First Draft
Comments Welcome
Inequality and Growth in Transition:
Does China’s Rising Inequality Portend Russia’s Future?
Pradeep Mitra
[email protected]
Ruslan Yemtsov
[email protected]
World Bank
January 2006
This paper has been prepared for the Annual Bank Conference on Development Economics
(ABCDE) Beyond Transition at St. Petersburg, Russia on January 18 and 19, 2006. Views
expressed are the authors’ and do not necessarily represent those of the World Bank.
Abstract
This paper reviews the literature on inequality dynamics in the transition to a market
economy in order to answer the question about the relevance of China’s fast growing
inequality for transition economies in Eastern Europe and the former Soviet Union. It relies
for its empirical analysis on data produced for the recent World Bank study “Growth,
Poverty and Inequality in Eastern Europe and the Former Soviet Union: 1998-2003” [World
Bank (2005b)] and its predecessor "Making Transition Work for Everyone: Poverty and
Inequality in Europe and Central Asia” [World Bank (2000)].
The paper shows that an increase in inequality in transition, as predicted by a number
of theoretical models, in practice differed across countries, with inequality evolving at
different speeds depending on the relative importance of its key drivers, viz., changes in
wage distribution, employment, entrepreneurial incomes and safety nets. In some countries,
the process of reaching a new, higher long-term level of inequality is taking longer than
expected; while in others, there has been a clear overshooting of inequality levels compared
to the likely final outcome of the transition process.
Available data suggest that rapid increases in wage inequality in Russia, driven in part
by arrears, with some moderation in the later years of the transition, have so far determined
the pace of inequality dynamics. In that respect, Russia is different from economies in
Central and Eastern Europe, where changes in employment and compensating adjustments in
the social safety nets have played a greater role. We also contrast key features of inequality
in Russia in the context of other transition economies and compare them to trends in
inequality observed in China, arguing that the latter’s experience of rapidly rising inequality
is to a large extent a developmental, rather than a transitional phenomenon, and therefore has
limited relevance for predicting changes in inequality in Russia.
The paper attempts to assess the likely evolution of Russian inequality in the future.
Given current underlying structural characteristics of these economies, it argues that the
process of transition to a market economy in Eastern Europe and the former Soviet Union has
not yet been completed, and transitional dynamics, as well as, increasingly, other global
factors will influence further evolution of inequality.
To make a more informed judgment on possible long-term paths of inequality
dynamics in Russia, the paper applies the framework developed in the World Development
Report 2006 on “Equity and Development” [World Bank (2005a)]. The discussion of “nontransitional” components of inequality, such as equality of opportunities, preferences for
equality or global drivers of inequality in the context of Russia and other countries in the
Region finds that they do not all point in one direction. Some factors, such as a preference
for greater equality or universal access to public services, tend to moderate inequality
through a variety of channels, while others, such as increases in the returns to skills in a
global economy, or constraints to migration tend to magnify it. While discussing the
possible outcomes of these complex interactions, it is important to remember that many of
these long-term factors are not given once and for all; and that they can be influenced by
policy.
2
Introduction
Since 1999, regional output growth rates in Eastern Europe and the former Soviet Union
(ECA) have consistently exceeding the world average. The prospects to accelerate growth to
match the extremely strong performance of China do not seem implausible, especially for the
fast growing resource rich CIS economies. Figure 1 depicts GDP per capita evolution in
purchasing power parity terms for China and Russia and hints at growth in Russia since 1999
having resembled the pace of GDP change in China.
Figure 1. China and Russia: GDP per capita in PPP (2000 $)
China
Russia
12,000
10,000
$ per capita
8,000
6,000
4,000
2,000
0
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Source: The World Bank WDI (2005)
Deep transitional recession in the CIS during the 1990s was associated with sharply
increasing inequality (Svejnar [2002]). Meanwhile, inequality rose significantly in the fast
growing East Asia region: in Vietnam the Gini index, the most commonly used measure of
income inequality, grew by 2.3 percent a year between 1993 and 2002, while in China it went
up by 2.0 percent a year between 1990 and 2001 (World Bank [2005e]). For countries in
Eastern Europe and the former Soviet Union which share a common legacy with former
command economies of China and Vietnam, this may be a harbinger of things to come. Will
improved economic performance in Russia come at the price of a further widening of income
disparities?
Inequality is an important determinant of poverty in any country, and it is also a significant
factor which influences the dynamics of poverty and its responsiveness to growth. The higher
is inequality the lower is the effect of growth on poverty. On examining variations in
changes of poverty levels across a sample of developing countries for the 1980s and 1990s,
Kraay (2004) found that growth accounted for just over two-thirds of the changes in relative
incomes, while distributional change accounted for the rest. The impact of inequality
changes on poverty becomes much larger if one examines shorter time intervals
(Bourguignon [2003]).
3
But inequality is also essential to the operations of market economy by providing incentives
for effort and risk taking. It is therefore important to separate inequality of outcomes from
inequality of opportunities (WDR 2006). Countries in transition which have emerged from a
history of repressed inequality face a complex task of finding the right balance between the
growth-enhancing aspects of inequality while managing the scale of inequality of
opportunities. In carrying out that task, it is important to understand the drivers of inequality,
and assess direction of trends.
This paper focuses on Russia. It first compares Russia to other countries in ECA in terms of
inequality.1 The paper then looks at the available evidence on inequality in China. The
treatment of China and ECA countries in this paper is somewhat different. For the most part,
primary micro data are used for ECA to assess and decompose trends in inequality. Only
published sources are used for China: in terms of data availability China remains far a field
for the scholars.
In terms of definitions, the paper follows the mainstream literature by looking at inequality as
measured by the indicator as closely related to the standards of living as possible. The paper
relies on recent advances in the measurement of inequality in ECA, as exemplified by the
World Bank regional study ‘Growth, Poverty and Inequality in Eastern Europe and the
Former Soviet Union: 1998-2003’ (see Annex 6). However, such comparable indicators are
available for only a limited subset of countries and years. The paper is organized in four
sections. Section One sets the stage by introducing data and providing a broad picture of
changes in inequality. Section Two summarizes insights from theoretical models of
transition. Section Three discusses main decompositions results. Section Four concludes
with outlook for inequality changes, policy implications and further research agenda.
I.
Stylized facts about inequality in transition
Former planned economies had suppressed inequality, and the transition to a market
economy was associated with a set of factors which had the potential to rapidly increase the
variation in individual incomes and consumption. How far can the transitional increase in
inequality go? Are countries now, 15 years after the start of transition, converging to some
common level of inequality that prevails in the long-run in market economies? 2
Different sets of inequality data from transition countries support conflicting views on these
issues and it is necessary to venture into the details of inequality measurement to provide
meaningful answers. International comparisons of inequality are always prone to gaps (see
Atkinson, 2003 for a summary). It is therefore important to make sure that likes is compared
1 In this paper “ECA Region” is a group of 27 ex-socialist countries of Eastern Europe and the Former Soviet
Union. These countries with a few exceptions (Albania and countries of the former Yugoslavia) formed the
“Soviet bloc”.
2 Ravallion (2001), quoting Benabou (2000) argues that countries are expected to converge to the same
distribution and proposes a test for such convergence, but due to data limitations transition economies have not
been fully incorporated in his analysis.
4
with like. A concerted effort has been made to ensure comparability for data reported on
Figure 2 (see Annex 6).
Figure 2 suggests that median inequality in the ECA region is lower than in the rest of the
developing world. However, it is broadly comparable to the median inequality in OECD
countries. Figure 2 also suggests that ECA region shows the full spectrum of inequality
outcomes, from fairly unequal to fairly equal. Despite an apparently common legacy and
common circumstances of transition, the outcomes appear to show great variation across
countries.
Figure 2: Consumption Inequality in China, Russia and ECA in an International
Perspective
0.60
LAC
AFR
0.45
MENA
EAP
SAR
ECA
HIC
PAK
BGD
LKA
NPL
INDu
ISR
YEM
DZA
EGY
JOR
DJI
MAR
IRN
TUN
NIC
JAM
BRZ
PAN
MEX
ITA
FRA
TWN
UK
ISR
ESP
US
GRC
BLR
HUN
AZE
UKR
BGR
KGZ
ROM
ALB
POL
RUS
TJK
LTU
KAZ
ARM
MDA
GEO
IDN
TUR
VNM
MNG
LAO
TMP
CHNu
PHL
THA
KHM
ETH
MUS
BEN
NER
MLI
MRT
GHA
BFA
SEN
CIV
KEN
CMR
ZAF
0.15
MDG
0.30
Note: HIC stands for High Income Countries which includes France, Greece, Israel, Italy, Spain, Taiwan,
UK and USA; Russia and China (only urban data used for reasons explained in WDR 2006) are highlighted.
Source: World Bank, WDR 2006 and Inequality in LAC, and Luxembourg Income Study Working Papers 338
and 341 for HIC countries.
•
Why has inequality increased in very different degree across countries in
transition? What implications does this have for where these countries will end up
in the long run?
5
If we were to use similar data from China on this graph, we will observe it fitting one of the
highest peaks (with consumption Gini by 2002 of around 0.45), but most researchers agree
that the official definition of consumption in China does not allow direct comparisons with
other countries.3 But there is little doubt that China fits into higher inequality cluster (Gini of
0.4 for per capita incomes), exhibiting levels similar to more unequal countries in ECA in
mid-1990s and 1990s (Table 1 and Annex 1, Table A1.3).
China has experienced a far longer period of growth than the ECA countries and it also
experienced a period of rising inequality (Figure 3). By 2002 real GDP per person in China
was 4 times above the level observed in 1985. ECA countries as a group just returned to the
pre-transition level of GDP per capita just in 2004 [World Bank (2005b)]. This leads us to
the second question which motivates this paper:
•
Would faster growth in transition countries in ECA be accompanied by increasing
inequality on a scale similar to China’s?
Atkinson (2003) shows that transition economies were not alone in experiencing growing
inequality: there has been significant change in the distribution of income in many OECD
countries. Figure 3 shows the extent of inequality increases in China, Russia, US, UK,
Mexico, Sweden, Poland, Georgia and Hungary, using for each country what is believed to
be the most reliable indicator of living standard dispersion.
3 Chinese Statistical Office makes all investment expenditure (including new housing construction and durable purchases)
part of the "consumption expenditures". It also fails to properly account for regional price differences. This biases
consumption Gini upward and makes Gini index for consumption not comparable to data from elsewhere. Ravallion and
Chen (1996) using the survey micro data for from 4 provinces to correct for cost-of-living indices, conduct simulations to
purge the consumption from these items, and show that Gini index for properly calculated consumption is lower by as much
as 4 points for pooled household data from these 4 provinces (from 29.4 in 1990 to 24.8).
6
Figure 3. Gini index changes between 1980s and 2000s
Mid 1980-s
Early 2000s
50
Gini index
40
30
20
China
Russia
Hungary
Poland
Georgia
Mexico
Austria
Sweden
USA
United
Kingdom
Source : ECAPOV2, Ravallion and Chen for China, LIS (www.lispoject.org) for other countries.
Note levels of Gini index are not comparable across countries as different concepts and definitions of welfare
are used. In ECA for 2002s: current consumption per capita without housing rental values, correcting for
regional price differences and without outliers. Data for early 1980s come from published sources and refer to
total expenditures, not correcting for price differences. China – total incomes per capita, correcting for price
differences. OECD – per equivalent adult total money incomes, correcting for regional price differences without
outliers.
Figure 3 demonstrates fairly well that inequality increase in transition economies occurred
against a backdrop of a global increase in inequality (with however important variations
across countries, Atkinson 2003). The increase in inequality in transition economies
coincided with similar increases around the world. This leads us to the third question
motivating our analysis:
•
Is rising inequality in ECA and in China primarily a reflection of global factors or
it is a transitional phenomenon?
Answers to these three questions would help us to understand prospects for inequality
changes. To get responses we need good data. Figures 2 and 3 are based on datasets that
often give one or just two points per country. They therefore ignore potentially complex
dynamics which may shed some light on three questions this paper is concerned with.
Published official data on distribution of income provide extended data series.
Table 1 presents the evidence using UNICEF dataset of social statistics based on national
published sources and data for China painstakingly constructed by Ravallion and Chen using
published tabulations for household survey by the State Statistical Bureau (SSB).
7
Table 1. Gini indices for per capita incomes
Armenia
Azerbaijan
Belarus
Bulgaria
Croatia
Czech Rep
Estonia
Georgia
Hungary
Kazakhstan
Kyrgyz Rep.
Latvia
Lithuania
Macedonia
Moldova
Poland
Romania
Russia
Slovenia
Slovak Rep.
Tajikistan
Turkmenistan
Ukraine
Uzbekistan
China
19871990
0.269
0.345
0.233
0.245
0.251
0.197
0.240
0.313
0.214
0.297
0.308
0.240
0.248
0.349
0.267
0.255
0.232
0.259
0.220
0.186
0.334
0.308
0.240
0.351
0.289
1991
1992
1993
1994
1995
1996
1997
1998
1999
0.570
2000
0.344
0.280
0.340
0.440
0.253
0.384
0.244
0.357
0.249
0.366
0.235
0.326
0.228
0.395
0.270
0.350
0.258
0.370
0.230
0.361
0.430
0.254
0.253
0.345
0.333
0.239
0.354
0.231
0.242
0.330
0.246
0.35
0.353
0.310
0.350
0.265
0.274
0.260
0.227
0.289
0.282
0.365
0.285
0.398
0.347
0.369
0.360
0.290
0.409
0.250
0.320
0.312
0.381
0.328
0.302
0.375
0.302
0.237
0.470
0.326
0.309
0.367
0.420
0.334
0.305
0.381
0.305
0.249
2002
2003
0.428
0.247
0.332
2001
0.537
0.373
0.245
0.333
0.246
0.370
0.212
0.361
0.232
0.389
0.231
0.385
0.237
0.393
0.250
0.253
0.259
0.272
0.267
0.249
0.351
0.29*
0.234
0.402
0.469
0.268
0.411
0.321
0.332
0.399
0.414
0.327
0.355
0.377
0.326
0.298
0.398
0.298
0.262
0.334
0.299
0.399
0.299
0.249
0.470
0.437
0.345
0.310
0.394
0.310
0.264
0.435
0.341
0.353
0.396
0.353
0.263
0.382
0.358
0.357
0.34*
0.436
0.353
0.349
0.398
0.22*
0.267
0.342
0.379
0.318
0.34*
0.411
0.356
0.352
0.404
0.22*
0.299
0.282
0.288
0.290
0.277
0.271
0.364
0.385
0.395
0.343
0.354
0.360
0.331
0.342
0.367
0.330
0.376
0.365
0.351
0.350
0.354
Source: Data from UNICEF TRANSMONEE 2005 edition [www.unicef-icdc.org/research], except for selected countries and years form
ECAPOV I, Milanovic (1997), Poverty Assessments for Aremnia, Georgia, Uzbekistan, Ukraine, Tajikistan, and Eurostat (2005). Note: For
Russia 1992 and earlier years data refer to total incomes, for later years – only to Money incomes; * data are form Eurostat and rely on a
OECD per equivalent equivalence scale. China- data from Ravallion and Chen (2004).
The data show a rapid increase in inequality in the middle-income and low-income CIS,
followed by some moderation. The new member states of the European Union (the EU-8),
on the other hand, seem to experience a more gradual but steady increase. Table 1 also
demonstrates a wide range of dynamics within each group. For example, the Baltic States
experienced inequality paths similar to Russia’s, while Belarus resembles more those of the
EU-8.
By construction, data in Table 1 are not fully comparable. But to what extent they are?
Annex 1 illustrates a wide range for the inequality estimates for Russia, Hungary, Poland and
China coming from different sources and using different definitions of welfare. There are
also well known biases regarding officially published data for the pre-transition and early
transition years (Atkinson and Micklewright 1992). Published data on income distribution
should be treated with great care for at least six different reasons:
8
First, published data from different countries rely on different imputation and adjustment
procedures. In Ukraine, for example, significant and rather unusual imputations are
undertaken with the reported in-kind components. In some countries total incomes include
imputed rents, whereas in others they do not. As demonstrated in Annex 1 this might have
large impact on the level of inequality. Table 2 provides a summary for China and Russia
showing that levels, trends and direction of change can be affected.
Table 2 Effect of housing rent imputations on income Ginis for China and Russia
Urban China
No owner occupied housing rents imputations
Imputing owner occupied housing rents+ subsidies
Russia
1995*
2001/2**
1992***
2003****
0.283
0.332
0.323
0.318
0.417
0.354
0.436
0.338
Sources: * Ravallion and Chen (2004) based on SSB 1995 and 2001, official survey ** Khan and Riskin (2004) based on
CASS 1995 and 2002, large unofficial survey, *** Russia 1992 data are from the RLMS calculated by Buckley and
Gurenko(1998), and 2003 by Ovcharova et al and based on NOBUS survey See ANNEX 1 for detailed list of sources.
Clearly, controversy surrounding Chinese data with and without housing rents and subsidies
imputations offers an important lesson for ECA countries and for Russia in particular. Own
housing is an important factor influencing the distribution of living standards and its
omission from the measurement distorts the data and policy analysis.
Second, in all EU-8 economies, wages account for over 60 percent of household incomes. In
contrast, among the low income economies of the CIS, wages represent less than 15 percent
in some cases. State transfers are a much more important component of income in EU-8,
where they comprise 25 to 30 percent of total incomes; their importance among the
economies of the low income CIS has shrunk drastically: state transfers represent less than 10
percent in Moldova and Georgia. Such compositional effects have serious implications for
the accuracy of measuring inequality. Both wages and transfers can be measured quite
accurately by household surveys. In contrast, other sources of income (more dominant in the
low income CIS) are notoriously hard to measure. To the extent that these other sources of
income are measured with large error by survey data, we should discard data presented in
Table 1 for low income CIS: Armenia, Georgia, Kyrgyz Republic and Moldova.
Third, there are also serious issues of under-reporting and non-response. Richer households
tend to be increasingly missed by sample surveys. Countries practice different degree of
adjustments to correct for non-response. Such corrections rely on a number of assumptions
which can undermine comparability. In Russia, unlike in any other country, the increasing
gap between reported incomes and estimates from macroeconomic sources is arbitrarily
assigned to the top decile of households as “undeclared” incomes (World Bank 2005d),
limiting comparability to any other income distributional data.
9
Fourth, correction for regional price differences is not a normal practice in many statistical
offices.4 Fifth, the use of equivalence scales is also not converging towards a single
standard.5 Finally, as pointed out already, the definitions of income vary greatly. These
factors combined can easily cumulate to differences in the value of Gini index over 5
percentage points across countries and time periods.
Despite these limitations data reported in Table 1 are often used to produce “stylized facts”
on inequality dynamics in transition (Ivashenko, 2002) and draw far-reaching conclusions.
The lack of consistency of the existing data prompted the project which aimed at the creation
of comparable and consistent inequality statistics across ECA countries based on household
consumption per capita (see Annex 6). Results are presented in Table 3. We make a
distinction between consistent data obtained from micro datasets and data on consumption or
expenditure inequality from other sources which are not as comparable.6 Table clearly
shows that there are discontinuities, and the evidence is of variable quality. But overall the
data tell a very clear story: (1) there were rapid increases in inequality in middle-income CIS,
followed by some stabilization, or even moderation afterwards; (2) there was much more
gradual increases in CEE with continued change to the most recent observation point; (3)
there was a wide diversity of experience, even among countries within the same grouping.
It is important to note here that comparing data in Tables 3 and 1 that income-based and
consumption-based measures of inequality appear to be fairly consistent in some cases,
typically in EU-8. In the low income CIS, and in some middle –income CIS and SEE they
are clearly not, and we should rely primarily on consumption-based measures of distribution.
4 When such corrections arte practiced, they tend to reduce inequality as measured by Gini by between 1-3
percentage points. Ravallion and Chen (1996) and Chen and Wang (2002) demonstrate profound impact of cost
of living (COL) adjustment on measures of inequality for China.
5 The use of Eurostat equivalence scale rather than per capita with ECA household structure typically reduces
the value of Gini index by about 2 percentage points.
6 Because for earlier transition years available micro data almost exclusively represent already computed
indicators, and no primary detailed records survived, it was impossible to recreate comparable consumption
figures from scratch.
