Employment and earnings by industry before Covid-19

(1)

Andrew R. Donaldson and Aidan J. Horn

Working Paper Series Number 2 77

(2)

About the Author(s)

Recommended citation

Donaldson, A.R., Horn, A.J. (2021). Employment and earnings by industry before Covid-19.

Cape Town: Southern Africa Labour and Development Research Unit, University of Cape Town. (SALDRU Working Paper Number 277)

ISBN: 978-1-928516-39-2

© Southern Africa Labour and Development Research Unit, UCT, 2021 Working Papers can be downloaded in Adobe Acrobat format from

http://opensaldru.uct.ac.za. A limited amount of printed copies are available from the Office Administrator: SALDRU, University of Cape Town, Private Bag, Rondebosch, 7701, Tel: (021) 650 1808, Fax: (021) 650 5697, Email: [email protected]

Andrew Donaldson: Senior Research Associate, Southern Africa Labour Development Research Unit (SALDRU), University of Cape Town

Aidan Horn: Junior Researcher, Southern Africa Labour and Development Research Unit,

University of Cape Town.

(3)

Employment and earnings by industry before Covid-19

Andrew R. Donaldson and Aidan J. Horn

Saldru Working Paper 277

University of Cape Town July 2021

Abstract

Employment levels and the distribution of earnings by industry in 2010 and 2019/20 are examined in this paper, illustrating trends over this decade before the impact of Covid-19 and the accompanying economic downturn.

Drawing on both the Quarterly Labour Force Survey (QLFS) and the Quarterly Employment Statistics (QES) aggregates, we provide estimates of the distribution of earnings consistent with the System of National Accounts (SNA) income and production aggregates.

We draw attention to similarities and differences between the QLFS, QES and SNA data sources, and note differences in the implicit trends over the 2010-2020 decade. We provide distributions of gross earnings within eleven employment and industry sectors, consistent with the national accounts compensation of employees’

aggregates adjusted to include earned income attributable to employers and the self-employed in unincorporated enterprises.

We find evidence that the national accounts have under-estimated growth in earnings since 2010, and that the levels of both nominal and real GDP in recent years are understated. Nonetheless we find that QLFS estimates of earnings have to be raised by about 50 per cent in order to generate earnings levels consistent with the national production accounts. The adjustments required vary considerably by industry. We compile uprated earnings distributions by industry in two ways: aligned with industry-specific SNA aggregate earnings, and uniformly uprated to align with aggregate SNA earnings.

Both employment and earnings were severely disrupted by the 2020 Covid-19 economic shock. At the time of writing (early 2021) the economic recovery path is far from clear. This paper provides sectoral benchmark data from official sources against which the recovery might be assessed, but also indicates that there are substantial discrepancies between the available measures of earnings by industry.

(4)

Acronyms

C19-TERS Covid 19 Temporary Employer/Employee Relief Scheme

CoE Compensation of employees

GDP Gross domestic product

GOS Gross operating surplus

LMDSA Labour Market Dynamics in South Africa

NEDLAC National Economic Development and Labour Council

NIDS-CRAM National Income Dynamics Study Coronavirus Rapid Mobile Survey PALMS Post-Apartheid Labour Market Series

QES Quarterly Employment Statistics QLFS Quarterly Labour Force Survey

SA South Africa

SALDRU Southern Africa Labour and Development Research Unit SARS South African Revenue Service

SIC Standard Industrial Classification SNA System of National Accounts StatsSA Statistics South Africa

TERS Temporary Employer/Employee Relief Scheme UIF Unemployment Insurance Fund

(5)

1. Introduction

Growth in the South African economy over the past decade has been slow, contributing to a rise in the official unemployment rate from 23.7 per cent in 2009 to 28.7 per cent in 2019 (averaged over the year).¹ In the first quarter of 2020, with the economy in recession, unemployment reached 30.1 per cent.

