OPINION

How coherent are the 2020 production and labour market data (II)

Charles Simkins compares estimates from different sources

How coherent are the 2020 production and labour market data (II)

4 November 2020

INTRODUCTION

This brief is the eleventh in a series discussing developments in production, the labour market and macroeconomic policy since the beginning of 2020[1]. It is the companion piece promised in the tenth brief.

PRODUCTION COMPARISON

Here the comparison is between monthly production statistics and quarterly national accounts. Brief 5 describes a method for integrating and rendering coherent information from the two sources in order to provide a picture as up to date as possible, with quantitative data presented in the annexure to the brief. Brief 9 updates Brief 5 by reporting monthly production statistics for August. In November, we shall have monthly production statistics for September, and in December, national accounts for the third quarter. Table 1 in the Annexure to this brief compares indices of monthly production statistics to indexed estimates of value added reconciled to GDP estimates. It indicates that adjusting monthly production indices to second quarter value added from the GDO estimates makes little difference to the raw monthly indices based on the production figures alone. In so far as they can be compared, trends in production and value added are coherent and, as shown in Brief 9, indicate, at the level of output, an economy which had broadly returned to its February level by August. This is in contrast to the view that recovery is still to come and needs to be stimulated by additional demand side measures.

EMPLOYMENT COMPARISON

Here, the sources to be compared are the Quarterly Labour Force Survey, the Quarterly Employment Statistics and the NIDS-CRAM statistics. Brief 8 contains a comparison between the NIDS-CRAM estimates and the WLFS estimates. It shows that the NIDS-CRAM estimates of employment are higher than those of the QLFS. In part, this may be explained by the longer reference period in the NIDS-CRAM study. If one has been employed in the last week, one has necessarily been employed in the last month, but the converse is not true. Expressed in index terms, the fall in employment between the first and the second quarters was sharper in the NIDS-CRAM data but, given sampling error, the difference may not be statistically significant. The NIDS-CRAM data indicate a sharp drop between February and April, followed by partial recovery in June, though the April estimate is rendered uncertain by a relatively high incidence of missing data.

In order to bring the Quarterly Employment Statistics in the picture, one has to divide employment as estimated by the QLFS and NIDS-CRAM into formal and informal categories. Brief 10 shows how this is rendered problematic by definitional differences between the sources. Table 2 shows how vexed the comparison is. One ends up with a large jump in formal sector employment between April and June as recorded by NIDS-CRAM, which might partly be a result of a transfer from the ‘not working but have a job to return to’ category to the ‘formal, non-agricultural category’. There is also a discrepancy between the recorded falls in formal employment between Q1 and Q2 from the QLFS and QES. Both are smaller than the fall in all employment recorded by the QLFS suggesting a sharper fall in informal than in formal employment. One has to ask what the purpose of the formal/informal distinction is? Is it to identify precariousness, as the QLFS and part of NIDS-CRAM suggests? If that is Statistics South Africa’s key concern, why is the approach not followed in the QES? Is to identify the extent of the tax net as another part of the NIDS-CRAM approach and some questions in the QLFS imply? Is it merely a device to narrow the gap between QLFS and QES estimates? Harder thinking about the purpose of the formal/informal distinction is required as a necessary input to more coherent measurement.

HOURS WORKED COMPARISON

The NIDS-CRAM data for April and June enable identification of three categories of the employed: those who actually did an hour of work in the month, those that did no work but were nonetheless paid for one or another reason (mostly those who were not able to work because of the lockdown) and those that did not work and were not paid, but had a job to return to. The information is shown in Table 3, which shows the extent of reabsorption into the actually working category between April and June. Table 3 also compares hours worked by week between NIDS-CRAM and the QLFS by people who worked an hour or more in the reference period. The QLFs yields higher estimates of the length of the working week than NIDS-CRAM and also a much small dip in it between Q1 and Q2. No reasons for this difference come immediately to mind.

OCCUPATIONAL DISTRIBUTION COMPARISON

Table 4 groups occupational categories among the employed as reported by NIDS-CRAM and the QLFS. More detailed occupational categories are reported by both sources, but small sample size in NIDS-CRAM mean that standard errors in the more detailed estimates are likely to be very high. The comparison is made more difficult by the missing data as well as the unavailability of the classification in the February data. The changes in the occupational groups are therefore not comparable. The NIDS-CRAM data reflect a recovery in employment between April and June, whereas the QLFS reflect the deterioration between and Q2. However both changes reflect greater volatility in unskilled compared to skilled employment, a plausible feature.

