\\PS 2 SHG POLICY RESEARCH WORKING PAPER 2846 Is India's Economic Growth Leaving the Poor Behind? Gaurav Datt Martin Ravallion The World Bank Development Research Group Poverty Team May 2002 Poi.Ic v RESEARCH WORKING PAPER 2846 Abstract There has been muchi debate about how muchl India's growth would have genierated a rate of poverty redtiction poor havc shared in the econonic growth unleashed by that was double India's historical trend rate. States with economic reforms in the 1990s. Datt and Ravallion argue relatively low levels of initial rural development and that India has probablx mainitained its 1 980s rate of humani capital development were not well-sulted to poverty reduction in the 1990s. However, there is reduce povertv in response to economic growth. considerable diversity in performance across states. This The study's results are consistent with the view that holds some important clues for understanding wlhy achieving higher aggregate economic growth is only one economic growtlh has not done more for India's poor. clement of an effective strategy for poverty reduction in India's economic growth in the 1990s hias not been India. The sectoral and geographic composition of occurrinig in the states where it would have the miost growth is also important, as is the need to redress impact on poverty nationially. If not for the sectoral and existing inequalities in human resource developmeint and geographic imbalanice of growth, the nationlal rate of between rural and urban areas. This paper-a product of the Poverty Team, Development Research Group-is part of a larger effort in the departmient to better understand the relationshlip between economilic growth and poverty. Copies of the paper are available free from the World Bank, 1818 H StreetNW, Washington, DC 20433. PleasecontactCatalinaCunanan, room MC3-542, telephonc 202-473-2301, fax 202-522-1151, email address ccunanan@)worldbank.org. Policy Research Working Papers are also posted on the Web at http://econ.worldb-ank.org. Tlle authors may rbe contacted at gdattrxworldbank.org or mravalloll(ioworldbank.org. May 2002. (29 pages) Thve Policy Researcb Workiiig Paper Series dusseminates the fRedings of ea ork c i progress to aff colarage the exchange of Ideas abolit developm^enzt issiwes. Ai7 objective of tbe series Is to) get the fiiidiiigs oitt qiilckly, even If tbe presenztationas are less than fiilly polished. 'I he papers carry the niames o)f the aitthors and shouild be cited accordinigly. The finzdings, interpretationis and coniclusions expressed nil this paper are entirely those o)f the azthtors. Thvey do niot iiecessarilv represeitt the tvietv of the W/orld Bank, its Execi(tiv e Directors, or the coiiiitries tbev represenit. Produced bv the Researcli Advisory Staff Is India's Economic Growth Leaving the Poor Behind? Gaurav Datt and Martin Ravallion1 I These are the views of the authors, and need not reflect those of the World Bank or any affiliated organization. For comments or other forms of help, the authors are grateful to Kaushik Basu, Angus Deaton, Monica Jain, Valerie Kozel, Peter Lanjouw, Vijayendra Rao, T. N. Srinivasan and seminar participants at the Delhi School of Economics, the Indian Statistical Institute, New Delhi, and the World Bank. The support of the Bank's South Asia Poverty Reduction and Economic Management Network is gratefully acknowledged. 1. Introduction More of the world's income-poor live in India than any other country. Using an international poverty line of $1 per day (measured at a 1993 purchasing power parity exchange rate), about one third of the poor in the mid-1990s lived in India.2 What happens to poverty in India is quantitatively important to the world's overall progress in fighting absolute poverty. Thus the recent signs of sustainably higher growth in India may offer encouraging news for poverty reduction. In the 196 Os and 1970s, the real annual rate of GDP growth in India was 3.4 percent, implying a per capita annual growth rate of barely 1 percent. Growth rates in national output since the mid-19,30s have been appreciably higher on average. In the 1 990s, average consumption per capita (as measured in the national accounts) has grown at an annual rate of 3.0 percent, implying about a one-third increase in consumption per capita over the decade. It appears plausible that the economic reforms carried out by India in the 1990s are the main cause of this higher growth. (Ahluwahlia, 2002). Experience prior to the 1990s suggests that economic growth in India has typically been poverty reducing. Using data from 1958 to 1991, Ravallion and Datt (1996) find that the elasticity of the incidence of poverty with respect to net domestic product per capita was -0.75, and that with respect to private consumption per capita was -0.9 (Table 1). The higher absolute elasticities for measures of the depth and severity of poverty in Table 1 indicate that those well below the poverty line have benefited from macroeconomic growth, as well as those near the poverty line. Nor is there any convincing evidence that economic growth in India prior to the 1990s has tended to be associated with rising overall inequality (Bruno et al., 1998). These observations clearly refute claims that pre-1990s growth in India tended to leave the poor behind. However, the 1990s are more contentious. Some observers have argued that poverty has fallen far more rapidly in the 1990s than previously (see, for example, Bhalla, 2000). Others have argued that poverty reduction has stalled, and that the poverty rate may even have risen (for example, Sen, 2001). So what has happened in India in the 1990s? Has poverty continued to fall with growth, or has the nature of the growth p-rocess changed, such that the poor have been left behind? This 2 This calculation is based on the World Bank's Global Poverty Monitoring database (http://www.worldbank.org/research/povmonitor/). 2 paper tries to answer those questions. We do not attempt to assess the impact of India's macroeconomic reforms of the 1990s on poverty, since this would require identification of the counter-factual of what would have been experienced in the 1990s without the reforms. Rather, our aim is to describe what has happened to poverty in India in the 1990s. In the course of the discussion, we will learn about the proximate causes of changes in India's poverty rate. Moreover, although this discussion is India-specific, it illustrates themes that are often encountered in the analysis of poverty in low-income economies, including difficult issues of survey design and comparability, and the proximate factors underlying the responsiveness of poverty to economic growth. 2. Measuring Poverty in India Many surveys have thrown light on the dimensions and causes of poverty in India, ranging from village-level studies to national surveys. However, by far the most important tool for monitoring poverty since the 1960s has been the Household Consumer Expenditure Surveys conducted by the National Sample Survey (NSS) Organization. Various methods have been used to measure poverty with the NSS data. Figure 1 presents our estimates of the poverty rate in India since 1958. These are population-weighted averages of the poverty measures for urban and rural areas of 14 major states of India (not including Jammu and Kashmir, for which there are data problems). The estimates use household consumption expenditure per person as the indicator of individual welfare, and use the urban and rural poverty lines developed by India's Planning Commission (Government of India, 1979). This poverty line was about 15 percent higher in urban areas than rural areas. We have adjusted these urban and rural poverty lines over time and space using price indexes for the different states of India.3 We should emphasize that the state-specific poverty lines implied by our price indices differ from the current poverty lines used by the Planning Commission, despite their common starting point in the original 1979 3 For a description of our approach to the data, see Ozler et al., (1996) and Ravallion and Datt (2001, Appendix). A compilation of the data and description of sources can be found at: http://www.worldbank.orgZRoverty/data/indiapaper.htm). For further details on the construction of the price indices, see Ozler, Datt and Ravallion (1996), Datt (1997), and Datt and Ravallion (1998a). 