0.08
Mining or Construction 0.06
Manufacturing 0.12
Utilities, Communications, or Transportation 0.10
Service 0.15
Occupational classification of head:
Professional 0.23
Service Worker 0.12
Skilled Labor 0.18
Household Characteristics
Number of adults not in the labor force 1.22
Number of Children 0.58
City Residence (1 if an urban household) 0.70

Variable Standard Deviation

Dependent Variables:
Household After Tax Earnings 1615.25
Household Social Transfers 423.33
Household Productive Activity 717.69
Household Disposable Income 2288.01
LnEarnings 3.45
Ln Social Transfers 2.72
Ln Productive Activity 1.98
Ln Income 0.70
Demographic Characteristics
Gender (1 if head is female) 0.49
Nationality (1 if head is Estonian) 0.47
Age (age of head) 16.37
Age Squared 1663.22
Elderly (1 if head is over age 65) 0.44
Disabled (1 if head is disabled) 0.15
Secondary Education 0.44
Some Training Beyond Secondary 0.46
College+ 0.36
Employment Characteristics
Unemployed (1 if head is unemployed) 0.48
Public (1 if head works in a publicly owned facility) 0.46
Self(1 if family has self-employment income) 0.20
Part Time (1 if head has part-time earnings) 0.22
Industry classification of head:
Agriculture, Forestry or Fishing 0.28
Mining or Construction 0.24
Manufacturing 0.33
Utilities, Communications, or Transportation 0.29
Service 0.36
Occupational classification of head:
Professional 0.42
Service Worker 0.38
Skilled Labor 0.38
Household Characteristics
Number of adults not in the labor force 1.00
Number of Children 0.49
City Residence (1 if an urban household) 0.46

Source: The Household Income and Expenditure Survey (HIES), the Statistical Office of Estonia. Income per equivalent adult is measured as household disposable income/household size**.5. The weighted sample size for the regression analysis is 606,332.

Lisa Wilder, Assistant Professor
Bowling Green State University

Mary Ellen Benedict, Associate Professor
Bowling Green State University

Marie Viies
Institute of Economics

(*) Thanks to an anonymous reviewer and session participants at the 1997 Missouri Valley Conference, to Reet Madre, Teet Rajasalu and to the staff at the Statistical Office of Estonia. Send all correspondence to Mary Ellen Benedict at the above address.3