On Identification of Rural Labour Households in NSS Surveys

This note deals with the definition of agricultural labour and rural labour households in the quinquennial rounds of National Sample Surveys of Consumer Expenditure and of Employment and Unemployment in India. Identification of rural labour households and agricultural labour households in the NSSO Surveys of Consumer Expenditure and of Employment and Unemployment is important because, since 1977-78, data for these households are used for the quinquennial Rural Labour Enquiry reports published by the Labour Bureau. Rural Labour Enquiry reports, the most important official source of statistics on agricultural and rural workers in India, deal with consumer expenditure, days of employment, wages and earnings, and indebtedness among rural labour and agricultural labour households. Data on consumer expenditure for these households are also used for determining the weighting diagram used in calculation of the Consumer Price Indices for Agricultural Labour (CPIAL) and for Rural Labour (CPIRL).¹

Problems in identification of rural labour households have been discussed in the literature, most recently, in Dhar (2013)\cite{nsdras2013}. NSSO Surveys of Consumer Expenditure, and of Employment and Unemployment, identify different types of households on the basis of primary source of net household income.² Dhar (2013)\cite{nsdras2013} points out that income criterion has been used to identify agricultural/rural labour households since the second Agricultural Labour Enquiry, conducted in 1956-57. Dhar (2013)\cite{nsdras2013} further points out that NSSO does not use any clearly specified statistical method to obtain either the absolute levels of income of households or the shares of income from various sources in order to assign households to specific occupational categories. Instead, the method of classification of households relies on the “perceptions of the enumerator and respondent” regarding contribution of different activities to income of the household.

Investigators are required to somehow ascertain contribution of different occupations to net income of the household, and then, on the basis of a specified method, classify the household into different categories. Instruction manuals for the surveys do not provide specific questions that the investigators are supposed to ask to get information on contribution of different sources to net income of households. It is implicit in the instructions that somehow the respondents will be able to tell the investigators either the actual levels of income from different sources, or, at least, the share of income coming from different sources.

Such an assumption is likely to result in gross inaccuracies in categorisation of households. Dhar (2013)\cite{nsdras2013} argues that classification based on “self-perception” may result in under-estimating the number of rural labour households as some of these households may report themselves as self-employed.

Most casual workers, peasants and households engaged in own-account household enterprises do not maintain any accounts, and, given low-levels of education among these sections of the population, are not even capable of maintaining rigorous, consistent accounts. Expecting respondent from such households to report net income from different sources is unrealistic. Any rigorous estimation of household income, or contribution of different activities to household income, has to be based on collection of detailed data on gross output, costs of different inputs used in production, prices, and receipt of wages, and derivation of household income from them (Rawal, 2008 \cite{vikasincome2008}). In addition to collection of detailed data from households, estimation of household income requires detailed methods of imputation for costs of input that are not procured from the market and value of output that is not sold in the market \cite{vikasincome2008}. While large amount of data required for such estimations can not be collected as part of consumer expenditure or employment-unemployment surveys, the problem is not solved by leaving it to the respondent to figure out how to answer vaguely formulated questions about household income.

Dhar (2013) \cite{nsdras2013} has discussed the long-standing problem of lack of a methodology for obtaining accurate estimates of contribution of different activities to household income. Focus of the present note is on a new problem in the methodology of classification of rural labour households. This problem has arisen on account of changes in the method that is used to convert the information on income shares into a scheme for classification of households. These changes have been brought about in the surveys conducted as part of the 68th round of NSSO, data from which have been released recently.³

In this note, I show that, in an attempt to correct some gaps in the method used for classification of rural households, NSSO has created serious conceptual problems in the scheme of categorisation. Changes in the method of classification introduced in the 68th round have made it impossible to separately identify agricultural labour and rural labour households in data from the 68th round survey. Instead of correcting the way agricultural labour households and rural labour households were identified prior to the 68th round, NSSO has created new categories of households that are not only incomparable with the categories of agricultural labour households and rural labour households in earlier surveys, but have serious conceptual problems.