Many households and businesses were therefore in distressed or vulnerable circumstances when lockdown was announced on 23 March 2020 in response to Covid-19. In recognition of the need to compensate for its impact on earnings and livelihoods, the lockdown decision was accompanied by employment-targeted relief measures.² Recent research has drawn on South African household and labour market survey data, and statistics of income derived from tax sources, to highlight labour market and poverty dynamics (Zizzamia, Schotte and Leibbrandt 2019), trends in top incomes (Bassier and Woolard 2020), and firm-level determinants of formal sector earnings (Bhorat, Oosthuizen, Lilenstein and Steenkamp 2017). We complement this work in seeking to reconcile labour market survey data with the national production accounts.

This paper provides a summary of the available statistics on employment and earnings by industry before Covid- 19, in part to provide benchmark data to assist in assessing the labour market impact of Covid-19 and the economic lockdown, and the contribution of relief measures to ameliorating these impacts. It outlines provisional estimates of the size distribution of earnings by industry in 2010 and 2019/20 consistent with gross earnings estimates, comprising the aggregate compensation of employees and mixed income of households reported in the national production accounts.³

Further work in this project provides baseline “pre-Covid” projections of earnings and employment by industry for 2020-2022, estimates the loss of earnings and employment in 2020 Q2, Q3 and Q4, and suggests possible recovery paths over the period ahead. This work also provides a database to simulate or model contributions and benefit costs of extended unemployment insurance, social security reform options and of alternative employment-targeted relief or wage support programmes.

Underlying this project is extensive data analysis. All output for the project and source code is shared at https://social-insurance.saldru.co.za/ters. The scripts in our repository enable researchers to reproduce or utilize our methods.

1 Unemployment (narrowly defined) as recorded in the Quarterly Labour Force Survey is South Africa’s official

unemployment rate. In 2019, unemployment broadly defined, including discouraged work-seekers, averaged 36.6 per cent.

2 The Covid-19-TERS (C19-TERS) programme was gazetted by the Minister of Employment and Labour on 26 March and took effect with the commencement of lockdown on 27 March. It was implemented through the Unemployment Insurance Fund and its design relied in part on consultations with organized business and labour through a NEDLAC task team. The C19-TERS benefit drew on principles agreed between the NEDLAC parties following the 2018 Jobs summit, which included a

commitment to use the resources of the UIF to assist companies in distress and to retain jobs. A Temporary

Employee/Employer Relief Scheme (TERS) to give effect to this Jobs Summit agreement came into effect on 11 December 2019. This original TERS scheme has remained largely in abeyance, whereas the C19-TERS benefit has been widely utilized.

3 Our approach draws on the comparative review of production and labour market developments in 2020 set out in a series of “briefs” by Charles Simkins published by the Helen Suzman Foundation (https://hsf.org.za/publications/hsf-briefs).

(6)

2. Employment by industry: QLFS estimates 2010-2020

Our analysis begins with a sectoral breakdown of employment derived from the Quarterly Labour Force Survey (QLFS), broken down by formal non-agricultural and total employment, together with the accompanying earnings distributions released by Statistics SA as the Labour Market Dynamics in South Africa (LMDSA) dataset.

Table 1: Industry divisions (based on the Standard Industrial Classification (SIC), version 5) 1 Agriculture, hunting, forestry and fishing

2 Mining and quarrying 3 Manufacturing

4 Electricity, gas and water supply

5 Construction, including residential and non-residential building and infrastructure works 6 Wholesale and retail trade, motor and household repair services, hotels and restaurants 7 Transport, storage and communication

8 Financial intermediation, insurance, real estate and business services 9 Community, social and personal services:

Government services (national and provincial departments, extra-budgetary institutions, municipalities, higher education and research organisations)

Non-government services (religious, cultural and sporting organisations, entertainment services, health and social work, hairdressing and other personal services)

Domestic service (household employment)

We use a 9-sector breakdown of employment, with “community, social and personal services” further separated into government, non-government and household (domestic service) sectors.⁴ The scope of each industry, consistent with the Standard Industrial Classification (SIC) divisions as applied by Statistics SA, is summarized in table 1.

