NEWS & ANALYSIS

Has Statistics SA under-estimated the employment recovery?

Charles Simkins says thesis that Covid will lead to a sharp decrease in employment is open to question

Has Statistics SA under-estimated the employment recovery?

6 April 2021

INTRODUCTION

South Africa’s low labour absorption rate and high unemployment rate have been with us for many years and are well known. A number of commentators, relying solely on Statistics South Africa’s Quarterly Labour Force Survey (QLFS), have argued that the COVID-19 epidemic has made things sharply worse.[1] But the QLFS is not the only source of information on the labour market. Statistics South Africa also publishes Quarterly Employment Statistics (QES) and national accounts (GDP). There are also data from the National Income Dynamics Study Coronavirus Rapid Mobile Survey (NIDS-CRAM) for February, April, June and October 2020. All the relevant information up to and including the fourth quarter of 2020 have now been published, and it is the purpose of this brief to review it, concentrating on output and employment.[2]

ANALYSIS

The analysis proceeds by considering the table below, row by row. All Rand amounts are in current prices, unless otherwise stated. All indices are set at 100 in the first quarter of 2020.

The first six rows present data from the national accounts. Row 1 presents value added at (current) basic prices and Row 2 converts Row 1 into an index. Row 2 shows that value added at current basic prices more than recovered to the first quarter (Q1) level in Q3 and increased further in Q4. Row 3 presents value added at constant basic prices and Row 4 the corresponding index, indicating that real value added more than recovered to the Q1 level in Q4. Row 5 presents aggregate compensation of employees in current prices and Row 6 the corresponding index. Comparing Row 6 with Row 2, one sees that aggregate compensation increased more than value added between Q1 and Q4.

So the question arises: given that real value added had fully recovered to the Q1 level by Q4 and that compensation of employees had risen more than value added, how could it be possible that the level of employment dropped between Q1 and Q4? If the national accounts are correct, it could only be so if nominal compensation per employee had risen by 8.1% between Q1 and Q4, and that is implausible. Row 7 presents average gross earnings[3] in current prices from the QES and Row 8 the corresponding index, indicating a rise of 3.3% between Q1 and Q4.

It is true that the QES does not cover all employees: workers in agriculture, private households and the informal sector are excluded. In Q1, the QES covered 62.2% of the workers counted by the QLFS. But it is hard to imagine that wages in sectors not covered by the QES rose much faster than in sectors that were covered, and certainly not as fast as the 17.5% needed to bring the average for all workers up to 8.1%.

Output and employment, 2020

 

Row

Q1

Q2

Q3

Q4

Gross Domestic Product

Value added at basic prices (R million)

1

1,119,208

985,196

1,135,078

1,190,229

Value added index (Q1=100)

2

100.0

88.0

101.4

106.3

Value added at constant 2010 basic prices (R million)

3

696,485

593,969

678,836

702,864

Value added index at contant basic prices

4

100.0

85.3

97.5

100.9

Compensation of employees (R million)

5

593,760

556,320

597,919

642,015

Compensation Index (Q1=100)

6

100.0

93.7

100.7

108.1

Quarterly Employment Survey

Average wages per month (Rand)

7

22395

21488

22588

23133

Wage index (Q1=100)

8

100.0

95.9

100.9

103.3

Quarterly Labour Force Survey (Indices Q1=100)

All employed

9

16,384,000

14,148,000

14,691,000

15,023,000

Indices (Q1=100)

All employed

10

100.0

86.4

89.7

91.7

Formal employment outside agriculture, households

11

100.0

89.2

91.4

93.0

Agriculture and private households

12

100.0

82.7

88.4

92.0

Informal employment outside agriculture, households

13

100.0

78.1

84.1

86.3

Quarterly Employment Survey

Formal employment outside agriculture, households

14

10,195,106

9,506,902

9,563,480

9,640,464

Formal employment index (Q1=100)

15

100.0

93.2

93.8

94.6

NIDS-CRAM

All employed

16

17,308,854

14,465,669

 

