Economics of TransitionVolume 7 (2) 1999, 299–341

© The European Bank for Reconstruction and Development, 1999.Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA.

Explaining the increase ininequality during transition 1

Branko MilanovicPolicy Research Department, The World Bank, 1818 H Street NW, Washington D.C., 20433USA, E-mail: bmilanovic@worldbank.org

Abstract

This paper attempts to explain the increase in inequality that has been observed inall transition economies by constructing a simple model of change in compositionof employment during the transition. The change consists of the ‘hollowing-out’of the state-sector middle class as it moves into either the ‘rich’ private sector orthe ‘poor’ unemployed sector. The predictions of the model are contrasted withthe empirical evidence from annual household income surveys from six transitioneconomies (Bulgaria, Hungary, Latvia, Poland, Russia and Slovenia) over theperiod 1987–95. We find that the most important factor driving overall inequalityupwards was increased inequality of wage distribution. The non-wage privatesector contributed strongly to inequality only in Latvia and Russia. Pensions,paradoxically, also pushed inequality up in Central Europe, while non-pensionsocial transfers were too small everywhere and too poorly focussed to make muchdifference.

JEL classification: D31, D63, H42, P21.Keywords: transition, inequality, wage distribution.

1 The findings, interpretations, and conclusions expressed in this paper are entirely those of the author.They do not necessarily represent the views of the World Bank, its Executive Directors, or the countriesthey represent. I am grateful to Jan Rutkowski, Simon Commander, Ehtisham Ahmad and participants atthe IMF Conference ‘A Decade of Transition: Achievements and Challenges’ held in Washington D.C.,February 1–3, 1999 for very helpful comments.

MILANOVIC300

1. Introduction

To explain the change in inequality that has occurred in all transition economies,2we first need to present the economic changes in terms of what happened to theincomes of different social classes, and then to translate this into the ‘language’ ofpersonal incomes. This is, of course, a common problem encountered whenmoving from social classes and thus factoral income distribution to personalincome distribution (for a recent overview see Atkinson, 1995). The ‘translation’ iscumbersome but necessary if we want to combine (i) a discussion of thoseeconomic forces that shape income distribution with (ii) a look at how these forces‘played themselves out’ in the arena of personal income distribution. To look into(i) we have to use the standard economist’s ‘toolkit’ which is designed to studythe behaviour of individuals who are either workers or capitalists or farmers ortransfer-recipients. In doing this we ‘tag’ individuals by assuming that they haveonly one source of income. This is the approach I use in Sections 2 and 3 where Ipresent first, a simple model of factoral income distribution during the transitionand second, a numerical illustration of change in inequality.

‘Tagging’ is, of course, wrong: a capital-owner may (and presently often does)work; a workman may own stocks and engage in entrepreneurial activity; apensioner may work part-time or lease some of his assets. Thus when we move topersonal income distribution, the point of view changes: we no longer ‘tag’individuals but simply add all their sources of income, adjust total income forpersonal and household characteristics (household size, age of children etc.), andthen rank everybody according to his or her adjusted household income. Peopleare no longer workers, capitalists or pensioners; they just have incomes, some ofwhich may be in the form of wages, profit or pensions. In terms of disaggregationof an index of inequality we go from the disaggregation by recipients todisaggregation by income sources. The latter is the point of view I adopt inSection 4, where I present empirical evidence on income inequality and factorsthat were responsible for its increase. In the concluding Section 5, I contrast theconclusions drawn from the empirical evidence with the conclusions drawn fromthe model and numerical illustrations in Sections 2 and 3.

2. Change in sector shares and factor incomes

Consider the following simple model of an economy before and after transition.

2.1 Pre-transitionThe economy is composed of two sectors. First, a small private sector whose sizeis limited through various legal restrictions (e.g., a limit on the number of workers

2 See Milanovic (1998, Chapter 4).

THE GROWTH OF INEQUALITY 301

that can be employed; no access to credit; banned sectors of operation, etc.); andsecond, a large state sector. The private sector (consisting mostly of the self-employed) maximizes the average product per worker ( papl )3 where subscript pdenotes the private sector. It employs OP workers (Figure 1). Since its growth islimited by legal barriers, it is not allowed to maximize the papl curve. Theaverage income in the private sector is therefore y.

Since there is no unemployment, the state sector must employ the rest of thelabour force (NP)4 and its wage ( sw ) is established at the point where the demandfor labour ( smpl ) intersects the vertical P line.5 Normally, we would expect thatsince yws < , labour would flow to the private sector. This is precluded becauseof the legal limits on the private sector size.

Figure 1. Private and state-sector equilibrium before transition

There is no taxation of the private sector. The entire surplus of the state sector(the area between the smpl line and sw ) is taxed and is used to pay pensions (and

3 All per capita magnitudes are written in lower case.4 The state-sector share should be read from right to left.5 A more usual assumption (see e.g., Blanchard and Keeling, 1996; Commander, Tolstopiatenko andYemtsov, 1997, Appendix 1) has been to assume that the private sector equalizes the marginal product oflabour and the wage, while the state sector maximizes the average product of labour. This was rationalizedby arguing that the private sector was a ‘normal’ private sector, and the state sector was labour-managed.However, this approach fails to acknowledge that almost the entire private sector before the transition wasself-employed or co-operative, and that average product was therefore a better maximand. Further, only inYugoslavia, and to some extent in Poland, could the state sector be described as labour-managed.

O N

y

P Labour

Wage

sw

papl smpl

MILANOVIC302

other social transfers). Writing the number of private and state-sector employeesrespectively as pN and sN , and the total number of pensioners as T (for transfer-

recipients), we can write the identity (1) where the LHS shows the output fromthe production side and the RHS its distribution. For simplicity, total tax revenues

ssNτ (tax per state-sector employee times number of state-sector employees) areassumed to pay only for total pensions (average pension or transfer t times T).

tTsNswpyNsNssNswpyN

taxes

++=++��

τ

Note that since sN and pN and the papl and smpl curves are known, the

entire LHS is known. Since T is known (by assumption), t can be directlyobtained. (The other way to look at it is to realize that total output of the statesector must be exhausted for wages and pensions. Once sw is determined, t thenbecomes known.)

2.2 TransitionThe transition is defined as the removal of legal restrictions on the private sector.6Moreover, the private sector is no longer a self-employment sector whichmaximizes the average product per worker but becomes a ‘normal’ private sectorwhere firms hire labour until pp wmpl = . The private-sector demand for labour

now becomes pmpl (Figure 2). State-sector demand for labour ( smpl ) shifts

downward as demand for state-sector output declines. If we assume wage rigidity(wages stay at the pre-transition level sw or do not decline sufficiently) therewould be some unemployment ( uN ).7 The employment share of the private sectorhas increased, and that of the state sector declined.

The entire surplus of the state sector is still taxed away to pay for pensions.However, this is no longer sufficient to maintain pensions unchanged in realterms both because the state sector has shrunk, and, in addition to (an unchanged)number of pensioners, the government must also pay unemployment benefits to

uN individuals. The private sector must now be taxed. Not all of its surplus (thearea between the pmpl curve and pw ) is taxed, but only a portion α. The rest of

the surplus is capitalists’ income.8

6 In a similar model, Blanchard and Keeling (1996) define the transition as a removal of subsidies receivedby the state sector. Under both scenarios, the relative sizes of the private and state-sector change in favourof the private.7 Wages in the two sectors must now be the same.8 We can further assume that capitalists invest at least a portion of their surplus thus shifting the demandfor labour curve upward and increasing the demand for labour. If state-sector capital depreciation is equalto the user cost of capital, there would be no net investments in the state sector, and its demand for labour

(1)

THE GROWTH OF INEQUALITY 303

Figure 2. Private and state-sector equilibrium after transition

Now, on the production side instead of two ‘classes’ (the self-employed andworkers in the state sector), we have three ‘classes’: workers in the private sector,capitalists, and workers in the state sector. This yields:

( )������

taxes

sssspp bNNwb1Nw ατα +++−+

where b = private-sector profit; αb is taxed away from capitalists and (1 – α)b theykeep. The last two terms in (2) represent total taxes that are assumed to pay forpensions and unemployment benefits. For simplicity, the average amount ofunemployment benefit is assumed equal to the average pension.

On the distribution side, this yields:

( )usskkpp NTtNwNyNw ++++

where ky = the average net income of capitalists, kN = the number of capitalists,and all other symbols as before. If we compare equation (3) and the RHS ofequation (1), we note that the distribution side has changed during the transitionin three ways: (i) the self-employment sector has now split into private-sectorworkers and capitalists; (ii) pN has increased while sN has shrunk (compare

would remain unchanged. Thus dynamically, the proportions between the private and the state sector willcontinue to shift in favour of the former. For a similar mechanism see Blanchard and Keeling (1996).

O NLabour

Wage

(2)

(3)

swpwuN

)old(mpls

)new(mpls

pmpl

pN sN

MILANOVIC304

Figures 1 and 2), leading to an increase in private-sector income, and a decrease instate-sector income; (iii) the unemployed have joined the pensioners as transfer-claimants.

3. Changes in inequality

To study the overall change in inequality, it is not sufficient to consider only whathappened to factor shares. This is because the factor shares are distributed withdifferent degrees of inequality.

3.1 Before the transitionBefore transition, state-sector wages were more equally distributed (among therecipients of state-sector wages) than was the self-employment income (amongthe recipients of the latter). In other words, the Gini coefficient of Ws was lowerthan the Gini of Y (upper case letters indicate total amounts of wages, or private-sector income).9

Take the RHS of equation (1), depicting the distribution of pre-transitionincomes, and assume (ultra restrictively) that state-sector wages ( sw ) andpensions (t) are distributed equally per capita (and that we study per capita incomeinequality). Then, the only source of inequality may seem to come from inequalityin the distribution of Y. However, this is not the case because there would also besome inequality associated with the differences in average incomes of workers inthe state sector, pensioners and the self-employed. The overall income Gini (GINI)can be written:

( ) ( ) ( )[ ] Lpptwpptyppwy1

pGpGpGGINI

twstywys

tttwwwyyy

+−+−+−

++= +

µ

πππ

where µ = average overall income, iG ’s are the Gini coefficients across recipientsonly of self-employment income ( yG ), state-sector wages ( wG ) and pensions ( tG );

π’s are income shares of the three types of income; p’s are the population shares.The first three terms on the RHS of (4) represent that part of total inequality whichis due to inequalities within groups; the next term accounts for inequality causedby the differences in groups’ mean incomes. Finally, L – the overlapping term –shows what part of total inequality is due to the fact that some people belongingto a group with a higher than average income (say, the self-employed) have lower

9 The entire private (self-employment) income before the transition is denoted by Y. After the transition,we still denote the entire private-sector income by Y; however, it is now equal to Wp (private-sector wages)plus B (capitalists’ profit).

