Poverty and growth in Romania: *

Transcription

1 Poverty and growth in Romania: * Roberta Gatti INDEX Executive summary 1. Introduction Methodological note Decline and growth in Romania: macroeconomic context and growth patterns for the period Poverty and Growth How much did growth contribute to poverty changes? Results from the growth-redistribution decomposition The distribution of gains from growth: growth incidence curves Are poverty changes due to intra sectoral (regional) poverty reduction or to population shifts? Evidence from sectoral and regional decomposition of poverty Sources of growth Poverty projections Conclusions..78 Appendix..80 * I thank Nicholas Burnett, Catalin Pauna, Peter Lanjouw, Cem Mete, Lucian Pop, Emil Tesliuc and, in particular, Michael Lokshin for useful conversations. Denis Nikitin provided outstanding assistance. 55

2 Executive summary In the past seven years the Romanian economy was characterized by high-income volatility. The country experienced a deep recession between 1997 and 2000, marked by high fiscal and quasifiscal deficits that were monetized at the price of high inflation. However, the past two years recorded an exceptional recovery, which was the result of a combination of factors, including an election-driven expansion in demand, an increase in exports and investment, and disinflation. Over the whole period, growth (decline) rates in the economy were highly differentiated spatially both across rural and urban areas and across the eight large regions. Urban areas were marginally more hit by the recession throughout the period , but benefited substantially more than rural areas from the recent growth in 2001 and As the poverty is more prevalent in rural areas, this trend contributed to perpetuate poverty. The analysis of the relative contribution of growth and inequality changes on poverty changes highlights that growth has a strong impact on poverty dynamics. However, movements in redistribution have also been important in determining changes in overall poverty. In the period (mostly recession years), poverty increased and would have done so even more substantially had the distribution of income not moved favorably for the poor. Conversely, poverty decreased between 2000 and However, in these last two years income distribution moved unfavorably for the poor. In particular, in , this pattern was driven by the distribution of growth in rural areas. The occupational structure of the population is an important dimension along which to assess the impact of growth on poverty changes. Between 1995 and 2000, people moved progressively to occupational categories that acted as buffer against the recession and were characterized by increasing poverty. These population shifts accounted for a substantial share of poverty changes during the period. Conversely, population shifts across occupations played a very limited role for poverty changes in Although poverty differentials across regions are substantial, we observe very little, if none, net relocation of the population across regions or across urban and rural areas. Although the economic recovery in the past two years has been notable, the extent to which growth is sustainable in the medium and long run is a relevant issue. There is consensus that macroeconomic stability and fiscal discipline are crucial to this end. At the same time, reforms such as improving governance and the investment climate and revitalizing rural economies can address structural growth determinants and, especially in the case of rural development, lay the ground for a path of equitable growth. Projections under different growth scenarios suggest that, provided growth persists, poverty reduction can be achieved consistently. 56

3 1. Introduction Romania s economic growth recovered in 2000 after three years of consistent recession that were marked by sustained high inflation and macroeconomic instability. In this period, poverty dynamics tracked income quite closely. In light of these developments and of the commitment of the current government to poverty reduction, an in-depth analysis of the role of growth in poverty reduction in Romania is particularly relevant. In order to provide insights into these issues, this paper presents a number of statistical exercises based on the rich set of unit-level data from recent Romanian household surveys (Romanian Integrated Household Survey RIHS and Household Budget Survey - HBS). The work is organized as follows. First we describe dynamics and spatial patterns of growth. We then quantify and discuss the relative contributions of growth and changes in inequality to poverty changes (growth-redistribution decomposition) and describe the patterns of growth across different strata of the population so as to identify the extent to which growth was pro-poor (growth incidence curves). Since Romania has witnessed over time a substantial shift of the population across occupation, we also quantify the extent to which intra-sectoral changes in poverty and population shifts across sectors contributed to overall changes in poverty (sectoral decomposition). In this context we examine how some potentially important sources of growth for the Romanian economy could affect poverty outcomes. Finally, we discuss poverty predictions for the years based on different assumptions on growth scenarios. 1.1 Methodological note For consistency with existing work on poverty in Romania (see for example Tesliuc et al. 2001) and with other sections of the report for which this paper is background work, all of the results discussed in the following sections are based on a measure of consumption per adult equivalent. As robustness checks, the analysis was also applied to consumption per capita. The related results are not reported, but divergences in findings are noted. The appendix briefly discusses of advantages and limitations of the consumption per adult equivalent measure. For simplicity of exposition, emphasis is given throughout the report to headcount measures of poverty for extreme and total poverty lines. However, substantial differences in results based on poverty gap or poverty severity measures are noted. 2. Decline and growth in Romania: macroeconomic context and growth patterns for the period Throughout the late nineties, the Romanian economy was characterized by cycles of booms and busts that were the result, amongst other things, of a slow and erratic implementation of reforms (see figure 1). 57

