_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ V~~~~~~~\A/PS 20 7L POLICY RESEARCH WORKING PAPER 2077 Change in the Perception Russia experienced a precipitous drop in real of the Poverty Line during income from March 1993 to Times of Depression September 1996. As the percentage of the *" objectively" poor (those with Russia 1993-96 income below the official poverty line) increased, the percentage of the Branko Milanovic "asubjectively" poor (those Branko Jovanovic who felt poor) decreased. Perception of the subjective poverty line went down even faster than real incomes. The World Bank Development Research Group Poverty and Human Resources March 1999 POLICY RESEARCH WORKING PAPER 2077 Summary findings During Russia's economic transition real income declined situation, the percentage of the subjectively poor precipitously for most of the popUlation. How were decreased more or less in step with a reduction in Russians' perceptions of the minimum income level people's real income. Only larger-than-usual income needed to survive affected by such a rapid decline in decreases were needed to jolt the population - that is, their incomes? to keep the percentage of the subjectively poor Based on data collected from repeated surveys of unchanged. individuals during the period from March 1993 to The percentage of the self-assessed poor was always September 1996, Milanovic and Jovanovic find that the lower than the percentage of the poor according to the subjective estimate of that minimum income for an adult "social" subjective poverty line. This suggests that Russian decreased by about 1.7 percent each month. pockets of the population regarded their own income as This sharp reduction in the subjective poverty line adequate although in the public perception they were meant that proportionately fewer people felt poor. poor. However at all times at least 60 percent of the This in turn suggests two mechanisms for adapting to population considered itself poor. worsening circumstances: 1) a reduction in what people In other words, the percentage of the "subjectively perceive to be the minimum income needed for survival poor" tended to decline as the perception of the needed and 2) the existence in the population of pockets of minimumn was reduced. In this somewhat unusual people who demand even less than others. This paper - a product of Poverty and Human Resources, Development Research Group - is part of a larger effort in the group to study the social effects of transition to a market economy. The study was funded by the Bank's Research Support Budget under research project "Changing Ideas about Poverty in Russia" (RPO 681-42). Copies of this paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Criselda Argayoso, room MC3-568, telephone 202-473-3592, fax 202-522-1153, Internet address cargayosoaworldbank.org. Policy Research Working Papers are also posted on the Web at http://ww-w.worldbank.org/html/dec/Publications/Workpapers/home.html. Branko Milanovic may be contacted at bmilanovic@worldbank.org. March 1999. (32 pages) The Policy Research -Working Paper Series dbsseminates the findcngs of work in progress to encourage the exchange of ideas about develo p-mnet issues. An objective ouf the series is to get the findings ouit quickly, even if the presentations are less than fully polished. Tlle papers carry the names of the authors and sbouild be cited accordingly. Tke findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. Thev do not necessarily represent the view of the lX'orld Bank, its Execuitive Directors, or the countries they, represent. Produced by the Policv Research Dissemination Center CHANGE IN THE PERCEPTION OF THE POVERTY LINE DURING THE TIMES OF DEPRESSION: RUSSIA 1993-96 Branko Milanovic and Branko Jovanovic' JEL Classification Code: 132, P2 Key words: Russia, subjective poverty, transition Development Research Group, The World Bank, Washington, D.C. and Texas A&M University. Email: bmilanovic@worldbank.org. Comments from Jeanine Braithwaite, Christine Jones, and Misha Lokshin are gratefully acknowledged. The research was financed by the World Bank Research grant RPO 681-42. 1 Introduction In the course of its transition to a market system, the Russian economy has experienced a series of shocks. It has experienced a sharp fall in output: its 1997 GDP is almost a third less than it was ten years ago. It has suffered from rapid and continuing inflation: over the period under study here (March 1993 to September 1996), the price level increased 46 times. It has witnessed the appearance of open unemployment, affecting some 10 percent of the labor force by 1997. Real wages and pensions have declined by a half compared to their pre-transition levels, and delays in their payment have become endemic. A few, who have either been enterprising and lucky, or politically well-connected, have amassed considerable fortunes. As a consequence, income inequality has increased by an unprecedented speed (the Gini coefficient has risen four to five times faster than it has during the 1980's in the United States2), as has the number of families living in poverty.3 Against the background of such rapid, and economically, generally unfavorable developments, the population's views about what constitutes poverty, and what it considers to be a minimum income "needed to make ends meet", must have evolved as well. Because the decline in income was sharp, it enables one to see, within a very compressed time period, how conceptions of wellbeing and deprivation respond to abrupt changes in income. For most people in most countries, these factors remain relatively constant over considerable periods of time. It is therefore difficult to observe the impact of changes in external circumstances on the formation of attitudes or expectations. The Russian experience allows us to explore the impact of abrupt changes in circumstances. In addition, the question of what the population views as a minimum acceptable income has obvious political implications: if most of the population feels poor, it is unlikely to support the reforms. This paper will explore how the perception of the poverty line, among the population as a whole, has changed in Russia over the period 1993-96. Section 1. The Model: Estimating the Subjective Poverty Line In the literature on the subjective-welfare estimation the usual specification defines the minimum income necessary for a family (MYf) to make ends meet as a dependent variable,4 and, in its most parsimonious formulation, total household income (Yf) and family size (n), as explanatory variables (e.g. Hagenaars and van Praag, 1985; van Praag and Van der Saar, 1989). InMYf = fct(lnY,f,lnn) (1) 2 See Milanovic (1998, p. 40ff). 3See, for example, Braithwaite (1997), Glinskaya and Braithwaite (forthcoming), Milanovic (1998), Lokshin and Popkin (1998), Ovcharova, Turuntsev and Korchagina (1997). 4MYf may be considered a point on a household cost function related to a specific welfare level umin. 2 The minimum income necessary for a family to make ends meet is obtained from the so-called Minimum Income Question (MIQ) such as "what do you consider as an absolute minimum net income (per period of time) for a household such as yours?" (see Flik and van Praag, 1991, p.320). Obviously, the family size influences positively the minimum income or its "subjective poverty line" (SPL)-the terms will be used interchangeably. In addition, the actual level of family income, which may be regarded as a proxy for family's "perrnanent income", influences positively SPL. The rationale is that families accustomed to a higher standard of living will, everything else being the same, have higher aspirations and hence higher estimate of what "their" minimum income is. This was termed by van Praag (1971), "the preference drift", and its value, in a double-log formulation such as (1), will lie between 0 and 1. If the preference drift equals 0, then the subjective poverty line becomes an absolute poverty line. At the other extreme, when the preference drift is 1, every increase in real income "exacts" the same percentage increase in what is perceived to be the poverty line. The poverty line then becomes fully relative. Not surprisingly, most research has yielded the values of the parameter drift between 0.4 and 0.7 (see, e.g. Flik and van Praag, 1991, p. 325; van Praag and Flik, 1992, p. 10) which accords well with our intuitive perception that as people get richer they set the necessary minimum higher, but do not raise it (in percentage terms) as much as their income goes up. Answers to the MIQ will yield a number of observations such as in Figure 1. We fit the regression based on these observations, and the intersection of the regression and actual income Yf (see point A in Figure 1), is defined as the "social" subjective poverty line. To see why this is so, notice that households to the left of A have an income that is below the regression line (that is, less than "society" deems needed). They are considered poor. On the other hand, all those lying to the right of A are not "socially" considered to be poor since their actual income is above the regression line-even if they may consider themselves to be poor (e.g. their required minimum income may lie at C, much above their own income). If we then write out (1) in log-linear form, ln M7l§ = ,8o +±3Iln Y1f +,B2 lnn and let MYf-=Yf, (I -/,i) ln Y, = o± +,32 In n (2) The elasticity of family size with respect to subjective poverty line (i.e. parameter 0 in the expression of the equivalent income, Y/n°) becomes 8i 1-/3 3 Figure 1. Determining the social poverty line Minimum income needed 45 degree line 0 ° / ° Regression 0/ 0 0 0. 