(4)

THE GROWTH OF INEQUALITY 305

incomes than some people belonging to a group with a lower than averageincome (say, pensioners).

Assuming (ultra-restrictively) that wG = tG = 0, and swy β= and swt γ=where β > 1 (see Figure 1), and γ < 1 (since the average pension is a fraction of theaverage wage), the overall inequality before the transition becomes:

( ) ( ) ( )[ ] Lpp1pppp1w1

pGGINI twtywysyyy +−+−+−+= γγββµ

π .

The typical population shares before the transition were approximately 60 percent of household heads employed in the state sector, 20 per cent self-employedor employed in the private sector, 20 per cent pensioners. Assuming further thatthe average pension was one-half of the average wage (γ = 0.5) and the averageself-employment income was one and a half times more than the average wage(β = 1.5),10 GINI becomes:

GINI = 0.06 Gy + 0.06 + 0.04 + 0.06 + L ≅ 0.18 + L

where we have assumed, based on some empirical evidence, that yG = 0.3.

L (overlapping) inequality can only stem from the self-employed who might‘overlap’ with other social groups. This is so because all pensioners and all state-sector workers have, respectively, swγ or sw : there can be no ‘overlapping’between them. Moreover, there is some evidence that the self-employed not only‘overlapped’ with some other recipients, but even bracketed them. The self-employed before the transition included both poor farmers, barely above thesubsistence level, and rich entrepreneurs who, thanks to connections and luckwere able to make large profits.11 These two extreme groups among the self-employed were particularly in evidence in countries that combined a large privateagricultural sector divided into many small plots like Yugoslavia, Poland andHungary, and also allowed some non-agricultural private enterprise.

Equation (5) shows that under the ultra-restrictive assumptions of equal wagesand equal pensions (for all state-sector workers and all pension recipients,respectively), the overall Gini in pre-transition countries would have beenbetween 18 and 20, where this extra ‘nudge’ above GINI = 18 comes from thepresence of both the relatively poor and the relatively rich among the self-employed. A Gini of approximately 20 is the value recorded in Czechoslovakia,the most egalitarian of the former socialist countries. Supporting indirectly theultra-restrictive assumption of equal wages and pensions, it was argued withsome plausibility by Vegernik (1986; 1993) that, before the transition, disposable

10 So that the overall average income µ (0.6*1+0.2*0.5+0.2*1.5) is equal to the average wage.11 The latter group even produced a few instances of ‘socialist’ millionaires like Jan Kulczyk, the founder ofInter Fragrances, a large Polish perfume exporter, or Erno Rubik of the Rubik cube fame in Hungary. Forthe evidence of a U-shaped share of the private-sector income in total income, see Milanovic (1992, p. 21).

(5)

MILANOVIC306

household income in Czechoslovakia could have been approximated well bymultiplying individual household demographic characteristics – number of theemployed, number of pensioners and number of children – with constants (whichare, by definition, equal for everybody).12 Implicitly, inequality of wage orpension distribution could be treated as negligible.

If we now abandon the ultra-restrictive assumptions of wG = tG = 0, andreplace wG = 0 with the empirically-based wG = 0.25;13 and similarly tG = 0 withthe empirically-based tG = 0.1514 (while still ruling out overlapping betweenworkers and pensioners: that is, still assuming that all workers have higherincomes than all pensioners), the GINI increases to about 27. With the overlapcomponent, it would be about 28. Most of the increase is due to inequality amongwage earners (see Table 1). Indeed the Gini coefficients between 24 and 28 wereobserved in Poland, Romania, Slovenia, Hungary, Bulgaria and most of therepublics of the former Soviet Union before the transition.15

Table 1. Breakdown of inequality before transition

Formula Value(in Gini points)

Inequality amongstate workers

9.0

Inequality amongthe self-employed

1.8

Wit

hin

-gro

up

Inequality amongpensioners

0.3

Differences inaverage incomes

≅ 16

Total withoutoverlapping

≅ 27

Overlapping 1–2 points

Total GINI about 28

Assumptions: wp = 60%, yp = 20%, tp = 20%; Ws = 1; pension/wage = 0.5; self-employmentincome/wage = 1.5; wG = 0.25; yG = 0.3; Gt = 0.15. Hence, wπ = 60%, yπ = 30%, tπ = 10%, µ = 1.

12 Household income, approximated by the formula c0 + c1*number of employed members + c2*number ofpensioners + c3*number of children, could explain more than 60 per cent of actual variation inCzechoslovak incomes (see Vegernik, 1993).13 Atkinson and Micklewright (1992, p. 81) give the earnings’ Ginis in 1986–87 as 20 in Czechoslovakia, 22in Hungary, 24 in Poland, 28 in the USSR. Similarly, Redor (1992 [1988]) gives the following values:Hungary (1980) 21; Poland (1980) 23; Czechoslovakia (1979) 20; USSR (1964) 24.14 Ginis of pensions (calculated across pension recipients) ranged in East European countries between 12and 18 (World Bank Development Research Group database).15 See Atkinson and Micklewright (1992, p. 113); also Milanovic (1998, p. 44).

( ) ( ) ( )[ ]twstywys pptwpptyppwy1 −+−+−µ

www pG π

ttt pG π

yyy pG π

THE GROWTH OF INEQUALITY 307

In conclusion, assuming ‘ideal-typical’ conditions of socialist economies beforethe transition and breaking down equation (4) we find that:(i) about 11 Gini points were due to within-group inequality where inequality

among state-sector workers was by far the most important (Table 1);(ii) about 16 Gini points of inequality were due to the differences in average

incomes between the self-employed, workers in the state sector andpensioners; and

(iii) the rest (1–2 Gini points) was due to ‘overlapping.’

3.2 TransitionFor simplicity, we combine the ‘new class’ of capitalists that has emerged duringthe transition with private-sector workers denoting both with the subscript y(equation 5), while subscript t now covers the pensioners and the unemployed.

( ) ( ) ( )[ ] Lpp1pppp1w

pGpGpGGINI

twtywys

tttwwwyyy

+−+−+−

+++=

γγββµ

πππ

What will happen to inequality during the transition? As a quick glance atequation (5) shows, there are three ways in which GINI could increase. First, theindividual Gini coefficients (of the state sector, private sector, or pensioners) canincrease while the population shares of different groups ( ip ’s) and their averageincomes stay constant. Second, the population shares might change with peoplejoining a more unequal (private) sector while individual sector Ginis and averageincomes stay the same. Third, the differences in average incomes might increase(e.g., the average income of the ‘rich’ private sector might go up while the averageincome of the ‘poor’ pensioners decreases) with individual Ginis and populationshares remaining constant. We can dismiss the third possibility in order toconcentrate on key changes during the transition as sketched in the model inSection 2: increase in individual-sector inequality, and change in the structure ofemployment. The effect of increased individual Ginis is straightforward. For thei-th type of recipient it is equal to:

iiii GpGGINI ∆π∆ = .

Note that a given change in the Gini coefficient among state-sector workerswill have a greater effect on total GINI than the same change in the other twosectors because the state sector weights are greater.16

16 The evidence (see Rutkowski (1996) for Poland; Vegernik (1994) for the Czech Republic; Vodopivec andOrazem (1995) for Slovenia; and Rutkowski (1995) for Bulgaria) shows not only that private-sector incomesbecome more unequally distributed during the transition (a thing which we might expect), but that state-sector wages’ inequality rises too.

(5)

MILANOVIC308

(6)

The effect on GINI of a transfer of workers from the state sector into theprivate sector will be more complicated. Rewriting GINI to express it explicitly asa function of wp , and setting the state-sector wage to 1 (numeraire), we obtain:

where we use yy pβπ = , ww p=π and wty pp1p −−= .17

When there is a movement of workers out of the state sector and into theprivate sector, differentiation of (6) with respect to wp yields:

( ) ( )( )[ ]

( )( )[ ] .dp

dLppp1

1pG2pG2

dp

dLp2p211

1pG2pp1G2

dp

dGINI

wwtywwyy

wwtwwwty

w

+−−−++−=

+−−−++−−−=

βµ

β

βµ

β

The sign of the within-group component of GINI (i.e., the sum of the first twoRHS terms in (7)) is ambiguous. Since the movement is to a more unequal and‘richer’ sector we know that yG > wG and β > 1. But offsetting this is the fact that

the sector which loses workers is larger than the sector which receives them, andthus yw pp > . It will therefore depend on specific numerical values whether the

within-inequality component is positive or negative. In our case, with yG = 0.3,

wG = 0.25, β = 1.5, wp = 0.6 and yp = 0.2, the component will be positive

(= 0.138), implying that the movement of labour out of the state sector and intothe private sector will decrease the within-sector component of GINI. (Rememberthat dw < 0.) The third term, the between-group component must, in our model,be negative since β > 1 and wty ppp +< while the sign of wdpdL is unknown.Therefore, it is the between-group component which explains the increase in GINIwhen there is movement from the state sector into the private sector.

The same analysis may be conducted for labour that moves out of the statesector into unemployment. There, however, the within-group component will bedefinitely positive (implying a decrease in GINI) because all three relevantinequalities are the same: tG < wG ; γ < 1; and wy pp < . The between-groupcomponent will also be negative.

In conclusion, the increase in GINI during the transition will be driven by:• Rising inequality within individual sectors (state workers, private workers).

