How important to India's poor is the urban - rural composition of growth?

J \ U d POLICY RESEARCH WORKING PAPER - 1399 - - .1u ' raIeconomicgru Ow-n-" How Important to Inciia's in 5- brogh -- ndi3-nt5-1;g :s. Poor is the Urban-Rural fargbtohe rura, and urapo y, %omposition of Growth? - ; Martin Ravallion Gau2rau Dau ib The World Bsik . Policy Research Departmnt- December 1994  POLICY RESEARCH WORKING PAPER 1399 Summary findings Views differ on how much India's poor have shared in Urban growth reduced poverty, but adverse the growth and contraction in the country's average distributional effects within the urban sector reduced the standard of living since independence. Some havc argued gains to the urban poor, and urban gro%vth had no that the rural growth that accompanied the green significant cffect on rural distribution. revolution in the 1970s and 1980s brought few gains to Rural growth was distribution-neutral within the rural the poor in the rural sector, while others have viewed sector and so brought sizable absolute gains to the rural agricultural growth as the key to rural poverty reduction. poor. Rural growth also had propoor distributional Views have also differed on how much urban growth has effects on urban poverty. benefited the poor. Identifying the nature of these intra- and inter-sectoral Ravallion and Datt used 33 household surveys effects reinforces the importance of rural growth to spanning 1951-91 to examine the relative importance to national poverty reduction. India's poor of both urban and rural consumption Future progress in fighting poverty in India will growvh. Among other things, they tested for spillover depend on both the rate of rural economic growth and effects between sectors: Does urban growth have the thie country's success in switching to a more propoor same effects on the rural distribution of consumption as process of urban growth. rural growth has on urban distribution? This paper - a product of the Poverty and Human Resources Division, Policy Research Department - is panr of a larger effort in the department to better understand the economywide and sectoral deteminant of progress in fighting poverty. Copies of the paper are available free from the World Bank, 1818 1- Street NW, Washingmn, DC 20433. Please contact Patricia Cook, room NS-061, extension 33902 (29 pages). December 1994. The PFolcy Resarch Workhg Paper Serics dmmate findings of work in proress to encwge the echane of ida abou development ses An objective of the srie is to get the findgsot guickly, aem ife prEsentatiare ls thanfily polisbed The papers cany die names of she authors and should be used ad citdaccrdi. The findngs. ihterpmtations, and condusionsarethe authos'own and should not be attributed to the World Bank. its Executie Board of Detora , or any of its neiber cotries Produced by the Policy Research Dissemination Center  How Important to India's Poor is the Urban-Rural Composition of Growth? Martin Ravallion and Gaurav DatC Policy Research Department, World Bank Abstract We use 33 household surveys spanning 1951-91 to examine the relative importance to India's poor of urban versus rural consumption growth. Urban growth reduced poverty, but adverse distibutional effects wiftiin that sector mitgated the gains to the poor, and urban growth had no significant effect on rural distrbution. Rural growth was distibution neutral withi that sector, and so brought sizable absolute gains to the rurAl poor; rural growth also had pro-poor distributional effects on urban poverty. The nature of these intra- and ir-sectoral effects ts reinforced the importan of rural growth to national poverty reduction. For their comments we dtank Lyn Squire, T.N. Srinivasan and Dominique van de Walle. These are the views of the authors, and should not be attnrbuted to the World Bank The support of the Bank's Research Committee (under RPO 677-82) is gratefully acknowledged.  I Introduction Views differ on how much India's poor have shared in the growth and contraction in the county's average stndard of living since Independence. Some observers have argued that the growth in average living standards in rural areas that accompanied the green revolution in the 1970s and '80s brought few gains to the poor within that sector, while others have pointed to agriculumal growth as the key to rural poverty reducton. Views have also differed on how much urban economic growth has benefitd the poor; for example, the optimism of many of India's post-indepdence planners that ihe country's (largely urban-based) indusialization would bring lasting longer-term gains to the poor has not been shared by many critics then and since. In all this, the cross-sectoral effects may be crucial to the dis nal outcomes. The fortunes of the poor in each sector are linked in various ways-through trade, migration, and sfers-tO the living standard of both poor and non-poor housebolds in the oxer sectr. This paper endeavors to throw