0 B Income The VCIOM (All-Russian Center for Public Opinion Research) data set available to us (see the discussion of the data set in Section 2), however, does not contain precisely the MIQ as explained above. Instead of asking the household head for his/her opinion on the minimum income for the entire family, the enumerator asks the question, "What income, in your opinion, constitutes the subsistence minimum per person at the present time?" This is a very general minimum income question, asking in effect for the household's view as to the minimum income for an adult (since person is likely to be interpreted as an adult person)-not what would be the minimum income per person for that family.5 This problem does not allow us to apply the theory sketched above in a straightforward fashion, but to use an alternative approach. Equation (3) shows the effective formulation based on the question as asked. We also introduce other control variables that may be relevant. 5If the latter were the case, the problem could be easily solved. If respondents are rational, there is no difference between asking them what is the minimum total income for their family and the minimum per capita income for theirfamily. The answers to the latter could then simply be multiplied by the number of family members to obtain the minimumfamily income. 4 ln AMY = fct(ln Y* ,age, age2, SETTLEMENT, REGION,time) (3) AMY represents answers to the minimum income for an adult question, Y* the "true" income level of the household (i.e. income per equivalent adult of a household); age of the respondent; size of the settlement, and region where the family lives.6 The crucial is "true" income variable. In trying to find out how people perceive, depending on their income, what is the minimum income for an adult in Russia, we have to find income Y* such that it accurately reflects household's economic welfare. This clearly is unlikely to be total family income since it does not take into account the number of people who share it. It could be a per capita income, or an income per equivalent adult which accounts for economies of size. Therefore Y* is defined as Y/n6 where 0 is a parameter for economies of size ranging from 0 (full economies of size) to 1 (no economies of size or per capita measurement). The problem is, of course, how to determine the right 0. We argue that the right 0 (0*) will be the one which would make the sign of the household size (n) variable introduced as an additional control in (3) statistically not significantly different from zero. The rationale is as follows. Once we identify the "true" household income, there is no reason why household's size or family composition will matter at all for what people regard as the minimum income for an adult in Russia. We shall therefore try different values of Y(0*), and choose the 0=0* that makes the coefficient on ln n in equation (3) equal to zero. Note also that for values 0<0*, we expect the coefficient on ln n to be negative, because large households' economic welfare is overestimated (they are not as rich as they seem). Their estimate of the minimum income (AMY) is therefore systematically biased downward, which in turn leads to a negative correlation between AMY and ln n and a negative regression coefficient. For values 0>0*, the opposite is true and we expect the regression coefficient to be positive (see Figure 4 below). Including the age and age2 variables accounts for the life cycle (parabolic) effect whereby the perceived needs increase until they reach a peak, and decrease thereafter. Since this variable captures the age of the respondent (not necessarily the age of the household head; see below Section 2), one must be careful with its interpretation. We capture the importance of the environment on the perception of poverty line by introducing the variables for the size of settlement and regional location. People living in big cities or richer regions (e.g. Moscow, St. Petersburg) will face higher prices and would be expected to pitch their poverty line higher. 7 The social reference (demonstration) effect may also be important in larger cities, as people seeing the wealth of others come to expect more. Living in a harsh climate might also increase one's perception of the necessary minimum income. 6 The dummy variables in (3) are written in upper case. 7Our data base is not deflated for regional price differences since regional CPIs are not available. 5 Finally, we introduce time in our model in order to capture the change in the perception of the poverty line over time. Our hypothesis is that the subjective poverty line will decrease as time passes on and people adapt to the new and worse conditions, and adjust their expectations accordingly. The period covered by our data spans 31/2 years, from March 1993 to September 1996, during which time the Russian population experienced a severe decline in real income. The decline is estimated at 14 percent based on our Survey results or almost 20 percent based on official (Goskomstat) monthly estimates of population income over the same time period (see Figure 2). 8 The question we ask is whether, in addition to the income effect, the mere passage of time, and realization of seemingly ever worsening circumstances, will lead the public to scale down its expectation of what the minimum "tolerable" income is. As the adaptation to the less fortunate circumstances proceeds, we would expect that the time variable will enter negatively in equation (3). Figure 2. Real population income, Q1/1993 to Q3/1996 (in constant March 1993 roubles; per capita; per month) 40.... . ................. 35E 30 o;Ks ~~~Official per capita ,I - incomei 25 L-_ n 20 - 0 1 5 Survey per capita 10 5. 1993 1994 1995 1996 0 1 5 9 13 quarters Sources: Survey per capita income calculated from VCIOM surveys. Official per capita income from monthly Goskomstat statistics. 8 Figure 2 also allows us to note that VCIOM Survey underestimates incomes by about 40 percent compared to the official data, but also that the underestimation diminishes with time. Most of the difference is due to the omission of income in kind from the Survey data. 6 Section 2. The Data We use the twenty-nine cross-sectional VCIOM (All-Russian Centre for Public Opinion Research-VCIOM in its Russian abbreviation) data sets covering the period from March 1993 to September 1996. The survey is a representative sample of Russian households conducted monthly (between March 1993 and January 1994) and approximately every second month since. Although most of the questions in the Surveys are concerned with the household (family), there were questions that targeted individuals. These variables include, among others, gender, age and education. In most surveys, such questions are targeted specifically to the head of the household. Here, however, the respondent is not necessarily the household head. The fact that the respondent need not be the household head might have an adverse effect on the accuracy of some data (for example, the respondent may not be fully aware of all the components of household income). The original data set consisted of 91,090 observations spread over 29 cross sections. The number of observations was reduced to 80,826 after omitting the observations that did not contain information on family income (total or by components). When the incomplete observations were omitted, individual cross sections contained between 3,626 (January 1994) and 2,034 (September 1996) observations. Although the reduction of the sample size over time was considerable it did not, according to the VCIOM staff, affect the representativity of the sample. The sampling procedures used were improved.9 The basic characteristics of the households and respondents surveyed are given in the Annex 2. The total family income variable was computed as a sum of income components: main income and income from the second job, income from private sector activities, pensions, other social transfers (family allowances, unemployment benefits, sickness benefits etc), stipends, alimonies, income from financial papers, income from sale of self-produced goods, and other monetary incomes."0 The all Russia monthly CPI, with March 1993 as a base, was used to deflate all the monetary variables. We assume that the inflation affects all regions equally since regional CPIs are not available. In real terms, the subjective minimum income for an adult (AMY) decreased dramatically between March 1993 and September 1996. It started by being higher than Rs. 35,000 in the early surveys and ended with Rs. 15,000 (see Figure 3).li The Ministry of Labor official minimum income for an adult (the prozhitochnyi minimum) remained constant in real terms at some Rs. 10,000.12 The gap between the two therefore steadily 91 owe this information to Jeanine Braithwaite. 