PofyRues adEls AN**1 WORKINJG ftAPEHS L World Dwdopment Report Offcs of the Vice President Development Economics The World Bank March 1991 WPS 637 Badcgound P,mr for the 1990 Worid Developmen Repotl Poverty in Poland 1978-88 Branko Milanovic As a result of Poland's economic crisis, which began in 1978, the proportion of Polish people living below the poverty line in- creased from 10 percent to almost 20 percent. Farm and mixed (farmVnonfarm) households weathered the crisis better than workers and pensioners - probably because farmers could vary their crops and workers in mixed households could choose between work in socialized industry or private agriculture. Thc Policy, Rcscarch, and Exterral Affairs Complex distnbutes PlR Working lapcrs todisseminaicthe fininng or WOk in prorm and to cncournge thc cxchangc of ideas among llank staff and aD others intcersted in devclopmncnt issucs. These papers carry the names of the authors, clect only their views, and should bc used and cited accordingly. The findings, interpreLauurs, and conclusions ae the authors' own. 'rhey should not bc stibuted to the World l3ank, its Board of Directors, it managemnct, or any of its mcmber couwntries. Policy, Rma,oh. wnd Extwmnl Ali| World Development Report WPS 637 This paper - a joint product o. the Socialist Economies Reform Unit, Country Economics Department and the Country Operations Division, Country Dcpartment IV, Europe, Middle East, and North Africa Regional Office - was written as a background paper forthc 1990 WorldDevelopmentReporton poverty. Copiesare availablc fre from th Wodd Bank. 1818 H SteNW. Washington.DC 20433. Please contacl t!e World Development Report office, room S13-060. extension 31393 (25 pages). The economic crisis that began in Poland in 1978 group for which the incidence of poverty de- significantly reduced the population's average creased was mixed households. incomes (about 20 percent by 1988) and in- creased the proportion of the population living Until the end of the period studied (1988), no below the poverty line by 10 percentage points. unemployment appeared. The wage bill was (It is significant that 3.1 million of the 7 million reduced by uniform cuts in real wages - so the estimated poor in Poland are the "new poor.") wage and the overall distribution of income rcmained practically unchanged. The real The composition of the poor has also incomc of pensioners' households decreased changed. Bcfore the crisis, most of thc poor almost as much as that of workers' households. lived in rural areas; now 70 percent of them livc in cities. This change occurred because of a Farm and mixed households weathered the sharp jump in poverty among workers in the crisis better than workers and pensioners. This socialized sector, whose real wages declined. was not so much because tcrms of trade bctween agriculture and industry improved, but because The most important dircct causc of increased farmers and mixed households had more flexibil- poverty in the second half of the 1980s was ity about economic decisions. Farmcrs could increased poverty in workers' households. The change the composition of their crops and mixed second most important cause was demographic: households could also vary their labor inputs in shifting to retirement, some workers' housc- between work in socialized industry and private holds joined the ranks of the poor. The only agriculturc. I The PRE Working Paper Scrics disseminatcs thc findings of work undCr way in thc Bank's Policy, Rcscarnh, and Extemal AffairsComplex. An objective of the scrics is to get these findings out quickly, even ifprcsentations arc icss than fully polished. The findings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by thc PRE Dissemination Center Poverty in Poland, 1978-88 by Branko Milanovic Table of Contents 1. Introduction 1 2. Changes in Poverty, 1978-88 3 3. Some Poverty Accounting 8 4. Factors behind Changes in Poverty Coefficients 11 5. Conclusions 15 Annex 1. The Definition of the Social Minimum (Poverty Line) 16 Annex 2. Transfers and Poverty: Some Poverty Accounting 21 1 POVERTY IN POLANP. 1978-88 Branko Milanovic 1. Introduction This paper considers the Issue of poverty In Poland In the period 1978-88. The first year of the period represents a benchmark year. It Is the year when Polish GDP peaked, and real incomes of the population were higher than at any time since. The decline In GDP continued until 1983. Since then the economy notched modest Increases. By the end of the period (1988), GDP per capita was 1.5 percent below its pre-crisis level, while the average standard of living (as reflected in real per capita income of the population) was 20 percent lower. It is important to study how economic stagnation affected the poor. It is generally felt that poverty expanded significantly. The appearance of soup kitchens In main cities of Poland in 1989 provides a tangible evidence of the degree of pauperization. In order to avoid possible misunderstanding we must state explicitly what are the premises and sources on which our analysis is based. First, the words "poverty" or "poor" should be understood only in their technical meaning. We classi'y as "poor" all people whose incomes are less than the social minimum calculated by the Institute of Labor and Social Affairs In Poland.2 This is a purely conventional definition, since It is generally held that the social minimum is higher than what most people In Poland would regard as uncontestable poverty. It is also higher than a pure existential minimum (or some measure based on a minimal calorie intake). Yet the social minimum, as defined by the Institute, 1 The first draft of this paper, covering the period 1978-87, was written as a background paper for the World Bank World Development Report 1990. The author acknowledges comments made, at various stages of the paper, by Bela Balassa, Lyn Squire, Aleksandra Posarac, Irena Topinska and Michael Walton. 2 We are dealing with Incomes and not expenditures. Incomes, however, are corrected for consumption requirements, so that we classify as poor a household whose income per consumption unit is less than some minimum. Classified as poor are obviously all persons in this household. 