3 Planning Commission poverty lines, because the Planning Commission uses a different set of spatial and temporal deflators to update poverty lines.4 Prior to the release in 200:t of the results from the 1999-00 survey round, a number of observers had looked at the numbers such as in Figure 1 and concluded that India's economic reforms were leaving the poor behind - in short that poverty reduction had stalled (Datt, 1999; Jha, 2000a). Some commentators have seen this as a damning criticism of the reform process. Others questioned the data. While the NSS has been a well-respected survey instrument, and a model for other countries, the seeming dichotomy between macroeconomic evidence on growth of consumption and the lack of commensurate poverty reduction in the NSS (prior to the latest survey round for 1999-00) led somne to doubt the reliability of the NSS as an instrument for monitoring poverty in India. 2.1 The Difference Between the NSS and the National Accounts It has long been clear that the macroeconomic data from India's national accounts define consumption differently from the NSS. Private consumption in India's national accounts includes expenditures by nonprolit organizations as well as households, while the NSS surveys only households. Consumption inI the national accounts also includes financial services and imputed rents for housing that are not found in the consumption numbers from the NSS. Comparing the nominal consumption aggregates from both sources over the period 1972-97, Sen (2001) finds that the consumption by households in India implied by the NSS accounts for 60-70 percent of the national private consumption implied by the national accounts, depending on the precise measures of consumptionL used. Moreover, the divergence between the NSS and the national accounts seems to be growing. We calculate that consumption rose an average of 0.74 percentage points faster per year (with a standard error of 0.10) in the national accounts than in 4 Comparisons of poverty measures between urban and rural areas in developing countries have been notoriously difficult, and different methods of setting the urban-to-rural differential in poverty lines can give radically different results. For example, a poverty measure based on cost of living differences can yield very different results from another commonly used method based on expenditures at which food-energy requirements are typically met (Ravallion, 1994). Elsewhere we have discussed this issue in the context of poverty measuremeni in India and argued that our urban-rural and inter-state differentials in poverty lines accord well with indqeendent estimates of cost-of-living differences facing the poor (Ravallion and Datt, 2002, Appendix). 4 the NSS data over 1972-97, though this difference was not significantly greater in the 1990s.5 This is clearly not a negligible difference in growth rates between the two main sources of data on aggregate consumption in India Some observers have assumed that the divergence is due entirely to underestimation of consumption in the NSS, and that this underestimation has been distribution-neutral - that the surveys get the mean wrong but inequality right. If one re-calculates India's poverty measures under that assumption - by scaling up all survey consumption levels by a fixed proportion to reach the national accounts levels- then one finds (of course) lower levels of poverty and higher rate of poverty reduction than indicated by the NSS (Bhalla, 2000; Srinivasan, 2000). However, there is no basis for assuming that the divergence between NSS and the national accounts is solely due to underestimation of consumption in the NSS. Nor is there any basis for the assumption that any underestimation by the NSS is distribution neutral, such that inequality is correctly estimated (Ravallion, 2000b). For example, household surveys are rarely considered a reliable source for measuring incomes of the rich. The rising income share of India's richest taxpayers that is found in tax records for the 1990s has not been reflected in the NSS consumption distributions, though it appears that this alone cannot account for the discrepancy between NSS and national accounts growth rates (Banerjee and Piketty, 2001). One way to explore the sources of divergence between the household data in the NSS and the national accounts data sources is to focus on specific categories of consumption. For example, food is about 60 percent of consumption on average. When one focuses on consumption of the food staples that figure most prominently in the budgets of the poor, there appears to be little or no divergence between the NSS and the national accounts data (Sundaram and Tendulkar, 2001a; Kulshreshtha and Kar, 2002). These calculations suggest that "correcting" for survey underestimation by scaling up all consumptions to reach the national accounts aggregate entails a substantial over-correction for the poor, and hence underestimation of the extent of poverty in India. 5 Two warnings are worth noting here. First, there are not a lot of data for drawing conclusions about whether trends changed in the 1990s. Second, there was a revision to the methods used for the national accounts data in the 1990s, in line with new intemational standards. The extent of divergence depends on whether one uses the new series (base 1993-94) or the old one (base 1980-81). The results in the text use the new series. Using the old series, we find that consumption in India's national accounts grows 0.55 percentage points faster per year (with a standard error of 0.07 points) from 1972-1997. The new methods for measuring the national accounts increased the rate of consumption growth in the 1990s. 5 2.2 Interpreting the 1999-00 NSS Data Given the concerns about whether poverty reduction had been stalling in the post-reform period, the release of the NSS data covering 1999-00 was keenly awaited. Based on these data, India's Planning Commission (2001) reported a sharp reduction in poverty in 1999-00 like that evident in our Figure 1. However, upon closer examination, one finds that the design of the NSS changed in 1999-00 - in ways that cast doubt on the comparability of the resulting poverty estimates with those from earlier rounds. When the NSS began in the 1950s, it used 30-day recall for consumption; that is, it asked people how much they had spent on various items in the previous 30 days. This changed with the survey done in 1994-95, and for this survey and the ones carried out in 1995-96, 1997 and 1998, the NSS administered two different consumption schedules to two independent sub-samples of households: one with the traditional 30-day recall, the other with multiple recall periods for different items: 7-day recall for food (food, pan, tobacco and intoxicants), 30-day recall for high-frequency nonfood (fuel and light, miscellaneous goods and services, non-institutional medical) and 365-day recall for low-frequency nonfood (educational, institutional medical, clothing, :Footwear and durable goods). These changes were not of serious concern, since one can still make consistent comparisons over time using the first schedule (as we have done in constructing Figure 1). The 1999-00 data from the NSS also included a far more worrying change. In that round, food consumption was obtained by both 7-day and 30-day recall for the same set of households, with the columns appearing side-by-side on the same page of the questionnaire. The numbers for mean of food consumption fronn the two recall methods in-the 1999-00 NSS