10 For details on how total family income variable was constructed see Annex 1. 1 Calculated as an individual-weighted average of AMY's over all households See Annex 3 for more details. 12The official minimum is composed of a given bundle of food and non-food goods. Its slight oscillations around Rs. 10000 at March 1993 prices are due to the fact that the CPI that we use to deflate the 7 diminished. At the beginning of the period, the average subjective minimum income for an adult was 3 and 1/2 and even four times higher than the official minimum; at the end of the period, the ratio was 1.7. The public perception of the minimum income for an adult Russian thus gradually became closer to the official minimum. Figure 3. The average subjective poverty line for an adult, the official poverty minimum for an adult, and average per capita income (per month; in March 1993 roubles) 40 35 A 30 um _ Su~~~~~~~~~~~bjective poverty line~ per adult-I 225 _A \_ *20 15 V Official minimum income for > _ . Inco~~~me per capita .'an adult 10 , ,__ , , , -, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 Survey number Note: Number of survey given on the horizontal axis. Survey years shown immediately above. The average subjective poverty line for an adult is the simple individual-weighted average of poverty lines (AMY) in the surveys; the official poverty line for an adult is MinTruda Rossii all-Russia official poverty line; the average per capita income is the average income from VCIOM surveys (same as in Figure 2). The composition of the households, as well as the demographic characteristics of the respondents, stayed roughly the same over time (see Annex 2 and Annex 3). The average household, over the entire survey period, consisted of 3.1 members, with 0.7 children. For comparison, according to the all-Russia official statistics for 1994, the average household size was 2.84 members (Goskomstat Rossii, 1995, p.28). The average age of the Survey respondent was 42.7 years, and he/she spent 11.2 years in school."3 59.2 percent of nominal monthly values of the official minimum might have at times increased faster or slower than the cost of the minimum bundle of goods. 3 The average duration of schooling of the population over 15 years of age calculated from the 1993 Russian Living Standard Monitoring Survey (RLMS) is a little over 9 years. 8 the respondents were women; according to the official 1995 statistics, women accounted for 53 percent of the Russian population. Most of the respondents (76.5 percent) lived in urban areas, a percentage quite close to the official 1995 statistics (73 percent). A plurality of respondents (46 percent) lived in cities with the population less than 100,000, followed by 25.1 percent of those who lived in cities with the population over a million. 14 4 In order to indicate the existence of possible outliers in the data, and especially in the variables total family income (Yf), and AMY, we create "flag" variables out], out2 and out3. All three variables were computed for each cross section separately. Variable out] takes value I if variables Yf and AAY both exceed mean plus 5 standard deviations, and zero otherwise. Using this rule the total of 701 observation (0.87 percent of the entire sample) was flagged as outliers. Variable out2 is equal to 1 whenever variables Yf and AMYf both exceed the corresponding upper confidence limit of 99.5 percentile, and zero otherwise. The total of 350 observation (0.44 percent of the entire sample) was flagged as outliers. Finally, we use the method developed by Hadi (1992; 1994) (using the hadimvo procedure in STATA) to compute the flag variable out3. The total of 2,634 observations is identified as possible outliers, which represents 3.28 percent of the total sample. While the results turned to be robust for the exclusion of the observations (households) flagged by out] and out2, they are somewhat sensitive when the observations flagged by out3 are excluded. 9 Section 3. Estimating the Subjective Poverty Line in Russia We first try to estimate the "true" household income using different values of 0. To do so, we run the basic model (3) including in addition to the variables shown there In n, as the control variable for household size. Figure 4 shows how the coefficient on In n changes as 0 in Y(0)* varies from 0 to 1. For 0=0.62, the coefficient becomes equal to zero.15 Figure 4. Coefficient of In n as function of the economies of size parameter 0 Da .. ....... ........... .. . .. ... . ... ....... . . .. .. .......... .. ..... . . .... .. . . .... . .. ... .. S I 01 02 03 04 0 0.S 07 0o 0s -0 12 ... . . ......... . .. . -.... ... .- s....... Tht,. We then move to the direct estimation of equation (3) using Y*= 0 The results are shown in Table 1. All the regressions are run with Huber (robust) variances to adjust for the fact that the observations are drawn from different time-clusters (that is, from the 29 surveys) and that variability of observations within each survey is less than if all observations were drawn at random from the population at large. 16 Nevertheless, as the t-values in Table 1 show, practically all the coefficients are significant at probability far greater than 99 percent. Elasticity of the subjective poverty line for an adult with respect to income is 0.144 for the overall sample and 0.132 in a regression that excludes Hadi outliers. 17 This It is statistically insignificantly different from zero for a few other values around 0.6, but takes its lowest values for 0=0.62. 16 That is, the variability of observations from the pooled cross-sections is less than if all 80,000 of our observations were drawn from one cross-section. 17 As mentioned before, the exclusion of other outliers (out] and out2) does not have an effect on the results and is not shown here. 10 is a significantly lower value than reported by Frijters and van Praag (1994) in their study of the former Soviet Union and Russia. For Russia in 1993 and 1994, Frijters and van Praag, report the preference drift value of 0.62 and 0.64 respectively (somewhat higher than the value they find for the Soviet Union in 1991: 0.41). However, these results are not entirely comparable. Frijters and van Praag used a variant of the so-called income evaluation question (IEQ)'8 in order to obtain the left-hand side variable ("the Leyden poverty line"), while in our case we have as the dependent variable what people consider to be a minimum income for an adult. Our preference drift is also significantly lower than the value found in some Westem countries. Flik and van Praag (1991, p.327), for example, report for the Netherlands a preference drift of 0.59; Hagenaars and van Praag (1985, p.151) report for a collection of West European countries a coefficient of about 0.54. A part of the difference may be due to a "richer" choice of control variables included here (regional and size of settlement dummies) as well to the introduction the time variable. In effect, if we run a very parsimonious formulation such as (1),"9 which is basically what Hagenaars and van Praag (1985) do, the preference drift increases from 0.14 to 0.23. This latter value (0.23) is almost identical to the preference drift obtained by Ravallion and Lokshin (1998, p.30). They use what they dub the Economic Ladder Question (ELQ) whereby individuals rank their own subjective level of living going from 1 (the poorest) to 9 (the richest). The rankings are, like in the rest of the subjective poverty literature, explained by the underlying differences in real income. The sample they use is a representative sample of the Russian population in 1996 and is obtained from a different survey than ours (the Russian Longitudinal Monitoring Survey). The fact that two independent studies, using two different surveys, come with very low values of the preference drift for Russia requires explanation. There are, we believe, two possible explanations. The first is the difference in formulations. It is possible to imagine that people's views will vary less with income when they are asked what they consider to be a minimum income for an adult in general (as in VCIOM survey) than when they are asked what is the minimum income for their own family. In the latter case, poor people may pitch their minimum fairly low, while the rich may find it hard to imagine living without a relatively high income. For an abstract (adult) individual 1 Under Income Evaluation Question methodology, a respondent is asked to write down what level of income his/her family would consider to be "very bad", "bad", "middling", "good" and "very good". The mean of the five answers is defined as the Leyden poverty line. (More on the methodology, see Hagenaars and van Praag, 1985; Flik and van Praag 1991). In addition, Frijters and van Praag (1994) upgraded the reported family incomes in order to take into account unofficial sources of income, again based on respondents' subjective perception of importance of informal income sources. Finally, a very high number of households (e.g. 2,668 out of 8,979 in the year 1991) were simply deleted from the sample due to missing observations, and the rest of the sample was reweighted (although the details are not given). It would thus appear that, possibly because the quality of the survey was wanting, a large degree of ad-hoc adjustments was made. 9 Without a control for household size, for the reasons explained above. 