10
Table 3. Gini index for comparable per capita consumption indicator
Country
Albania
Armenia
Azerbaijan
Bosnia
Belarus
Bulgaria
Croatia
Czech Republic
Estonia
Georgia
Hungary
Kazakhstan
Kyrgyz Republic
Latvia
Lithuania
Macedonia
Moldova
Poland
Romania
Russia*
Serbia
Slovenia
Tajikistan
Turkmenistan
Ukraine
Uzbekistan
19881992
19931995
1996
0.291
0.444
1997
1998
1999
2000
0.321
0.360
0.228
0.234
0.228
0.194
0.230
0.28
0.210
0.257
0.260
0.225
0.224
0.287
0.283
0.266
0.395
0.291
0.241
0.235
0.255
0.238
0.343
0.264
0.282
0.395
0.236
0.292
0.293
0.350
0.258
0.3700
0.232
0.327
0.537
0.310
0.373
0.299
0.353
0.523
0.316
0.323
0.340
0.268
0.404
0.405
0.317
0.277
0.353
0.279
0.248
0.376
0.386
0.250
0.313
0.393
0.259
0.339
0.397
0.254
2001
0.325
0.365
0.263
0.301
0.337
0.311
0.360
0.336
0.303
0.346
0.299
0.332
0.383
0.251
0.346
0.290
0.304
0.306
0.305
0.371
0.296
0.274
0.369
0.365
0.302
0.283
0.357
0.350
0.305
0.282
0.349
0.357
0.307
0.286
0.339
2002
0.319
0.310
2003
0.285
0.295
0.292
0.277
0.335
0.390
0.250
0.330
0.292
0.340
0.305
0.3678
0.345
0.320
0.294
0.338
0.292
0.330
0.391
0.318
0.276
0.350
0.325
0.3732
0.328
0.289
0.284
0.289
0.327
0.410
0.233
0.250
0.325
0.333
0.285
0.453
0.293
0.303
0.355
0.274
0.326
0.268
0.354
Sources: Figures in bold are from ECAPOV II, in italics – direct survey data estimates from other source
(ECAPOV I and PAs), other data are from WDI and Milanovic and are based on grouped data. Data for Poland
in italic are from Keane and Prasad (consumption per capita without durables) and refer to 1990 for 1989-1992,
Only figures from ECAPOV 2 are consistent across time Notes: * based on HBS.
In China, as demonstrated in Annex 1, Table A1.3, the range for different Gini estimates is
also wide. In addition, it is widely believed that urban inequality indices for China are
underestimated. This is because urban sample of the national survey includes only
permanent residents and migrants with permits (hukous) registered in urban areas. The
estimates of unregistered migrants differ and go up to 150 mln [check]. They are believed to
earn significantly lower salaries and their omission from the sample definitely underestimates
urban inequality.
Despite these limitations several important observations can be made using long time series
available to measure inequality in China. The evolution of inequality in China vis-à-vis
changes in real incomes (adjusted for cost of living differences) is portrayed on Figure 4
using extended time series constructed by Ravallion and Chen (2004).
11
Figure 4 China: Real per capita income and Gini index, 1981-2001
Real Mean Per Capita Income
Gini index
350
300
Index, 1981=100
250
200
150
100
50
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
Source: Ravallion and Chen (2004)
The first important observation is very uneven raise of the inequality with periods when there
were important reversals. Growth in China does not drive in itself higher inequality
(Ravallion and Chen 2004). Thus increasing inequality in China does not foretell raises in
income disparities in Russia just because of growth acceleration in the latter. There are other
forces at work, which are discussed below.
II.
Inequality in transition: insights from the theory
As pointed out by Flemming and Micklewright (1999) transition is "a dynamic process of
resource reallocation as well as a politico-economic process of reform, restructuring and
institution building". We start with a short description of drivers believed to affect the
distribution in transition. We then summarize predictions from highly stylized models of
transition to move to empirically-based explanations.
Main drivers of inequality in transition
The lost of eight major sets of factors affecting inequality in transition is given below:
• Wage decompression and the growth of private sector
• Restructuring , unemployment, reverting to subsistence economy
• Fiscal adjustment affecting Government expenditure and taxation, corruption
• Price liberalization, inflation and arrears
• Asset transfer, growth of property income
• Technological change and globalization
• Demographic transition and increased mobility
• Political process with closer reflection of preferences for equality
12
The first five factors are specific to transition, the latter three channels are drivers common to
all countries in the world. Their effects are usually long term. But as transition unfolded as a
long-term change process, the factors specific to transition interacted actively with each other
and with long-term drivers.
Driver 1. Changes in wage setting and the growth of private sector
Liberalization and decompression of wages is thought to be the major driver of inequality
dynamics in transition. This is because wages accounted for between 60 and 70 percent of
household incomes at the start of transition, and their distribution under the command system
was compressed. Among other consequences, this severely distorted returns to education.
Transition has changed wage setting in a number of key respects, reflecting tighter link
between productivity and wages and decentralization of wage setting. The most important
driver was the emergence of large private sector competing for workers with the State owned
firms. Very approximately, it appears that by 2004 [EBRD 2005] over 60 percent [check] of
GDP was produced in the private sector, with somewhat higher shares of private sector in
total employment. Table 4 below reports available data on Gini indices for wages based on
firm records.
Table 4 Gini coefficients for monthly wages in ECA, 1989-2001
EU-8
Czech Republic
Estonia
Hungary
Latvia
Lithuania
Poland
Slovenia
SEE
Serbia
Romania
Middle income CIS
Belarus
Russian Federation
Ukraine
Low income CIS
Armenia
Azerbaijan
Georgia
Kyrgyzstan
Moldova
1989
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
0.204
0.253
0.268
0.244
0.260
0.207
0.219
0.214
..
0.305
0.333
0.372
0.247
0.26
0.258
..
0.320
0.283
..
0.256
0.276
0.26
..
0.324
0.325
0.390
0.281
0.275
0.282
..
0.330
0.346
0.374
0.290
0.358
0.254
..
0.340
0.349
0.35
0.302
0.298
0.259
0.336
0.360
0.336
0.345
0.300
0.307
0.258
0.384
0.370
0.332
0.357
0.294
0.306
0.257
0.401
..
0.333
0.368
0.305
0.305
0.27
0.376
..
0.337
..
..
0.306
0.273
..
0.386
0.322
0.382
..
0.310
0.273
0.388
..
0.328
..
0.390
0.307
..
..
..
0.332
..
0.393
0.305
0.155
..
0.226
0.277
0.257
0.287
0.364
0.305
0.273
0.352
0.258
0.358
0.281
0.372
0.254
0.406
0.388
0.391
0.290
0.358
0.234
0.221
0.244
0.341
0.371
0.251
0.399
0.461
0.364
..
0.446
..
0.374
0.471
..
0.356
0.483
0.413
0.354
..
0.406
0.351
..
0.391
0.337
0.482
0.427
0.337
0.483
0.462
0.343
0.508
0.452
0.342
0.477
0.418
0.340
..
0.408
0.258
0.275
0.301
0.260
0.250
0.355
0.361
0.369
0.300
0.411
0.366
..
0.4
0.445
0.437
0.321
0.428
..
0.443
0.379
0.381
0.459
..
0.395
0.390
..
0.458
..
0.428
0.414
..
0.462
0.498
0.431
..
..
0.462
..
0.429
0.426
..
..
..
0.466
0.441
0.486
0.506
..
0.47
0.392
..
0.501
..
0.512
0.391
..
..
..
0.490
..
..
0.508
..
0.478
0.372
Source: UNICEF, TransMONEE Database, except for Russia, where Goskomstat published data is used to fill
TRANSMONEE data gaps and Serbia (data from Krstic) Note: distribution of gross monthly wage bill, with
bonuses, for full-time employees with the exception of Hungary, where net wages are used.
13
Table 4 data should be taken with some caveats. For Russia, for example, data refer to wage
bill just accruing to workers, not necessarily paid. With a spread of arrears in the mid-1990
this definition clearly understated inequality of take home pay (Lehmann and Wadsworth).
Series may behave in a somewhat strange way reflecting changes in the reporting procedures,
or sample included in the wage survey of firms. 7 With these caveats, Table 4 shows that for
most countries presented the inequality in wage distribution keeps going up during the whole
period. We can also say that wage inequality is lowest in CEE (maybe at the higher end of
the OECD countries), higher in the middle income CIS and highest in the low income CIS.
In many countries this increase occurred almost entirely in the early transition (before 1993).
Comparing early 2000s to 1993 there was a significant change only in Poland, Hungary,
Romania and low income CIS, but no change in the Czech Republic, Estonia, Slovenia and
Latvia (using 3 percentage points criteria to assess change propose by Atkinson (2003)).
There is some weak evidence for the presence of an inverse U-shape trajectory.
Given data uncertainties, it is important to supplement enterprise level data on wage
inequality with the survey based indices. Contrasting different data sources, enterprise level
and household survey-based, Figure 5 reports available Gini indices for Russia and Poland.
While dispersion between estimates is indeed very large, both the raise of inequality and
different patterns of increase are evident.
Figure 5 Gini index for monthly wages in Russia and Poland, various sources
Russia
Poland
Official wage, GKS
RLMS
Official, GUS
ISSP
HBS
LFS
ISSP
0.4
0.55
0.5
0.35
0.4
Gini
Gini
0.45
0.3
0.35
0.3
0.25
0.25
0.2
1989
1991
1993
1995
1997
1999
2001
2003
0.2
1989
1991
1993
1995
1997
1999
2001
Sources: ISSP: International Social Survey Program from Paternostro and Tiongson Russia: Russian Statistical
office (GKS), and RLMS – Russian Longitudinal Monitoring Survey, Lukianova for contractual wages
(cleaning out the effect of arrears). Poland: HBS – Household Budget survey reported net wages Keene and
Prasad, 2002 estimate –own; LFS –Labor Force Survey from Newell and Socha (2003).
Table A 2.1 in the Annex 2 shows values of Gini index for net wages reported by workers
for a wider set of countries from a comparable international survey, ISSP. It also shows that
7 The underlying data come from firm records, primarily large and medium size firms in the formal sector. Data
cover full time workers only and ignore therefore possible effects in inequality from hours adjustment.
14
contrary to the impression of steady raise in wage inequality over the period many countries
exhibit an inverse U-shape relationship. Considerable evidence exists that wage inequality
dynamics were influenced by the degree of labor market imperfections. These factors were
most noticeable in the increased spatial dispersion of wages without apparent tendency for
convergence.
For China increasing wage dispersion reflecting raising education premium and increased
divergence of wages across sectors, regions and occupations plays a similarly important role.
Like other transition economies, China had an extremely compressed wage structure in the
pre-reform period but experienced dramatic increases in the inequality. However, in China,
the level of inequality remained low until the early 1990s, more than a full decade after
economic reforms began (Li, 2003), and increased rapidly since then (Gustaffson et al [2001]
find surprisingly similar earnings profiles for Chinese and Russian urban workers in 1989).
Gustaffson and Li (2001) report that between 1988 and 1995 Gini index for urban earnings
increased from 24.0 to 30.4.
Driver 2. Restructuring, unemployment and subsistence economic activities: appearance
of a new class
The second major reason for increased inequality, as argued, for example by Flemming and
Mickewright (1999), was an emergence of a new class of unemployed with distinctly
different level of incomes its distribution. Central to transition is the closure and
restructuring of firms as resources are reallocated to more productive uses. A priori, there
was little insight on how the incidence of job loses and its distribution across households was
to look like. Ex-post there are important variations across countries and regions in the size of
labor reallocation and its speed.
Low income CIS countries experienced the sharpest contraction of state sector employment
with very weak compensatory growth of the new private sector, pushing many workers to
non-employment. Countries in EU-8 experienced fast restructuring and growth of the new
private sector, which was nevertheless lagging behind job destruction by the old firms
creating a sizeable unemployment pool. Finally, labor reallocation in SEE countries and
middle income CIS was more gradual but at the same time private sector growth much
weaker. Unemployed had only limited resort to the safety nets.
Unemployment benefits regimes intended to provide a floor to the welfare of those affected
by restructuring, but they displayed a significant amount of variation. The CEE and SEE
countries attempted for a while to compensate displaced workers with generous benefit
regimes (UB and early retirement), but soon were forced to scale back generous packages
confronting macroeconomic constraints. Low income CIS countries were not able to afford
any sizeable support to the new class of unemployed. As a result those left out of
employment resorted to subsistence farming or other survival economic strategies. Middle
income CIS countries while retaining the formal benefit schemes and early retirement as a
tool, were forced to set the level of benefits at very low level falling somewhere in between
the EU-8/SEE and low income CIS regimes.
15
Table 5 shows the dynamics of unemployment across a number of countries. It portrays
rapid increases in unemployment and significant persistence of unemployment in some
countries and gradual changes in other. In explaining the discussion of transitional dynamics
around Table 5, it might be useful to remind the reader that changes in measured
unemployment do not fully account for the extent of job destruction. Hungary is a good
illustration: comparatively low unemployment rate is on account of low participation (share
of working age population employed dropped between 1985 and 2002 from 80 to 55
percent). More generally, labor market developments during the transition have manifested
themselves in a combination of (1) open unemployment (as in the Table 4), (2) lower
participation and (3) low productivity employment.
Table 5: Survey unemployment rates in ECA for selected countries 1991-2003
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
11.1
15.3
16.4
12.8
11.1
12.5
13.7
16
17
16.4
19.5
16.8
12.7
13.2
13.2
14.8
14.5
14.5
10
9.9
11.4
13.6
16.1
15.8
14.8
14.3
3
8.2
10.4
10.1
8.2
6.5
7.4
10.4
11.8
10.5
8.8
8.4
7.2
4.1
2.6
3.5
3.2
2.9
3.5
5.2
6.5
8.7
8.8
8.1
7.3
7.8
6.6
7.6
9.8
10
9.7
9.8
12.2
13.6
12.6
10.3
10
7.4
9.3
11.9
10.7
10.2
9.9
8.7
7.8
7
6.4
5.7
5.8
5.9
3.9
8.7
16.7
18.1
19.4
14.8
14.1
14.3
14.4
13.1
12.4
10.6
0.3
1.3
4.4
3.8
17.5
16.4
14.1
13.2
14.6
16.4
17.4
13.8
12.4
11.8
14.3
16.4
16
14.9
13.2
8.6
10.2
13.4
16.4
18.5
19.8
19.2
8.2
8.3
9.1
9.1
7.4
7.3
7.1
7.6
7.4
7.2
5.9
5.9
6.7
0
5.3
6
7.8
9
9.9
11.2
12.3
12.6
9.8
8.9
8.6
8.3
Ukraine
Russia
Slovenia
Poland
CIS
Lithuania
Latvia
Hungary
Estonia
Czech
Republic
EU-8
Romania
Croatia
Bulgaria
SEE
0
0.2
0.4
0.4
5.6
7.6
8.9
11.3
11.9
11.7
11.1
10.2
Source: EBRD, Transition Report (various issues), and ILO Key Labour Market Indicators database for
Ukraine.
The accurate data on employment outcomes combining (1) and (2) are available starting only
in 1995. By 2002 only 50 percent of the working age population was employed in SEE, the
region most severely affected by delayed employment restructuring, followed by EU-8 with
60 percent employment rate, in contrast to middle income CIS where it stood at still low 65
percent. This is lower than in the EU-15 (70 percent).8
The scale of employment reallocation out of wage employment to various subsistence (or
low productivity) occupations in low income CIS countries was equally large. ECAPOV 2
data relying on household surveys show that even in by 2002-3 as much as 50 percent of the
population in Moldova were in families completely relying in subsistence farming or were
8 Using 15 to 64 range, ILO data.
16
unemployed. The share in Tajikistan was as high as 40 percent, in Georgia and Kyrgyz
Republic – 30 percent.
In several EU-8 countries the unemployment and non-participation took on some feature of
stagnant pool despite the early expectation of their imminent dissipation as economies grow.
As long as we believe that there is a potential of further increases in participation through
employment growth in the private sector, this transitional (and temporary) factor of
inequality dynamics continues to play a role. And some evidence form recent work (Poland:
Newell and Socha 2003) suggest that it may in fact start playing a dominant role as the wage
decompression driver wears itself out. In SEE and middle income CIS the restructuring of
the economy is still an unfinished agenda, and thus they remain under the influence of this
factor. Finally, in the low-income CIS to the extent that there is an uncertainty over the
future economic role of the subsistence farming and other economic activities developed by
the displaced worker as coping strategies, and the private sector is underdeveloped compared
to its longer-term proper role, the reallocation between sectors will continue and thus the
income distribution parameters are not settled.
Developments in China were quite distinct, but at the same time some parts of its economy
exhibited trends broadly consistent with ECA patterns. The primary feature of Chinese
economic development was a transition of workers from the informal occupations (primarily
agriculture) to formal sectors. This transition has well-known implications for inequality
described by the Kuznets curve. But the pace of this transition was highly uneven.
During the 1980s, rapid growth created an average of over 10 million formal jobs per year.
The pace of job creation accelerated further during the early 1990s. Employment prospects
facing Chinese workers improved remarkably, with millions transferring every year from
farming and other informal occupations into regular employment. But employment prospects
deteriorated dramatically after 1995 for some period. At the same time profound changes
occurred in the labor market institutions. China’s socialist tradition of lifetime tenure for
regular urban worker ended with labor market reforms, and in excess of 50 million workers
were dismissed or transferred to temporary status. Chinese cities started facing open
unemployment of the order of 8 percent (Rawski 2003).
This sequencing of reforms led urban areas of China to experience quite late in the process a
combination of effects from transition-induced changes with the pressure from migration
inflows typical of a developing country. Rural areas experienced effects from both large out
migration and the transitional factors associated with land reforms and greater market
penetration to rural economies (this is discussed in more detail under Driver 5 heading).
Driver 3. Changes in government expenditure and taxation
Key countervailing factor to the increase in inequality pushed up by the Driver 2 was the
effect of tax and transfer policy. Socialist system of social protection combined a universal
system of cash transfers with a substantial system of in-kind transfers delivered by a
combination of enterprises and government agencies. With a transitional recession a decline
in a real value of government social spending has been compounded by the reduction in
17
social spending by firms, but to a widely different degree across countries.9 The implications
differed depending on how severe was a reduction in the value of transfers and how well they
were targeted.
Table 6 provides some data on the evolution of total social security and welfare spending in
the GDP looking at early transition and late transition periods. Table6 shows that between
1993 and 2003 transfers increased as a share of GDP in most countries.
Table 6 Share of social transfers in GDP, %
Country
Azerbaijan
Belarus
Bulgaria
Georgia
Hungary
Kazakhstan
Kyrgyz Rep.
Latvia
Poland
Romania
Russia
Slovak Rep.
Ukraine
Uzbekistan
1987-1989
1990-1992
18.0%
15.7%*
9.9%*
8.0%
9.4%
8.2%
9.7%
17.4%
6.0%
13.4%
11.0%
1993-1998
4.3%
13.0%
11.0%
2.2%
11.8%
6.6%
3.0%
12.3%
18.7%
7.0%
9.0%*
12.2%
18.0%
10.2%
1999-2003
5.0%
14.5%
13.0%
4.3%
13.2%
5.4%
12.2%
14.5%
20.1%
10.0%
14.6%*
14.0%
15.0%
9.0%
*Including in-kind transfers in the form of privileges and enterprise transfers Source: PEM database for 1993-2003,
Milanovic (1999)for 1987-1992 and Forster and Toth.
Increase in transfers was seeing as a key inequality-reducing factor in CEE, as portrayed in
the theoretical and empirical studies by Aghion and Commander (1998), Forster and Toth
(1997), Garner and Terrell (1997), Keane and Prasad (2002a). Milanovic (1998) provides a
very simple intuition behind this assessment: the commitment of CEE Governments to ensure
stability of real pensions (compared to average wages in the economy), and the need to pay
unemployment benefits to a large number of workers. But the political economy of reforms
in the CIS resulted in a very different set of factors.
Rapid collapse of GDP in CIS and dwindling of Government revenues as a share of GDP,
resulted in a rapid fall of the total envelope of transfers. Once divided by the number of
claimants it was too small to sustain even the most basic subsistence levels of the transfer
recipients. As a result some countries drastically reduced their safety nets coverage, focusing
on the most needy (low-income CIS). Other countries aimed at retaining key benefits but
compressed levels to a simple per capita distribution among the claimants. That is largely a
story of the evolution of pensions in middle income CIS, representing the bulk of social
protection spending. But to what extent these changes affected the distribution of incomes
9 China’s government transfers as a share of GDP decreased from 0.35% in 1985 to 0.28% in 2001 (check), of
which 0.1 % of GDP represents “Minimum Guaranteed Livelihood program” targeted to some 20 mln. urban
poor (a tiny fraction of all urban poor – see Ravallion (2006). China also exhibited a particularly protracted
transition from the communist system of social safety net and the remnants of provision of public services and
safety nets through the employment remain strong to this day.