17 We assume for now that tp stays constant.

( ) ( )

( )( ) ( )( ) ( )[ ] Lpp1ppp1ppp111

pGpGpp1GGINI

twtwtwwt

ttt2

ww2

wty

+−+−−−+−−−

+++−−=

γγββµ

πβ

(7)

THE GROWTH OF INEQUALITY 309

This will increase the within-group component of total GINI.• Movement of labour out of the state sector and into the private sector. This

will increase the between-inequality component.However, movement of state-sector workers into unemployment reduces GINI

(lowering both the between- and within-group components).Table 2 summarizes the results. Since for simplicity we have assumed that

group mean incomes stay the same, that rules out one possible cause of increasedinequality.

Table 2. Why GINI changes during transition

Changing composition ofemployment

Increasing groupGinis

From state toprivate

From state tounemployment

Changing meanincomes

Within-groupinequality

Increase Ambiguous(probablydecrease)

Decrease

Between-groupinequality

No effect Increase Decrease

No changeassumed

Now let twenty per cent of the labour force move out of the state sector intothe private sector, and ten per cent of the labour force become unemployed ( uN =0.1). Also let inequality in both the state sector and the private sector rise by 10Gini points to 0.4 and 0.35, respectively. All the other factors, including β and γ,stay as before.18 The share of private-sector employment increases to 40 per centof household heads, and its share in total income to 57 per cent; the share of thestate sector in both employment and total income drops to 30 per cent; the shareof transfer-recipients (pensioners and the unemployed) increases to 30 per cent ofthe population and 14 per cent of income.19 The overall GINI without the overlapcomponent rises to about 34 (Table 3); with the overlap component it rises to 36–37. This is approximately the level of inequality recorded now in the Baltics andthe Balkans. Inequality in Central Europe is lower (about 25) and in Russia andUkraine higher (about 45).

18 Note that β = 1.5 sw is still possible despite the fact that sp ww = (see Figure 2). This is because the

private-sector income includes not only wages but net profits as well, i.e., the rectangle ppNw plus the

area (net profits) between pmpl and pw in Figure 2.19 Before the transition, overall income was 100 wage units (this can be seen from the RHS of equation 1).After the transition, the overall income is 105 wage units: 40 persons in the private sector earning one and ahalf times more than the average wage plus 30 persons in the state sector earning the average wage plus 30transfer recipients earning half of the average wage = 105 wage units.

MILANOVIC310

If we compare Tables 1 and 3 we can see the sources of increased inequality.The within-group inequality increased by less than 2 Gini points. The differencesin average incomes now account for 21 instead of 16 Gini points. This is primarilydue to the fact that the weight ( µty pp ), attached to the large gap between theaverage private-sector income and average pension, has increased. Note that thegreater between-group differences are not due to the changed average relativeincomes between the groups (β and γ were assumed constant), but to a changingcomposition among the employed, and between the employed and the transfer-recipients. In essence what occurred was the ‘hollowing out’ of the middle. Whilebefore the transition, 60 per cent of the household heads were employed at theaverage income (= state-sector wage) while 20 per cent each were either earningmore (the self-employed) or less (pensioners), the transition cut down this ‘middleclass’ to one-half of its previous size. Some of them moved to the private sectorand some became unemployed. Thus both the ‘rich’ and the ‘poor’ increased,while the middle decreased.

Table 3. Breakdown of inequality after transition

Formula Value(in Gini points)

Inequality amongstate workers

3.0

Inequality amongthe private sector

9.1

Wit

hin

-gro

up

Inequality amongtransfer recipients

0.6

Differences inaverage incomes

≅ 21

Total withoutoverlapping

≅ 34

Overlapping 2–3 points

Total GINI about 36–37

Assumptions: wp = 30 per cent, yp = 40%, tp = 30%; Ws = 1; pension or unemployment benefit/wage =0.5; self-employment income/wage = 1.5; Gw = 0.35; Gy = 0.4; Gt = 0.15. Hence, wπ = 29%, yπ = 57%, tπ= 14%, µ = 1.05.

Breaking down the increase in the Gini following the structure laid out inTable 2 allows one to see that the increased group Ginis and the changingcomposition of employment each contributed about 3.5 Gini points (Table 4).

In the empirical analysis in Section 4 we shall try to isolate the dominant forcesbehind the increase in GINI: was this the increase in within-group inequalities; or

( ) ( ) ( )[ ]twstywys pptwpptyppwy1 −+−+−µ

www pG π

ttt pG π

yyy pG π

THE GROWTH OF INEQUALITY 311

changing composition of population as between private and state sectors, andtransfer-recipients; or changing mean incomes of these groups?

Table 4. Breakdown of inequality change during transition

Increased groupGinis

Changingcomposition of

employment

Total GINI

Within-group inequality +3.5 –1.5 +2.0Between-group inequality 0.0 ≅ +5 ≅ +5Total without overlapping +3.5 ≅ +3.5 ≅ +7

Assumptions: wp = 30%, yp = 40%, tp = 30%; Ws = 1; pension or unemployment benefit/wage = 0.5;self-employment income/wage = 1.5; wG = 0.35; yG = 0.4; tG = 0.15. Hence, wπ = 29%, yπ = 57%, tπ =14%, µ = 1.05.

4. Inequality by sources of income 20

The analysis so far has been couched in terms of income recipients. This is aneasier way to proceed for the exposition of what happened during the transition.We can simply say that some persons who used to work in the state sector havetransferred to the private sector. But this ‘tagging’, as mentioned in theintroduction is not realistic (people have numerous sources of income), nor areincome distribution data normally presented in that way. Thus we need to moveto a study of income sources (wages, private-sector income etc.) rather thanindividuals, that is, to the disaggregation of GINI by factor incomes.

The formula for the decomposition of the GINI also gets simpler as the overlapterm disappears. We can now write gross income for each person as the sum ofwages (w), cash social transfers (t), and non-wage private-sector income (y). TheGini coefficient of gross income is equal to the weighted average of theconcentration coefficients Ci of the three individual sources (wages, transfers,private-sector income) where weights are their shares (Si) in total income(equation 8):��

20 Parts of this section are published in Chapter 4 of Milanovic (1998).21 The concentration coefficient captures both inherent inequality with which a given income source is

distributed (source Gini coefficient) and the correlation of that source with the overall income. Thus, aninherently unequal source like social assistance with a high Gini coefficient will have a low or negativecorrelation with overall income (because most of social assistance recipients are poor), and its concentrationcoefficient will be low or negative. When we use the term ‘concentration’ of the source we have in mindnot only its inherent inequality but also how it correlates with overall income. The exact definition of theconcentration coefficient of the source i is Ci = Gi Ri where Gi = Gini coefficient of the source, and Ri =

MILANOVIC312

CS + CS + CS = CS =G yyttwwii3

1=i∑

The change in the Gini between two dates (before and after the transition) canbe written as:

C S + S C + S C + S C + C S =G ii3

1=iyyttwwii3

1=i ∆∆∆∆∆∆∆ ∑∑

The first term on the RHS shows the change in Gini due to the changing sharesof different income sources; the next three terms show the change due to changingconcentration coefficients of individual income sources; and the last term is aninteraction term.

4.1 What happened to factor shares?Consider first what happened to Si’s during the transition. Table 5 shows theshares of wages, non-wage private income, pensions, and other social transfersbefore the transition (1987–89) and in 1995–96 in six transition economies: four inEastern Europe, two in the FSU. For simplicity of presentation, I have selectedonly the end-years for each country. The yearly data are shown in Appendix 1(top panels). All data are calculated from the countries’ Household Budget Surveys.The sample is limited to the countries where I had access to the fairly detailed22

successive annual or quarterly income distribution data.23

Wages are defined as all labour earnings including those from second jobs,fringe benefits (in cash or in kind) and they can come from either the state sectoror the private sector.24 (Private-sector wages were, of course, very rare before thetransition). Pensions include all types of pensions (old-age, survivor, invalidity).Other social transfers include all non-pension cash social transfers like familybenefits, unemployment allowance, sickness benefits,25 scholarships, socialassistance. Finally, non-wage private incomes are a mixed bag. They include self-employment net income, value of home consumption, private gifts andremittances from abroad, net interest,26 dividends, entrepreneurial income,

)]i(rank,icov[

)]income(rank,icov[ ratio of covariances between source i and ranking of recipients according to total income,

and source i and ranking of recipients according to source i. rank (.) is a rank function taking values from 1to N (total number of recipients). Since, in cov[i,rank(i)], both i and rank(i) uniformly increase, its value willbe greater or equal than that of cov[i, rank(income)]. Therefore, R ≤ 1. If the rankings according to totalincome and source i coincide, R = 1, and Cw = Gw.22 Individuals divided into deciles according to per capita household income and income composition byeach decile.23 The list of surveys and the discussion of data quality can be found in the Annex.24 In Poland though, from 1994 private-sector wages are shown separately. They are thus combined withother private-sector incomes.25 Some, however, might be included in wages (if paid out by enterprises).26 In cases of high inflation, however, this source is excluded. In HBSs, interest income is always shown innominal terms. But, in high inflation conditions most or all of nominal interest income compensates for the

(8)

(9)

THE GROWTH OF INEQUALITY 313

income from the lease or rental of assets, etc. This source does not include wagesearned in the private sector (except in Poland) but does, in principle, includedistributed business profits.

The data in Table 5 show:• The share of wages has declined in all countries. Its decline has been

sharper in Russia and Latvia (where initially wages were more importantas an income source) than in Eastern Europe. In the East Europeancountries, the share of wages declined by about 10 percentage points.27 InRussia, the wages’ share dropped by 25, and in Latvia by more than 30percentage points.

Table 5. Composition of population gross income before transition and ‘now’(1995 or 1996)* (in per cent; calculated from HBSs)

Wages Non-wageprivate income

Pensions Other socialtransfers

Countries (years)

pre ‘now’ pre ‘now’ pre ‘now’ pre ‘now’

Bulgaria (89–95) 57 47 22 31 17 18 5 4Hungary (87–93) 60 50 14 16 19 19 7 15Poland (87–95) 55 34 24 30 17 30 5 7Slovenia (87–95) 67 57 20 18 17 22 2 4Eastern Europe 60 47 20 24 17 21 5 8Russia (89–96) 74 49 5 27 8 18 7 6Latvia (89–96) 82 50 12 23 8 18 3 9FSU 78 50 9 25 8 18 5 7

*Except for Hungary (1993).Notes: The regional means are unweighted averages. Rows sum to 100. Non-wage private income forPoland includes private-sector wages in 1995.Sources: See Annex.

