ECONOMIC GROWTH, INCOME DISTRIBUTION AND POVERTY IN POLAND DURING TRANSITION Pierella Pacia Marcin J. Sasinb and Jos Verbeekc Abstract * This paper attempts to analyze the linkages between macroeconomic policies and economic growth variables, their movement over time, and their impact on poverty in the case of Poland. Poland, a middle-income country, is of particular interest because its data sources allow for a relatively detailed analysis of such developments, but also because of macroeconomic environment and the economic growth variables show a relatively sizable degree of variance. In addition, Poland has struggled in the last few years to reduce poverty while still experiencing positive economic growth. The paper will show that in Poland poverty-reducing growth depends heavily on the ability of the economy to generate jobs. During the early years of transition, net job growth was positive, but after the Russian crisis of 1998, productivity gains were accomplished mostly through labor shedding, consequently increasing poverty in Poland. In addition, the paper identifies how fiscal and social protection policies impact Poland's income distribution and poverty trends. World Bank Policy Research Working Paper 3467, December 2004 The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org. ____________________________ * This paper was presented at the ABCDE conference in Washington DC (2004). We thank its participants for helpful comments. a The World Bank. Europe and Central Asia Region. Poverty Reduction and Economic Management Unit. bTheWorld Bank. Europe and Central Asia Region. Poverty Reduction and Economic Management Unit. cTheWorld Bank. Africa Region. Poverty Reduction and Economic Management Unit. Introduction. In 1989 the first partially free elections in Poland's post-World War II history took place, and Poland became the first country in Central and Eastern Europe to reestablish democracy and embark on an economic and social transition to a market economy. A "shock therapy" program implemented during the early 1990s enabled the country to transform its economy into one of the most dynamic and robust in the region. Today, Poland's economy is by far the largest among the new EU-member states. It accounts for over 50 percent of the cumulative population of the 10 new member states1 and also about 50 percent of their cumulative GDP. The beginning of Poland's transition in 1990 was marked by exceptionally difficult macroeconomic conditions, which included high inflation, a large legacy of external debt, and a high black market foreign exchange premium. Saddled with a large part of the enterprise sector that was considered "value subtracting," Polish policymakers took huge risks by making the zloty convertible, fixing the exchange rate, and lowering import barriers. With privatization lagging behind, many predicted a crisis based on the notion that enterprises would not be able to cope with market conditions, which would lead to politically unacceptable mass bankruptcy and social upheaval. This did not happen. On the contrary, Poland turned out to be unique among the large front-running European transition countries in having an unbroken growth record once growth resumed after the initial output collapse. This paper attempts to analyze the linkages between macroeconomic and economic growth variables, their movement over time, and their impact on poverty and income distribution. Poland, a middle-income country, is of particular interest not only because its data source permit a relatively detailed analysis of such developments, and also because the macroeconomic environment and economic growth variables show a sizable degree of variance. In addition, Poland has struggled in the last few years to reduce poverty while still experiencing positive but subdued economic growth. A. A SUCCESSFUL MACROECONOMIC PERFORMANCE Economic growth has not been uniform over the last decade. After years of growth above 5 percent or above in the mid 1990s, growth declined substantially to reach a meager 1 percent in 2001, only to increase again in 2003 with good prospects for the coming years (see Figure 1 and Table 1). The slowdown in growth was accompanied by an increase in poverty and unemployment as is evident from Figure 1. This section provides a brief discussion of recent economic developments in Poland. The Polish Central Statistical Office's annual Household Budget Survey (HBS) micro-data sets (1994-2002), as well as publicly available real sector and monetary macro data, have been used to analyze linkages between the macroeconomic environment on the one hand and poverty, inequality, and consumption distribution trends on the other. 1The 10 new member states that joined the EU on May 1 2004 are the Czech Republic, Cyprus, Estonia, Hungary, Latvia, Lithuania, Malta, the Slovak Republic, Slovenia, and Poland. 2 Figure 1: Growth, Employment and Poverty, 1994-2002 Poverty change (%points) Employment (LHS, mn) 6.6 7% GDPgrow th (%y/y) No. of poor (RHS, mn) 16.0 6% mn 6.4 5% mn, 6.2 15.5 ment, 4% poor 6.0 3% 15.0 oyplmE of.oN 5.8 2% 5.6 1% Poverty increase 5.4 0% 14.5 -1% 5.2 Poverty decline -2% 14.0 5.0 1995 1996 1997 1998 1999 2000 2001 2002 1994 1995 1996 1997 1998 1999 2000 2001 2002 Source: Staff calculations based on GUS data. Note: LHS=left-hand side;RHS=right hand side Table 1: Selected Economic Indicators, 1997-2003 Main macroeconomic indicators 1997 1998 1999 2000 2001 2002 2003 GDP growth, % y/y 6.8 4.8 4.1 4.0 1.0 1.4 3.8 Private consumption growth, % y/y 6.9 4.8 5.2 2.8 2.1 3.3 3.1 Gross fixed investment growth, % y/y 21.7 14.2 6.8 2.7 -9.8 -5.8 -0.9 Unemployment rate, e.o.p., % 10.3 10.7 13.8 16.3 19.4 20.0 20.0 Poverty headcount, % of population 14.7 13.1 14.3 14,8 15.6 16.6 . . General gov't expenditures, % of GDP 43.5 41.9 42.7 40.5 43.3 44.0 46.3 General gov't budget balance, % of GDP -2.8 -2.4 -3.1 -3.0 -4.9 -5.9 -6.6 