THE TRAJECTORY OF RECOVERY

Over the last couple of decades, a taxonomy of recoveries has developed, each accompanied by a letter of the alphabet. A V-shaped recovery indicates a symmetrical pattern of decline and recovery, with a sharply defined and short-lived bottom. A U-shaped recovery is one with a vertical drop and a vertical recovery, but with a longer period at or near the bottom. A W-shaped recovery has a double dip, as might be occasioned by an epidemic with first and second waves. A J-shaped recovery is a steep drop followed by a slow rebound over a long period of time. An L-shaped recession is a steep drop followed by no recovery.  

The latest addition is a K-shaped recovery, in which the two diagonal arms of the K denote a bifurcation of the population into a group which moves up sharply and quite possibly to a better position than before, bolstered by stimulation of the economy, and a group which continues to encounter depressed, even worsening, conditions for a long time. One cause of this would be rapid growth in output and asset prices, accompanied by a lag in labour market recovery, ultimately to a level worse than before. The view that there will be such a lag forms part of the rationale for the Presidential Employment Stimulus[2], though that programme is also responding to the existence, which predates the coronavirus epidemic, of more than ten million adults who want work, but have it not.

Does the evidence discussed in this brief series support this view? On the whole, not. A V-shaped recovery in output has been accompanied by a rebound in employment. One acknowledges the limits of precision in estimates and that, to date, while we have evidence of output recovery up to August, labour market data are available only up to June. So the question is still unresolved to a degree and we need to wait for the third quarter QLFS and QES data to indicate whether or not our initial view is confirmed. Still, on theoretical grounds, why should a rapid output recovery after a temporary shock not be accompanied by an equally rapid employment recovery?

By Charles Simkins, Head of Research, HSF, 4 November 2020

TABLE 1

 

Mining

Manufacturing

Electricity

 

Raw monthly

Value added

Raw monthly

Value added

Raw monthly

Value added

 

production

index

production

index

distribution

index

 

index

 

index

 

index

 

 

(February=100)

(February=100)

(February=100)

(February=100)

(February=100)

(February=100)

January

113.2

113.2

88.9

88.9

100.0

105.4

February

100.0

100.0

100.0

100.0

112.5

100.0

March

113.8

113.8

100.9

100.9

113.5

102.8

April

70.1

71.4

49.4

50.3

56.6

83.2

May

90.5

92.2

73.5

74.9

84.3

99.9

June

92.4

94.2

88.7

90.3

101.6

108.1

July

107.0

109.1

95.6

97.3

109.5

113.2

August

126.0

128.4

98.9

100.7

113.3

110.4

TABLE 1 - CONTINUED

 

Trade - food, beverages and accommodation

Trade - wholesale, retail and motor

 

Raw monthly

Raw monthly

Value added

Raw monthly

Raw monthly

Raw monthly

Value added

 

production

production

index

production

production

production

index

 

index

index

 

index

index

index

 

 

(February=100)

(February=100)

(February=100)

(February=100)

(February=100)

(February=100)

(February=100)

 

Food and

Tourist

Food, beverages,

Wholesale

Retail

Motor

Wholesale,

 

beverages

accommodation

accommodation

 

 

 

retail, motor

January

101.4

96.9

99.4

94.2

96.7

99.4

95.7

February

100.0

100.0

100.0

100.0

100.0

100.0

100.0

March

80.3

58.0

70.4

101.3

105.3

85.0

100.6

April

4.6

1.8

3.3

55.4

52.3

16.0

49.5

May

13.7

1.5

8.3

79.2

92.1

51.7

80.0

June

41.1

4.7

24.9

93.4

94.3

86.4

92.8

July

49.8

7.8

31.2

103.5

92.4

93.1

98.6

August

58.1

16.1

39.5

101.1

101.4

94.5

100.4

TABLE 1 - CONTINUED

 

Transport (part)

 

Raw monthly

 

Value added

 

production

 

index

 

index

 

 

 

(February=100)

(February=100)

(February=100)

 

Freight

Passenger

Both

January

99.7

105.7

100.1

February

100.0

100.0

100.0

March

99.1

90.9

98.5

April

60.4

19.2

57.7

May

79.9

34.3

76.9

June

84.4

42.2

81.6

July

90.9

47.5

88.1

August

95.1

49.1

92.1

TABLE 2

Employment

 

 

NIDS-CRAM

QLFS

QES

Formal and informal sector

April

June

Q2

Q1

Q2

Q1

Q2

Formal (non-agricultural)