• In Eastern Europe, non-wage private-sector income increased by a fewpercentage points only.28 In Russia and Latvia, by contrast, this source of

depreciation of the principal. (Often, when real interest rates are negative, not even that is accomplished.)Strictly speaking, we would need to include as income only the positive real interest portion.27 If Polish private-sector wages were classified together with public sector wages, the wage share inPoland would increase to 50 per cent. This, in turn, would raise the mean East European wage share to 51per cent, almost exactly the same as in FSU.28 Before the transition, the entire private-sector income belonged there (as in our description of transitionin Section 2). After the transition, many of the successful self-employed and small enterprises grew into‘regular’ private firms. Their wage payments, as well as those of de novo private companies and theprivatized SOEs, are now included together with other wages (except in Poland). This explains why thissource of income – which now accounts for only a part of total private-sector income – has not grown moresubstantially. Another reason was a sharp drop in relative (in relation to overall country average) incomeof farmers after the transition.

MILANOVIC314

income increased its share by about 15 percentage points.• The share of pensions increased without exception in all countries. The

increase again was sharper in Russia and Latvia which started with a lowerpre-transition share, and in Poland where pensions now account for 30 percent of total income.

• Non-pension social transfers increased in all countries except in Russia.Their growth has been particularly dramatic in Hungary where in 1993they represented 15 per cent of population income (10 per cent in the formof various family benefits, 4 per cent as unemployment allowances).

• Overall, the income compositions in Eastern Europe, on the one hand, andRussia and Latvia, on the other, are more similar now than before thetransition. Wages in all countries are around one-half of gross income, non-wage private-sector incomes account for one-fourth of gross income and sodo cash social transfers. But because the initial starting point in Russia andLatvia was further from the current outcome, the change in incomecomposition in these two countries was more dramatic than in EasternEurope.

4.2 What happened to factor concentration coefficients?Consider next iC ’s from equation 8.• In all countries, the concentration coefficient of wages went up (Table 6).

The increase was substantial for the East European countries, averagingabout 10 points (or almost 50 per cent of the initial value), and truly‘gargantuan’ for Russia. The concentration coefficient of wages in Russia in1996 is 60. A concentration coefficient can increase either because thedistribution of the income source (wages in this case) becomes moreunequal or the correlation between that sources (wages) and income rises.Without additional information (i.e., individual data) we cannot distinguishbetween the two.29 Whatever the cause, higher concentration of wagesclearly puts an upward pressure on the overall Gini.

• Non-wage private income has kept the same concentration coefficient inEastern Europe overall; the changes in individual countries are small too.Before the transition, private-sector income had a substantially higherconcentration than wages in Bulgaria, Poland and Hungary (30’s vs. 20’s).Since wages’ concentration coefficients have risen, the two (wages, andnon-wage private-sector income) now have about the same concentration.In Russia and Latvia, private income’s concentration, like that of wages, hasincreased substantially.

• The average concentration coefficient of pensions in Eastern Europe hasincreased from 16 to 23 due to changes in Poland and Hungary. Theimprovement in the average ratio between pensions and wages in Eastern

29 See, however, an attempt in Section 5.

THE GROWTH OF INEQUALITY 315

Europe30 (see Table 7) has led to pension income (and pensioners) nowbeing distributed across the entire income spectrum – much more so than inthe past when they were more concentrated among the middle to poorersegments of the population.31 In Latvia, the differentiation of pensions andthus their concentration coefficient have declined following theintroduction of ‘flat’ pensions in 1992 (see Appendix 1, bottom panel forLatvia).

Table 6. Concentration coefficients before the transition and ‘now’ (1995 or 1996)*(in per cent; based on HBSs; individuals ranked by per capita gross income)

Wages Non-wageprivate income

Pensions Other socialtransfers

pre ‘now’ pre ‘now’ pre ‘now’ pre ‘now’

Bulgaria (89–95) 21 34 38 37 11 13 –6 2Hungary (87–93) 25 35 30 26 14 21 –13 –16Poland (87–95) 25 40 37 40 17 37 –10 –10Slovenia (87–95) 20 26 18 21 22 21 –4 –19Eastern Europe 23 32 31 31 16 23 –8 –11Russia (89–96) 28 60 18 56 –20 27 8 42Latvia (89–96) 23 41 16 43 34 9 –7 7FSU 25 50 17 50 – 18 – 25

*Except for Hungary (1993).Notes: The regional means are unweighted averages. – indicates that the country differences are so largethat averaging is meaningless.Sources: See Annex.

• The change in the concentration coefficient of other (non-pension) socialtransfers is important too, because the brunt of anti-poverty policy(particularly in conditions of massive income declines) falls on thesetransfers. By their nature, they are either explicitly targeted on the poor likesocial assistance, or implicitly so, like unemployment benefits. In EasternEurope, their targeting has improved from being very mildly pro-poor inabsolute terms (–8) to more strongly so (–11). Almost the entireimprovement in targeting is due to the introduction of unemploymentbenefits (Milanovic, 1998, pp. 108–14). Since (i) the unemployed are oftenamong the poor and are easily identifiable, and (ii) the rules for benefits’eligibility are relatively clear and are being observed by unemploymentoffices, unemployment benefits have been focused on the poorer segments

30 Except in Bulgaria.31 For example, in Slovenia, in 1983, 28 per cent of households with pensioners were in the lowest quintile;ten years later, the figure was only 22 per cent (see Stanovnik and Stropnik, 1998, p. 15).

MILANOVIC316

of the East European population.32 In Russia and Latvia, targeting of non-pension cash benefits has deteriorated. Particularly striking is thedevelopment in Russia where non-pension transfers have a concentrationcoefficient of 42 – apparently being targeted towards the rich rather thanthe poor. 33

Table 7. Average pension as percentage of average wage

1987 1997Bulgaria 44 32Hungary 55 56Poland 51 64Slovenia 55 66Eastern Europe 51 55Russia 37 34Latvia 40 35Former Soviet Union 39 35

Notes: The regional means are unweighted averages.Bulgarian data for 1996 instead of 1997.Source: World Bank (DEC) database.

4.3 Decomposing the overall GINI changeThe outcome of the changes in iS ’s and iC ’s is the change in overall GINI. Table8 shows the decomposed change in the Gini between a pre-transition year (1987or 1989) and 1995–96. The last column shows the overall increase in GINI betweenthe two end-periods. The pre-transition year, whether 1987 or 1989, is fixed by thedata availability. Since income distribution in the late 1980’s was stable, the exactyear does not matter much. This, however, is not the case for the choice of thetransition year because now inequality does change. To avoid ‘pinning’ allconclusions onto one end-point year, I present in Appendix 2, the year-after-yeardecomposition of the GINI change for the entire period. (The year-to-year levelGINIs are shown in Figure 3.)

We can draw four conclusions: First, the change in the composition of incomehas had very little to do with increased inequality. In the only country where itdid have a significant impact (Russia), it contributed to a reduction in inequality,i.e., the composition of income in 1996 was more favourable to equality than in1989.34 This is chiefly because social transfers, which were the most equallydistributed income source in Russia before the transition, increased their share in

32 Point (ii) for example does not hold for social assistance. Also, total spending on unemployment benefitsexceeds spending on social assistance in all East European countries.33 The same point is made by Commander and Lee (1998, p. 7) by who use Russian longitudinal data forthe period 1992–96.34 Had, of course, the concentration coefficients of various sources remained at their pre-transition levels.

THE GROWTH OF INEQUALITY 317

overall income (see Table 5). In Hungary too, the change in compositiondampened the overall increase in inequality because of the rising importance of(relatively equal) pensions and other transfers. In other countries, changed incomecomposition added or subtracted only about 1 Gini point to total inequality.

Second, higher concentration coefficients of wages (in all countries) drove theoverall Gini up. It was the most important factor behind the increase in inequality.The increased wage concentration was responsible for between 3.5 and 8 Ginipoint increase in Eastern Europe, and for huge 15–24 Gini point increases in bothLatvia and Russia. In the latter two countries, these large increases are due notonly to a greatly increased concentration coefficient of wages, but also to a veryhigh initial 1989 share of wages in income. Thus the weight attached to a moreunequal concentration of wages is greater (see the term Sw in equation 8) than ifthe original share were low.

Figure 3. Gini coefficients in Eastern Europe and former Soviet Union (based onindividuals ranked by annual household per capita gross income)

Sources: See Annex.

The increased non-wage private-sector income concentration was responsiblefor more than 4 Gini points increase in Russia and about 1.5 points in Latvia,while its impact was negligible in Eastern Europe.

Third, the effect of transfers on inequality was not uniform across thecountries. In Bulgaria and Slovenia, the concentration of transfers did not change(see Table 6). In Latvia, better targeting of transfers reduced overall inequality by

0

10

20

30

40

50

60

1987 1988 1989 1990 1991 1992 1993 1994 1995 1996

Russia

Latvia

Poland

Slovenia

Bulgaria

Hungary

MILANOVIC318

1.5 Gini points following the introduction of ‘flat’ pensions in 1992.35 In Polandand Russia, on the contrary, greater concentration of transfers (principallypensions) increased overall inequality. Non-pension transfers, because of theirsmall initial size, did not anywhere (except in Russia where their impact wasperverse) have much effect on the overall change in inequality.��

Fourth, for several countries (Poland, Slovenia, Latvia, Hungary) we observe agradual closing of the gap between the concentration coefficients of private-sectorincome, wages and pensions. This suggests that people at different income levelsderive approximately the same proportion of total income from different sources.The rich do not depend for a greater part of their income on wages than onpensions, etc. This is illustrated for Poland in Figure 4: public sector wages are aconstant proportion of gross income from the 4th decile onward; private-sectorincome contributes the same proportion of income for all deciles except the topand the bottom; the share of pensions varies within a fairly narrow range.