Consumer price inflation, % y/y 13.2 8.6 7.3 10.1 5.5 1.9 0.8 NBP refinancing rate, % 28.0 21.0 21.5 24.0 16.5 9.75 6.0 Current account deficit, % of GDP -3.7 -4.1 -7.6 -6.0 -2.9 -2.6 -2.0 External debt, % of GDP 34.6 35.2 41.6 39.8 37.7 41.7 47.6 e.o.p ­ end of period, 2003 ­ forecast or latest available Source: GUS data. The Factors Behind the Trends in Growth While many factors can account for Poland's growth record - including consistent economic policies despite frequent changes of government - the main economic factor has been that Poland managed the macro-micro policy linkages well.2 At its core, this meant maintaining a combination of hard budget constraints for enterprises, a competitive real exchange rate, and a post-privatization governance structure that allowed businesses, in particular small and medium- size enterprises, to flourish. By reducing fiscal deficits and placing public debt on a stable and sustainable trajectory (including making debt reduction agreements with the Paris Club), Poland achieved a macroeconomic environment that was conducive to growth and that allowed a gradual decline in inflation (see Table 1). However, it is important to note that during the early years of transition the restructuring process was aided by relatively easy access to social safety net programs for redundant workers, and that the dismantling of the overly generous social assistance 2See for example, unpublished memo by Brian Pinto (2002). 3 network of the communist era has not been tackled with the same vigor as have market-oriented reforms. The loss of export markets in the East as a consequence of the Russian crisis in 1998 costs Poland around 3 percentage points of GDP and triggered a new round of enterprise restructuring to curtail falling profitability. This time, the resulting improvements in productivity were brought about in large part through reducing employment. This reduction, together with the increased numbers of newcomers to the labor market owing to the baby boom of the early 1980s, led to significant increases in unemployment. As of December 2002, over 3.3 million people (20 percent of the labor force) were unemployed. These developments coincided with a shift in monetary policy in 1999, leading to the creation of an independent Monetary Policy Council (MPC). Inflation targeting became the main anchor of monetary policy, one of the means of facilitating accession to the EMU. When inflation rebounded in 2000 into double digits, the MPC progressively tightened monetary policy by raising its rediscount rate to a peak of 21.5 percent in August 2000. As a result, real lending interest rates increased to over 13 percent during 2001, up from around 10 percent during 1999- 2000. In combination with possible over-investment during the high growth years of 1996-99, this led to a significant decline in domestic demand, and especially in investment, from the second half of 2000 onwards. At the same time, inflation declined impressively from 2001 onwards to less than 1 percent in 2003. Figure 2: GDP Growth, Export, and Industrial Performance Real growth y/y, % GDP grow th Industrial Production and Export Growth 7% 12 30 6% 10 25 Real growth y/y, % 5% 8 20 4% h 6 15 ction 3% 4 10 grpowt produ 2% 2 5 port 1% Ex0 0 strial Indu 0% -2 -5 06 09 12 03 06 09 12 03 06 09 12 03 06 09 12 03 1999 1999 1999 2000 2000 2000 2000 2001 2001 2001 2001 2002 2002 2002 2002 2003 GDP growth, % y/y 4Qts moving average Industrial Output (RHS) Export index growth (LHS) Source: Staff calculations based on GUS data. The macroeconomic environment deteriorated progressively at the end of the last decade. While monetary policy tightened during the second half of 2000, fiscal policy eased considerably during 2001. Not only did the general government balance widen from ­3.0 percent for 2000 to -5.9 percent for 2002, but the structural deficit (i.e., corrected for business cycle effects) worsened by over 2 percent to over 5 percent of GDP by 2002 and 2003 (see Figure 3). This "tight monetary and loose fiscal" policy mix has additionally crowded out investment and moderated the potentially expansionary effect of fiscal policy. 4 Figure 3: Fiscal and Monetary Developments, 1996-2003 1 Structural De ficit (% of GDP) M one tary Conditions Inde x 2.5 0 2.0 -1 1.5 -2 1.0 -3 0.5 Monetary tightening -4 0.0 -5 -0.5 -6 -1.0 Monetary easing -7 -1.5 1996 -2.0 19971998 S tru c tu ra l 1999 -2.5 2000 2001 96M1 96M8 97M3 97M10 98M5 98M12 99M7 00M2 00M9 01M4 01M11 02M6 03M1 H e a d lin e 2002 2003 Headline Cyclical Structural B. DID POVERTY RESPOND TO GROWTH? It is clear that the high growth rates of the mid-1990s to 1998 facilitated a decline in poverty in Poland, which suggests high responsiveness of poverty to economic growth. From 1999 onward, however, this responsiveness appeared to weaken; poverty began to increase even though growth remained at 4 percent for 1999 and 2000 and being modestly positive thereafter. A Statistical Artifact? To fully evaluate the role of growth in poverty reduction, it is important to note that household consumption ­ as captured by the HBS ­ performed much more badly than the national accounts (NAS) would suggest, with average household consumption declining in the low ­ but positive -- growth years after 1998. This difference persists even when the figures for changes in aggregate private consumption from NAS are considered rather than GDP growth. This is not a phenomenon unique to Poland and may be explained by the unusually high investment accumulation over the 1990s, and by its relatively high export rate.3 Reconciling these two series is beyond the scope of this report, but it is important to note that the HBS shows declining values of average per capita consumption for the period 1999 ­ 2002 while the NAS data show moderate positive growth.Thus, the estimates of poverty, which are based on the HBS, show increases in poverty in these years despite the positive consumption growth rates shown by the NAS data. When the responsiveness of poverty to growth in average consumption (rather than to GDP growth) is considered, the analysis shows a very strong link between growth and poverty reduction. Over the last decade the average estimated elasticity of poverty with respect to growth in average consumption from the HBS data has been 3.6 with a peak of 4.11 in 1999.4 Given Poland's GDP level, this value compares very well with the estimates for other transition countries and with those typically found in countries outside of the region (Bruno, Ravallion, and 3Adams (2003). 