18
depended very much from the living arrangements of transfer recipients, and incomes of
other family members.
Improved targeting of social assistance benefits, on the other hand, was a sizeable factor of
inequality change only in a few EU-8 countries (especially in Hungary), because only in this
group experienced the expansion of social assistance programs in real terms.
In terms of taxation, the transition induced dramatic shift in the composition and incidence of
taxes (such as the introduction of value added tax) as well as declining tax compliance.
Unfortunately, the household survey data that we use in this paper have little usable
information on taxation and its incidence, but limited empirical evidence suggests that most
changes were in favor of greater equality, but with significant variation across countries and
time periods (Kattuman and Redmont find sharp changes over time in the effect of personal
taxes on distribution of incomes in Hungary, and Garner and Terell reveal significant
differences in the effect of taxes between the Czech and Slovak Republics, Commander and
Lee provide some evidence for Russia for mid 1990s that suggests dramatically smaller
effects from taxes on distribution than in the CEE).
At the same time, failures to move to the new taxes and ensure proper tax collection in the
early transition was a factor contributing to increased corruption in Russia and the rest of
the CIS (EBRD, 1998). Such failures relative to approved budgets either lead to inflation or
to public expenditure cuts often in the form of arrears - drivers which impact is discussed in
the next sub-section (Driver 4). Here it is important to mention that corruption has attracted a
lot of attention in the literature on inequality in transition. Using cross-country regressions,
several attempts were made to link corruption and the increased informality to inequality
(ECAPOV 1). But there was no empirical studies linking corruption to the inequality at the
micro level and the mechanism of such link is unclear. Both variables at the macro level are
clearly trended in the same direction in transition, but it is possible that they are caused by
different factors correlated with institutional reforms. It is nevertheless important to attempt
to trace the effect of the informal sector, which is in part driven by corruption, on inequality.
A priory, the effect should differ depending on the size of the informal sector which is known
to vary significantly across transition economies of CEE and CIS (Schneider).
As for the previous drivers therefore we observe significant variation across countries and
time periods and close inter linkages with other drivers.
Driver 4. Price liberalization, inflation and arrears
Transition has been associated with a loosening of controls on prices. As argued by
Flemming and Mickewright (1999) nearly all socialist economies embarked on the process of
transition with a substantial monetary overhang. Thus when prices were liberalized they
jumped, sometimes by factors of two or three and inflation rates tended to persist with
accommodating monetary policy stance. Experience from other high inflation episodes (e.g.
in Latin America) points to strong redistributive effects, mainly at the expense of the poor.
And aggregate data indeed indicate that the inflation tax in Russia appears to have had a
powerful effect. In 1992, for example, it has been estimated that households were hardest hit
19
by inflation, losing about 12 percent of GDP through this tax on financial assets (Commander
and Lee). This amounted to roughly a quarter of household income and is likely to have been
regressive, given the ability of richer households to shift into real assets or indexed-linked
financial assets. Similar if not larger redistributions took place in Georgia, Ukraine,
Bulgaria, Belarus and Georgia, but this factor affected to much lesser degree EU-8
economies. In a comparative perspective inflation is shown to have effects on income
distribution in China (Meng et al), but the order of magnitudes of effects is significantly
smaller than in ECA.
Arrears on pensions and social benefits payments appeared in the inflationary environment of
several countries in the CIS and SEE. According to survey data (RLMS) around 6o% of the
workforce in Russia were owed money at the height of the problem in 1998. Arrears were
concentrated in the bottom part of the distribution and in a highly inflationary environment
resulted in a cut of real wage in a highly disequalizing way, affecting measured returns to
education, gender gap, sectoral and regional dispersion of wages (Lehmann and Wadsworth
[2002]). Similar effects were found by Klugman (2000) for Uzbekistan.
It would be incorrect to assume that relative price changes always contributed to increasing
inequality. A good example is liberalization of agricultural prices in China vs ECA. In all
ECA countries this factor caused a dramatic decline in farm profitability and rural incomes.
Agricultural production and food consumption were heavily subsidized under the communist
system. Macro-economic reforms coincided with price liberalization and subsidy cuts in the
early years of transition. Reduced domestic demand with falling incomes was reinforced by
falling foreign demand and increased import competition with trade liberalization.
Agricultural terms of trade declined dramatically, i.e., between 40 and 80% (Rozelle and
Swinnen). That affected the relative position of rural areas and increased ceteris paribus the
income disparities.
In China too, price liberalization affected different groups of the population differently
(urban versus rural) and acted as strong factor explaining changes in inequality (Ravallion
and Chen 2004). But the direction of effect was opposite to ECA. Rozelle and Swinnen
(2004) have demonstrated that the positive effect of Chinese land reform on relative position
of rural areas was in part driven by a policy of price liberalization for agricultural goods
which resulted in rapidly increasing relative prices of agricultural commodities and positive
change in the terms of trade in favor of agriculture.
There were variations in the scale and pace of liberalization, and certain prices -- particularly
for housing -- have remained largely administratively determined for a long period in some
countries. This created an additional source of distortions for readily available measures of
inequality which typically in ECA do not use “shadow” prices to impute values of subsidized
goods. We therefore will ignore it in the subsequent discussion, but note that again we
discover a large range for possible outcomes associated with price reform-induced factors.
20
Driver 5. Asset transfer, growth of property income and change in the distribution of
factor incomes
The most visible sign of transition everywhere has been the large-scale transfer of previously
publicly owned assets into the hands of private agents. The increase in the share of
entrepreneurial income, and the share of families receiving financial incomes was an
immediate result common to all transition economies. In Russia, for example, the share of
property, interests and profits among households’ cash receipts has increased from around 4
percent in 1989 to 20 percent in 2003 (Goskomstat). These sources of income are known to
be inequalizing in the OECD countries (Milanovic 1998).
Privatization programs differed significantly across countries and across types of assets.
Birdsall and Nellis (2003) look at the impact of privatization programs on distribution across
a large swath of countries and conclude that many have worsened the distribution of assets
and income, at least in the short-run. But over time policy reform and elimination of
subsidies favored more equitable outcome, generating higher tax returns to the State and
leading to higher growth path with higher demand for skilled workers (thus perhaps greater
earnings dispersion). There were other effects. In ECA privatization generally favored
insiders, especially in the CIS, and often favored older generation (with some notable
exceptions), establishing a direct link between tenure (seniority) and claims on various assets.
Depending on the scale of the real asset transfer and the evolution of correlation between
incomes and age, this produced a change in the distribution of other income sources,
especially transfers. While before transition pensioners relied exclusively on pensions, new
sources of incomes described above emerged after the privatization and often changed the
relative income position of pension recipients. These illustrations reveal complex interaction
between different drivers and the danger to over-generalize.
Land privatization was an important reform redistributing vital asset in the low-income CIS
countries, with significant variation across countries. As much as 90 percent of arable land
was transferred to households on highly beneficial terms in Albania and Armenia, between
one half and three quarters in Romania, Estonia, Latvia, and Moldova one third in
Kyrgyzstan and Georgia, but only 10-20 percent in Russia, Kazakhstan, Ukraine and
Uzbekistan (Rozelle and Swinnen). This order of transfer was even larger in China where
200 mln. farmers now have land use rights for 99 percent of agricultural land. Despite initial
fears of potential regressive effects (Flemming and Mickewright 1999) restitution of land to
former owners adopted in several CEE and SEE countries was not shown empirically to have
a sizeable effect on the inequality (Macours and Swinnen).
In addition, significant privatization of housing has occurred, also largely favoring older
generations. In Russia, by early already 1996 nearly 50 percent of the housing stock was in
private hands a proportion which grew to 85 percent (check) by 2005. This is also shown to
enhance equality of outcomes, especially in middle income CIS countries. Imputing an
economic value to subsidized goods and assigning it to households in different parts of the
distribution usually leads towards reduction of inequality, but to a much greater degree in
transition economies than elsewhere (see Flemming and Mickewright 1999, Davies and
Shorrocks 2005 for a review, also Table 2), but all such studies are based on a single period
21
data and there are no studies known to us which look at dynamics aspects of this driver. This
is in contrast with China where a number of studies examined the effects of housing policies
in detail (Khan and Riskin for review).
Driver 6. Technological change and globalization
Technological change and modernization of the economy in a broad sense is a major factor
of inequality dynamics for all countries. The famous Kuznets hypothesis linked economic
development and technological change with inequality (Kanbur, see WDR 2006 for a short
review). In a more narrow sense the latest period of increasing inequality in OECD countries
is widely perceived as being linked to the change in the technology associated with the new
type of global production systems.
The reduced demand for less skilled and fixed supply of skilled labor in the short run means
a rise in the premium for skilled workers and a decline in the relative wage of unskilled
workers. This is taken as common explanation of rising wage dispersion.10 The reasons for
the shift away from unskilled to skilled workers are disputed. Globalization and technology
changes are most prominently featured and refer to the increase in international trade and the
advent of new wave of general purpose technology revolutionizing the entire production
system. But transition itself is also associated with the adoption of new technology and
organizational structures. Transition is also associated with trade liberalization exposing
technological obsolescence and associated loss of skills among workers (Aghion and
Commander 1999).
It is therefore inherently very difficult (if not impossible) to separate “transitional”
technological catch up with the market production systems from the global factor of
technological progress. The only insights offered by analysts are that transition economies
due to the combination of catch up and new technology absorption are experiencing faster
rates of technological change than other (developed and developing) countries. The same
type of argument can be applied to China as long as it inherited certain production systems
from the communist period.
How exactly this factor affects the distribution in the medium run is not only driven by shifts
in demand, but also by human capital endowments, adaptability of existing skills to the new
technology, labor market imperfections and supply of new skilled labor (Aghion and
Commander). It is argued that most likely, the wage dispersion will increase considerably
both between educational groups (cohorts) and within education groups and cohorts
(especially among those with firm-specific skills, such as vocational school).
Analysis of education premia and dispersion across education-age cohorts may be helpful to
gauge the size of effects (Kertesi and Kollo). It is important to mention that considerable
education expansion in most transition economies mitigated against this factor, even though
10 In some OECD countries effective minimum wage protection leads to higher unemployment rather then
decreasing wages for the unskilled workers with complex impact on earnings (reduced or stable dispersion)
versus household incomes (increasing inequality between employed and non-employed).
22
often accompanied by reduction of public spending on education.11 Tertiary enrollments
have increased everywhere starting in early transition. In Hungary, for example the number
of university graduates in the labor force tripled between 1985 and 2000 (Kertesi and Kollo).
This expansion is illustrated on Figure 6 showing the increase of the share of cohort entering
the labor market with higher education across countries ECA, with significant differences
across countries. Similar developments were observed in China, where the share of college
graduates in the urban labor force increased from 12 percent to 28 percent between 1988 and
2001 (Zhang).
Figure 6 Shifts in the relative supply of skills: Share of youth entering labor market
with university education
Belarus
Poland
Bulgaria
Romania
Hungary
Russia
Kazakhstan
Tajikistan
Moldova
30%
25%
20%
15%
10%
5%
0%
1995
1998
1999
2000
2001
2002
2003
Source: Mete et al. (2005) Note share of younger adults (26-35 year olds) cohort
Aghion and Commander (1998) discuss the long-term determination of inequality in
transition and argue that inequality persistence depends critically on the pace at which the
acquisition of skills takes place in the economy--and, hence, on the evolution of the
educational system. This is therefore a critically important long-term driver of inequality
with particularly important role in transition economies due to the intensity of change.
Driver 7. Demographic transition and increased mobility
Population aging that is experienced now by many countries, albeit to a different degree, has
profound implications for family formation, participation rates and thus affects the level and
dynamics of inequality. It also leads to distributional conflicts between older and younger
generations. Demographic transition and a move from a multi-generational extended
household to nuclear family are though to have a non-linear effect on inequality (Cowell).
11 As noted by Aghion and Commander. But private expenditures in education more than compensated for that
fall, and there was a considerable increase in the dispersion of quality of education (ECAPOV 2).
23
Demographic factors such as sharp changes in fertility and household formation (as discussed
by Svejnar 2002) in all countries in question were going at pace which was as rapid as
economic transformation. Family formation was found to matter empirically for Russia,
Hungary and Poland (see Annex 4) and hence had a significant effect alongside typical
transition factors. Similar, if not larger, effects were observed in China and generally were
inequality-reducing. Finally, migration also affects inequality in ECA, but the exact
magnitude and direction of change depends on the country (Pryadyilnikov).
Driver 8. Political economy factor: preferences for equality
Political economy factors are used to explain some persistent differences in the distribution
of income across countries (especially between OECD countries). As Benabou (1996)
pointed out, neoclassical growth model implies convergence in the distribution: countries
with the same fundamentals should trend towards the same invariant distribution of wealth
and pretax income. Key parameters defining this distribution are fiscal revenue composition,
provisions of public goods by the State and society’s inequality aversion (Kanbur and
Tuomala 2002). This factor ultimately determines the extent of fiscal redistribution that
takes place.
Empirical studies of preferences for equality (or inequality aversion) show remarkably stable
strong preferences for equality in EU-8 countries and somewhat variable, but not as strong
preferences in the CIS (Murthi and Tiongson, forthcoming). Democratization and a greater
role played in policy choices by the median voter will provide a stronger link between the
underlying preferences and the extent of redistribution and inequality reduction through taxes
and transfers that takes place.
How different drivers interact is very much a question of particular country circumstances, its
initial conditions, and most importantly, policy choices. Table 7 attempts to provide a
patchwork of factors of change simultaneously occurring in transition and to identify a
resulting vector of inequality change by sub-region.
Each of the drivers, especially the ones specific to transition (1-5) is operating through a
specific channel and can be traced through looking at the components of inequality in a
particular way. The last row of Table 7 shows a combined effect of all factors on inequality.
The ranking of potential for inequality increase is clear: smallest increases in EU-8 and SEE
(albeit for different reasons) to shaper ones in middle income CIS and most significant ones
in low income CIS. But given a complexity of the interactions between drivers the path of
uneven increase or even fluctuations would make the assessment at any given moment of
time difficult. It is also bound to produce different outcomes even with a group of countries.
24
Table 7 Hypothetical role of inequality drivers across subregions of ECA
EU-8
SEE
Mid-inc CIS
Low-inc CIS
1. Wage decompression and the growth of
private sector
↑↑↑
↑
↑↑
↑
2. Restructuring , unemployment/or reverting
to subsistence economy
↑↑
↑↑
↑
↑↑↑
3. Fiscal adjustment affecting the
Government expenditure and taxation
↓↓
↓
↑↓
↑↑
4. Price liberalization, inflation and arrears
↑
↑↑
↑↑↑
↑↑↑
5. Asset transfer and the distribution of factor
incomes
↑↑↑
↑
↑
↑↓
6. Technological change and expansion of
knowledge economy
7. Demographic transition and increased
mobility
↑↓
↑
↓
↓
↓
↓
↓
↑
8. Preferences for equality
↓
↑↓
↑
↑
Combined effect
↑
↑
↑↑
↑↑↑
↑- Inequality increasing, ↓- inequality decreasing, number of arrows shows the strength of a driver.
Models of restructuring
Aghion and Blanchard (1994) proposed a theoretical model of transition dynamics describing
the reallocation of productive resources in transition. The transition is formalized as a
reallocation of labor and capital across state and private sectors (and unemployment12 as a
transient step between the two). The model is concerned with the labor allocation, labor
incomes and transfers and, hence, can provide the paths (short- and medium-term dynamics)
of inequality. It takes the total labor available for economic activities as fixed.
In this framework, we can immediately isolate several discrete ways in which the transition -summarized by the dynamic reallocation of resources from the state to the private sector -can affect the income distribution. The model produces a rich set of trajectories and paths
12 This state can also be re-interpreted as subsistence or informal sector employment (Radulovic 2003).
25
depending on the key parameters. Simulations using versions of the model and their
applications to study cross-country variation are described in detail in Commander and
Tolstopiatenko (1996).13
In a number of modifications of the model some simulations were presented which provide a
set of simple benchmarks and even used to assess the empirical evolution of inequality
(Aghion and Commander). Such trajectories were associated with stylized path of expected
inequality dynamics in CEE versus CIS. In the CIS case with a very ungenerous benefits
regime, low initial values for closure and restructuring, unemployment remains quite low
throughout the transition. The reallocation of labor to the private sector is protracted.
Inequality rises gradually and steadily to high level. More generous benefit regime with
higher probabilities of restructuring alike to Central Europe leads to unemployment peaking
at higher levels, but given fast movements of workers into the private sector and generous
floor in the form of unemployment benefits, inequality rises to reach a hump at a lower level
than observed in the first scenario, and then recedes (Figure 7).
Figure 7. Simulation results from the restructuring model with different configuration
of parameters: two typical trajectories for inequality indices
“CIS”
“CEE”
0.5
0.5
0.45
0.45
Gini
0.4
0.35
0.35
0.3
0.3
0.25
Inequality
Inequality
0.4
Theil_L
0.2
0.25
Theil_L
0.2
0.15
0.15
0.1
0.1
0.05
0.05
0
Gini
0
0
5
10
15
20
25
30
35
40
45
Tim e
0
5
10
15
20
25
30
35
40
45
Tim e
Source: Commander, Simon and Andrei Tolstopiatenko (1997).
13 The economy consists of two sectors, state and private, and three basic labor market states: state
employment, private employment and unemployment. The model assumes standard production functions and a
given distribution of workers by skills and productivities, as well as utility functions. It endogenizes the
decision of workers/managers to restructure (i.e. change the way the operational surplus is shared within a firm)
their firm by assuming that they depend on the difference between values of staying in different sectors
compared to the value of being unemployed and to the value of being in the unrestructured firm. In addition a
key parameter, probability of closure, is determined by exogenous institutional and financial factors. There are
also exogenously chosen policy parameters, such as taxation and unemployment benefits, which makes this
model somewhat less ‘closed’ than in the optimal taxation models as analysed in Kanbur and Tuomala (2002),
where both taxations and benefits are chosen simultaneously by a benevolent dictator to reflect societal
preferences.
26
The strongest point of the model is conceptualization of restructuring: not as one–time shift
in the behavior of agents, but as a whole array of outcomes with different degree of rent
appropriation by insiders in partly restructured enterprises. The empirical studies of ten years
of transition revealed the coexistence of new, partly restructured and unrestructured firms as
a defining feature of transition (World Bank 2002e). This introduces an additional source of
variability (and hence, inequality), adequately captured by the model.
However, contrasting model’s predictions with empirical evidence, we see a surprising
reversal of patterns between CEE and CIS. The inverse U-shape trajectory of inequality
empirically seems to emerge not in the CEE, but in some CIS countries. This failure to
predict the actual dynamics of inequality may be a result of important limitations of the
model, or it may be an effect of mitigation by offsetting policy measures.
Several limitations of the model make it a poor predictor of inequality outcomes. This is
essentially a model of labor reallocation in the transition and the capital (or mixed) income is
omitted, by-passing one of the most important features of transition. Rural self-employment
in particular, or informal subsistence economy, an important sector in many economies
especially in low-income CIS, is not incorporated in the framework. The modeled inequality
is also essentially driven by across-sectoral variations. The parameters of the distribution
within each sector are taken as exogenous (and constant). This greatly simplifies the
representation of the drivers of inequality. Incorporation of such effects could lead to even
more variability in the inequality dynamics.
Despite all these limitations, the ability of the model to portray large variation of shapes and
levels of inequality dynamics is instructive. It suggests that there is probably no single
“transition” story as far as inequality dynamics are concerned. Judging by the own criteria of
the restructuring model, the transition is far from being over in ECA countries as
restructuring continues in many CIS and SEE countries and (transitional ) unemployment
pool has not been receding in EU-8 economies.
Where does China fit in this framework? The model helps to understand why factors
affecting inequality in China are different and could not fully be accounted for (even in the
most stylized features) by the restructuring model. The assumption of fixed labor supply and
its reallocation between essentially modern (socialist and capitalist) production activities
contradicts well known stylized factors about the role of large rural economy in China. But
at the same time to the extent that China‘s transition involves reallocation of labor from State
owned enterprises to fully private (or foreign owned) entities, the restructuring model is
relevant.
Accounting for factors of inequality change
Even the model of restructuring with its great simplifications is a complex system that
produces solutions only through the numerical simulations. Inequality decomposition
framework helps to disentangle empirically different factors affecting distribution
simultaneously in a simpler, yet descriptive way.