4 Figure 1 Real GDP Growth Rates: Source: World Bank estimates. The expansion of 1996 an electoral year was followed by a substantial output collapse that lasted until the end of the decade. During that period the restructuring of many of the large state owned enterprises (SOEs) proceeded very slowly. SOEs generated substantial losses, which resulted in the accumulation of substantial quasi-fiscal deficits. In turn, monetization of these deficits generated an inflation rate among the highest in transition countries. On the other hand, poor fiscal management combined with controls on bank deposit rates contributed to a substantial credit crunch which brought investment to a halt. Overall, growth performance was disappointing (GDP in 2001 had not yet regained its 1990 value in real terms), partly due to the slow pace at which structural reforms proceeded in the country, to macroeconomic instability, and to an unfavorable investment climate characterized by widespread corruption (see discussion later). However, over the last two years, the economy has experienced an exceptional combination of disinflation and strong growth. In the second half of 2000, an election-driven expansion in domestic demand and a depreciation-led increase in export (mainly re-export of processed inputs from EU countries) boosted growth to among the highest levels in the region. Growth was then sustained by an increase in investment which benefited from an expansion of credit. Inflation was also brought down from 36% in mid-2001 to 19% in late The dynamics of income over time is not the only interesting dimension along which to study growth, as growth rates in the country differ substantially along spatial and occupational dimensions. For example, the consumption decline was relatively less pronounced in rural areas. At the same time, subsequent growth was substantially lower than average in rural areas, suggesting that the rural economy is to some extent isolated from economic volatility. However, a simple comparison of relative differences of growth and decline in urban and rural areas would indicate that rural areas participated less in the growth recovery than they were shielded from recession (figure 2). 58

5 Figure 2 Growth in mean consumption per adult equivalent 15.0% 10.0% 5.0% 0.0% % -5.0% -10.0% -15.0% -20.0% Year Urban Rural Total Source: Household Budget Survey. When disaggregating growth rates at the regional level, the West and South-West regions experienced the strongest overall decline in output. Figure 3 plots cumulative growth rates at the regional level for the sub-periods and , where regions are sorted by increasing levels of average consumption in Although it is not easy to identify trends with such a limited number of observations, the decline in consumption appeared more pronounced in the regions with lower poverty rates the West and Bucharest. In the following two years of growth the substantially lower than average growth in the South-West is notable. 2 A simple growth regression can summarize growth trends across regions over the whole period. When regressing regional growth on time dummies (not reported), average consumption (C), share of rural population (RURAL) and percentage of unemployed population (UNEMPLOYED), the results suggest that growth (recession) was higher (lower) in regions with initial lower average consumption (thereby indicating that a standard converge effect was at work) and with a lower share of rural and unemployed population. Results are reported below (tstatistics in parentheses). 3 Growth it = *C it *RURALl it *UNEMPLOYED it (-5.30) (-5.08) (-2.50) 1 Note that the ranking of regions according to average consumption levels in 1995 and 2000 was the same. 2 A number of possible explanations are consistent with this finding.: the ceased severance payments to miners, some convergence effect (the South -West started with relatively lower poverty in 2000), the influence of the embargo on the neighboring Serbia and Montenegro. 3 The coefficients are estimated with OLS on a small panel of growth for the sub-periods ; , for the eight macro regions. Standard errors are corrected for clustering at the regional level. I indices regions and t time. 59

6 Figure 3 Growth rates in mean consumption per adult equivalent 20% 10% 0% -10% -20% -30% -40% North-East South South-East South-West North-West Centre West Bucharest Regions are sorted by increasing levels of average consumption per adult equivalent Source: HBS. 3. Poverty and Growth The extent to which economic growth contributes to poverty reduction has been the object of increasing interest in the policy debate. A standard view is that continued economic growth is associated with reductions in poverty rates. Measures of elasticity of poverty to growth are often quoted as prima facie evidence of the impact to growth on poverty. The evidence for Romania suggests that elasticity of poverty to growth has been substantial in the past (1996 and 1997) and has decreased since then almost uniformly for both extreme and total poverty. The elasticity of poverty to growth has been around two (in absolute value) for , a value which is in line with the impact of growth on poverty in other countries, but decreased somehow in 2002 (see figure 5). 60