Section 2 of the note discusses the method used for categorisation of rural households prior to the 68th round. Section 3 discusses the changes introduced in the 68th round and their implications for identification of rural labour households. Section 4 concludes with suggestions about improving the methods of identification of rural labour households in NSSO surveys.

It would be useful start by quoting at length from the Instructions to Field Staff for Round 66 of the NSSO Surveys.

1.9.5 Household type: The household type, based on the means of livelihood of a household, is decided on the basis of the sources of the household’s income during the 365 days preceding the date of survey. For this purpose, only the household’s income (net income and not gross income) from economic activities is to be considered; but the incomes of servants and paying guests are not to be taken into account.

In rural areas, a household will belong to any one of the following five household types:

• self-employed in non-agriculture (SENA)
• agricultural labour (AL)
• other labour (OL)
• self-employed in agriculture (SEA)
• others (OTH)

For urban areas, the household types are: self-employed (SE), regular wage/salary earning (RWS), casual labour (CL), others (OTH).

1.9.5.1 Procedure for determining household type in rural sector: For a rural household, if a single source (among the five sources of income listed above) contributes 50% or more of the household’s income from economic activities during the last 365 days, it will be assigned the type corresponding to that source.

1.9.5.2 For a household to be classified as ‘agricultural labour’ (AL) or ‘self-employed in agriculture’ (SEA) its income from that source must be 50% or more of its total income. If there is no such source yielding 50% or more of the household’s total income, it will be assigned one of the remaining types (SENA, OL or OTH) according to the following procedure.

1.9.5.3 To be classified as self-employed in non-agriculture (SENA), the household’s income from that source must be greater than its income from rural labour (all wage-paid manual labour) as well as that from all other economic activities put together (a three-way division is to be considered here).

1.9.5.4 A household not assigned one of the types SENA, AL or SEA will be classified as other labour (OL) if its income from rural labour (all wage-paid manual labour) is greater than that from self-employment as well as that from other economic activities (again a three-way division). All other households will be classified under “others” (OTH).

Source: Instructions to Field Staff, Volume-I, Design, Concepts, Definitions and Procedures, Socio-Economic Survey, NSS 66th Round, (JULY 2009 - JUNE 2010), National Sample Survey Organisation.

It is clear from above that if a household derived at least half of the income from own cultivation, it was classified as “Self-employed in agriculture”. Similarly, if the household derived at least half of its income from agricultural labour, it was classified as an agricultural labour household. If a household did not derive 50 per cent income from agricultural labour, but its share of income from wage labour (agriculture and non-agriculture) was higher than the share of income from own cultivation and from any non-agricultural business it had, then the household was classified as “other labour” household. Agricultural labour households and other labour households together constituted rural labour households.

One limitation of this method was that salaried non-manual workers were not separately identified and were classified as part of the residual category “others”, which, in addition, would have included households dependent primarily on pensions, remittances and other forms of cash transfers, as well as destitute households.

Another problem with this scheme of classification was that rural labour households for which agricultural labour did not contribute 50 per cent of household income but contributed more than the contribution of non-agricultural labour were classified as “other labour” and not as agricultural labour households. For example, if 40 per cent of income of a household came from agricultural labour, 20 per cent from non-agricultural labour and the rest from other sources, the household would have been classified as “other labour” household and not as agricultural labour household. As a result, “other labour” households could include households that derived more income from agricultural labour than from non-agricultural labour.

Let me now quote at length from the Instructions to Field Staff for the 68th round.

1.8.5 Household type:

The household type, based on the means of livelihood of a household, is decided on the basis of the sources of the household’s income during the 365 days preceding the date of survey. For this purpose, only the household’s income (net income and not gross income) from economic activities is to be considered; but the incomes of servants and paying guests are not to be taken into account.