The trends in reported QLFS employment by industry are illustrated in figure 1 – extended to show the impact of Covid-19 and lockdown in 2020 Q2-Q4. The sectoral shift of employment towards services over the past decade is clear, as is the decline in manufacturing employment and comparatively slow growth in trade and related activities and domestic household employment. It is apparent also that in construction and both government and non-government services there was employment growth until about 2015 and more sluggish performance thereafter.

The data indicate a marked increase in agricultural employment in 2015. This is not supported by sectoral output or earnings trends and is in our view an artefact of variation in the survey’s coverage and weights.⁵ In the analysis below, we adjust agricultural employment up by 20 per cent for the period prior to 2015, which yields a more plausible balance between employment and earnings growth than the unadjusted QLFS series. We also adjust

4 We separate the QLFS community, social and personal services sector into government and non-government services using the survey’s “public sector” and “private sector” fields.

5 We are advised that the increase in QLFS agricultural employment in 2015 resulted mainly from a change in the QLFS weights, which was not extended backwards in the QLFS data base. We acknowledge the assistance of Wandile Sihlobo, Johann Kirsten and Louw Pienaar in interpreting the agricultural employment data.

(7)

mining employment upwards for the full period, as explained in section 4 below. These adjustments raise total employment in 2010 by 2.2 per cent, and in 2019/20 by 0.3 per cent.

Taking into account these adjustments to agriculture and mining numbers, QLFS employment estimates for the four quarters of 2010 and 2019Q2-2020Q1 are summarized in Table 2, together with annualised employment growth estimates for the intervening period. Although not shown in this table, it should be noted that both aggregate and sectoral employment as measured by the QLFS were broadly stable over the 2018 to 2020 Q1 period. Total QLFS employment in 2019/20 was 99.9 per cent of the 2018 quarterly average. Within sectors, slight declines in manufacturing, utilities and construction offset moderate increases in agriculture, mining and services.

Real GDP in 2020 Q1 (seasonally adjusted) was within 0.5 per cent of its 2018 average level.

In the analysis which follows, we use QLFS earnings distribution data for 2018 (in 2020 prices) together with employment levels by industry in 2019/20, as broadly indicative of the position in the first quarter of 2020.

Although not ideal, this seems reasonable in the light of the comparatively flat trends in the overall economy and sectoral output and incomes over this period.

(8)

Figure 1: QLFS Employment by industry, 2010-2020

(9)

Table 2: QLFS Employment by industry: 2010 and 2019/20 (2019Q2-2020Q1), and average growth per annum

Employment (000s)¹

2010 2019/20 Employment

growth per annum 2010-2019/20⁴ Formal

and Informal

Total

Formal non- agricultur

al

Formal and Informal

Industry sector Employees

Employers &

Self-

employed Total

Formal non-

agricultural Total

1. Agriculture² 812 812 75 887 1.0%

2. Mining and quarrying³ 502 458 463 3 466 -0.9% -0.8%

3. Manufacturing 1,860 1,525 1,522 246 1,768 -0.7% -0.5%

4. Electricity, gas and water 95 126 129 1 130 3.3% 3.4%

5. Construction 1,117 882 985 375 1,360 1.1% 2.2%

6. Wholesale and retail trade 3,111 2,174 2,303 1,098 3,401 0.7% 1.0%

7. Transport, storage and communication 823 688 822 181 1,003 1.4% 2.2%

8. Finance and business services 1,770 2,281 2,195 358 2,553 3.8% 4.0%

9. Community, social and personal services

Government services 1,838 2,230 2,290 2 2,292 2.4% 2.4%

Non-government services 1,064 1,039 1,097 362 1,459 2.8% 3.5%

Domestic (household) service 1,253 1,298 14 1,312 0.5%

Total 14,246 11,403 13,916 2,715 16,631 1.6% 1.7%

1 Quarterly average estimates, derived from the QLFS datasets. These include employed persons older than 65, and so are about 1.3% higher than the Statistics SA published estimates of employment within the population aged 15-64.