7 Figure 5 Elasticity of poverty (headcount) to growth Extreme poverty Total poverty Source: HBS, author s calculations There are limitations on relying on the calc ulated measures of growth elasticity to understand the impact of growth on poverty. A measure of elasticity of poverty to growth captures only the sensitivity of measured poverty to changes in mean income while the relevant growth rate for poverty reduction is the growth for poor and, in particular for the headcount ratio, the growth occurring in the neighborhood of the poverty line. In this context, it is important to note that certain patterns of growth might be associated with increased inequality that can offset in part or even completely the impact of growth on poverty (El-laithy et al. 2003). Two distinct but related methodologies can address in more detail the impact of growth on poverty. First, the growth-redistribution decomposition introduced by Ravallion and Datt (1992) allows us to quantify the relative importance of average income growth versus changes in income distribution for poverty changes. Second, the construction of growth incidence curves gives a visual representation of the extent to which economic growth has reached different strata of the population, including the poor and very poor. These methodologies and the respective results for Romania are described in sections 3.1 and 3.2. Section 3.3 discusses results from a decomposition of poverty across occupations where we quantify the relative within- and across- occupation contributions to overall poverty changes. We perform a similar exercise for regions and across urban and rural areas. 3.1 How much did growth contribute to poverty changes? Results from the growth - redistribution decomposition With the sole exception of 1996 an electoral year the period between 1995 and 2000 was characterized by a long and persistent recession also an electoral year marked the end of the recession with a spectacular 8.4% growth recorded in per-adult equivalent consumption. Figure 6 depicts annual growth rates in consumption per adult equivalent over the period. Poverty dynamics mirrored these trends quite closely. Poverty, however measured, started increasing in 1996, peaked in 2000, and then decreased (see detailed discussion in Pop and Tesliuc, 2003). Because of the apparent structural break in growth trends in 2000, it is useful to discuss the relative impact of growth and redistribution on poverty changes for the sub-periods and separately. 61

8 Figure 6 Growth in mean consumption per adult equivalent 10.0% 5.0% 0.0% % -5.0% -10.0% -15.0% -20.0% Source: HBS, author s calculations According to the methodology developed by Ravallion and Datt (1992), a change in poverty between two dates can be decomposed into three components: (i) the change in poverty due to a change in mean income while holding income distribution constant the growth component; (ii) the change in poverty due to a change in income distribution while holding mean income constant the redistribution component, and (iii) a residual capturing the interaction between the growth and redistribution component. Tables 1 and 2 report the results of such decomposition for yearly changes in poverty as well as for the two sub-periods and We computed the decomposition for the headcount measure of extreme and total poverty for the country as a whole and for rural and urban areas. In the period (and especially in ), output decline had a substantial impact on poverty increase. However, income distribution moved in favor of the poor reflecting a compression of income distribution that is often observed during recessions, thereby mitigating the adverse effects of the output decline. 62

9 The growth-redistribution decomposition methodology (Ravallion and Datt 1992) Consider a poverty measure P t that can be fully characterized in terms of the poverty line (z), the mean income of the distribution (µ t ) and the Lorenz curve representing the structure of relative income inequalities (L t.). In this context, the level of poverty can change due to a change in mean income relative to the poverty line, or to a change in relative inequalities. In particular, the growth component of a change in the poverty measure is defined as the change in poverty due to a change in mean while holding the Lorenz curve constant at some reference level L r. The redistribution component is the change in poverty due to a change in the Lorenz curve while the mean income is constant at the reference level µ r. Formally, a change in poverty over dates t and (t+n) can be written as P P = P( z / µ, L r ) P( z / µ, L r ) + P( z / µ r, L t + n ) P( z / µ r, L t ) R t + n t t + n t + n + This decomposition is not usually exact. A residual (R) exists whenever the poverty measure in not additively separable between µ and L. The sectoral decomposition methodology (Ravallion and Huppi, 1991) Changes in a poverty measure P t can be decomposed into an intra-sectoral effect (measuring the contribution of poverty changes within sectors controlling for the sectors population share in the base period, n i1 ); a population shift effect (capturing how changes in the population shares of sectors contributed to changes in poverty) and an interaction effect (which captures the correlation between the two). The decomposition of poverty changes between period 1 and period 2 can therefore be written as follows: P 2 P 1 = ( P i2 P i1 ) * n i1 + ( n i2 n i1 ) * P i1 + ( P i2 P i1 ) * ( n i2 n i 1 ) where i indices. 63

