The April to June Quarterly Labour Force Survey, A cautionary note

30 September 2020

One would have expected a substantial rise in unemployment between the first and second quarters as a result of the lockdown. But Statistics South Africa’s (“SSA”) Quarterly Labour Force Survey (“QLFS”) reports a substantial drop in the official unemployment rate: from 30.1% in the first quarter(“Q1”) to 23.3% in the second (“Q2”)[1], the lowest measured unemployment rate since the QLFS started in 2008. A discussion of this result will follow, after a comment on the reliability of the survey.

RELIABILITY OF THE SECOND QUARTER QLFS

Normally, the QLFS data are collected in face-to-face interviews. But SSA suspended face-to-face data collection for all its surveys on 19 March 2020 as a result of the COVID-19 pandemic. It changed the mode of collection for collecting Q2 QLFS data to Computer Assisted Telephone Interviewing. However, not all dwelling units on the sample had contact numbers. Moreover, among those who supplied contact numbers, some of the contact numbers were found to be invalid, some calls were not answered, and some households in the sampled dwelling units indicated that they were no longer residing at the dwelling units they had occupied during the first quarter. The result was the national response rate (responding households as a percentage of households in the sample) was markedly lower in the second quarter than in the first: 57.1% as against 87.7%.

Not only was there a drop in the response rate, the use of telephonic interviews introduced bias in the second quarter QLFS because the characteristics of the telephone and non-telephone households differ. For example, there is a significantly higher unemployment rate for the non-telephone households as compared to the telephone households. SSA corrected for the bias by imputation based on first quarter data, but imputation is never as good as collection of information. 

For both these reasons, the Q2 QLFS is less reliable than its predecessors. It is not SSA’s fault. It has done the best it could under the circumstances and, on the whole, it has reported adequately on its methods, though its analysis of non-response could have been more detailed. It warns that comparisons with previous quarters should be made with caution. It would help if SSA publishes the full second quarter data set on its Nesstar site, so that analysts can probe the issues more fully than is possible using the tables in the Statistical Release.

Nonetheless, a cautious comparison between the Q1 and Q2 QLFS results can be made. It must be made if one is to understand the most one can from the QLFS about the impact of the epidemic and the lockdown.

EMPLOYMENT

Employment statistics are the least paradoxical element in the Q2 data.

It is important to pay attention to definitions. 

Employed persons are those aged 15–64 years who, during the reference week (the week for which they reported information), did any work for at least one hour, or had a job or business but were not at work (i.e. were temporarily absent). Employed persons are divided into four categories: those in the formal non-agricultural sector, those in the informal sector non-agricultural sector, those employed in agriculture and those employed in private households.

Informal employment identifies persons who are in precarious employment situations, irrespective of whether or not the entity for which they work is in the formal or informal sector. Persons in informal employment comprise all persons in the informal sector, as well as employees in the formal sector, and persons working in private households who are not entitled to or receive basic benefits such as pension or medical aid contributions from their employer, and who do not have a written contract of employment.

Informal sector.

The informal sector has two components:

i) Employees working in establishments that employ fewer than five employees, who do not deduct income tax from their salaries/wages; and

ii) Employers, own-account workers and persons helping unpaid in their household business who are not registered for either income tax or value-added tax.

Table 1 presents estimates of employment Q1 and Q2.

Table 1

In total, 2.2 million jobs are estimated to have been lost in Q2. The drop in second quarter GDP is estimated at 13.2%[2], close to the drop of 13.6% in employment. But one must be careful here. Some of the lost output was a result of people in employment not working. The Q2 QLFS reports that there were 8.0 million persons who were expected to work and actually did some work during the national lockdown and a further 0.2 million who were expected to work, but could not. This means that 6.2 million employed people did no work during their reference week and they would have contributed to the drop in output as well as the people who lost their jobs. The drop in the number of people who worked (from 15.4 million to 8.0 million, or 48.1%) was considerably greater than the drop in employee compensation at basic prices in the economy (12.5%)[3] between Q1 and Q2. Why was this?

Part of the answer lies in the fact that some of the people who did not work in Q2 were nonetheless paid. SSA reports that there were 11.5 million employed persons who continued to receive pay during the lockdown. Assuming all people who worked were paid, this means 56% of the employed who did not work received pay. About one in five of all employed had a reduction in their pay/salary during the lockdown. Another part of the explanation may be that the people who lost their jobs in Q2 were less skilled and therefore paid less in Q1, contributing less to the drop in employee compensation between Q1 and Q2. Testing that hypothesis would require access to the full Q2 QLFS data set. Further work is required when more data become available.

Also to be noted is that production varied quite sharply within the second quarter in sectors covered by SSA’s monthly production estimates as the figure below shows[4].