11 in a country, their opinions may not be so far apart. Second, a low preference drift may also suggest a relative homogeneity of people's perceptions, as people on the top of the income scale do not evaluate the minimum income needed "to make ends meet" so much higher than the poor. The homogeneity, in turn, can be explained by the relatively recent "explosion" of income inequality which means that people who had more or less same incomes only recently will not suddenly diverge very much in their perception of the poverty line. Clearly, in countries (as in Western Europe) where income differences have historically been greater and where income mobility was less (in the sense that people with current high incomes probably had high incomes five or ten years ago), the perception of the poverty line may differ significantly between the rich and the poor. But in a country, like Russia, which, until recently was very egalitarian, and was then subjected to an almost random and huge income shock, which made some people's income increase manifold and other people's incomes drop significantly, perceptions of the minimum income would still be relatively similar. The economies of scale parameter (0) is, as mentioned before, 0.62. This result too is in sharp contrast with Frijters and van Praag (1994) finding. They report elasticity of the farnily poverty line with respect to household size to be 0.2. They claim that this low value (compared to Western Europe) "reflect[s] the relatively cheap and good facilities for child care in the USSR, still existing in 1991" (1994, p.10). However, they do not mention that 0 is composed of two elements. One is economies of size, that is, how minimum needs increase with the number of household members (regardless of whether they are children or adults). The second element is the cost of children. Now, while the low cost of child care might have pushed 0 down under the socialist regime, the first, and a more important element, economies of size, pushed it up (see Lanjouw, Milanovic, and Paternostro, 1998). This is because public or semi-public goods, like utilities, rent etc. were cheap relative to private goods. The share of spending on public goods was routinely much below the corresponding values in market economies (e.g. rent spending accounted for a few percents of total expenditures, while its share in market economies is 15-20 percent). As shown in Dreze and Srinivasan (1995, p.27), the parameter for the economies of scale is bounded from above by the share of spending on private goods. Since this share was high in socialist economies, so must have 0. This can be understood intuitively too: if public goods are practically free and households spend their income only on food, there would scarcely be any economies of size. Thus, Frijters and van Praag (1994) contention that a low 0 in the USSR is appropriate is wrong. However, again, our economies of scale parameter is not much different from the one reported in Ravallion and Lokshin (1998). In the already mentioned study, they find the economies of scale parameter to be 0.42 (Ravallion and Lokshin, 1988, p.30) with a standard error of 0.148. 12 Table 1. Regression results Dependent variable:ln subjective minimum income for an adult (AMY) (1) Basic equation (2) = (1) without (3) = (2) with Gini with Huber Hadi outliers coefficient (robust) variances Ln equivalent 0.144 0.132 0.132 income (Y*) li (24.0) (22.3) (22.3) Age 0.016 0.017 0.017 (14.0) (17.1) (17.1) Age2 -0.0002 -0.0002 -0.0002 (-18.9) (-20.5) (-20.6) Small towns and -0.062 -0.065 -0.066 villages (popul. (-4.2) (-4.3) (-4.4) Under 100,000) Towns (between 0.064 0.056 0.056 100,000 and 1/2 (4.0) (3.4) (3.4) million) Medium size cities 0.058 0.059 0.0.58 (between 1/2 and 1 (4.7) (4.8) (4.9) million) Northern region -0.243 -0.220 -0.223 (-12.0) (-11.7) (-10.7) Central and Black -0.330 -0.307 -0.311 Earth (-13.9) (-13.0) (-10.9) North Caucasus -0.225 -0.210 -0.210 (-5.7) (-5.6) (-5.5) Volga-Vyatka -0.324 -0.291 -0.295 _ (-13.1) (-12.3) (-10.6) Volga -0.256 -0.236 -0.239 (-6.9) (-6.9) (-7.0) Urals -0.194 -0.179 -0.182 (-7.8) (-7.5) (-6.7) West Siberia -0.150 -0.131 -0.132 (-6.4) (-6.0) (-5.6) East Siberia and 0.035 0.030** 0.028** Far East (1.6) (1.4) (1.3) Time -0.017 -0.017 -0.017 (-17.7) (-18.0) (-17.3) _ Regional Gini -0.054** coefficient (by (-0.4) survey) Constant 2.889 2.822 2.847 (82.7) (97.8) (40.0) Sample size 79,595 76,965 76,965 R2(adjusted) 0.189 0.191 0.191 F value 210.6 246.9 243.4 Note: t-values given in parentheses (under the coefficients). All coefficients are significant at the 1 percent level, except those with * which are significant at the 5 percent level, and ** =not significant. For size of settlement, the omitted category is larger cities (population over 1 million). For the regions, the omitted variable is Moscow-city. I/Defined as YIN062. 13 The parabolic age effect implies that the subjective poverty line rises with age until a certain point, after which needs decrease. The peak obtains at around 40 years of age, some 4 1/2 years later than reported by Frijters and van Praag (1994, p. 11). However, since the variable captures respondent's age, it may not be representative of the household age composition. The dummy variables adjust for the size of the settlement where the family lives, and region. For the size of settlement, the omitted category is larger cities (with over 1 million population). The subjective poverty line is lower in small towns and villages. Surprisingly, the perceived minimum income for an adult is higher in towns and medium size cities than in the very large metropolitan areas. We would expect that the "needs" increase monotonically with the size of settlement, be it because the cost of living is higher or the demonstration effect is greater. The absence of this regularity for the large metropolitan areas may be due to the fact that some of the effect is picked up by the regional variables. For the regional variables, Moscow-city is the omitted category. Of course, subjective needs in all other regions except East Siberia and Far East are less than in the city of Moscow.20 Compared to Moscow, the subjective poverty line is lower (under ceteris paribus conditions) by between 13 percent in West Siberia, and 30 percent in Central and Black Earth and Volga-Vyatka regions. In East Siberia and Far East, the subjective needs are about the same as in Moscow. High poverty line in East Siberia and Far East is explicable by the harshness of the climate (which requires higher housing and energy expenditures) and its remoteness which means that prices of consumption goods are higher. We discuss the difference between the regional subjective and official poverty lines in Section 4 below. The variable time, measured in months with March 1993 as a starting point, shows how the subjective poverty line for an adult has changed through downscaling of people's 21 expectation. In principle, we would expect this effect to operate through the income variable-lower income would, through preference drift, reduce the subjective poverty line. But in conditions of a rapid decline in real income as in Russia 1993-96, expectations are apparently downscaled even faster. Thus passage of each month (after March 1993) reduced the subjective poverty line by 1.7 percent. After more than three years of depression (by the Fall of 1996), the public's perception of the minimum per 20 Moscow-city does not include the Moscow region, which is a part of the Central and Central Black Earth region. 