10 Table 1. Growth-inequality decomposition of changes in extreme poverty (Headcount) between Welfare measure: consumption per adult equivalent Period P(t) P(t+1) DP0 DPGR dis DPDIS gr Residual Nationwide Urban Rural Legenda: P(t): poverty at time t; P(t+1): poverty at time t+1; DP0: absolute change in the headcount DPGR dis: change in the headcount measure due to growth keeping inequality constant DPDIS gr: change in the headcount measure due to inequality keeping inequality constant. Values are rounded to the third decimal. 64

11 Table 2. Growth-inequality decomposition of changes in total poverty (Headcount) between Welfare measure: consumption per adult equivalent. Period P(t) P(t+1) DP0 DPGR dis DPDIS gr Residual Nationwide Urban Rural Legenda: P(t): poverty at time t; P(t+1): poverty at time t+1; DP0: absolute change in the headcount DPGR dis: change in the headcount measure due to growth keeping inequality constant DPDIS gr: change in the headcount measure due to inequality keeping inequality constant. Values are rounded to the third decimal. 65

12 The contribution of growth to poverty changes was also sizeable in had redistribution kept constant, poverty reduction would have been much more substantial on account of economic growth alone. Nonetheless, during this period and, in particular between 2001 and 2002, income distribution moved adversely for the poor, thereby partially hampering the impact of growth on poverty. 4 Most of the shift in income redistribution that was adverse for the poor was concentrated in rural areas. Interestingly, income redistribution moved favorably for the poor in urban areas between 2000 and Although the relative magnitude of the effects differs somehow, the decomposition of total poverty changes returns a similar picture. Results of the decomposition are also similar when we use the poverty gap as the poverty measure. Over the two sub-periods the poverty gap first increased and then decreased and growth and redistribution components moved in the same direction for changes in poverty gap and for changes in the headcount. However, the relative magnitude of the two components suggests that the shift in the redistribution was relatively more effective in reducing the relative distance of the poor from the poverty line than to get people out of poverty. 5 The reported trends are robust to using consumption per capita as an alternative measure of welfare. 3.2 The distribution of gains from growth: growth incidence curves Because of the pattern of substantial economic decline followed by strong growth in the past two years, assessing the extent to which the gains from aggregate economic growth and the losses from contraction are distributed across households is especially interesting in the case of Romania. Growth-incidence curves are useful to this end, as they plot the growth rate in consumption for individuals ranked according to their consumption (Ravallion and Chen, 2002). 6 Figure 8 reports growth incidence curves for the sub-periods and The vertical axis measures cumulative growth (or decline) for the whole period, while people on the horizontal axis are ordered by increasing levels of per-adult equivalent consumption. The two vertical lines identify the share of the population in extreme and total poverty, respectively. Note that a downward sloping incidence curve indicates that people in poorer quantiles of the population have benefited from growth (or lost from output decline) more (less) than the average. By 2000, per adult equivalent consumption was about 80% of consumption in During this period people in the poorer quantiles of the income distribution were hurt relatively less by the recession than people in the higher quantiles. This trend was fairly uniform across urban and rural areas and in year-by-year dynamics. 7 This evidence is consistent with the evolution of inequality for the period the Gini coefficient decreased from a value of 0.31 in 1995 to a value of 0.28 in 4 The interpretation of the tables is straightforward. For example, the change in the headcount for extreme poverty between 2000 and 2002 was (DP0). Had income distribution kept constant, poverty would have decreased more substantially, precisely by 0.04 (DPGR dis). Conversely, had growth kept constant, poverty would have increased by 0.01 percentage points (DPDIS gr). 5 Results are not reported but are available upon request. 6 H More precisely, Ravallion and Chen (2002) define the rate of pro-poor growth as t 0 g ( p) dp / H t, i.e. the integral of the growth rates by quantile g(p), where p indexes income quantiles up to the headcount index H t weighed by the headcount itself. The authors show that this measure is consistent with the Watts index of poverty. 7 An exception was the pattern of growth between 1999 and 2000, when poor in urban areas were hit by output decline more than people in urban areas ranking above the 30 th percentile in expenditure. Year-byyear growth incidence curves are not reported but are available upon request. 66