4 These estimates are in line with the value of 3.5 given in World Bank (2002). 5 Squire,1998; Ravallion and Chen,1997).5 To sum up: a 1 percent growth in mean consumption brings about a 3 percent reduction in the headcount poverty rate which is equal to about 0.5 percent of the total population or (multiplying by 38.2 million) 185,000 people. Given the relatively high degree of elasticity and the optimistic predictions for growth in 2004, the prospects for poverty reduction in Poland are good. Assuming inequality to increase at the average rate experienced over the last five years, poverty should fall by 6.4 percent, lowering it by 1.5 percentage points to 15.1 percent in 2004 (see table 1). But Inequality Matters, Too However, inequality also rose steadily through the second half of the 1990s (see Figure 4). In 2002 the Gini Figure 4: Inequality, 1994-2002 coefficient for consumption inequality was 0.28, which 0.29 Gini coeff, LHS 0.14 was moderately high compared to other CEE countries 0.28 Mean log deviation, 0.13 0.27 (see, for example, World Bank, 2000 and 2002). Given RHS 0.12 0.26 the formerly centrally planned economic systems in CEE 0.11 0.25 countries and their subsequent economic development, it 0.10 0.24 is not surprising that growth is associated with rising 0.09 0.23 0.08 inequality. In general, however, CEE countries constitute 0.22 0.07 an outlier in the overall picture ­ any simple universal 0.21 0.20 0.06 link between growth and inequality has been flatly 1994 1995 1996 1997 1998 1999 2000 2001 2002 rejected by recent studies (see, for example, Ravallion Source: Household budget survey and Chen, 1997). Table 2 decomposes the overall inequality ­ measured by the Theil mean log deviation index - into two components: (i) differences between groups and (ii) within group inequalities. The focus is on variations within (and between) (i) regional groups (96 groups = 16 voivodships crossed with 6 town sizes), (ii) educational groups (3 groups: higher, secondary, other) and (iii) age categories. If, for example, mean consumption was the same for all regions (i.e., the region variation was 0), then the numbers in the respective row in the table would be 0 and all inequality could be entirely attributed to inequalities within each region. Analogously, if everyone within a given region had equal consumption (i.e., within-region variation was zero) but regions differed in mean consumption, the "between" inequality component would be 100 percent. The table shows that inequality caused by variation within groups is the main component of total inequality for all groups analyzed. However, differences between different educational groups are significant and growing component of overall inequality and are differences between different regions. On the other hand, differences across different age groups were and remain minimal. Thus, regional inequalities clearly increased during 1994-2002, not only in absolute terms6 but also in relative terms. Or, equivalently, regional inequalities rose even more than total inequalities and the role of regional inequalities in overall inequality increased. The increase in the primary education was an even more substantial signifer rising inequality. 5 Just to give few examples, the growth elasticity of poverty for Romania was estimated at 3 in 1996 and 1997 but has declined since than to below 2 (World Bank, 2003). An average elasticity of 3.7 was also estimated for Croatia (World Bank ,2002). However, the limited information available suggests much lower responsiveness in other countries. Estimates of around 2 were derived for Russia (World Bank, 2002) and less than 1 for Armenia, Azerbaijan and the Kyrgyz Republic (World Bank, 2003). 6See Chapter 3 of Poland Living Standard Assessment, World Bank (2004) for a discussion of regional poverty and prices 6 Table 2: Decomposition of Theil Inequality Index (Inter-Group Inequality), 1994-2001 1994 1995 1996 1997 1998 1999 2000 2001 Theil index 0.100 0.095 0.097 0.107 0.110 0.114 0.120 0.120 Decomposition Within groups 0.092 0.087 0.089 0.097 0.097 0.100 0.104 0.106 Regional Between groups 0.008 0.008 0.008 0.010 0.013 0.014 0.016 0.014 Share of between groups 7.9% 8.3% 8.1% 9.4% 11.5% 12.5% 13.0% 11.6% Within groups 0.087 0.083 0.084 0.092 0.093 0.095 0.097 0.097 Education Between groups 0.013 0.012 0.130 0.015 0.017 0.020 0.023 0.023 Share of between groups 13.0% 12.9% 13.4% 13.8% 15.9% 16.5% 19.5% 19.1% Within groups 0.098 0.093 0.095 0.105 0.107 0.111 0.118 0.118 Age Between groups 0.002 0.002 0.002 0.002 0.003 0.003 0.002 0.002 Share of between groups 1.7% 2.2% 2.2% 2.0% 2.5% 2.4% 1.5% 1.7% Source: Authors' calculations based on HBS. Generally speaking, when inequality increases (if other things are kept constant), poverty normally increases as well, thus dampening the beneficial effect of growth on poverty. Fortunately, changes in the shape of the distribution of consumption are much slower than the change in the mean. Therefore, the positive growth impact should outweigh the negative effect of a widening of the distribution itself. Decomposing Trends in Poverty Having described the trends in growth and in inequality, it is important to analyze the effects of these trends on poverty. Table 3: Decomposition of Changes