round are quite similar - far more so than in the four previous experimental rounds, in which different households got different recall schedules (Visaria, 1999; GOI, 2000). Putting both 7-day and 30- day recall questions side-by-sid.e on the same page of the questionnaire probably promotes convergence; interviewers and respondents naturally would tend to cross-check or validate the response based on oxie recall period with that based on the other. By contrast, spending on low- frequency nonfood consumption items - typically accounting for about 20 percent of the average consumption. This change could increase or decrease the poverty count (while the longer recall period will tend to give a lower mean it will probably also give a lower variance). The 30- 6 day recall period was only used for the high-frequency nonfood items, accounting for the other 20 percent of aggregate consumption. It turns out that the way one interprets the 1999-00 NSS data depends heavily on whether one uses 7-day or 30-day recall for food expenditures. If one uses the 30-day recall estimates for food and ignores the other differences in the 1999-00 data, then the consumption distributions for that year imply a sizable reduction in poverty. The Planning Commission's (2001) estimates along these lines indicate that the national poverty rate fell by about 10 percentage points between 1993-94 and 1999-00, from 36 percent in 1993-94 to only 26 percent in 1999-00. If instead one compares the 7-day estimates for 1999-00 with the 7-day estimates from the previous four experimental rounds, then one gets an increase in poverty. The comparison suggests an increase of 2 percentage points in the rural poverty rate between 1994-95 and 1999-00, and an increase of 5 percentage points in the urban poverty rate (Visaria, 1999; Sen, 2001). Is it possible to work with the data in the 1999-00 NSS round in a way that produces estimates more comparable to those from earlier rounds? Deaton (2001 a) attempts to do so by exploiting the fact that some goods in the 1999-00 data - accounting for about one-fifth of mean consumption -used the same 30-day recall period as in previous survey. Deaton makes two key assumptions. First, he assumes that the survey results for the goods with the common 30-day recall period were unaffected by the change in survey design. Secondly, he assumes that the distribution of total consumption conditional on consumption of the common-recall goods has not changed over time and so can be inferred from the 1993-94 round (which was of course uncontaminated by the change in survey design). These assumptions allow him to generate an estimate of the distribution of total consumption as if there had been no change in survey design. Using the Planning Commission's (2001) official poverty lines, Deaton (2001 a) finds that the rural poverty rate fell from 37.2 percent in 1993-94 to 30.2 percent in 1999-00, while urban poverty fell from 32.6 percent to 24.7 percent. After weighting these reductions by the urban and rural population shares, Deaton's estimates imply that the national poverty rate fell from 36.2 percent in 1993-94 to 28.8 percent in 1999-00 - a decline of 1.2 percentage points per year. Using a similar method, Tarozzi (2001) finds a similar rate of decline in poverty. Deaton 2001(b) uses an altemative price deflator developed by Deaton and Tarozzi (1999) which leads to a lower estimate of the poverty rate but a similar estimate of the decline in the poverty rate in the 1990s. 7 Of course, these "corrections" are only as good as the identifying assumptions on which they are based, as Deaton points out. The assumptions here imply that, at a given level of total consumption, demand for the goods with the common recall period must not change over time because of changes in tastes, relative prices or survey design. It is known that the stnrcture of relative prices changed during this period (Sen, 2001). There will be an under- (over-) estimation of the level of poverty in 1999-00 if the underlying changes in tastes and prices entail that demand for the goods with the common recall period increased (decreased) over time at any given level of total spending. Nor is it obvious that the changes in survey design would leave the results for the "30-day goods" unaffected. Deaton (2001 a) and Tarozzi (2001) find indirect supportive evidence for their identifying assumptions using the intermediate "thin" NSS sample surveys between 1993-94 and 1999-00. Of course, if one accepts these intermediate surveys for validation purposes, then one must presumably accept their implied poverty measures, in which case the puzzle remains as to why poverty fell so sharply in just one or two years. 2.3 Other Surveys for Measuring Poverty in India Other surveys have been used to measure poverty in India in the 1990s. These surveys have their own problems, but the generally point to the conclusion that India has experienced a non-negligible reduction in the aggregate incidence of poverty in the 1990s. For example, the National Sample Survey Organization also carried out Employmnent- Unemployment Surveys (EUS) in 1993-94 and 1999-00 that included consumption modules, which were not contaminated by the mixture of recall periods within one survey, as in the expenditure survey for the 1999-00 NSS. The consumption module in the EUS was abridged compared with the standard Consumer Expenditure Surveys. However, past surveys have demonstrated that abridged modules tend to report lower levels of consumption, so the EUS may provide a lower bound to the degree of reduction in poverty over the period 1993-94 to 1999-00. Sundaram (2001) analyzed the consumption distributions from the EUS and found a annual rates of poverty reduction was 0.50 points per year for rural areas and 0.27 points per year for urban areas. So we can be reasonably confident that poverty incidence has in fact fallen.6 6 Sundaram and Tendulkar (2001b) take this comparison of the EUS with the NSS a step further and argue that the 30-day recall numbers from the 1999-00 NSS round expenditure survey are comparable witl previous 8 Yet another survey is the Market Information Survey of Households (MISH), carried out by the National Council of Allied Economic Research (NCAER). This annual survey was started in the mid-1980s for the purpose of assessing the market for various consumer goods. MISH was not designed for obtaining reliable consumption (or income) aggregates. For example, MISH does not include food consumption, so one cannot create consumption-poverty measures directly comparable to those from the NSS. The survey also asks for "total household income from all sources" in the form of a single question. Naturally this is a very difficult question to answer, and it is far from clear that the answers would be consistent or accurate, given the ambiguity in what "income" means (not least in rural areas of a developing country) and the influence of subjective factors and respondent knowledge of family income.7 However, working within these limitations, Lal et al. (2001) compare a series of poverty measures from MISH using a poverty line that gives the same poverty rate as the Planning Commission's 1987-88 estimate. Using the same deflators as the Planning Commission, Lal et al. report a decline in India's poverty rate from 39 percent in 1987-88 to 26 percent in 1997-98- clearly a steeper decline than indicated in Figure 1. While none of these data sources or methods can be considered conclusive on their own, it is compelling that they at least point to the same qualitative result that there was a non- negligible decline in India's poverty rate during the 1 990s. We will revisit this assessment once we have looked more closely at what has been happening at state level. 