Dimensions of Size and Quality of Employment 2004-05 to 2009-10

Dealing with. Divergences in Population Estimates from the Population Census and the NSS Surveys

In moving from the labour force/workforce population ratios to the size and structure of the workforce we start with the well-recognised fact that, for a variety of reasons, the survey-based estimates of aggregate population are significantly below those based on interpolations of the Census count of populations over two successive Population Censuses or, estimates based on official Population projections starting from the nearest Population Census. At the all-India level, the census-based estimate(s) of the total (rural plus urban and males plus females) population for the mid-point of the survey year(s) are higher than the survey-based estimate by 122.7 million on 1st January 2005 (the mid-point of the Survey year 2004-05) and by close to 167 million on 1st January 2010 (Table 6.7).

The problems are not fully solved by multiplying the total all-India WPR by the Census-based estimate of total all-India population. Problem arises from the fact that the extent of divergence between the Census-based and the Survey-based estimates of total population varies significantly as between the rural and the

Table 6.7 All-India population by gender and rural-urban locations as on 1st January 2005 and 2010: NSS estimates and census-based estimates. All-Indian population (in ‘000s)

Notes Census based estimates are based on interpolation from the 2001 and 2011 Population Census counts by Gender and rural urban location

NSS Estimates have been computed from Unit Record Data of the NSS 61st and 66th Round Employment Unemployment Surveys

urban-areas, and also as between males and females in both segments for 2005 and at least in urban areas for 2009-10.

In view of the foregoing, a widely used procedure for estimating the size and structure of workforce at the all-India level is to derive the estimates of workforce for the four population segments separately and aggregate them as necessary for estimates by gender and rural-urban location. For each of the four population segments, the estimates of labour force/workforce are derived by multiplying the Census-based estimates of population for the mid-point of the Survey-year by the corresponding (all-ages) LFPR/WFPR.^[1] This is the procedure used by the Planning Commission. Suresh and I, in our papers on employment in India have also followed this route.^[2]

To follow up on a comment by Professor T.N. Srinivasan^[3] at my presentation of an earlier version of this paper (which had followed the procedure just outlined) at this Conference, it was decided to explore, as a first step, the extent of divergence between the Census-based and the survey-based estimates of population, separately for the four population segments, at the level of individual States and Union Territories for which survey-based estimates are available.

Not surprisingly (see Appendix Tables A.6.1 and A.6.2 for the two Survey-years), the extent of divergence between the Census-based and the Survey-based estimates of Population varies sharply across states, by gender, and rural-urban locations as well as between the two years. Thus, for the survey-year 2004-05 (NSS 61st Round), for rural areas of Kerala the Census-based estimate of male (female) population was only 95 (92) percent of the Survey-based estimate. In contrast, for rural Bihar, the Census-based estimates are higher by 25% for males and by 26% in respect of females. For the smaller states, the variation across states is much sharper. The ratio of Census-based estimates to the survey estimates varied from 3.0 (Nagaland) to 0.73 (Chandigarh).

In the case of urban areas, among the larger states, the excess of Census-based estimate over the corresponding survey estimate varies from just 7% (Himachal Pradesh) to 68% (Jharkhand) for males and from 16% (Himachal Pradesh) to 64% (Jharkhand) for females. A similar picture is seen for 2009-10.

Following the logic underlying the procedure used by us (as well as by the Planning Commission), it was decided to build-up the all-India estimates of workforce, unemployed, students and others outside the labour force by multiplying the state/segment/gender specific Census-based population estimates by the respective survey-based estimates of the proportion of workers, the unemployed, the students and non-student population outside the labour force to the (survey estimate) of population. The all-India estimates are then derived separately for the four population segments, by aggregating the estimates derived as above across States and Union Territories. The corresponding all-India estimates of ratios—such as WPR, LFPR etc.—are then derived by dividing the level estimates so-derived, by the Census-based population totals. These ratios will in effect be a weighted average of the state specific LFPR/WFPR etc., weighted by the share of the State in the respective all-India (Census-based) estimate of population. By the same token, the survey based estimates of all-India LFPR/WFPR etc. represent a weighted average of state specific ratios weighted, this time, by the ratio of the survey-based estimate of the population of the state under reference to the survey-based estimate of the all-India population.

If we compare the all-India per 1000 distribution of population by broad activity status categories on UPSS that is derived by aggregation from state-level estimates (with Census-based population estimates for the individual States) against the direct survey based estimates for all-India (which are based on the survey-based estimates of population for the individual states) we get a pleasant surprise: the two-sets of activity-status ratios are very close to one another, with differences, where present, occurring at the third decimal place. (See Table 6.8).

The reason for the closeness of the two sets of ratios turns out be, unsurprisingly, a very close correspondence of the population shares of the individual states in the

Table 6.8 Per 1000 distribution of population by activity status on UPSS obtained by aggregation from state level estimate with survey-based and census based population weights: all india, 200405 and 2009-10

Per 1000 Distribution by Activity Status

WPR Worker-population ratio

PUE Proportion unemployed

SPR Student-population rates

Other OLF Non-students outside the labour force

Source Computed from Unit Record Data for NSS 61st and 66th Round Employment Unemployment Surveys and Interpolation of the population Counts from the 2011 and 2011 Population Censuses all-India population as between the census-based and the survey-based estimates of population. (See Appendix Tables A.6.3.R and A.6.3.U for 2004-05 and A.6.4.R and A.6.4.U for 2009-10).

This result provides a partial support^[4] to the procedure used by us (earlier) and by others for deriving the all-India estimates of the size and structure of the workforce, the unemployed, the student population, and of non-students outside the labour force.

• [1] The use of LFPR/WFPR for all-ages would be appropriate provided the age-distributions from theNSS Employment-Unemployment Surveys are at least broadly in line with those from the Censusor Census-based population projections (see Sundaram 2007b).
• [2] See, Sundaram and Tendulkar (2002), and Sundaram (2007a, b).
• [3] The central burden of Professor Srinivasan’s comment was that there was no automatic presumption that the “excess” (of Census—count over the Survey-based estimate of) population hadthe same labour force characteristics as those captured by the survey.
• [4] The support is only partial because the state-specific level and ratio estimates are still based on thesurvey design and multipliers that are based on population estimates of (typically) an earlier (ratherthan the current) Population Census—in the case of rural areas—and on the latest Urban FrameSurvey (UFS) for the urban areas. These have not been altered to reflect the results of the 2011Population Census. Also, if part of the explanation for the divergence between the Census & NSSSurvey estimates of population arise from possible inadequate capture in the Survey of largehouseholds (say, with 7 or more members), or due to any other reason, we cannot adjust for thesame.