Figure 4. Public sector wages, private-sector income, and pensions as share ofgross income by decile, Poland 1995 (in per cent)

Source: Household Budget Survey.

35 Note the steep downward-sloping line for the concentration coefficient of pensions in Appendix 1 forLatvia (bottom panel): the concentration coefficient of pensions decreased from 34 in 1989 to –4 in 1995.36 This conclusion differs to some extent from Cornia’s (1994, p. 39) observation that ‘the relativeimportance of redistribution [via transfers] has grown...Targeting of these [social] transfers has generallyimproved or remained sufficiently progressive.’

10.0

20.0

30.0

40.0

1 2 3 4 5 6 7 8 9 10

Deciles (acc. to gross per capita income)

Public wage

Pensions

Private sector income

Table 8. Decomposition of the change in the Gini coefficient between 1995 or 1996 and before the transition(in Gini points)*

Due to:Change in concentration of:

Out of which:

Country (end years)Change in

compositionof income

Wages Socialtransfers Pensions Non-pension

transfers

Non-wageprivatesector

Interactionterm

Overall Gini change(between the end-years)

Bulgaria (89–95) +1.6 +7.8 +0.9 +0.4 +0.4 –0.4 +0.3 +10.0 (from 21.7 to 31.7)Hungary (87–93) –2.7 +5.5 +1.2 +1.4 –0.2 –0.6 –1.1 +2.2 (from 20.7 to 22.9)Poland (87–95) –1.3 +7.9 +3.3 +3.4 –0.1 +0.6 –0.1 +10.6 (from 25.0 to 35.6)Slovenia (87–95) –0.3 +3.6 –0.5 –0.1 –0.4 +0.4 –0.7 +2.6 (from 19.8 to 22.3)Russia (89–96) –6.2 +23.6 +6.0 +3.7 +2.3 +4.3 +2.2 +29.9 (from 21.9 to 51.8)Latvia (89–96) –1.8 +15.0 –1.5 –2.0 +0.5 +1.4 –3.1 +10.0 (from 22.6 to 32.6)

*Except for Hungary (1993).

MILANOVIC320

The ‘stylized’ facts of transition are illustrated well by the example of Bulgariain Appendix 1. The rising concentration of wages (from around 20 to 35)contributed strongly to inequality. The concentration coefficient of private-sectorincome which was already high before the transition, remained at the same levelwhile the share of private-sector income in the total increased. The rising share ofprivate-sector income thus also pushed up the overall inequality. Pensions’concentration and share both remained unchanged leaving overall inequalityunchanged. Finally, non-pension transfers were too small (less than 5 per cent oftotal income) to make any difference to the overall GINI. In conclusion, the twomain causes of increased inequality were the rising concentration of wages, andthe growing share of private-sector incomes.

Polish and, to a lesser degree, Slovenian results illustrate a slightly differentstory (see Appendix 1 for the two countries). Although wage concentrationincreased markedly, the most dramatic developments were in the area of socialtransfers: their rising concentration, and rising share in overall income. In 1995,pensions had about the same concentration coefficient as wages and non-wageprivate-sector income.

Russia represents a unique case of a country where all income sources'concentration coefficients are higher now than before the transition: therefore theyall pushed up overall inequality (see Appendix 1 for Russia, bottom panel).

5. The conclusions: contrasting the model and the empiricalevidence

The model and numerical simulations from Sections 2 and 3 led to severalpredictions regarding the changes in factor shares and inequality during thetransition. We shall consider three predictions: regarding (i) the change in incomeshares, (ii) increased inequality with which different income sources aredistributed, and (iii) the ‘hollowing out’ of the middle class (state-sector workers).

First, we expect, of course, a declining share of state-sector income. The modelalso ‘predicts’ that, after the transition, the share of transfers in total income willbe greater. This is because the government will have to pay unemploymentbenefits in addition to unchanged relative pensions.37 Indeed, this is what theevidence from six transition economies in Table 5 confirms. There is not a singlecountry where the share of wages did not decline and the share of cash socialtransfers did not increase.

Second, one cause of increased inequality, according to our model, lies inincreased within-group inequality. In the empirical part, we saw that theconcentration coefficient of wages ( wC ) increased everywhere and that it was themost important element driving up GINI. A higher wC may be due either to a

37 That is, in relation to state-sector wages.

THE GROWTH OF INEQUALITY 321

higher Gini coefficient of wages (which is what we would expect from our model)or to the increase in the correlation coefficient between wages and overall income.But as wages became a smaller proportion of overall income, it is very unlikelythat the correlation coefficient between wages and income increased. So, most ofthe increase in the concentration coefficient of wages must have been due to thegreater inequality among wage-earners.

Third, according to our model, the most important source of increase in the Gini is the ‘hollowing out of the middle’, that is the movement of state-sectorworkers into either ‘rich’ private sector activity or ‘poor’ unemployment. Weknow that the share of wages and, in particular, state-sector wages, in grossincome has decreased (see Table 5). But that could have happened either becausewages relative to the mean population income went down, or because the numberof wage-earners decreased. In our numerical calculations, we assumed that peoplemoved out. As Table 9 shows, the wage-to-income ratio stayed constant or evenslightly increased during the transition. Thus, the declining share of wage incomemust indeed have been due to the movement of people out of the state sector, thatis, to the ‘hollowing out of the middle.’ But did some of them go into the ‘rich’private sector, while some joined the ‘poor’ unemployed, as we have initiallyassumed? To find out more about that we need longitudinal data, and anemphasis on the issues of polarization rather than inequality — both topics thatare beyond the scope of this paper.

Table 9. Ratio between wage and mean householdincome (per month)

1987* 1996**

Bulgaria 1.27 1.56Hungary 1.01 1.03Poland 1.87 1.86Slovenia 1.62 1.60Eastern Europe 1.44 1.51Russia 1.48 2.02Latvia 1.34 1.92Former Soviet Union 1.41 1.97

Note: Wage data from official statistics. Mean household incomefrom budget surveys. Data for Russia in 1996 exaggerate the wageratio because the official statistics record as wages both paid andunpaid (but due) wages, while the household budget surveysinclude only the paid part. * Year 1989 for Bulgaria, Hungary, Russia and Latvia (the sameyear as in the Gini decomposition in Table 8).** Year 1995 for Bulgaria, Hungary and Slovenia.

MILANOVIC322

Annex. Description of the surveys used and data problems

Description of the surveys usedFor all East European countries and Latvia, the data were obtained from theofficial surveys conducted by the countries’ statistical agencies (CSO). For Russia,the 1989 data is from the official survey; the 1992, 1994, 1995 and 1996 surveys arethe World Bank and Goskomstat Rossii jointly sponsored Russian LongitudinalMonitoring Survey, RLMS (see Table A1 below).

For Poland, Bulgaria and Slovenia all survey instruments are the same, that isthe surveys for each individual country have exactly the same design year afteryear (barring some improvements: e.g., the Polish surveys became fullyrepresentative in 1993). For Hungary, the 1987 and 1993 survey instruments arethe same (Household budget survey that is normally conducted once every twoyears, but whose 1991 results were not published). The 1991 survey is amicrosimulation of a large 1987 Income survey conducted by CSO. For Latvia, the1989 survey is a Living Standard Survey;38 the 1992–93 surveys, however, are theunrepresentative quota-sample Soviet Family budget surveys. Finally, for 1995–96, Iuse the new and representative New Latvian Household Survey. For Russia, the 1989data come from the old Soviet survey; the 1992–96 data are from therepresentative RLMS (Rounds 1, 4, 6 and 7).

Almost all surveys are annual. The shortest ones are the 1995 and 1996 Latviasurveys, and RLMS Rounds 1, 6 and 7 which are quarterly. Out of the total of 37surveys used, 30 are annual, two are semi-annual, and five are quarterly.

Out of 37 surveys used, I had access to individual data for seven. For all othersI used the grouped data. The number of published groups varied between 10 and20. The income groups were formed according to CSOs’ definitions of per capitaincome. In two countries (Slovenia and Bulgaria) this led to some problemsbecause the CSO-defined income included items that did not belong to income. InSlovenia, it included net withdrawals from saving accounts and personalborrowing; in Bulgaria, sales of assets and insurance compensations. These itemshad to be deducted from the CSO-defined income in order to obtain actualdisposable income. Performing this operation on the grouped data (in distinctionto individual data) implies that the measured income inequality becomesunderestimated because we no longer, strictly speaking, estimate the Ginicoefficient of disposable income but the concentration coefficient of disposableincome. The problem is negligible in the case of Bulgaria because the ‘wrong’items account for less than 1 per cent of the CSO-defined income; in Slovenia,however, they account for about 8 per cent of the CSO-defined income.

For Poland, Bulgaria, Latvia and Russia, for a number of years, the incomeconcept used is gross rather than the disposable income.39 However, personal

38 Living standards survey was conducted once every five years.39 Disposable income = gross income minus payroll and direct personal taxes (PIT).

THE GROWTH OF INEQUALITY 323

income taxes were minimal because gross income excludes payroll taxeswithdrawn at source which represented the largest chunk of personal taxes. In allcases, the difference between disposable and gross income was less than 1 percent, and thus using either of the two concepts would produce the same results.This, of course, has changed now with the introduction of a more substantial PITsystem in Hungary and Poland. Finally, in all cases except Hungary in 1993,home-consumption is included in income.

The components of disposable income are standard. Disposable income isequal to all wage earnings (from primary and secondary jobs, etc.) plus cash socialtransfers plus income from property and entrepreneurship plus received giftsplus value of home consumption. It excludes payroll and PIT taxes.

Comparing pre-transition and transition years: what are the biases?

The comparison between Polish, Bulgarian, and Slovenian survey results over theperiod 1987–95 is straightforward and warranted. All the data essentially comefrom the same surveys, and no dramatic changes in the refusal rates (they wentup though) or under-reporting (it went up too) occurred. The same is, to a largeextent, true for Hungary whose 1987 and 1993 survey instruments are the same.