in Poverty, 1995-2002 (percentage points) Poverty change 1995 1996 1997 1998 1999 2000 2001 2002 Total -0.4% -1.5% -0.3% -1.6% 1.3% 0.5% 0.8% 1.0% Due to growth 0.4% -2.1% -1.1% -2.0% 0.5% 0.4% 0.6% 0.6% Due to inequality -0.9% 0.3% 0.8% 0.5% 0.8% 0.0% 0.2% 0.5% Source: Staff calculations based on GUS's data. The results of a decomposition of changes in poverty for Poland into their growth and inequality components are presented in Table 3 and Figure 57 It is evident that economic growth was a main driving force in poverty reduction. Two periods are readily distinguishable. During Poland's high growth years, 1994-98, the economic growth effect (5 percent) far outweighed the negative effect of the increase in inequality (-1 percent), resulting in a cumulative decline in poverty of about 4 percentage points. During the subsequent period, 1998-2002, and because of a decline in mean consumption, lack of growth and increased inequality jointly resulted in a cumulative increase in poverty of about 3 percentage points for that period. 7For a description of the methodology used, see Bourguignon (2002). 7 Figure 5: Decomposition of Changes in Poverty,1994-2002 3% 2% 1% 0% -1% -2% -3% -4% -5% total poverty change due to grow th -6% due to inequality 1994-1998 1998-2002 1994-2002 Note: The 2002/1994 change is approx. a sum of 1998 ­2002 and 1994-1998. Source: Staff calculations based on GUS data.. For further illustrative purposes, one can look at the consumption level and its dynamics at different parts of the consumption distribution. It is interesting to compare mean incomes/consumption across the ranked population. Inverting the cumulative distribution of consumption at a p-th percentile, one can infer consumption levels at this percentile. This line ­ depicted in the left panel of Figure 6 ­ is called a quintile function, or, sometimes, a Pen's parade (Pen, 1971). For example, individuals at the 70th percentile had in 2002 a consumption level of almost PLN750 per month. The left panel of Figure 6 presents three curves for three different years (1994, 1998, and 2002). Between 1994 and 1998 the line shifted upward almost in parallel, implying a rise in consumption across the entire population. However, between 1998 and 2002 the line generally shifted downward (except for the higher percentiles). The magnitude of this decline was clearly greater for the lower percentiles. The consumption level of the poor returned to, or fell below, its 1994 level. The right panel in Figure 6 presents the same story but from a dynamic perspective. The period 1995-98 was generally favorable for all the consumption groups,8 but in 1999 an across- the-board deterioration emerged, with only the top quintiles maintaining a barely positive level of consumption growth (except in 2001). 8There could be two explanations for the 1997 performance of the lower quintiles IN 1997: a tightening of social benefits eligibility, and massive flooding in the summer. 8 Figure 6: Consumption and Growth in Mean Consumption by Percentiles/Quintiles, 1995-2002 1339 The Pen's parade: Consumption growth for selected 7.1 consumption by percentile quintiles 1130 t 6.9 1994 5% 1 953 adul 1998 2 6.7 4% 5 804 eq./h 2002 3% 679 6.5 ontm/ 2% 572 6.3 NLP ` 1% 483 6.1 0% 407 5.9 -1% 344 5.7 -2% 290 percentile 5.5 -3% 0 10 20 30 40 50 60 70 80 90 100 1995 1996 1997 1998 1999 2000 2001 2002 Source: Staff calculations based on GUS data. Figure 7: Growth Incidence Curve It is straightforward to extend the idea of the Growth incidence: consumption Pen's parade to show the change in mean increase by percentile consumption (rather than ITS level) for various 15% percentiles of the consumption distribution during the period of interest. This approach, 10% which uses the so-called growth incidence curve 5% (gic), is meant to show how gains from aggregate economic growth are distributed 0% among the population.9 Figure 7 shows growth incidence by percentile. -5% 1998/1994 It should be noted that there was almost no -10% 2001/1994 change in the consumption level at the 15th 2001/1998 percentile between 1994 and 2002. However, -15% the consumption level at the 90th percentile and 0 10 20 30 40 50 60 70 80 90 100 higher grew by over 10 percent over this period. Note: The 2002/1994 change is approximately a sum of 2002/1998 and 1998/1994 These results are consistent with what has Source: Staff calculations based on GUS data. already been identified. For example, the curves are (almost) regularly increasing, meaning that inequality has risen at all levels of distribution. During 1994-98 the curve was above zero everywhere, indicating that poverty has fallen, whatever the poverty line and whatever (reasonable) poverty measure is adopted. Although, 9Note: The curves should not be interpreted as a growth in consumption of an average p-percentile Mr. X during that period. The cross-sectional character of the exercise implies that the graph rather depicts the difference in mean consumptions of two groups of individuals that happen to be in the same percentile at both ends of the period of interest. We are comparing the poor in year A with the poor in year B, not asking whether they are the same people, thus (leaving aside mobility within the distribution) we underestimate the improvement for individual poor people and overestimate the growth of rich people. 