3. Disaggregating Poverty and Growth by State With some states of India larger than all but a few countries, one cannot be satisfied looking solely at all-India aggregates, as in Figure 1. We will first show that there is considerable diversity across states in rates of poverty reduction underlying Figure 1. We will then argue that rounds. They note that for a number of food consumption item groups, the mean expenditure from the EUS, despite the abridgement, is very sirmilar to that from the published 30-day recall numbers from the Consumer Expenditure Surveys. (If the published 30-day recall numbers were "contaminated" they should have been doubly higher, both due to a more detailed consumption module and a shorter recall.) This argument is suggestive but we do not think the case is fully proven. For the items where there is a substantial difference between the EUS and NSS estimates, the bulk of this difference may be still due to artificially higher 30-day estimates in the NSS rather than the abridgement of the EUS. In this case, the use of an unadjusted 30-day distribution with the 1999-00 data would overstate poverty reduction 7 Unfortunately, the micro data from MISH are not currently available to researchers outside the NCAER, so it has not been possible to further explore the survey results, as it has been with the NSS. 9 this offers some important insights into how recent economic growth has affected the incidence of poverty nationally, and what actions might be needed to assure a more pro-poor growth process in the future. 3.1 Geographic and Sectorcal Pattern Of Growth Let us look first at the gec,graphic pattern of India's growth process in the 1990s. The states that have had the slowest levels of growth in per capita GDP in the 1990s are two states with the lowest level of per capita GDP in the 1980s and the two states with the highest level of per capita GDP in the 1980s (Purnjab and Haryana). The high growth rates in the 1990s have been in the middle-income states. This pattern can be seen in Figure 2, which plots the mean annual growth rate from 1992-93 to 1999-00 against mean real GDP per capita over the 10 years 1980-89. However, ignoring the two states with highest GDP per capita in the 1980s, there is a strong positive relationship between level of per capita GDP in the mid-1980s and growth rate in the 1990s; that is, there is divergence in per capita GDP among all but the richest states of India. Including the two richest states, I1o simple linear relationship exists. It is not clear that Figure 2 tells us much about where the Indian economy is heading in the longer-term. In theoretical models, a decrease in the costs of trade can in some cases first lead to divergence between two trading regions, and then later to convergence (Baldwin et al., 2001). However, the regional irrLbalance evident in the 1990s growth process will be an important factor in the following analysis. It is also notable that agriculture as a whole has lagged the non-agricultural sector in the 1990s; while India's aggregate CDP grew at a rate of 6.7 percent per annum over the period 1993-94 tol999-00, agriculture and allied services grew at only 3.2 percent per annum. The importance of rural economic growth, and agricultural growth in particular, to poverty reduction in India has long been recognized.8 8 Empirical evidence linking measures of poverty to agricultural output can be found in Ahluwalia (1978), Bell et al., (1994), Ravallion and Datt (1996), Datt and RavaUion (1998b). 10 3.2 Trends in Poverty by State Let us first consider the evolution of rural and urban poverty rates. In India, as in most developing countries, the incidence of poverty has historically been higher in rural areas than urban areas. Prior to the 1990s there was little sign of a trend in the ratio of rural-to-urban poverty incidence. From 1960 to 1990, for example, the ratio of the rural poverty rate to the urban poverty rate in India hovered in the range of 1.1-1.2. However, over the 1990s, the ratio of rural-to-urban poverty rates has leaped up to 1.4 as shown in Figure 3. The 1990s have seen rising rural incidence relative to urban. Table 2 summarizes the trend rates of poverty reduction by state underlying Figure 1. The trend rates of poverty reduction have varied greatly across states, with Kerala the highest rate of poverty reduction (both as a proportion and in percentage points per year) and Assam the lowest, with Bihar close behind (and Janimu and Kashmir, in the linear case). Kerala's performance relative to Bihar is notable; around 1960, Kerala had one of the highest poverty rates, along with Bihar. By the mid- 1 990s, Kerala's poverty rate was almost half that of Bihar. 3.3 Cross-State Tests of Hypotheses about Poverty Reduction In understanding how India's growth process in the 1990s has affected national poverty, it is important to note that India's states vary enormously in terms of initial conditions that are relevant to how much impact economic growth will have on poverty (Ravallion and Datt, 2002). Those differences also lead one to expect that the sectoral composition of growth will matter more in some states than others; in a state with high literacy, for example, there could be more scope for reducing poverty through non-agricultural growth. We also need to allow for differences in other covariates of poverty. With these features in mind, our equation for studying how economic growth has impacted on poverty takes the following form. The dependent variable is the headcount index of poverty in state i at date t. The first explanatory variable is real non-agricultural product per head of the population in state i at date t. A second explanatory variable is a measure of average farm productivity, namely the real value of agricultural output per hectare. A third explanatory variable takes state government development spending into account. A fourth variable is the 11 inflation rate, which is known to have an adverse short-term adverse effect on the real wage rate for unskilled labor (Datt and Ravallion, 1998b). Consistently with Datt and Ravallion (1998a,b), we found that the fit of this model was improved if we used the current plus lagged values of non-agricultural product, farm yield, and lagged values of state development spending. Since we are interested here in modeling the evolution of poverty measures over time (rather than the level of poverty), we also include a dummy variable for each state in all regressions. To allow for any omitted (timne-trended) variables we also include state-specific trends. These variables are defined more precisely and justiiied more thoroughly in Ravallion and Datt (2002).9 We began by allowing all coefficients to vary across states. However, we found that we could not reject the null hypothesis (at the 10 percent level or better) that the coefficients on farmn yield, government spending, and. inflation were constant across states. Thus, we imposed the restriction that these coefficients, were constant, while allowing the coefficient on non- agricultural output per person to vary across states.1° We retained the state fixed effects and time trends. Table 3 presents the results of this restricted model. The model's explanatory power is clearly good; it accounts for over 90 percent of the variance in poverty incidence across states and over time. What do we learn from these results? Higher farm yields and higher development spending reduce the incidence of poverty, and the coefficients are highly significant. Higher non-agricultural output per person lowers poverty in all states. Higher inflation increases poverty. We also find significant positive time trends for 10 of the 15 states, which means that after adjusting for other factors, there was an upward trend on poverty rates over this time. Such a trend could arise from population pressure on agricultural land availability at given yield per acre or they could reflect an adverse distributional effect of population growth on poverty, as 9 We initially estimated the mc,del with an ARI error term, allowing for the uneven spacing of the surveys when estimating the autoregression coefficient (following the method in Datt and Ravallion, 1998a). However, the autoregression coefficient was not significantly different from zero so we set it to zero to simplify the estimation method. 