Size of land holding, living standards and employment in rural Western India, 1972-73

Size of Land Holding, Living Standards and Employment in Rural Western India, 1972-73 SWP459 World Bank Staff Working Paper No. 459 May 1981 Prepared by: Pravin Visaria, Consultant Development Research Center Copyright (D 1981 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. The views and interpretations in this document are those of the author , 0 1J2 and should not be attributed to the World Bank, to its affiliated K organizations, or to any individual acting in their behalf.   The views and interpretations in this document are those of the author and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting on their behalf. WORLD BANK Staff Working Paper No. 459 May 1981 SIZE OF LAND HOLDING, LIVING STANDARDS AND EMPLOYMENT IN RURAL WESTERN INDIA, 1972-73 This study analyses the relationship between the size of land holding possessed by a household and the household's level of per capita expenditure in rural areas of Gujarat and Maharashtra States of India from October 1972 to September 1973, on the basis of the 27th Round of the Indian Nationa3 Sam- ple Survey. According to this study, although land is usually considered to be the most critical determinant of living standards in rural India, differences in per capita land can explain only a very small proportion of variance in the monthly per capita expenditure of households. The likely reasons for these findings are noted; they are hypotheses for further research. The incidence of poverty, measured on the basis of a norm adopted by the Indian Planning Commission in the early 1960's, declines with an increase in the size of land holding possessed; but it remains significant even among households with ten acres or more of land, particularly in V'aharashtra. As expected, the incidence of poverty is the highest among agricultural labour households and the lowest among rural white collar households. The study examines various characteristics of the households or popula- tion according to the size of land holdings. Noteworthy results include the finding that the differences in the illiteracy rates of the population aged 15 and over, and in the school enrollment rates for ages 5-19, according to size of holding, are smaller than is generally believed, particularly in rural Gujarat. On the other hand, the structure of employment shows signifi- cant differences according to the size of land holding possessed by the house- hold; and the incidence of unemployment varies inversely with the size of land holding, with significantly large differences. Prepared by: Pravin Visaria, Consultant Development Research Center Copyright 63 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A.   Acknowledgements Thanks are due to Professor V.M. Dandekar, Chairman of the Governing Council of the National Sample Survey Organization, and P.B. Buch, M.A. Telang and S.M. Vidwans, former or current Directors of the Bureaus of Economics and Statistics in Gujarat and Maharashtra, for providing me with an opportunity to study the 27th Round state sample data analysed in this paper. Several basic tables used for the preparation of this paper were compiled at the Gujarat Computer Centre in Gandhinagar, India, with the cooperation of a large staff working under the leadership of P.B. Buch. Others who have helped in the statistical work include R. Murti Pemmarazu, Shyamalendu Pal, Robert E. Sterrett, Jr. and Cynthia Hwa. The editorial advice of Vivianne Lake has improved the presentation of the material in- cluded in this paper. I am grateful to all of them. I have greatly benefitted from many discussions with Montek S. Ahluwalia, B.S. Minhas and T.N. Srinivasan at various stages of the prepara- tion of this paper. Others, whose advice was particularly useful during the earlier stage of preparation of data for analysis include Professor M.L. Dantwala, my teacher and colleague at the University of Bombay, and Sudhir Bhattacharyya of the N.S.S. Organization. Thanks are due to S.K. Sanyal of the NSS Organization for his comments on an earlier draft of the paper. However, the responsibility for all errors rests with me alone. Pravin Visaria   TABLE OF CONTENTS Page No. I. INTRODUCTION AND SUMMARY 1 A. The Data and Their Limitations . . . . . . . . . . . 1 B. An OutZine of the Paper and Summary of Results. . . . . . 3 II. EARLIER STUDIES ON THE RELATIONSHIP BETWEEN LAND AND POVERTY 8 III. SIZE OF LAND HOLDING, HOUSEHOLD OCCUPATION AND POVERTY 17 A. Land Holding and PCE DeciZe . . . . . . . . . . . . 17 B. Land Holding and HousehoZd Occupation. . . . . . . . . 17 C. HousehoZd Occupation and PCE DeciZe . . . . . . . . . 19 D. Size of Land HoZding and the Proportion of Poor Households. . 19 E. Distribution of the Poor HousehoZds by Occupation and Size of HoZding . . . . . . . . . . . . . . . 28 IV. RESULTS OF MULTIVARIATE ANALYSIS 30 A. MuZtipZe Regression AnaZysis. . . . . . . . . . . . 30 B. ResuZts of MuZtipZe CZassification AnaZysis. . . . . . . 35 C. Factors UnderZying the Weak ReZationship Between Land and Poverty. . . . . . . . . . . . . . . . . 37 V. SALIENT CHARACTERISTICS OF HOUSEHOLDS OR POPULATION BY SIZE OF LAND HOLDING 40 A. Demographic Characteristics . . . . . . . . . . . . 40 1. Average Household Size . . . . . . . . . . . . 40 2. Average Household Size by Household Occupation . . . . 42 3. Age Composition of Household Heads . . . . . . . . 42 4. Age Distribution of the Population . . . . . . . . 43 5. Sex Composition of Population . . . . . . . . . . 43 6. Proportion of Female Heads . . . . . . . . . . . 44 B. Literacy, Education and School EnroZZments . . . . . . . 45 1. Literacy and Educational Attainment . . . . . . . . 45 2. School Enrollment Rates . . . . . . . . . . . . 47 C. Economic Characteristics . . . . . . . . . . . . . 49 1. Labour Force Participation Rates . . . . . . . . . 50 2. Labour Force Participation Rates by Occupation and PCE Decile . . . . . . . . . . . . . . 52 3. Incidence of Unemployment . . . . . . . . . . . 53 4. Unemployment by Age . . . . . . . . . . . . . 58 5. Incidence of Unemployment by Occupation and PCE Decile. 59 6. Status of Class of Worker Distribution of the Employed. 60 7. Occupational Distribution of Workers. . . . . . . . 61  Page No. VI. SOME IMPLICATIONS OF FINDINGS 63 ANNEX 1. DATA ON THE SIZE DISTRIBUTION OF LAND HOLDINGS IN GUJARAT AND MAHARASHTRA--1972-73: AN ASSESSMENT 65 Definitions. . . . . . . . . . . . . . . . . 66 The Data. . . . . . . . 67 Stratification Scheme for the Rural Ze . . . .73 The Discrepancy Between the Expected and Actual Size of the SompZe . . . . . . . . . . . . . . . 74 ANNEX 2. MULTIPLE REGRESSION ANALYSIS: SOME SUPPLEMENTARY RESULTS 76 Cropping Intensity as a Dummy VariabZe. . . . . . . . 76 Regression ResuZts for Different Regions . . . . . . . 76 ANNEX 3. DISTRIBUTION OF HOUSEHOLDS BY SIZE OF HOUSEHOLD LAND HOLDING AND PER CAPITA LAND 83 STATISTICAL APPENDIX . . . . . . . . . . . . . . . . . 