9 technically speaking, the growth could be described as moderately pro-rich (higher growth at the higher quintiles), it brought significant benefits to the poor as well. During 1998-2002, the situation changed radically: only the highest decile managed to maintain its consumption level and the reduction was hardest for the poorest percentiles. In addition to the discussed indicators, Ravallion and Chen (2003) derive an intuitive, straightforward index to measure pro-poor growth: it is equal to mean growth (change in (h) consumption) for the poor:10 = gic(p)dp (h) p where (h) is a percentile where 0 the poverty line is drawn (equal to the poverty headcount). This index can then be compared with an average growth to infer whether the growth is pro-poor. Table 4 presents the comparison. Table 4: Pro-poor Growth Growth 1995 1996 1997 1998 19999 2000 2001 2002 Average -0.7% 3.5% 2.1% 4.2% -0.9% -0.9% -1.0% -1.0% Pro-poor 0.3% 3.2% -0.9% 3.2% -2.1% -2.8% -2.3% -2.1% Source: Staff calculations based on GUS data. C. WHAT IS BEHIND THE INCREASE IN CONSUMPTION INEQUALITY? Changes in inequalities in consumption can be the result of a number of factors. Four important ones are: (i) changes in the distribution of original income; (ii) changes in the progressivity of the tax and benefit system; (iii) changes in the distribution of savings; and (iv) changes in relative prices of the basket of goods bought by households at different points along the income distribution. Unfortunately, the Polish HBS has very limited information on household savings. Thus, the analysis below focuses on the other three factors identified above. Changes in the Distribution of Original Income and its Components. Original income refers to the income that the household has before paying taxes and receiving benefits (other than old-age pensions). As shown in Table 5, over the period 1994-2002 inequality in original income, as measured by the Gini coefficient, increased by around 7 percent, from 0.38 in 1994 to 0.41 in 2002. The increase was surprisingly moderate considering the degree of transformation the country had experienced, particularly since it occurred entirely after 1998 with inequality in original income fluctuating around a stable value in the high growth years of 1994- 98. Table 5 also shows the main sources of inequality and the main factors behind the increase. Inequalities in labor income were and remain, a major source of total income inequality. In addition their relative role in overall inequality has increased over time until 2001 (from 70.5 percent to 72.3 percent) although it appears to have declined back to 70 percent in 2002.11 The combination of increased inequality and the growing size of the component attributable to labor 10 This index, as Ravallion and Chen (2003) note, has an additional desirable property of being based on an (h ) "axiomatically appropriate" poverty measure, namely, the Watts index:W = ln( H / c( p )) dp. 0 11The nature of a decline in 2002 is not clear. 10 income dispersion resulted in a increase in the concentration coefficient for this component income of 8 percent between 1994 and 2002, higher than for overall income and for any other component. To conclude, an important source of the increase in inequality of consumption over the 1994- 2002 period as discussed above can be found in the growth of inequality in original income. The main force behind this increase has been the growth of inequality in labor income, with is reflected in both a higher concentration coefficient for this type of income and the increased role played by this component in the overall measure of inequality. Table 5: Income Inequality and Decomposition by Income Sources, 1994-2002 Gini Coefficients: 1994 1995 1996 1997 1998 1999 2000 2001 2002 Original Income* 0.379 0.377 0.386 0.395 0.366 0.377 0.395 0.395 0.407 of which: Labor income 70.5% 68.9% 70.8% 71.6% 72.1% 73.7% 74.1% 72.3% 69.9% Old age pension 15.1% 16.5% 15.9% 16.0% 16.4% 16.4% 15.3% 17.8% 18.8% Income from farm 10.1% 11.6% 9.3% 8.9% 7.6% 6.4% 6.2% 5.7% 6.6% Other income 4.3% 3.0% 4.1% 3.5% 3.9% 3.5% 4.4% 4.3% 4.7% Disposable Income** 0.308 0.304 0.312 0.323 0.293 0.298 0.311 0.308 0.317 Concentration Coefficients***: Labor Income 0.389 0.382 0.396 0.402 0.377 0.398 0.420 0.414 0.421 Old age pension 0.394 0.408 0.404 0.410 0.380 0.385 0.372 0.407 0.423 Income from farm 0.332 0.358 0.326 0.371 0.290 0.273 0.292 0.287 0.338 Other income 0.307 0.254 0.320 0.289 0.293 0.265 0.304 0.280 0.295 Method: Decomposition of the Gini coefficient into components, ranking by total income. * Before taxes and benefits. ** After taxes and benefits *** Household ranked by original income. Source: Authors' calculations based on HBS. The Role of Taxes and Social Benefits While increased inequality in original income is clearly an important component of the dispersion of consumption, it is not the full explanation. As the second part of Table 5 shows, the growth in the Gini coefficient for original income (7 percent) was just over half of the increase in the Gini coefficient for consumption (13 percent). Thus, other factors must be at play. Amongst these factors, of potential importance are changes in the tax and benefit system which, for any given distribution of original income, may affect the distribution of the income that households actually have to spend (i.e., their disposable income). The top section of Table 6 shows that the concentration index for taxes increased by 32.8 percent in the period 1994-2002, which suggests that the tax burden has become more concentrated. At the same time, the social benefits system evolved toward the poor more strongly. This would suggest that since the mid-1990s changes in the tax and benefit system have counteracted the increased inequality in original income, leading to a much smaller growth of inequality in disposable income (only 3 percent, from 0.308 to 0.316). 