some new empirica light on the intra- and cross-sectoral effects of urban and rural economic growth on poverty in India, by an econometric analysis of new time series data spnniing 40 years. We quantify te effects of changes in average consumption on poverty within each of the urban and rural sectors. But an important part of our motivation is also to test for the existence of spilover effects between sectors; does urban growth have the same effects on rual distution as rural growth has on urban distribution? To help motvate our empirical ust, the following section descrbes various ways that cross-sectoral spillover effects might occur. In section 3, we briefly describe our data, comprising our own estimates of a consistent tme series of poverty meaures for urban and rural areas of India for 1951-91. This is followed by a discussion of our econometric model in section 4. before presenting the results in section 5. Our conclusions are summarized in section 6. I  2 Cross-sectoral spillover effects of growth on poverty For the class of additively decomposable poverty measures, national poverty is a population-weighted sum of rural and urban poverty. The direct iInpact of urban (rural) growth on national poverty is thus limited by its population share. However, in principle, it is also well recognized that growth and contraction in the affluence of one sector or region of an economy can have pervasive spillover effects elsewhere, with potentially wide ranging implications for poverty reduction. An insne of this is the often heard view that an important cause of urban poverty in developing counties is rural poverty. By this view, the fortunes of the urban poor are closely linked to their rural countrparts through various forms of interaction with the effect that poverty is in part "shared". The vast urban shlms of many third world cities are (by this view) smply the urban analogue of the deprivation (often on a larger scale) in the rural hinterland) The existence of such cross-sectral spillover effects implies tat the total impact of growth in one sector on aggregate poverty can exceed or fall short of its direct effect. It also impies that the signficance of the urbma-rural compositon of growth for poverty goes beyond what is implicit in the sectoral population sire. We will be concerned with empiically assessing the diredon and magnitude of such cross-sectoral distributional effects. There are a mber of ways in which spill-over effects between urban and rural distributions can occur: Labor mobility between the two sectors can yield an equilbrium relationship between the real wages of similar workers, entailing some degree of "horizontal integration" in the earings and income distributions-the living standards of people in different sectors but at similar levels of living are caulsally related. Even without labor mobility, such ' For a survey of the literature on poverty in developing countries, including comparisons between urban and rural poverty, see Lipton and Ravallion (1994). 2  integration can also arise through trade in goods; the living standards of households in different sectors but sharing similar factor endowments will tend to move together to the extent that trade in goods eliminates differences in factor costs at the margin. Transfer behavior can also produce horizontal integration through income sharing of related households living in different sectors. All such effects may operate either through changes in the sector's own mean, or changes in the sector's distribution around the mean, which may be due to growth and contraction in the other sector's mean. Plainly, the existence of horizontal intgration suggests that changes emanating from one sewtor may well have powerful effects on the absolute levels of living in another sector. Tie linkage can occur at any level; when it exists amonst the poor in diferent sectors one can interpret it as 'shared poverty', a term borrowed from Geertz (1963) (who used it in an inlra- zral context). When the degree of horizontal integration varies by the level of livig, one can also expect growth or contraction in one sector to induce shift in the Lorenz curve in the other sector. There is no a priori reason to expect the integration to be uniform at all levels. And there is at least one good reason to expect that it will not be: distributions of absolute levels of lving in different sers tend to overlap imprfety i.e., they share a positive density over certin (compact) intervals of the range of living stadards, but not others. The urban sector of a developing country will often incude an elte which simply has no conlerpart in the rural sectr.2 When combined wilth shared poverty in the overlappMg intrval of te distribution, this can have stong implications for how an increase in incomes in one sector will spill over to affec both average levels of living, and iequalities widlin other sectors. 