10 We also tested whether the irnplied elasticities of poverty with respect to non-agricultural output per capita had changed over time. To do this we split the data into two periods, before and after 1980, and tested for stability. We found no significant difference in the elasticities between the two periods. The hypothesis of no change in elasticities is readily accepted (probability value of 0.27). 12 argued by van de Walle (1985). However, it might also reflect rising under-reporting of consumption in the NSS, as is thought to be the main source of the divergence between the NSS-based consumption estimates and those from the national accounts statistics. Upon closer examination, a population based explanation appears more likely. The positive time trends are not just a 1990s phenomenon, as has sometimes been claimed. On the contrary, the conditional trends are the strongest for the 1970s - when there was not much puzzle about the relationship between macroeconomic growth and poverty - but weaker for the 1 980s and 1 990s. In addition, Ravallion and Datt (2002) show that the time trends disappear or switch sign if one allows for population growth or total population effects. Perhaps the most striking finding from Table 3 is the variation in the elasticity of poverty with respect to non-agricultural output. The (absolute) elasticities vary from a low of 0.26 in Bihar to a high of 1.33 in Kerala. (Notice that the elasticities are twice the estimate from Table 3 because non-agricultural product enters as the sum of the current and the lagged values.) The next section will explore the implications of this variation in the non-agricultural output elasticities of poverty for understanding India's progress in poverty reduction during the 1990s. 3.4 India's Pattern of Growth and Aggregate Poverty Reduction Growth in large states with high levels of poverty is what matters most for aggregate poverty reduction. Table 4 compares the growth rates in non-agricultural output over 1993/94- 1999/00 with our estimates of the poverty-weighted elasticities of poverty incidence to non- agricultural economic growth based on Table 3. (Note that the elasticities from Table 3 have to be weighted by the states' shares of total poverty; the weighted elasticity then gives the impact on national poverty of growth in a given state.) There is no sign that the rates of growth were higher in the states where growth would have had greater impact on national poverty. Over the 14 major states, the correlation coefficient between the growth rate in non-agricultural output per person from 1993-94 to 1999-00 and the weighted (absolute) growth elasticity of poverty is -0.10, which is not statistically significant at any reasonable level. Figure 4 plots the estimates from Table 4. It is clear that the non-agricultural growth has not been concentrated in the states where it would have had the greatest impact on poverty nationally. A more pro-poor geographic pattern of growth in India's non-agricultural economy would have required higher growth in states such as Bihar, Madhya Pradesh, Orissa and Uttar Pradesh. Nor has 13 the geographic pattern of agricultural growth been particularly pro-poor. The states with higher growth in agricultural yields were not the key states with higher share's of India's poverty. Indeed, there is a mild negative correlation, although not statistically significant. Another way to perceive how the India's growth has not been sectorally and geographically distributed in such a way as to most benefit its poor is based on the regression results presented earlier in Table 3. Those results were based on data from 1960-94. By using the actual performance of the various explanatory variables from 1993-94 to 1999-00, one can estimate how far the poverty level should have dropped by 1999-00, if the historical relationships between these variables had continued to hold. When we carry out this exercise, we predict that the incidence of poverty in India falls from 39.1 percent in 1993-94 to 34.3 percent in 1999-00, implying a rate of reduction of about 0.8 percentage points per year." A rate of poverty reduction of 0.8 percentage points per year is slightly higher than the historical average for India. The average rate of decline in the poverty rate implied by the entire series of national measures in Figure 1 is 0.65 percentage points per year (with a standard error of 0.1 1). However, an annual rate of decline in the poverty rate of 0.8 percentage points per year is lower than one would have expected given India's growth rate in the 1990s and the historical elasticity of the national poverty rate to aggregate growth. As mentioned earlier, the elasticity of the poverty rate with respect to changes in per capita net national product, based on data from 1958 to 1991, was -0.75 (Ravallion and Datt, 1996). India's actual growth rate in net national product per capita was 4.8 percent per annum between 1993-94 and 1999-00, implying that the poverty rate would have fallen by 1.3 points per year over that period. Similarly, if one uses our model in Table 3 to estimate a counterfactual in which farm and non-agricultural sectors and all states have the same growth rate, given by the national rate, with all else remaining the same, then we predict rate of poverty r eduction of 1.2 percentage points per year. If not for the sectoral and geographic pattern of growil, India's macroeconomic growth rate in the 1990s would have delivered a rate of poverty reduction roughly double the historical trend. "1 The post-sample projections, including discussion of confidence intervals, are discussed further in Datt et al. (2002). 14 3.5 Why Does Economic Growth Benefit the Poor More in Some States? One of the most striking results presented earlier in Table 3 is how much the elasticity of poverty to non-agricultural output has varied across states. Why does poverty respond so much less to economic growth in some states than in others? A plausible explanation, with some support from cross-country regressions, is that certain types of initial inequalities can severely impede the prospects for growth-mediated poverty reduction (Ravallion, 2001). For example, pervasive credit market imperfections and greater initial inequality of assets (particularly of land) can mean that the growth that does occur is less poverty reducing. In addition, low basic education attainments are often identified as an impediment to the ability of the poor to participate in opportunities for economic growth. The initial income disparity between urban and rural sectors can also limit poverty reduction through growth in a dualistic labor market environment. This argument echoes a long-standing view (though not a dominant one in recent development thinking) that rural underdevelopment constrains prospects for poverty reduction through industrialization (for example, Clarke 1940). Initial urbanization however could have a positive influence on the poverty impact of non- agricultural growth by enhancing the poor's access to markets and infrastructure. One could also argue that higher initial farm yields and nonfarm product will promote tighter labor market conditions and help to bid up wages as economic growth increases demand for labor. One can test to see if initial conditions can explain the differences in the elasticity of poverty with respect to growth rates. In Ravallion and Datt (2002), we show that a number of conditions around 1960 -- the average farm yield, the ratio of urban to rural average consumption, the share of the rural population that is landless in the state, the state's infant mortality rate and the literacy rate - are significant predictors of the elasticity of poverty with respect to growth. Table 5 presents the key results which show that non-agricultural economic growth was less effective in reducing poverty in states with "poor" initial conditions in terms of rural development (in both absolute terms and relative to urban areas) and human resources. Thus, low farm productivity, low rural living standards relative to urban areas and poor basic education and health all inhibited the prospects of the poor participating in growth of the non-agricultural sector. 