86  LIST OF TABLES Page No. II. EARLIER STUDIES ON THE RELATIONSHIP BETWEEN LAND AND POVERTY Table 1. Percentage of Poor Households in Rural India by Size Class of Operational Land Holdings, 1958-59 12 Table 2. Results of Multiple Regression Analysis of Per Capita Consumption with Size of Landholdings and Household Size as Explanatory Variables, India: 1958-59 14 Table 3. Results of Multiple Regression Analysis of Per Capita Consumer Expenditure (ACE) with Per Capita Land (in Acres) and Household Size as Explanatory Variables, Rural Gujarat and Maharashtra, 1972-73 16 III. SIZE OF LAND HOLDING, HOUSEHOLD OCCUPATION AND POVERTY Table 1. Distribution of Rural Households by Size of Land Holding Possessed by the Household and Household Occupation Group, Gujarat and Maharashtra, 27th Round of the NSS, 1972-73 20 Table 2. Percentage of Poor Households by Size of Class Land Holding Possessed, Rural Gujarat, 1972-73 21 Table 3. Percentage of Poor Households by Size Class of Land Holding Possessed, Rural Maharashtra, 1972-73 22 Table 4. Percentage of Households Below the Poverty Line by Land Possessed and Household Occupation, Rural Gujarat and Maharashtra, 1972-73 25 Table 5. Regression Coefficients for the Percentage of Poor Households with the Average Size of Land Holding as an Explanatory Variable, Rural Gujarat and Maharashtra, 1972-73 27 Table 6. Percentage Distribution of Poor Households by Household Occupation and Size of Landholding, Rural Gujarat and Maharashtra, 1972-73 29 IV. RESULTS OF MULTIVARIATE ANALYSIS Table 1. Coefficients of Correlation between Selected Variables According to Household Occupation (Major Source of Income), Rural Gujarat and Maharashtra, 1972-73 31 Table 2. Quadratic Specifications for Monthly Per Capita Expenditure by Household Occupation Groups, Rural Gujarat and Maharashtra, 1972-73 33  LIST OF TABLES (continued) Page No. Table 3. Results of Multiple Classification Analysis of Monthly Per Capita Expenditure (MPCE) in Rural Gujarat and Maharashtra, 1972-73 36 V. SALIENT CHARACTERISTICS OF HOUSEHOLDS OR POPULATION BY SIZE OF LAND HOLDING Table 1. Selected Characteristics of Households or Population by Size of Land Holding Possessed, Rural Gujarat and Maharashtra, 1972-73 41 Table 2. Correlation and Regression Coefficients Showing the Association Between (i) The Proportions of Illiterates and Educated Up To Primary School Among Persons Aged 15 and Over, and (ii) The Size of Land Holdings, Rural Gujarat and M4aharashtra, 1972-73 46 Table 3. Correlation and Regression Coefficients Showing the Association between School Enrollment Rates and the Size of Land Holding, Rural Gujarat and Maharashtra, 1972-73 48 Table 4. Correlation and Regression Coefficients Showing the Association Between (i) Labor Force Participation Rates (Based on Usual Activity) and (ii) The Size of Land Holdings, Rural Gujarat and Maharashtra, 1972-73 51 Table 5. Correlation.iand Regression Coefficients Showing the Association Between (i) the Incidence of Unemployment in Terms of Current Activity and Persondays and (ii) the Size of Land Holding, Rural Gujarat and Maharashtra, 1972-73 54 Table 6. Quadratic Speicifications for the Incidence of Unemploy- ment in Terms of Persondays with the Size of Land Holding as the Explanatory Variable, Rural Gujarat and Maharashtra, 1972-73 56 ANNEX 1. Table 1. Distribution of Rural Households of Gujarat and Maharashtra According to the Size of Land Holding, 26th and the 27th Rounds of the NSS, 1971-72 and 1972-73 68 Table 2. Average Size of Holding in Rural Gujarat and Maharashtra, 1971-72 and 1972-73 70 Table 3. Distribution of Households Not Operating Any Land by Size of Owned Land Holding, in Rural Gujarat, Maharashtra, and India, 1972-73 (26th Round) 72  LIST OF TABLES (continued) Page No. ANNEX 2. Table 1. Quadratic Specifications for Monthly Per Capita Expenditure by Household Occupation Groups, Rural Gujarat and Maharashtra, 1972-73 77 Table 2. Distribution of Households by Household Occupation (Major Source of Income), Rural Gujarat and Maharashtra, 1972-73 79 Table 3. Average Per Capita Expenditure, Per Capita Land and Household Size by Region and Household Occupation, Rural Gujarat and Maharashtra, 1972-73 80 Table 4. The R for Monthly Per Capita Expenditure with Specified Explanatory Variables, by Household Occupation and Different Regions, Rural Gujarat and Maharashtra, 1972-73 81 ANNEX 3. Table 1. Distribution of Households by Size of Household Land Holding and Per Capita Land, Rural Gujarat and Maharashtra, 1972-73 84   I. INTRODUCTION AND SUMMARY This paper is complementary to the paper on "Living Standards, Employment and Education in Western India, 1972-73", prepared earlier. It is concerned with only the rural areas of the two states of Gujarat and Maharashtra and examines the relationship between the size of land holding and rural living standards as well as some of the salient charac- teristics of the households (and/or population) with holdings in different size classes. A. The. Dat,a dnd Thaiv*.'Zitat&n,. The data examined here were gathered from the "state samples" canvassed in the 27th Round of the National Sample Survey (NSS) from 5543 house- holds in rural Gujarat and 5329 households in rural Maharashtra.-/ The comparability of the data available for analysis with other sources of information was examined in Annex 4 to the earlier paper. However, the available 27th Round data on land holdings have been compared with those provided by the 26th Round conducted in 1971-72 in Annex 1 below. It is shown there that although the 27th Round questions on land possessed were intended to collect information on operational holdings, the data actually gathered seem to pertain to ownership holdings; otherwise, non-agricultural holdings, mainly house sites or adjacent land, were included. Whatever the reason, the proportion of households with land holdings of up to 0.49 acres seems to be overstated; when grouped data are used, such misspecification of land holdings would affect the category of households with up to one acre 1/ The number of sample households for rural Maharashtra has been stated as 5532 in the earlier paper. Three households were dropped from the sample during subsequent editing.  -2- of land. These limitations might partly explain our finding that although land is usually regarded as the most critical determinant of the living standards in rural India, differences in per capita land can explain only a very small proportion of the variance in the monthly per capita expenditure (NPCE) of households. 