21 An alternative formulation is to use survey dummies. The results are given in Annex 5. Up to August 1993 (survey no. 6), the coefficients are positive, indicating an increasing subjective poverty line. Then for a few months they are not significantly different from zero before turning consistently negative, suggesting a decreasing subjective poverty line as time goes on. 14 capita income was about 1/2 of what it would have been with the same real income in the beginning of the period (Spring of 1993).22 Finally, in variant 3 (Table 1), we introduce a measure of income inequality (regional Gini coefficient) to account for a possible increase in the subjective poverty line due to higher inequality-an influence found in Hagenaars and van Praag (1985) and explained by the demonstration effect (greater inequality and therefore presence of higher incomes invites people to pitch their poverty lines higher). We calculate the Gini coefficient for per capita income for each region and for each survey (see Annex 3), and include it in the regression. However, we find no evidence that inequality influences the subjective poverty line. 22 We introduced (time)2 variable to check if the time effect may be subsiding as surveys progressed. It was found not significantly different from zero. 15 Section 4. Comparison of regional "subjective" and official poverty lines We have already seen (see figure 3) that the subjective poverty line for an adult was several times higher than the official poverty line (prozhitochnyi minimum) for an adult although the gap between the two diminished. A different question is how the structure (rankings) of the regional-official and subjective-poverty lines differ. Table 2 shows the rouble amounts for the official and subjective regional poverty lines in 1996. As we would expect subjective poverty lines are always higher, but the extent of how much higher they are differs between the regions: the official poverty line is less than half of the official one in the North Caucasus, but is almost two-thirds of the subjective line in the North. Table 2. Official and subjective regional poverty lines in 1996 (in 000 of March 1993 roubles) Regions (2) (3) Subjective Ratio (2):(3) Official poverty line poverty line North 8.30 12.7 0.65 Central and Central Black Earth 6.55 11.3 0.58 North Caucasus 6.17 12.9 0.48 Volga-Vyatka 6.82 11.5 0.59 Volga 6.84 12.4 0.55 Urals 7.47 13.4 0.56 West Siberia 8.79 14.1 0.62 East Siberia and Far East 9.06 16.8 0.54 Moscow 10.3 16.3 0.63 Note: Official poverty lines calculated from Goskomstat Rossii (1997, Table 2.7 and Table 4.20). Subjective poverty lines calculated from variant 2 (Table I above). The subjective poverty lines cover the period January-September 1996. 16 The implication of these regional differences is that the official poverty lines do not accurately reflect population perception of the differences in subjective needs between the regions.23 Figure 5 shows that if Moscow-city poverty lines, both subjective and official, are set at 100, relative subjective poverty line for all but one region (Northern Russia), are higher than the official. This suggests a pro-Moscow bias in setting of the official poverty lines. For example, the official poverty line for an adult in the Caucasus is 40 percent below that of Moscow; but the public perception of the minimum there is that it should be only 20 percent below the Moscow subjective poverty line (Figure 5). Figure 5. Regional poverty lines: subjective and official (Moscow-city=100), year 1996 120 Subjective 100 - ----O fficial- -- - - - -- - - - - -- - -- - -- - - -- - - 40 -- - --- ---- -- - -------il -- --- - 40 -- 23 This despite the fact that the correlation coefficient between the official and subjective poverty lines is 0.85. 17 Section 5. How many people are poor? In this section, we look at the proportion of the poor where "poor" are defined according to three criteria. The first criterion defines as "subjectively poor" those households that are poor according to their own assessment, that is households whose view of the minimum income for an adult is greater than their actual adult equivalent income (AA4Yf>Y*r for a given family). This criterion leads to inconsistencies, in the sense that two identical households with the same incomes may be classified as respectively poor and non-poor depending on how they perceive own wellbeing. Furthermore, we impose a "social" equivalence scale (0=0.62) which may not correspond to the household own equivalence scale. This is why the second criterion, the "socially subjectively poor" is, as discussed in Section 1, used instead. This criterion defines as poor those households whose current income per equivalent adult (Y*f using 0=0.62) is less than the social subjective minimum income (per adult) for such a household predicted from the regression 3 (the variant with Huber-robust variances and excluding Hadi outliers). Finally, under the third criterion, the poor are those whose current income per equivalent adult (Y*f using 0=0.62) less than the official all-Russia poverty line (per working adult). Figure 6 shows the share of households who are poor according to the three criteria. We can make several conclusions. First, an extremely high percentage of the population (almost always greater than 60 percent) is subjectively poor-whatever (subjective) criterion is used. This is consistently higher than the percentage of the poor according to the "objective" criterion of the official poverty line. Second, there is a clear tendency for the "subjective" poverty headcounts to decrease with time. This is not surprising because we have already noted a sharp decrease in the subjective poverty line with the passage of time. As the subjective poverty line decreased faster than population real income, fewer people assessed themselves as poor. This explains how the percentage of the "socially" subjective poor went down from 90 percent of individuals in March 1993 to less than 60 percent at the time of the last Survey while real average per capita income decreased by 14 percent. 18 Figure 6. The share of the poor individuals in total population according to three concepts of poverty 1,0 .oo ................. ..... ---- -M--- ........... . . . .... ....... . .... ............ ... . . .. _ . . .. ..... ... .................... . ..... ......... . ...... . . ..... .. . ... . ....... 0~9 42 - ~~~~~~~~~~~Accordinig to "society" 0.80 X 0.70- l *' t According to their own i :E 0.60. ____, _ _ 0 u0.50 w According to oftidal povertyline */\ 0.40 -_ _ _ _ _ _ _ _ _ _ _ _ _ 030 e 0.20 0 5 10 15 20 25 30 35 40 Months (March 1993=1) Note: All shares are individual-based. Third, since real income declined while the official poverty line remained the same, the percentage of those with income less than the official poverty line increased from a third of the population in 1993 to more than one-half in 1995 before heading down toward 40 percent in the late 1996.24 It is thus somewhat ironic that while the decrease in real income has made more people poor according to an "objective" and fixed yardstick, the same reduction in real incomes has reduced people's perception of the minimum income they need in order to survive and has made fewer of them feel poor. This is why the decrease in the percentage of the self-assessed poor coincided with the decline in real income (Figure 6). The decline in the percentage of the self-assessed poor decelerated only between mid-1994 and mid-1995 (surveys 16 to 24) when the current real income took a further sharp dip: it required a larger than usual decrease in real income for the percentage of the self-assessed poor to stay constant. One could say that there are two ways to make fewer people feel poor: to augment their real incomes fast or to reduce their 24 This percentage may not be compared with the percentage of the poor from the official Goskomstat statistics that ranged between 22 and 31 percent over the same period (Goskomstat Rossii, 1998, p.79), or the percentage obtained from the Russian Living Standards Measurement Survey (see calculations by Jeanine Braithwaite in World Bank 1998, p. 5) because income in these cases is defined to include non-cash sources while VCIOM income includes only cash sources (see footnote 8 above). 19 incomes equally fast. In Russia, unfortunately, it was the second alternative that happened. Figure 6. Real per capita income (March 1996=100) and the percent of self-assessed poor 140 ..... .... . . ._. . . . ..... . . ... . -- . ....... .......... ..9 ........ .. 90.0 80.0 120asesetpo 70.0 100 V 60.0 0 0.0 50.0 80~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 50. 60 ...140.0 30.0 40 20.0 20 10.0 1993 1994 1995 1996 0 i 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 Survey number Note: real per capita income calculated from VCIOM surveys. Fourth, the use of the "social" subjective poverty line yields in all but two surveys higher poverty headcounts than own assessment (Figure 5). That means that some households that are "socially" considered poor, do not view themselves as such.25 This, in turn, indicates that there are households who are located in the triangle OAB (Figure 1). They are to the left of the social poverty line (AB), and are thus "socially" poor. However, their own assessment of minimum income is less than their actual income. This may indicate the presence of the much-discussed pockets of social resilience and patience that are often associated with the Russian population. 25 A caveat is in order there. Since we assume that all households have the social equivalence scale reflected in 0=0.62, it could well be that some households whom we classify as poor according to their own view, may in fact have a lower 6, and thus do not regard themselves as poor. The opposite classification mistake is possible with the non-poor households whose 0 is greater than 0.62. 