However, this somewhat optimistic assessment needs to be qualified. There isa change for which the users of surveys, and possibly the ‘producers’ of surveystoo, could not control. It is the change that accompanied the transition, and hasnothing to do with the survey design per se. We can call it systemic or underlyingchange. Generally speaking, refusal rates have increased during the transition andparticularly among the rich; coverage of wage and social transfer income that wasnearly 100 per cent before the transition has deteriorated (as earnings reported bythe households which in the past used to be double-checked against the enterpriseor pension authorities’ records are no longer so checked); the omission orinadequate coverage of informal (and illegal) sector income before the transitionhas now become an even greater problem as such incomes have increased inabsolute and relative terms. These are some of the problems for which we, asusers of the surveys, cannot correct. The bottom line effect of these systemicchanges – assuming an unchanged survey design – is that incomes are now moreunderestimated than in the past. The direction of the bias in terms of inequality isless clear. In the past, surveys underestimated inequality by not accounting formany fringe benefits and perks received by the elite.40 Today, they mightunderestimate it by not covering those with high incomes who refuse toparticipate.

But, it is up to each researcher to decide how strong an emphasis he or shewishes to place on these systemic (underlying) changes; how much he or shebelieves that they vitiate all pre-post comparisons. I would tend to believe that the

40 Subsidies were not included either; yet their effect was (with the possible exception of housingsubsidies) to reduce inequality.

MILANOVIC324

underlying change in Eastern Europe was not of such a magnitude as to render,after appropriate caveats, the comparisons of inequality before and after thetransition unreliable. On the other hand, the argument that such comparisons aremuch less reliable can be, I think, made with respect to some of the republics ofthe former Soviet Union. In the Soviet case, not only was the underlying changemuch more profound (witness the explosion of the informal sector), but the initialsurveys were fundamentally flawed because they were not random surveys butbasically surveys of the families of the employed (see the discussion below),among which the ‘average’ households (e.g., both parents employed, and having1 or 2 children) tended to be over-sampled. To these households a quota of thepensioners and students living outside their homes was added. The pre-transitionsurveys in the former Soviet Union were biased, left out large segments of thepopulation, and tended to show higher average incomes and lower inequality.Thus the recorded change in inequality in Latvia and Russia is almost certain toappear larger than the actual change.

Survey biases before the transitionWhat can we say more formally about the survey biases before the transition? Weshall consider four areas: survey design, under-reporting of income, the use of percapita versus equivalent units, and annual versus quarterly data.

The very fact that these caveats are listed here indicates that we cannot domuch (or anything) to remedy them. Yet they are worth listing for two reasons: toprovide some caution when it comes to the interpretation of the results, and todelineate the areas that most clearly need improvement in future work.

Survey design

This is the problem of sampling inadequacy. The household surveys that we usehave been criticized, rightly, for several biases. The Eastern European surveyswere sample surveys. However, in several countries (e.g., Poland), they were notdesigned to be representative of the entire population but rather of individualsocioeconomic groups (SEGs). This was probably the product of a Marxist view ofsociety as composed of social classes and concern with intergroup equity. Thedata were thus representative of workers’ households (in the state sector) or ofpensioners, but they could not be easily combined to obtain an accurate picturefor the whole population, essentially for two reasons. First, the sample shares ofthe groups that were included were not always proportional to their shares in thepopulation (e.g., there were too many workers and not enough pensioners) andthe results were not corrected for systematic differences in the rate of refusal toparticipate in the surveys. Second, some groups were left out of the surveysentirely. These groups included both those with high incomes (self-employedentrepreneurs, Army and police personnel) and those with low incomes (theinstitutionalized population, the unemployed). Income distribution was truncatedat both ends.

THE GROWTH OF INEQUALITY 325

The Soviet data were even more problematic. Not only could data for SEGs notbe combined, but the surveys were not based on a sample technique but onselecting households at their place of work (the so-called ‘branch [of production]approach’). Workers and farmers were chosen by their managers and asked to co-operate with statistical authorities. The results were biased: the employed weresystematically over-represented in relation to the non-employed (to correct someof the bias a quota of pensioners and students was added);�� workers in largeenterprises and with a longer work record were more often selected that thoseworking in small firms and with a shorter work record. Since the selectioncriterion was employment, larger households were under-sampled.�� The surveywas essentially a panel – with the same households staying in the sample yearafter year – but to further complicate matters, it was not explicitly designed as apanel, and the identification numbers of the households were not systematicallymaintained. The panel nature of the survey further biased the results: sincehouseholds were supposed to stay in the sample indefinitely, the share of olderworking households,�� presumably with higher than average earnings, was toohigh.

In conclusion, there were two kinds of biases. First, a bias toward samplingthat is representative of various pre-defined socio-economic groups but not of thepopulation as a whole with the result that people that could not ‘fit’ into any ofthe main social groups were likely to be left out, and these people were often atgreater risk of poverty than the average citizen. Second, a bias toward the‘average’ or ‘normal’ households that existed only in the household surveys thatfollowed the so-called ‘branch principle’. These are Soviet and, it seems,Romanian surveys.�� The sampling selection was skewed in favour of ‘average’enterprises, ‘average’ workers in terms of earnings, ‘average’ skills, ‘average’family size (a couple with one or two children), etc. Thus even within a givensocial group (workers in state enterprises) income distribution was truncated.

Under-reporting of income

The second problem has to do with income. The use of income, rather thanexpenditure, data tends to underestimate ‘true’ welfare. This is because peopletend to hide their sources of income and thus to under-report them.�� They are

41 The pensioners households were simply ‘added on’ i.e., statistical offices would be asked to add a quotaof pensioners which was often below their true share in the population.42 In order to have unbiased results, the probability of selection of a larger household should beproportionately greater than the probability of selection of a smaller household. But when the criterion ofselection is employment and the participation rates are high, the two households (e.g., one with two adultsand three children, and another with two adults only) will have approximately the same probability ofbeing selected.43 Once households stopped working, they generally tended to drop out.44 In addition, Romanian results were doctored to such an extent that they are worthless for the yearsbefore 1989. Bulgaria also followed ‘the branch principle’ in the 1970’s, but abandoned it later.45 The under-reporting problem exists in market economies too. It is particularly severe for self-employment and capital income. Atkinson, Rainwater and Smeeding (1995, Table A3) find, using LIS data,

MILANOVIC326

less careful when asked to remember their expenditures. An example of thistendency is shown in Figure A1 below, which gives income and expenditure databy ventiles (5 per cent of recipients) for Poland in 1993. Individuals are ranked onthe horizontal axis according to their level of household (per capita) income. Aninteresting fact is revealed by the situation of the lowest income ventile. Thereported expenditures of the lowest ventile are twice its income and are equal tothe expenditures of the fourth ventile. This indicates a possible measurementproblem: the people in the lowest income ventile are in reality not very differentfrom those who are (according to income) significantly better off. It seems thatthey either severely under-report their income or that their permanent incomesubstantially diverges from their current income. But in any case, in our statisticsthey would be counted as poor. We thus impart an upward bias to the povertyrates. Generally speaking, the countries that have a greater share of the informalsector (‘gray economy’) and small-scale private sector will be more affected.��

Their data will systematically show lower incomes and higher poverty than thedata in countries in which an overwhelming share of income is earned in the statesector, or in the wage-reporting (and thus tax-paying) private sector, or isreceived in the form of social transfers. Inter-temporal comparisons will beaffected too. As the share of the gray economy rises with the transition, theproblem becomes more serious. To offset this, however, improvements have beenmade in the survey techniques and greater effort is now made to include such‘gray’ sources of income. For example, all countries except those that still stickwith the Soviet-type surveys, now include the self-employed in their HBSs.47

Hungarian statistical authorities have been imputing tips, ‘fees’ and ‘black’income. Finally, while the gray income often remains illegal (in the sense thatpeople do not pay taxes on it), there is no longer a political compulsion to ignoreit. The political compulsion existed in the past, with both households and theenumerators being keenly aware that such sources of income are not only illegalbut also ‘politically incorrect.’ Both preferred to look the other way and ignore allnon-official sources of income. This source of bias is now gone.

Per capita versus equivalent units

International comparisons of poverty are – to make an understatement – complicated. In the particular context of transition economies, there are at leastseveral problems that must be mentioned. First, the use of a per capita poverty lineexaggerates poverty in any country compared to using an equivalent-scalederived poverty line. This is because with an equivalence scale the needs ofadditional family members (often children) do not rise in proportion as they do

that self-employment income is underestimated (compared to the national accounts data) by between 10per cent (Canada) and 60 per cent (W. Germany). Property income in almost all countries (US, UK, Italy,Germany, Finland, Canada and Australia) is underestimated by a half.46 Eighty-four per cent of individuals belonging to the lowest income ventile are individual farmers,farmers-workers, and the self-employed (outside agriculture).47 In the past, they were left out of the surveys.

THE GROWTH OF INEQUALITY 327

when we use a per capita measure. By implication then, the use of a per capita lineexaggerates even more poverty in countries with larger average family size.�� Forexample, Marnie and Micklewright (1993) compare Uzbekistan and Ukraineusing the same Soviet Family Budget Surveys (FBS) for 1989. They find that thelarger household size in Uzbekistan accounts for 14 out of the 38 percentage pointdifference in the headcount index between the two republics. There are severalreasons why we use the per capita line: the data in all countries are published inthat form (rather than being adjusted for household composition by the use ofequivalence scales); the economies of scale in consumption under socialism weretypically less than in market economies because the main source of sucheconomies of scale (housing, utilities, etc.) were heavily subsidized and this stillremains true although to a lesser extent; the use of per capita poverty comparisonallows us to move easily from such per capita comparisons to GDP per capitacomparisons.

Quarterly versus yearly data

A final problem concerns the time period of data collection. Normally, the surveysare designed in such a way that households report (i.e., keep track of, or recall)income and expenditures for a quarter or a month. These data are then ‘blown up’for the entire year.�� Under conditions of high inflation, however, the datacollected in different months represent wholly different real quantities of goodsand services and cannot be summed up unless they are adjusted for inflation. Theadjustments are sometimes not made by statistical agencies, and at times they aremade inadequately (if, for example, inflation is understated). Under suchconditions, quarterly data on income or expenditure distribution, from which wecalculate poverty figures for several FSU countries, are to be preferred becausethey refer to a shorter time period and imply about the same command over realgoods and services. The usual drawback of the short-period data, namely, thatthey overestimate income inequality and poverty (there are more people withextraordinarily low and high incomes the shorter the time period), is then of lessimport than the advantage that the same reported money amounts representapproximately the same real quantities of goods and services.