2 allows only for a very mdinimal satisfaction of human needs. A more detailed description of the minimum Is provided in Annex 1. The social minimum represents that level which at a given time and in a given environment Is deemed Indispensable for decent livlng. This is the rationale for treating the minimum as the "poverty line". The poverty line must consequently be understood as relevant only In a specific context, limited both in space and time: the Poland of the 1980's. Since the line is constant in real terms, It allows us to chart relatively well how the extent and the composition of poverty changed during the last ten years. The paper Is not concerned with economic and sociological characteristics of the poor per se. It Is also beyond the scope of the paper to study the route by which people fall Into poverty, and how different specific subgroups (e.g. single mothers, school drop-outs, unskilled people In the countryside) are affected. This requires much more detailed micro analysis. The approach adopted here is more of a "broad-brush" kind. We use only published sources and: (1) estimate the extent of poverty in the last ten years; (2) study how composition and incidence of poverty In the four main social groups (workers', mixed, farmers', and pensioners' households) has changed and; (3) propose some general, relatively simple and Intuitive, explanations of the macro-economic factors that influence changes in poverty. We are concerned only with the "head-count" or poverty incidence measure.3 This Is partly determined by the nature of the task ("How many pe-ple are (have become) poor?"), and is partly chosen for reasons of simplicity. The structure of the paper reflects these objectives. Section 2 charts 1- evolution of poverty. In Section 3 we presents some "poverty accounting". This is an attempt to disentangle demographic and migrational effects from purely economic effects. We shall be concerned with households who in 3 Terms "ppoverty coefficient" and "poverty incidence" are used interchangeably. 4 For example, if population growth rates are higher In low income groups, then an increase in population, with everything 3 the last ten years have joined the ranks of the poor. These are "the new poor" and to find out who they are. Is, for political and social reasons, particularly important. Section 4 presents some econometric evidence on poverty, viewing the percentage of the poor In a social group as determined by two variables: average Income of the group and inequallty of Income distribution within the group. 2. Changes in Poverty, 1978-88 Total percentage of people classifled as poor In 1987 and 1988 is almost twice as high as at the onset of the crisis In 1978. As mentioned before, Polish real GDP then reached its peak. Between 1979 and 1982 GDP ner capita decreased by 24 percent. The decline was without precedent in post-war Europe. Starting from 1983 relatively slow recovery began with the result that in the last year of the period under study (1988) GDP per capita was only slightly below the 1978 level. Real income of the population as obtained from Household Surveys was 20 percent lower than In 1978 (see Figure 1). It Is therefore not surprising to find that while the share of the poor in total population was u 10 percent in 1978-79, since 1982 It has been less than 17 perce..t only once. Figure 1 Real per capita Incomes and Poverty Coefficients 19781 1060 1980 1982 1964 1966 30 26 20 16 10 6 0 20 40 0 80 1t0 120 Powrly Coeftiolont (In %) RMl Inroma (1970-100) else the same, Increases the percentage of the poor; or a transfer of popul^'-n from "high-poverty" groups or areas to "low-poverty" groups oa .. ea reduces the overall poverty incidence. 4 The overall (country-wide) poverty coefficient Is the outcome of two effects: different poverty coefficients for different social groups, and varying shares of social groups in the sample. These data are presented in Tables 1 and 2 Consider first the structure of the sample. If we compare only the end years of the period (1987-88 vs. 1978-79) we can see that the structure of population as between urban and rural households Is practically ..nchanged. Rural population (mlxed and farmers) accounts for Table 1. Poverty Coefficients, 1978-88 (share of the poor in tota: group's population) a/ Workers hlxed Farmers Pensioners Total 1978 6.4 9.5 14.9 20.8 9.2 1979 6.1 12.8 16.7 17.1 9.7 1980 7.8 10.6 17.2 23.7 11.1 1981 11.4 11.4 16.4 29.2 13.9 1982 17.3 15.8 20.9 35.7 19.8 1983 '9.1 13.4 29.7 49.0 23.7 1984 19.0 12.9 25.1 39.3 21.9 1985 17.3 11.3 19.5 32.4 19.1 1986 17.0 9.4 19.2 25.4 17.3 1987 25.2 12.6 21.4 27.6 22.7 1988 14.8 8.0 14.4 25.9 15.2 a/ Coefficients are calculated In terms of total group's population (individuals in a group not households). Table 2. The Structure of the Sample, 1978-88 (in percent of total sample) Workers Mixed Farmers Pensioners Total 1978 61.7 17.1 13.3 7.8 100 1979 61.2 16.1 13.9 8.9 100 1980 60.5 15.8 13.9 9.8 100 1981 60.3 15.5 14.0 10.2 100 1982 61.6 12.7 13.7 11.9 100 1983 61.4 13.8 10.9 14.0 100 1984 61.1 13.2 10.8 14.8 100 1985 60.5 13.0 10.9 15.6 100 1986 55.5 16.9 13.0 14.7 100 1987 52.5 18.4 14.7 14 ' 100 1988 52.1 18.4 14.2 15.2 100 slightly over 30 percent of the sample, about 2-3 percentage points more than In the beginning of the period. The composition of the rural population Is broadly unchanged as both the share of 5 farmers' and mixed households went up by about 1 percentage point. The situation among urban households is different. The importance of workers' house, :lds has decreased from more than 60 percent of the sample to _bout 52 percent; conversely, the share of pensioners has increased from 8-9 percent to 15 percent.5 The last fact, namely Increasing share of pensioners, points to the first cause of Increased poverty. Since pensioners' households consistently have the highest Incidence of poverty, an increase in their share drives the overall poverty coefflilent up. Poverty incidence among pensioners has Increased from less than 20 percent (in the beginning of the period) to 25-26 percent. While in the beginning of the petriod pensioners contributed about 1.6 points to the overall poverty coefficient (this is the product of the group's poverty coefficient and Its share in the sample; see notes to Taole 3), this increased to 4 points. Pensioners thus alone account for 2.4 percentage point increase in the overall poverty coefficient. 