2 Similarly, when comparing comties at very different levels of development there may be no countepr in the richer country to the poorest stra of the other country. 3  Combining these observations, we postulate that the observed level of poverty in each sector depends on the mean consumptions in both sectors. 3 Data The extent to which the poor share in a rising average standard of living has been a source of great debate in India, as elsewhere. Mucb of the debate has been informed by little more than anecdotal observations, or by systematic anwiyses of small non-representative samples. Fotunately, a sufficiently long time series of reasonably comparable and nationally representative consumption surveys exists for hIdia to permit a systematic empirical investigation of the issue (Govemment of India, 1990; Bhauaharya et al., 1991); indeed, India is the only developing country for whih one can say that 3.1 The consumption distions We use a new time se of poverly measures for rural and urban India over the period 1951 to 1991. This is based on cuption distribuons from 33 household surveys conducted by the National Sample Survey (NSS) Organization, beginnig with the 3rd round for August to November 1951, we use distibuins up to the 47th round for July to December 1991? This series significantly improves upon the what bas been the-most widely-used time series on povery measures in India to date4, due to Abluwalia (1978. 1985). The Ahluwalia series was a rural poverty series giving estimates of the headcout index and Sen's poverty measure for 13 rounds 3 The first tWo rounds of the NSS covered rual areas only. 4 Including Gffin and Ghose (1979), Saitb (1981), van de WaLe (1985), Abluwalia (1985), Desai (1985), and other papers in the collecdon edited by Mellor and Desai (1985). 4  spanning 1956-57 to 1977-78. Our new series provides a sectorally-disaggregatzed consistent time series for the entire period 1951-1991 on a range of poverty measures within the Foster, Greer, Thorbecke class (more on this later). Datt (1994) describes in detail how this series was estinated, so we will be brief here. A number of intrinsic limitations of these data should be noted: 1) Poverty is measured solely in terms of consumpton, though it is a comprehensive measure, following sound and consistent survey and accounting practices. The underlying NSS data do not include incomes, though it can be argued that current consumption is a better indicator of living standards than current income.5 Noneteless, there are various wnon-income dimesions of well-being that this measure cannot hope to capture, and we say nothing here about how responsive thes ohr dimensions may be to growth.' -i) We are solely ,oonered with the effects on poverty of growfh in average consumption; in particular, we do not look at the direct effects of COMe grow. That distinction may be important, since the existece of inter-temporal consumption smoothing behavior may make poverty (m terms of consption) less responsive to income growth tan consumption growth, at least in the short-term While current c on may well be a good predictor of the trend in cine,7 deviations from current income must be expected, and the respon of poverty easures to changes in current income may be of indeendent interesL s Particularly in this setting; for an overview of the arguments why see Ravallion (1994). Using village panel data from India, Cbaudhuri and Ravallion (1994) find that current consumption and mcome are better indicators of chronic poverty than other measures tested, though the choice between consumption and income is less clear. Even so, it can be argued that current consumption is the better indicator of current level of living. 6 For further discussion and references see Anand and Ravallion (1993). 7 For supportive evidence (for the US) see Cochrane (1994).  However the nature of our data-notably that incomes were not surveyed, and that survey periods canot be readily mapped into an annual time series for comparison with national accounts or other data-pretty much deterrnmines the choice. iii) It should be noted that we do not decompose the sources of growth any further than the urban-rural spiit. We do not distinguish (for example) technical progress from expanding input usage. The NSS data do not allow such a breakdown, and other data sources are not easily integrated with the NSS survey rounds.8 iv) The average sample size over the 33 NSS surveys is 10,988 households in urban areas and 18,691 in rural areas. However there is considerable variation over time. The urban samples range from 514 to 58,162 whle for mrual areas the range is 1,361 to 99,766. In both cases, the smallest sample was in 1953 (though different rounds), while the largest was for 1977- 78. From 1955 on, all samples exceeded 1,000. v) We only usc the classification of 'urban' and 'rural' areas built into the NSS tabulations.' However, over such a long period some rural areas would naturally have become urban areas.?