17,184,485

Index (Q1=100)

17

100.0

83.6

 

99.3

Quarterly Employment Survey

Total earnings (R billion)

18

729,388

644,719

680,570

740,897

Earnings index (Q1=100)

19

100.0

88.4

93.3

101.6

Consumer price index (Feb 2020=100)

20

100.0

98.9

101.2

101.8

Rows 9 to 13 present data from the QLFS. Row 9 presents the number of employed persons and Rows 10 to 13 present the data in index form with components of employment separately identified. Rows 14 to 15 present employment data from the QES. They are in approximate agreement when it comes to indices of employment in the formal sector excluding agriculture and private households in Q4 (QLFS: 93.0, QES: 94.6). The QLFS data suggest that employment in agriculture and private households decline about as fast as formal employment outside these sectors between Q1 and Q4, with a greater decline in informal employment[4]. On the other hand, Rows 16 and 17, which present NIDS-CRAM data[5], indicates that employment in Q4 was very nearly back to the Q1 level (Q4 Index: 99.3)

Finally, rows 18 to 20 considers earnings information from the QES. Aggregate nominal earnings were higher in Q4 than in Q1 and aggregate real earnings were a tiny bit lower (Q4 index: 99.8). Aggregate real earnings according to the national accounts were higher (Q4 index: 106.2).

CONCLUSION

Where does all this leave us?

When it comes to employment, the national accounts and NIDS-CRAM[6] are in one corner, and the QLFS and QES are in the other. Data from the former suggest that employment in Q4 had recovered to its Q1 level; data from the latter suggest that the Q4 employment was 5-7% lower in Q4. When it comes to earnings, both the national accounts and QES indicate that real earnings were no lower in Q4 than in Q1, and possibly appreciably higher.

This means that the thesis that the COVID epidemic thus far will lead to a sharp decrease in employment and earnings over the medium term, and possibly longer, is open to reasonable doubt. It follows that policy towards the labour market and income support should be approached with caution, especially in current fiscal circumstances. Of course, the epidemic is not over yet, and careful attention should be paid to new data as they emerge.

Charles Simkins, Head of Research, Helen Suzman Foundation. 

Footnotes:

[1] See, for instance, Dominic Brown and Greg Dor, Loss of 1.4m jobs proves ‘recovery’ is more myth than fact, Daily Maverick, 2 March 2021 and Karen Singh, Covid job crisis will take years to cure, Mercury, 3 March 2021

[2] Employed people are defined as:

QLFS - those who have worked for an hour in the reference week or have a definite job to return to

QES – those who received payment (in salaries, wages, commission, retainer, piece rates; or payments in kind) for any part of the reference quarter

NIDS-CRAM - those who have worked for an hour in the reference month or have a definite job to return to

[3] Gross earnings are payments for ordinary-time, standard or agreed hours during the reference period for all permanent, temporary, casual, managerial and executive employees before taxation and other deductions for the reference period. This includes salaries and wages; commission if a retainer, wage or salary was also paid; employer’s contribution to pension, provident, medical aid, sick pay and other funds; allowances; etc., but excludes earnings of sole proprietors or partners of unincorporated businesses; commission where a retainer, wage or salary was not paid; payments to subcontractors and consultants who are not part of the enterprise; and severance, termination and redundancy payments. Gross earnings are the total sum of the earnings including performance and others bonuses; overtime payments for the three months of the reference quarter

[4] Not too much weight should be placed on this last point. Estimates of informal employment are generally less stable than in other sectors.

[5] The February estimate is taken as representing the position in Q1, the mean of the April and June estimates the position in Q2 and the October estimates in Q4. No data were collected by NIDS-CRAM in Q3.

[6] In Ihsaan Bassier, Joshua Budlender and Andrew Kerr, Why the employment numbers differ so vastly in the Quarterly Labour Force Survey and NIDS-CRAM, Daily Maverick, 25 February 2021, the authors seem to have lost confidence in the NIDS-CRAM estimates once the Q4 QLFS data appeared. They may have been wrong to do so.