6 This explains a quarter of the overall increase. The second important cause of increased poverty has to do with workers' households. They display two essential characteristics: d 'ning share In total population and rising poverty coefficient. The second feature is not unique to workers: poverty coefficients for all social groups except for mixed households increased. Workers households, however, were the most severely affected. Probability of living In a poor worker household has tripled: the poverty coefficient increased from little over six percent before the crisis to 25 percent in 1987 and 15 percent in 1988. Developments among workers' households thus account for 6.5 percentage point increase in the overall poverty coefficient: they explain more than two-thirds of the total increase. Combined urban households (workers and 5 Comparison between the end and the beginning of the period always refers to years 1987-88 and 1978-79. The average of two years is taken to even out sharp yearly fluctuations. 6 The peak in terms of pensioners' contribution was reached in 1983, when extremely high poverty incidence (49 percent) and a high share (14 percent) combined to make pensioners' contribution to total poverty almost 7 percent. 6 pensioners) therefore explain 95 percent of the overall increase in poverty. Table 3. Factors Explaining the Change in Poverty 1987-88 versus 1978-79 Contributions Workers MlxSed Farmeus Pensioners Total 1978-79 3.87 1.85 2.15 1.58 9.45 1987-88 10.42 1.90 2.58 4.01 18.91 Change +6.55 +0.05 +0.43 +2.42 +9.46 Relative b/ contribution(%) 69.3 0.5 4.5 25.6 IlnO Poverty effect ci +8.41 -0.14 +0.28 +0.65 9.21 Population eff.d _0.59 +0.20 +0.14 +1.25 1.01 Interaction term -1.27 -0.02 +0.02 +0.52 -0.75 Total +6.55 +0.05 +0.43 +2.42 +9.46 a/ The product of the group's share in total population and its poverty coefficient. b/ Group's contribution to poverty divided by the overall change in the poverty coefficient. c/ Calculated on the assumption that the group's share In total population is the same as in 1978-79, and that only Its poverty coefficient has changed. d/ Calculated on the assumption that the group's poverty coefficient is the same as in 1978-79, and that only Its share in total population has changed. The mechanism leading to the increased contribution to poverty has been different for workers' and for pensioners' households. For workers, the cause lies in increased poverty within the group; for pensioners, it was principally their rising share in total population. In total (for all social groups) increased poverty within the groups accounts for 9.2 out of 9.5 percentage points increase in poverty (Table 3). However, population change (including demographic and migrational effects) also contributed to increased poverty. This was almost entirely due to transfer from workers' to pensioners' households, which Is in effect a movement from a low-poverty to a high-poverty group. Rising share of pensioners came about not only because of demographic trends but was also due to government decision to 7 lower the mandatory retirement age by five years in 1983. The decision was motivated by fear of widespread unemployment following the introduction of market-oriented refcrms In 1982 and 1983. The first two conclusions about the changes in poverty are: (1) Thu most important direct cause of greater overall poverty in the second half of the 1980's Is Increasing poverty among workers households. (2) The second most Important cause Is of a predominantly migrational or demographic character. Some of workers' households experienced, due to retirement, a decline In their lrcome and joined the ranks of the poor. These two effects (shown In bold In Table 3) account for the entire change in poverty. All the other effects cancel out. Among rural households the crisis did not have such dramatic effects. Poverty among farmers increased by about 2.5 ,ercentage points (from 15.5 to 18 percent). Mixed households represent an exception to generalized increase in poverty. They are the only group whose poverty coefficient In 1987-88 is (slightly) lower than before the crisis. From 1982 they display the lowest po-erty incidence of all groups. At about the same time their average per capita Income begins to equal or to exceed that of workers' households.7 This leads to the third conclusion: (3) The only group that experienced decrease in the incidence of poverty were mixed households. From the early 1980's both farmers and mixed households' average per capita incomes are higher than workers'. However, higher degree of inequality, particularly among farmers, Is responsible for the fact that these higher average Incomes are not translated Into equivalently lower poverty coefficients. 8 3. Som Poverty Accounting Table 4 shows total number of the poor in the period 1978-88. It Is obtained by applying calculated poverty coefficients to the estimated rural and urban population. Table 4. Tot&. Number of the Poor a/ (in 000 of people) Wgrkers Mixed Farmers Fensioners Urban Ruril Total 197'z 1154 793 967 472 1627 1760 3386 1979 1094 1025 1151 441 1536 2177 3712 19b0 1396 830 l188 687 208: 2018 4101 1981 2055 887 1147 893 2948 2034 4982 1982 3108 1121 1597 1244 4351 2718 7069 1983 3400 1106 1931 1990 5390 3037 8427 1984 3385 1050 1675 1697 5082 2725 7807 1985 3072 914 1322 1483 4555 2236 6792 1986 3042 790 1240 1204 4246 2030 6277 1987 4491 1041 1401 1388 5879 2441 8321 1988 2654 663 921 1358 4012 1534 5596 1978-79 1124 909 1059 466 1582 1968 3549 1987-88 3572 852 1161 1373 4945 2012 6958 Change +2448 - 57 + 102 + 907 +3363 + 44 +3409 Relative contribution to total Increase U.j 71.8 -1.7 3.0 26.6 98.7 1.3 iCo a/ The number of the poor in workers' households calculated as follows: percentage share of workers' households In total urban households (from the Surveys) times total urban population (from the demographic macro data) times poverty coefficient for workers' householdr The same procedure is used for other social groups. Total estimated number of people living below the poverty line rose from about 3.5 million before the crisis to 7 million in 1987-88. The Increase is entirely concentrated in urban areas. Almost 2.5 million of the new poor belong to workers' households and about 0.9 million are pensioners (Table 4). The average poverty incidence In u.ban households went up from 7.8 percent to 21.5 percent. The position of rural households did not worsen: total number of the poor in mixed households slightly decreased, while among farmers It Increased by only 100,000. Poverty 9 coefficient for the rural population as a whole remained practically constant: 13.3 percent In 1978-9 and 13.7 percent in 1987-8. Different evolution of poverty among urban and rural households completely altered the plcture of -erty. While before the crisis total number of the rural po exceeded the number of the urban poor, the ratio now stands at approximately 2.5-to-I In favor of the urban poor. The emergence of significant urban poverty has far-reaching consequences for economic policy (e.g. towards whom should the main thrust of welfare policy be directed? will Increased unemployment, due to reorganization of the economy, be easily absorbed? etc), as well as for social stability. A political system can, ceteris paribus, cope more easily with rural than with urban poverty. Rural poverty Is often "buried" In the countryside, while urban poverty is highly visible. Urban citizens are also politically more active and Influential among other reasons because they are closer to the centers of power. Poland presently enters tht painful process of Industrial restructuring and transition to market system, In which urbe.