13 2000 and indicates that the recession was accompanied by a compression in the overall structure of income distribution. It also lends support to the view that in this period social protection policy was effective in dampening the impact on the poor of the structural reforms initiated in the late nineties. Finally, it is useful to note that the patterns of the growth incidence curves are consistent with the evidence we discussed on the partial effects of growth and redistribution on changes in poverty. Between 2000 and 2002 output grew by a cumulative 11.9% but the trend in the distribution of growth rates across quantiles of the population reversed when compared to the previous years. People below the 50 th percentile, and in particular the poor, did not appear to share the benefits of economic recovery. Overall this pattern appears to be driven by the distribution of growth of the population in rural areas, where not only growth was substantially lower than the national average, but the poor seemed to be particularly excluded from it. Conversely, growth in urban areas was substantial (a cumulative average growth of 15%) and the poor fared relatively better than the rest of the urban residents, as they benefited from growth rates between 15 and 17%. When disaggregating the evidence into yearly growth patterns, the extent to which the poor and the lower quantiles of urban population benefited from growth appears to be driven by the developments in , and is most likely due to increased social spending (pension recorrelation and heating subsidies) and patronage that coincided with the peak of the electoral campaign and was mainly targeted to urban areas. The pattern of the growth incidence curve for urban areas in highlights that gains from growth in urban areas were concentrated among those below the extreme poverty line while people in the neighborhood of the total poverty line enjoyed less than average growth. 8 A further dimension along which one could gain insights into the distribution of growth across quantiles of the population is to disaggregate growth incidence curves by economic sector. When we do so, we find that growth patterns were quite uniform across sectors in (with the poor being hit less by the recession in all sectors). Conversely, between 2000 and 2002, growth was substantially below the (sectoral) mean for all the pensioners and the unemployed in poverty and for agricultural workers in extreme poverty. 8 This evidence is consistent with the results from the growth-inequality decomposition which indicate that in (the year that seems to be driving this pattern) inequality moved relatively more favorably for the urban poor in extreme poverty than for the urban poor in poverty. 67

14 Figure 8. Growth-Incidence Curves Cumulative growth in consumption per adult equivalent Nationwide Nationwide Urban Areas 14 Urban Areas Rural Areas Rural Areas Notes: the vertical lines indicate the share of people in extre me (leftmost line) poverty and total poverty. 68

15 Figure 9a. Cumulative growth in consumption per adult equivalent , by sector, Agriculture Pensioners Industry Unemployed and others Services 69

16 Figure 9b. Cumulative growth in consumption per adult equivalent by sector, Pensioners Agriculture Unemployed and others 10 Industry Services 70

17 3.3 Are poverty changes due to intra sectoral (regional) poverty reduction or to population shifts? Evidence from sectoral and regional decomposition of poverty The structure of occupation in Romania has changed substantially over the past seven years. 9 Following the structural adjustment policies initiated in 1997, the share of employees dropped significantly in 1998 and decreased consistently since then. The flow out of employment translated partly in higher unemployment (which peaked in 2000) and in an increased number of pensioners, and partly in a higher share of agricultural and non-agricultural self-employment that acted as a buffer for the effects of the recession. The reported number of self-employed can be taken to be a good proxy for the share of the population engaged in the informal economy. Within this category, the increase of the share of people employed in agricultural was particularly significant. 10 The rapid growth in the number of pensioners has been a constant throughout the nineties. It reflected both the increase in the number of disability pensions which, thanks to lax rules, were often used as coping mechanisms for the unemployed, as well as a sharp increase in retirees in 2001 the transition period before the implementation of the more restrictive Law on Pension Reform (World Bank, 2002). Figure 9 depicts the shares of the population in different occupational categories for 1995, 2000, and Figure Trends in occupational structure Employee Self-empl. and employer Self-employed agriculture Unemployed Pensioner Other 9 Although, for simplicity, we use the term sectoral throughout this section, we computed the decomposition across occupational categories (employees, employers, self-employed in agriculture, etc.). This approach is more likely to allow us to capture movements of workers from the formal to the informal sector. 10 The share of people in agriculture increased substantially in 2000 and then marginally in the following two years of growth (2001 and 2002), partly as a result of the continued process of land restitution initiated with the land reform in the early 1990s. 11 Note that, because of the structure of the household budget survey, which collects information only at the household level, the occupational structure of the population is constructed by attributing the same occupation of the head of the household to all family members. 71