0 B. An OtZine of the Paper and Summart.w' 6f Results Section II reviews briefly the earlier studies on the relationship between land and poverty. that were based on the analysis of the NSS data collected in the Eleventh and Fourteenth Rounds (conducted during 1956-57 and 1958-59, respectively). The results of the multiple regression analysis of per capita expenditure in rural areas of Gujarat and Maharashtra show that the responsiveness of expenditure to an increase in per capita land possessed by a household was almost the same in 1972-73 (in real terms) as that esti- mated for rural India as a whole in 1958-59. In relative terms, however, per capita expenditure in the two states seems more responsive to household size than to per capita land. Section III of the paper begins with an examination of the dis- tribution of households with land holdings of different sizes according to MPCE deciles. It reports the incidence of poverty or the proportions of poor households by the size of land holdings for (a) "regions" within the states, and (b) occupation (or the major source of income) of the household. The distribution of the poor households by size of land holding and house- hold is also shown. The incidence of poverty is the highest among agricultural labour households and the lowest among white collar households. The agricultural labour households account for only 30 to 40 percent of all (rural) house-  -3- holds below the poverty line in the two states. Over 40 percent of the poor households in Maharashtra and almost 55 percent of those in Gujarat are cultivating households of which 27 to 35 percent (forming about 14 to 15 percent of all poor rural households) have land holdings of 10 acres or more. Thus, while the incidence of poverty declines with an increase in the size of land holding, it remains significant even among households with 10 acres or more, particularly in Maharashtra. The interregional differences in the proportions of the poor households show the Eastern region of Gujarat (with a predominance of tribal population) to be the worse off; the Saurashtra region (with its agriculture dominated by the cash crop of groundnuts) had the luwest proportion of poor households. The incidence of poverty was almost equal in the other three regions. In Maharashtra, the Marathawada region, generally regarded as the most backward and having a very low percentage of cultivated land under irrigation, had the lowest proportion of poor households; the Bhandara and Chanda districts with a fair amount of irrigation facilities followed. The other four regions had almost the same proportion of-poor. Section IV and Annex 2 report the results of the multiple regres- sion analysis of MPCE with per capita land and household size as the main explanatory variables and irrigation status or cropping intensity in the village as the dummy variables. 'The percentage of variance in NPCE (i.e. R ) explained by the quadratic'regression-is no more than 20 percent for rural Gujarat and 11 percent for rural Maharashtra, when all households are considered'together, and 22 and 8 percent, respectively, if the regression analysis is restricted to households reporting cultivation as their major source of income.  -4- There are significant differences in the R values for different regions within the two states. For the Southern Plains of Gujarat, and Bhandara and Chanda in Maharashtra, 36 and 30 percent of the variance in MPCE is explained by the variables considered in the regression. (The * values for cultivating households are higher, 46 and 35 percent, respectively.) But the dummy variables have a noteworthy impact only in three of the eleven regions (see Annex 2). Results of a multiple classification analysis of MPCE are also shown in Section IV. When the NSS region, the household occupation, and the age, (general) education, usual activity (a combination of status-and farm/ non-farm work) and "skill" of the household head are added to per caDita land and household size as the explanatory variables, the R for NPCE increases to 33 percent for rural Gujarat and 16 percent for ruial Maharashtra. Clearly, further work is necessary to identify better predictors of the consumption level. 2 The R values appear to be smaller than might be expected on a priori grounds. The likely reasons are the heterogeneity of land and the non-availability of data on irrigation and on cropping intensity status of the land holding of each individual household. Such data at the village!level are not a good proxy for the household level information. Among other things, the available PCE data do not include any imputation for the rental value of owner-occupied dwellings. (The 1961 Census data show almost 90 percent of the rural households in the two states to be owners of their dwellings.) Insofar as the large landholders tend to have larger and better quality houses, their expenditure is underreported to a greater extent. The response errors, non-availability of data on remittances and/or the extent of out-  -5- migration as well as the drought conditions during the first three quarters of the survey year are also likely to lower the explanatory power of the land variable. Section IV of the paper reports on various characteristics of the households or population (such as size, literacy, school enrollment rates, the level of labour force participation and unemployment, as well as the occupational and status distribution of workers) according to the size of land holdings. This analysis retains the size of land holding possessed by the household as the classificatory variable partly because most of the special programmes devised to assist the rural poor (such as the Small Far- mers' Development Agencies - SFDA's - and the Marginal Farmers and Agricul- tural Labourers' Agencies - MFAL) use the criterion of household land holding rather than per capita land. (Annex 3 indicates that if the households were grouped into some combination of deciles and quintiles, the ranking of the bottom and the top groups of households, having less than one acre or more than 20 acres, would remain unchanged but many of the households in the intermediate size classes of holding would be ranked differently.) Both in Gujarat and Maharashtra, the average household size in- creases with the size of land holding. The median age of the household head also tends to rise with the size of land holding, probably because when households split up, the division of land among children leads to smaller land holdings for the younger successors. But there is little systematic association between the size of land holding and the age composition (the dependency ratio) or sex ratio of the population. Female heads of households are, however, overrepresented among the landless or small land holders, per- haps because a majority of them are widows.  -6- The differences in the illiteracy rates of the population aged 15 and over, and the school enrollment rates from ages 5-19, according to size of holding, are smaller than is generally believed, particularly in Gujarat. It is interesting to find that land holding no longer seems to have a strong influence on school enrollments. The labour force participation rates for all ages together do not show any systematic variation with the size of land holding, probably because there are no systematic differences in the dependency ratio associated with land. The incidence of unemployment (measured in terms of week-long unem- ployment, and particularly that in terms of the persondays of unemployment), however, varies inversely with the size of land holding, with significantly large differences. The differences in the incidence of unemployment are partly a reflection of the sizeable differences in the distribution of the employed according to class of worker and occupation. Consistent with the general impressions, the proportion of employees (and particularly "casual labourers" among whom the incidence of unemployment is much higher than the average in rural areas l/) varies inversely with the size of land holding whereas that of the self-employed and family helpers rises with the size of land holding. Essentially, these same relationships are reflected in the occupational dis- tribution of workers: the proportion of cultivators steadily rises whereas that of agricultural labourers falls as the size of land holding increases. These differences in the occupational and status composition of workers as well as the incidence of unemployment should indeed be captured by 1/ Pravin Visaria, "Poverty and Unemployment in Rural India: An Analysis of Recent Evidence,"'World Development, forthcoming.  -7- such variables as "household occupation" and the "usual activity" of the household head. Some of the characteristics of the households and popula- tion by household occupation have been examined in Sections III and V. These data confirm the expected overrepresentation of the white-collar households in the top deciles and the overrepresentation of agricultural labour house- holds in the bottom deciles. However, when the household occupation and the usual activity of the household head are considered jointly with some other variables in the multiple classification analysis, the two variables ranked very low among the explanatory variables of variance in MPCE in rural Gujarat. It is hoped that further work will help to identify more important deter- minants of MPCE. In conclusion, the relatively low explanatory power of per capita land (admittedly undifferentiated land in terms of quality or the extent of irrigation and cropping intensity) with respect to the variance of monthly per capita expenditure has come as a surprise. Pending further work, how- ever, the study suggests the need for great caution in accepting the size of land holding owned and/or operated by a household as an index of its economic status and in relying upon land reform as the means to ameliorate poverty and reduce inequality in rural areas. The need to standardize the measure of land size, taking account of the quality and productivity of land, is obvious. This point has also been recognized in several earlier studies on the relationship between land and rural inequality. It is hoped also that future surveys of the NSS will try to collect more information on owned and operated land holdings as well as on other sources of income together with consumption expenditure in order to improve our understanding of the deter- minants of rural living standards.  -8- II. EARLIER STUDIES ON THE RELATIONSHIP BETWEEN LAND AND POVERTY The relationship between the size of land holding and the level of per capita expenditure or poverty has been studied earlier on the basis of the NSS data collected in the Eleventh and Fourteenth Rounds (conducted) during the periods August 1956-March 1957 and July 1958-June 1959, respec- tively.) The Eleventh Round tabulation of households possessing cultivated land (including leased in land, and therefore relating to operational holdings) by levels of per capita expenditure was examined first by B.S. Minhas.-/ He showed that the proportion of people below the poverty line was only 40 percent among those with operational holdings of less than one cent of an acre-/ but between 83 and 72 percent among households with operational land holdings of 0.01 to 4.99 acres. Minhas' estimate of the proportion of the poor among households with less than one cent of an acre (40 percent) is not quite correct because rural households not possessing (operating) any cultivated land were excluded from 1/ B.S. Minhas, "Rural Poverty, Land Distribution and Development Strategy: Facts and Policy," Indian Economic Review (New Series), vol. 5, no. 1 (1970). Minhas had applied the Eleventh Round estimates of the propor- tion of poor households in each (operational) land holding class to his estimates of the proportion of persons in each class. However, within each land holding class, the average size of households is inversely related to the level of per capita expenditure. This is certainly evi- dent in our data for Gujarat and Maharashtra. Minhas' procedure would therefore involve an understatement of the number of poor persons in different size classes of holdings. Of course, his estimates of poverty based on the consumption expenditure data of the rural population or agricultural labourers are unaffected by the use of the Eleventh Round data. 2/ The number of sample households with operational holdings of up to 0.005 acre in the Eleventh Round was probably no more than 12 (0.21 percent to the total after the weights were applied) and some scepticism about their per capita expenditure distribution is in order. One cent of an acre equals one hundredth of an acre.  -9- the Eleventh Round sample used for tabulation.-/ Such non-land operating households comprise essentially (a) the agricultural labour households not operating any land,-/ (b) the households of artisans and (c) non-agricultural white collar workers. The first two of these groups are known to suffer from a higher incidence of poverty than the land-operating households.-/ If the data for the entire sample had been tabulated, the difference in the pro- portion of the poor among households with operational holdings of up to 0.005 acre (including the landless) and those with 0.01 to 4.99 acres would have been less than shown by Minhas. It has not been possible to work out correct estimates because the proportion of the poor households (with a monthly per capita expenditure of Rs. 15 or less - in 1956-57 prices) in the total rural 1/ The data pertained to 5495 sample households from 1858 villages and excluded 1780 households which did not possess any cultivated land. The latter constituted 31.5 percent of the (weighted) total number of rural households and 24.4 percent of the rural population during July 1955 - June 1956 (the fixed reference period). See: The National Sample Survey, Report No. 140, Tabtes with Notes on Some Aspects of Agriculture in India: Eleventh Round, August 1956-February 1957 (RuraZ) (New Delhi, 1969), pp. 3-4, 12-13, 19-21. The relevant NSS tables are confusing because they label the households possessing less than 0.01 acre of land as those with "0.00" land and do not make it explicit that households not operating any land are excluded. 2/ Some 57 percent of the agricultural labour households surveyed in the Eleventh and Twelfth Rounds of the NSS did not operate any land. See: The NSS, No. 46, Tables with Notes on Consumer Expenditure of Agricul- turaZ Labour Households in Rural Areas: EZeventh and Twelfth Rounds, August 1956-August 1957 (Delhi, 1961), p. 42. The agricultural labour households were defined as those for whom "agricultural wage employment constituted the major source of income". 