National Council of Applied Economic Research
Opinion: G C Manna.
While the methodology for the Household Consumption Expenditure Survey is more refined now, the survey needs to address some methodological challenges.
The National Sample Survey (NSS) Office released the key results of the Household Consumption Expenditure Survey (HCES) 2022-23 in late February. These primarily include all-India estimates of the average household monthly per capita consumption expenditure (MPCE) for rural and urban areas, its distribution bybroad item groups for food and non-food categories, the variation in the average MPCE of households with different standards of living (by appropriately grouping them into 12 ‘fractile classes’ of MPCE), and the trend in the composition of MPCE since the 1999-2000 survey (55th round of the NSS). So far as the State-level estimates are concerned, the factsheet gives only estimates of average MPCE — total of food and non-food items — for each State and Union Territory (UT) for rural and urban areas.
As the earlier available results pertain to 2011-12, the latest results have bridged the data vacuum of more than a decade on an important subject that also goes into the compilation of poverty estimates. While the methodology for the HCES is more refined now, there are still challenges that need to be addressed. This will ensure that apart from producing much firmer estimates of the average MPCE, the new series has an in-built mechanism to maintain comparability of the current estimates with those of the earlier ‘quinquennial series’ available from 1972-73 to 2011-12.
Changes and implications
One change in the new HCES is the updated item coverage, which has been done keeping in view the latest consumption behaviour.
Another significant change is the splitting of the single questionnaire into three parts covering food items, consumables and services items, and durable goods. The three questionnaires have been used at random in a selected household during three separate monthly visits contrary to the past when the team would visit a household just once.
Using the single questionnaire during one visit often resulted in long interviews. As a result, respondents were fatigued and there was a possibility of under-reporting consumption expenditure, particularly in respect of items like durable goods which were placed towards the end of the questionnaire. While the latest change will help us derive more reliable estimates of the average MPCE, we are also now unable to compare the current estimates of the average MPCE, and the share of poor that may be derived from it based on the survey data, with the estimates of the past, given the likelihood of under-reporting of household consumption expenditure in the previous surveys.
A third change is in the method of stratification of villages and urban blocks for the purpose of sampling. While in HCES 2011-12, every district was considered as a basic stratum for rural and urban areas, the new HCES considers a State/UT as the basic stratum. While every district with some minimum sample allocation got represented in both the rural and urban samples of the 2011-12 survey, the new HCES does not ensure the same. Such a change does not affect the generation of State-wise estimates.
There is also a change in stratification of households. All the households of a selected village/urban block are classified into three groups depending on a criterion. The criterion in rural areas is possession of land and in urban areas it is possession of four-wheeler cars for non-commercial use on the date of the survey. The total sample of 18 households with proportional representation from the three groups have been selected. Given that the proportion of urban households possessing four-wheelers is as low as 6% in States like Andhra Pradesh, Bihar and West Bengal as per the National Family Health Survey-5 (2019-21), adequate number of rich households in the sample may not get ensured in such States as intended. It is worthwhile to note that the said stratification in the HCES 2011-12 was based on the average MPCE of the households with the top 10%, middle 60% and bottom 30% forming the three strata, and a sample of two, four, and two households, respectively, was allotted to these strata.
Methodological issues
The splitting of the questionnaire and visiting a sample household thrice now have led to non-comparability of the current estimates with those of the past, although the current estimates appear to be much firmer ones. One way to address this issue is to replicate the traditional approach of ‘one schedule with a one-time visit to a household’ in an independent random sample of households to be drawn from the same villages and urban blocks of at least one of the panels for which the fieldwork is yet to commence. In this context, it is important to note that the sample in the new HCES for a year is in the form of 10 panels, each comprising consecutive three months, with an equal number of sample villages/urban blocks allotted to each panel. This add-on module will facilitate generation of two independent estimates of the average MPCE and other associated correlates based on the current approach vis-à-vis the earlier one. This information will be useful to study the extent of divergence between the two alternative estimates of MPCE and build a comparable series.
Further, to ensure adequate representation of rich households, it will be worthwhile to develop a frame of such households based on administrative data and a random sample of households drawn from this frame for enquiry on consumption expenditure of rich households through a dedicated survey. This database in conjunction with those from the HCES will be useful to derive an improved distribution of households by their average MPCE.
G.C. Manna is Professor at the Institute for Human Development (IHD), New Delhi, and a senior adviser at the National Council of Applied Economic Research. He was earlier the Director General of the Central Statistics office and the National Sample Survey Office and a member of the National Statistical Commission. Views are personal.