- population Is likely to be the most affected. The two startirg conditions --large number of the urban poor and a very strong trade union movement -- render this process more difficult. Particularly important question is how many of the roor are the "new poor", that is people who before the crisis lived above, and are now below the poverty line. We turn to this question next, by trying to estimate their number. If we divide all the population, and accordingly all the poor, Into two groups, agricultural and urban, we can write the number of the poor In the agricultural sector In period 1 (PA1) as equal to their number In period 0 (PA0) plus increase of the poor 8 It should be mentioned that thei.: Is In Poland also a strong farmers lobby. It draws non-negligible portion of Its strength from the shared feeling that private agriculture was treated inimically by the authorities until the early 1980's. Farmers lobby has been able to commit all recent governments to the parlty policy whose aim Is to equalize Income of farmers with income of workers In the state sector. The lobby seems to be well-represented across the political spectrum: among "Rural Solidarity" and United Peasant Party (formerly allied with Communists) as well as among some technocrats in the current government. 10 In agriculture due to population growth (n pa) plus the new poor in agriculture (NPA) minus transfers of the poor from agriculture to urban areas (t ):9 a PA, = PA + npa + NPA- ta (1) Similar equation for urban households shows that the number of the poor In urban areas In period 1 (PU ) Is equal to their number in the previous period (PU 0) plus Increase of the poor as result of population growth (n pu) plus the new poor in urban areas (NPU) plus people who migrated from the agriculture and are now poor (T a), where a = the percentage of transferees who are poor and T a=total transfers from agriculture to urban areas. PU = PU + n + NPU + T (2) 1 o pu a Using averages for the 1978-79 (the beginning of the period, t=0) and 1987-88 (t=1) we can write equations (1) and (2):10 2013 = 1969 + 160 + NPA - 0.15 Ta = 1969 + 160 + NPA - 0.15 (1347) (la) and 4946 =1582 + 129 + NPU + 0.22 T = 1582 + 129 + NPU + + 0.215 (1347) (2a) where 1347 = estimated total transfers from rural areas, and npa and npu are calculated assuming that the population growth rate among the poor Is the same as the overall rate. We further assume that transfers are not exactly uniform across Income groups, but rather biased toward low Income agricultural households. Consequently, the share of the poor In agricultural transfers (15 percent; see equation (la)) somewhat exceeds their share In agricultural population in the beginning For a more complete explanation of the methodology see Annex 2. 10 All data In thousands. 11 Total transfers are estimated as the difference between what the rural populatior would be at the end of the period (with a population growth rate of 0.79 percent p.a.) and its actual size. Increase in the number of the poor due to population growth is calculated by applying to the overall population growth in rural and urban areas the initial poverty coefficients. In a more detailed study, if population growth is inversely related to income, this calculation could be corrected. I of the period (13 percent). The percentage of transferees who are poor in cities is assumed to be the same as the average level of poverty in urban areas at the end of the period (21.5 percent). From the two equations we obtain NPA = 86 and NPU = 2945. This means that there are only 86,000 new poor in rural areas, and almost 3 million new poor In the urban areas. Total increase In the number of the urban poor is composed of 3 million new urban poor, 290,000 rural migrants, and 129,000 people who were born in the already poor households. It is significant that more than 3.1 million out of the total number of 7 million of the poor are the new poor 13, i.e. people who before the crisis lived above the poverty level, and have now fallen below It. 4. Factors behind Changes in Poverty Coefficients One of objectives of a study of poverty is also to link observed changes in incidence of poverty to macroeconomic variables. This is Important because regularities of this kind, if established and found sufficiently robust, allow us to make conclusions about the impact of various macroeconomic measures on poverty. To take an extreme example, suppose that we are interested In assessing the Impact on poverty of a reduction In real wages. That impact will vary in function of the importance of wages in total income of a social (or income) group, inequality of the wage distribution, participation rates etc. The importance of the impact may thus fluctuate between fairly minimal and substantial. Policy implications of one or another conclusion are quite different. In this section we shall try to relate changes in poverty coefficients of urban and rural population to 12 In rural areas the accounting is as follows: there are 86,000 new poor plus 160,000 born in already poor families = 246,000. Out of these, 202,000 (15 percent times 1,347,000) migrated to cities, which yields a net Increase of 44,000. 13 This figure is composed of: 2.945 million new poor in cities + 86,000 new poor in rural areas + (290-202) thousand new poor due to migration from rural to urban areas = 3.119 million. 12 ..r..* )nomi,- variables. 4 The most natural candidates are: (1) iveraiiw real inccme of a group, and (2) the within-the-group Gini -oefii Aent as an indicator of' the pattern of distribution. We -an expect that the t'irst variable be negatively, and the second, positivelv, r-elated to poverty. .'he results are displayed in T'able 5 15 A one percent uniform reduction in zeal income of urban and rural households is associated with respectively 1 and 1.6 percent increase in the incidence of poverty (income elasticities of 1 and 1.6). This means that relatively more people are bunched around the poverty line in the case of rural population. The distribution term is statistically significant only in the equation for rural households. 