27
There are two ways in which inequality can be decomposed: by income sources, and by
socio-economic population groups (see Annex 5 for details). In this paper, following
Milanovic (1998) we combine both. In making this analysis, we use both the household
incomes and consumption.
Income components
The contribution of each components of income to total inequality can be obtained from the
product of the concentration coefficient for each component and their respective weights in
total income (see Annex 5). Concentration coefficients in turn depend on how unequally an
income source is distributed and how closely it is correlated with total incomes.
It is easy to generalize this framework to extend it to typical drivers of inequality in
transition. Table 8 puts together for three representative countries shifts in the shares of
income sources and concentration coefficients. We add to Milanovic (1998) original simple
framework an extra group of countries that was omitted from his analysis: the low income
CIS. The parameters have been chosen to reflect the observed structure of income inequality
pre- and after- transition and broadly match inequality outcomes. The main income sources
are taken to represent key drivers: wage income, pensions, social transfers and non-wage
income which is a combination of all other income sources, ranging from in-kind subsistence
income, farm incomes, remittances to property income and incomes from self-employment
and entrepreneurial activities.
This stylized framework is helpful to understand how inequality levels changed in transition.
It shows the relative importance of structural shifts versus own-inequality effects.14 By
assuming constant concentration coefficients (fixed at their pre or post transition levels), we
see the effect of changing income structure. And it is, as noted by Milanovic(1998), a minor
factor. Everywhere is accounted for less than 20 percent of the inequality change, and it
acted in fact as an equalizing factor in CEE and middle income CIS. Only in low income
CIS the collapse of wages resulted in a sizeable effect towards increasing inequality from the
changing income structure (about 3 Gini points).
Turning to within-income source distributional dynamics, we can make three observations:
First, income from labor is the main source of livelihood, and distribution of earnings is the
main driver of overall inequality. But the shape of the distribution is also determined by the
concentration of other income sources, including transfers and capital incomes. Higher
concentration coefficients of wages (with the exception of low-income CIS countries) drove
the overall Gini up. It was the most important factor behind the increase in inequality,
accounting for over 50% of the inequality change. In low income CIS receipts from selfemployment, farming and remittances (assembled under non-wage private income) was the
main driver.
14 Note that whenever the concentration coefficient of income source k is greater than the overall Gini
coefficient, an increase in the income source k (holding everything else constant), will increase inequality.
28
Second, the effect of transfers on inequality was not uniform across the Region, and changes
were mostly driven by changes in the size and the distribution of pensions. Non-pension
transfers, because of their small initial size, were not expected anywhere to have much
impact on the overall change in inequality.
Table 8 Stylized decomposition of income inequality by sources before and right after
the transition by country groupings in ECA.
Region
Wages
PreAfter-
Non-wage private
PreAfter-
Pensions
PreAfter-
Shares of incomes, %
24
17
21
23
8
18
70
10
5
CEE
CIS Middle I
CIS Low I
60
78
50
50
53
20
20
9
30
CEE
CIS Middle I
CIS Low I
23
25
30
32
52
55
Concentration coefficients, x100
31
31
16
23
17
43
0
20
30
50
-5
5
CEE
CIS Middle I
CIS Low I
62
93
64
58
66
24
28
7
38
CEE
CIS Middle I
CIS Low I
+2.2
+8.0
-4.0
Contribution to inequality, %
27
12
18
24
0
9
76
-2
1
Other soc.transfers
PreAfter-
Total
PreAfter-
5
5
10
7
6
5
100
100
100
100
100
100
-8
0
0
-12
11
-5
22.3*
21.0*
23.5*
27.4*
41.7*
46.0*
-2
0
0
-3
2
-1
100
100
100
100
100
100
Contribution to change of inequality, (in points of Gini)
+1.2
+2.1
-0.4
+8.4
+3.6
+0.7
+26.0
+0.8
-0.3
+5.1*
+20.7*
+22.5*
* Gini indices, or Change in Gini, points
Source: CEE and CIS Middle income based on Milanovic (1998) (CEE and FSU), Low income CIS –own estimates.
Third, to account for major source of changes in inequality in CIS we need to disband a very
heterogeneous “private non-wage income” source. In particular it would be important to
separate survival type activities, new entrepreneurial incomes and incomes from property.
This simple numerical exercise suggests that the restructuring in a narrow sense, i.e. shifts
between sources of livelihood in itself does not result in large changes in the inequality.
What matters more is within-income component of inequality.
Given its potential to summarize the sources of inequality change, it is very surprising that
there are only a few studies which use this framework (see Commander at al., Shutz,
Kattuman and Redmont, Kyslitsyna). Annex 3 provides summaries of available
decompositions from different countries. Annex 4 calls for caution by reporting dramatically
different decompositions for the same survey and the same year depending on whether
income or consumption is used are target variable to explain.
29
Socio- economic groups
Decomposition of Gini by income sources is informative, but it suffers from a considerable
problem in the context of transition: poor quality of reported income data. In addition, the
shape of the distribution can change not only as a result of changes in income streams. It is
also determined by the evolution of the size and composition of households and by incidence
of work and joblessness among household members.
Total inequality can also be represented are the sum of inequality from within each of the
groups and part of the inequality coming from differences in means between these groups
(see Annex 5). These groups can be defined to represent key actors of transition affected by
the redistribution. Decompositions of inequality by groups allow us to move to the indicator
of consumption inequality, which is superior in terms of data quality.
Milanovic (1998) applies this approach and starts from employment restructuring between
state and private sectors. The transition is modeled as a shrinking of the state sector
accompanied by the emergence of the private sector and an increase in the variation of
incomes both across and within sectors (part of which is determined through a state budget
constraint). The parameters of the distribution are defined by the share of state sector
employment, the number of private sector workers (tagged as self-employed), the share of
pensioners and the unemployed (transfer recipients), their relative incomes and distribution
within each ‘sector’. The values were calibrated to fit the generally observed patterns of
inequality dynamics in the pre-transition period and immediately after the transition.
Tables 9.1 and 9.2 summarize the example provided by Milanovic (1998) based on some
stylized representation of inequality patterns. If we compare these tables, we can see the
structure of the process of increasing inequality. The shift of workers from the relatively
egalitarian state sector to the private sector increases inequality. The greater between-group
differences are adding some 5 Gini points to inequality. The greater inter-group differences
are due to a changing composition of population (increased size of both high and lowincome groups), which is inequality-increasing (Milanovic calls this effect “hollowing of the
middle class”). The rest of increase comes from greater inequality within the private sector.
Milanovic also observed that inequality may display considerable inertia, therefore one needs
to control to the initial pre-transition inequality to allow pre-transition parameters of the
distribution to survive the transition (such as the schedule of wage distribution in the state
sector). It is important to note that this representation is highly stylized and ignores
differences across groups of countries in ECA. It also disregards the emergence of a large
class of agricultural workers engaged in subsistence farming – a major consequence of
restructuring with inadequate safety net. Thus, “classes” can be specified in a different way
to address the heterogeneity of countries experiences better.
30
Table 9. Hypotehtical pre- and after-transition distribution of incomes by main
groups (Gini)
Source : Milanovic (1998)
Preliminary conclusion from this review
Different drivers affected the distribution in different ways and their simultaneous operation
and interaction is sufficiently rich to allow a wide variety of outcomes across countries and
across time. A number of these factors are directly or indirectly influenced by policies
leaving a large space for policy-induced variation in inequality. It is also clear that the
transitional factors in ECA have not yet fully played their role.
There is a disconnect in the transition literature between the theory, which emphasized
changes in the structure, i.e. between-sectoral variations, and the empirical importance of
shifts within sectors of the economy. The importance of within-sectoral, or within –income
sources of inequality make it difficult to reduce the explanation of the increase of inequality
in transition to a simple accounting exercise and focus back our attention on the story of
channels of redistribution.
31
Policies play a very active role in shaping the inequality in ECA and in China. Transition is
very far from “pre-determined” mechanical change in the distribution, and policy drivers
need to be analyzed in details as forces which may explain the differences in inequality
outcomes between countries and time periods.
The main contribution of this paper is therefore to fill this gap in terms of providing group
decompositions of inequality changes based on comparable and consistent consumption data.
These results will be presented in section III.2 after we discuss results from income based
decompositions borrowing from the existing empirical studies.
III.
Decomposing inequality change in transition: Russia, China and ECA
experiences confronted
Applying the framework developed in the previous section, we first look at the evolution of
the structure of income over time, the respective concentration coefficients of the
components of income and the contribution of each component to inequality in Russia,
Poland, and China (Section III.1). Following this, section III.2 presents group-based
decompositions using data on consumption inequality.
Following Figure 3 which depicted joint evolution of growth and inequality in China, Figure
8 contrasts the evolution of GDP per capita and inequality in Russia and Poland using the
best available data.
Figure 8 Poland and Russia: Real per capita GDP and Gini index, 1990-2003
Poland
Real GDP Per Capita
Russia
Gini index
GDP Per Capita
Gini Index
175
175
150
Index, 1990=100
Index, 1981=100
150
125
125
100
100
75
75
1985
50
1987
1989
1991
1993
1995
1997
1999
2001
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
Source: For Poland : Keane and Prasad (2002a) for 1985-1997 (Gini for consumption per capita without durables), own estimates for Gini
based on regional data archive (see Annex 6), Russia: simulations based on published expenditure distributions for 1990-1996 and own
estimates for consumption Gini based on regional data archive (see Annex 6)
The evolution of GDP per capita and the Gini index follows very different paths in the two
countries with clear division into sub periods: (i) the early transition stage in Poland with a
noticeable fall in GDP but limited change in the Gini index, followed by impressive growth
and fast increase in the Gini index; and (ii) in Russia a prolonged recession has accompanied
by a sharp increase in the Gini index, followed by rapid growth with some moderation of
inequality. In neither case is there any apparent association between economic growth and
32
changes in inequality. Furthermore, these cases are sufficiently different to provide some
lessons about differences in the impact of different drivers on inequality.
III.1
Decomposition of inequality by income sources
Figure 8 makes it very clear that to understand changes of inequality we need to go beyond
the simple before- and after- taxonomy presented in Section II and differentiate between
periods of economic decline and periods of growth. A set of tables below provides available
evidence on the role of income sources as drivers of inequality for Poland, Russia and China
over an extended period of transition.
Tables 10 and 11 reproduce the major elements of Table 7 showing the structure of incomes
and its change over time, the concentration coefficient of each income component and its
change over time.
Table 10 Poland: Contribution of Income Sources to Total Inequality, 1987-2002
1987
1994
1998
Income structure: percent
2002
54
56
55
46
48
47
8
8
8
10
7
4
24
26
24
5
6
9
6
4
8
100%
100%
100%
Inequality: Concentration Coefficients
1994-1987
2002-1994
-6
-9
+3
-3
+7
+0
+1
+1
+1
+0
-6
+0
+4
+2
Work income
Of which wages
“entrepreneurial”
Income from farm
Old age pension
Social transfers
Other income
TOTAL
60
55
5
13
17
5
5
100%
Work income
Of which wages
Entrepreneurial
Income from farm
Old age pension
Social transfers
Other income
0.260
0.330
0.388
0.431
0.251
0.302
0.350
0.394
0.360
0.488
0.613
0.650
0.415
0.390
0.471
0.575
0.171
0.175
0.204
0.263
-0.100
0.080
-0.017
-0.011
0.340
0.283
0.450
0.263
Decomposition: Gini index, contributions
+27%
+20%
+35%
-6%
+3%
+180%
-17%
+31%
+30%
+33%
+47%
+50%
-86%
-7%
Gini, per capita inc.
Work income
Of which wages
Entrepreneurial
Income from farm
Old age pension
Social transfers
Other income
0.250
0.156
0.138
0.018
0.054
0.029
-0.005
0.017
+0.030
+0.022
+0.001
+0.021
-0.015
+0.013
+0.009
+0.000
+0.063
+0.059
+0.046
+0.013
-0.016
+0.021
-0.005
+0.004
0.280
0.178
0.139
0.039
0.039
0.042
0.004
0.017
0.320
0.217
0.168
0.049
0.033
0.053
-0.001
0.018
0.343
0.237
0.185
0.052
0.023
0.063
-0.001
0.021
Source: Milanovic for 1987, ECAPOV I for 1994-1998 and Poverty assessment Staff calculations based on CSO’s data for
2002.
33
It is useful to recollect here that the product of incomes share (upper part of each Table) and
concentration coefficient (middle part of each Table) equals the contribution of an income
component to the Gini index. The sum of these contributions equals the Gini index, which
gives us a very convenient way to show the role of each income component in the overall
change of the Gini index (bottom part each Table).
Table 11 Russia: Contribution of Income Sources to Total Inequality, 1989-2004
1989***
Work income
Wages
“entrepreneurial”*
Income from farm
Old age pension
Social transfers
Other income
Total income
Work income
Wages
“enterpreneurial”*
Income from farm
Old age pension
Social transfers
Other income
Gini for income per capita
1992
1996
1998
2004**
Income structure: percent
79
67
48
55
62
74
61
34
49
54
5
6
14
6
8
4
8
15
11
8
8
10
18
20
17
7
6
2
2
2
2
9
17
13
11
100
100
100
100
100
Inequality: Concentration Coefficients
0.285
0.540
0.679
0.540
0.515
0.280
0.531
0.644
0.514
0.454
0.360
0.633
0.764
0.750
0.925
0.300
0.350
0.440
0.573
0.375
-0.200
-0.140
0.111
0.025
0.094
0.086
0.317
0.500
0.450
0.150
0.200
0.833
0.512
0.492
0.373
Decomposition: Gini index, contributions
0.22
0.47
0.51
0.45
0.41
1998-1989
2004-1998
-24
-25
+1
+7
+12
-5
+10
+7
+5
+2
-3
-3
+0
-1
+90%
+84%
+108%
+186%
+113%
+425%
+146%
-5%
-12%
+23%
-35%
+276%
-67%
-24%
+0.206
-0.024
Of which:
Work income
0.225
0.362
0.326
0.297
0.319
+0.072
Wages
0.207
0.324
0.219
0.252
0.245
+0.045
“enterpreneurial”*
0.018
0.038
0.107
0.045
0.074
+0.027
Income from farm
0.008
0.028
0.066
0.063
0.030
+0.055
Old age pension
-0.016
-0.014
0.020
0.005
0.016
+0.021
Social transfers
0.006
0.019
0.010
0.009
0.003
+0.003
Other income
0.004
0.075
0.087
0.059
0.041
+0.055
Method: decomposition of the Gini coefficient into components, ranking by total per capita income
+0.022
-0.007
+0.029
-0.033
+0.011
-0.006
-0.018
Source: 1992-1998 from Commander et al based on RLMS, 2004 ** - own estimates based on RLMS data. ***1989Milanovic based in HBS,* includes in-kind and cash incomes from non-agricultural self-emplyment, informal work and
property income.
When we look at the Gini measure of income inequality by the end of the period, we observe
significant differences, with Poland’s income Gini around 0.34, and Russia’s 7 points larger,
at 0.41. We are interested not just in describing the differences and similarities between two
countries, but to explain how Russia’s excess income inequality compared to Poland’s. This
may offer some clues in assessing the future evolution of inequality in Russia.
34
The direction of changes in the income structure in Russia and Poland is generally similar
and reflects transitional drivers: falling share of wages, increase in entrepreneurial incomes
(profits and self-employment), and increase in the share of pensions among income sources.
The changes are consistent across periods of economic decline and growth, with wages and
transfers moving in opposite direction. But the outcomes are very different.
Was the change in the composition of the source of income which created these differences?
Surely, Poland experienced a much steeper rise in social transfers (including pensions) than
Russia. Russia followed a clear V-shaped trajectory for changes in wages, with a big dip and
then some recovery, while Poland exhibited an L- shaped trajectory. To see to what extent
these differences matter, it is insufficient to simply compare actual changes across countries,
because the observed change is a complex result of interactions between drivers pulling
inequality in different directions. What is needed is a counterfactual. However, producing a
fully satisfactory counterfactual distribution is difficult and requires building a model of
household income (Bouguignon et al), which goes beyond the scope of this paper.
Performing the simulations using base period concentration coefficients or structures is
feasible. It is of course a purely hypothetical counterfactual, because share and concentration
often change for the same reason.
With this caveat, using simple simulations, we see that the differences in the pace of
structural change in income sources do not fully explain inequality differential between
Russia and Poland. In fact if we adopt Poland’s income structure and apply it to Russian
concentration coefficients, the Gini index in Russia becomes just 1-0.5 points lower than it
actually was. Polish inequality will be a bit higher (by 2-4 Gini points) if Russia’s structure
of incomes is adopted.
The results are strikingly different for the simulations focusing on the impact of changing
concentration coefficients. Application of end-period concentration coefficients to Russia’s
original income structure would result in inequality exceeding its actually observed level by
at least 5 percentage points. For Poland the result is striking: application of Polish endperiod concentration coefficients to the original income structure would result in a Gini of
about 0.45, - a level observed in Russia and 10 points higher than the actual outcome in
Poland. This shows therefore that changes in the structure and changes in the concentration
coefficients offset each other, more so in Poland, but the factors of increasing inequality
within income sources clearly dominate. Among these sources of change three factors need
to be mentioned.
First, among income sources the key role of earnings as drivers of changes in inequality is
evident for both countries. Wages contribute around 25 Gini points to inequality in Russia,
and 18.5 in Poland; the difference between these contributions, 6.5 Gini points is almost the
entire difference between Poland and Russia’s Ginis (0.34 and 0.41 respectively). At the
same time, the concentration coefficient for wages in Poland is surprisingly large and similar
to Russia’s -- 0.4 compared to 0.45-- despite dramatically different level of Gini indices for
wages (as depicted on Figure 5 or Table 4 for example). This is due to the different degree
of polarization of labor incomes: in Poland a significantly larger share of households does not
35
receive any wage income (47 percent in 2002) compared to Russia (35 percent), reflecting a
more sizeable adjustment in employment in Poland as opposed to Russia.
The second driver of inequality changes are transfers, pensions and other social benefits. In
both countries changes in the distribution of pensions played a significant role as contributors
to the increase in inequality. But since their concentration coefficients are below that of
market income sources, this expansion reduced inequality compared to potential levels. Had
there be no increase in transfers in Poland, inequality would have been fully 3 Gini points (or
10 percent) higher. The effects could have been more progressive had there been no change
in the concentration coefficients of pensions.15 Other social transfers, on the other hand,
played a dramatically different role in Poland and Russia: the failure to target social benefits
in Russia as shown by a rapid increase in concentration coefficient in early transition is in
striking contrast to Poland. Data reported in Annex 4 for Hungary show a case where social
transfers dampened a potential increase in inequality even more. Data for Hungary also
show the critical importance of changes in tax incidence: a progressive tax system
compensated sharp increases of earnings inequality in early transition and moderated
inequality afterwards. Unfortunately no consistent data are available to analyze effects of
taxes on distribution in Russia or Poland.
The third broad driver is private sector growth combined with increasing informality. The
latter is difficult to measure with precision since we are dealing here with reported incomes,
which are known to underestimate informal incomes significantly [Yemtsov (1999)].
Informal incomes in various guises feature in different parts of the income spectrum: in farm
income (as in-kind consumption from own land plots), in entrepreneurial income (as many
businesses are not registered or as “side” wages reported as result of “free-lancing”), or
“other income” (especially in the CIS where this term is often used as a euphemism for not
fully legal or untaxed income). In terms of sheer size its effects were large. It is also quite
remarkable that in the post-crisis period in Russia some of the sources with strong informal
component show a fall in the concentration coefficients (farming and other incomes).
15 This increase might seem somewhat counterintuitive, as all transfers are regarded often as factors mitigating
against inequality increase (Keane and Prasad [2002a]). Paradoxically, it was kargely a result of the increased
pensions and greater reliance on pension payments by recipients. Before transition inadequate pension payments
were often supplemented by individual work post-pension age, and their recipients were as likely to be in the
bottom of the distribution as in the top. After the changes in pension policy and indexations most pensioners
moved to the middle, while having to forego additional earning with tighter labor markets. This created a
stronger positive correlation between income level and pensions, hence larger concentration coefficients.
Gustaffson and Nororozhkina (2005) used a unique study of households made in Taganrog in 1989 and a
follow-up study made in 2000 to arrive at the same conclusions: the main beneficiaries from expanding public
transfers have been households in the middle of the income distribution, and that also positively contributed to
the increase of income inequality in Russia.