18 Figure 10 depicts consumption growth by occupational category for the sub-period and Figure 10 Growth in mean consumption per adult equivalent 40.0% 30.0% % 20.0% 10.0% 0.0% -10.0% -20.0% Employee Self-employed non-ag. Self-employed ag. Unemployed Pensioner Other Total -30.0% -40.0% Not surprisingly, self-employed in non-agricultural activities (a category which also includes employers) were subject to substantial volatility self-employed were hit more strongly than average by the recession and benefited more than average from the recovery. This trend is reversed among agricultural workers, who were relatively more isolated from the recession It is interesting to couple this information with changes in the structure of poverty by occupational categories. The share of employees in extreme poverty has declined substantially since 1995, so that in 2002 less than 10% of the poor were employees. Conversely, the share of agricultural workers in extreme poverty increased from less than 25% in 1995 to around 32% in Finally, the share of pensioners among the poor increased marginally, but steadily, reaching 33% in 2002 (see figure 11). 72

19 Figure 11 40% 35% 30% 25% 20% 15% 10% 5% 0% Occupational structure of the poor (extreme poverty), % of all poor Year employee employer & self-empl. non-ag self-employed agr. unemployed pensioner other These trends lend support to the interpretation that people have progressively moved towards occupational categories with increasing poverty. In this context, it is interesting to quantify with precision the extent to which sectoral growth and inter-sectoral shifts of population contributed to poverty reduction (increase). The sectoral decomposition introduced by Ravallion and Huppi (1991) allows us to disentangle the intra-sectoral effects from the effects of population shifts across sectors. The related evidence is reported in table 3 for changes in the headcount for extreme and total poverty. Results for the sectoral decomposition of changes in the poverty gap and poverty severity (not reported) return a similar picture. In the period , extreme poverty (as measured by the headcount) increased. A substantial part of this increase (around 60%) was due to increases in poverty within sectors, particularly among pensioners and agricultural workers. Shifts of population across activities accounted for 30% of the poverty increase. This reflected the strong decrease in the share of employees, which resulted in an increase in the share of agricultural worker, pensioners, and unemployed. The sign of the population-shift effect indicate that people moved towards sectors with increasing poverty over the period. In the following two years ( ), almost all of poverty reduction was due to intra-sectoral effects and was concentrated among pensioners (who benefited from the recorrelation exercise), employees, and workers in the informal, non-agricultural sector. Population shifts across occupational categories indicate that people moved towards occupations were poverty was decreasing at a lower rate, but this effect was of limited magnitude. Results for the sectoral decomposition of total poverty are similar, although it is notable that in an increase in total poverty among employees contributed substantially to the overall change in the indicator. A similar exercise can be carried over to investigate whether population shifts across regions and urban and rural areas have significantly contributed to changes in poverty rates. The evidence suggests that most poverty reduction (or increase, as it is the case for the period ) was due to intra-regional effects. Moreover, no significant contribution to poverty changes was due to 73

20 movements of people across urban and rural areas. Most of the poverty change in both subperiods was due to changes in poverty in urban areas. Results are similar for decomposition of the headcount ratio as calculated with the total poverty line as well as for the poverty gap for extreme and total poverty. Table 3. Sectoral and regional decomposition of changes in poverty EXTREME POVERTY TOTAL POVERTY Period Poverty in year Poverty in year Absolute change % change Activity Pop. share, year1 Abs. Change * % of change in pov. Pop. share, year1 Abs. change % of change in pov. Pop. share, year1 Abs. Change % of change in pov. Pop. share, year1 Abs. change % of change in pov. Employees Self-employed non-agr Self-employed agr Unemployed Pensioner Other Intra-sectoral effect Population-shift effect Interaction effect Region North-East South-East South South-West West North-West Center Bucharest Intra-sectoral effect Population-shift effect Interaction effect Urban/Rural Urban Areas Rural Areas Intra-sectoral effect Population-shift effect Interaction effect Total change in poverty *This column reports absolute changes in poverty in each sector. These changes sum up to the intra-sectoral effect. Indicate the absolute value and share of poverty change due to movement of people across sector. For example, in extreme poverty increased by 4.4 percentage points. Of these, 2.7 percentage points were due to increase in poverty within sectors (61.18%), while 1.3 percentage points (30.25%) were due to shifts of the population towards activities were poverty increased. 74