14 In order to increase the number of observations the data set for urban population is composed of 11 annual observations for workers and 11 annual observations for pensioners' households. The same applies to rural population which is composed of farmers' and mixed households. 15 For income we are using real wages in the socialized sector or real pensions (annual averages) rather than average real income of workers' (pensioners') households as given in Household Surveys. Correlation coefficient between the two is very high: 0.95. The f'irst type of data (average wage or pension) is a macro variable avai.lable with less than a month delay; the second Is available only with 1.5 to 2 years delay. For policy forecasts It is therefore easier to use average wage or pension. 13 Table 5. Determinants of Povert, Dependent variable: log percentage of the poor -2 period constant income distrib. R DW term term term (F) (SE) Urban households a. so 1978-88 10.607 -1.009 0.127 0.864 1.74 (0.000) (0.000) (0.791) (43.21) (0.189) Rural households so a. *- 1978-88 8.402 -1.609 2.009 0.833 1.61 (0.001) (0.000) (0.000) (53.33) (0.136) Notes: Equations are of the log form: log (POOR) = Bo + B1 log (income) + B2 log (distribution). Autoregression coefficient is statistically significant at less than 1 percent in the first equation; it Is not statistically significant in the second. Number of observations is 21 for the first, and 22 for the second equation. Income is in 1978 constant zlotys (wages and pensions for urban households; real per capita household income from Surveys for rural households). Distribution term is the Gini coefficient for each social group as calculated from the samples in the Household Surveys. Data in brackets below regression coefficients show levels of significance at which the null hypothesis is rejected. It is important to be able to tell what are the likely effects on poverty of changes in some key macro variables. For urban households this is relatively easy since real wages and real pensions, as shown in previous equations, have an unambiguous and measurable effect on poverty. The situation is different for rural households. Only the use of real per capita income of farmers' and mixed households (obtained from Surveys) yields meaningful results. Agricultural terms of trade (TOT) and real revenues of agricultural households (AGROR, compiled by the Central Statistical Office) are only very loosely related to the income data from the Surveys (YFARMR) and thus to poverty incidence among farmers, POORF (see Table 6). It means that TOT and AGROR are bad predictors of farmers' income. Unfortunately, the Survey data on farmers' income are available only at annual intervals, and cannot be used for short-term policy forecasts. 14 Table 6. Correldtion Coefficients, 1978-88 TOT AGROR YFARIR lOT - AGROR 0.929 - YFARMR 0.430 0.525 - POORF -0.391 -0.418 -0.806 Thlz opens up a following problem. While for workers' and pensioners' households there was no inconsistency between macro (wages and pensions) data and Survey data, inconsistency Is quite visible in the case of farmers' households. Survey data show that incomes of farmers did not decline as much as AGROR or TOT Imply (18 versus 26 percent, both compared to 1978). Moreover, after 1983, Surveys point to a steady Increase in farmers' per capita real Incomes, while AGROR and TOT data show stagnation or mild decline (see Figure 2). If Survey data are more reliable, the divergence can be explained by an Increase in revenues from non-conventional sources (including the "second economy") which are not captured by macro data. It is also possible that farmers, being (unlike workers) private entrepreneurs, have succeeded to avoid as sharp a decline In their incomes, as suggested by the terms of trade, by displaying greater flexibility in their production decisions. Figure 2 FARMERS: Terms of Trade, Real Income from Surveys and Agricultural Income 140 -fl 1 Terms of Trodo 120- / A gre Ineomo 1976 1979 198O 1981 1982 1983 1984 1966 1966 1967 1966 107s.100 15 5. Conclusions. The economic crisis that started in Poland in 1978 brought about a significant reduction of average incomes of the population (about 20 percent by 1988), and an increase in the percentage of people living below the poverty line (by about 10 percentage points). The composition of the poor also changed: while before the crisis most of them lived In rural areas, majority of the poor (70 percent) are now city-dwellers. The change in composition was due to a severe increase In poverty among socialized sector workers whose real wages declined. Until the end of the period under study (1988) no unemployment appeared. The wage bill was reduced by uniform cuts In real wages with the result that the wage as well the overall income distribution remained practically unchanged. Real Income of pensioners' households decreased almost as much as that of workers. On the other hand, farmers' and mixed households weathered the crisis much better than the other two groups. The explanation behind their relatively good performance seems to lie in a greater flexibility that these households had when undertaking economic decisions (farmers could change crop composition while mixed households could, in addition, vary their labor inputs between the work in socialized industry and private agriculture) rather than In better terms of trade between agriculture and industry, 16 Annex 1. The Definition of the Social Minimum (Poverty Line) For the poverty line we use data on the "social minimum" calculated by the Institute of Labor and Social Affairs attached to the Polish Ministry of Labor and Social Policy. The social minimum is calculated several times per year (generally, quarterly) and for the year as a whole. We use the average annual value since data on household Incomes also refer to the year as a whole. Social minimum was calculated for the first time in Poland in 1980. The change in methodology In 1983 renders only the data for the period 1983-88 mutually comparable. The social minima calculated for 1980 and 1981 are, In real terms, somewhat higher than those calculated after the change. In order to keep an absolute standard of measurement, we have extended back to 1978 the real value of the social minimum In 1982 (see Figure Al