3/ The average size of households not possessing or operating any cultivated land was 3.8 persons higher than the 2.6 reported for households having operational holdings of up to 0.005 acres, who were included among those for whom the tabulation was done. The proportion of the poor varies in- versely with the size of the household and would be higher among the former.  -10- sample for the Eleventh Round was only 52.41/, lower than the 71.9 percent estimated for households possessing some cultivated land.-/ If both figures are correct, the proportion of the poor among households which did not possess any cultivated land would be no more than about 10.1 percent, an implausibly low figure.- Unfortunately, it is not possible to examine the data in any greater detail. Subsequently, in 1971 Dandekar and Rath have used the data collected in the Eleventh Round of the NSS during 1956-57 to examine the proportion of the poor households among small land-owners (excluding those operating less than 0.01 acre of land) and the data from both the Eleventh and the Twelfth Rounds for a similar estimate for agricultural labour households. They con- cluded that "in spite of these rather old and fragmentary data, it is clear that the rural poor consist predominantly of the agricultural labour house- holds and small land holders with cultivated holdings of less than 5.0 acres and particularly less than 2.5 acres."k-/ According to them an important cause of rural poverty is the "lack of land resources."-/ 1/ The NSS, No. 77, Tables with Notes on Consumer Expenditure: Eleventh Round: August 1956-February 1957 (Delhi, 1963), p. 14. 2/ The NSS, No. 140, TabZes with Notes on Some Aspects of Agriculture in India: Eleventh Round, August 1956-February 1957 (RuraZ). 3/ The proportion of agricultural labour households with a monthly per capita expenditure of Rs. 15 or less during 1956-57 was about 70.9 percent. It is not possible to identify the distribution of agricultural labour house- holds not operating any land by per capita expenditure. See: NSS, No. 46, Tables with Notes on Consumer Expenditure: EZeventh Round: August 1956- February 1957 (Delhi, 1963), pp. 33-34. 4/ V.M. Dandekar and Nilkantha Rath, Poverty in India (Poona: Indian School of Political Economy, 1971), p. 16. The two groups were presumed to include also the "village artisans progressively thrown out of their traditional employment." 5/ Ibid., pp. 12-13.  -11- Vaidyanathan has presented the Eleventh Round data on the per capita expenditure and the size of operational land holding (land possessed) to emphasize that the "rich households" (defined in terms of their per capita expenditure) "do not consist exclusively of big land holders nor are all big land holders 'rich'."1' His point is corroborated by a special tabulation of the Fourteenth Round data (for about 98 percent of the total rural sample households surveyed during 1958-59); it shows the distribution of households by land operated and the level of per capita consumption expenditure. Although the Fourteenth Round data used by Vaidyanathan were not weighted, they were more comprehensive than the Eleventh Round data and covered the households not operating any land.-/ Table 1 shows the proportion of the poor households from the Fourteenth Round, defined again as those with a per capita expenditure of Rs. 15 or less, for each land holding class.-/ Evidently, poverty was most prevalent among households not operating any land or operating holdings of less than 0.5 acres. The incidence of poverty was only slightly lower among households operating holdings of between 0.5 and 1.0 acres. Holdings 1/ A. Vaidyanathan, "Inequalities in Living Standards in Rural India," in Poverty and Income Distribution in India, T.N. Srinivasan and P.K. Bard- han eds. (Calcutta: Statistical Publishing Society, 1974). 2/ The proportion of rural households operating land up to 0.49 acres was 28.65 percent in Vaidyanathan's 14th Round table, but almost 37.0 per- cent according to the Eleventh Round (31.52 percent did not operate any land; and of the remaining 68.48 percent households, 7.96 percent oper- ated land holdings up to 0.49 acre). Vaidyanathan has noted that "bulk of missing households (for whom data were not available to him) appeared to be in the lower income group." Even if all of them had operational holdings of less than 0.5 acre, the Fourteenth Round estimate of the class falls short of that from the Eleventh Round or the Sixteenth Round (35.95 percent). 3/ An adjustment for the price rise between 1956-57 and 1958-59 (about 7.2 percent according to the index number of wholesale prices) would raise these figures; but unless one can use different price indices for different per capita expenditure classes, the observed relationship between land and the proportion of the poor would not be affected.  -12- TABLE 1 PERCENTAGE OF POOR HOUSEHOLDS IN RURAL INDIA BY SIZE CLASS OF OPERATIONAL LAND HOLDINGS, 1958-59 Size of Household Operational Percentage of Poor Households Percentage Distribution Number Holding (with NPCE of Less than Rs. 15 of Households of Sample (Acres) in current prices) Poor Households All Households Households up to 0.49 51.7 37.3 28.6 2127 0.50 - 0.99 48.8 6.2 5.1 375 1.00 - 2.49 40.7 19.4 18.9 1402 2.50 - 4.99 35.2 16.8 18.9 1400 5.00 - 7.49 32.0 7.5 9.3 693 7.50 - 9.99 34.3 4.4 5.1 379 10.00 - 12.49 32.5 3.5 4.2 314 12.50 - 14.99 18.9 0.8 1.6 122 15.00 - 19.99 24.7 2.0 3.3 243 20.00 - 24.99 16.9 0.7 1.8 130 25.00 - 29.99 20.9 0.5 0.9 67 30.00 - 49.99 19.6 0.8 1.6 122 50.00 & More 7.8 0.1 0.7 51 All 39.7* 100.0 100.0 7425 * The proportion of poor households in rural India in 1958-59, with due account taken of the weights, was 41.9 percent. See: The NSS, No. 102, TabZes with Notes on Consumer Expenditure Fourteenth Round, July 1958 - June 1959 (Delhi, 1966), pp. 20-21.  -13- of between 2.5 and 12.5 acres of land implied almost the same incidence of poverty (around 32-35 percent), which declined to between 17 and 25 percent of households with holdings up to 50 acres. Even among households with more than 50 acres of land, about eight percent appeared to be poor. Given the positive association between land holding and household size, the households with large land holdings probably enjoy some economies of scale in consump- tion and, therefore, a uniform poverty line may overstate the proportion of the poor among larger land holders. However, the inverse relation between the holding size and productivity per acre, partly attributable to the greater availability and use of irrigation facilities on smaller holdings, is a relevant explanatory factor. Vaidyanathan also analysed the Fourteenth Round data through a multiple regression analysis of per capita consumption with the size of land holdings and household size as explanatory variables, separately for four occupation groups of farmers, farm workers, professional and administrative workers and others. The results of his regressions are reproduced in Table 2. The R2 values or the percentages of explained variance are not shown by Vaidyanathan but all the coefficients are significant at the one percent level. The results of a multiple regression analysis of per capita expen- diture in terms of per capita land and household size for rural households in Gujarat and Maharashtra in 1972-73 can be compared with those of Vaidyanathan for rural India as a whole in 1958-59.11 Taking advantage of the relatively large sample for the two states, the regressions have been estimated separ- ately for the regions within Gujarat and Maharashtra, by household occupation 1/ It seems likely that Vaidyanathan's analysis used the size of household land holding rather than per capita land as one of the explanatory variables.  -14- TABLE 2 RESULTS OF MULTIPLE REGRESSION ANALYSIS OF PER CAPITA CONSUMPTION WITH SIZE OF LANDHOLDINGS AND HOUSEHOLD SIZE AS EXPLANATORY VARIABLES, INDIA: 1958-59 Regression Coefficients of Land Family holdings size 1. Farmers 0.422 -1.204 2. Farm workers 0.335 -1.377 3. Professional and administrative workers 0.782 -1.523 4. Others 1.11 -2.338 Source: A. Vaidyanathan, "Inequalities in Living Standards in Rural India," in Poverty and Income Distribution in India, T.N. Srinivasan and P.K. Bardhan eds. (Calcutta: Statistical Publishing Society, 1974), p. 228.  -15- (determined on the basis of the major source of income). While the regional results are discussed elsewhere, Table 3 summarizes the estimates for the two states. The regression coefficients for land variables in our analysis are generally higher, except for households pursuing manual non-agricultural occupations. The coefficients for household size are also higher, except for cultivator households in rural Maharashtra. To a considerable extent, this seems to be the effect of inflation; the wholesale price index (1961-62 = 100.0) had increased from about 90.2 in 1958-59 to 207.1 in 1972-73 and 252.4 in 1973-74 (April to March in all cases). The percentage of explained vari- ance in per capita expenditure is rather low, particularly for rural Maha- rashtra. It is necessary to take account of other variables, as is done in Section IV below. In relative terms, per capita expenditure in rural Gujarat and Maharashtra seems more responsive to household size than to per capita land. Depending on the household occupation group, a one percent decrease in the average household size would raise per capita expenditure by between 0.22 and 0.38 percent, whereas a similar increase in average per capita land would increase per capita consumption by between less than 0.01 and 0.16 percent. (Except for cultivator households in rural Maharashtra, the absolute value of the elasticity of consumption with respect to household size exceeds that with respect to per capita land.) On the whole, our findings for the two states are broadly compar- able to Vaidyanathan's results for rural India as a whole during 1958-59. Section IV below reports other results of a multiple regression analysis based on a quadratic specification and including irrigation status of the village as a dummy variable.  TABLE 3 RESULTS OF MULTIPLE REGRESSION ANALYSIS OF PER CAPITA CONSUMER EXPENDITURE (ACE) WITH PER CAPITA LAND (IN ACRES) AND HOUSEHOLD SIZE AS EXPLANATORY VARIABLES, RURAL GUJARAT AND MAHARASHTRA, 1972-73 Zero Order Regression Coefficients of Coefficient Correlation of Coefficients Number of Household Per Capita Household Constant Determination PCE PCE Sample Occupation Land (PCL) Size (HS) Term . (R2) & PCL & HS Households Rural Gujarat All 5.31 -2.56 62.36 0.16 0.34 -0.25 5543 (60.6) (42.0) (149.5) Cultivation 4.95 -2.11 60.32 0.16 0.37 -0.26 2184 (48.1) (27.2) (100.9) Agricultural 3.68 -2.76 54.20 0.17 0.18 -0.38 2186 Labour (13.7) (31.2) (106.3) White Collar 25.35 -5.60 95.40 0.43 0.60 -0.30 336 Work (33.9) (16.2) (48.8) Other 0.59 -3.59 71.33 0.10 0.04 -0.32 837 (0.9) (20.6) (69.3) Rural Maharashtra All 2.48 -1.77 50,06 0.07 0.20 -0.17 5329 (49.6) (41.2) (188.2) Cultivation 2.44 -0.65 44.53 0.05 0.23 -0.10 1742 (37.0) (9.6) (91.8) Agricultural 2.92 -2.52 47.69 0.15 0.13 -0.37 2420 Labour (22.5) (58.6) (198.7) White Collar 6.59 -4.65 89.89 0.10 0.14 -0.25 305 Work (12.0) (17.7) (58.4) Other 0.57 -3.33 58.10 0.14 0.05 -0.37 862 (3.6) (35.8) (114.6) Note: Figures in parentheses show the 't' values.  -17- III. SIZE OF LAND HOLDING, HOUSEHOLD OCCUPATION AND POVERTY Given the predominance of agriculture in the rural economy of India, the scarcity of land available to the agricultural labour households and small land holders seems to be an obvious, important cause of their poverty. If cultivable land were abundant, the rural households would face no constraint regarding that key factor in agricultural production. However, the extent to which the possession and size of land holding determines the prevalent consumption levels in rural areas is an open question which is examined below and in the next section. The discussion below is based on the size of land holding possessed by the household, whereas per capita land is used as the explanatory variable in the next section. A. Land Holding and PCE Decile Tables Al and A2 in the Statistical Appendix show the decile dis- tribution of the households reporting possession of land holdings in different size classes as well as the land holding distribution of households in each PCE decile. It is evident that the rural landless are not predominantly in the bottom deciles whereas among the households with 20 acres or more of land, some five percent in rural Gujarat and eight percent in rural Maharashtra were in the bottom two deciles. As expected, however, those with large land holdings were overrepresented in the top PCE decile.-/ B. Land HoZding and HousehoZd Occupation It can be argued that the relationship between land holding and consumption or poverty should be examined for cultivating households. Among 1/ As shown in Section IV, the correlation between per capita land and the per capita consumption level of a household was no more than 0.34 in rural Gujarat and 0.20 in rural Maharashtra.  -18- various ways of defining a cultivating household, one convenient approach is that followed by the NSS, which designates the major source of income of a household during the year preceding the time of the survey as its occupation.1/ Of course, when the economic activities of a household change over a year and/or when there is more than one worker undertaking different activities, an identification of the major source of income may require some detailed questions. Since the instructions to the NSS investigators do not prescribe any detailed questions, the data recorded in regard to the household occu- pation probably involve a certain degree of approximation. In the absence of specific data on the earnings of non-employee workers, however, there is no alternative to relying on the recorded information on the household occu- pation. In order to ensure an adequately large sample size, a distinction is made between four classes of households: those deriving a major part of their income from (a) cultivation; (b) agricultural labour; (c) white collar work (including medical and public health workers, teachers, village officials and other administrative managerial, professional and technical workers, clerical workers, traders, money lenders and other sales workers); (d) other workers (including agricultural workers not engaged in cultivation, artisans and craftsmen, barbers, launderers and other service workers, etc.). 