36
Decomposition results reported in Table 12 for China use the same framework and the same
approach. The data are however from a non-official survey as reported by Khan and Riskin
(2004) and depict a picture of inequality change which is slightly different from that
represented in the official series.16 Unfortunately, since Chinese micro data are not
available for our analysis, it is not possible to test for differences in definitions etc. and it
becomes necessary to rely on specification as they are provided by the authors and assess
factors driving the inequality. These can be compared with the only study decomposing the
inequality relying on the official data for rural China using a more common set of factors by
Ravallion and Chen for an earlier period (See Annex 4).
Table 12: China: Contribution of various components to inequality, urban and rural,
1995-2002
1995
Work income
Wages
“entrepreneurial”*
Income from farm
Old age pension
Other income
Total income
80
78
2
15
4
100
Work income
Wages
“enterpreneurial”*
Net income from farm
Old age pension
Other income
0.198
0.195
0.291
Gini for income per capita
0.206
0.248
0.261
Urban
2002
95-02
Income structure, %
78
-2
74
-4
4
+2
18
+4
3
-1
100
Concentration coefficients
0.245
+23.9%
0.254
+30.2%
0.085
-70.7%
0.247
-0.5%
0.358
+37.3%
Decomposition
0.247
+0.041
1995
Rural
2002
95-02
37
25
11
53
49
34
14
45
+12
+9
+3
-7
9
100
5
100
-4
0.583
0.652
0.432
0.211
0.422
0.392
0.492
0.174
-27.7%
-39.9%
+14.0%
-17.3%
0.565
0.739
+30.8%
0.379
0.323
-0.055
Of which:
Work income
0.158
0.191
+0.033
0.215
0.205
-0.010
Wages
0.151
0.188
+0.036
0.165
0.134
-0.031
“enterpreneurial”*
0.007
0.003
-0.003
0.050
0.071
+0.021
Income from farm
0.111
0.079
-0.032
Old age pension
0.037
0.045
+0.009
Other income
0.011
0.010
-0.000
0.053
0.040
-0.013
Method: decomposition of the Gini coefficient into components, ranking by total per capita income
* Includes property income Source: Khan and Riskin (2004) based on CASS, the results are recalculated to
purge them out of the imputed rent and housing subsidies effects
16 Table 2 already emphasized the source of differing inequality estimates for China: inclusion of housing rents
and housing subsidies in kind for owner occupied housing. Exclusion of these items undertaken by authors to
produce Table 12 brings the distribution dynamics fro urban areas back to its common trend as exemplified by
Ravallion and Chen (2004), but for rural areas the result remains different: slight decline as opposed to a slight
increase in the official series. Note that our recalucations of concentration coefficients are based on assumption
of no-reranking while housing imputations and subsidies are removed, which may not be too problematic given
that this source represents a secondary source of income.
37
We first focus on the urban panel. Clearly, a number of factors were operating in a very
similar way to Poland and Russia and therefore represent a common transition story. As
noted by Khan and Riskin (2004) during the period under review, a number of significant
changes in the composition of urban income took place. There was a fall in the share of
wages and a rise in the share of incomes for the retirees (which includes payments to laid-off
workers). The share of income from individual enterprise also increased sharply although it
remained low in absolute terms and had surprisingly low and falling concentration coefficient
compared to ECA, which is suggestive of low-productivity activities most often by the poor.
In China, given underdeveloped safety nets, the dynamics are dominated by increased
dispersion of wages to a much greater extent than in Poland and Russia. This is the
continuation of a long-term trend. Wages were highly, and almost certainly artificially and
inefficiently, equalizing in 1988 with a concentration ratio of only 0.178 (Khan and Li). It
rose – gradually as implied by the slow pace of reforms - to 0.198 in 1995 and to 0.245 in
2002. Clearly, the rapid growth of private, foreign and mixed-ownership enterprises,
contributed to this increase, but slow restructuring of SOEs is likely to have arrested the pace
of change. At this juncture it is useful to be reminded that by 2003 still over 80 mln out of
250 mln. urban employed in China were working in SOEs (China Statistical Yearbook 2003).
Khan and Riskin include housing subsidies and imputed rents in their decomposition and
show that housing subsidy in kind and rental value of owned housing became substantially
less dis-equalizing as homeownership became widespread, undoing the effect of increased
wage dispersion. Since rental values imputation remain a source of controversy and not done
in Poland and Russia, we do not use this result in Table 12.
A mixture of transitional and typical development factors are at work as far as rural incomes
are concerned, Equally important contributions to the inequality dynamics were the
distribution of farm income, and income from wage labor. The first was already a strongly
equalizing source of income back in 1995, but much more strongly equalizing in 2002. The
Gini ratio of land distribution is remarkably low by the standards of all available international
estimates and is a result of equitable de-collectivization in China. Even this low inequality is
almost certainly due to regional differences in land endowment rather than inequality in any
region with given land endowment per rural household.
The reduction in the dis-equalizing effect of wages is perhaps partly explained by the rapid
growth in rural employment, outmigration and some reduction in the regional inequality of
access to wage employment, most likely all developmental rather than purely transitional
factors. Khan and Riskin note that wages derived from local TVEs (more or less a
transitional phenomenon) are far more dis-equalizing than other wages for 2002.
The effect of non-farm incomes according to different datasets (which exclude housing rents)
are considered to be large enough to outweigh the positive effects of equitable access to land
(which is though to be diminished by the continuous worsening of terms of trade). Thus
Benjamin, Brandt and Giles (2005) using a panel of 100 villages in rural China conclude that
inequality worsened considerably between 1995 and 1999 due to the falling role of
agriculture in rural areas and increased inequality in non-farm employment opportunities.
38
Luo and Zhu (2006) using data from the Living Standards Measurement Study in the
provinces of Hebei and Liaoning arrive at the opposite conclusion supporting Khan and
Riskin’s views. Decomposing the Gini index they find that two parts of non-farm incomewage and self –employment – had dramatically different concentration coefficients and
contributed to inequality in very different ways. Modeling household choices and assuming
that nonfarm income is a potential substitute for farm income they show that nonfarm activity
reduces rural income inequality by raising the income of poor households to a larger extent
than that of rich households.
Ravallion and Chen (2001) present an alternative analysis, the only full decomposition by all
income sources available based on official SSB data, but they focus only on rural areas from
4 provinces (see Table A 4. in the Annex 4). They differentiate further within components
of rural income and demonstrate that 104% of the increase in the Gini index over the period
1985-90 can be attributed to farm grain income, while 61% was attributable to income from
collectives (including TVEs) enterprises. Smaller positive contributions to rising inequality
came from self-employment in industry and construction (42%), labor earnings (36%) and
services (32%). Against these positive contributions to rising inequality, there were large
inequality-decreasing effects from private transfers (mostly remittance, -131% of the increase
in inequality) and other farm income (-65%).
Ravallion and Chen (2001) also use regression-based approach to decompositions
demonstrating that the increased returns to education were pushing inequality up, while
greater access to education was inequality reducing. Fixed farm assets were a noticeable
driver of inequality change, but access to farm land was relatively equal, and consistent with
results reported by Khan and Riskin (2004), dampening the effect of the raise in returns.
Location factors were shown to play a key role. Indeed, the changing differences between
mountainous areas and the plains accounted for 52% of the increase in the Gini index
between 1985 and 1990. This analysis already opens up the issue of identifying factors
determining the change in the distribution of an income source that we will look at in the
next sections, focusing on wages and transfers.
This brief overview of a vast literature on inequality in China has so far focused in turn on
urban areas (exemplifying the transition story), and rural areas (more developmental story)
separately. But the differences between urban and rural areas and changes in these
differences account for an unusually large share of inequality in China. These differences are
driven in part by transition factors, but mostly by the process of development. Looking at
these changes in detail and comparing them with those in ECA countries will help us assess
whether the experience of rising inequality in China is a preview of Russia’s future. To do
this we will use group-based decompositions. But before we turn to section III.2 we need to
look in more detail at the key driver of inequality and the “excess” inequality source for
Russia, viz., inequality of earnings.
39
Wage inequality
Extremely high wage inequality in Russia is a big part of the explanation of its “excess”
inequality compared to Poland and even urban China. Why has inequality in Russia
increased so rapidly? Several reasons were advanced to explain this phenomenon.
Arrears as reported by Lehmann and Wadsworth (2001) were responsible for up to a third for
the overall inequality in wages. At the peak of wage arrears in November 1998 as much as 64
percent of workers were owned back wages, and the Gini for actually paid wages was as
high as 0.58. By 2004 the share of workers in arrears fell to 15 percent (data are from
RLMS), and Gini index for paid wages fell to around 0.44, i.e. by just less than a third, as
predicted by the simulations undertaken by Lehmann and Wadsworth. But even at this level,
wage inequality appears to be high, much higher than in Poland or other CEE economies.
Therefore this factor may be responsible for producing an inverse U-shape evolution of
inequality for wages in Russia, but it does not explain the excess inequality in the distribution
of earnings in Russia.
Macroeconomic policies. Countries experiencing higher inflation rates or higher volatility of
inflation rates are normally characterized by higher earning dispersion. Russia had a
protracted period of instability, triple and double digits inflation rates and lacked any
established system of indexation. Adjustments to real wages were not synchronized and led
to the emergence of large differentials. However plausible this explanation is, it is
inconsistent with observed relative stability of real sectoral wages.
Wage policies. It is well known [see World Bank (2005c)] that minimum wage policies were
distinctly different in CEE, which also avoided increases in wage inequality and CIS where
inequality in wages has sharply increased. In CEE and SEE, minimum wages are set at
around 40 percent of the average wage; in contrast, in Russia its accounts for 10 percent of
the average wage. This allowed firms to maintain low-paid jobs that otherwise would have
been economically unviable, so that low minimum wage contributed to higher wage
dispersion. This has been a very important policy induced factor.
Sectoral differences, including private/public sector. Mean private sector wages in the early
transition in Russia were double the average wage in the public sector (see Commander et
al.), but subsequently the private sector premium dropped to some 20-30 percent (according
to recent study by Gimpelson et al.). This remains larger that in most CEE countries where
premia have largely disappeared, but is not a strong enough factor to drive excess wage
inequality in Russia. On the other hand, dispersion of average real wages across industries is
believed to be large. But it has not increased in Russia since 1990, and therefore cannot be
used to explain wage inequality growth (using Goskomstat data on mean wages by sector for
the period 1991-2004 the coefficient of variation of real wages remained stable).
Regional divergence. As opposed to relatively stable sectoral and inter-industry wage
differentials, regional variation of real wages (relative to national average) has almost tripled
in Russia between 1995 and 2003. Segmentation of labor markets is a common feature of
40
many transition economies, but in Russia this dispersion takes particularly extreme forms due
to institutional, infrastructure and geographical realities,
Regulations or lack of regulations. To explain very large divergence of wages in Russia
across regions, references are often made to the highly decentralized system of wage
determination in Russia (Flemming and Micklewright based on Layard). Wages are
determined at the level of enterprises and are extremely sensitive to changes in performance
of a firm, without any regulatory framework or institutions of organized wage bargaining
which are instrumental in bringing about wage equalization across sectors or regions in CEE
and SEE economies. That brings about greater flexibility of real wages in Russia which,
rather than employment changes, become the main channel of adjustment. It also leads to a
greater dispersion of wages across firms. Income policies were also pervasive in many CEE
countries in early transition. Wages were in several countries subject to controls and in
others to taxes (such as the Popiwek in Poland) related to the rate of increase of nominal
wage rates, enterprise wage bills or of average earnings. It had clear impact on the
distribution of wages in early transition and moderated increases in earning dispersion early
on.
Institutions of the labour market, including discrimination. Other factor contributing to wage
dispersion in Russia is its large informal sector. It is believed that wage data used for
assessment of wage inequality in Russia are severely distorted (both in the household and
firm surveys). The main reason is significant hidden cash component of compensations to
employees to avoid taxation of social contributions. But if anything this factor is likely to
make “real” distribution of earnings even more unequal thus it does not explain why Gini for
formal Russian wages looks so high (assuming that as elsewhere highly paid workers have
greater possibilities to be compensated “under the table”). Discrimination is sometimes
called upon as an explanatory factor, but the analysis of gender gap, which is normally
reflective of price discrimination in the labor market, reveals no particular problem in Russia
compared to other transition economies [see World Bank (2005c)]
Returns to human capital. As far as the empirical evidence regarding this driver is
concerned, both country-specific and cross-country studies find that returns to education
increased from the “pre-transition” period to the “early transition” period. The meta-study
by Fleisher and others (2004) also suggests that the sharpest increases in returns to education
took place during the early transition (see Annex 2). Paternostro and Tiongson (2005)
examine what has happened to the skills premium in transition economies through the late
1990s, or the period thereafter (through 2002 or 2003) using ISSP data (internationally
comparable survey). Russia does not stand out as having particularly large or distorted
pattern of returns compared to CEE. While the direction of the change of education premia
coincides with wage inequality increases, it is believed to play a less important role in Russia
than in Hungary or Poland. This factor therefore cannot be used to explain excess inequality
of Russian wages.
To summarize, several factors are likely candidates for an explanation of the persistently
large wage dispersion observed in Russia. Most of them are linked to the institutional and
regulatory framework of labor markets and thus are policy-induced.
41
To sort out the relative importance of various factors affecting the distribution of wages
simultaneously, researchers often rely on Fields decompositions of variance of wages into the
factors explaining the level of wages of individual workers (Mincer earning functions based
see Fields [2002] for details). Such decompositions for a number of ECA countries and for
China (Table 13) yield interesting results.
Table 13 shows that despite a significant increase in the education premia in Russia, it
remains a much less important driver of the overall inequality of wages than in Hungary and
Poland (but very similar to its size in China). At the same time, Table A2.3 in Annex 2
suggests that the size of education premia in Russia is fully in line with other transition
economies. Therefore it is not the suppressed role of human capital in wage determination
that makes this factor look small in Russia compared to Hungary and Poland, but the
extremely high dispersion of wages, in both explained and residual components. The
absolute contribution of the education premium to inequality in Russia is no different from
that in the more advanced transition economies of Central Europe.
There is a significant difference between all ECA countries with urban China in the role of
job experience, suggesting much faster pace of depreciation for firm-specific skills in the
former. The role of location factors on the other hand is significant in both Russia and
China, as well as that of private-public wage differential.
Table 13: ECA: Decomposition of wage inequality by factors of earnings, percent
Measured by variance of monthly wages
Poland
1995
Hungary
1997
Russia
1996
Urban China
2001
1988
1995
% Total
Explained
% Total
Explained
% Total
Explained
% Total
Explained
% Total
Explained
% Total
Explained
Variance
Variance
Variance
Variance
Variance
Variance
Variance
Variance
Variance
Variance
Variance
Variance
Education
11.5
34.9
25.9
61.7
2.7
8.1
3.5
11.9
0.3
1.0
3.0
Job Experience
3.8
11.6
2.1
5.0
1.5
4.6
0.2
0.7
16.5
52.4
9.1
Gender
6.8
20.5
3.8
9.1
6.5
20.0
4.9
16.7
2.2
7.0
1.9
Ownership
(private/public)
-0.6
-1.9
0.0
0.0
1.0
3.0
3.4
11.6
3.2
10.2
3.8
Sector
8.7
26.3
4.8
11.3
7.5
23.0
5.7
19.4
1.2
3.8
2.7
Location
2.9
8.6
5.5
13.1
13.5
41.3
11.7
39.8
8.1
25.7
14.3
Unexplained
66.9
57.9
67.3
70.6
68.5
65.1
Total
100.0
100.0
100.0
100
100.0
100.0
Sources: Poland and Hungary (Rutkowski, 2001); Armenia (Yemtsov, 1999). Georgia (World Bank. 1999); The
Russian Federation (Lehman & al., 1999a) and Lukianova for 2001 (2004) China: Knight and Song (2003) Location
includes urban/rural and region.
A study by Gustaffson et al (2001) comparing wage inequality in Russia and China in 1989
relied on a unique dataset for Russia which offers truly comparable wage data for pre- and
post- transition. The study shows that while wage inequality as measured by Gini in the pretransition period was very similar (around 0.27) in urban China and urban Russia, its
determinants differed somewhat. Most importantly, the returns to education were negligible
in China (but not in Russia). Subsequent developments of education wage premium had
slightly different pace: in China the increased education premium was a stronger driver
42
8.6
26.1
5.4
10.9
7.7
41.0
behind wage increases (albeit from lower base), while in Russia it played a less prominent
role in explaining wage inequality dynamics.
Market imperfections and distortions seem to be as pervasive in China as they are in Russia.
John Knight and Lina Song (2003) demonstrate it by using comparable urban wage survey
data from China. They show that while productive characteristics were increasingly
rewarded in China as marketization occurred, discrimination and segmentation and
corresponding inequalities also grew. Location factors continue to be a key determinant of
wage dispersion in China as they are in Russia. They conclude that the move towards a
fully-fledged labour market in China was by no means complete.
The analysis of regional differences which play the dominant role in explaining the
inequality change in China can be illustrated with urban/rural differences suggesting
significant impediments to the operation of market forces in China. Shi, Sicular and Zhao
(2002) explore the question of rural-urban inequality in greater detail for nine different
provinces using the China Health and Nutrition Survey (CHNS). Once they have taken into
account differences in living costs, the authors conclude that the apparent labor market
distortion in the form of registration system and other impediments for migration amount to a
rate of apparent taxation on rural wages of 81 percent. Shi (2002) finds that 28% of the rural
urban wage difference can be explained directly via the coefficient on the registration.
To the extent that impediments to migration reflect distortions inherited from the command
economy, the large role of regional factors as drivers of wage inequality in China and Russia
is a phenomenon of transition. Gradually market reforms and more competitive labor market
will lead to greater equalization of wages across regions.
We now turn to a more in-depth analysis of spatial and other group-based factors of
inequality.
III.2
Decomposition of inequality by groups
To what extent can inequality be explained by inequality between groups, such as rural
residents versus city dwellers, or high school graduates versus those with less than high
school education, or working families versus jobless households? The choice of these
partitions is not to any degree random. The labor market groups tend to capture the key
dimension of transition per se: emergence of new social classes and changing distribution
within the classes. Differences by education attainment help to assess the magnitude and
dynamics of inequality effects related to technological change. Finally, partition by location
(region) helps to assess the degree of “imperfections” as a driver of inequality.
To examine this further we decompose inequality into the contribution of inequality
‘between’ groups and inequality ‘within’ groups using the Theil entropy measure of
inequality, which, unlike Gini is fully decomposable (no overlap component). The share of
within-group inequality is the product of inequality within the group and the share of the
group. So, the share of a particular group to overall inequality may change either because
inequality within the group has changed or because the share of the group in the total
population has changed. The sum of the within- and between-group contributions equals 1.
43
Annex 3 presents a set of results from available studies using group decompositions for
Hungary, Poland and Russia. Graphs show the contribution of each component to the overall
inequality. Well in line with priors regarding the increase in education premia we see
increasing and large contribution of differences between education groups. Results are
consistent with wage story both in the magnitude of effects across countries and in terms of
their direction. We can look at this factor using our consistent series of consumption data
and expand the country and time coverage.
But alongside the rise in differences across different levels of education attainment, the
importance and persistence of location effects for Russia, and even in Hungary and Poland, is
striking. We therefore start with urban-rural dimension of inequality.
Urban-Rural
Figures 8 and 9 provide a series of snapshots of the extent of both between urban and rural
and within inequality in ECA and China. Several observations can be made.
Figure 8 Gap Between Urban and Rural Areas in ECA, 1993-2002
1981
1993
1996
1998
2000
2002
Per capita consumption premium in urban areas
+90%
+80%
+70%
+60%
+50%
+40%
+30%
+20%
+10%
+0%
-10%
-20%
Moldova
Armenia
Poland
Hungary
Romania
Russia
China
Source: Own estimates based on ECA Regional Archive, China – data on real incomes from Ravallion and Chen (2004), all
means include cost of living adjustment.
First, looking at the evolution of urban rural gap we observe no clear pattern associated with
transition: in some countries it went up, in others down. Clearly China stands out as a country
with an extremely large gap. In ECA following turbulent transition in some countries the gap
was actually reversed due to availability of some sources of livelihood in rural areas at a time
when unviable enterprises were restructured or closed in industrial cites). Not surprisingly as
reported by Shorrocks and Wan (2004) China has the highest between urban and rural areas
44
component of inequality in the world (37 percent in 2000). It is also a particularly dynamic
component with changes originating both from the change of the relative mean (Figure 8)
driven by terms of trade shifts (Ravallion and Chen) as well as rapid dynamics of migration:
the share of urban in the population in China went from just over 20 percent to 40 percent
between early 1980s and 2000. Only a few low income economies in ECA have a similar
share of rural population, and none displays so clear cut a trend.