20 Conclusions In the three-and-a-half years (March 1993 to September. 1996) covered by the VCIOM surveys of the Russian population, real per capita income decreased by between 15 and 20 percent. This came on top of severe income contraction in 1991 and 1992. Thus, the Russian population experienced one of severest peacetime depressions in the 20t century. At the same time, income inequality substantially increased. What happened, under these rather exceptional conditions, to the public perception of minimum income needed to "make ends meet"? We would expect that the subjective poverty line would decrease too. Indeed, the time variable was found significant as each month lowered the subjective poverty line by 1.7 percent. Thus, after more than three years of depression, the public's perception of a minimum income for an adult to survive was about 1/2 of what it would have been with the same real income in the beginning of the period. Yet the cross-sectional "preference (income) drift" parameter was relatively low, at slightly less than 0.15: each percent of real income decrease would, on average, reduce public perception of the poverty line by 0.15 percent. Some of the "sluggishness" is due to the inclusion of the time variable-a mere passage of time amid seemingly never improving circumstances led the population to downscale its expectations. However, even after dropping the time variable, "income drift" remains low (0.23) in comparison to West European countries, where it ranges between 0.4 and 0.7. This seems to suggest a relative homogeneity of people's perception of the subjective poverty line (for an adult Russian). Those on the top of the income scale do not evaluate the minimum income needed to survive so much differently than the poor. This is in turn explicable by either (or both) the poverty line question formulation that implicitly addressed the needs of an adult, or relatively recent "explosion" of income inequality. The question formulation might have influenced the answers in the sense that the rich and the poor individuals might differ less when asked to assess how much an abstract person needs in order to survive than when asked how much they themselves need. The recent increase in inequality might mean that people who had more or less same incomes until only recently will not suddenly diverge very much in their perception of the poverty line. We also find that subjective needs vary as function of the region. The poverty line is the highest in East Siberia and Far East, and Moscow city. The poverty line in other regions is less between 13 percent (West Siberia) and 30 percent (Central and Black Earth, and Volga-Vyatka) than in Moscow-city. These differences are smaller than the differences in the official regional poverty lines. This suggests the existence of a pro- Moscow bias in the setting of the official poverty lines. A very high percentage of the population (always in excess of 60 percent) considered itself poor using the "social" subjective poverty line. The percentage of the subjectively poor tended to decline as the minimum income itself was reduced. We thus faced a somewhat unusual situation that the percentage of the subjectively poor decreased more or less in step with reduction in people's real income. Only larger than usual income decreases were "needed" to jolt the population-that is to keep the percentage of the poor unchanged. It is also noteworthy that the percentage of the self-assessed poor was always 21 lower than the percentage of the poor according to the "social" subjective poverty line. This suggests the presence of the pockets of the population who regarded own income as adequate, while, in the view of the public perception of the minimum income, they were deemed "poor." These last two findings-the decline in the percentage of the subjectively poor as real income went down, and the lower percentage of the self-assessed than "socially" subjective poor-suggest two mechanisms of adaptation to the worsening circumstances: (1) reduction of what people perceive to be a minimum income needed for survival, and (2) the existence of very modest (less demanding) pockets of the population. 22 REFERENCES Braithwaite, Jeanine (1997), "The Old and the New Poor in Russia" in Jeni Klugman (ed.), Poverty in Russia, Washington, D.C.:World Bank. Dreze, Jean and Srinivasan, P.V. (1995) 'Widowhood and Poverty in Rural India: Some Inferences from Household Survey Data", Discussion Paper 33, Centre for Development Economics, University of Delhi. Flik, Robert J. and Bernard van Praag (1991), "Subjective Poverty Line Definitions', De Economist, vol. 139, No. 3, 1991, pp. 311-33 1. Frijters, P. and B.M.S. van Praag (1994), "Estimates of poverty ratios and equivalence scales for Russia and parts of the former USSR", mimeo. Glinskaya, Elena and Jeanine Braithwaite (1998), "Poverty, inequality, and income mobility in Russia 1994-96", American Population Association paper, forthcoming. Goskomstat Rossii (1995), Rossiiski Statisticheskiy Ezhegodnik 1995, Moscow: Goskomstat Rossii. Goskomstat Rossii (1998), Rossiya v tsifrah 1998, Moscow: Goskomstat Rossii. Hadi, A.S. (1992), "Identifying multiple outliers in multivariate data", Journal of the Royal Statistical Society, Series B 54: 761-771. Hadi, A.S. (1994), "A modification of a method for the detection of outliers in multivariate samples", Journal of the Royal Statistical Society, Series B 56: 393-396. Hagenaars, Alvi J. M. and Bernard M.S. van Praag (1985), "A Synthesis of Poverty Line Definitions", Review of Income and Wealth, Series 31, No. 2:139-54, June. Lokshin, Michael and Barry M. Popkin (1998), "The emerging underclass in the Russian Federation: Income dynamics 1992-96", mimeo. Lanjouw, Peter, Branko Milanovic and Stefano Patemostro (1998), "Poverty in the Transition Economies:A Case of Children Pitted Against the Elderly?", mimeo, Washington: World Bank, September. Milanovic Branko (1998), Income, inequality, and poverty during the transition from planned to market economy, World Bank. Ovcharova, Liliana, E. Turuntsev and I. Korchagina (1997), "Bednost': gde porog", Voprosi ekonomiki, No.7, 1997. 23 van Praag, Bernard (1971), "The Individual Welfare Function in Belgium: An Empirical Investigation", European Economic Review, vol. 2, pp. 227-269. van Praag, Bernard and Nico Van der Saar (1988), "Empirical Uses of Subjective Measures of Well-being: Household Cost Functions and Equivalence Scales", Journal of Human Resources, vol. 23, No.2, p. 193-210, Spring. van Praag, Bernard and Robert J. Flik (1992), "Subjective poverty", Foundation for Economic Research, Rotterdam, Research Institute for Population Economics, Final report in the framework of European Eurostat Project "Enhancement of family budget surveys to derive statistical data on least privileged groups." Ravallion, Martin and Michael Lokshin (1998), "Subjective Economic Welfare", mimeo. World Bank (1998), "Poverty Policy in Russia: Targeting and the Longer-term poor", Eastern Europe and Central Asia region, December 4, 1998. 24 Annex 1. The construction of the income variables The VCIOM data set contains a number of income variables that are measured on both individual and household level. Two reported income variables are individual: individual main income (main inc) and individual income from the second job (sj_inc). Household (family) income components include family main job income (mj_inc), family income from the second job (sj_inc2), income from private sector activities, pensions, other social transfers, stipends, alimonies, income from financial papers, income from sale of self-produced goods, and other monetary income. The total family income (tot inc) variable is also included in the data set and it is supposed to be equal to the sum of family income components,. This however is rarely the case. In a number of cases, the tot_inc was reported missing although the income components were available. Also, in a number of cases even though all the income components were missing, the tot inc took a positive value. Furthermore, there were inconsistencies in reported individual and family main income (i.e. between main inc and mj_inc), as well as between individual and family second job variables (i.e. sj_inc and sj_inc2). For that reason, we chose to recompute the total family income (variable tot) as a sum of family income components, that is as a sum of mj_inc, sj_inc, and income from private sector activities, pensions, benefits and subsidies, stipends, alimonies, income from financial papers, income from sale of self-produced goods, and other monetary income. This was done as follows: . The individual main income variable before April 1994 (the number of survey, n_survey=13) corresponds to the variable main_inc, and to the variable main in2 thereafter (the two variables have same definition, only the name has been changed). Thus, for n_survey > 13, we replace the value of main_inc with main_in2. However, there is no data available for November 1995 since there was no question concerning the main income in that survey, and therefore we do not have observations on the main individual income for that survey. * Family main income (mj_inc) and income from the second job (sj_inc2) variables are to be at least equal to the corresponding variables for the individual. Therefore, where the data on family income was missing or less than the individual income, we replace the value of the family income with the observation on individual income. In cases where all the income components are missing, we replace our tot with VCIOM computed total income (tot_inc). Also, where tot t 3000 a.000000 00 0-000 0 W oNo N,Og>^^oo t 9 n6o n Q A oS_Noc .o- ^*N , _oooooceo^oooeooco_ooo_oz°-g>o,f°°^oooocecooo,-ooo°°,o°°goo--oo 0 e3 p o o o o 0o oo eoo eoeo oooo ooo oe So o o oe- e-e-N °ocO ° 000 _°S oooo o o ooc 000ooo co~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Sooooo>! ocsoRooco-o,0oooooo-ooooct- -ooo oOo-ocncogoooooooonooooo o0000 o 00n - c co _- ee Z o° Zc 00o 00o 0-N - r-^c 00000 oooOG--OcowN w 0 w_ Gc e -0 n b-0-0-0-0-0-0-0-0-0-0-0-0-0-00-00--0-0-e--c-0-0-0-0-0-0-wo wNww cw 00 00 e0Ncce-* e00c-ce0c e0R 0 0 0 0 - cae-N e o NeNenere>e>er xeNe e a_^e>er e c0 00 0 e r0e r0e r e o0wc o1 0- - 00 N °_c_°_°_°S _°_°_°_°_° °_°_°_°_°_ _°_°_°_°_° °_°_ e_ c_ cO _cO N _0 0 0 0 0 G0 0 0 0 00 0 e 0 0 00 c0es~0 00 -O OO ee w000tGr O^ mgo X-os^nS _ o No. ooooooooo ood66 oooo6666 6o66d6 66 C OtONONO^ON£t .>S _ 00 _O O N e 00 0 00 0 0 Oe c 0 eO 0 0 0 00 - 0 0 0 e 0Fo ,--,co-no-oo ^. r- O c c O co _r EO_ccO_ 0_ _ wc_ r000 00 wvwww _wr wG_a E wO °- 00000000w000g0w080E00000000^000200000000000000000000000000° . :, OccoocoN 6660 66ooooooood6666doooooooo66do6660000ooodo6666600 ' e > ~ ~ ~ ~ ~ ~ 0 ~ ~ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0X UJ u o o 8 - 0 0 N -Z _- 0 0 >ttsW C =1 0 R - 0 r 2 n -- 0 zc c 0 - 0 - < ~ ~~ ~~~ ~ 0V. 0 0 0 0 C~ 2 o 0 0 - ANNEX 2. MAIN DESCRIPTIVE STATISTICS BY QUARTER Education categories shares Survey No Year Month Shart of women Age educI educ2 educ3 educ4 educ5 educ6 educ7 educ8 educ9 1 93 _II 0.5915 41.8213 0.0508 0.0759 0.0299 0.0460 0.1621 0.1013 0.2685 0.0324 0.2331 (0.4916) (14.7297) (0.2196) (0.2649) (0.1704) (0.2094) (0.3686) (0.3018) (0.4433) (0.1770) (0.4229) 2-4 93 q_2 0.5960 41.8582 0,0564 0.0782 0.0262 0.0449 0.1574 0.0922 0.2764 0.0328 0.2354 (0.4907) (14,9096) (0.2308) (0.2686) (0.1599) (0.2071) (0.3642) (0.2894) (0.4473) (0.1780) (0.4243) 5-7 93 q_3 0.6079 43.1043 0.0672 0.0857 0.0269 0.0433 0.1561 0.0978 0.2736 0.0319 0.2175 (0.4883) (15.6351) (0.2504) (0.2799) (0.1619) (0.2035) (0.3630) (0.2971) (0.4458) (0.1758) (0.4126) 8-10 93 q_4 0.6007 43.4179 0.0624 0.0800 0.0266 0,0447 0.1443 0.1026 0.2794 0.0312 0.2288 (0.4898) (15.4647) (0.2420) (0.2713) (0. 1609) (0.2067) (03514) (0.3034) (0.4487) (0.1738) (0.4201) 11-12 94 o_1 0.5975 42.6296 0,0530 0.0784 0.0232 0.0434 0.1464 0.1089 0.2781 0.0294 0.2392 (0.4904) (15.0716) (0.2241) (0.2688) (0.1505) (0.2038) (0.3535) (0.3115) (0.4481) (0.1690) (0.4266) 13-15 94 q2 0.6023 42.3752 0.0616 0.0850 0.0260 0.0396 0.1575 0.0997 0.2856 0.0329 0.2121 (0.4895) (15.2648) (0.2405) (02788) (0.1590) -(0. 1951) (0.3643) (0.2997) (0.4517) (0.1784) (0.4089) 16-17 94 q_3 0.6051 42.3560 0.0610 0.0854 0.0271 0.0399 0.1414 0.0998 0.2896 0.0367 0.2192 (0.4889) (15.2015) (0,2393) (0.2795) (0.1623) (0.1958) (0.3485) (0.2997) (0.4536) (0.1881) (0.4137) 18 94 q_4 0.6002 41.9052 0.0584 0.0717 0.0300 0.0379 0.1434 0.0964 0.2777 0.0474 0.2371 (0.4900) (14.8391) (0.2346) (0.2580) (0.1705) (0.1911) (0.3505) (0.2951) (0.4479) (0.2126) (0.4254) 19-20 95 q_1 0.5628 42.7652 0.0606 0.1014 0.0198 0.0436 0.1494 0.0910 0.2864 0.0320 0.2158 (0.4961) (15.2811) (0.2387) (0.3019) (0.1395) (0.2041) (0.3565) (0.2876) (0.4521) (0.1760) (0.4114) 21 95 q_2 0.5876 41.5867 0.0530 0.0777 0.0216 0.0382 0.1568 0.1078 0.2740 0.0368 0.2341 (0.4924) (15.5974) (0.2241) (0.2678) (0.1453) (0.1917) (0.3637) (0.3102) (0.4461) (0.1884) (0.4235) 22-23 95 q_3 0.5721 42.9184 0.0559 0.0841 0.0247 0.0423 0. 1504 0.0977 0.2632 0.0460 0.2357 (0.4948) (15.8818) (0.2298) (0.2776) (0.1553) (0.2013) (0.3575) (0.2970) (0.4404) (0.2095) (0.4245) 24 95 q4 0.5610 42.6338 0,0643 0.0807 0.0269 0.0299 0.1525 0.1046 0.2715 0.0334 0.2362 (0.4964) (16.2003) (0.2453) (0.2725) (0. 1618) (0. 1703) (0.3596) (0.3062) (0.4449) (0.1797) (0.4248) 25-26 96 q_l 0.5651 43.5239 0.0470 0.0828 0.0250 0.0444 0.1394 0.0980 0.2754 0.0419 0.2460 (0.4958) (15.3605) (0.2118) (0.2756) (0.1561) (0.2059) (0.3464) (0.2974) (t.4468) (0.2004) (0.4307) 27-28 96 qu2 0.5618 43.4090 0.0624 0.0873 0.0242 0.0490 0.1409 0.1098 0.2684 0.0359 0.2222 (0.4962) (15.9206) (02420) (0.2823) (0.1536) (0.2160) (0.3479) (0.3127) (0.4432) (0.1860) (0.4158) 29 96 q_3 0.5791 42.5874 0.0560 0.0874 0.0285 0.0354 0.1287 0.0987 0.2721 0.0398 0.2534 (0.4938) (15.4034) (0.2300) (0.2825) (0. 1664) (0.1847) (0.3349) (0.2984) (0.4452) (0.1955) (0.4351) 1-10 93 0.6006 42.7148 0.0610 0.0808 0.0269 0.0444 0.1534 0.0980 0.2757 0.0320 0.2277 (0.4898) (15.3037) (0.2394) (0.2726) (0.1618) (0.2061) (0.3604) (0.2973) (0.4469) (0.1760) (0.4193) 11-18 94 0.6012 42.3951 0.0585 0.0815 0.0258 0.0407 0.1488 0.1022 0.2833 0.0344 0.2249 (0.4897) (15.1434) (0.2346) (0,2736) (0.1586) (0.1975) (0.3559) (0.3029) (0.4506) (0.1822) (0.4175) 19-24 95 0.5704 42.5827 0.0582 0.0875 0.0231 0.0399 0.1516 0.0987 0.2734 0.0381 0.2296 (0.4950) (15.7123) (0.2341) (0.2826) (0.1501) (0.1957) (0.3587) (0.2983) (0.4457) (0.1914) (04206) 25-29 96 0.5665 43.2921 0.0551 0.0856 0.0253 0.0445 0.1379 0.1029 0.2719 0.0390 0.2378 (0.4956) (15.6012) (0.2281) (0.2797) (0.1572) (0.2062) (0.3448) (0.3039) (0.4450) (0.1937) (0.4258) Education categonies: educl: pimay and loss than pnima educ4: tech college and less than 2nda.- educ7: vocational college. secondas or vocaional eduation educ2: incomplete socond.e educ5: complete secondarc cith a diploma edacS: 3-4 yeas" of university educ3: comploto soco.da.. dsi a diplonta edac6: technical college and seconda,y edac9: unieraity completed ANNEX 3. INCOME, POVERTY LINE AND SOME STATISTICS BY SURVEY Survey Year Month/ Total real family income Per capita real family income Subjective per adult poverty line Size of Children Share of Gini coeff. No Quarter onom quartile all top quartile ottom quartile all top aurtile ottom auartile all top quartile household per household pensions (all Russia) 1 93 March 10.802 38.373 95 579 3.546 13.282 36.544 32.066 3 306 0.800 46.4 2 93 April 10.069 34.060 80521 3.365 11.534 28.862 32.192 3.285 0806 -- 42.7 3 93 May 10.420 42.154 112.501 3.954 14.855 39.573 37.803 3091 0.716 0076 47.0 4 93 June 9762 37.875 92.956 3.914 13445 33.680 36.063 3.103 0.691 0.100 42.4 5 93 July 8.866 37.961 100.633 3.737 14 157 38.862 35.746 3.014 0.674 0.113 46.5 6 93 August 7.699 34.617 91.403 3.311 12611 33.329 32.676 3.017 0662 0.155 46.1 7 93 September 8.310 34.621 86.559 3538 12.379 30699 30.888 2998 0.654 0.183 44.0 8 93 October 7.342 32.408 83.053 3 146 11.433 28604 29.523 3.009 0,665 0.200 44 1 9 93 November 7 424 32.685 82 325 3 242 11.531 28.296 30.922 2.990 0.662 0.216 42.8 10 93 Decemnber 7 244 28 916 65 434 3 020 10 128 23 977 28.861 3034 0 663 0.212 40.5 11 94 January 7196 35303 97.913 2.983 12121 32.907 27.911 3.067 0691 0200 49,7 12 94 March 7.585 34644 89836 3.200 11965 31017 25.754 3.100 0726 0.204 45.7 13 94 April 7280 33.129 84.205 2.960 11.252 28,252 24.395 3.170 0.726 0.208 447 14 94 May 7.518 31.911 78.303 3.014 10.788 26.023 25.221 3.164 0.743 0.200 41.6 15 94 June 7.334 32.546 81282 2.913 11.056 27187 25.774 3.181 0.739 0.219 43.7 16 94 July 7486 36.701 100.436 2,981 14.301 42.406 23.729 3.117 0.767 0.212 49.6 17 94 September 7992 36909 95441 3.086 13.080 35.037 25.921 3.199 0.752 0202 46.1 18 94 November 7,800 31473 77.629 2.995 11.222 28.417 22.607 3.137 0.705 0.205 42.9 19 95 January 6.011 26.331 65.746 2.552 9.257 22.856 18.508 3.061 0.745 0.246 43.5 20 95 March 6.103 23.090 54.120 2343 8.114 19.033 16.821 3.107 0.700 0.227 41.3 21 95 May 6.136 26.562 66.864 2.486 9.316 24,105 19.044 3.093 0.747 0.217 46.1 22 95 July 6.343 26.674 68.990 2,449 9.772 26.140 18.770 3.072 0.628 0.215 464 23 95 September 6.706 27.113 66.537 2.603 9.412 23.145 18.834 3.103 0.665 0.233 428 24 95 November 6.335 20.409 41.339 2.539 7.302 14.686 17.195 3.025 0.627 0.241 32.7 25 96 January 6.751 27.247 65.269 2.698 9.416 22.374 18.828 3.079 0.644 0.227 42.9 26 96 March 6.525 27.489 67.299 2.417 9.637 23.592 17.187 3.125 0669 0.216 45.2 27 96 May 6.364 27 174 71171 2.485 9.924 26.286 14.709 3.055 0,667 0.227 47.8 28 96 July 6.817 27.009 66.001 2 670 9.744 24.109 15.508 3.026 0.606 0.249 44.7 29 96 September 6.692 27.193 65.822 2.531 10.031 25.228 15.181 3.076 0.685 0.218 45.3 Mean 31.468 11.140 26.245 3.097 0.697 . 