9.8.1 For a comprehensive study on levels of living, one needs to examine different facets of levels of living separately before attempting to construct any overall index like the Human Development Index. Consumption, especially in the household sector, is one of the most important facets of levels of living. Household consumption data collected from nationwide samples of households in several rounds of NSS since 1950-51 have been extensively used for studies on levels of living in India and for the measurement of absolute poverty, including studies on disparities across States and regions, sectors, socio-economic classes etc. Size distributions of consumption may also be used for measurement of inequality.
Consumption Surveys carried out for Selected Classes of Population
9.8.2 Enquiries on family/household income and expenditure were sometimes conducted in pre-independence India in connection with the construction of cost of living index numbers. After independence, the first landmark was the series of Family Living Surveys (FLS) conducted by the Labour Bureau, during 1958-59 for industrial workers in mining and plantation centres. The weighting diagrams obtained from these surveys, were used for the centre-wise CPI numbers for industrial workers on base 1960 compiled by the Labour Bureau. Payroll sampling was adopted in a majority of the centres, while tenement sampling was followed in the remaining centres. Sample sizes varied across the centres, being larger for centres with larger or more heterogeneous population. Data on family income and expenditure were collected for all sample households, and imputed rental of owner-occupied or rent-free dwellings were included.
9.8.3 The next series of family living surveys for working class households was conducted in 1971 in 60 centres. Increased use was made of payroll sampling as against tenement sampling. However, only family income and expenditure data were collected. The third series of surveys, called family income and expenditure surveys, for working class population was conducted in 1981-82 in 70 centres in connection with CPI numbers on base 1982 = 100 compiled by the Labour Bureau. The scope was widened to include workers in railways, ports and docks, electricity undertaking and distributing establishments, etc. The 1982 series of CPI numbers of industrial workers is under revision on the basis of family income and expenditure survey conducted in 78 selected centres during 1999-2000.
9.8.4 As regards the middle class (or urban non-manual employees), the first comprehensive family living survey was conducted in 1958-59 by CSO in 45 selected centres. The latest series of such enquiries was conducted in 59 urban centres in 1982-83. Weighting diagrams based on these surveys were used by the CSO to compile centre-wise CPI numbers on base 1960, later 1984-85. Fieldwork for all the enquiries mentioned above was done by the NSS.
9.8.5 The first all-India Agricultural Labour Enquiry (ALE) was conducted by the Government of India in 1950-51. Data on income, consumption and other aspects were collected for each sample household through 12 monthly visits. The Second ALE was conducted by the Labour Bureau using NSS as its field agency during 11th and 12th NSS Rounds (August 1956-August 1957). The weighting diagrams were used to compile State-wise CPI numbers for agricultural labour with 1960-61 as base. Rural labour enquiries were conducted in a similar manner during NSS 18th (February 1963-January 1964) and 25th (1970-71) Rounds. Standard NSS methodology was followed in these enquiries. For the current series of CPI (AL) on base July 1986 - June 1987, the weights were derived from NSS 38th Round consumer expenditure data relating to January-December 1983.
Household Consumption Expenditure Surveys
9.8.6 The NSS is a multi-purpose socio-economic enquiry of all-India coverage, carried out in the form of successive "rounds", the duration of a round being typically one year. Data on household consumer expenditure was collected in almost all the rounds, since 1950-51, from nationwide probability samples of households, except for a decade or so beginning with 1974-75. However, starting from 1972-73, quinquennial surveys have been conducted on a large sample of households, in 1977-78, 1983, 1987-88, 1993-94 and 1999-2000. Annual surveys on a smaller sample have been conducted in the intervening years since 1986-87. One thus gets time series data on different facets of household consumption in rural and urban sectors by States and UTs.
9.8.7 The sampling design has provision for two or more independent sub-samples. The divergence of sub-sample estimates indicates the margin of uncertainty associated with the combined sample estimate. In recent rounds, attempt has been made to over-sample relatively prosperous areas and also relatively affluent households in different areas.