2 Reported QLFS employment in agriculture is adjusted up by 20 per cent, taking into account the weighting adjustment of agricultural employment effected from 2015 onwards.

3 Formal employment in mining and quarrying is taken from the Quarterly Employment Statistics averages for both 2010 and 2019/20.

4 Employment growth is calculated as the 2010-2019/20 annual compound growth rate (for 9.25 years).

3. The distribution of earnings within industries: QLFS 2010 and 2018 estimates

The Quarterly Labour Force Survey (QLFS) collects gross earnings data, reported separately from the regular QLFS datasets in an annual Labour Market Dynamics in South Africa (LMDSA) publication.⁶ Some earnings data is collected in bands, and imputed values are added for missing data. There are acknowledged shortcomings of the QLFS earnings data and apparent inconsistencies in coverage over time, and so the results reported here should be interpreted with caution. Our analysis relies on version 3.3 of the Post-Apartheid Labour Market Series (PALMS) maintained by Datafirst, University of Cape Town.⁷

Figure 2 shows the trend in monthly real earnings by industry as “box-plot” estimates of the distributions, representing the 25^th, 50^th (median) and 75^th percentiles in each year. Reported earnings levels are inflated by

6 The QLFS earnings distributions are summarized in the Labour Market Dynamics in South Africa (LMDSA) reports, together with descriptive statistics and analysis of trends in employment and unemployment. The latest available LMDSA dataset is for 2018.

7 Andrew Kerr and Martin Wittenberg, A guide to version 3.3 of the Post-Apartheid Labour Market Series (PALMS). Datafirst, University of Cape Town. October 7, 2019. See also Andrew Kerr, Earnings in the South African Revenue Service IRP5 data, UNU_WIDER Working Paper 2020/62. May 2020. Also: Martin Wittenberg, Analysis of employment, real wage and productivity trends in South Africa since 1994, International Labour Office Conditions of Work and Employment Series No.

45. 2014.

(10)

the consumer price index to 2020 average prices.⁸ We present these results as broadly indicative, not definitive, given the known shortcomings of the QLFS earnings data. These estimates suggest an upward trend in real earnings levels in agriculture and household sector, a decline in private services and broadly stable trends in other sectors during the decade after 2010. Declines are apparent at the median and 75^th percentile in mining, manufacturing, transport and communication and the finance sectors. A widening in the dispersal of earnings in government service employment is apparent after 2014.

Figure 2: LMDSA Real earnings distributions by industry, 2010-2018

8 Earnings data in this paper are all reported in 2020 prices, indexed to the headline consumer price index, which increased by 63.9 per cent between 2010 and 2020, by 7.5 per cent between 2018 and 2020, and by 2.2 per cent between 2019/20 and 2020.

(11)

In tables 3-6 we summarise the LMDSA distributions of real earnings within each industry sector for 2010 and 2018, after adjustments to address identified data quality and completeness concerns.⁹ Tables 3 and 4 set out the results for all employed persons, and tables 5 and 6 show the distributions for formal employment only.

The 2018 earnings distributions (in 2020 prices) by industry provide approximations of the annualised earnings distribution between and within industries in 2019/20 prior to the Covid-19 economic downturn. The implicit assumption is that the size distribution of real earnings within each sector remained unchanged between 2018 and 2020 Q1.

