Rural poverty in India, 1973-86

C (c /l R 6.v (/r Policy, Research, and External Affairs WORKING PAPERS Welfare and Human Resources Population and Human Resources Department The World Bank October 1990 WPS 526 Rural Poverty in India, 1973-86 Nanak Kakwani and Kalinidhi Subbarao Growth (trickle-down) and poverty alleviation (pull-up) programs are not substitutes for -ach other, but complements, the Indian data on poverty show. The Pohcy. Research, and External Affairs Complex distnbutes PRE Working Papers to disseminate the fundi -gs of woik in progress and to enoourage the exchange of ideas among Bank staff and all others interested in development issues. These papess carry the names of the authors. relect only their views, and should be used and cried accordingly. The findings, interpretations, and conclusions are the authors' own. They should not he atinbuted to thc World Bank, its Board of Directors, its management, or any of its member countries.  Policy, Research, and External Affairs | Welfare and Human Resources WPS 526 This paper- a product of the Welfare and Hurman Resources Division, Population and Human Resources Department - is part of a larger effort in PRE to understand better the impact of general and targeted policies on poverty. Preliminary results of this study were reported in the paper"Poveny and Its Alleviation in India," in Economic and Political Weekly, 1990. This version was extended to cover 1986-87 and was substantially revised to accommodate the new evidence. Copies are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Brenda Rosa, room S9-137, extension 33751 (79 pages). The effects of economic growth can trickle down Just as increased inequality hurts the - but it rarely happens automatically, conclude ultrapoor disproportionately, so a decline in Kakwani and Subbarao, after assessing the inequality benefits the ultrapoor more than the impact of consumption growth on India's poor poor. From 1983 to 1987, growth was high and and ultrapoor between 1973 and :986. there was almost no change in inequality be- tween states. The growth effect dominated a Conversely, growth's beneficial effects on substantial decline in poverty. the incidence of poverty can, but nec d not, be offset or even nullified by increased inequality of Between 1973-74 and 1986-87, rural poverty consumption. In India, in W'73-77, they were. declined substantially. The incidence of poverty declined from 60.6% to 41.5% and its severity The policy response - a series of (the gap between the poverty line and an average antipoverty (consumption-equalizing) interven- poor person's income) fell from 18.8% to 10.5%. tions since the mid-1970s, aimed at raising the Even the absolute number of poor declined by income and consumption levels of the poor and about 37 million. The poverty ratio has become the ultrapoor- was basically sound. more responsive to (I) growth and (2) changing inequality in consumption, except in Bihar and In 1977-83, average consumption grew West Bengal. slowly but inequality of consumption fell in many states - and poverty and the poverty gap Both growth and poverty alleviation efforts were reduced more than in the earlier period. contributed to this success, conclude Kakwani Why is not clear, but the role of direct interven- and Subbarao. But on the whole investments tions cannot be minimized. and performance in health, education, and nutrition are unimpressive. It is their impression Program effectiveness is clearly weaker . that the social policies that can raise the capabili- the poorer states, however, and needs to be ties of the Indian people have generally been strengthened. Employment programs especially relegated to the background in Indian -which substantially increased rural employ- policymaking. ment and income growth -require more effort in Bihar and West Bengal. The PRE Working Paper Series disseminates the findings of work under way in the Bank's Policy. Research, and Extcmal AffairsComplex. An objective of the serics is to get these findings outquickly, even if presentations are less than fully polishied. The findings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. P'roduced by the l'RE Dissemination Center  Rural Poverty in India: 1973-86 by Nanak Kakwani and Kalinidhi Subbarao* Table of Contents 1. Introduction 1 2. Methodology 4 3. Trends in Inequality and Poverty: All India Rural 12 4. State-Wise Trends in Average Consumption Per Capita 18 and Inequality 5. Correlates of Consumption Per Capita and Inequality 23 6. Regional Trends in Poverty 29 7. Impact of Economic Growth and Inequality on Poverty 38 8. Responsiveness of Poverty to Growth and Inequality 46 Changes 9. Poverty Alleviation Policies 52 A. Policies for Labor Absorption 52 i. Agriz;,ltural Growth and Poverty 52 ii. Nonagricultural Growth and Poverty 55 iii. Direct Anti-Poverty Interventions 55 B. Nutrition, Health, and Other Social Sector 63 Interventions i. Nutrition 63 ii. Health 66 iii. Education 68 10. Concluding Remarks 74 References 77 * We are most grateful to Michael Lipton, Inderjit Singh, Wil Bussink, and Martin Ravallion for providing us with detailed written comments on earlier drafts of this paper. We would also like to thank Pranab Bardhan and Jacques van der Gaag for their comments and suggestions. We are indebted to Maria Felix for providing excellent secretarial assistance at all stages of this work.  