0.201 44.3 1 93 qtI 10.802 38.373 95.579 3.546 13.282 36.544 32.066 3.306 0.800 -- 46.4 2-4 93 q_2 9.965 38.034 95.020 3.695 13.281 34.312 35.365 3.159 0.736 0.089 44.0 5-7 93 q_3 8.184 35.716 92.481 3.542 13.040 34.480 33.079 3.010 0.663 0.151 45.6 8-10 93 q_4 7.228 31.339 78.191 3.097 11.032 27.013 29.769 3.011 0.663 0.209 42.5 11-12 94 n_I 7.422 34.982 94.632 3.083 12.045 31.999 26.860 3.083 0.708 0.202 47.7 13-15 94 q_2 7.361 32.527 81.533 2.937 11.031 27.272 25.132 3.172 0.736 0.209 43.4 16-17 94 q_3 7.630 36.804 97.148 3.030 13.695 38.782 24.817 3.158 0.760 0.207 47.8 18 94 q 4 7.800 31.473 77.629 2.995 11 222 28.417 22.607 3.137 0.705 0.205 429 19-20 95 q_I 6.061 24.701 60.649 2.426 8.682 21.153 17.659 3.084 0.721 0.236 42.4 21 95 q_2 6.136 26.562 66.864 2.486 9.316 24.105 19.044 3.093 0.747 0.217 46.1 22-23 95 q_3 6.311 26.886 68.166 2.531 9.598 24.628 18.801 3.087 0.646 0.224 44.6 24 95 qt4 6.335 20.409 41.339 2.539 7.302 14.686 17.195 3.025 0.627 0.241 32.7 25-26 96 q_I 6.604 27.366 66.137 2.558 9.524 22.846 18.022 3.102 0.656 0.221 44 1 27-28 96 q_2 6.505 27.091 69.085 2,593 9.833 25.323 15.112 3.040 0.636 0.238 46.2 29 96 q_3 6.692 27.193 65.822 2 531 10.031 25.228 15.181 3.076 0.685 0.218 45.3 Mean 30.631 10.861 26.245 3.103 0.699 0.205 44.1 1-10 93 8.427 35.297 90.451 3.420 12.514 32.381 32630477 3.082 0.698 0158 446 11-18 94 7.449 34.141 88.933 3.021 11.975 31.485 25.301104 3.137 0.729 0.206 45.4 19-24 95 6.169 25.110 61.386 2.484 8.896 21.926 18.240928 3.077 0.682 0.229 41.4 25-29 96 6.610 27.220 68.443 2.555 9.750 24.523 16.279124 3.072 0.654 0.227 45.2 Mean 30.442 10.784 26.245 3.092 0.691 0.205 44.2 Note: Per capita real family income is calculated using number of persons per household as weights. ANNEX 4. REGIONAL GINI COEFFICIENTS (BY SURVEY) Regionk 1 2 3 4 5 6 7 8 9 coeff. Survey Northern Central-BICaucasusVolga-VyaVolga Urals W. SiberiaFar East Moscow st.deviat. mean of var 1 34.7 38.3 62.7 36.0 41.3 39.1 39.0 39.6 43.5 7.9 41.6 18.9 2 40.6 33.6 52.0 37.6 40.3 31.5 43.0 37.6 36.0 5.7 39.1 14.5 3 43.4 45.3 43.8 35.5 40.3 43.3 39.1 42.3 49.9 3.8 42.5 9.0 4 35.2 35.4 48.8 35.8 35.6 31.5 43.2 41.4 46.0 5.5 39.2 14.0 5 43.9 36.5 38.6 39.1 38.7 40.8 49.8 42.8 52.0 5.0 42.5 11.8 6 51.8 34.5 35.5 38.8 40.3 34.8 47.3 41.3 53.6 6.8 42.0 16.3 7 44.4 32.9 41.8 40.5 40.3 37.2 45.2 47.5 39.8 4.1 41.1 10.1 8 45.5 41.8 43.7 32.9 33.6 34.6 45.4 44.4 47.3 5.4 41.0 13.1 9 50.2 30.5 47.1 34.9 37.7 38.7 38.3 37.5 42.1 5.7 39.7 14.3 10 36.2 35.2 37.7 39.0 41.0 37.8 39.8 36.5 43.3 2.4 38.5 6.2 11 42.8 36.9 43.2 34.3 40.6 35.9 39.1 39.6 67.4 9.4 42.2 22.2 12 37.6 33.8 50.4 30.8 40.5 47.5 58.9 39.7 36.5 8.4 41.7 20.1 13 41.1 40.9 42.0 43.0 36.8 45.5 38.6 43.7 41.7 2.5 41.5 5.9 14 36.5 33.5 40.6 40.0 42.4 40.0 37.6 40.8 39.3 2.5 39.0 6.5 15 41.7 37.5 45.0 38.2 42.6 38.3 37.4 45.4 36.4 3.3 40.3 8.1 16 42.6 50.9 60.9 39.1 37.9 39.5 35.1 44.2 42.1 7.4 43.6 17.0 17 44.2 37.9 43.1 42.0 45.4 40.3 35.4 52.1 40.4 4.5 42.3 10.7 18 35.8 37.8 43.5 38.8 42.0 37.7 35.0 39.5 45.2 3.2 39.5 8.2 19 39.4 38.0 51.8 34.0 37.8 41.9 42.9 37.6 33.7 5.2 39.7 13.1 20 39.7 35.4 36.0 29.3 33.7 38.5 37.5 41.3 42.9 3.9 37.2 10.5 21 46.4 37.8 49.3 42.9 39.3 34.4 37.5 40.4 46.3 4.7 41.6 11.2 22 40.3 35.3 41.0 30.9 39.8 36.5 47.8 28.5 48.9 6.5 38.8 16.7 23 39.2 36.3 42.5 41.2 34.7 37.4 43.4 35.3 42.5 3.2 39.2 8.1 24 26.6 33.0 29.0 27.7 30.9 28.5 28.8 32.1 28.4 2.0 29.4 6.8 25 27.7 36.6 34.0 38.8 41.2 36.8 35.7 42.2 44.9 4.8 37.5 12.7 26 40.7 38.4 45.7 33.4 45.0 35.9 29.5 48.5 40.6 5.8 39.7 14.6 27 32.2 33.7 39.6 36.9 56.2 40.3 46.5 39.6 48.9 7.3 41.6 17.5 28 36.4 41.4 49.4 35.6 36.2 39.6 41.2 40.1 41.4 3.9 40.1 9.8 29 37.7 39.3 44.7 34.1 33.7 38.4 41.2 42.1 44.8 3.8 39.6 9.7 Annex 5. Regression results with cross section dummies Dependent variable: In subjective minimum income for an adult (AMY) (1) Basic (2) = (1) without (4) = (3) with equation with Hadi outliers Gini coefficient Huber (robust) variances Ln family income 0.139 0.127 0.123 1/ (39.0) (36.7) (35.1) Age 0.016 0.017 0.018 (18.5) (20.6) (21.4) Age2 -0.0002 -0.0002 -0.0002 (-23.9) (-26.0) (-26.9) Small towns and -0.062 -0.065 -0.064 villages (-7.9) (-8.8) (-8.5) (population under 100,000) Towns (between 0.066 0.577 0.058 100,000 and 1/2 (7.5) (6.9) (7.0) million) Medium size 0.060 0.060 0.060 cities (between l/2 (5.7) (6.0) (6.0) and 1 million) Northern region -0.243 -0.221 -0.222 (-24.1) (-23.4) (-23.5) Central and -0.330 -0.309 -0.311 Black Earth (-29.4) (-29.1) (-29.3) North Caucasus -0.225 -0.211 -0.214 (-18.3) (-18.0) (-18.4) Volga-Vyatka -0.324 -0.292 -0.294 (-25.5) (-24.1) (-24.2) Volga -0.257 -0.238 -0.241 (-21.8) (-21.3) (-21.6) Urals -0.194 -0.180 -0.180 (-18.0) (-17.7) (-17.8) West Siberia -0.147 -0.128 -0.130 (-12.4) (-11.4) (-11.6) East Siberia and 0.034 0.027* 0.029* Far East (2.7) (2.3) (2.5) Gini coefficient -0.033 (26.7) Survey 2 -0.003** -0.006** 0.115 (-0.2) (-0.4) (7.5) Survey 3 0.116 0.085 0.065 (6.6) (5.2) (3.9) Survey 4 0.144 0.135 0.266 (8.5) (8.5) (18.8) 5 0.123 0.097 0.093 (7.1) (5.9) (5.7) 6 0.626 0.515 0.061 (3.7) (3.2) (3.8) 7 0.026** 0.012** 0.091 (1.5) (0.8) (6.4) 8 -0.014** -0.033* 0.042 (-0.8) (-2.1) (2.8) 9 -0.021** -0.022** 0.097 (-1.2) (-1.4) (6.7) 10 -0.060 -0.057 0.134 (-3.5) (-3.6) (9.8) 11 -0.099 -0.096 -0.204 (-5.8) (-6.1) (-11.2) 12 -0.122 -0.116 -0.093 (-7.4) (-7.4) (-6.1) 13 -0.173 -0.168 -0.113 (-9.9) (-10.1) (-7.2) 14 -0.123 -0.122 0.035* (-7.0) (-7.3) (2.3) 15 -0.099 -0.098 -0.011 ** (-5.7) (-6.0) (-0.8) 16 -0.195 -0.199 -0.302 (-11. 1) (-I11.9) (-16.1) 17 -0.098 -0.078 -0.068 (-5.6) (-4.7) (-4.1) 18 -0.249 -0.236 -0.122 (-14.0) (-13.8) (-7.9) 19 -0.383 -0.377 -0.281 (-19.5) (-19.9) (-15.9) 20 -0.455 -0.444 -0.278 (-24.5) (-24.8) (-17.3) 21 -0.395 -0.399 -0.390 (-21.9) (-23.4) (-23.1) 22 -0.393 -0.390 -0.393 (-21.9) (-23.0) (-23.2) 23 -0.370 -0.371 -0.256 (-21.2) (-22.4) (-17.2) 24 -0.445 -0.447 Dropped (-24.9) (-26.5) 25 -0.391 -0.386 -0.274 (-22.4) (-23.3) (-18.4) 26 -0.455 -0.440 -0.401 (-25.4) (-25.5) (-24.3) Survey 27 -0.638 -0.640 -0.686 (-35.6) (-38.1) (-38.9) 28 -0.580 -0.585 -0.531 (-32.0) (-34.2) (-32.8) 29 -0.615 -0.611 -0.575 (-33.7) (-35.4) (-34.5) Constant 2.774 2.720 1.204 (106.1) (110.1) (22.7) Sample size 79,595 76,965 76,965 R2 0.462 0.489 0.490 F value 1661.1 1782.5 1745.7 Note: t-values given in parentheses (under the coefficients). All coefficients are significant at the 1 percent level, except those with * which are significant at the 5 percent level, and ** =not significant. For size of settlement, the omitted category is larger cities (population over 1 million). For the regions, the omitted variable is Moscow-city. 1/ Defined as Y/NO.62. Policy Research Working Paper Series Contact Title Author Date for paper WPS2059 Financial Intermediation and Growth: Ross Levine February 1999 K. Labrie Causality and Causes Norman Loayza 31001 Thorsten Beck WPS2060 The Macroeconomics of Delayed Daniel Kaufmann February 1999 D. Bouvet Exchange-Rate Unification: Theory Stephen A. O'Connell 35818 And Evidence from Tanzania WPS2061 A Framework for Regulating Hennie van Greuning February 1999 A. Thornton Microfinance Institutions Joselito Gailardo 80409 Bikki Randhawa WPS2062 Does Financial Reform Increase Oriana Bandiera February 1999 A. Yaptenco or Reduce Savings? Gerard Caprio, Jr. 38526 Patrick Honohan Fabio Schiantarelli WPS2063 The Practice of Access Pricing: Tommasso M. Valletti February 1999 G. Chenet-Smith Telecommunications in the United 36370 Kingdom WPS2064 Regulating Privatized Rail Transport Javier Campos February 1999 G. Chenet-Smith Pedro Cantos 36370 WPS2065 Exporting, Externalities, and Howard Pack February 1999 C. Bemardo Technology Transfer Kamal Saggi 31148 WPS2068 Flight Capital as a Portfolio Choice Paul Collier February 1999 A. Kitson-Wafters Anke Hoeffler 33712 Catherine Pattillo WPS2067 Multinational Firms and Technology Amy Jocelyn Glass February 1999 L. Tabada Transfer Kamal Saggi 36896 WPS2068 Quitting and Labor Turnover: Tom Krebs February 1999 T. Gomez Microeconomic Evidence and William F. Maloney 32127 Macroeconomic Consequences WPS2069 Logit Analysis in a Rotating Panel Patricio Aroca GonzAlez February 1999 T. Gomez Context and an Application to William F. Maloney 32127 Self-Employment Decisions WPS2070 The Search for the Key: Aid, David Dollar March 1999 E. Khine Investment, and Policies in Africa William Easterly 37471 WPS2071 The World Bank's Unified Survey Jos Verbeek March 1999 M. Galatis Projections: How Accurate Are 31177 They? An Ex-Post Evaluation of US91-US97 Policy Research Working Paper Series Contact Title Author Date for paper WPS2072 Growth, Poverty, and Inequality: Quentin T. Wodon March 1999 J. Badami A Regional Panel for Bangladesh 80425 WPS2073 Politics, Transaction Costs, and the Antonio Estache March 1999 G. Chenet-Smith Design of Regulatory Institutions David Martimort 36370 WPS2074 Light and Lightning at the End of Antonio Estache March 1999 G. Chenet-Smith the Public Tunnel: Reform of the Martin Rodriguez-Pardina 36370 Electricity Sector in the Southern Cone WPS2075 Between Group Inequality and Quentin T. Wodon March 1999 J. Badami Targeted Transfers 80425 WPS2076 Microdeterminants of Consumption, Quentin T. Wodon March 1999 J. Badami Poverty, Growth, and Inequality in 80425 Bangladesh