11 Table 6: The Role of Taxes and Social Benefits in Inequality Trends, 1994-2002 1994 1995 1996 1997 1998 1999 2000 2001 2002 Concentration Coefficients*: Taxes 0.311 0.307 0.318 0.328 0.322 0.367 0.386 0.400 0.413 Benefits -0.270 -0.316 -0.318 -0.334 -0.374 -0.367 -0.371 -0.365 -0.362 Progressivity**: Taxes -0.070 -0.072 -0.071 -0.069 -0.045 -0.012 -0.010 0.002 0.004 Benefits -0.650 -0.695 -0.707 -0.71 -0.72 -0.746 -0.767 -0.762 -0.771 Method: * Household ranked by original income. ** Kakwani index. Source: Authors' calculations based on HBS. An alternative way of looking at the effect of fiscal policy on poverty is to consider its redistributive effect. This can be done by comparing the Lorenz curve for original income with the concentration curves for taxes and benefits. Figure 8 compares the redistributive impact of the system in 1998 and 2002. The comparison points to some interesting conclusions. The first surprising result is that in 1998 the concentration curve for taxes dominated the Lorenz curve for original income (i.e., it was above the Lorenz curve for all points in the income distribution). This suggests that the overall tax system was regressive (i.e., that the poor paid more than the better-off as a proportion of their income). However, since the distance between the two curves was very small, the magnitude of the redistribution involved was very limited. In 2002 the situation for the poorest individuals had not changed. The share of taxes that they paid exceeded their share of income, which pointed to a redistribution away from the poor. However, in 2002 the concentration curve for taxes crossed the Lorenz curve, making impossible to derive firm conclusions about the overall level of redistribution. Figure 8: Progressivity of the Tax and Benefit System in 1998 and 2002 100 100 1998 2002 90 latot 90 80 of 80 latot 70 %, 70 of ount 60 60 %, 50 50 ount 40 amevitalu 40 30 muc 45 degree line 30 45 degree line Original income 20 20 amevitalumuc Original income Taxes 10 B enefits Taxes 10 Benefits 0 cumulative % of population 0 cumulative % of population 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 It is in cases like this that it becomes important to use an index of progressivity such as the Kakwani index, which measures (twice) the area between the Lorenz curve for pre-tax income and the concentration curves for tax liabilities and received benefits.12 Table 6 above gives the values of the Kakwani indexes for taxes and benefits over the period 1994-2002. These values 12 Kakwani (1977). 12 confirm the intuitive conclusions derived from Figure 8. In 1994 the Polish direct tax system was slightly regressive but the degree of regressiveness declined over time until the system became slightly progressive in 2001. The benefits system, on the other hand, was always highly pro-poor and grew increasingly so over the period under consideration (the Kakwani index increased by nearly 2 percent). However, much of the change occurred between 1994 and 1998 (an increase of 14 percent). The latter is a most interesting result, as it suggests that households are increasingly dependent on social transfers to finance their consumption expenditures. Although work income is the most important source of financing for expenditures, it clearly varies across the expenditure distribution (see Figure 9). For 2002, its share varies from 41 percent for the first decile to 68 percent for the tenth decile. Even for the bottom quintiles, income from work remains the main source of financing consumption expenditures (see Figure 10). Figure 9: Comparison of Income Composition: by Decile in 2002 (left) and over 1994-2002 for the Poor (right) Income Composition by Decile, 2002 100% Income Sources of the Poor 350 90% PLN / month/ eq. adult 80% 300 other 70% 250 agric 60% 200 50% o.a.pension 40% own food 150 30% social transf100 20% 50 10% work 0% 0 1994 1995 1996 1997 1998 1999 2000 2001 2002 d1 d2 d3 d4 d5 d6 d7 d8d9 d1 0consumption deciles Source: Staff calculations based on GUS data. Figure 10: Poverty and Social Transfers, 1994-2002 Social transfers per capita % of population relying on transfers 104 23% to close their poverty gap 102 22% 100 21% 98 20% 96 19% 94 18% 92 90 17% 1994 1995 1996 1997 1998 1999 2000 2001 2002 19 19 19 19 19 19 20 20 20 94 95 96 97 98 99 00 01 02 Source: Staff calculations based on GUS data. 13 However, Figure 10 shows that the number of people relying on social transfers to close their poverty gap increased steadily after 1998, approaching 20 percent in 2002. This means that headcount poverty would have been 20 percentage points higher if social assistance had not been available as a safety net. Even though social transfers (excluding old age pensions) have a direct poverty reducing effect, the system is extremely costly and inefficient. Calculations based on Household Budget Surveys estimate the total amount of social transfers in 2002 at 6.3 percent of GDP, which is close to what government statistics report. However, 56 percent of this amount (or 3.6 percent of GDP) is not reaching the poor: 26 percent is in excess of the amount necessary to bring household consumption to the poverty line level (called "overflow"), while 30 percent (an equivalent of 2 percent of GDP) goes directly to non-poor households (see the left-hand panel of Figure 11). Although benefit tightening has been tried on several occasions, the amounts leaking were of a similar order of magnitude throughout the analyzed period. Figure 11: Targeting of Social Transfers (Excluding Old Age Pensions) Who receives social transfers ? Social transfers per capita by consumption deciles % of resources going to: poor non-poor overflow d1 d10 d9 d8 d7 26% 44% esl d6 deci no d5 d4 30% pti um d3 d2 cons PLN / d1 month / 0 eq. adult 50 100 150 Source: Authors' calculations based on HBS The right-hand panel of Figure 11 illustrates the incidence of social transfers (excluding old age pensions) from an ex post rather than an ex ante perspective and reveals a disturbing pattern: there is very little difference in the actual nominal per capita amount of social transfers across the expenditure distribution. Whatever quintile is evaluated, whether rich or poor, each (equivalent) individual receives on average 120-140 PLN per month. And all deciles except the 9th and 10th receive higher per capita transfers than the poorest decile. If these resources were applied with perfect targeting, rather than being spread