21 4. Sources of growth The recovery of the Romanian economy that started in the late 2000 was impressive. In this context, an important issue is the extent to which growth is sustainable in the medium run. There is consensus on the fact that macroeconomic stability and responsible fiscal policies are crucial preconditions for sustained the growth over time. However, structural development in a number of specific areas, such as improvement of the investment climate and the revitalization of the rural economy, can lay the foundations for equitable growth in the long run. The cross-country literature as well as microeconomic studies highlight the importance of good investment climate as an engine of growth. Many studies have shown that measures of good governance, enforced property rights, and low corruption are causally associated across countries with higher growth in the short run (see for example, Knack and Keefer, 1995 and Mauro, 1995) and with higher income per capita in the long run (see Kaufman et al., 1999, and Hall and Jones, 1999). Results from microeconomic evidence confirm the strength of this association. For example, using firm-level data for five countries, Dollar et al. (2003) find that indicators of investment climate (such as number of inspections per year, number of days to clear customs or to obtain a phone line) are important determinants of firm productivity and growth. Moreover, good institutions are believed to promote FDI flows (see World Bank, 2003) Strengthening institutions is a particularly relevant issue in the Romanian economy. According to Transparency International (TI), in 1998 Romania was among the top third most corrupted countries in the world (ranking 61 st out of 85 countries in the TI index of lack of corruption, with a Corruption Perception Index of 3/10 where 10 indicates absence of corruption). To the extent that comparison of the index over time are informative, Romania s corruption index had fallen by 2002 to 2.6, with the country ranking 77 th out of 102 countries. 12 Findings from a very detailed World Bank report that surveyed households, enterprises, and public officials also indicates perceptions of a high level of corruption surrounding the implementation and design of laws and regulations. Firms reported corruption to be a substantial obstacle to their business. An analysis of the patterns of informal payments revealed that corruption has important distributional consequences: poorer households appeared paying larger share of their income than richer families in the form of bribes and the data suggested that access to health care might be precluded to poor households who cannot afford to make unofficial payments. (World Bank, 2000). Rationalizing the regulatory burden can also be highly beneficial, especially in an economy that is witnessing a progressive move of the work force from the formal to the informal sector. This need for institution building has long been recognized and governance reforms are at the core of the World Bank support to the private sector development and to the strengthening of accountability in the public sector (see CAS, World Bank, 2002). Poverty in Romania has an important regional dimension and is particularly more entrenched in rural areas (see table 4). Rural poverty is a reflection, amongst others, of the collapse of agricultural production and of the absorption of displaced industrial workers. Substantially slower growth rates (and only marginally lower decline rates of output during recession times) account for the persistence of poverty in rural areas. In this context, devising strategies to revitalize the rural economy can accomplish the twofold objective of tackling lack of growth for a large share of the population as well as addressing an important dimension of inequality within the country. The World Bank Country Assistance Strategy has moved along these lines and targeted a large 12 For information on how the TI index is constructed see 75

22 share of its interventions to rural areas by, amongst other things, financing capacity building and infrastructure activities (see CAS, World Bank, 2002). Table 4. Urban and rural poverty and growth rates Extreme Poverty (headcount) Growth rate Rural Urban Total Poverty projections In this section we discuss results from projections for the headcount ratios for extreme poverty for the years using two alternative methodologies. First, we compute poverty using predicted future consumption obtained by applying a uniform growth rate to the whole structure of current consumption (base projections). Second, we assume a balanced growth path across occupational categories and, while keeping sector shares constant at the existing level (2002), we apply to each sector the growth rate that is compatible with the assumed overall growth rate. On this basis we then compute poverty rates (sectoral growth).. We first implement a simple historical validation of the two procedures. In particular, (i) we apply to the 2001 consumption structure the average growth rate between 2001 and 2002 to predict consumption in 2002 and then calculate poverty (base); (ii) using 2001 sector shares in output and ratios of growth by occupational categories, we predict consumption in 2002 and calculate headcount ratios (sectoral projections). We then compare the poverty measures obtained in (i) and (ii) to actual poverty measures. In the base projection, the source of forecast error lies in applying to every individual in the distribution the same growth rate (while we know from growth incidence curves that in this period the poor benefited from growth less than richer people). The base projections turn out to be very close to the actual poverty rate, although, as one would expect, they slightly underestimates poverty. In sectoral projections both hypotheses of fixed sector shares and balanced growth are potential sources of forecast error. However, by diversifying the growth rate by occupation, we expect these projections to be more accurate than base projections (see table 5). Table 5. Historical validation for base and balanced sectoral growth projections, Extreme Poverty Headcount in 2002 Total Poverty Headcount in 2002 Predicted Actual Error Predicted Actual Error Per capita consumption growth rate Base (3.3%) % % Sectoral (3.3%) % % 76