Rural Poverty In India: 1973-861/ 1. INTRODUCTION Much has been written about poverty in India over the last two decades. Initially attention was concentrated on estimating poverty. Recently, more efforts have gone into explaining its variations, across regions and over time. While controversy still prevails over both estimation< and explanation, several insights have been gained - above all, that faster agricultural growth tends to reduce poverty. The central issues are two: Does overall economic growth reduce poverty, i.e., is there a "trickle-down" mechanism? Has economic growth been accompanied by increased inequality of income or consumption? In the past analysts have examined these issues by looking at the distribution of assets (land) or income or consumption in the high and slow-growing districts/states, often using single-year cross-section data. To understand the .mpact of economic growth on poverty, we should measure separately the impact on poverty, over time, of changes in average income (consumption) and in its inequality. This paper, therefore, examines trends in the growth of consumption and its distribution and assesses its relative impact on the poor and the ultra poor, over time and across states of India. During the period 1973-87, Indian GNP per person has grown by about /, Preliminary results of this study were reported in a paper entitled "Poverty and its Alleviation in India" in Economic and Political Weekly 1990. The present version is extended to cover 1986-87 and has undergone substantial revisions in view of the new evidence.  - 2 - 20 percent. The share in CNP of investment and current public expenditure has grown; the average private consumption per person has also grown annually at an average rate of 2 percent. If that had been so for all the households, it would have resulted in a significant reduction in poverty. However, if the distribution of per capita consumption had worsened, the impact of growth on poverty would be much less. The important questions, therefore aret How has the distribution of per capita consumption changed over time across the states, and to what extent has it nullified the beneficial impact of growth on poverty? How have the poor and ultra poor fared in the wake of changes in consumption and its distribution? What have been the regional patterns and to what degree can these patterns be explained considering our knowledge of the structural characteristics of these states and the state-wise performance indicators of anti-poverty programs? Using the National Sample Survey data for sixteen major states of India the paper addresses the above questions, in Part I. This is followed by an analysis of India's poverty alleviation strategies (Part II). The potential indirect role of agriculture and manufacturing, as well as the contribution of direct poverty alleviation interventions are analyzed, againbt the backdrop of the results pertaining to poverty and inequality in Part I. In particular, we examine the extent to which the regional allocation of funds for direct poverty alleviation programs is sensitive to the regional distribution of the poor and the ultra poor, and whether the recent evidence  on the impact of these programs is consistent with the observed patterns in poverty and inequality. So far the debate on poverty has focused exclusively in terms of "income" or "expenditure" needed to command the critical minimum basket of goods. This concentration of attention on "income" has resulted in the neglect of other influences -- :iotably education and health -- which expand and ensure the long-run capabilit-.es of the population. This paper also makes a modest attempt to relate these capabilities to the observed patterns in poverty at the state level. Thus, concluding sections of this paper deal with the trends in social sector expenditures and their outcomes, and assesses the performance of different states in celation to their perceived needs. The last section ties together the main findings and draws some inferences for policy.  