Comparison: before and after the transition

If we assume that the ‘best’ achievable household survey should be:(i) representative for the country as a whole, (ii) refer to annual income; (iii) use

48 In a recent paper Coulter, Cowell and Jenkins (1992) show that the poverty headcount charts a U-shapedpattern, first decreasing and then rising, as equivalence scale moves from 0 (full economies of scale: totalhousehold income alone counts) to 1 (per capita calculations). The same results are obtained by Forster(1993, p. 21) in an empirical study of 13 OECD countries.49 In order to increase the response rate (which was one of the main sources of bias), Polish householdsurveys began to require households to keep track of their income and expenditures for one, instead ofthree, months. The response rate increased from 65 to 80 per cent (see Kordos, 1994).

MILANOVIC328

disposable income; (iv) include home-consumption, and (v) have income‘correctly’ defined, and consider the transition-year surveys, no survey fulfils allfive conditions but several come close. Bulgarian and Slovenian surveys have aslightly incorrect definition of income for which I could not fully adjust as I didnot have the individual data; Hungary’s 1993 survey does not include home-consumption in income;50 for Poland in 1993, I had semi-annual instead of annualdata. Russian and Latvian 1994–96 surveys are quarterly. One may also besomewhat skeptical regarding the claim that the Russian RLMS survey is self-weighted.

(Note, however, that even the ‘best’ achievable survey would still containsome possible biases – which are present in all the surveys, including in those indeveloped countries. The ‘best’ survey would still probably understate the twotail-ends of income distribution (the poorest and the richest) who are typicallyunder-surveyed, and it would also underestimate some sources of income likethose from property, which are routinely underestimated by up to 50 per cent indeveloped countries,51 and entrepreneurship.)

Table A2 shows where the surveys fall short of the five requirements listedabove, and what this implies in terms of the bias when estimating inequality.

In the case of Hungary, the absence of home-consumption will lead to a slightincrease in inequality, because home-consumption is generally greater for thepoorer households. In Poland, too, there should be a slight overestimate of theincrease because the pre-transition surveys did not include the entire population(i.e., left out some well-off segments).

For Bulgaria, the use of gross income instead of disposable will reducemeasured inequality (to the extent that personal income taxes are progressive).However, since PIT is very small, as most taxes are deducted at source, thedownward bias is negligible.

The bias is more serious for Latvia and Russia. The old Soviet survey could notsatisfy more than two (annual data and inclusion of home consumption) out ofthe five requirements listed above.52 The choice of households to participate in thesurveys was biased. The new improved surveys suffer from too short a period ofobservation (a quarter).

50 It includes it in expenditures, but it could not have been separated from other expenditures and thuscould not be added to income.51 The so-called ‘non-earned’ income is understated by about 40 per cent (compared to national accountsstatistics) by the US Current Population Survey (see Michel, 1991, p.185).52 In addition. since the difference between gross and net income was negligible, they could be said tosatisfy the condition (iii).

Table A1. Characteristics of the surveys

Country(number ofsurveys)

Source of data;survey conductedby:

Periodcovered

Period ofanalysis

Data reported in: Representativesurvey

Incomeconcept

Incomeincludeshomeconsum-ption

Otherproblemswith incomeorexpendituredefinitions

Poland (8) CSO, Householdbudget surveys(Budzetygospodarstwdomowych)

1987–93;1995

1987–92 and1995 yearly;1993 first sixmonths.

Grouped data for the period 1987–92(published in the annual Budzetygospodarstw domowych, CentralStatistical Office, Warsaw).Individual data for 1993 and 1995.

Fullyrepresentativesince 1993;before policy,Army and non-agriculturalprivate sectoromitted.

Disposableafter 1993;gross before(the differencebetween grossanddisposablewas minimal;less than 1%)

Yes None

Hungary (3) CSO, Householdbudget surveys for1987 and 1993;Income surveymicrosimulationfor 1989

1987, 1989,1993

Yearly. For 1987, grouped data as publishedin Csaladi Koltsegvetes 1987, CentralStatistical Office, Budapest, 1989, pp.78–9, 102–3, 126–7. For 1989, detailedgrouped data from Kupa and Fajth(1990). For 1993, individual dataavailable.

Yes Disposable No None

Slovenia (9) CSO, Householdbudget surveys.

1987–95 Yearly. Grouped data. For 1987–91 datapublished in Anketa o potrosnjidomacinstava, Federal statisticalOffice, Belgrade (all data presentedby republics). For 1992–95 personalcommunication by the Statisticaloffice of Slovenia (Mrs. IrenaKrizman and Mrs. AlenkaKajzer).Some results published alsoin Statisticne informacije, and Rezultatiraziskovanja st.684/1997 StatisticalOffice of Slovenia, Ljubljana.

Yes Disposable Yes Incorrectdefinition ofincome(adjustmentmade)*

Table A1. Characteristics of the surveys (continued)

Country(number ofsurveys)

Source of data;survey conductedby:

Periodcovered

Period ofanalysis

Data reported in: Representativesurvey

Incomeconcept

Incomeincludeshomeconsump-tion

Otherproblemswith incomeorexpendituredefinitions

Bulgaria (7) 1989–94 CSO,Household budgetsurveys; 1995 Gallupsurvey

1989–95 1989–94 yearly;1995 first sixmonths.

For 1989–94 grouped data reportedin Biudzeti na domakinstvata vRepublika B’lgariya, NationalStatistical Institute, Sofia. For 1995,individual data available.

Yes 1989–94 grossincome (thedifferencebetweengross anddisposableminimal).1995disposable.

Yes Slightlyincorrectdefinition ofincome(adjustmentmade)*

Latvia (5) CSO, 1989 Livingstandards survey;1992–93 Familybudget surveys;1995–96 New Familybudget surveys

1989, 1992,1993, 1995,1996

1989, 1992–93annual; 1995–96quarterly

Grouped data. For 1989 datareported in Survey of LivingStandards, Riga: GoskomstatLatviiskoi SSR, 1990; for 1992–93reported in Gimenes budzets, Riga:CSO; for 1995–96, personalcommunication by the LatvianStatistical Offices (Mr. EdmundsVaskis).

Not before 1995(the Soviet-typebranch-basedsurvey).Representativesince 1995.

1989–93 grossincome (thedifferencebetweengross anddisposableminimal).1995–96disposable.

Not in1992–93.Yes forother years.

1989: slightlyincorrectdefinition ofincome.

1992–93moneyincome only.

Russia (5) 1989 CSO; 1992,1994, 1995, 1996RLMS

1989, 1992,1994

1989 annual;1992 quarterly1994 (Oct. 1993–Feb. 1994),1995–96quarterly

For 1989 and 1992, grouped datareported respectively in Popkin(1992, Table 5) and Popkin (1993,Table 20). For 1994, 1995 and 1996,individual data available.

1989 notrepresentative(the Soviet-typebranch-basedsurvey); 1992–96surveysrepresentative.

1989 grossincome (thedifferencebetweengross anddisposableminimal).1992–96disposableincome.

Yes 1989 slightlyincorrectdefinition ofincome.

Note: CSO = country’s statistical office. HBS = Household Budget Survey. FBS = Family Budget Survey. RLMS = Russian Living Standards MonitoringSurvey. *Disposable income calculated by deducting some revenue items from the income concept used by the CSO (and according to which theindividuals and households were ranked).

THE GROWTH OF INEQUALITY 331

Table A2. Survey defects and inequality bias: pre-transition and transition years

Pre-transition Transition Change in inequalitybias

Poland Incomplete coverage ofrecipients

Semi-annual in 1993; butannual in 1995

Slight overestimate ofthe increase

Hungary No home consumptionin 1993

Slight overestimate ofthe increase

Slovenia Income definitionproblem

Income definitionproblem

None

Bulgaria Minor income definitionproblemGross instead ofdisposable income

Minor income definitionproblemGross instead ofdisposable income

Negligible

Latvia Quota, biased sample Quarterly instead ofannual data

Overestimates increase(strongly)

Russia Quota, biased sample Quarterly instead ofannual data

Overestimates increase(strongly)

Figure A1. Income and expenditures per capita, Poland 1993 (individuals rankedby income per capita).

0

500

1000

1500

2000

2500

3000

3500

4000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Ventiles

'000

zlo

ty

Expenditure

Income

MILANOVIC332

References

Atkinson, Anthony B. (1994), ‘Seeking to Explain the Distribution of Income’, DiscussionPaper WSP 106, Centre for Economics and Related Disciplines, London School ofEconomics, December.

Atkinson, Anthony B. and John Micklewright (1992), Economic Transformation in EasternEurope and the Distribution of Income, Cambridge: Cambridge University Press.

Atkinson, Anthony B., Lee Rainwater and Timothy Smeeding (1995), ‘Income Distributionin Advanced Economies: Evidence from the Luxembourg Income Study’, LuxembourgIncome Study (LIS) Working paper No. 12, Luxembourg, October.

Blanchard, Olivier and J. Keeling (1996), ‘A Numerical Model of Transition’, HarvardUniversity, Cambridge, Mass., mimeo, May 30, 1996, p. 20.

Commander, Simon, Andrei Tolstopiatenko and Ruslan Yemtsov (1997), ‘Channels ofRedistribution: Inequality and Poverty in the Russian Transition’, paper prepared forthe EBRD Conference on Inequality and Poverty in Transition Economies held inLondon on May 23–24.

Commander, Simon and Une Lee (1998), ‘How Does Public Policy Affect the IncomeDistribution? Evidence from Russia, 1992–1996’, August, EBRD, London, mimeo.

Cornia, Andrea Giovanni (1994), ‘Income Distribution, Poverty and Welfare inTransitional Economies: A Comparison between Eastern Europe and China’, InnocentiOccasional Paper, Economic Policy Series No.44, Florence, Italy: UNICEF.

Coulter, Fionna A. E., Frank A. Cowell and Stephen P. Jenkins (1992), ‘Equivalence ScaleRelativities and the Extent of Inequality and Poverty’, The Economic Journal, 102,pp. 1,067–82.