9.8.8 Consumer expenditure comprises all expenditure incurred by the household on domestic account including consumption out of home-grown produce (which is imputed at producer prices) and out of wages in kind, gifts, loans, free collections, etc. (which are imputed at prevailing local retail prices). For item-groups like durables and miscellaneous goods and services, there is no concept of consumption and actual expenditure incurred is recorded. Expenditure on repair and maintenance of residential house is also included. However, imputed rental of owner-occupied and other rent-free dwellings is excluded. Transfer payment in kind like loans, advance and gifts are excluded. Second hand purchases were also left out, until very recently.
9.8.9 The reference period varied to some extent in the early rounds. Both 'last week' and 'last month' were tried for food and some other item-groups in the 4th and 5th Rounds (April 1952-March 1953) of NSS. The results of these rounds and some special field experiments led to the choice of 'last month' or 'last 30 days' for all items of consumption from the 7th Round (November 1953-March 1954). However, 'last 365 days' was also used in addition to 'last 30 days' for clothing, footwear and durable goods from 32nd Round (1977-78) onwards, to eliminate the effects of seasonality and other transitory factors; and education and medical expenses were added to this list in 50th Round (1993-94). But only data relating to the reference period of 'last 30 days' was used for obtaining the main estimates and accepted as the sole basis for measuring the incidence of absolute poverty in India.
9.8.10 Both 'last week' and 'last month' were used in different half- samples of households in the annual enquiries of NSS Rounds 51 through 54 (July 1994-June 1998) for items of food, pan, tobacco and intoxicants, and the results showed that the week estimates were about 30 per cent higher than the month estimates for these items. Further, for some non-food items, 'last month' and 'last year' were used in the two half-samples of households in these NSS rounds, and for a few of these items - education, medical care (institutional) and durables - the last month estimates were much higher than the last year estimates.
9.8.11 In NSS 55th Round (1999-2000), two different reference periods - 'last 7 days' and 'last 30 days' - were employed to collect data on consumption of items of food, pan, tobacco and intoxicants, for every sample household. The use of two reference periods for collecting information from the same household was found to have narrowed down the differences between week-based and month-based estimates for food, pan, tobacco and intoxicants, and this raised serious questions about the inter-temporal comparability of both sets of estimates and the derived estimates of absolute poverty in the country.
9.8.12 Consumption of home-grown produce was imputed at local retail prices up to the 8th Round (July 1954-March 1955) but at ex-farm/ex-factory prices from the 9th Round (May-November 1955) onward.
9.8.13 The data were generally collected through a "Consumer Expenditure Enquiry" schedule in most of the rounds, but a major change was made during the 19th (July 1964 - June 1965) through the 25th Round (July 1970 - June 1971), when an "Integrated Household Survey" schedule was canvassed for collecting comprehensive data on various productive activities of the household along with data on consumer expenditure, employment and unemployment and other aspects. The blocks for recording consumer expenditure data in this schedule were also somewhat different from the corresponding blocks in the "Consumer Expenditure Enquiry" schedule of the earlier rounds. The latter schedule was brought back from 26th Round onwards. Further, during Rounds 10 (December 1955-May 1956) to 14 (July 1958-June 1959), the "Consumer Expenditure Enquiry" schedule had been expanded to an "Income and Expenditure" schedule by adding a few blocks for recording the receipts and disbursements of the household during the reference month.
9.8.14 A Pilot Survey was conducted in five States of India, including the four big cities, during January-June 2000, to compare the data collected for last week and last month reference periods with the data collected through daily visits and measurements using containers to the same household for 3 or 4 consecutive days, with a view to determining the appropriate reference period for collecting data on items of food, pan, tobacco and intoxicants. The preliminary results of this survey seem to indicate that 'last month' estimates are, on the whole, better than 'last week' estimates
9.8.15 Concepts, definitions and procedures have remained more or less uniform over the rounds. Indeed, comparability over time and across regions seems to be a great merit of NSS household consumption expenditure data.
9.8.16 Doubts have sometimes been raised about the reliability and validity of NSS consumer expenditure data, which, were collected by the interview method from respondents, many of whom were illiterate and/or did not keep any accounts. Apart from conscious distortions of data made by the respondents, recall biases could play a significant part. Much empirical work based on NSS consumer expenditure data needs to assume that the biases in the data are constant over time or across regions, socio-economic classes, etc.