1/ To some extent, the basic stratification scheme of the NSS sample also attempts a similar distinction between (a) rural labour households, (b) those self-employed in non-agricultural activity, and (c) others. (See Annex 1.) However, no tabulation is made separately for these substrata.  -19- Table 1 shows the distribution of households according to the size of land holding and household occupation.-/ It shows that the landless households include a significantly larger proportion of households whose occupation was white collar work and "other" activities (crafts, etc.). C. Household Occupation and PCE DeciZe Table A.3 in the Statistical Appendix shows the decile distribution (in terms of PCE) of households (with the intermediate six deciles grouped into three quintiles) in each occupation group. It confirms the expected overrepresentation of the white collar households in the top decile. Table A.4 in the Statistical Appendix gives further information on the households of each occupation that fall in each decile. The inverse relation between MPCE decile and the household size holds for each occupation group, but the inter-decile range of variation in household size is the lowest for cultivating households. On the other hand, the inter-decile range of variation in PCE, like the total household expenditure, is the lowest for the agricultural labour households. D. Size of Land Holding and the Proportion 6f Poor_HouAZhoZds Tables 2 and 3 show the proportion of the poor households (defined as those with a per capita expenditure of Rs. 15.00 or less in 1960-61 1/ As would be expected, a comparison of these data with the occupational distribution of workers in Tables A28 and A29 in the Statistical Appendix shows a broad correspondence, with the proportion of cultivator house- holds exceeding the percentage of cultivators among workers (of both sexes together) in each size of land holding class and a contrary tendency in the proportion of agricultural labour households. Likewise, the pro- portion of households with white collar occupations exceeds that of white collar workers. It is quite likely that the income of a white collar worker or a cultivator ~in a household would exceed that of other workers, if any, engaging in agricultural labour or some other miscellaneous work. But cultivation and white coll.ar work also carry a certain prestige which may cause certain overstatement of such activities as the major source of household income. However, the biases introduced by such factors are believed to be relatively minor and unlikely to affect an examination of the incidence of poverty by size of land holding and household occupation.  TABLE 1 DISTRIBUTION OF RURAL HOUSEHOLDS BY SIZE OF LAND HOLDING POSSESSED BY THE HOUSEHOLD AND HOUSEHOLD OCCUPATION GROUP, GUJArAT AND MAHARASHTRA, 27TH ROUND OF THE NSS, 1972-73 Puraza C'4jarat -hur- -Ao.rashtrc Size rf Land Holding HOUSEHOLD OCCUPATION GROUP HOUSEHOLD OCCLTATION GROUP Pcssessed by the House- Agricultural Agriculeural Ho'd (in acres) Wh.ite Collar Cultivators Labor Others. All White Collar Cultivators Labor Others All (A) in Each 'land Holding Size No Land 16.7 3.6 44.0 35.7 100.0 17.0 0.3 56.5 26.2 100.0 0.01- 0.99 14.4 3.4 52.2 29.9 100.0 9.5 5.9 60.4 24.2 100.0 1.00- 2.49 4.4 64.5 20.5 10.6 100.0 2.0 45.5 39.0 13.5 100.0 2.50- 4.99 1.5 86.9 7.5 4.0 100.0 2.6 58.7 31.6 7.1 100.0 5.00- 7.49 2.3 92.1 2.9 2.7 100.0 3.1 68.8 22.5 5.6 100.0 7.50- 9.99 1.5 94.4 2.6 1.5 100.0 3.4 79.4 14.4 2.9 100.0 10.00-19.99 1.1 95.9 1.7 1.3 100.0 2.1 83.8 9.4 4.6 100.0 20.00 & Above 0.5 98.6 0.7 0.3 100.0 2.4 91.6 3.0 2.9 100.0 All 5.8 62.8 19.2 12.2 100.0 5.9 46.4 34.4 13.3 100.0 (B) ln Sich HIoucehold Orcxaqtion Gro zo No Land 23.4 0.5 18.8 24.1 8.2 34.7 0.1 19.7 23.7 12.0 0.01- 0.99 53.5 1.2 59.0 53.3 21.6 36.5 2.9 39.8 41.4 22.7 1.00- 2.49 7.4 9.9 10.3 8.5 9.7 4.3 12.3 14.2 12.7 12.5 2.50- 4.99 4.2 22.1 6.3 5.3 16.0 5.8 16.6 12.1 7.0 13.1 5.C0- 7.49 4.7 17.6 1.8 2.7 12.0 5.9 16.4 7.2 4.6 11.0 7.5C- 9.99 1.6 9.4 0.8 0.8 6.3 3.1 9.2 2.3 1.2 5.4 10.00-19.99 3.2 25.1 1.4 1.8 16.4 4.7 23.7 3.6 4.6 13.1 20.00 & Above 0.7 14.2 0.3 0.2 9.1 3.9 18.9 0.8 2.1 9.6 All 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0  TAIBLE 2 PERCENTAGE OF POOR HOUSEHOLDS BY SIZE OF CLASS IAND HOLDING POSSESSED, RURAL GUJARAT 1972-73 CMION Sit* of Land Holding Northerm Soutbern Dry (Acres) All Eastern Plains Plains Areas Saurashtra (A) PCE CRITERION (PEi? CAPI1A FXPEVPITURE OF Rs. 15.00 OR LESS AT 1960-61 PRICWS AND Rs. 38.90 AT 2972-73 PRICES) Nil 36.8 41.1 43.8 * 30.7 31.9 Less than 1.00 44.0 52.2 42.7 46.7 46.2 31.9 1.00 - 2.49 39.8 52.2 34.8 35.3 47.5 38.4 2.50 - 4.99 41.2 57.8 33.5 33.0 49.6 24.6 5.00 - 7.49 37.3 50.7 25.6 36.8 38.9 27.9 7.50 - 9.99 29.9 63.2 16.4 25.6 29.8 14.1 10.00 - 19.99 24.4 54.2 20.2 22.3 23.1 15.3 20.00 & More 13.5 __ 13.1 6.8 19.6 10.6 All 34.7 53.5 31.8 35.6 30.2 22.1 (H) THE CRITERIONI (TOTAL HO.SEiOLD EXPENDITURE OF RS. 87.00 OR LESS AT 1960-61 PRICES AND Rs. 225.60 OR LESS AT 1972-73 PRICES) Nil 59.1 95.4 67.2 * 60.8 44.8 Less than 1.00 62.2 70.9 62.8 66.1 71.1 43.8 1.00 - 2.49 59.7 62.9 64.3 51.9 82.2 38.1 2.50 - 4.99 55.2 75.1 48.8 40.8 55.1 38.7 5.00 - 7.49 37.9 55.2 28.] 32.7 38.8 21.9 7.50 - 9.99 28.5 57.6 18.9 15.4 44.5 8.5 10.00 - 19.99 18.5 32.5 13.7 19.1 23.7 12.1 20.00 & More 7.9 -- 8.6 -- 11.2 6.6 All 43.5 61.6 45.3 45.6 38.6 25.1 (C) DISTRIBUTION OF ALL HOUSEHOLDS BY SIZE OF LAND HOLDING Nil 8.2 1.0 14.0 0.2 15.7 11.8 Less than 1.00 21.7 21.9 14.4 37.2 11.1 20.9 1.00 - 2.49 9.7 11.5 14.3 14.1 3.2 1.9 2.50 - 4.99 16.0 25.3 22.0 13.9 6.9 7.1 5.00 - 7.49 12.0 19.4 12.0 11.0 7.3 8.5 7.50 - 9.99 6.3 7.0 6.0 5.0 7.6 6.3 10.00 - 19.99 16.4 11.9 10.4 13.8 22.8 26.7 20.00 & More 9.1 1.7 5.8 3.8 25.2 16.1 Not Reported 0.8 0.3 1.1 1.0 0.2 0.8 All 100.0 100.0 100.0 100.0 100.0 100.0 Sanple Households 5543 832 1575 923 650 1563 * lie nu.MbCL of seM;,le h11IScholds wais Loo snall to permiL estimaLes. NOTE: The poverty line in terms of 111W. is obtained by multiplying the average Lize of a household in rural Gujarat (5.8 persons) with the norm of a per capitn e.xpenditure of Rs. 15.00. hl'e regional dfffercnces in the average ?iz- of hon.chi5,( ! 11.1wso 1 f h','1 taken i o accoln't k-c.ilse the TIPHE el I trion is used only for an illustratio:t.  -22- TABLZ 3 PERCENTAGE OF POOR HOUSEHOLDS BY SIZE CLASS OF LAND HOLDING POSSESSED, RURAL MAHARASHTRA, 1972-73 REGION Nagpur Division Size of except Land Holding Poona Khandesh Maratha- Bhandara Bhandara (Acres) All Konikan Division & Nasik wada 6 Chanda & Chanda (A) PCf CRITERION (PER CAPITA EXPENDITURE OF Re. 15.00 OR LESS AT 1960-61 PRICES AND Rs. 40.60 OR LESS AT 1972-73 PRICES) Nil 60.0 46.3 61.3 69.4 57.9 58.1 36.5 Less than 1.00 63.6 66.5 60.6 73.9 59.8 66.4 52.1 1.00 - 2.49 66.6 68.5 64.7 77.4 55.5 68.0 63.0 2.50 - 4.99 59.4 55.2 59.6 69,9 38.8 63.5 69.5 5.00 - 7.49 58.8 70.6 56.2 58.2 47.9 66.4 58.9 7.50 - 9.99 57.5 78.5* 75.1 54.9 51.0 49.9 45.9 10.00 - 19.99 47,7 34.2* 57.9 47.8 38.7 50.6 27.4 20.00 & More 36.1 3.1* 44.0 52.5 29.9 30.9 35.3 All 57.3 60.9 59.5 63.6 46.5 58.3 53.8 (B) THE CRITERION (TOTAL HOUSEHOLD EXPE/DITUR OF RS. 79. 