Figure 9 Urban and Rural Areas: Gini index in selected ECA countries and China
China: Gini Index for Urban and Rural areas
Russia: Gini Index for Urban and Rural areas
Russia Rural
China Rural
Russia Urban
China Urban
0.500
0.450
0.400
0.350
0.300
0.250
0.150
0.200
1993
1994
1995
1996
1997
1998
1999
2000
2001
1981
2002
Moldova: Gini Index for Urban and Rural areas
Moldova Rural
1987
1989
1991
1993
1995
1997
1999
2001
Armenia: Gini Index for Urban and Rural areas
Armenia Rural
0.500
0.400
0.400
0.300
0.300
0.200
Armenia Urban
0.200
1994
1995
1996
1997
1998
1999
2000
2001
2002
1993
Poland: Gini Index for Urban and Rural areas
Poland Rural
1994
1995
1996
1997
1998
1999
2000
2001
2002
Hungary: Gini Index for Urban and Rural areas
Poland Urban
Hungary Rural
0.500
0.500
0.400
0.400
0.300
0.300
0.200
1993
1985
Moldova Urban
0.500
1993
1983
Hungary Urban
0.200
1994
1995
1996
1997
1998
1999
2000
2001
2002
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
Note : Solid line – Rural Gini, Broken line – Urban Gini, consumption per capita with cost of living adjustment.
Source: Own estimates, ECA data archive China Ravallion and Chen (2004), income per capita
45
Looking at the evolution of within-urban and within rural components, it is found that there
are no transitional paths associated with it in Poland and Hungary, and some clear effects
from the land ownership or use rights reforms which led to a decline in Armenia, Moldova
and to some extent Russia. In any event, the rural Gini of 0.25 in China in 1980 reflects the
result of an equitable land distribution during de-collectivization. A similar redistribution of
land ownership occurred in low income CIS countries such as Armenia and Moldova which
also have labor-intensive agriculture, and therefore it is reasonable to expect that the effects
are likely to have been equitable. Table 14 also shows a large fall in within-rural area
inequality in these countries, as opposed to Georgia where land reform was much less
comprehensive and a virtually flat trajectoryin Hungary and Russia.
Going even broader, Table 14 reports a first panel of data (A) on the role of differences
between urban and rural areas using full Theil decomposition approach. There were
significant changes in population shares: in Hungary the share of rural areas dropped from 39
percent to 35 percent of population during the period under review, in Latvia the share of
Riga increased from 33 to 38 percent, and in Tajikistan the share of rural areas dropped from
78 to 73 percent in just 5 years. Using initial shares as counterfactuals we can see that these
demographic shifts had an impact on inequality: typically they were dampening the increase
in inequality. In Latvia, for example, had the initial distribution of population by location
stayed the same, inequality would have been 15 percent higher and in Tajikistan 12 percent
higher by 2003.
The between-component went up everywhere except in Hungary. Within capital city
inequality increased everywhere, again with the exception of Hungary and stayed virtually
the same in Russia. This shows a reversal of the trend of the early to mid-1990s when there
were significant shifts into subsistence agriculture from capital and secondary cities as
enterprises in those latter areas became unviable in countries where safety nets were
ineffective. This would have been expected to have been inequality-increasing. The reversal
is clearly observed in Tajikistan and Moldova, but impact is practically nil in Russia.
Summing up, location factors play a relatively minor role as driver of inequality and poverty
reduction in the transition economies of ECA compared to China and the rest of the
developing world. In general, income in rural areas is lower than income in urban areas.
Capital cities have much higher incomes and thus are expected to have more favorable
conditions for poverty reduction, although higher inequality there may diminish their
potential in reducing poverty with a given growth rate. But it is unlikely that this driver will
strongly influence the dynamics of inequality in Russia going forward. In contrast to China,
where its sheer size (over a third of the national inequality) and dynamics make it one of the
leading causes for changes in inequality, the differences between urban and rural areas in
Russia explain less than 3 percent of national inequality and within – rural inequality is
influenced by one-time factors. We therefore need to focus on within-urban drivers as key
factors that will determine future inequality dynamics.
46
37%
34%
23%
0.189
0.189
0.184
33%
37%
31%
1.193
0.976
0.824
of which within Capital city
of which within other urban areas
of which within rural areas
Capital city
Other urban areas
Rural areas
Capital city
Other urban areas
Rural areas
Capital city
Other urban areas
Rural areas
0.126
0.206
0.217
Poland
1998
2002
0.167
0.178
Romania
1998
2002
29%
49%
19%
3%
100%
31%
58%
5%
5%
100%
28%
51%
14%
7%
100%
42%
56%
3%
100%
31%
58%
11%
100%
38%
43%
19%
0.125
0.124
0.194
0.189
0.198
0.187
0.190
0.195
0.266
35%
48%
17%
41%
55%
4%
40%
52%
8%
Population shares
0.116
0.124
0.127
45%
55%
0.170
0.157
46%
54%
0.154
0.160
Theil entropy index for per capita consumption
28%
36%
30%
5%
100%
Decomposition of Theil inequality measure :
0.149
Hungary
1993
2002
27%
68%
6%
0.258
0.177
0.182
32%
57%
8%
3%
100%
0.205
0.698
0.852
1.357
0.885
0.993
1.251
0.897
1.024
1.149
0.834
1.093
1.420
0.805
1.088
1.386
0.898
1.085
0.783
1.181
0.944
0.980
1.504
0.801
1.065
1.148
26%
68%
6%
0.168
0.181
0.178
19%
70%
7%
4%
100%
0.186
Russia
1997
2002
Real means, relative to national mean per capita consumption =1.00
30%
32%
38%
0.189
0.177
0.241
16%
19%
50%
16%
100%
100%
6%
0.254
0.198
Of which between locations
Theil entropy measure
Latvia
1998
2002
A. By location (for inequality measured by consumption per capita)
Table 14 Decomposition' of Inequality: Share of between and within group inequality in Theil entropy index
0.873
0.998
1.467
63%
19%
17%
0.218
0.230
0.212
50%
18%
22%
9%
100%
0.240
0.911
0.882
1.427
63%
19%
18%
0.184
0.176
0.214
51%
14%
27%
9%
100%
0.209
Moldova
1998
2002
47
0.962
1.032
1.385
78%
16%
6%
0.131
0.158
0.151
69%
18%
9%
3%
100%
0.142
0.952
0.999
1.370
73%
18%
9%
0.169
0.213
0.217
62%
20%
15%
4%
100%
0.190
Tajikistan
1999
2002
7%
15%
54%
4%
20%
0.158
0.189
0.267
0.183
23%
58%
4%
15%
0.814
0.980
0.902
1.377
Of which within primary education
Of which within second education
Of which within vocational education
Of which within tertiary education
Within primary education
Within secondary education
Within vocational education
Within tertiary education
Primary education
Secondary education
Vocational education
Tertiary education
Primary education
Secondary education
Vocational education
Tertiary education
100%
Decomposition:
of which between education group
0.198
Theil entropy measure
23%
40%
5%
19%
12%
0.206
0.217
19%
24%
26%
18%
13%
100%
20%
21%
19%
14%
26%
100%
Decomposition:
100%
0.126
Poland
1998
2002
13%
19%
25%
32%
11%
100%
0.167
13%
18%
25%
25%
20%
100%
0.178
Romania
1998
2002
12%
52%
7%
30%
0.188
0.113
0.110
0.136
28%
17%
49%
7%
0.208
0.163
0.184
0.165
0.195
0.158
0.130
0.181
11%
38%
27%
24%
13%
37%
28%
22%
Population shares
0.125
0.109
0.095
0.105
8%
25%
28%
40%
0.169
0.139
0.139
0.155
8%
20%
33%
39%
0.153
0.139
0.138
0.145
Theil entropy index for per capita consumption
22%
38%
5%
23%
11%
100%
0.149
Hungary
1993
2002
22%
34%
40%
5%
0.176
0.207
0.193
0.256
22%
39%
31%
5%
2%
100%
0.201
1.669
0.761
0.944
0.644
1.483
0.962
1.138
0.844
1.365
0.962
1.079
0.824
1.619
0.817
1.150
0.827
1.647
0.692
1.263
0.789
1.621
0.915
1.084
0.884
1.804
1.123
0.981
0.789
1.174
0.997
0.904
0.902
1.217
0.986
0.880
0.794
24%
34%
39%
3%
0.164
0.177
0.176
0.159
27%
37%
29%
2%
5%
100%
0.181
Russia
1997
2002
Real means, relative to national mean per capita consumption =1.00
19%
11%
48%
21%
0.224
0.195
0.189
0.171
28%
6%
34%
9%
22%
100%
0.254
Latvia
1998
2002
B. By education of the household head (for inequality measured by consumption per capita)
1.502
1.092
0.922
0.842
11%
23%
35%
31%
0.203
0.220
0.237
0.219
14%
23%
32%
24%
8%
100%
0.240
1.514
1.061
0.930
0.866
11%
19%
40%
30%
0.189
0.189
0.214
0.166
15%
18%
38%
21%
8%
100%
0.209
Moldova
1998
2002
48
1.239
1.034
0.942
0.922
13%
27%
26%
34%
0.148
0.136
0.141
0.126
17%
27%
24%
28%
4%
100%
0.142
1.221
1.108
0.902
0.919
18%
18%
42%
22%
0.178
0.176
0.175
0.202
21%
18%
35%
22%
4%
100%
0.189
Tajikistan
1999
2002
0.233
0.221
0.339
0.196
within "informal" subsistence
within LM group of non-working
62%
5%
26%
6%
of which within LM group of wage earners
of which within LM group of self-empl
of which within LM group of subsistence
of which within LM group of non-working
20%
8%
17%
7%
LM group of subsistence farmers
LM group of non-working /transfer rec.
1.037
0.865
0.963
0.929
0.990
0.903
LM group of self-employed
LM group of subsistence farmers
LM group of non-working / transfer rec.
0.891
0.693
1.116
1.055
8%
14%
12%
67%
6%
7%
12%
60%
4%
89%
0.180
0.156
0.203
0.184
Source: authors’ calculations based on ECA regional data archive)
1.042
1.020
LM group of wage earners
Russia
9%
7%
LM group of self-empl
Relative means
62%
68%
Russia
8%
18%
9%
62%
1%
98%
LM group of wage earners
Percentage population
0%
of which between LM groups
Decomposition:
Percentage to base period
100%
0.200
0.172
within "informal" of self-empl
Russia
0.207
0.194
within "formal" wage earn/entr
0.193
0.214
0.218
Theil entropy measure
Russia
1999 2002
1997
0.167
0.157
0.192
0.173
0.186
5%
15%
26%
37%
5%
87%
5%
24%
31%
39%
1.050
0.853
0.911
1.181
1.163
0.829
0.908
1.157
Kazakhstan
7%
29%
28%
36%
Kazakhstan
7%
21%
27%
40%
5%
100%
Kazakhstan
0.213
0.176
0.227
0.204
0.214
Kazakhstan
2001 2003
0.325
0.354
0.247
0.205
0.271
13%
27%
37%
19%
1%
97%
12%
20%
45%
23%
0.785
1.212
0.942
1.046
0.916
1.081
0.941
1.086
Georgia
10%
20%
50%
20%
Georgia
11%
29%
41%
17%
3%
100%
Georgia
0.368
0.339
0.239
0.228
0.279
Georgia
1999 2002
C. Detailed decomposition by household labor market status
0.243
0.134
0.213
0.208
0.209
14%
17%
16%
33%
7%
87%
13%
38%
16%
32%
1.161
0.784
1.152
1.222
1.071
0.774
1.095
1.193
Moldova
9%
48%
7%
36%
Moldova
8%
29%
10%
43%
9%
100%
Moldova
0.196
0.187
0.277
0.239
0.240
Moldova
1998 2002
0.240
0.160
0.170
0.210
0.187
8%
17%
49%
51%
0%
125%
5%
17%
42%
35%
1.050
0.956
1.170
1.071
0.931
0.955
1.018
1.011
Tajikistan
0%
74%
15%
11%
Tajikistan
1%
66%
19%
13%
2%
100%
Tajikistan
0.433
0.138
0.159
0.171
0.149
Tajikistan
1999 2002
0.149
0.133
0.243
0.142
0.149
16%
10%
8%
80%
2%
116%
13%
13%
18%
56%
1.006
0.911
0.824
1.059
1.023
0.899
0.816
1.077
Romania
12%
13%
14%
61%
Romania
11%
11%
14%
60%
3%
100%
Romania
0.126
0.121
0.156
0.118
0.128
Romania
1998 2002
0.114
0.089
0.096
0.104
0.107
3%
66%
19%
2%
34%
2%
65%
8%
1%
25%
27%
3%
15%
56%
0.921
0.845
1.089
1.022
0.968
1.041
1.172
0.970
Hungary
26%
1%
8%
66%
Hungary
124%
100%
Hungary
0.092
0.086
0.077
0.084
0.087
Hungary
1993
2002
21%
4%
46%
48%
1%
119%
17%
2%
37%
43%
1.092
1.249
0.967
0.988
1.010
1.263
0.988
0.992
Poland
15%
2%
43%
41%
Poland
15%
3%
43%
38%
1%
100%
49
0.180
0.202
0.189
0.171
0.181
Poland
0.147
0.146
0.157
0.145
0.152
Poland
1998 2002
Regional factors
While the urban/rural dichotomy in Russia clearly is nowhere near comparable to China’s, the role
of regional differences is enormous and puts Russia back in a position which is very comparable to
that of China. The analysis using official per capita income data series undertaken by Yemtsov
(2002), shows that between-regional factors accounted for about a third of the overall inequality in
Russia by the year 2000. The analysis also identifies an increase in the between-regions
component as a key driver of the inequality change between 1995 and 2000. Lack of convergence
across Russian regions in mean real incomes is also presented as a major factor of inequality
change going forward by Dolinskaya and by Fedorov.
Quite revealing is the role of resource rich regions and Moscow in driving the inter-regional
income disparities. When they are “removed” from the national distribution, the betweeninequality component contribution to rising inequality vanishes (Yemtsov 2002). While these
factors may well reveal lack of fiscal redistribution, and market failures they may also simply
represent statistical artifacts.
They may reflect the idiosyncrasies of Russian statistical system which continues to face serious
difficulties in measuring population incomes (over 40 percent of the wage bill in the national
accounts is reported as “hidden” compensations with unknown or very hypothetical distribution
across income groups and regions). The official household surveys in Russia do not collect
information on incomes. Instead they use the reported expenditures and savings (known to be
severely distorted and prone to measurement error) to construct a would-be indicator of incomes at
the household level and then compares it to the macro-based estimates of population incomes
(themselves prone to measurement problems). It then assigns all of the gap between survey-based
and macro-based average incomes to the top decile of the distribution. This is likely to distort the
representation of the actual extent of the regional inequalities, because the financial system
operates in a way which serves to concentrate money balances (used to assess changes in the
money incomes at the macro level) in a few financial centers (World Bank 2005 d)
Use of direct survey measurements – on consumption or on incomes (from RLMS) shows
strikingly smaller role of regional variation: about 15 percent of the overall inequality can be
ascribed to the between regional differences in means (ibid). Thus, the persistency of regional
factors is evident, but their role as drivers of inequality change is not.
Sector
Close to urban –rural story is a process of structural change in transition. Most CEE countries
have witnessed a fall in the number of blue-collar manufacturing jobs and an increase in white
collar service sector jobs. In contrast, in low income CIS countries, deindustrialization was
associated with an increase in agricultural employment. In the Czech Republic, market services as
a share of total employment increased by about 5 percentage points during the transition, while
manufacturing’s share fell by about 3 percentage points. In contrast, in the Kyrgyz Republic,
market services gained little as a share of total employment. However, there was a dramatic fall in
the share of manufacturing (about 17 percentage points) and a huge shift toward agriculture, which
increased its share by some 20 percentage points.
50
It is not clear whether the significant increase in the share of agricultural employment in the CIS
countries is only temporary or represents a more profound and long-lasting reversion toward
employment patterns more typical of countries with relatively low income per capita. But it was a
factor that influenced the dynamic of inequality in ECA in general and specifically in Russia. Just
between 1998 and 2002 inter-sectoral differences increased their contribution to the overall
inequality in Russia from 2 to 6 percent, but remained stable in Poland at around 6 percent. The
services share of the overall inequality expanded in both Russia and Poland, but it's within sector
inequality differs: in Poland it is the most unequal sector, while it is second in Russia after
manufacturing. 17 The previous discussion of the uneven pace of enterprise restructuring in Russia
as a key driver of excess inequality is again confirmed by this findings. Agriculture does not seem
to play an active role in these countries.
In contract to this, Ravallion and Chen (2004) demonstrate that agriculture sector growth in China
has been inequality reducing in both rural and urban areas whereas secondary and tertiary sector
growth have been inequality increasing both nationally and in urban areas. Gap premium between
the modern a d traditional sector increased from 20 percent to 35 percent, driving inequality up.
That again highlight the nature of inequality increase at the national level in China as being
primarily a developmental phenomenon.
What does this imply for Russia’s inequality outlook? There is clearly a potential for increasing
inequality following the Polish pattern of inequalizing growth of the service sector. But at the
same time over the medium term a powerful effect may come from both the moderation of the
inequalities within the manufacturing sector and its secular decline.
Labor market
In Panel C of Table 14 we focus on the market for labor where the bulk of incomes are earned,
dividing up households into groups characterized by wage employment, entrepreneurial activities,
subsistence activities, and non-employment (retirement, unemployment, and so on).18 These
partitions reflect purely transitional factors.
A few broad generalizations emerge. First, between-labor market-groups inequality has had little,
if any, role to play in explaining changes in inequality in the period under consideration.
Second, the growth of entrepreneurship has been the most significant factor pushing up inequality
in most countries. This is because as a group it is associated with higher inequality in outcomes
than wage employment or subsistence activities, and its share in total population has been rising.
There are however exceptions to this finding, notably Romania and Georgia, where a decline in the
17 Results of sectoral decompositions are not reported in Table 14 but available from authors on request.
18 The definitions used area as follows. Dependence on: (i) wage employment: no working members who are selfemployed and minimal income from self-production (<5 percent); (ii) entrepreneurial activities: at least one adult in
self-employment but minimal income from self-production (<5 percent); (iii) subsistence activites: at least one adult
in self-employment and significant income from self production (>5 percent); (iv) non-employment : no adult in
employment or self-employment.
51
share of households characterized by entrepreneurial activity has resulted in a falling contribution
of this group.
Third, the rise in the ‘contribution’ of the non-employed (transfer recipients) is an important factor
behind rising inequality particularly in EU-8 and SEE. The rise is owed to growing inequality
within this group accompanied, in some cases, by the rising share of this group. Growing
inequality amongst the non-employed may be a reflection of the increasingly poor opportunities
for those who are unemployed or out of the labor force to sustain their standard of living relative to
pensioners, and can be related to the failure to raise the share of the employed in total population.
Beyond these generalizations, how different factors come together is very much a country-specific
matter. In Russia, in particular where overall inequality has somewhat receded during the period
we look at, the main factor is the shift from self-employment (whether entrepreneurial or
subsistence) to wage employment accompanied by a decline in inequality among wage earners.
One factor explaining this decline is the reduction in arrears which has been a feature of the
economic recovery in the CIS. Wage arrears were regressive in impact, driving up inequality
among wage recipients (Lehmann and Wadsworth, 2001). It is therefore not unlikely that arrears
reduction has been beneficial to equality, but it is also likely that this factor has come to an end.
In Moldova, too, overall inequality declined. However, this is not due to changing shares of
different groups, but a decline in within-group inequality for all major groups i.e. wage employees,
entrepreneurs, and subsistence farmers. The reduction in wage inequality may be on account of
arrears reduction. However, the reduction in inequality among agricultural self-employed and
rural residents engaged in subsistence farming is a likely outcome of somewhat delayed, but
equitable land reform. In contrast, in Poland and Romania upward pressure from non-workers has
been reinforced by rising inequality among wage earners. This is no doubt related to the further
decompression in wages in these countries (World Bank 2003a, 2004, 2005a).
Education
Panel B in Table 14 presents the evolution of differences across education groups. Looking first at
the share of between component we observe a picture strikingly different from that provided by the
urban/rural divide. Not only it is much larger, it is also clearly and consistently increases
throughout the region.
In Latvia, Poland, Romania and Tajikistan, overall consumption inequality as measured by the
Theil entropy index went up. However, by education of household head, in no country did the
between-component fall and there were large increases in the between-component in Latvia,
Poland and Romania and also in Russia where however the contribution of the between-component
is relatively low.