below). The social minimum is calculated for for workers' and pensioners' households.i6 For farmers, social minimum was calculated only once, In 1982. According to the researchers In the Institute and some indirect evidence on price levels, the social minimum for rural (both farmers and mixed) households is 10 to 20 percent below the minimum for workers' households. The social minimum includes expenditures for seven types of goods and services: food, clothing and footwear, housing, hygiene and protection of health, culture and education, transport, and, finally, an additional 10 percent (of total expenditures for the previous six groupings). The last item is supposed to be used to defray unanticipated expenses. Among different types of expenditures by far the most Important Is food (between 50 and 55 percent of the total depending on the social group and size of the household). The second most Important Is housing (about 17-18 percent). This percentage Is generally less than In similar minima calculated for Western countries, because of heavily 16 It is calculated for one- and four-person workers' households, and for one- and two-person pensioners' households. In the analysis we use the minimum for one-person household (one adult male = one consumption unit) as our standard. 17 subsidized rents and energy prices in Poland (at least in the period covered by our analysis). The minimum also assumes that the household lives in a municipal apartment where rents are most heavily subsidized (in 1987, subsidy on rent and maintenance amounted to about 80 percent of operating costs ). Although only a quarter of all households in Poland (a third in urban areas) live in such apartments, it is not evident that, for the purposes of the social minimum, the use of this assumption entails a significant under-estimate of the actual housing costs. First, because tenants In cooperative apartments (about one-fifth of all apartments in cities) receive a subsidy which is not much less than that received by households living In municipal apartments. The subsidy takes the form of soft loans or direct covering of maintenance expenditures. Second, if privately-owned apartments (about 40 percent of the total housing stock in cities) are entirely paid up and owned, they do not Involve any monetary costs in addition to maintenance. If apartments are not yet paid up, low interest rates charged on loans render again housing expenditures less than they would be in a market environment. The only category of households for whom the assumption used in the construction of the minimum represents a gross underestimate of expenditures are households that rent apartments from private owners at market rates. A micro analysis of poverty would be needed to determine how many of these households are poor. The social minimum differs from the "existential minimum". The social minimum is a more subjective measure since it includes needs which are considered indispensable at a certain level of development but cannot be shown to be necessary for physical survival. The social minimum incorporates, as its name indicates, a certain "social" consensus about the minimum needs that vary between different societies or, for the same society, between different points in time. For a relatively short period, it is probably acceptable to keep the social minimum fixed in real terms. This is moreover so since the standard of living in Poland was stagnant in the 1980's, and there was little need to revise 17 See Poland: Subsidies and Income Distribution, The World Bank report, November 14, 1989, p. 30. 18 the socially acceptable minimum. Some evidence on the relationship between the social and existential minimum in Poland is provided by comparing the figures supplied by the Institute for Social Affairs with the survival minimum calculated by the "Solidarity" experts. In 1984 and 1985. some researchers from (the then illegal) "Solidarity" conducted calculations on the biological (existential) minimum. This minimum was obtained from the observations of expenditures of 26 poor urban families. The cost of the basket of goods needed for the biological survival was 30 to 35 percent less than the social minimum. Figure Al shows the evolution of the social minimum in current US dollars and in real zloty amounts (1978 prices). By construction, real social minimum Is constant In the period Figure Al Social Minimum in Real Terms (1978.100) and in Current US$ 110 90 60- 30 1978 1979 1980 1981 1962 1963 104 196 1986 1sa6 1968 Dollars - Reel zloty Per eapits and per meoth In 4-mmber woMr 1'w khod 19 1978-82. 18 In calculations of poverty incidence, a necessary complement to the poverty iine are Income statistics. We are using Polish yearly Income and Expenditure Surveys conducted by the Central Statistical Office (GUS). Surveys cover, depending on the year, between 9,000 and 30,000 households. They are widely used both in Poland and abroad, and are reliable. All households in the Surveys are divided Into four social groups: workers, farmers, mixed (worker-farmer households) and pensioners.19 The first and the last group are urban households; farmers and mixed households are rural households. Households in each social group are broken down according to the household size: from 1 to 6 persons for workers, farmers and mixed, and from 1 to 3 persons for pensioners. Finally, all households are grouped Into 7 or 8 (depending on the year) Income groups defined according to per capita household income. For each "cell", i.e. at the Intersection of each social group, income group, and household 18 The reported rate of retail price Increase (RPI) for 1980 and 1981 was raised by respectively 10 and 15 percentage points in order to account for widespread shortages of goods In presence of price controls. Actual prices at which transactions took place were often much higher than the official prices. Intensity of shortages Is reflected In the discrepancy between the price increase of agricultural products sold on the free market and the RPI. In the period 1976-80, the difference between rates of growth of the two indexes was about 5 percent (in favor of free agricultural pilces). In 1980 and 1981, however, the official RPI increased by respectively 9.4 and 21.2 percent, while free agricultural prices rose by 32.3 and 56.4 percent. By adjusting RPI by respectively 10 and 15 points, we implicitly assume that purchases at higher (free market) prices accounted for about 40 percent of total purchases. As If to underline the abnormal situation in 1981-82, for the next five years free market agricultural prices increased, on average, slower than the RPI. 19 Mixed households are those where at least one member Is employed In agriculture while others work outside of agriculture. Definitions of other households are self-evident. When there are both pensioners and workers In a household, the household Is classified In one or other group according to the dominant source of income. Not included in the Surveys are those employed In the private sector outside agriculture, the military and the police. Surveys thus cover approximately 90 percent of all households. 