What the numbers don’t tell us: Looking beyond standard measures of employment during an economic lockdown


Rosa Abraham

“In the context of unique circumstances like the pandemic-induced economic lockdown, standard employment metrics do not convey the full extent of employment distress. Supplementing reported activity status with other metrics such as earnings as well as hours worked will provide a better understanding of the extent and nature of work participation in the economy.”

The COVID-19 pandemic has raised a unique set of challenges for the assessment of the impact on labour markets and measurement of employment outcomes. In a practical and immediate sense, the pandemic has severely constrained the activities of ongoing as well as scheduled surveys. At a more conceptual level, it has raised several pertinent questions on how well existing measures can capture the full impact of such a shock on the labour market. Based on our experience in conducting large in-person as well as phone surveys, and our work with secondary data, we suggest revisions in existing measures to better capture the impact of the shock on the labour market. We point towards the need for cautious interpretations of standard measures in these times of disruption as well as the use of other statistics to get a better understanding of the true state of employment.

P.C. Mohanan and Aloke Kar, in their piece in The India Forum, highlight some of these issues describing the  inadequacy of conventional metrics of employment in the context of an economic lockdown. These metrics, they argue, allow for instances where workers might temporarily be out of work due to sickness/strike/other reasons.  However, they point out that under the context of an economic lockdown when production units have shut down fully or partially, there is no reasonable surety of the workers being able to resume their jobs. Therefore, according to them, counting these off-duty employees among the employed can result in an over-estimation of employment numbers.

In this piece, we  extend on this discussion, arguing that besides the pitfalls of such an ‘expectation-based’ (Mohanan & Kar 2020) concept of employment, the ‘identity-based’ nature of employment statuses can further mis-estimate employment numbers. Employment statuses, as they are asked in most surveys, enquire about an individual’s perception of their economic activity status. This perception is often conflated with their own identity as a worker, and may not always coincide with their actual activity.  Although many of these measurement issues existed prior to the pandemic, these have been further brought to the fore in light of the unique circumstances of the pandemic.

Relying on ‘identity-based’ measures of employment

Most labour market surveys determine an individual’s employment status in terms of their self-reported work ‘identity’,  or their ‘activity status’. Such identity-based measures tend to group individuals into statuses based on what the respondents identify as their primary activity, irrespective of their actual level of engagement with such activity.  Relying on self-reported activity status is problematic even under normal circumstances. Such a measurement approach attributes a status to an individual based on their perceptions of their activity status which is often conflated with their ‘identity’ rather than the actual activity therein.

For instance, farmers are likely to report themselves as being engaged in farming even though they may not have actually worked in their farm or earned an income from it during the period in question. Similarly, women not engaged in typical wage or salaried work, will identify as engaged in unpaid domestic work although they may have undertaken some economic activity or worked with the family enterprise during that time. For instance Deshmukh et al (2019)  find that survey questions that ask about individuals’ actual engagement with economic activity (rather than their self-reported status ) identifies more women as ‘working’.[1]

The problem with relying on self-reported status is apparent even in ‘normal’ times. This discrepancy is further amplified in the context of the massive economic shock to the labour market as a result of the lockdown.  It is also likely that the deviation between the self-reported activity and actual activity is not randomly distributed, rather is systematically biased or skewed for certain demographics, and in such cases, the impact on such groups may be particularly mis-estimated. This is the case for farmers or say, businesspersons who are likely to report themselves as continuing to be employed since they still identify themselves as being in this work, despite not earning nor having worked in their farm or business for any day during this period.

Indeed, this is evident from both the Centre for Monitoring Indian Economy (CMIE)’s Consumer Pyramid Household Survey (CPHS) data as well as from the Azim Premji University COVID Livelihoods Survey, both of which examine the labour market impacts of the lockdown. In both cases, farmers and non-agricultural self employed persons reported the least employment loss. Recent headlines using CMIE data also suggest massive job losses among the salaried workers.[2] But does this mean that self-employed workers are less severely impacted? Surely if their actual days worked or earnings made while being ‘self employed’ are taken into consideration, it is quite possible that they may have been just as severely affected. An ‘augmented’ employment status that takes into account the number of days worked and the earnings made from work used alongside a standard measure of employment would therefore be more informative.