9.8.17 Another consideration is the lack of strict inter-temporal comparability of NSS consumption expenditure data collected in recent decades. Indian respondents seem to have become less and less cooperative every year, either because they are busier than before, or because they have become cynical about or resistant to all enquiries into their private affairs. This seemed to be more true of urban respondents than for their rural counterparts. A questionnaire that should take 2½ to 3 hours of interviewing may have to be filled up in a very hurried manner. This may have vitiated long-range comparisons of levels of living and poverty based on NSS consumer expenditure data.
9.8.18 Abridged versions of the detailed consumer expenditure schedule have been used quite successfully in many countries under a Programme launched by World Bank in 1979. The NSSO also has acquired encouraging experience by canvassing the abridged consumer expenditure schedule as part of the Employment-Unemployment enquiry in NSS 55th Round. A similar finding was obtained in the NSS 52nd Round.
9.8.19 The statistical agencies of different State Governments have been participating in the NSS programme and canvassing the same questionnaires in matched samples of households in their respective States, following identical concepts, definitions and procedures. Results from the Central Samples and the State Sample(s) have occasionally been compared. The main purpose of the programme is to pool the two Samples and obtain dependable estimates for regions within the States. It may be noted that the sample size for all-India gradually increased from a few thousand households in the early rounds to 20 or 30 thousand households or even more (Central Sample only) in the later rounds. In the recent quinquennial surveys, the sample size was well above 100,000 households for the Central Sample alone.
9.8.20 The NSS household consumption expenditure data have been extensively used in studies on levels of living and disparities in levels of living, in cross-sectional and time series analysis of consumer behaviour, in studies on incidence of taxation, etc. In recent periods, these data have become extremely important as the basis of measurement of absolute poverty in the country.
Issues Of Concern
9.8.21 The results of the on-going Pilot Survey on Reference Periods launched by NSSO need be utilised to choose the appropriate reference period for collecting data on food, pan, tobacco and intoxicants in future NSS enquiries on consumer expenditure. Further work need to be done to choose the appropriate reference period for non-food items like durables, education, medical care, etc. Pilot studies may be made to collect panel data for understanding changes over time. Revisit to the same sample of households for data collection after the lapse of few years could also be explored.
9.8.22 Any change in the methodology of NSS household expenditure enquiries runs the risk of loss of comparability with the past, but the present situation may require changes. Keeping this in view, the NSS questionnaires for consumer expenditure should be shortened in such a way that it should not take more than 1-1½ hours of interview time, especially in urban areas.
9.8.23 In order to reduce the non-sampling errors in canvassing a lengthy schedule, the NSS should make efforts to reduce the length of the of consumer expenditure schedule by analysing the past data. Further, in doing so, the possibility of dividing the full sample of households into two matched half-samples and the use of one schedule with detailed food items but abridged non-food items in one half-sample, and of another schedule with abridged food item but detailed non-food items in the other half-sample should also be explored
9.8.24 The State sample data should be processed regularly within a reasonable time after completion of fieldwork and attempts should be made to obtain and utilise pooled estimates by combining Central and State samples. Efforts need also be made to generate separate estimates for the different NSS regions for all the rounds, especially the quinquennial rounds.
9.8.25 Techniques of small area estimation may be adopted to extract maximum amount of information from the collected data, keeping the goal of district level estimates in view.
9.8.26 For monitoring of changes in the levels of living, dissemination of consumption data through regular NSS Reports may be done as expeditiously as possible. These Reports may include estimates of per capita physical consumption of a few items besides cereals, like pulses and products, liquid milk, sugar and gur, etc.
9.8.27 Poverty measurement and poverty analysis has come to occupy a central place in policy-making in the country. NSSO should provide relevant information expeditiously to Government agencies like the Planning Commission and also to interested researchers for undertaking poverty-related studies. Further, the NSSO may take up special tabulations/exercises periodically to facilitate studies like calculations of calorie intake by households in different segments of the population, especially by those below the poverty line.
9.8.28 To facilitate the building of models explaining the poverty status of individual households, the NSS questionnaire should also record relevant characteristics of sample villages and blocks. In the past, such characteristics were not always recorded or made available.