23 The results of this exercise suggest that, in the short term, these methodologies might produce reasonably good projections of poverty rates. However, as the horizon lengthens, projections are bound to becomes less informative, not last because the assumption of constant sector share and balanced growth across sectors might be less adequate. Projections are presented for growth in consumption per adult equivalent (see discussion in the appendix for the choice of welfare measure) and for a number of different growth scenarios. The benchmark growth rate for is taken from IMF projections on per capita consumption growth. 13 These growth projections are constructed using consumption from the National Accounts. It is useful to note that there exists a substantial drift between consumption data reported by national accounts and the data (see table 6). This divergence has been noted in other contexts and has been linked to a possible poor representativity of the survey and/or methodological discrepancies in the construction of the consumption aggregate (see for example Tesliuc, 2001). Table 6. Drift between National Accounts and Household Budget Survey consumption aggregates Household consumption National Accounts per capita % annual change, real terms Household Budget Survey per capita % annual change, real terms Source: National Accounts (NA) and Household Budget Survey. Per capita consumption growth from NA is calculated based on the hypothesis of annual population growth of 0.1%. Alternative assumptions are also adopted, corresponding to an optimistic scenario (sustained 5% annual growth over the whole period), two middle ground scenarios (one where growth is sustained at the rate of 4%, the other where growth continues at the current annual 3.3%), and a low growth scenario (where growth is at annual rate of 2.5% throughout). Table 7 reports the projections for the headcount ratio for both extreme and total poverty. 13 As discussed in the appendix, consumption growth of per capita and per adult equivalent consumption track each other very closely. 77

24 Table 7. Poverty Projections, Growth rate Scenarios Extreme Poverty Total Poverty IMF's Projections of p.c. growth rates* Growth projections Base 9.36% 8.22% 7.17% 6.20% 5.42% 25.82% 23.17% 20.73% 18.82% 16.77% Sectoral 9.60% 8.72% 7.95% 7.35% 6.78% 25.98% 23.84% 22.00% 20.01% 18.56% 2002 growth rate Growth projections 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% Base 9.91% 8.84% 7.99% 7.18% 6.36% 26.73% 24.75% 22.74% 20.75% 19.25% Sectoral 10.16% 9.19% 8.59% 7.95% 7.44% 26.77% 25.09% 23.53% 22.01% 20.34% 5 Percent Growth projections 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% Base 9.35% 7.98% 6.84% 5.63% 4.72% 25.75% 22.73% 19.88% 17.51% 15.31% Sectoral 9.57% 8.59% 7.77% 7.01% 6.59% 25.95% 23.54% 21.12% 19.14% 17.41% 4.0 Percent a Growth projections 4% 4% 4% 4% 4% 4% 4% 4% 4% 4% Base 9.64% 8.55% 7.48% 6.49% 5.58% 26.35% 23.82% 21.43% 19.45% 17.49% Sectoral 9.86% 8.94% 8.21% 7.57% 6.97% 26.42% 24.42% 22.61% 20.56% 19.08% 2.5 Percent Growth projections 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% Base 10.15% 9.34% 8.67% 7.93% 7.33% 27.28% 25.72% 24.14% 22.64% 21.13% Sectoral 10.30% 9.46% 8.82% 8.43% 7.93% 27.27% 25.42% 23.99% 22.71% 21.62% Notes: actual extreme poverty rate in 2002 = 10.91%; actual total poverty rate in 2002 = 28.89%. *Per capita growth rates are computed assuming a constant annual population growth of -0.1%. a The 2001 CAS projected per capita consumption growth at 4% for Conclusions In this work we analyzed the role of economic growth for poverty reduction for the years in Romania. During this period the country experienced a long recession followed by a substantial economic recovery. Poverty dynamics mirrored mean income developments quite closely. However, we have shown that changes in income inequality were important determinants of changes in poverty. In particular, during the period (mainly of output decline), income distribution moved favorably for the poor, thereby hampering the impact of the recession on poverty. Conversely, in the following period the poor and the extreme poor benefited from growth less than richer strata of the population. This pattern appeared to be partly driven by income dynamics in rural areas, where growth was substantia lly lower than the mean and skewed towards people above the 50% percentile in the consumption distribution. 78