50 OR LESS AT 1960-61 PRICES OR RH. 215.20 OR LESS AT 1972-73 PRICES) Nil 80.4 72.9 76.4 80.3 77.3 93.1 73.7 Less than 1.00 78.2 82.7 73.8 79.7 69.6 84.3 77.9 1.00 - 2.49 75.9 82.3 68.8 88.8 65.2 74.0 77.0 2.50 - 4.99 64.4 50.9 64.1 72.6 60.5 80.2 56.5 5.00 - 7.49 60.6 54.2 59.9 56.2 54.6 71.4 54.8 7.50 - 9.99 51.8 61.2* 49.6 46.2 69.6 59.3 8.0 10.00 - 19.99 46.4 29.2* 43.2 38.2 43.0 60.2 37.9 20.00 & More 24.9 3. 1* 26.6 20.6 20.8 33.6 17.8 All 63.8 71.8 61.4 62.7 53.6 72.4 58.9 (C) DISTRIBUTION OP ALL flOUSEHOLDS BY SIZE OF LAND HOLDING Nil 12.0 8.7 15.0 18.8 8.9 11.1 3.3 Less than 1.00 22.7 28.6 18.3 14.3 22.7 28,8 26.2 1.00 - 2.49 12.5 29.0 14.9 9.2 4.8 6.1 20.1 2.50 - 4.99 13.2 20.0 13.6 13.3 8.2 10.4 21.0 5.00 - 7.49 11.0 7.7 12.1 11.1 11.7 12.4 5.7 7.50 - 9.99 5.4 1.2 4.4 9.4 5.7 5.8 7.6 10.00 - 19.99 13.1 ].3 12.6 15.1 19.1 15.4 10.3 20.00 & More 9.6 1.0 8.8 8.4 t8.4 10.1 5.7 Not Reported 0.5 2.4 0.2 0.3 0.6 -- -- All 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Samuple Households 5329 671 1779 584 1029 916 350 * The number of sample houselholds was: less than 1J. NOTL: 1lie poverty line iti terms of TiiE is; obtained by multiplying LIhC average size of a houselinld in rural Kl,hrashtra (5.3 persons) with the nrm of a per capiLa e::p-sdltssre of Rs. 15.00. IThe rqgional differences in the average sizC Of a ho1:sje'i,d have not becit tatVs.v into accotint because the TIlE crIter,rn 1oi oisedI only for an il iwstrrat len.  -23- prices l/) according to size classes of land holding. As was seen earlier in the NSS data of 1956-57, the percentage of households below the poverty line was lower among the landless than among landed households with less than 7.5 acres of land in rural Gujarat and less than 2.5 acres of land in rural Maharashtra. Households with land holdings from 2.5 to 10.0 acres in Maharashtra had almost the same proportion of the poor as the landless. Even among households owning more than 20 acres, some 14 percent in rural Gujarat and 35 percent in rural Maharashtra were below the poverty line. Prima facie, the estimated proportions of the poor among households with holdings of 10 acres or more are surprisingly large and contrary to general impressions. Perhaps our impressions overlook the positive asso- ciation between the size of land holding and household size, and are based on the total expenditures of the household. Tables 2 and 3 also show, therefore, the estimated proportions of poor according to a criterion based on the total household expenditure (THE) with the poverty line obtained by multiplying the norm of per capita expenditure by the average household size in rural Gujarat and Maharashtra. According to the THE criterion, the proportion of the poor among households with 10 acres or more is smaller than that according to the PCE criterion. However, the ranking of households according to PCE deciles is substantially different from that in terms of the THE decile and 1/ The consumer price index for agricultural labourers in rural areas of Gujarat and Maharashtra has been used to estimate the poverty line relevant for 1972-73 -- Rs. 38.90 for Gujarat and Rs. 40.60 for Maharashtra. The Bureau of Economics and Statistics in Gujarat has suggested, however, that the cost of living in rural areas of the state was higher than in rural India as a whole and that to ensure the desired minimum level of food intake, the poverty line should be set at a per capita consumer expenditure of Rs. 18.25 in 1960-61 prices. (See: Government of Gujarat, Incidence of Poverty in Gujarat (An Analysis Based on the Data on Consumer Ezpenditure), Ahmedabad, 1973) The price level in rural Maharashtra is also estimated to be higher than in rural India as a whole, bat the differential is less than observed in rural Gujarat. (See: G.S. Chatterjee and N. Bhattacharya, "Between States Variation in Consumer Prices and Per Capita Household Consumption in Rural India, ' Sankhya, Vol. 36, Series C, Pts. 2 & 4, pp. 337-368.) A higher poverty line would lead to a higher estimate of the proportion of the poor.  -24- the extent of overlap between the poor households in terms of PCE and THE is not likely to be high.-/ In order to understand the factors that might explain the presence of poor households even among those with land holdings of 20 acres or more, the estimates in Tables 2 and 3 distinguish between five regions in rural Gujarat and six regions in rural Maharashtra. It is evident that the pro- portion of households with holdings of more than 20 acres was significantly higher in the "dry areas" (25 percent) and in the Saurashtra region (16 per- cent) of Gujarat, or in relatively dry Marathawada (18 percent) in Maharashtra (compared with the average of 9 percent in each of the two states). The proportion of poor households among those with 20 acres or more was also significantly higher in the dry areas of Gujarat than the state average for that class size, but that was not the case in Saurashtra or Marathawada. Table 4 shows the proportion of households below the poverty line by size of land holding and household occupation. According to Table 4, which shows the proportion of households below the poverty line by size class of land holding and household occupation, the incidence of poverty was signif- icantly lower among households occupied in white collar work and "other" activities. It is evident that the lower incidence of poverty among the landless than among landed households reflects the occupational distribution of the former. Within the ranks of cultivating households in rural Maharashtra, the incidence of poverty declines steadily (though rather slowly) with an l/ For further details on this point, see: Pravin Visaria, "Poverty and Living Standards in Asia: An Overview of the Main Results and Lessons of Selected Household Surveys," Working Paper No. 2 of the Living Standards M1easurement Study (LSMS) (Development Research Center, The World Bank, forthcoming).  -25- TABLE 4 PERCENTAGE OF HOUSEHOLDS BELOW THE POVERTY LINE BY LAND POSSESSED AND HOUSEHOLD OCCUPATION, RURAL GUJARAT AND MAHARASHTRA, 1972-73 Land Possessed White Collar Agricultural (acres) All Work Cultivation Labor Other RURAL GUJARAT No Land 36.8 8.8 14.6@ 48.7 37.6 0.01 - 0.99 44.0 20.9 34.1 57.9 32.1 1.00 - 2.49 39.9 57.9 33.6 61.0 29.6 2.50 - 4.99 41.2 27.3 40.6 49.8 44.3 5.00 - 7.49 37.3 13.1 37.4 46.4 46.3 7.50 - 9.99 30.0 37.2* 30.1 34.1 8.9* 10.00 - 19.99 24.3 N* 24.5 29.9 29.5 20.00 + more 13.2 N* 13.4 N* N* All 34.7 19.9 30.2 54.8 33.5 RURAL MAHARASHTRA No Land 60.0 35.5 -$ 72.1 50,4 0.01 - 0.99 63.6 37.2 67,8 67,1 64.0 1.00 - 2.49 66.6 50.1 64.6 69.0 68.9 2.50 - 4.99 59.4 41.0 55,4 70,1 51.4 5.00 - 7.49 58.8 48.0 54.3 70,7 72,9 7.50 - 9.99 57.5 49.5* 55.1 71,5 63.5 10.00 - 19.99 47.7 37.1 46.1 65.3 46.8 20.00 + more 36.1 1.8