Mean reported earnings estimates within industries are substantially higher than the medians (50^th percentile), as expected – this result is particularly striking for the (comparatively small) electricity, gas and water sector (“utilities”), wholesale and retail trade and for the financial and business services sector. These estimates suggest that mean earnings levels have increased in agriculture, the household sector, and the high-paying utilities sector, while decreasing in mining, construction and transport. In all sectors, the QLFS data (which includes part-time workers) show substantial numbers in 2018 at earnings levels below the official minimum wage. At the 25^th percentile for the whole distribution, reported earnings is R2,364 a month (in 2020 prices), and the 25^th percentile is below R3,500 in all sectors except mining and utilities.

Measured by the ratio of earnings at the 90^th to the 10^th percentile, earnings dispersion widened somewhat in all sectors except agriculture over this period. Earnings disparity widens mainly because of declines in 10^th percentile earnings – increases at the 90^th percentile occur only in utilities, trade, finance and household employment. The tables show declines in top-end (99^th percentile) earnings estimates in enterprise-based industries, alongside increases in the Eskom-dominated utilities sector and in government services. However, the sample sizes above this level are small: these estimates carry wide margins of error.

9 Modified versions of the 2010 and 2018 LMDSA datasets are used for this paper, with adjustments for outliers and imputation of missing data, amongst other adjustments, consistent with the approach adopted by Kerr, Lam and Wittenberg (2019) for the 2017 PALMS version of the LMDSA data. See also Kerr and Wittenberg (2017). Combined data for the year as a whole, after adjustment for inflation, are used, rather than for a single quarter, to take advantage of the larger sample.

(12)

Table 3: Mean earnings and selected earnings quantiles by industry (2010) – QLFS total employment

Sector

Mean Monthly earnings (2020 rands)

Earnings percentiles - 2010 (2020 rands per month)

Ratio of 90^th - 10^th percentile 10th 25th 50th 75th 90th 99th

1. Agriculture 4,187 1,140 1,629 2,107 2,856 8,102 42,578 7

2. Mining and quarrying 13,012 3,132 5,255 8,211 15,966 26,277 72,922 8

3. Manufacturing 9,935 1,764 2,856 5,267 10,589 22,179 81,457 13

4. Electricity, gas and water supply 14,767 244 4,887 9,854 18,735 29,668 93,611 12

5. Construction 9,232 1,319 2,463 4,213 8,102 18,312 95,800 14

6. Wholesale and retail trade 8,396 1,303 2,431 4,073 8,102 17,203 69,681 13 7. Transport and communication 12,733 1,764 3,241 6,320 13,960 24,723 82,410 14 8. Finance and business services 13,556 2,281 3,421 6,569 16,205 28,466 108,609 12 9. Community and personal services

Government services 14,449 2,273 5,267 11,496 19,104 26,066 65,928 12 Non-government services 12,060 1,319 2,490 5,702 14,781 25,712 110,194 20 Domestic (household) service 2,123 702 1,123 1,648 2,472 3,529 8,751 5

Total¹ 10,002 1,319 2,431 4,861 11,496 22,112 81,025 17

After taking into account upward adjustments in employment in agriculture and mining.

Table 4: Mean earnings and selected earnings quantiles by industry (2018) – QLFS total employment

Sector

Earnings percentiles – 2018 (2020 rands per month)

1. Agriculture 4,561 1,064 2,183 3,126 3,831 6,599 37,376 6

3. Manufacturing 10,330 1,419 2,801 5,126 10,585 21,543 64,628 15

4. Electricity, gas and water supply 21,847 2,154 4,524 10,643 26,485 38,139 425,712 18

5. Construction 8,482 1,064 2,364 4,433 8,314 17,268 63,857 16

Government services 13,713 798 2,585 8,617 18,553 28,006 70,937 35 Non-government services 12,216 1,059 2,154 4,789 11,529 24,774 70,934 23 Domestic (household) service 2,831 742 1,293 2,119 3,178 4,257 14,237 6

Total¹ 10,200 1,077 2,364 4,523 10,643 22,404 66,783 21

After taking into account upward adjustments in employment in mining.

