Uniform response rates across occupation groups
by Mahesh Vyas
CMIE classifies households into broad occupation groups based on the composition of occupations of members of households. There are five broad categ ories of occupations. These are business persons, salaried employees, daily wage workers, farmers and others.
In 2014, households of business persons in the sample accounted for about 20 per cent of all households. Sample households of salaried employees and of daily wage earners accounted for about 31 per cent and 32 per cent, respectively. Farmer households in the sample were about 11 per cent of all households. The remaining 6 per cent of the sample was an assortment of households.
In five years, by 2019, although the sample did not change much, the composition of households by their broad occupation had changed significantly. The share of households of business persons increased from 20 to 28 per cent reflecting an increase in self-employment. The share of salaried employees declined from 31 to 26 per cent reflecting the persistent fall in good quality jobs in India. The share of households of daily wage labourer declined from 32 to 22 per cent. This could mean a migration to becoming farmers or entrepreneurs. The share of farmer households has increased from 11 to 15 per cent.
In spite of these structural changes in the composition of households by broad occupation groups, the response rates show no significant change between 2014 and 2019. Response rate among households of business persons has remained stable at around 86 per cent except for a brief fall to around 82 per cent in 2016. The same is true in case of salaried employees. The response rates among daily wage labourers and farmers saw no such fall in 2016 and the response rates were broadly stable around 87-88 per cent till 2019.
In Wave 19 that was administered during January through April 2020, when the average response rate dropped to 64.4 per cent, the response rates among business households and salaried employee households was a shade higher than the average at 66.4 per cent and 65.3 per cent, respectively. Correspondingly, it was a shade lower than the average among daily wage labourer households and farmer households at 63.9 per cent and 63.4 per cent, respectively. The dispersion of response rates across broad occupation groups is evidently narrow enough to build confidence that there was no systematic bias in the survey execution during the beginning of the year of the pandemic. This holds true for Wave 20 as well.
In Wave 20 administered during May through August 2020, the response rate had taken a severe beating because of the lockdowns. It fell to 43.8 per cent. During this wave, the response rate of salaried employee households was lower at 40.9 per cent. Others were higher. Daily wage labourer households yielded a response rate of 44.2 per cent, farmers 45.5 per cent and business households 46.6 per cent. Evidently, the fall in response rates is reasonably uniform across broad occupation types.
By Wave 21, we see a small bias in favour of households of business persons. This small bias persisted even in Wave 22. The average response rate during Wave 21 was 70.6 per cent. But, the response rate among business households was higher at 76.3 per cent. The response rate among farmer-households was the lowest 66.7 per cent. The other two broad occupation groups had response rates of a little over 69 per cent, very close to the overall average. In Wave 22, the overall response rate was 73.4 per cent. Households of business persons yielded a response rate of 77.7 per cent while farmer households yielded a response rate of 70.1 per cent.
This small bias shows in the form of a small increase in the dispersion of response rates across broad occupation groups. The range between the maximum and minimum response rates that average at 3.5 from Wave 2 through Wave 19, shot up to 9.6 in Wave 21 and was 7.6 in Wave 22.
CMIE classifies households in CPHS by the density of educated persons in a household by taking into account the level of education of the members of the households. Prima facie the data suggests that since 2014, the proportion of households with no literate person has declined very sharply from over 12 per cent in 2014 to less than 2 per cent in 2021. This has helped improve the proportion of households with members with education till matriculation but it has not improved the proportion of households with graduate members.
In Wave 20, the May-August 2020 wave when execution was severely impacted by the lockdown, the response rate in all-graduate, graduate-dominated and all-less-than-matriculate households yielded disproportionately low response rates. These response rates were 37 per cent, 37 per cent and 38.2 per cent, respectively when the average response rate was 43.8 per cent. These response rates improved in the subsequent two waves. Response rates of the all-graduate and graduate-dominated households recovered in Wave 21 and maintained the improvement in Wave 22 while the response rate for all-less-than-matriculate households improved partially in Wave 21 and then completely in Wave 22.