In rural areas, a household will belong to any one of the following six household types:

• self-employed in agriculture
• self-employed in non-agriculture
• regular wage/salary earning
• casual labour in agriculture
• casual labour in non-agriculture others

For urban areas, the household types are:

• self-employed
• regular wage/salary earning
• casual labour
• others

1.8.5.1 Procedure for determining household type in rural sector: The broad household types in rural areas to be used in this round are self-employed, regular wage/salary earning, casual labour and others. A household which does not have any income from economic activities will be classified under others. Within each of the broad category of self-employed and casual labours two specific household types will be distinguished, depending on their major income from agricultural activities (sections A of NIC-2008) and non-agricultural activities (rest of the NIC-2008 sections, excluding section A). The specific household types for self-employed are: self-employed in agriculture and self-employed in non-agriculture. For casual labour, the specific household types are: casual labour in agriculture and casual labour in non-agriculture. In the determination of the household type in the rural areas, first the household’s income from economic activities will be considered. Rural household will be first categorized as ‘self-employed’, ‘regular wage/salary earning’ or ‘casual labour’ depending on the single major source of its income from economic activities during last 365 days. Further, for those households which are categorized either as self-employed or casual labours, specific household types (self-employed in agriculture or self-employed in non-agriculture and casual labour in agriculture or casual labour in non-agriculture) will be assigned depending on the single major source of income from agricultural or non-agricultural activities.

It may be noted that the household types and the procedure suggested here for rural areas are different from those followed in NSS surveys till now.

Source: Instructions to Field Staff, NSS 68th Round, (JULY 2011 - JUNE 2012), National Sample Survey Organisation.

The new scheme of classification separately identifies households that are primarily dependent on regular salaries/wages.

The first problem with the new scheme of classification is that it does not separate non-manual employees from manual wage labourers. Instead, it distinguishes casual wage labourers from regular salaried/wage employees on the basis of regularity of payment of wages/salaries. As a result, agricultural labourers and non-agricultural labourers are not separately identified. The survey instead identifies “casual” agricultural workers and “casual” non-agricultural workers. Manual workers who are hired on a long-term contract are now classified as households dependent on regular salaried/wage earnings. In particular, according to this classification, farm servants (long-term workers in agriculture) would not be classified as part of agricultural labourer households but as part of regular salaried households. Similarly, households dependent on earnings of monthly-paid manual non-agricultural labourers (for example, domestic workers, shop assistants, drivers, cleaners, gardeners) would now be classified as salaried employee households.

As a result of this, there is no clear way of identifying agricultural labour and rural labour households. The data can be used, at best, to identify casual agricultural labour and casual rural labour households.

Secondly, the new scheme of classification involves a two-stage classification. In the first stage of classification, households are categorised into self-employed, regular wage/salary earning, casual labour and other households. In the second stage of classification, self-employed households are classified into those who are primarily self-employed in agriculture and those who are primarily self-employed in non-agricultural enterprises. Similarly, in the second stage, casual labour households are further classified into casual agricultural labour households and casual non-agricultural labour households. There is no second-stage classification of regular wage/salary earning households and other households.

Given that casual workers and monthly-paid workers are separated in the first stage, this two-stage classification has another obvious problem. Suppose a household earns 35 per cent of its income from casual agricultural labour, 25 per cent of its income from long-term agricultural labour done by some members of the household, and 40 per cent of its income from own-cultivation of land. By the new classification, the household will be classified as self-employed in agriculture despite the fact that 60 per cent of its income comes from agricultural labour. Such a household would have been classified as an agricultural labour household by the classification method used before the 68th round.

In other words, the new two stage classification scheme under-counts rural labour households. Some rural labour households are separately classified along with salaried non-manual workers into a new category of regular salaried/wage earning households. On the other hand, some households that are primarily dependent on rural labour are classified now as self-employed (presumably primarily in agriculture but also in non-agriculture).

Table 1 shows proportion of rural households divided into different categories of household type for NSSO large-sample rounds starting 1993-94. The table shows that the proportion of casual agricultural labour households from the 68th round survey was 4.6 percentage points lower than the proportion of agricultural labour households in the 66th round survey conducted just two years earlier. Proportion of “other labour” in the rural population increased consistently from 1993-94 till 2009-10. It is interesting to note that the proportion of casual non-agricultural labour households from the 68th round survey was 1.3 percentage points lower than the proportion of “other labour” households in the 66th round survey. Consistent with the explanation given above, proportion of households that were self-employed in agriculture from the 68th round survey was 2.4 percentage points higher than the corresponding proportion from the 66th round survey.