Changes according to education group are informative. The contribution of the within-component
in primary education fell everywhere, mostly reflecting their fall in the population. The relative
position of this group sharply deteriorated. Groups with specific skills (vocational education) lost
in relation to other groups, especially in fast restructuring economies. At the same time the withincomponent in tertiary education went up everywhere, and particularly in Latvia, Russia and
Tajikistan. In fact, the contribution of the within-component in tertiary education was among 20 –
25% in most cases, reflecting a fast pace of technological change but also possibly increased
52
dispersion of education quality and revealed differences in adaptability. Overall Russia lags
behind CEE countries in both size and intensity of changes.
This brings us back to the importance of restructuring for understanding inequality changes. “Fast
restructuring” in ECA is clearly associated with increasing differences across groups and richer
groups becoming the stronger drivers of inequality. Protracted restructuring is reflected in the
between group component being small, but within groups component increasing. This qualifies
significantly the analysis presented in the World Development Report 2006 [World Bank (2005a)]
which shows a relatively minor role of between –education group differences in ECA and implies,
based on global experience, that this might point to lower inequality of opportunities. The
interpretation of the between component can be very different for these groups of countries
because of transition dynamics and should not be always interpreted in a positive light.
IV.
What are the possible determinants of inequality changes in Russia going
forward? Conclusions and policy implications.
How then should one answer the question that forms the subtitle of this paper, viz., does China’s
rising inequality portend Russia’s future? The answer of this paper is: no. First, as Ravallion and
Chen (2004) have shown, faster growth in China did not drive higher inequality. Hence an
acceleration of growth in Russia and other CIS countries will not automatically generate higher
inequality. Second, a significant determinant of China’s inequality derives from the rural-urban
divide, viz., migration from the former to the latter and rapid changes in the sectoral composition
of output, a classic development phenomenon for which there is no obvious Russian analogue.
All that said, the paper has found there is no single driver of inequality in the different transition
countries that have been the subject of our analysis and that different drivers combine to create a
complex patchwork which is rich enough to allow a wide variety of outcomes across countries and
over time. Inasmuch as transition-related factors have not played themselves out in the countries
of Eastern Europe and the former Soviet Union, the lessons to be drawn from an analysis of
Chinese experience need to be borne in mind in thinking about the evolution of inequality in
Russia.
A dominant driver of inequality common to Central Europe, China and Russia has been wage
decompression. While the share of wages has declined in the ECA transition economies and more
modestly so in urban China, their concentration coefficient, which depends both on how unequally
wage incomes are distributed and how closely they are correlated with total income, has increased
significantly in all cases. The analysis showed that wages have become less unequally distributed
in Russia in the late 1990s and early 2000s, reversing the trend of increasing inequality in earlier
years. Although this has moderated in the last few years, that is due to a reduction of wage arrears
which, however, is a one-time phenomenon. Could wage inequality increase further? This could
happen to some extent, reflecting increases in education premia in Russia. But an important policy
issue in Russia is to reduce the size of the informal economy by creating better jobs through
improvements in the investment climate which would encourage the entry of new firms and the
restructuring and exit of unviable firms, accompanied by more active use of the social safety net to
manage the resulting job turnover [World Bank (2005c)]. Such reforms would have the effect of
53
mitigating the disequalizing effect of wages. On balance, it is not unreasonable to expect that
continuing market-oriented reforms would prevent further increases in inequality.
Our analysis has shown that location is an important determinant of wage inequality in Russia.
While this might remain persistent, it need not aggravate an increase in inequality. On the
contrary, to the extent such inequality has roots going back to central planning, it can be mitigated
through freer movement of goods and labor. In addition, depending on societal attitudes to
inequality, intergovernmental fiscal transfers can play a role as well.
The focus in this paper has been on income disparities as they emerged in transition. Good health
status, educational achievement, or living conditions jointly reflect the nature of the public
interventions in these sectors and the ability of households to invest in human and physical capital.
As argued in the World Development Report 2006 increasing disparities in the non-income
dimensions of well being between different groups, and between poor and non-poor, affect income
inequality in the long run.
The analysis undertaken for the report Growth, Poverty and Inequality in Eastern Europe and the
former Soviet Union, 1998—2003 [World Bank (2005b)] demonstrates that there has been a
decline in the quality of education across the region. It suggests that while the region may be able
to live off its previous investments, these are eroding rapidly. The failure to maintain human and
physical capital is resulting in environments that are not appropriate for effective education
services and that are reflected in declining performance. This is particularly observable in rural
areas and among poor households, who typically face the worst conditions. In Russia, as in the
other ECA transition countries, policy interventions that improve the quality of education services
are essential if the decline is to be arrested and reversed. This will require fundamental reforms in
public service delivery.
Finally, it is important to remember that the evolution of inequality in Russia is not pre-determined
by endowments and existing trends and can be influenced by policy. Those policies would include
(i) reforms of the investment climate that encourage entry of new firms and restructuring and
closure of unviable enterprises, together with the use of targeted safety nets to facilitate the
associated job turnover, and (ii) policies which permit freer movement of labor and goods to offset
the legacy of central planning with respect to location; and (iii) strengthening service delivery in
education where the evidence suggests a continuing decline in quality.
Looking forward, it is likely that transition-related factors will become less important in the
evolution of inequality in Russia compared to factors such as technological progress, global
changes in skills premia, the effects of demographic changes and migration. To the extent that
China’s income distribution is influenced by its greater integration in world markets, its experience
is relevant for Russia in pointing to the role of such long-term factors. That analysis remains to be
done.
While the paper, on assessing the available evidence, has argued that further increases in inequality
in Russia are not inevitable, it is worth noting that it identifies several gaps in our understanding of
inequality on which future research might profitably focus. Such research would include an indepth exploration of (i) on non-income dimensions of inequality and inequality of opportunities;
54
(ii) the role of technological change and globalization; (iii) housing policies, subsidies and imputed
rents and (iv) the effect of tax policies on the distribution of income.
We end by mentioning that ongoing joint work between the Russian government, DFID (UK) and
the World Bank aims at putting the Russian household survey data in the public domain, and hence
offers the prospect that significant advances in understanding the nature and dynamics of
inequality in transition will be made in the near future. Such an approach of open access to data
will help policy makers and the development community monitors developments in poverty and
inequality associated with policy changes undertaken in the Russian Federation.
55
Bibliography
Aghion, P. and O. Blanchard (1994). ‘On the Speed of Restructuring’ (mimeo), MIT: Cambridge MA, and EBRD: London.
Aghion, Philippe and Simon Commander On the dynamics of inequality in the transition The Economics of Transition Volume 7: Issue 2,
July 1999
Alexeev, Michael and Clifford Gaddy, (1993), “Income Distribution in the U.S.S.R. in the 1980s”, Review of Income and Wealth, Series 39,
pp23-36
Atkinson, Antony B., and John Micklewright. 1995. “Economic Transformation in Eastern Europe and the Distribution of Income.”
Cambridge: Cambridge University Press.
Atkinson, Antony (2003) Income Inequality in OECD Countries: Data and Explanations, Revised version of paper prepared for the CESifo
conference on “Globalization, Inequality and Well-Being” in Munich, November 8-9, 2002.
Benabou, Roland (1996) “Inequality and Growth”. In Ben Bernanke and Julio Rotemberg (eds.) NBER Annual, Cambridge: MIT Press.
Birdsall N. and Nellis J. (2003) Winners and Losers: Assessing the Distributional Impact of Privatization, World Development
Bourguignon, Francois, (1979), “Decomposable Income Inequality Measures”, Econometrica, 47, 4, pp901-920
Bourguignon, François , Francisco H. G. Ferreira and Phillippe G. Leite ()Beyond Oaxaca-Blinder: Accounting For Differences In Household
Income Distributions Across Countries.
Bourguignon, F. (2003), “The growth elasticity of poverty reduction: explaining heterogeneity across countries and time periods”, in T.
Eicher and S. Turnovski, (eds), Inequality and Growth: Theory and Policy Implications, MIT Press, Cambridge, Mass.
Buckley, Robert and Eugene Gurenko, (1997), “Housing and Income Distribution in Russia: Zhivago’s Legacy”, World Bank Research
Observer, 12, 1, pp19-32
Campos, Nauro F., and Dean Jolliffe. 2003. “After, Before and During: Returns to Education in Hungary (1986–1998).” Economic Systems
27: 377–90.
Chen, Shaohua, Yan Wang (2002) China's Growth and Poverty During the 1990s, Trends between 1990 and 1999. the World Bank Policy
research working paper 2651
Commander, S. and A. Tolstopiatenko (1996). ‘Restructuring and Taxation in Transition Economies’ Policy Research Working Paper 1625,
World Bank, Washington DC, July.
Commander, S. and A. Tolstopiatenko (1997). ‘Unemployment, Restructuring and Inequality in the Transition’ (mimeo) EBRD: London.
Commander, Simon and Une Lee (1998) How does public policy affect the income distribution? Evidence from Russia, 1992-1996, EBRD
and The World Bank, Mimeo.
Commander, S., A. Tolstopiatenko and R. Yemtsov (1999). ‘Channels of Redistribution: Inequality and Poverty in the Russian Transition’,
Economics of Transition 7(2):411–47.
Coricelli, Fabrizio. 1998. Macroeconomic Policies and the Development of Markets in Transition Economies. Budapest: Central European
University Press.
Cowell, F. and S.Jenkins (1995) “How Much Inequality Can we Explain? A Methodology and an Application to the United States”,
Economic Journal, Vol.105: 421-430.
Cowell, F (2000) "Measurement of Inequality." Handbook of Income Distribution. Volume 1. Handbooks in Economics, Vol. 16. Eds.
Anthony B. Atkinson and Francois Bourguignon. Amsterdam; New York and Oxford: Elsevier Science, North-Holland, 2000. 87166.
Davies J. and T. Shorrocks Wealth Holdings in Developing and Transition Countries. Memo 2005
Démurger, Sylvie, Martin Fournier and Li Shi (2005) Urban income inequality in China revisited (1988-2002). French Centre for Research
on Contemporary China, Hong Kong, mimeo,http://www.hiebs.hku.hk/working_paper_updates/pdf/wp1119.pdf
56
Fields, G. (2002) “Accounting for Income Inequality and Its Change: A New Method, with Application to the Distribution of Earnings in the
United States”, Cornell University, mimeo, October. Available at: http://www.ilr.cornell.edu/extension/files/20021024104432pub785.pdf
Flabbi, Luca, Stefano Paternostro and Erwin R. Tiongson (2005) Returns to Education in the Economic Transition: 1986-2002. The World
bank. Mimeo.
Fleisher, Belton M., Sabirianova, Klara, and Wang, Xiaojun. (2004) “Returns to Skills and the Speed of Reforms: Evidence from Central and
Eastern Europe, China, and Russia,” William Davidson Institute Working Paper Number 703.
Flemming, J. and J.Micklewright (1999) “Income Distribution, Economic Systems and Transition”, Ch.14 of A.B.Atkinson and
F.Bourguignon (Eds.) Handbook of Income Distribution, Elsevier Science BV.
Förster, M., D. K. Jesuit, and T. M. Smeeding (2003) “Regional Poverty and Income Inequality in Central and Eastern Europe.” United
Nations University/World Institute for Development Economics Research (UNU/WIDER) Discussion Paper 2003/65.
Garner, Thesia and Katharine Terrell, (1997), “The Distribution of Welfare in Transition Economies: Market versus Policy in the Czech
Republic and Slovakia”
Gorodnichenko, Yuri, and Klara Sabirianova Peters 2004. “Returns to Schooling in Russia and Ukraine.” Discussion Paper 1325, IZA, Bonn
Gottschalk, Peter, and Timothy M. Smeeding. 1997. “Cross-National Comparisons of Earnings and Income Inequality.” Journal of Economic
Literature 35 (June): 633–87.
Gustafsson, Bjorn, Li Shi, Ludmila Nivorozhkina, and Katarina Katz (2001) Rubles and Yuan: Wage Functions for Urban Russia and China
at the End of the 1980s*. Economic Development and Cultural Change. University of Chicago.
Gustafsson, Bjorn, Li Shi (2001) The Anatomy of Rising Earnings Inequality in Urban China. Journal of Comparative Economics 29, 118–
135.
Gustafsson Bjorn, and Ludmila Nivorozhkina (2005) How and Why Transition made Income Inequality Increase in Urban Russia: a Local
Study.
Haltiwanger, John, Harmut Lehmann, and Katherine Terell (2003) “Symposium on Job Creation and Job Destruction in Transition
Economies.” Economics of Transition, Vol. 11: pp. 205-219.
Ivaschenko, Oleksiy (2002) Growth and Inequality: Evidence from Transitional Economies. Department of Economics, Gothenburg
University, Paper Presented at 2001 UN WIDER Conference in Helsinki
Kakwani, Nanak, (1994), “Income Inequality, Welfare and Poverty in Ukraine”, Policy Research Department, Working Paper
Keane, M. and Prasad, E. (2002 a) Inequality, Transfers and Growth: New Evidence from the Economic Transition in Poland. IZA
Discussion paper Discussion Paper No. 448 March 2002
Keane, M. and Prasad, E. (2002 b) Changes in the Structure of Earnings During the Polish Transition. IMF Working Papers 135.
Krueger, D and Perri, F (2002) “Does Income Inequality Lead to Consumption Inequality?”, NBER Working Paper No 9202.
Kanbur, R. and M. Tuomala (2002) ‘Understanding the Evolution of Inequality during the Transition: The Optimal Income Taxation
Approach’ (mimeo), Cornell University: Ithaca.
Kanbur, R and Zhang, Xiaobo (2001) Fifty Years of Regional Inequality in China: A Journey through Revolution, Reform and Openness.
CEPR Discussion Paper No. 2887, Centre for Economic Policy Research, London.
Kattuman, Paul and Gerry Redmont (2001) Income Inequality in Early Transition: The Case of Hungary 1987–1996. Journal of Comparative
Economics 29, 40–65
Kertesi, Gábor and János Köllő (2001) “Economic transformation and the revaluation of human capital– Hungary, 1986–1999.” Budapest
Working Papers on the Labour Market BWP No. 2001/4. Budapest: Institute of Economics, Hungarian Academy of Sciences.
Khan, Azizur R., and Carl Riskin (2004) “China’s household income and its distribution, 1995 and 2002.” Mimeo,
http://www.economics.ucr.edu/seminars/spring04/05-28- 04AzizurKhan.pdf
57
Klugman, J. (1998). Wage in Transition: The case of Uzbekistan. PhD dissertation, Australian national university1998.
Knight, John and Lina Song (2003) Increasing urban wage inequality in China Extent, elements and evaluation. Economics of Transition
Volume 11 (4), 597–619
Kraay, A. (2005), "When is Growth Pro-Poor? Evidence from a Panel of Countries", Journal of Development Economics, forthcoming.
Lehmann, H., and J. Wadsworth. (2001) “Wage Arrears and the Distribution of Earnings in Russia.” Working Paper 421, WDI.
Li, Shi (2000) “Changes in Income Inequality in China Transition.” Mimeo, Chinese Academy of Social Sciences, 2000.
Lopez, Umberto (2005) Pro growth, pro poor: Is there a trade off?. World Bank Mimeo.
Lukianova, Anna (2005) Wage Differentiation and Phenomenon of Working Poor in Russia. The World Bank Mimeo.
Meng, Xin (2000) “Economic Restructing and Income Inequality in Urban China.” Research Paper, Australian National University, 2003.
Meng Xin, Robert Gregory and Youjuan Wang ( 2005) Poverty, inequality, and growth in urban China, 1986–2000, Journal of Comparative
Economics 33 (2005) 710–729
Milanovic, B. (1998). Income, Inequality, and Poverty during the Transition from Planned to Market Economy, World Bank: Washington
DC.
Milanovic, B. (1999a). ‘Explaining the Increase in Inequality during the Transition’, World Bank Policy Research Working Papers 1935,
World Bank: Washington DC.
Newell, A. (2001) “The Distribution of Wages in Transition Countries”, IZA DP No.267, March.
Newell, A., and M. Socha (2005) “The Distribution of Wages in Poland, 1992–2002.” Discussion Paper 1485, IZA, Bonn.
Popova, D. (2004) Dynamics of income inequality in the context of globalization. In : Russia’s accession to WTO: Imaginary and Real
Consequences by Ovcharova L et al. IISP Working Papers WP4/2004/01. М.oscow, Pomatur (In Russian)
Ravallion M. and Chen. S. (2004) China’s (Uneven) Progress Against Poverty. World Bank Policy Research Working Paper Series.
Ravallion, M. (2001). ‘Inequality Convergence’, World Bank Policy Research Working Papers 2645, World Bank: Washington DC.
Ravallion M. and Chen. S. (1996) Data In Transition: Assessing Rural Living Standards In Southern China China Economic Review Volume
7(L) 1996
Rozelle, Scott, and Johan F. M. Swinnen (2004) “Success and Failure of Reform: Insights from the Transition of Agriculture.” Journal of
Economic Literature 42 (June): 404–56.
Rutkowski, J. (1996a) ”High Skills Pay Off: The Changing Wage Structure During the Economic Transition in Poland”, Economics of
Transition, Vol.4: 89-112.
Rutkowski, J. (1998) “Welfare and the Labor Market in Poland”, World Bank Technical Paper No. 417.
Rutkowski, Jan. (2001) “Earnings Inequality in Transition Economies of Central Europe: Trends and Patterns During the 1990s,” SP
Discussion Paper No. 0117 (Washington: The World Bank).
Shorroks, Anthony (1980) “The Class of Additively Decomposable Inequality Measures”. Econometrica, Vol.48, No.3, pp. 613-25.
Shorrocks, Anthony, (1982), “Inequality decomposition by factor components”, Econometrica, 50, 1, pp193-211
Svejinar J. (2002) Transition Economies: Performance and Challenges. Journal of Economic Perspectives, Vol 16, No1, pp. 2-28
Szulc, Adam (2003) Inequality During The Economic Transition In Poland: 1993 -1999 Evidence. Warsaw School of Economics, Institute of
Statistics and Demography. Mimeo
58
Tóth, István György (2003) Assessing the effects of economic transition on income distribution: the case of Hungary, 1987-2001. Paper
presented at Fifth Annual Global Development Conference “Understanding reform” New Delhi, India January, 28 2004
Večerník, J. (1995) “Changing Earnings distribution in the Czech Republic: Survey Evidence from 1988-1994”, Economics of Transition,
Vol.3: 355-371
World Bank (2005a) WDR World Development Report 2006, Equity and Development.
World Bank (2005b) ECAPOV II. Growth, Poverty and Inequality in Eastern Europe and Former Soviet Union.
World Bank. (2005c) Jobs Report “Labor Markets in Europe and Central Asia: Toward Creating More and Better Jobs.” Washington, DC:
The World Bank, processed.
World Bank (2005d) Russian Federation: Reducing Poverty through Growth and Social Policy Reform. Washington, DC: The World Bank.
World Bank (2005e) “Pro-Poor Growth in the 1990s: Lessons and Insights from 14 Countries” Washington, DC: World Bank.
World Bank (2004) Poland Growth, Employment and Living Standards in Pre-Accession Poland. Washington, DC: The World Bank
World Bank (2003a) Kyrgyz Republic Enhancing Pro-poor Growth. Washington, DC: The World Bank
World Bank (2003b) Romania Poverty Assessment. Washington, DC: The World Bank
World Bank (2002a) Armenia Poverty Update. Washington, DC: The World Bank
World Bank (2002b) Georgia Poverty Update. Washington, DC: The World Bank
World Bank (2002c) Hungary Long-Term Poverty, Social Protection, and the Labor Market. Washington, DC: The World Bank.
World Bank ( 2002d) Slovak Republic: Living Standards, Employment, and Labor Market Study. Washington, DC: The World Bank
World Bank (2002e) Transition: The First Ten Years. Analysis and Lessons for Eastern Europe and the Former Soviet Union. Washington,
DC: World Bank.
World Bank (2000) ECAPOV 1. Making Transition Work for Everyone: Poverty and Inequality in Europe and Central Asia, World Bank:
Washington DC.
World Bank (1999) "Georgia. Poverty and Income Distribution". Volumes I and II. Poverty Reduction and Economic Management Unit,
ECA Region (May). Washington, D.C.
Wu, Ximing and Jeffrey M. Perloff (2004) China’s Income Distribution over Time: Reasons for Rising Inequality. CUDARE Working
Papers (University of California, Berkeley) paper 977.
Yemtsov R. (1999) Georgia: Income Distribution. IZA Discussion Papers.