more or less uniformly over the distribution, they would be more than double the amount needed to eradicate poverty . The conclusion, therefore, has to be that Poland has a costly and inefficient system for distributing social transfers; less than half of which serves to bring household consumption up to the poverty line level. A major proportion of these resources is not going to the most needy. However, it might be the case that poverty reduction is not the only objective of this redistribution, i.e. some of the transfers serve an insurance function and are not necessarily designed to improve the well-being of the poorest. Another 14 objective ­ given the even spread of these resources ­ could be the improvement of the well- being of the "median voter" in order, for example, to win support for government policies and maintain social consensus. Inflation and Exchange Rate Movements and Expenditure Composition It should be borne in mind that inflation is a regressive tax, the burden of which is typically borne disproportionately by the poor. For example, the poor tend to hold most of their financial assets, if any, in the form of cash. In addition, when inflation reaches a certain threshold it negatively affects investment and output growth and therefore reduces employment opportunities. The composition of inflation is also frequently regressive: for example, it is well known that food prices often tend to rise faster than the overall CPI ­ a situation that disproportionally affects poorer households. Large exchange rate depreciations also tend to hurt the poor more than others, as, again, their consumption basket tends to contain a larger component of tradables than that of richer households; moreover, such depreciations normally negatively affect inflation. The composition of food and non-food expenditures changed somehow between 1998 and 2002 (see the left panel of Figure 12) with every income group spending less of its consumption on food in 2002 compared to 1998. During this period the food price index changed 15 percentage points less than the overall CPI (right panel of Figure 12), indicating that less of one's income needed to be used to provide for one's daily bread and butter. Apparently, the income effect outweighs the substitution effect. The divergence in average and food price indexes means that food prices remain important in poverty measurement.13 Figure 12: Composition of Food and Non-food Expenditures and CPI and Food Price Developments, 1998-2002 Food as % of consumption 70 16 CPI and Food Price developments 1998-2001 135 14 60 130 1998 2002 12 50 10 125 8 40 120 6 30 115 4 110 20 2 0 105 10 -2 100 0 -4 Annual CPI e.o.p. Annual Food e.o.p 1 2 3 4 5 6 7 8 9 10 -6 CPI-Index (RHS) Food-Index(RHS) 95 Consumption deciles July: 1998 1999 2000 2001 2002 The same situation applies to developments in the real effective exchange rate and the composition of tradables versus non-tradables in each decile's expenditure pattern (see Figure 13). The appreciation of the zloty during the latter part of the period induced all expenditure groups to spend less on tradables and more on non-tradables, indicating once again that the income effect dominates the substitution effect. 13 For further discussion of this issue, see Chapter 3 of Poland Living Standard Assessment, World Bank (2004). 15 Figure 13: Composition of the Expenditures of Tradables versus Non-tradables and REER Development, 1998-2001 35 Non-tradables as % of consumption 25%Real effective exchange rate developments 135 1998-2001 30 1998 2001 20% 130 125 25 15% 120 115 20 10% 110 5% 15 105 0% 100 10 95 -5% 90 5 -10% De M Jun- Sep- De M Jun- Sep- De M Jun- Sep- De M Jun- Sep- De 85 0 1 2 3 4 5 6 7 8 9 10 c-97 ar-9 98 01 8 98 c-98 ar-9 99 9 99 c-99 ar-0 00 0 00 c-00 ar-0 1 01 c-01 Consumption deciles Annual Change (e.o.p) (LHS) REER Index (RHS) D. POLICY RECOMMENDATIONS In the first section we argued that macroeconomic stability is a key component of a growth- promoting environment and therefore, is indirectly a foundation for any successful poverty reduction strategy. The link is that macroeconomic stability encourages investment and promotes productivity growth and employment creation. In this sense, macroeconomic stability is a public good. Since the late 1990s Poland's macroeconomic environment has been relatively volatile and, unfortunately, characterized by inappropriate loose fiscal/tight monetary policy mix. A labor-shedding restructuring of firms that was not matched with an environment conducive to employment creation by the private sector (particularly by small and medium enterprises) on the demand side, as well as unchecked access to social safety nets and the recent baby boom on the supply side, have contributed to a dramatic worsening of labor market conditions, a consequent rise in unemployment, and a resulting increase in poverty. A return to robust growth (the main factor in poverty reduction) requires a stable macroeconomic environment with a balanced fiscal-monetary policy mix. To prevent the re-emergence of economic imbalances, fiscal consolidation and the reduction of overall deficits are needed to complement the monetary easing since 2000. A parallel reduction in the size of the government budget would provide greater space for private sector development. Together with an additional and necessary redirection of government expenditures from consumption to investment, these policies would significantly strengthen Poland's growth prospects. In order for growth to be poverty-reducing, the link between economic development and labor market improvement must be strengthened significantly (or perhaps simply re-established). In other words, the growth environment must be made more labor-friendly. In the short to medium term this would involve a necessary reduction in the tax wedge (payroll taxes) as well as an increase in the flexibility of the labor market. In