For wage and salaried workers too, there is a case for using such augmented measures. Many wage workers are likely to have worked but not received any pay for the work done.  Or it may be the case that salaried workers may not have worked for any days at all but may have continued to receive their pay. Standard measures of employment are likely to interpret such instances as ‘temporary absences’ as Mohanan and Kar (2020) argue, and will include these workers among the employed when these very jobs may no longer exist.  In both cases, a simple reporting of employment status will fail to capture any of these outcomes. As Mahesh Vyas (2020) of the Centre for Monitoring Indian Economy points out in his piece in the Business Standard, the assumption that employment implies work and work implies employment is now broken, and in these changed circumstances, standard interpretations may not always hold.

Accounting for previous activity status

Relying on activity status during these times of extreme disruption has another shortcoming.  The status of an individual conveys their activity at a particular point of time, and in that sense, is a static measure. It does not take into consideration the previous status of that individual. It is possible that many workers having lost their jobs would report themselves not as ‘unemployed’ rather as out of the labour force. This is particularly likely amongst women, who having lost their jobs are less likely to report themselves as looking for work, rather as engaged in domestic work. For instance, in the Azim Premji University COVID Livelihoods survey, we asked respondents what their activity status was in February, as well as in April during the lockdown. In our sample, nearly 5 percent of male respondents reported that they were unemployed in the month of February. In April, this number had risen to 44 percent. However, if we were to include those men who were previously employed but reported having exited the labour market (and would, therefore not show up in standard measure of unemployment), the unemployed share increases to 52 percent. We find a similar increase from 45 percent to 52 percent in the case of women.

The CMIE-CPHS dataset reveals a similar pattern. The share of respondents reporting as being unemployed rose from 5 percent in December 2019 to 18  percent  in April 2020. However, if we account for previous status, i.e., if we include those who were previously working but now exited this labour force as among the ‘unemployed’,  this number rises to 54 percent.

Table A reports the activity status in April 2020 of those who were employed in December 2019. 61 percent of men continued in employment, while about 32 percent reported that they were unemployed. About 7 percent of men had withdrawn from the labour market entirely. For women, about 32 percent  were able to continue in employment in April 2020. Notably, only 21 percent  of women, who were previously employed, and had now lost employment, reported themselves as being unemployed. Instead, a large proportion of previously employed women now reported themselves as being ‘Unemployed, and not looking for a job’. Therefore, an important impact of the crisis is the shrinking of the labour force which in fact hides a large share of individuals, particularly women, who have lost employment and withdraw from the labour market.

Table A: Employment status in April 2020 of those in the workforce in December 2019, by gender

Men Women
Employed 61.1 31.6
Unemployed 31.9 20.6
OOLF 7.1 47.8

Source: CMIE-CPHS 

This loss of work in times of disruption is manifested not just in reported unemployment but also through other non-employment statuses. This was seen also in the case of the demonetisation shock as well. Soon after the demonetisation announcement, CMIE data showed a decline in unemployment rates, as well as a shrinking of the labour force participation rates for both men and women.[3]

Augmented measures of employment

Identity-based measures tend to organise individuals into roles/activities that represent their identity. It says less about their actual level of engagement with this activity. It is therefore important to look beyond the reported activity status and augment such measures with other variables including days of work and earnings. For instance, in their survey of 8500 workers, researchers at the Centre for Economic Performance (Bhalotia et al 2020) estimated 15.5 percent of the sample were unemployed as per the standard understanding of unemployment.[4] However, if those who did zero hours of work during this period were also included, the share of unemployed rose to 21.7 percent. Including those who did not receive any pay or financial assistance during this period further increased the effective unemployment rate to 52 percent. Along similar lines, Mahesh Vyas (2020) points out that employment statistics need to be used ‘in conjunction with other data including time use and wages’, and together these can provide a comprehensive understanding of labour market conditions.

In the context of unique circumstances like the pandemic-induced economic lockdown, standard employment metrics do not convey the full extent of employment distress. Supplementing reported activity status with other metrics such as earnings as well as hours worked will provide a better understanding of the extent and nature of work participation in the economy.

Notes

[1] Deshmukh, Neerad, Sonalde Desai, Santanu Pramanik and Dinesh Tiwari. 2019. ‘Improving Measurement of Women’s Work Participation: Sensitivity of Work Participation Rates to Questionnaire Design.’ NCAER Data Innovation Centre Measurement Brief No. 2020-1. NCAER, New Delhi.

[2] https://economictimes.indiatimes.com/news/economy/indicators/five-million-salaried-people-lost-jobs-in-july-their-ballooning-numbers-a-source-of-worry-cmie/articleshow/77610432.cms

[3] CMIE results on unemployment rate and labour force participation

[4] Bhalotia, S., Dhingra, S., Kondirolli, F. 2020. City of Dreams no More: The Impact of Covid-19 on Urban Workers in India. Paper No 008