2. METHODOLOGY To analyze poverty, we need to measure the economic welfare of each individual in the society. Although income i- widely used to measure economic welfare, it has many serious drawbacks.21 One major drawback of using income as a measure of economic welfare is that it may have substantial fluctuations which are averaged out in the long run. Therefore, it has been suggested that consumption is a better indicator of the actual economic position of a household than its current income.3/ The National Sample Surveys provide reasonably comparable time series data on the levels and distribution of household consumption expenditures. These data are available in grouped form, giving for each group: (a) the estimated number of persons, and (b) the average consumer expenditures in rupees per person. The monthly per capita expenditure levels are generally grouped into 13 to 14 experditure classes. To estimate poverty from such data, one needs to employ some intrapolation device. A commonly used procedure is to fit a density function to the entire consumption range and then compute poverty measures from the parameter of the fitted function. Most of the Indian studies have employed a two-parameter lognormal distribution (Minhas, Jain, Kansal and Saluja, 1987). The difficulty with this approach is that the lognormal distribution tends to overcorrect the positive skewness of 2/ For a detailed discussion of this issue see Kakwani (1986). 3" It would be more appropriate to use permanent income as a proxy for welfare or capability to escape poverty. Since there exists no reliable measure of permanent income, we have used household consumption expenditure, which is regarded as a more suitable measure of the household's economic welfare than the current income.  -5- the income distribution and thus fits poorly to the actual data. In the present paper we used a general interpolation device proposed by Kakwani (1980). This method utilizes, within each expenditure range, a separate, continuous differentiable function which exactly fits to the data points. The inequality and poverty measures are then coniputed by linking this function. We used a polynomial of the third degree to represent the Lorenz curve within each income class, except the first and the last open-ended classes. For the first and last expenditure ranges, a Pareto curve is used as a further refinement.4/ In this paper we used per capita household expenditure as a measure of household economic welfare.5/ Expenditure comorises all expenditures incurred by the household exclucively on domestic account including consumption out of home-grown produce or out of transfer receipts like gifts, loans, etc. The expenditure on household enterprises is excluded from consumer expenditure. Consumption out of home-grown produce is evaluated at a,/ It is not clear from the NSS whether the food given to poorer households (often permanent laborers) by richer households (large rural landowners and employers) gets "counted into" the consumption of the richer households. If this is so, the estimates of poverty and inequality will be overestimated. The fitting of the Pareto curve for the first and last expenditure ranges does not remedy this situation. The curve is used only as an interpolation device because information on individual households is not available. A better measure of household welfare will of course be the per equivalent adult consumption which corrects for the differing needs of adults and children. But this measure could not be employed because the NSS data were available only in grouped form (the groups formed on the basis of per capita household expenditure). We could have remedied this only by asuming that the ranking of households by per capita consumption is the same as that by consumption per equivalent adult. This assumption which is unlikely to hold will result in more serious estimation errors.  -6- ex-farm rates. Ideally, one should include the depreciated value of consumer durables which are in stock in the household. Because of non-availability of information on stock values, the monetary value of consumer d"rables acquired during the reference period is included in the total expenditure. This may in some cases distort the results on consumption inequality. When the index of household welfare is constructed, the next step is to determine the welfare of the individuals in the tiouseholds. In this paper ineividual welfare was derived by assigning every ..dividual in a household a welfare value equal to the per capita consumption for that household (Kakwani 1986). If there are severe intrahousehold inequities ;n the distribution of food and non-food items, poverty