 

 

 

11,282,000

10,064,000

10,196,000

9,548,000

Informal (non-agricultural)

 

 

 

2,921,000

2,280,000

 

 

Formal

6,448,205

9,131,672

7,789,939

 

 

 

 

Informal

2,820,195

4,218,069

3,519,132

 

 

 

 

Agriculture

 

 

 

865,000

799,000

 

 

Private households

 

 

 

1,316,000

1,005,000

 

 

No job, no business, not active,

3,142,901

1,679,364

2,411,133

 

 

 

 

paid activity to return to

1,409,063

81,869

745,466

 

 

 

 

Missing data

 

 

 

 

 

 

 

 

 

13,820,364

15,110,974

14,465,669

16,384,000

14,148,000

 

 

All employment

 

 

 

 

 

 

 

Index (February=100)

 

 

 

100.0

86.4

 

 

Index (April=100)

100.0

109.3

 

 

 

 

 

Formal sector excluding

 

 

 

 

 

 

 

agriculture

 

 

 

 

 

 

 

Index (February=100)

 

 

 

100.0

89.2

100.0

93.6

TABLE 3

 

 

 

NIDS-CRAM

 

QES

 

 

Q1

April

June

Q2

Q1

Q2

Employed

 

 

 

 

 

 

Worked

 

10,014,610

13,148,264

11,581,437

 

 

Not worked, paid

 

1,906,954

897,896

1,402,425

 

 

Not worked, not paid

 

1,898,799

1,064,814

1,481,807

 

 

 

 

 

 

 

 

 

Total

 

13,820,363

15,110,974

14,465,669

 

 

Per cent distribution of employed

 

 

 

 

 

 

Worked

 

72.5%

87.0%

80.1%

 

 

Not worked, paid

 

13.8%

5.9%

9.7%

 

 

Not worked, not paid

 

13.7%

7.0%

10.2%

 

 

Total

 

100.0%

100.0%

100.0%

 

 

Hours worked per week

 

 

 

 

 

 

Average hours per week - job

38.6

35.5

38.8

37.2

 

 

Average hours per week -business

28.2

15.3

26.7

21.0

 

 

Average hours per week, no job,

 

 

 

 

 

 

no business, active

25.5

21.8

34.0

27.9

 

 

Average hours all

36.7

25.7

33.3

29.5

42.9

41.9

TABLE 4

 

 

 

 

NIDS-CRAM

 

QLFS

 

 

 

 

April

June

Q2

Q1

Q2

Employed by broad occupational class

 

 

 

 

 

Managers, professonals, technicians

3,741,582

3,918,908

3,830,245

3,749,000

3,573,000

Clerical, services, agriculture, trades, operators

5,448,207

5,844,083

5,646,145

7,821,000

6,577,000

Elementary, domestic workers

2,224,187

2,645,918

2,435,053

4,810,000

3,936,000

Other

9,543

20,089

14,816

1,000

62,000

Missing data

2,396,844

2,681,978

2,539,411

 

 

Total

 

 

13,820,363

15,110,976

14,465,670

16,381,000

14,148,000

Per cent change

 

 

 

 

 

 

Managers, professonals, technicians

 

 

 

 

 

Clerical, services, agriculture, trades, operators

 

4.7%

 

 

-4.7%

Elementary, domestic workers

 

7.3%

 

 

-15.9%

 

 

 

 

19.0%

 

 

-18.2%

[1] The ten preceding briefs are Charles Simkins, (1) Decision making in a time of uncertainty, 11 June, (2) The Adjustment Budget and beyond, 30 June, (3) Has the Supplementary Budget betrayed the promise of a R 500 billion stimulus package? 15 July, (4) Austerity and a permanent income shock, 15 July, (5) The implications of the second quarter Gross Domestic Product data, 11 September, (6) (with Charles Collocott) July production statistics: an indication of a V-shaped recovery? 28 September, (7) The April to June Quarterly Labour Force Survey: a cautionary note, 30 September, (8) The National Income Dynamics Study’s Coronavirus Rapid Mobile Survey: the labour market in the first and second quarters of 2020, 14 October, (9) August production estimates and April to June Quarterly Employment Statistics, and (10) How coherent are the 2020 production and labour market data? I – The basis for assessment.

[2] Announced in the recent presidential address to a joint sitting of parliament and documented in Building a society that works: Public investment in a mass employment strategy to build a new economy, available at http://www.thepresidency.gov.za/documents