Yemtsov, R. (2003), “Quo Vadis Inequality and Poverty Dynamics Across Russian Regions, ” UN WIDER Discussion Paper No. 2003/67.
Zhu, N and X. Luo (2006) Nonfarm Activity and Rural Income Inequality: A Case Study of Two Provinces in China. The Worlld Bank Policy
Research Working Paper
59
Annex 1. Is there such things as “the” inequality index for a country?
Atkinson (2003) argues that the answer to that question is typically ‘no”. Figure A1.1 below provides
evidence for UK, for instance.
Figure A1.1 Gini index for income per equivalent adult in United Kingdom
Source: Atkinson (2003)
Transition countries display a particularly large range of variation depending on the source, definitions use
and degree of adjustments. Comparing Tables 1 and 5 in the text one can ascertain that the situation is the
same in ECA. Table A1.1 gives an extended range of Gini indices for per capita incomes and expenditures`
Table A1.1. Russia: Gini indices per capita from various sources.
1992
Money incomes: official series
0.29
1995
0.38
1996
0.40
1997
0.38
1998
0.40
1999
0.40
2000
0.39
2001
0.40
2002
0.40
2003
0.40
HBS (based on survey data)
Money income, nominal
0.47
0.46
0.46
0.45
0.46
Money income, COL adjusted*
0.45
0.43
0.42
0.41
0.42
0.45
0.40
Cash Expenditures, nominal
0.45
0.46
0.45
0.44
0.45
0.45
Cash Expenditures, COL adjusted *
0.43
0.42
0.41
0.41
0.41
0.41
Consumption, nominal
0.37
0.39
0.37
0.36
0.37
0.37
Consumption, COL adjusted *
0.35
0.36
0.34
0.33
0.33
0.33
RLMS
Total incomes, nominal**
0.49
Total incomes, COL adjusted **
0.47
Total expenditures, nominal
0.45
Total expenditures, COL adjusted *
0.39
0.48
0.45
0.56
0.51
0.47
0.42
0.42
0.42
0.51
0.45
0.41
0.4
0.4
0.41
0.51
0.48
0.46
0.45
0.44
0.45
0.45
0.47
0.45
0.44
0.42
Consumption, nominal
0.48
0.42
0.4
0.39
0.39
0.40
Consumption, COL adjusted *
0.41
0.39
0.38
0.37
0.39
NOBUS
Money incomes, nominal
0.44
Money incomes, COL adjusted *
0.43
Consumption , nominal
0.36
Consumption, COL adjusted *
0.34
Total incomes with imputed rent, COL adjusted,*,**
* (deflated using subsistence minimum) ** with in-kind consumption of farm products, no imputed rent Source; Russia PA with authors’ additions
based on direct data estimates for NOBUS and published data for RLMS
0.34
60
The Table shows that for the most recent period for different definitions of welfare and different data
sources one could obtain almost any figure in a large range between 0.34 and 0.45, - a difference that would
put Russia in different “clubs”. Importantly better defined and more accurate measures from larger surveys
point to the lower part of the range. What is also important is that dynamics once assessed with different
series exhibit different paths.
Figure A1.2. illustrates three major sources on inequality dynamics that are available over time. They
depict very different dynamics: Figure A1.2. demonstrates differences in dynamics. In Russia the variation
of estimates is extremely large.
Figure A1.2. Russia: Dynamics of Gini index for per capita incomes and
expenditures from various sources, 1992-2004
0.5
Officially published,
nominal per capita
incomes
0.45
0.4
Per capita
expenditures in
RLMS, direct data est.
0.35
HBS/NOBUS* per
capita consumption
corrected for regional
price differences
0.3
0.25
1992
1994
1996
1998
2000
2002
2004
.
Sources: Goskomstat, Poverty Assessment (World Bank )
Flemming and Mickelwright (1999) provide an evidence that for many countries, even with the most
developed statistics, inequality indices differ even for the same year. Figure A1.3 illustrates the impact of
using different sources on measured inequality in incomes in Hungary
Figure A1.3 Hungary Gini index for per capita incomes from HBS and TARKI survey
Source: TARKI panel results from Galasi (1998, Table 1) HBS results from Toth.
61
Poland is also no exception for differences in inequality estimates even using the same data. Data in the
Table A1.2 report data from just one source, HBS and rely on consumption per capita. Seemingly slight
modifications of the definition of consumption produce changes which are significant by all possible
standards.
Table A 1. 2 Consumption inequality(Gini***) in Poland 1994 - 2001: an overview
1994
1995
1996
1997
1998*
1999
2000
2001
CSO
0.318
0.312
0.326
0.336
0.334
0.336
0.340
0.341
CSO reg**
0.313
0.307
0.321
0.330
0.327
0.329
0.333
0.333
WB0**
0.279
0.275
0.278
0.284
0.279
0.287
0.290
0.290
Notes:
*Some corrections of expenditure concepts have been introduced by GUS in 1998 (see: Published HBS 1998, Methodological Notes, p. XXV).
**
Updated regional deflators for 1994-2000; first version of deflators for 2001.
***
Gini coefficients are for per capita aggregates, weighted with persons (no sample weights have been used).
****
Consumption in constant prices (deflator: annual CPI), including imputed rents and flow of service form durables.
Source: Poland PA, World Bank (2002)
Finally Table A1.3 shows a significant spread of figures for Gini for per capita incomes obtained by authors
of various studies. Even when based on the same years and the same data, the extent of adjustment and the
ambiguity of price indices creates noticeable differences ion the exact measure of Gini. But trends are all
similar and point to the same direction.
Table A 1.3. China: Comparison of Gini coefficients for per capita incomes from various studies
Rural
Urban
Data
1988
1995
2001-2
1988
1995
2001-2
Ravallion and Chen (2004)
SSB
0.297
0.334
0.365
0.211
0.283
0.323
Wu and Perloff (2004)
SSB
0.300
0.338
0.343
0.201
0.221
0.269
Li (2000)
SSB
0.301
0.323
0.230
0.280
Khan and Riskin (1998, 2004)
CASS
0.338
0.416
0.233
0.332
Gustaffson and Li (1999)
CASS
0.228
0.276
0.234
0.282
Wagstaff (2005)
Meng (2003)
CHNS
CASS
0.395*
0.375
0.318
0.419*
Note : SSB –State statisitical bureau based on household budget survey, CASS- Economics Institute of the Chinese Academy of Social Sciences
Survey * All China, 1989 to 1997.
62
Annex 2. Review of the Literature of Wage Inequality in ECA
Table A. 2 .1 Hourly Wages: Gini Coefficient from ISSP survey
1986 1991 1992 1993 1994 1995 1996 1997
Bulgaria
Czech
Republic
Hungary
Latvia
Poland
Russia
Slovak
Slovenia
0.26
0.27
0.28
0.27
0.26
0.31
0.24
0.33
0.27
0.25
0.29
0.37
0.30
0.40
0.30
0.34
0.34
0.30
0.29
1998
0.31
0.26
0.30
0.34
0.27
0.49
0.27
0.31
0.25
0.29
0.32
0.27
0.42
0.31
0.27
0.30
0.25
0.39
0.30
0.25
0.30
0.34
0.28
0.23
0.28
1999
2000
0.31
0.28
0.24
0.31
0.41
0.26
0.42
0.24
0.29
0.43
2001
2002
0.32
0.23
0.31
0.29
0.29
0.43
0.26
0.24
0.29
0.32
0.29
0.43
0.25
0.31
Source: ISSP data and Bank staff calculations.
Note: The calculations are based on hourly wages (levels, not logs). The sample of workers includes men and women age 18 to 65 and excludes the
self-employed.
Table A 2.2: The Returns to Education in Russia from various studies
Brainerd
Brainerd
Brainerd
S&G
S&G
1991
1993
1994
1985
1990
Years of education
0.031*
0.066*
0.067* 0.028* 0.039*
Years of education (with controls)
Dataset
vTsIOM
vTsIOM
vTsIOM RLMS RLMS
S&G
1996
0.081*
0.079*
RLMS
S&G
1998
0.091*
0.091*
RLMS
S&G
2000
0.093*
0.094*
RLMS
S&G
2002
0.092*
0.097*
RLMS
Sources: Brainerd, E. “Winners and Losers in Russia’s Economic Transition”. American Economic Review, Vol 88(5), 1998, pp.1094-1116. (S& G)
Sabrianova, K; Gorodnichenko, Y. “ Returns to Education in Russia and Ukraine: A Semi-Parametric Approach…
Table A2.3
Source : Fleishner et al.
63
Annex 3. Inequality decompositions by groups: Available evidence from existing sources
Figure A3.1 Relative importance of between-groups inequality over time: Russia, Poland and Hungary
Russia
X 2001
IX 1992
10.2
Region
8.7
Employment
6.6
Age
2.6
Gender
1.9
Education
Family Size
Dependency
Employment
6.2
Education
6.2
Age
6.1
Location (U/R)
5.8
6.3
Location (U/R)
Dependency
1.3
Size
1.1
Gender
0.9
1.3
0.9
15.2
Region
Hungary
2003
1992
17.9
Education
15.1
Employment
Age
3.1
Dependency
Dependency
2.9
Age
Gender
14.0
Employment
12.0
Location (U/R)
6.0
Location (U/R)
27.0
Education
1.9
Gender
5.0
3.0
1.0
Poland
1994
2002
13.0
Education
7.9
Region
4.0
Employment
Age
1.7
19.1
Education
12.2
Region
6.4
Employment
Age
1.7
Decompositions is Hungary using Mean Log Deviation index and income based on CSO and TASKI data, Poland –consumption per equivalent
adult and HBS data, Russia – disposable resources per equivalent adult and RLMS data.
Sources: Toth, Sultz, World Bank (2002) – Poland PA, Popova
64
Annex 4 Russia, Hungary and China: Income-based decompositions by income sources
Table A 4.1 Russia: Contribution of Various Income Sources to Total Inequality measured by per capita incomes
(without imputed rent), 2003
Sources:
Share of incomes, % Concentration coefficient, C Contribution to inequality,
Gini index for per capita incomes
0.432
100%
100%
Work income
66%
0.59
90%
Of which “enterprenuerial”
1%
0.64
1%
Old age pension
24%
0.02
1%
Social transfers
3%
0.20
1%
Income from farm
5%
0.42
4%
Other income
3%
0.47
3%
Source: NoBUS data by Ovcharova and Popova.
Table A.4.2 Russia: Contribution of Various Income Sources to Total Inequality measured by per capita incomes
(with imputed rent), 2003
Sources:
Share of incomes, % Concentration coefficient, C Contribution to inequality,
Gini index for per capita incomes
0.336
100%
100%
Work income
50%
0.42
62%
Of which “enterprenuerial”
1%
0.64
1%
Pensions and other public transfers
17%
0.18
9%
Private transfers
4%
0.26
3%
Income from farm
6%
0.04
1%
Imputed rent and other income
23%
0.36
25%
Source: NoBUS data by Ovcharova and Popova.
Table A4.3 Russia: Contribution of Various Income Sources to Total Inequality as measured by the expenditures
(“diposable resources”), 2003
Sources:
Share, %
Concentration coefficient Contribution to inequality
Gini per capita dispos. Res.
0.338
100%
100%
Work income
39%
0.35
40.4%
Of which “enterprenuerial”
1%
0.57
1.7%
Old age pension
22%
0.12
7.8%
Social transfers
7%
3.9%
Income from farm
6%
0.30
5.3%
Other income
3%
0.47
4.2%
Gap expenditures and income
23%
0.54
36.7%
Source: NoBUS data by Ovcharova and Popova.
Table A4.4. Decomposition of the change in the Gini coefficient for Taganrog from 1988 to 2000 by income components
1988
2000
Income source
Share of
Concentra Contributi Share of
Concentr Contributi Change
income
tion
on to Gini income
ation
on to Gini 1988-2000
Earnings income
82.8
0.285
107
53.2
0.457
70
+0.008
Public transfers
15.7
-0.095
-7
23.7
0.055
4
+0.028
Allowances
1.3
-0.154
-1
8.2
0.195
5
+0.019
Subsidiary incomes
0.2
0.990
1
4.6
0.608
8
+0.027
Private transfers
10.2
0.440
13
+0.045
Gini
0.220
0.346
+0.126
Note: Disposable equivalent household income is computed using the equivalence scale (1 0.5 0.7). s –share of income, C- concentration coefficient.
Source: Gustaffson and Novorizhkina.
65
Table A 4.5
Note using median log deviation as inequality measure. Total income inequality =1, according to the measures employed, I first fell from 0.207 in
1987 to 0.195 in 1991 – a statistically significant but negligible fall, and then increased monotonically to 0.242, which is both significant and abovethershold change.
Source: Kattuman and Redmont (2001)
Table A 4.6. Rural China: Decompositions by factors
Source: Ravallion and Chen (1998)
66
Annex 5 Inequality measures used in the paper: short technical summary
There are many ways to measure inequality because inequality has many aspects, and it is hard to present a
complex phenomenon of inequality in a single number. The literature relies mainly on three types of
inequality measures: (i) quantile (such as decile) ratios and other similar measures; (ii) Gini coefficients;
and (iii) Theil inequality measures and other entropy-indices.
Quantile ratios are straightforward indicators of inequality that are easy to interpret. The most common is
the decile ratio is the 90/10 ratio, which is the equivalent consumption at the 90th percentile of the
distribution divided by the equivalent income at the 10th percentile. Quantile ratios are insensitive to
outliers either in the very top or very bottom tail of the distribution. They do not reflect what happens in
other parts of the distribution.
The Gini coefficient is given by:
G=
2
μ n2
n
⎛
∑ ⎜⎝ r
i
i =1
−
n + 1⎞
⎟c i ,
2 ⎠
where there are n individuals indexed by i, their consumption or income is given by ci, mean consumption
or income is denoted by μ, and where ri is household’s i rank in the consumption or income ranking (i.e. for
the household with lowest consumption ri equals 1 while for the household with the highest consumption ri
equals n). The Gini coefficient is bounded between 0 and 1, with 0 indicating absolute equality and 1
indicating absolute inequality. The Gini coefficient is especially sensitive to changes in inequality in the
middle of the equivalent consumption distribution.
Another widely used class of inequality indicators is the generalized entropy class developed by Theil.
Within that class, most commonly used is the Theil entropy index,
E (1) =
1 n ci ⎛ c i ⎞
∑ ln⎜ ⎟
n i =1 μ ⎜⎝ μ ⎟⎠
or Theil mean log deviation index,
E (0) =
1 n ⎛μ⎞
⎛1 n ⎞ 1 n
ln⎜⎜ ⎟⎟ = ln⎜ ∑ ci ⎟ − ∑ ln(ci )
∑
n i =1 ⎝ ci ⎠
⎝ n i =1 ⎠ n i =1
Both measures are zero for perfect equality. For complete inequality (one person consumes everything),
E(0) goes to infinity while E(1) reaches nln(n). The two Theil inequality measures differ in their sensitivity
to inequality in different parts of the distribution. The entropy measure, E(1), is most sensitive to inequality
in the top range in the distribution while the mean log deviation measure, E(0), is most sensitive to
inequality in the bottom range of the distribution.
Inequality can be decomposed along two dimensions. One can decompose total inequality into the
contribution of each component of income (labor income, self-employment income, state transfers and so
on). This decomposition can be performed using the Gini index. The second way of decomposing
inequality is to decompose it into inequality within population subgroups and between subgroups. This
decomposition can be performed using the Theil indices.
67
Following A. Shorrocks, (1982) the contribution of each income source is the product of a concentration
coefficient for that income source and the fraction of that income source in total income. More formally,
Gk* , the concentration coefficient for income component k is given by:
Gk* =
2
μ n2
⎛
n
∑ ⎜⎝ r
i
i =1
−
n + 1⎞
⎟ y k ,i
2 ⎠
where yk,i is component k of the income of individual i, mean total income is denoted by μ, and ri is
household’s i rank in the ranking of total income. The Gini component is a weighted sum of the
concentration coefficients:
μk * K
G=∑
Gk = ∑ S k Gk*
k =1
k =1 μ
K
where S k = μ k / μ is the share of the component k in total income. The percentage contribution of income
source k to total income equality is found as:
Pk = S k
Gk*
× 100%
G
The expression above gives the overall contribution of income source k to inequality.
Decomposition by population groups allows us to look more closely at the causes of inequality. The Gini
coefficient does not lend itself well to a decomposition by population groups. For that purpose, the Theil
generalized entropy class of inequality measures is used. Following F. Bourguignon (1979) and A.
Shorrocks (1980), we decompose total inequality into a component that is due to inequality across
population subgroups, and into a component that is due to inequality within these subgroups. This
decomposition can be performed for various population groupings. In this paper we rely on Theil entropy
measure.
Let the population be divided into m mutually exclusive and exhaustive subgroups. Let the population
share of the jth group in the population be given by wj , and the consumption share by vj. For the Theil
entropy measure the decomposition is :
m
m
⎡ v j ⎛ v j ⎞⎤
E (1) = ∑ v j E (1) j + ∑ w j ⎢ ln⎜ ⎟⎥,
⎜ ⎟
j =1
j =1
⎣⎢ w j ⎝ w j ⎠⎦⎥
where E(1)j is the Theil entropy measure calculated for all individuals in subgroup j. The first summation is
a weighted average (using consumption shares as weights) of the entropy measures calculated for the
subgroups. Hence, this first term gives the component of overall inequality that is due to inequality within
subgroups. The second summation is the entropy measure calculated on mean consumption of each
subgroup (and weighting each subgroup by its population share). Hence, this second term gives the
component of inequality that is due to between-group differences.
68
Annex 6. Data: Comparable consumption figures for ECA countries.
To arrive at the internationally comparable assessment of ineuqality this paper uses ECA regional data
archive. The archive contains primary unit record data from recent household surveys for 24 countries in
ECA spanning the period 1998-2004. This data were used to construct a comparable indicator of living
standards across all countries for the recent World Bank report “Growth, poverty and inequality in Eastern
Europe and the Former Socviet Union: 1998-2003” and for the WDR 2006. The choice of consumption
rather than income to measure inequality was dictated by practical considerations. Income data remain
particularly difficult to collect in transition countries. In contrast, practice has shown that consumption data
can be gathered with a great degree of precision. Survey consumption modules have become more detailed
over time, and better capture various dimensions of consumption including informal payments etc.
In relying on consumption of goods and services by a household as the measure of living standards, there
were a number of conceptual and practical issues that needed to be addressed. First, unlike food, consumer
durables and housing are consumed over a long period of time. It is customary, therefore, to include the
imputed value of the consumption flow associated with the possession of consumer durables (including
housing) but exclude the expenditure on the purchase of the these goods. However, for the ECA region,
data availability limits the application of this approach to all countries. We did not, therefore, include
estimates of flow of services of durables, nor have we added in durable purchases or rents. Second, when
consumption is used as a measure of well-being, higher consumption should indicate a higher level of wellbeing. For most consumption items this correspondence is reasonable. However, for some categories, such
as health expenditures, this correspondence is questionable. As a result health expenditures were not
included as a part of consumption (discussion in Deaton and Zaidi 2002). Third, given the significance of
spatial differences, we adjusted for spatial price differences using survey-data-based Paasche price indices
using the same set of information in all countries. In the case where data were collected over a long period
of time, it was also necessary to adjust for changes in prices over time. Quarterly CPI (IMF) indices were
used to compute real values. Fourth, households in ECA cope with poverty by relying on an array of nonmarket strategies including producing their own food and engaging in reciprocal exchange with other
households and institutions. We used a consistent approach in assigning a monetary value to these
components of consumption. Fifth, to adjust for differences in household composition we took the simplest
approach and used the per capita scale. Sixth, the same procedure, which conforms to methods used in other
international household survey data depositories such as the Luxemburg Income Study, was used to clean
the data of outliers across all data sets. As we have followed a consistent approach across all data sets, we
are reasonably confident that differences across countries in the final consumption measure are due to
differences in the primary data and are not owed to the method of aggregation.
The constructed estimate of real per capita consumption has several shortcomings which reflect some
persistent data problems in ECA. First, it ignores the differential impact of price increases on the poor and
non-poor. No price indices for low income groups are routinely available in ECA which would allow us to
address this issue. Second, over time, there has been considerable deterioration in response rates in many
countries. Countries deal with this problem in different ways which may have (as yet unknown)
implications survey based poverty and inequality measures.
For other countries the Figure 2 resorts to the available per capita consumption, which may not exactly the
degree of comparability (with the exception of Columbia and Vietnam where exactly the same methodology
was applied and Mexico and United States, for which identical definitions were used to pick up indicators
from published data).
69