Forster, Michael F. (1993), ‘Comparing Poverty in 13 OECD Countries-Traditional andSynthetic Approaches’, , Luxembourg Income Study (LIS) Working paper No. 103,August.

Kordos, Jan (1994), ‘Outline of the Current Budget Surveys in Poland’, Warsaw, Poland,mimeo.

Kupa, Mihaly and Gaspar Fajth (1990), ‘Hungarian Social Policy Systems and Distributionof Income of Households’, Hungarian Household Panel Survey (TARKI) Budapest,mimeo.

Marnie, Sheila and John Micklewright (1993), ‘Poverty in Pre–Reform Uzbekistan: WhatDo Official Data Really Reveal’, European University Institute, Florence, Italy, mimeo.

Michel, Richard C. (1991), ‘Economic Growth and Income Equality since the 1982Recession’, Journal of Policy Analysis and Management, 6(2), pp. 181–203.

Milanovic, Branko (1992), ‘Income Distribution in Late Socialism: Poland, Hungary,Czechoslovakia, Yugoslavia and Bulgaria Compared’, Socialist Economies ReformUnit, Income Distribution during the Transition, paper No. 1, Washington D.C.: TheWorld Bank.

Milanovic, Branko (1998), Income, Inequality and Poverty during the Transition from Planned toMarket Economy, Washington, D.C.: The World Bank.

Popkin, Barry (1993), ‘Poverty in the Russian Federation: Demographics and Coverage byCurrent Support System’, Carolina Population Center, University of North Carolina,North Carolina, Chapel Hill, mimeo, September 23, 1992.

Redor, Dominique (1992; first published in French 1988), Wage Inequalities in East and West,Cambridge: Cambridge University Press.

THE GROWTH OF INEQUALITY 333

Rutkowski, Jan (1995), ‘Labour Markets and Poverty in Bulgaria’, A background paper forthe World Bank poverty assessment, World Bank, Washington D.C., mimeo, October.

Rutkowski, Jan (1996), ‘High Skills Pay off: The Changing Wage Structure duringEconomic Transition in Poland’, Economics of Transition, 4(1), pp. 89–112.

Stanovnik, Tine and Nada Stropnik (1998), ‘Impact of Social Transfers on Poverty andIncome Inequality in Slovenia: A Comparison between the Pre-Transition and the Post-Transition Period’, SOCO Program, Ljubljana, September 1998.

Vegernik, Jiri (1986), ‘Mzdova a prijmova diferencijace: analyza nekterych problemu’,Institute for Philosophy and Sociology, Prague, mimeo.

Vegernik, Jiri (1993), ‘Recent Changes in Household Income Distribution: the CzechRepublic 1988–1993’, Czech Academy of Sciences, Prague, mimeo.

Vegernik, Jiri (1994), ‘Changing Earnings Inequality under the Economic Transformation:The Czech and Slovak Republics in 1984–92’, Czech Academy of Sciences, Prague,mimeo.

Vodopivec, Milan and Peter Orazem (1995), ‘Winners and Losers in Transition: Returns toEducation, Experience and Gender in Slovenia’, The World Bank Economic Review, 9(2),May.

MILANOVIC334

Appendix 1Bulgaria, 1989–95

Composition of income (in per cent)

Concentration coefficients (in per cent)

-20.0

0.0

20.0

40.0

60.0

1989 1990 1991 1992 1993 1994 1995

Labour

Private

Pensions

Other transfers

0.0

20.0

40.0

60.0

1989 1990 1991 1992 1993 1994 1995

(per cent)

Labour

Private

Pensions

Other transfers

THE GROWTH OF INEQUALITY 335

Hungary, 1987–93

Composition of income (in per cent)

Concentration of coefficients (in per cent)

-40.0

-20.0

0.0

20.0

40.0

60.0

1987 1989 1993

Labour

Private

Pensions

Other transfers

0.0

20.0

40.0

60.0

1987 1989 1993

(in per cent)

Labour

Private

Pensions

Other transfers

MILANOVIC336

Poland, 1987–95

Composition of income (in per cent)

Concentration coefficients (in per cent)

0.0

20.0

40.0

60.0

1987 1988 1989 1990 1991 1992 1993 1994 1995

Labour

Private

Pensions

Other transfers

-2 0 .0

0 .0

2 0.0

4 0.0

6 0.0

1 98 7 1 98 8 1 98 9 1 99 0 1 99 1 1 99 2 1 99 3 1 99 4 1 99 5

P riv ate

L a bo u r

P e n sio n s

O th er tra n sfe rs

THE GROWTH OF INEQUALITY 337

Slovenia, 1987–95

Composition of income (in per cent)

Concentration coefficients (in per cent)

0.0

20.0

40.0

60.0

80.0

1987 1988 1989 1990 1991 1992 1993 1994 1995

Labour

Private

Pensions

Other transfers

-40.0

-20.0

0.0

20.0

40.0

1987 1988 1989 1990 1991 1992 1993 1994 1995

Labour

Private

Pensions

Other transfers

MILANOVIC338

Russia, 1989–96

Composition of income (in per cent)

Concentration coefficients (in per cent)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

1989 1992 1994 1995 1996

Labour

Pensions

Other transfers

Private

-40.0

-20.0

0.0

20.0

40.0

60.0

80.0

1989 1992 1994 1995 1996

Labour

Private

Pensions

Other transfers

THE GROWTH OF INEQUALITY 339

Latvia, 1989–96

Composition of income (in per cent)

Concentration coefficients (in per cent)

0.0

20.0

40.0

60.0

80.0

100.0

1989 1992 1993 Q3:95 Q4:96

Labour

Private

Pensions

Other transfers

-20.0

0.0

20.0

40.0

60.0

1989 1992 1993 Q3:95 Q4:96

Labour

Private

Pensions

Other transfers

MILANOVIC340

Appendix 2

Decomposition of annual changes in the Gini coefficient

BULGARIA

base year 1989 1989 1990 1991 1992 1993 1994 1995

Shares 0.5 1.3 1.7 1.7 3.4 1.6Wages 0.6 –1.9 5.7 7.1 4.3 7.8Non-wage private 0.9 1.7 2.3 2.9 3.4 –0.4Social transfers –1.1 –0.4 –1.5 –0.0 0.1 0.8 Pensions –1.1 –0.3 –1.7 –0.4 –0.2 0.4 Non-pension 0.0 0.0 0.2 0.4 0.2 0.4Total 0.9 0.8 8.2 11.7 11.1 9.7Residual 0.2 1.1 0.1 –0.1 1.4 0.3

Gini (level) 21.7 22.8 23.6 30.0 33.3 34.2 31.7

POLAND

base year 1987 1987 1988 1989 1990 1991 1992 1993 1994 1995

Shares 0.0 –1.1 –0.8 –2.6 –3.3 –2.3 –1.5 –1.3Wages –0.0 3.0 2.2 2.3 3.9 4.1 8.0 7.9Non–wage private 1.3 2.1 –0.4 –3.3 –3.8 –2.7 1.3 0.6Social transfers –1.9 –3.0 –0.3 1.9 1.5 3.6 3.5 3.3 Pensions –1.6 –3.2 –0.4 1.9 1.8 3.9 3.2 3.4 Non–pension –0.3 0.2 0.1 0.0 –0.3 –0.3 0.3 –0.1Total –0.6 1.0 0.8 –1.6 –1.8 2.7 11.3 10.7Residual 0.2 0.2 –0.1 1.5 2.3 2.3 –0.1 –0.1

Gini (level) 25.0 24.4 26.0 25.5 24.7 25.5 29.8 36.2 35.6

SLOVENIA

base year 1987 1987 1988 1989 1990 1991 1992 1993 1994 1995

Shares –0.1 0.0 –0.1 –0.2 –0.1 –0.5 –0.5 –0.3Wages –0.7 3.5 –0.7 2.2 2.3 3.2 1.5 3.6Non–wage private 0.4 0.5 –0.2 1.7 1.6 2.2 1.5 0.4Social transfers –2.8 –2.3 1.4 –1.4 –0.7 –1.1 –0.6 –0.5 Pensions –2.4 –2.5 1.2 –1.2 –0.5 –0.8 –0.2 –0.1 Non–pension –0.3 0.2 0.2 –0.1 –0.2 –0.3 –0.4 –0.4Total –3.1 1.7 0.3 2.5 3.0 3.9 1.8 3.3Residual 0.7 1.3 –0.2 –0.2 –0.2 0.4 –0.1 –0.7

Gini (level) 19.8 17.4 22.9 19.9 22.1 22.6 24.1 21.5 22.3

THE GROWTH OF INEQUALITY 341

HUNGARY

base year 1987 1987 1989 1993

Shares 1.7 –2.7Wages 1.3 5.5Non–wage private 1.7 –0.6Social transfers –0.7 1.2 Pensions –0.9 1.4 Non–pension 0.2 –0.2Total 4.0 3.4Residual 0.2 –1.1

Gini (level) 20.7 24.8 22.9

LATVIA

base year 1989 1989 1992 1993 1995 1996

Shares –2.2 –2.7 –1.9 –1.8Wages 1.2 4.2 16.4 15.0Non–wage private 1.2 0.9 1.3 1.4Social transfers –2.1 –1.8 –2.6 –1.5 Pensions –2.2 –1.9 –2.9 –2.0 Non–pension 0.1 0.1 0.3 0.5Total –1.8 0.5 13.3 13.1Residual 2.7 0.7 –3.8 –3.1

Gini (level) 22.6 23.5 23.9 32.0 32.6

RUSSIA

base year 1989 1989 1992 1994 1995 1996

Shares –5.2 –5.1 –7.5 –6.2Wages 20.2 17.8 19.7 23.6Non–wage private 5.4 3.0 5.3 4.3Social transfers 7.0 4.3 5.2 6.0 Pensions 4.5 3.9 2.6 3.7 Non–pension 2.5 0.4 2.6 2.3Total 27.4 19.9 22.6 27.7Residual 5.1 3.7 3.5 2.2

Gini (level) 21.9 54.3 45.5 48.0 51.8