This pattern of change between the 66th round and the 68th round survey is consistent with the explanation that it is likely to have been driven by the change in methodology of classification.

Notes:

• a. Refers to casual agricultural labour households only
• b. Refers to casual non-agricultural labour households only
• c. Includes 9.6 per cent of households dependent on regular salaried/wage earnings

Classification of rural households requires a consistent estimation of contribution of different activities to income of households and a conceptually sound method of using this information to classify households.

Most rural workers are either engaged in household enterprises or work as hired workers under casual contracts. Given this, rural households do not have any need to maintain, and are often incapable of maintaining, detailed account of their economic activities. In such a situation, accurate measurement of contribution of different activities to household income has to be based on collection of extremely detailed data on processes of production and other economic activities that households are engaged in. In addition, such a measurement would require consistent methods of dealing with intermediate products of household enterprises as well as commodities that are produced and self-consumed by households. Measurement of incomes would also require a consistent method of imputation of value of labour and owned assets deployed in the economic activities of households.

The NSSO has, so far, shied away from the challenge of designing a survey for collection of reliable and accurate data on household incomes.⁴ The need for NSSO to seriously engage in collection of statistics on household incomes cannot be over-emphasised.

However, since Surveys of Consumer Expenditure and Surveys of Employment and Unemployment are already extremely time consuming, additional collection of detailed data on household income as part of these surveys is not likely to be feasible. In view of this, devising methods that can be used for reliable measurement of share of income from different sources as part of these surveys, without incorporating a full-blown survey of incomes, is even more challenging. NSSO needs to address this problem rather than leave it to ingenuity of investigators to figure out a method of determining share of income coming from different sources.

A starting point for improving collection of information on contribution of different economic activities to household income would be to include a reasonably clear statement of what is meant by net incomes in the instruction manuals. The manuals should clearly specify costs that are expected to be accounted for in the measurement. For rural households, it would be particularly important to have a detailed discussion of the concepts to be used for agriculture, income from animal resources and other non-agricultural businesses. ⁵

Further, instruction manuals should have a clear discussion of the difficulty in determining the primary source of income for cultivators who combine cultivation on small operational holding of land with wage labour. It is important to sensitise investigators to this problem so that extra care is taken in determining income shares for such households.

Finally, the NSSO schedule should require the investigators to record the share of different types of activities in total household income along with household type, so that statistical testing could be done on relationship between these shares and other variables to determine reliability of classification of households.

Since the 68th round surveys, NSSO has adopted a new method for classification of rural households. This note shows that the new scheme for classification of rural households is conceptually flawed. The new scheme results in an under-counting of rural labour households as only casual labour households separately identified and long-term manual workers are classified as part of a new category that includes salaried non-manual employees along with long-term manual workers. Further, the scheme of classification is such that some of the manual worker households are likely to get classified as self-employed despite their being primarily dependent on wage labour.

NSSO must make a correction and adopt a conceptually consistent method of classification of households. It is important that the method of classification of rural households is revised before the next large-sample survey so that a consistent classification is used for the next Rural Labour Enquiry. I would like to propose that the correct scheme of classification should be on the following lines.

In the first stage, households should be divided between rural labour households, households self-employed in agriculture, households self-employed in non-agricultural enterprises, households primarily dependent on non-manual salaried employment, and other households. This classification should be based on which of the following is the most important source of income: manual wage labour, self-employment in agriculture, self-employment in non-agricultural enterprises, non-manual salaried employment, or any other.

In the second stage, households primarily dependent on manual wage labour should be sub-divided into agricultural and other labour households on the basis of whether agricultural labour or non-agricultural labour provides a higher share of total household income.

Conceptually, the definition of rural labour households in the proposed method is same as the definition of rural labour households used prior to the 68th round survey. As was discussed in Section 2, a problem with the method of classification of households used prior to the 68th round was that some households with greater income from agricultural labour than from non-agricultural labour were wrongly classified as “other labour” households rather than as agricultural labour households. The proposed method would correct that anomaly.