and inequality will bvth be underestimated. This problem could not be corrected because of non-availability of information concerning the intra-household distribution of resources. Once we have decided upon a suitable index of economic welfare for individuals, the next step is to find a threshold welfare level below which an individual is poor. In the present paper we have used two poverty lines. The Planning Commission (1979) has defined poverty threshold as the per capita monthly expenditure of Rs49.09 in rural India at 1973-74 rural prices. Following the Planning Commission, we adopted a round figure of Rs5O as our first poverty line. This poverty line corresponds roughly to the per capita expenditure which a typicall--structured, typically-spending household exactly meets its per capita daily requirement of 2,400 calories in rural areas. Our second poverty line is equal to the per capita consumption of households of less than 80 percent of the Planning Commission's poverty threshold of Rs5O. We define the households whose per capita consumption is below Rs4O per month  -7- as "ultra-poor"; those whose physical personal maintenance is unstable (Lipton 1988). The ultra-poor cut off paint of Rs4O corresponds closely to the poverty line used by Bardhan (1970) and Dandekar and Rath (1971), which is Rsl5 at 1960-6i prices (this line at 1973-74 rural prices is equivalent to the per capita expenditure of Rs42.5). To compare poverty across different tir.e periods, one needs to adjust the distributions given in current prices for price changes over time. Minhas (1970) and Dandekar and Rath (1971) used the national income deflator, which has been criticized by Bardhan (1974). This deflator does not reflect the prices which people actually pay for their consumption goods. We have followed Bardhan in using the Consumer Price Indices for Agricultural Labourers (CPIAL) prepared by the Labor Bureau. This index, constructed on the basis of the monthly retail prices of 75 consumer items collected from selected rural centers, uses the consumption pattern of rural agricultural labour households observed in 1956-57. A major objection that has been raised against this index is that the agricultural labor households constitute only about 30 percent of the total rural population and the remaining 70 percent of the rural households which constitute a large number of poor small farmers may hav;. a quite different consumption pattern (Minhas, et al 1987). This objection is, of course, relevant but Bardhan (1974) has observed that the consumption patterns of agricultural labor households are not very different from those of the other household groups in the rural areas. Recently, Minhas, et al (1987) have constructed a new price index which depicts the movement of consumer prices for the entire rural population. An attractive feature of this index is that it is based on the  -8- consumption patterns observed in more recent years than tha CPIAL. This index has been made available only recently for individual states. In the present paper we decided to use the consumer price index for agricultural labourers because it is the nnly index available for the most recent years 1986-87 for which the poverty estimates are presented here. Also since majority of the poor in rural India are agricultLral labourers, their consumption bracket will be closer to that of the poor.-/ Minhas, et al (1987) have applied the index to compute the incidence of poverty in the entire rural and urban areas of India for 5 different periods ranging from 1970-71 to 1983. They performed the poverty computations on the All India expenditure distributions which are published by the NSS separately for rural and urban areas. These aggregate distributions are derived from the state distributions by means of population weighted averages. This procedure of obtaining the aggregate distributions is nor satisfactory because of widely observed differences in price levels in different states. The expenditures in current prices given for each state have to be adjusted for state-wise price differences before they can be aggregated. Since this procedure was not adopted in the Minhas, et al study, their All India estimates of rural poverty are inaccurate notwithstanding the advance made on the construction of a new price index. To obtain comparable poverty levels across different states, we need 6/ Gaiha (1989) points out that since agricultural labourers are typically net buyers of food, the CPIAL can be expected to provide a close approximation to the prices confronting the net buyers of food among the rural poor.  -9- relative price levels in the rural areas of different staces in a given year. Bhattacharya and Chatterjee (1974) worked out these state-wise price relatives for the year 1963-64, which alco formed the basis for the Bardhan (1973) study. These price relatives were for the entire rural population. In the present paper we have used the price relatives (obtained from Minhas 1989) which were based on the weighting diagram of the middle group of the rural population. These estimates were considered more appropriate for estimating Table 1: Consumer Price Index For Agricultural LBouRERts and State Specific Price Relatives Price States Relatives Consumer Price Indices 60-61 73-74 77-78 83 86-87 ANDHRA PRADESH 100.9: 242 297 426 484 ASSAM 110.8 260 315 508 606 BIHAR 102 337 341 545 594 GUJARAT 112.1 246 285 429 530 HARYANA .104.7 273 332 497 602 JAMMU-KASHMIR 104.2 263 342 505 618 KARNATAKA 98.6 275 308 490 557 KERALA 106.6 276 317 531 665 MADHYA PRADESH I 93.8 309 345 506 585 MAHARASHTRA 105.5 276 320 482 58S ORISSA 97.9 322 351 555 616 PUNJAB 104.7 273 332 497 602 RAJASTHAN 103.3 284 327 459 565 TAMIL NADU 108.2 242 306 496 575 UTTAR PRADESH 94.6 303 336 506 593 WEST BENGAL 116 276 321 512 607 ALL INDIA 100 283 323 511 578 Source: For Price Relatives: Minhas and Jain (1989) and for Consumer Price Indices :Monthly Abstract of Statistics  - 10 - Table x State Specific Poverty Lines for Rural Areas (Rupees per person per month at current prices) Poor Ultra Poor States 73-74 77-78 83 86-87 73-74 77-78 83 86-87 ANDHRA PRADESH 43.2 53.0 76.0 86.3 34.5 42.4 60.8 69 ASSAM 50.9 61.7 99.5 118.7 40.7 49.3 79.6 94.9 BIHAR 60.8 61.5 98.3 107.1 48.6 49.2 78.6 85.7 GUJRAT 48.7 56.4 85.0 104.9 39 45.1 68 83.9 Haryana 50.5 61.4 92.0 111.4 40.4 49.1 73.6 89.1 Jammu-Kashmir 48.4 63.0 93.0 113.8 38.7 50.4 74.4 91 KARNATAKA 47.9 53.7 85.4 97.0 38.3 42.9 68.3 77.6 KERALA 52.0 59.7 100.1 125.3 41.6 47.8 80 100.2 MADHYA PRADESH 51.2 57.2 83.9 97.0 41 45.7 67.1 77.6 MAHARASHTRA 51.5 59.7 89.9 108.3 41.2 47.7 71.9 86.6 ORISSA 55.7 60.7 96.0 106.6 44.6 48.6 76.8 85.2 PLinjab 50.5 61.4 92.0 111.4 40.4 49.1 73.6 89.1 RAJISTHAN 51.9 59.7 83.8 103.2 41.5 47.8 67 82.5 TAMIL NADU 46.3 58.5 94.8 109.9 37 46.8 75.8 87.9 UTTAR PRADESH 50.7 56.2 84.6 99.2 40.5 44.9 67.7 79.3 WEST BENGAL 56.6 65.8 104.9 124.4 45.2 52.6 83.9 99.5 ALL INDIA 50.0 57.1 90.3 102.1 40 45.6 72.2 81.7] the incidence of rural poverty (Minhas 1989). Using these price relatives and the CPIAL available at the state l1evel (see Table 1) we worked out the state- specific poverty lines at the current prices for the years 1973-74, 1977-78, 1983, and 1986-87 (see Table 2).7/ The poverty line varies substantially across the states as well as over time, due entirely to differences in prices. For instance, West Bengal Z/ We performed calculations for both 1972-73 and 1973-74 years but in the present paper the results are presented only for 1973-74 because this was a normal agricultural year.  - 11 - has the highest poverty line for each of the 5 years; it means that West Bengal is the most expensive state for the poor to live. It also had the highest inflation rate particularly over the 1977-78 to 1983 period. The incidence of rural poverty for each state was first computed using the state-specific poverty lines given in Table 1. The poverty for All India rural level was then derived from the rural poverty levels computed at the state levels. The trends in inequality and poverty at the All India rural level are discussed in the next section.  - 12 - 3. TRBNDS IN INEQUALITY AND POVERTY: ALL INDIA RURAL Having decided upon the poverty line, we next compute poverty indices which would measure the intensity of poverty. The head-count measure, while widely used, is a crude poverty index because it does not take account of the income-gap among the poor. If the degree of misery suffered by an individual is proportional to the income shortfall of that individual from the poverty line, then the sum total of these shortfalls may be considered an adequate measure of poverty. Such a measure is called the poverty gap ratio and can be written as: g = fzg(x)f(x)dx = H(z-p*)/z (3.1) where g(x) = ( x), z being the poverty line, f(x) is the density function of income x, H is the head-count ratio and p* is the mean consumption of the poor* The measure g will provide adequate information about the intensity of poverty if all the poor are assumed to have exactly the same income, which is less than the poverty line. In practice, the income among the poor is unequally distributed and, therefore, g cannot be an adequate measure of poverty. To make g sensitive to the income inequality among the poor Sen (1976) proposed a poverty measure which led to a large theoretical literature on the measurement of poverty.81 The main difficulty with Sen's measure is 8/ See for instance Kakwani (1980), Clark, Hetmning and Ulph (1981), Foster, Greer and Thorbecke (1984) and Takayama (1979).  - 13 - that it is not additively decomposable. It is, therefore, awkward to compute its value for the All India rural population using the state level rural expenditure distributions. Moreover, additively decomposable poverty measures are uieful because they allow assessment of the effects of changes in sub- group poverty upon total poverty. In 1968, Watts proposed an additively decomposable poverty measure which can be obtained by substituting g(x) = (logz-logx) in (3.1): W = fz(logz-logx)f(x)dx (3.2) Although this is not a well-known measure, it is simple to compute and has all the important attributes: it satisfies Sen's monotonicity and transfer axioms and also Kakwani's (1986) transfer-sensitivity axiom. It is also closely related to income inequality. If, instead of z, we use i, (the mean income of the entire population), and evaluate the integral (3.2) over the whole range of x, we obtain: T = ro(logp-logx)f(x)dx (3.3) which is one of Theil's (1967) two inequality measures. An attractive feature of this measure is that it can be expressed as the sum of the between and within group inequalities. In the present exercise we have used only additively decomposable poverty and inequality measures. These measures have been employed to derive poverty and inequality in each of the 16 major states of India. The results  - 14 - Table 3: Per Capita Real Consumption,lnequality and Poverty Rural India 1973-74 to 1986-87 Annual Growth Rates(%) 73-74 77-78 83 73-74 73-74 77-78 83 86-87 to to to to Indicators 77-78 83 86-87 86-87 Per Capita Real Consumer 51.6 56.5 62.5 67 2.3 1.9 2 2 Expenditure Theil's Inequality Measure Within state inequality 12.3 15.6 13.6 14.5 6.1 -2.5 1.8 1.3 (93.2) (91.2) (88.3) (92.9) Between state inequality 0.9 1.5 1.8 1.1 13.6 3.4 -13.1 1.6 (6.8) (8.8) (11.7) (7.1) Total inequality 13.2 17.1 15.4 15.6 - 6.7 -1.9 0.4 1.3 (100) (: 00) (100) (100) Poor Head-count ratio(%) 60.5 56.2 47.7 41.5 -1.8 -2.9 -3.9 -2.9 Number of poor(millior's) 270.7 271.1 253.4 233 0 -1.2 -2.4 -1.1 Poverty gap ratio(%) 18.8 17.4 13.4 10.5 -1.9 -4.6 -6.7 -4.4 Ultra poor Head-count ratio(%) 41.2 38 29.6 23.8 -2 -4.4 -6 -4.1 Number of poor(millions) 184.7 183.5 157.6 134 -0.2 -2.7 -4.5 -2.4 Poverty gap ratio(%) 10.6 9.9 7.1 5 -1.7 -5.9 -9.5 -5.6  - 15 - obtained from individual states have then been aggregated to obtain the All India situation.9/ Table 3 presents the empirical results on inequality and poverty measures aggregated over the rural areas of 16 major states. Average consumption per capita rose by 2.3 percent per annum froni 1977-78 and then the growth rate stabilized at around 2 percent per annum. The total inequality in per capita consumption (measured by Theil's index) rose dramatically from 1973-74 to 1977-78, but fell from 1977-78 to 1983. However, from 1983 to 1986-87, inequality did not change lo/ significantly.- Between state inequality contributed 6.8 percent to the total inequality in 1973-74, which means that the within-state inequality is the major determinant of the inequality at the All India level. This suggests that policies aimed at redistribution of income and assets within the poorer states such as Bihar deserve greater emphasis. The share of between-state inequality grew, however, from 6.8 percent to 11.7 percent between 1973-74 and 1983, but fell dramatically to 7.1 percent between 1983 and 1986-87. Thus, the between state disparities in average living standards have shown a tendency to decrease over the most recent period. But this phenomenon is 9/ It is worth repeating that this procedure, unlike the procedures adopted by many previous researchers, appropriately takes account of the regional price variations while deriving poverty at the All India level. lo/ It needs to be stressed that these are rural growth and inequality trends. These changes trigger off changes in urban growth and inequality - an aspect not examined in this paper.  - 16 - accompanied by an increase in the within-State inequality, and, consequently, the net increase in the total inequality was negligible. Aggregate poverty in the sixteen major states has declined quite substantially from 1973-74 to 1986-87. The percentage of poor has decreased from 60.5 in 1973-74 to 41.5 in 1986-87. The pover:y gap ratio which takes into account not only the percentage of poor but also the consumption gaps of the poor, fell even faster. Even the absolute number of rural poor has declined substantially, from 270.7 million in 1973-74 to 233 million in 1986- 87. These figures demonstrate India's substantial achievements in poverty reduction, particularly in the 1980s. Further, poverty ratios for ultra poor fell to a greater extent than for the poor. From 1983 to 1986-87, the head-count ratio for the poor fell by 3.9 percent, but for the ultra poor by 6.0 percent. The trends in poverty gap ratio and Watts measures are similar. In general, the reduction in poverty is more for the ultra-poor than for the poor, and the magnitudes of reduction in poverty are higher after 1977-78 than before. This is interesting, because 1973-74 to 1977-78 was a period of higher growth (2.3 percent) with increasing inequality; 1977-83 was characterized by somewhat slower per capita growth of consumption (1.9 percent) but with a substantial decrease in within-state inequality. This decrease in inequality was the major factor behind the substantial reduction in poverty from 1977-78 to 1983. After 1983, the inequality did not change significantly, but growth at 2 percent led to an even greater reduction in poverty. While no significant changes in the pattern of growth would be observed as between 1973-74/1977-78 and 1977-78/1983, the latter period  - 17 - witnessed a spate of direct anti-poverty interventions. If this strategy were working with reasonable efficiency, one would expect a reduction in consumption inequality after 1977-78. This is indeed noticeable during the period from 1977-78 to 1983 in Table 3. Can it be attributed to the anti- poverty programs which started around 1977-78? Why then was there not a further reduction in inequality observed after 1983? To answer this and other related questions we need to analyze poverty trends at a disaggregated (state) level. This is done in the following sections.  - 18 - 4. STATE-WISE TRENDS IN AVERAGE CONSUMPTION PER CAPITA AND INEQUALITY State-wise trends in average consumption per capita are shown in Table 4. From 1973-74 to 1977-78, all states except Assam and Maharashtra showed a positive trend. That trend continued from 1977-78 to 1986-87 except for Bihar from 1977-78 to 1983 and Karnataka from 1983 to 1986-87. Kerala's performance has been consistently good; its per capita consumption Increased at an annual rate of 3 percent for the entire period (1973-74 to 1986-87). This is not surprising because Kerala received substantial overseas Table 4:Trends in Rural Real Consumption Per Capita at 1973-74 Prices Annual Growth Rates (%) Per Capita Consumption 73-74 77-78 83 73-74 States "(Rupees/month) to to to to 73-74 77-78 83 86-87 77-78 83 86-87 86-87 ANDHRA PRADES 58.7 65.8 76.5 81.6 2.9 2.8 1.9 2.6 ASSAM 51.1 48.3 56.5 58.9 -1.4 2.9 1.2 1.1 BIHAR 46.1 46.7 46.2 54.8 0.3 -0.2 5 1.3 GUJARAT 56 62.3 71.6 72.8 2.7 2.6 0.5 2 HARYANA 71.8 75.2 82.9 .89.2 1.2 1.8 2.1 1.7 JAMMU-KAS. iMIR 54.8 57.9 69.4 75.9 1.4 3.3 2.6 2.5 KARNATAKA 54.6 60.5 68.3 66.6 2.6 2.2 -0.7 1.5 KERALA 53.3 62.2 72.2 78.1 3.9 2.7 2.3 3 MADHYA PRADES 49.1 52.4 59.4 62.9 1.6 2.3 1.6 1.9 MAHARASHTRA 50.8 48.5 61.1 62.8 -1.2 4.3 0.8 1.6 ORISSA 38.3 43.2 51.5 53.6 3.1 3.2 1.1 2.6 PUNJAB 74.8 93.2 93.1 95.1 5.7 0 0.6 1.9 RAJASTHAN 61.7 70.6 76.2 76.2 3.4 1.4 0 1.6 TAMIL NADU 51.6 54.1 57 63.7 1.2 1 3.2 1.6 UTTAR PRADESH 50.6 60 60.3 70 4.4 0.1 4.4 2.5 WEST BENGAL 42 45.1 48.6 55.9 1.8 1.4 4.1 2.2 ALL INDIA 51.6 56.5 62.5 67 2.3 1.9 2 2  - 19 - remittances during this period.ll/ Table 5 presents the ranking of states by rural per capita consumption. The results show that Kerala has substantially improved its relative position from 9 in 1973-74 to 13 in 1986-87. The two largest states Madhya Pradesh and Utter Pradesh have also improved their relative position. The states whose relative position has deteriorated are Assam, Bihar, Gujarat, Rajasthan and Karnataka. The remaining states maintained more or less the same ranking. Table 5: Ranking of States by Rural Real Per Capita Consumption States Ranks 73-74 77-78 83 !,6-87 ANDHRA PRADESH 13 13 14 14 ASSAM 7 4 4 4 BIHAR 3 3 2 2 GUJARAT 12 12 11 10 HARYANA 15 15 15 15 JAMMU-KASHMIR 11 8 10 11 KARNATAKA 10 10 9 8 KERALA 9 1 1 12 13 MADHYA PRADESH 4 6 6 6 MAHARASHTRA 6 5 8 5 ORISSA 1 1 1 1 PUNJAB 16 16 16 16 RAJASTHAN 14 14 13 12 TAMIL NADU 8 7 5 7 UTTAR PRADESH 5 9 7 9 WEST BENGAL 2 2 3 3 / Whether or not these rural trends reflect state average real consumption per head depends on the trend in rural-urban inequality by state - an aspect not examined in this paper.  - 20 - We now turn to trends in inequality in rural per capita consumption. These are shown in Tables 6 and 7. Table 6 presents the levels and annual percentage change in inequality measured by the Gini index, and Table 7 by Theil's measure.121 of Table 6: State-wise Patterns in Inequality/Per Capita Consumption Gini Index 1973-74 to 1986-87 Annual Growth Rates (0/) Gini Index (