20 size, we have Information on the average number of adult equivalent consumption units.20 For example, the lowest Income group among workers' households with three members would have, on average. 2.2 consumption units. This allows us to consider as poor only those households whose income per consumption unit falls short of the social minimum !similarly defined for one consumption unit).21 If the social minimum falls between lower and upper Income bound of an individual group, households in that group are proportionally allocated among the poor and non-poor households. For example, If lower and upper Income bounds (per consumer unit) are 100 and 200, and the social minimum Is 150, one-half of households In that group will be considered poor and one-half non-poor. We Implicitly assume uniform distribution of households within each Income group. For each year we thus obtain 21 Individual poverty coefficients: 3 for pensioners and 6 for each of the other three groups. These 21 coefficients represent the "building blocks" from which composite poverty coefficients for each social group and the country as a whole are computed. 20 The weighting scheme Is as follows: adult male (18 years of age or more) = 1; male between 14 and 17 years of age = 0.85; adult female = 0.85; female between 14 and 17 years = 0.75; for children, weights are as follows: between 12 and 13 years of age = 0.7; between 8 and 11 = 0.6; between 3 and 7 years - 0.5; 2 years = 0.40; 1 year = 0.3; less than 1 year = 0.25. 21 Strictly speaking, this Is only approximately correct. We treat each "cell" which contains different households as practically a single household. Problems may arise In the following case. Distributions given In Surveys rank all households according to their per capita Income. Rankings of Individual households according to Income per consumer unit would be different. It may happen that a household belongs to an income group which as a group (based on its average Income per consumer unit), Is classified as poor, while that particular household, If It were ranked according to its own Income per consumer unit, would be non-poor. This Is a problem common to all studies that use grouped, Instead of Individual, Income data. 21 Annex 2. Transfers and Poverty: Some Poverty Accounting Let all the population be divided into two groups: agricultural (A) and urban (U). If we then denote the poor In the agricultural sector in period 1 by PAI; Increase in the number of the poor In agriculture as result of population growth (people born In families that are already poor) by npa; transfers of the poor between agriculture and urban areas by ta; total transfers from agriculture to cities by ra; and finally, the nuber of the new poor (people who berore were non-poor and are now poor) In agriculture by NPA, we can write the following identity, where a1 Is the poverty coefficient In agriculture: PA PA + n + NPA - t 1 o pa a a1 ----- (_----------------------- A1 Ao n a - Ta In equation (1), the numerator shows that the total number of the poor in agriculture In period 1 Is equal to their number In period 0 plus increase of the poor due to population growth and decline in Income minus transfers of the poor from agriculture to urban areas. The denominator shows the same relationship for the overall agricultural population (Al=total agricultural population in period 1 and n a=population growth rate in agriculture. Similarly, for the urban households we have relation (2): PU1 PU + n + NPU + aT 1 ----- = ---------------------------- (2) U1 Uo + nu + Ta In equation (2), total number of the poor in urban areas (in the numerator) is equal to their total number in the previous period (PU0) plus increase of the poor due to population growth (n pu) and due to the decline in income ("the new poor", NPU) plus those who transferred from agriculture and are now (or remained) poor. The coefficient a denotes the percentage of transferees who are poor. The denominator of (2) simply shows that total urban population in period 1 is equal to the population in period 0 plus natural growth of urban population (n u) plus transfers from 22 agriculture. We would normally know the following variables: PAI' PAO, As, A0, nat Pul PUo Ul, U0, and n u. Ta can then be calculated. From the assumptions (or information) on population growth as function of income level, we can also get npa and npu. The simplest assumption would be to take that population growth among the poor Is the same as among the non-poor. We are then left with two equations and four unknowns: PAI =PA0 + np +NPA -ta 3 it Po pa a() PU -P0 +npu + NPVU + aT a (4) The unknowns are the new poor among agrlcultural and urban households, poor transferees to the urban sector (t a), and the percentage of all transferees who are now poor In the urban sector (a). If we assume that the transfers from the agro-sector to the urban are uniform across Income groups, the percentage of the poor transferees Is equal to their share in total agricultural population, and the equation (3) becomes: PAI = PAo + npa + NPA- aTa We are now concerned with values taken by coefficient a in equation (4). If transferees do as well as the average person in the urban areas a would be equal to u . If they do, on average, as well as they did before the transfer a=aO. Finally, they can even temporarily do worse than in agriculture and then a=a0+C where C is some (positive) coefficient of adjustment. The shorter the time horizon, the greater C as some people who were not poor In agriculture fall below the poverty level in urban areas, accepting a temporary decline In living standards In expectation of a medium-run improvement (beyond what they would get If they stayed In villages). The normal range of a is shown below. Then for different values of a we can calculate the new poor among urban population. UI a0 a0+C 23 Up to now we have assumed that the transfer function Is uniform across income