15 Amongst the conditions that are found to matter significantly to prospects for pro-poor growth, the role played by initial literacy is particularly notable. India's relatively poor performance in expanding literacy is well known (for example, Dreze and Sen, 1995; World Bank, 1997; PROBE Team, 1999; Swaminathan and Rawal, 1999). Our results reveal just how costly low educational attainmnent has been to India's poor, by retarding their capacity to participate in the opportunities that come with economic growth. For example, more than half of the difference between the elasticity of the headcount index of poverty to non-agricultural output for Bihar (the state with the lowest absolute elasticity) and Kerala (the highest) is altributable to Kerala's substantially higher initial literacy rate (Ravallion and Datt, 2002). Women's literacy matters no less than men's; indeed, women's literacy is a slightly more significant predictor of the elasticity of poverty with respect to economic growth. The need to combine human resource development with economy-wide policies favorable to growth has been well recognized in discussions of policies for fighting poverty (for example, World Bank, 1990). The revealed importance of human resource development as a precondition for pro-poor growth in India reinforces the concerns of Dreze and Sen (1995) and others that rapid poverty reduction in India will require more than economic reformi. The key message emerging from recent research is that achieving a policy environment conducive to growth interacts multiplicatively with human resource development; doing just economic reform or just human resource development one may achieve very little in terms of poverty reduction, but doing both can take a nation a long way. 4. Conclusion Our own estimates and our review of alternative estimates in the recent literature lead us to the conclusion that India has probably maintained its 1980s rate of poverty reduction in the 1990s. Our results suggest that the incidence of poverty has been falling at a little less than one percentage point over the main post-reform period. Using very different methods, other researchers obtain estimates of one point per year or higher. While none of the (multiple and methodologically diverse) attempts that have been made to assess the extent of bias in poverty measures for 1999/00 can be considered fully convincing on its own, they do point to significant poverty reduction in the 1990s though the question of acceleration of poverty reduction in this 16 decade remains contentious. Our own investigation fails to provide any evidence in support of such an acceleration. However, the basic question of measuring India's poverty rate has turned out to be harder to answer than it needed to be, because of difficulties with coverage and comparability of the survey data. There are lessons here for India and other countries about the need for assuring that potentially valuable experimentation and innovation with key survey instruments do not compromise their ability to provide reliable monitoring of how living standards are changing over time. There is no reason why such a high price needs to be paid for innovation in survey design; by simply assuring that, for a time, the same survey instruments is maintained for a sizable sub-sample (parallel to experimental samples) one can achieve both goals. Our investigation also emphasizes the considerable diversity in performance across states, with important clues for understanding why economic growth has not done more for India's poor. Our results suggest that, by and large, the (farm and non-farm) growth in India during the 1 990s has not been occurring in the states where it would have the most impact on poverty nationally. If not for the sectoral and geographic imbalance of growth, we estimate that national rate of growth would have generated a rate of poverty reduction that was double India's historical trend rate. We also find large differences across states in the poverty impact of any given rate of growth in non-agricultural output. States with relatively low levels of initial rural development and human capital development were not well-suited to reduce poverty in response to economic growth. Our results are thus consistent with the view that achieving higher aggregate economic growth is only one element of an effective strategy for poverty reduction in India. The sectoral and geographic composition of growth is also important, as is the need to redress existing inequalities in human resource development and between rural and urban areas. 17 References Ahluwalia, Montek S. 1978. "Rural Poverty and Agricultural Performance in India", Journal of Development Studies, 14(3): 298-323. Ahluwalia, Montek S. 2002. "Economic Reforms in India: A Decade of Gradualism," Journal of Economic Perspectives (forthcoming). Baldwin, Richard E., Philippe Martin and Gianmario I.P. Ottaviano. 2001. "Global Income Divergence, Trade and Industrialization: The Geography of Growth Take-Offs." Journal of Economic Growth 6(1): 5-38. Banerjee, Abhijit and Thomas IPiketty. 2001. "Top Indian Incomes, 1956-1998," mimeo, ENS- CEPREMAP Paris. Bell, Clive and R. Rich. 1994. 'Rural Poverty and Agricultural Performance in Post-Independence Indlia," Oxford Bulletin of Economics and Statistics 56: 111-133. Bhalla, Surjit. 2000. "Growth and Poverty in India - Myth and Reality", available from http://www.oxusresearch.com/economic.asp. Bruno, Michael, Martin Ravallion and Lyn Squire. 1998. "Equity and Growth in Developing Countries: 'Old and New Perspectives on the Policy Issues", in Income Distribution and High-Quality Growth (edited by Vito Tanzi and Ke-young Chu), Cambridge, Mass: MIT Press. Cashin, Paul and Ratna Sahay. 1996. "Internal Migration, Center-State Grants and Economic Growth in the States of India," IMF Staff Papers 43(1): 123-171. Chen, Shaohua and Martin Ravallion. 2001. "How Did the World's Poor fare in the 1990s?", Review of Income and 'Vealth, Vol.47(3), September 2001, pp.283-300. Clarke, Colin. 1940. The Conditions of Economic Progress, London: Macmillan. Datt, Gaurav. 1999. "Has Poverty in India Declined since the Economic Reforms?," Economic and Political Weekly 34 (December 11-17). Datt, Gaurav and Martin Ravallion. 1992. "Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s." Jouirnal of Development Economics, 38: 275-295. _ and .1997. "Macroeconomic Crises and Poverty Monitoring: A Case Study for India." Review of Development Economics, 1(2): 135-152. and .1998a. "Why Have Some Indian States Done 18 Better than Others at Reducing Rural Poverty?", Economica 65: 17-38. and .1 998b. "Farm Productivity and Rural Poverty in India", Journal of Development Studies 34: 62-85. Datt, Gaurav, Valerie Kozel and Martin Ravallion. 2001. "A Model-Based Assessment of India's Progress in Reducing Poverty During the 1990s," mimeo, World Bank. Deaton, Angus. 2001a. "Adjusted Indian Poverty Estimates for 1999-00," mimeo, Research Program in Development Studies, Princeton University. . 2001b. "Computing Prices and Poverty Rates in India, 1999-00," mimeo, Research Program in Development Studies, Princeton University. Deaton, Angus and Alessandro Tarozzi. 1999. "Prices and Poverty in India", mimeo, Research Program in Development Studies, Princeton University. DeLong, J. Bradford. 