IMAGE1

The dip in production was sharpest during Level 5 lockdown and recovered partially as the levels decreased in stringency. So it would be useful to know in when the reference periods were for each household and it is desirable that this information be included in the full Q2 data set when it is released.

UNEMPLOYMENT AND THE LABOUR FORCE

It is when these variables are considered that paradoxes become apparent. Again, start with definitions:

Unemployed personsaccording to the official definition are those (aged 15–64 years) who:

a) Were not employed in the reference week; and

b) Actively looked for work or tried to start a business in the four weeks preceding the survey interview; and

c) Were available for work, i.e. would have been able to start work or a business in the reference week; or

d) Had not actively looked for work in the past four weeks, but had a job or business to start at a definite date in the future and were available.

A discouraged work-seeker is a person who was not employed during the reference period, wanted to work, was available to work/start a business but did not take active steps to find work during the last four weeks, provided that the main reason given for not seeking work was any of the following: no jobs available in the area; unable to find work requiring his/her skills; lost hope of finding any kind of work.

Persons in underemployment (time-related)are employed persons who were willing and available to work additional hours, whose total number of hours actually worked during the reference period were below 35 hours per week.

Under-utilized labourcomprises three groups that are defined as follows: persons who are underemployed, persons who are unemployed, and persons who are discouraged.

Unemployed persons according to the expanded definitionare those (aged 15–64 years) who:

are unemployed, plus those who were available to work but are either discouraged work-seekers or have other reasons for not searching.

The labour forcecomprises all persons who are employed plus all persons who are officially unemployed.

Thelabour force participation rateis the proportion of the working-age population that is either employed or unemployed.

The unemployment rateis the proportion of the labour force that is unemployed.

Table 2 reports the relevant statistics for persons aged 15-64.

Table 2

What has happened here? SSA has the following to say:

In spite of the massive decline in employment, the number of discouraged work-seekers, like the number of unemployed, decreased by 447 000, and the number of people who were not economically active for reasons other than discouragement increased by 5.6 million between the two quarters, resulting in a net increase of 5.2 million in the not economically active population.

The movement was proportionately more for the unemployed than for the employed, which resulted in a significant decrease of 6.8 percentage points in the unemployment rate to 23.3%.

It should be noted that during Q2 the travel restrictions as a result of the national lockdown played a role in people not actively looking for work.

But there are strange features of this approach. The ‘other non-economically active’ category should be reserved for people who are not working and do not want to work. But many of the people who lost their jobs and the unemployed in Q2 who stopped looking for work still wanted it. They were simply prevented from looking for work by travel restrictions or concluded that it was pointless trying to find new work during the lockdown. We can make an estimate of how many by comparing the people who did not work for reasons other than discouragement between Q1 and Q2 (F in Table 2). The increase between the two quarters was 2 684 thousand. By contrast, the increase in other non-economically active on the expanded definition between Q1 and Q2 was 2 919 thousand. There was a decrease in the number of discouraged workers by 447 thousand, a surprising result. This means that more than half of the total increase in people not in the labour force on the official definition left the labour market altogether between Q1 and Q2, and they must have reported that they did not want to work. This is a second surprising result, and accounts for the fact that the expanded unemployment rate rose less than one might have expected. More effort should have gone into thinking about these issues and producing appropriate explanations and tabulations.

CONCLUSIONS

The Q2 QLFS Statistical Release does contain new information about the labour market in the second quarter. The release of the full data set will allow further analysis. However, the data in Q2 are less reliable than in preceding periods and the information it will need to be checked against the National Income Dynamics Study’s CRAM data as well as the Quarterly Employment Series for the second quarter, due out on 15 October. SSA also needs to provide a more carefully worked out analysis of labour market status in Q2 than contained in its Statistical Release, and publish it as soon as it can. It will cover an exposed flank.

This brief is the second word on the labour market in Q2[5]. It is by no means the last.

By Charles Simkins, Head of Research, HSF, 30 September 2020

[1] Statistics South Africa, Quarterly Labour Force Survey, Quarter 2: 2020, Statistical Release P0211, 29 September 2020

[2] See Charles Simkins and Charles Collocott, July Production Statistics: an indication of a V-shaped recovery?, Helen Suzman Foundation Brief, 28 September 2020

[3] See Charles Simkins, The implications of the second quarter Gross Domestic Product data, Helen Suzman Foundation Brief, 11 September 2020

[4] The solid line shows the production index (February 2020=100) for sectors covered by monthly production series. The circular dot indicates value added in Q2 compared with Q1 for covered sectors and the triangular dot compares value added in all sectors. All these variables should be referred to the left hand vertical axis. The dotted line indicates the lockdown level and it should be referred to the right hand axis.

[5] The first was in the brief of 11 September, citied above.