the longer run, policies are needed that will close the mismatch of skills between labor demand and supply, which means promoting investment in human capital and education. Recent World Bank studies, particularly the Public 16 Expenditure and Institutional Review (World Bank, 2003) and the Labor Market Study (World Bank, 2001), have explored the agenda and the prospects for policy reforms in these areas. Their recommendations are still applicable. Price stability is an essential component of a poverty reduction policy. Inflation predominantly hurts the poor, who have little opportunity to protect their assets. For this reason, Poland's low inflation in general, and particularly its low food price inflation should be perceived as a genuine achievement of recent years. In addition, the decline in nominal and real interest rates brought about by the monetary easing should work to the advantage of the poor by increasing their capacity to smooth consumption. Fiscal policy works best if it is pro-poor. Since the poor have a higher propensity to spend, leaving more resources with them during hard times provides an effective tool against cyclical downturns of the economy, contributes to the well-being of the vulnerable, and feeds back positively to overall macroeconomic stability. The tax component of fiscal policy appears to perform its anticipated stabilization role. Any changes in the system to improve its pro-poor impact should be considered carefully, so that they do not further increase the distortions that taxes bring to the economy, and especially to ensure that they do not increase the overall tax burden. The elimination of exemptions (tax expenditures are directed mainly toward the rich) complemented by an overall reduction in payroll taxes in at least a budget neutral way is a plausible recommendation. Although Poland's system of social transfers plays an important role in poverty reduction it is more than costly and inefficient. Over half of the resources used is not going to the most needy. If these funds were targeted properly, rather than being spread more or less uniformly over the entire consumption distribution, they could more than eradicate poverty. Therefore, improving the targeting of social transfers is an immediate necessity. Poor targeting not only leaves many of the vulnerable without the help of the state; it also skews incentives to work and poses a significant burden on government finances. As has been demonstrated, Polish society is still changing. The Polish economy is growing but at the same time becoming more polarized. Therefore, there is a strong case for good poverty/social monitoring that involves going beyond aggregations and applying more complex analytical tools, such as various decompositions, growth incidence, benefit incidence, etc. Such monitoring can provide both the public and policymakers with a better knowledge of whether and how society is changing, and the ability to identify the sources of such changes. It can also make it possible to assess winners and losers, and, in particular, it can become a basis for judging whether current and future inequalities are acceptable. Moreover, an ex ante assessment of the impact of social programs on income distribution and poverty could become a valuable policy tool. Despite a very modestly positive GDP growth of about 1 percent in 2001 and 2002, poverty has been increasing. In 2003, Poland stands at the threshold of EU accession with all its related challenges and opportunities, while at the same time economic recovery is accelerating. Therefore, actions are required to improve the pro-poor impact of economic growth and to make this growth work for Polish society in general and for its poor in particular. 17 REFERENCES: Adams, R. (2003), Economic Growth, Inequality and Poverty: Findings from a New Data Set, World Bank Policy Research Working Paper 2972, Washington, D.C. Bourguignon, F. (2002), "The Growth Elasticity to Poverty Reduction: Explaining Heterogeneity across Countries and Time Periods," in T. Eicher and S. Turnovsky (eds.), Growth and Inequality, Cambridge, Mass: MIT Press. Bruno, M., M. Ravallion and L. Squire (1998), "Equity and Growth in Developing Countries: Old and New Perspectives on the Policy Issues," in Vito Tani and Ke-Young Chu (eds.), Income Distribution and High Growth, Cambridge, Mass: MIT Press. Kakwani, N.C. (1977), Measurement of Tax Progressivity: An International Comparison. Economic Journal, 87, 71-80. Pen, J. (1971), Income Distribution, New York: Praeger Publishers. Pinto, B. (2002), Poland ­ Macro-Micro Policy Linkages, Unpublished World Bank Mimeo. Ravallion, M. (1995), "Growth and Poverty: Evidence for Developing Countries in the 1990s," Economics Letters, 48. Ravaillon M. (2001), Measuring Aggregate Welfare in Developing Countries: How Well Do National Accounts and Surveys Agree?, World Bank Policy Research Working Paper 2665, Washington, D.C. Ravaillon, M., and S. Chen (1997), "What Can New Survey Data Tell Us about Recent Changes in Distribution and Poverty?", World Bank Economic Review, 11(2). Ravaillon M., and S. Chen (2003), "Measuring Pro-poor Growth," Economic Letters, 78. Shorrocks, A. (1982), "Inequality Decomposition by Factor Components," Econometrica, 50 (1). Theil, H. (1967), Economics and Information Theory, North Holland, Amsterdam. World Bank (2000), Making Transition Work for Everyone: Poverty and Inequality in Europe and Central Asia, World Bank, Washington D.C. World Bank (2001), Poland's Labor Market: The Challenge of Job Creation, World Bank, Washington D.C. World Bank (2002), Transition: The First Ten Years, World Bank, Washington D.C. World Bank (2003), Poland: Towards a Fiscal Framework for Growth: A Public Expenditure and Institutional Review, World Bank, Washington D.C. World Bank (2004), Poland Living Standard Assessment: How far is Warsaw from Lisbon, World Bank, Washington D.C. 18