2001. "India Since Independence: An Analytic Growth Narrative," in Dani Rodrik (ed.) Economic Growth: Analytical Country Narratives. Princeton, NJ: Princeton University Press. Dreze, Jean and Amartya Sen. 1995. India: Economic Development and Social Opportunity. Delhi: Oxford University Press. Foster, James, J. Greer, and Erik Thorbecke (1984). A Class of Decomposable Poverty Measures, Econometrica 52: 761-765. Ghosh, Jayati. 2000. "Poverty Amidst Plenty?", Frontline 17(5), March 4-17 (available from www.the-hindu,com/fline/fll 705). Government of India. 1979. Report of the Task Force on Projections of Minimum Needs and Effective Consumption. New Delhi: Planning Commission. Government of India. 2000. Choice of Reference Periodfor Consumption Data, National Sample Survey Organization, March. Gupta, S.P. 1995. "Economic Reform and its Impact on Poor," Economic and Political Weekly 30 (22): 1295-1313. Jha, Raghbendra. 2000a. "Reducing Poverty and Inequality in India: Has Liberalization Helped?," Working Paper 204, World Institute of Development Economics Research.. . 2000b. "Growth, Inequality and Poverty in India. Spatial and Temporal Characteristics," Economic and Political Weekly 35 (March 11): 921-28. Joshi, V. and I. Little. 1996. India's Economic Reforms: 1991-2001; New Delhi: Oxford 19 University Press. Kulshreshtha, A.C., and Aloke Kar. 2002. "Estimates of Food Consumption Expenditure from Household Surveys and National Accounts," Central Statistics Office, Government of India. Lal, Deepak, Rakesh Mohan and I. Natarajan. 2001. "Economic Reforms and Poverty Alleviation. A Tale of T'wo Surveys," Economic and Political Weekly 36 (March 24): 1017-28. Natarajan, I. 1998. India Market Demographics Report, National Council of Applied Economic Research, New Delhi. Ozler, Berk, Gaurav Datt and Martin Ravallion. 1996. "A Database on Poverty and Growth in India," mimeo, Policy Research Department, World Bank. Planning Commission. 2001. "Poverty Estimates for 1999-00," htts ://planninecommi ssion.nic.in/Prfebthtm Ravallion, Martin, 1994. Poverty Comparisons. Chur, Switzerland: Harwood Academic Press. . 1996. "How Well Can Methodology Substitute for Data? Five Experiments in Poverty Analysis," World Bank Research Observer, 11: 199-222. . 2000a. "What is Needed for a More Pro-poor Growth Process in India?", Economic and Politicai' Weekly, 35 (March 25-31): 1089-1093. . 2000b. "Should Poverty Measures be Anchored to the National Accounts?" Economic and Political Weekly, 34 (August 26), 3245-3252. . 2001. "Growth, Inequality and Poverty: Looking Beyond Averages," World Development, 29(11): 1803-1815. Ravallion, Martin and Benu Bidani. 1994. "How Robust is a Poverty Profile?" World Bank Economic Review, 8(1): 75-102. Ravallion, Martin and Gaurav Datt. 1996. "How Important to India's Poor is the Sectoral Composition of Economic Growth?" World Bank Economic Review 10(1): 1-25. __ _ and _ . 2002. "Why Has Economic Growth Been More Pro-Poor in Some States of India than Others?", Journal of Development Economics, in press. 20 Sen, Abhijit. 2001. "Estimates of Consumer Expenditure and its Distribution: Statistical Priorities after the NSS 55th Round," Economic and Political Weekly 35 (December 16): 4499-4518. Srinivasan, T.N. 2000. "Growth and Poverty Alleviation: Lessons from Development Experience," mimeo, Yale University. Sundaram, K. 2001. "Employment and Poverty in 1990s: Further Results from NSS 55th Round Employment-Unemployment Survey, 1999-00," Economic and Political Weekly, (August 11), 3039-3049. Sundaram, K., and Suresh D. Tendulkar. 2001 a. "NAS-NSS Estimates of Private Consumption for Poverty Estimation: A Disaggregated Comparison for 1993-94," Economic and Political Weekly, (January 13), 119-129. and . 2001b. "Recent Debates on Data Base for Measurement of Poverty in India: Some Fresh Evidence," mimeo, Delhi School of Economics. Swaminathan, Madhura and Vikas Rawal. 1999. "Primary Education for All", in Parikh, Kirit S. (ed.) India Development Report 1999-00, New Delhi: Oxford University Press. Tarozzi, Alessandro. 2001. "Estimating Comparable Poverty Counts from Incomparable Surveys: Measuring Poverty in India," mimeo, Princeton University. van de Walle, Dominique. 1985. "Population Growth and Poverty: Another Look at the Indian Time Series Evidence", Journal of Development Studies 34(4): 62-85. Varshney, Ashutosh. 1999. "Mass Politics or Elite Politics? India's Economic Reforms in Comparative Perspective" in Jeffrey D. Sachs, Ashutosh Varshney and Nirupam Bajpai (eds) India in the Era of Economic Reforms, New Delhi: Oxford University Press. Visaria, Pravin. 1999. "Poverty in India During 1994-98: Alternative Estimates," mimeo, Institute of Economic Growth, Delhi, India, 1999. 21 Table 1: Elasticities of national poverty measures to economic growth in India 1958-91 EIeadcount index Poverty gap index Squared poverty gap Elasticity with respect to index Mean consumption from -1.33 -1.88 -2.26 national sample surveys (15.19) (12.83) (10.22) Mean private consumption -0.90 -1.36 -1.67 from national accounts (4.23) (3.98) (3.45) Mean net domestic -0.75 -1.15 -1.45 product from the national (3.68) (3.59) (3.27) accounts Note: Absolute t-ratios in parentheses. The headcount index is the percentage of people below the poverty lines discussed in the text. The poverty gap index is the mean distance below the poverty line as a proportion of the poverty line, counting the nonpoor as having zero poverty gap. The squared poverty gap indices is the measure proposed by Foster et al. (1984) in which the proportionate poverty gaps are weighted by themselves, to reflect the extent of inequality amongst the poor. The elasticities are based on regressions of first differences of the log poverty measures against first differences of the log consumption or net product per person using 33 surveys spanning 1951-91 for estimating the elasticity with respect to the surveys-based mean consumption, and 23 surveys spanning 1958-91 for estimating elasticities to consumption or income from the national accounts. The estimates based on the national accounts included a correction for differences in deflators in the form of an additional regressor, namely the difference in the rates of inflation implied by the consumer price index and the national income deflator. All regressions comfortably passed residual diagnostics tests for serial correlation, functional form, normality, and heteroscedas&icity. Source: Ravallion and Datt (1996). 22 Table 2: Unconditional trends in poverty incidence by state 1960-2000 Log-linear Linear Annualproportionate rate of Annual rate of change change (°/) (% points) Sate 1960-2000 Excluding the 1960- Excluding 55th round 2000 the 55th (1999/00) round (1999/00) Andhra Pradesh -2.38 -2.18 -1.02 -0.99 Assam -0.06 0.06 -0.06 -0.01 Bihar -0.32 -0.11 -0.20 -0.09 Gujarat -2.02 -1.57 -0.88 -0.77 Jammu & Kashmir -1.02 -1.02 -0.29 -0.29 Karnataka -1.54 -1.11 -0.70 -0.57 Kerala -3.26 -2.73 -1.45 -1.37 Madhya Pradesh -0.80 -0.63 -0.41 -0.34 Maharashtra -1.31 -1.01 -0.64 -0.53 Orissa -1.55 -1.59 -0.76 -0.77 Punjab and Haryana -2.96 -2.55 -0.70 -0.67 Rajasthan -1.49 -1.15 -0.67 -0.57 Tamil Nadu -1.92 -1.51 -0.85 -0.74 Uttar Pradesh -1.11 -0.88 -0.49 -0.41 West Bengal -2.29 -1.97 -0.87 -0.81 Total -1.43 -1.18 -0.65 -0.58 Note: Log-linear estimates are based on regressions of the log of the headcount index on time, while the linear estimates are based on regressions of the level of the headcount index on time. t-ratios in parentheses. Table 3: Regressions for state headcount indices of poverty in India, 1960-94 Independent variable Parameter t-ratio estimate Real agricultural output per hec.tare of net sown area: current + lagged -0.097 -3.50 Real per capita state development expenditure: lagged -0.128 -2.16 Inflation rate 0.392 4.71 Real non-agricultural output per person: current + lagged (NAG) Andhra Pradesh -0.141 -1.31 Assam -0.361 -2.30 Bihar -0.130 -2.02 Gujarat -0.289 -2.36 Jammu & Kashmiir -0.369 -3.29 Karnataka -0.332 -2.73 Kerala -0.665 -4.02 Madhya Pradesh -0.320 -3.83 Maharashtra -0.251 -2.34 Orissa -0.290 -4.63 Punjab and Haryana -0.426 -2.09 Rajasthan -0.270 -3.24 Tamil Nadu -0.272 -2.03 Uttar Pradesh -0.337 -4.14 West Bengal -0.511 -5.56 Time trend x 10-2 Andhra Pradesh 0.223 0.25 Assam 3.088 2.21 Bihar 1.530 3.72 Gujarat 1.575 1.75 Jammu & Kashmir 3.302 3.21 Karnataka 2.223 2.42 Kerala 2.500 2.16 Madhya Pradesh 2.611 4.48 Maharashtra 2.006 2.44 Orissa 1.266 2.38 Punjab and Haryana 2.339 1.24 Rajasthan 1.164 2.23 Tamil Nadu 1.545 1.46 Uttar Pradesh 2.172 3.80 West Bengal 0.979 1.94 Root mean square error .0937 R 2 0.923 24 Test for common non-ag. growth 1.59 elasticities across states: F(14,238) with p- (0.08) value in () Test for common time trends across states: 1.12 F(14,238) with p-value in () (0.34) Note: The dependent variable is the log of the headcount index of poverty (percentage below the poverty line) by state and date. All other variables are measured in natural logarithms. A positive (negative) sign indicates that the variable contributes to an increase (decrease) in the headcount index. The estimated model also included state-specific intercept effects, not reported in the Table. The number of observations used in the estimation is 272, using an unbalanced panel data for 15 states. 25 Table 4: Did the growth occur where it would have the most impact on poverty? (1) (2) (3) (4) (5) (6) State Headcount Share of Elasticity of Share % increase in % increase in index for national poverty to non- weighted non-agricultural farn output per 1993/94 poverty agricultural elasticity output per capita hectare 1993/94 (%) output (2)x(3) 1993/94-1999/00 1993/94- (Table 1) 1999/00 AndiiaPr&ish 29.5 6.' -0.281 =O.017 43.2 21.9 Assam 44.5 3.1 -0.722 -0.023 13.2 27.8 Bihar 60.3 16.7 -0.259 -0.043 29.3 0.5 Gujarat 33.7 4.4 -0.577 -0.025 38.1 29.9 Jammu & Kashmir n.a. n.a. -0.738 n.a. n.a. n.a. Karnataka 37.4 5.3 -0.663 -0.035 53.4 11.9 Kerala 28.8 2.6 -1.330 -0.034 73.5 64.0 Madhya Pradesh 44.0 9.2 -0.641 -0.059 18.1 26.8 Maharashtra 43.2 10.8 -0.502 -0.054 27.4 11.4 Orissa 40.3 4.0 -0.580 -0.023 26.3 36.7 Punjab 21.4 2.5 -0.852 -0.021 43.2 14.0 Rajasthan 43.3 6.1 -0.539 -0.033 44.3 64.5 Tamil Nadu 34.9 6.0 -0.544 -0.032 58.9 -6.4 Uttar Pradesh 40.1 17.8 -0.674 -0.120 37.3 26.1 West Bengal 25.9 5.5 -1.022 -0.056 49.4 83.0 Note: Output measures in (5) and (6) are deflated by the Consumer Price Index for Agricultural Laborers, for consistency with other data in the model. Table 5: How state-level initial conditions affect the elasticity of poverty with respect to non-agricultural output Increase in absolute elasticity of the state-level headcount index with respect to non-agricultural output per person ... ... due to an a 10% increase in ... Initial female literacy rate 1.53 (6.9) Initial urban-rural mean consumption disparity - 1.66 (2.3) Initial percent of rural landless households - 0.72 (2.8) Initial infant mortality rate - 1.01 (2.2) Initial yield per hectare 0.27 (2.4) Note: Absolute t-ratios in parentheses. The estimates are based on a regression for the log of the headcount index of poverty (across states and over time) on the same variables as in Table 3 with the difference that the state dummy variables interacting with non-agricultural output in Table 3 were replaced by variables describing initial conditions in the state around the beginning of the time period interacted with non-agricultural output. State effects and time trends were also included. For further details see Ravallion and Datt (2002). Figiure 1: Poverty Incidence in India, 1960-2000 70 - -Ce 60 - u g 0 5 x (D -c) 40 - 30 - 20 --, . . 1960 1970 1980 1990 2000 Figure 2: Growth divergence in the 1990s across all but the richest two states 8 0 0) 0) N 0) 0) 6 0 -c a) 0~~~~~~~~~ 400 00 2 CD Mean GDP per capita in the 1980s 28 Figure 3: Ratio of rural to urban poverty 1.5 C V_ o 1.4- a 1.3 CL Cu m 1.2 0 4- 1.1 - 1.0-,T 1960 1970 1980 1990 2000 Figure 4: Did the non-agricultural growth occur in the states where it would have had the most impact on poverty nationally? 14 cooO .0. 120 1 0 0 8 0~~~~~~~~ 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~ 0 co-O 6 4 -0. 0)4 C 0)~~~~~~~~~~~~ 0) ~ ~ ~ ~ ~ ~ 00 4- 2 - 0 o ~-0.14 -0.12 -0.10 -0.08 -0.06 -0.04 -0.02 0.00 Share-weighted elasticity of poverty to non-farm output, 1993/94 29 Policy Research Working Paper Series Contact Title Author Date for paper WPS2820 The Privatization of the Russian Igor Artemiev April 2002 V. Joseph Coal Industry: Policies and Processes Michael Haney 32155 in the Transformatcin of a Major Industry WPS2821 Income, Wealth, ard Socialization Daniel Lederman April 2002 P. Soto in Argentina: Provccative Responses 37892 from Individuals WPS2822 An Econometric Analysis of the David Mckenzie April 2002 C. Mendoza International Bank For Reconstruction 80599 and Development's Creditworthiness WPS2823 Real Exchange Rale Uncertainty Luis Serven April 2002 P. Soto and Private Investment in Developing 37892 Countries WPS2824 Trade Policy and Labor Services: Maurice Schiff April 2002 P. Flewitt Final Status Options for the West 32724 Bank and Gaza WPS2825 Demand for Imports in Venezuela: Mario A Cuevas April 2002 M. Geller A Structural Time Series Approach 85155 WPS2826 Potential GDP Growth in Venezuela: Mario A Cuevas April 2002 M Geller A Structural Time Series Approach 85155 WPS2827 Learning to Export: Evidence from Marcel Fafchamps April 2002 E. Khine Moroccan Manufacturing Said El Hamine 37471 Albert Zeufack WPS2828 Beyond Oaxaca-Blinder: Accounting Franqois Bourguignon April 2002 F Bourguignon for Differences in Household Income Francisco H. G. Ferreira 31056 Distributions across Countries Phillippe G. Leite WPS2829 The Spatial Distribution of Poverty Nicholas Minot April 2002 R. Bonfield in Vietnam and the Potential for Bob Baulch 31248 Targeting WPS2830 Labor Market Implications of Lodovico Pizzati April 2002 L. Pizzati Switching the CurrE-ncy Peg in a 32259 General Equilibriurri Model for Lithuania WPS2831 Health Outcomes in Poor Countries Limin Wang April 2002 L. Wang and Policy Options Empirical Findings 37596 from Demographic and Health Surveys WPS2832 Poverty and Survival Prospects Adam Wagstaff April 2002 H. Sladovich of Vietnamese Children under Doi Moi Nga Nguyet Nguyen 37698 WPS2833 School Choice, Student Performance, Emiliana Vegas April 2002 H Sladovich and Teacher and School Characteristics: 37698 The Chilean Case WPS2834 Investor Protection, Ownership, and Charles P. Himmelberg April 2002 K. Labrie the Cost of Capital R. Glenn Hubbard 31001 Inessa Love Policy Research Working Paper Series Contact Title Author Date for paper WPS2835 A Decade of Fiscal Transition Asad Alam May 2002 A. Panton Mark Sundberg 85433 WPS2836 Ethnic Minority Development in Bob Baulch May 2002 E. Khine Vietnam A Socioeconomic Truong Thi Kim Chuyen 37471 Perspective Dominique Haughton Jonathan Haughton WPS2837 Reform, Growth, and Poverty David Dollar May 2002 E. Khine in Vietnam 37471 WPS2838 Economic Mobility in Vietnam Paul Glewwe May 2002 E. Khine in the 1990s Phong Nguyen 37471 WPS2839 Marketing Externalities and M. Shahe Emran May 2002 F. Shilpi Market Development Forhad Shilpi 87476 WPS2840 Public Spending and Outcomes: Andrew Sunil Rajkumar May 2002 H. Sladovich Does Governance Matter? Vinaya Swaroop 37698 WPS2841 Contractual Savings in Countries Gregorio Impavido May 2002 P. Braxton with a Small Financial Sector Alberto R. Musalem 32720 Dimitri Vittas WPS2842 Financial Sector Inefficiencies Pierre-Richard Agenor May 2002 M. Gosiengfiao and the Debt Laffer Curve Joshua Aizenman 33363 WPS2843 A Practical Guide to Managing David Scott May 2002 L. Yeargin Systemic Financial Crises: A Review 81553 of Approaches Taken in Indonesia. the Republic of Korea, and Thailand WPS2844 Money Demand in Venezuela Mario A. Cuevas May 2002 M Geller Multiple Cycle Extraction in a 85155 Cointegration Framework WPS2845 The Spatial Division of Labor Marcel Fafchamps May 2002 F. Shilpi in Nepal Forhad Shilpi 87476