groups. If, more realistically, we assume that It is an Increasing function of the difference between the average income in cities (the aspiration Income) and the actual Income of a household In agriculture, migration from low income agricultural households would be proportionately greater. This Is shown In Figure A2. Figure A2 Migration as Function of a Difference in Urban and Agricultural Incomes Income Income pattern Migration In agriculture function Average urban Income yu Income groups The gain (decrease in the number of the poor) due to transfers occurs if g, percentage of the agricultural poor in the total number of transferees (t a/Ta), Is greater than a. 1l would be greater than ai If transfers are not Independent of the level of income (but are positively related). a would be greater than u1 if we assume that transferees do not as well as the average urban household (at least between the two points In time which we consider). The "normal" constellation of the variables would then be as follows: gain UI cc a0 However, If (as we observe In East European countries) urban incomes decine significantly so that u1 Increases, while the situation In agriculture remains as before, we could have: 24 gain uI ~~~~a0 The gain is now reduced both because of Increase in a and because fewer migrations take place since decline In average urban income reduces the Intensity of transfers. The worst situation could occur If the decline In urban incomes is accompanied by an increased inequality, so that uI goes up significantly and becomes greater than a0. Since migration decisions are, by assumption, based on average urban Income (and average urban income is still greater than the average rural income and the income of a significant portion of the rural population) migration would continue even if the total number of the poor increases. The analysis can be formalized: [G al (5) T a g [y -ya A (6) f= [G,uYU (7) Equation (5) shows that the percentage of the poor among agricultural transferees is a (positive) function of the degree of inequality (the Gini coefficient) in agriculture. Equation (6) shows that total transfers are a (positive) function of the difference between the average urban and rural income. Equation (7) shows that the percentage of transferees that are (or remain) poor in urban areas Is a (positive) function of Inequality In cities, (negative) function of average urban income, and a (positive) function of C=coefficient of adjustment. Combining these three equations into one, we can write the Net Decrease in the number of the Poor (NDP) as NDP = (O - a) T = 11h(G)- f(G.y.UC)) Ta (Ga) f(Gu'yu )a A =(h(Ga ) - f(GuY-u,C) g[-yu Ya I AO 25 Whether NDP will be positive or negative will, of course, depend on the sign of f-, since we can normally assume that Ta is greater than zero. NDP will be greater the greater G aand yu and the smaller G , C and Ya. The relationship with respect to y is particularly strong because a rise in the average urban Income increases transfers from agriculture and also reduces the percentage of the people In urban areas who fall below the poverty level. However, If Inequality in cities increases simultaneously with the average income, it could happen that the sign of $-a turns negative, and In this case higher urban Incomes, by stimulating rural exodus, might lead to an increase In total number of the poor.22 The situation In Poland between 1978 and 1988 was characterized by practical absence of the urban-rural Income gap (which stemmed the flow of transfers), generally unchanged inequality in the urban areas, and slightly Increasing inequality in rural areas (which might explain that transfers were still positive). While In 1978-79 the average poverty coefficients for rural and urban households were respectively 0.13 and 0.08, there was a reversal In 1987-88 with coefficients equal to 0.14 and 0.22. The difference $- was thus quite possibly negative23 which meant that transfers could have led to an Increase In poverty. 22 The same Is true if the adjustment coefficient, C, increases. 23 Note that for 13 the relevant poverty coefficient is the one in the beginning of the period, 0. 13, and for a one at the end, 0.22. PRE Working Paper Series Contact 1Tn6 ALAhor Do for WPS620 Have Commercial BanKs Ignored Sule Ozler March 1991 S. King-Watson History? 33730 WPS621 Sensible Debt Buybacks for Highly Enrica Detragiache March 1991 S. King-Watson Indebted Countries 33730 WPS622 How Factors in Creditor Countries Sule Ozler March 1991 S. King-Watson Affect Secondary Market Prices Harry Huizinga 33730 tor Devlopin Coumr Deb WPS623 World Bank-Supported Adjustment Vttorio Corbo March 1991 A. Oropesa Programs: Country Performance Patricio Rojas 39075 and Effectiveness WPS624 Choosing Policy Instruments for Gunnar S. Eskeband March 1991 A. Bhalla Pollution Control: A Review Emmanuel Jimenez 37699 WPS625 How Trade and Macroeconomic Ramon Lopez March 1991 K. Cabana Policies Affect Economic Growth 37946 and Capital Accumulation in Developing Countries WPS626 The Macroeconomics of the Public William Easterly March 1991 R. Luz Sector Deficit: The Case of 39059 Colombia WPS627 The Role of Institutions in Poverty Sharon L. Holt March 1991 WDR Office Reduction: A Focus on the 31393 Productive Sectors WPS628 Tl,e Indonesian Family Planning Dov Chernichovsky March 1991 0. Nadora Program: An Economic Perspective Henry Pardoko 31091 David De Leeuw Pudjo Rahardjo Charles Lerman WPS629 An Atheoretic Evaluation of Success Patrick Conway March 1991 D. Ballantyne in Structural Adjustment 37947 WPS630 The Spirit of Capitalism and Long-run Heng-fu Zou March 1991 A. Bhalla Growth 37699 WPS631 The Macroeconomics of the Public Riccardo Faini March 1991 R. Luz Sector Deficit: The Case of 34303 Morocco WPS632 The Macroeconomics of the Public Carbos AHredo Rodriguez March 1991 R. Luz Sector Deficit: The Case of 34303 Argentina PRE Working Paper Sories Contact ALtor pager WPS633 The Macroeconomics of the Public Virabongse Ramangkura March 1991 R. Luz Sector Deficit: The Case of Bhanupongse Nidhiprabha 34303 Thailand WPS634 Trends in Developing Country Bela Balassa March 1991 WDR Office Exports. 1963-88 31393 WPS635 Exchange Rates and Foreign Trade Bela Balassa March 1991 C. Cuskelly- in Korea Young 39413 WPS636 Economic Integration in Eastern Bela Balassa March 1991 C. Cu Ally- Europe Young 39413 WPS637 Poverty in Poland, 1978-88 Branko Milanovic March 1991 WDR Ofice 31393 WPS638 Researching the Trade-Productivity James Tybout March 1991 D. Ballantyne Link: New Directions 37947