(13)

Table 5: Mean earnings and selected earnings quantiles by industry (2010) – QLFS formal employment only Mean

Monthly Earnings (2020 rands)

Ratio of 90^th -

10^th percentil

e Sector

10th 25th 50th 75th 90th 99th

1. Agriculture 4,362 1,296 1,648 2,118 2,944 8,211 48,615 6

3. Manufacturing 10,351 1,971 3,202 5,692 11,404 22,992 81,457 12

5. Construction 9,756 1,604 2,770 4,562 8,900 19,550 90,652 12

Government services 14,699 2,444 5,702 11,893 19,446 26,277 68,061 11 Non-government services 13,846 1,648 3,258 7,417 16,423 27,548 117,299 17

Total¹ 11,812 1,945 3,141 6,241 14,480 24,634 82,410 13

After taking into account upward adjustments in employment in agriculture and mining.

Table 6: Mean earnings and selected earnings quantiles by industry (2018) – QLFS formal employment only

Sector

Mean Monthly Earnings (2020 rands)

Earnings percentiles – 2018 (2020 rands per month)

1. Agriculture 4,659 1,293 2,352 3,178 3,878 6,568 37,700 5

3. Manufacturing 10,987 1,613 3,173 5,476 10,913 22,918 65,027 14

5. Construction 9,478 862 2,728 4,820 9,043 20,399 65,480 24

Government services 14,059 826 2,693 9,005 18,989 28,375 71,091 34 Non-government services 14,379 1,173 2,649 5,457 14,321 27,990 74,146 24

Total¹ 11,970 1,377 3,033 5,601 13,096 26,109 70,937 19

After taking into account upward adjustments in employment in mining.

Particularly striking is the widening of earnings inequality in government service employment. Earnings at the 10^th percentile was lower in government service employment than in all other sectors other than household employment in 2018, and fell by more than 60% over the 2010-2018 period. This is in part be accounted for by the introduction of the Community Work Programme, which by 2018 employed about 240,000 people for two days a week at wages below R100 a day. It is possible that temporary, part-time work and special employment programmes account for declines in earnings at the 10^th percentile in other sectors too. An increase in informal sector employment might also underlie this finding.

(14)

A comparison of tables 3 and 4 with tables 5 and 6 indicates that reported mean earnings is somewhat higher in formal than in total employment in all sectors. The ratio of earnings at the 90^th to the 10^th percentile is lower in formal than in total employment in most industries, though significantly higher in 2018 in construction and transport – industries in which informal employment is high. Mean monthly earnings reported in the 2018 QLFS is about 18 per cent higher for the formally employed than for total employment. This difference is largely a consequence of the exclusion of domestic workers from the formal employment category.

These results suggest that employment growth over the 2010 to 2018-2020Q1 period was accompanied by broadly stable average real earnings across the economy together with some widening in sectoral earnings inequality. But these cannot be regarded as robust findings, bearing in mind the data quality limitations in the underlying QLFS datasets. We proceed by comparing the implied aggregate QLFS earnings data with two alternative sources, and we then present upwardly adjusted QLFS earnings distributions aligned with national accounts estimates.

4. Employment and earnings: QES estimates 2010-2020

For the formal economy, the Quarterly Employment Statistics (QES) datasets provide an alternative source of evidence on employment and aggregate earnings.

The QES derives from an enterprise-based survey, which excludes both agriculture and domestic household employment. The QES collects data from a narrower set of formal enterprises than is covered in the QLFS formal sector definition, and reports numbers employed over the quarter whereas the QLFS questionnaire relates to employment in the previous two weeks. The effective sample coverage of formal employment is larger in the QES than the QLFS and its reported trends are likely to be more reliable at industry level.¹⁰

Table 7 summarises employment by industry reported in the QES, for 2010 and 2019/20 (2019 Q2-2010 Q1). The QES indicates growth in formal employment of 2.2 per cent a year between 2010 and 2019/20, with increases especially in the trade, finance and non-government services sectors.

The QES reports lower employment in 2019/20 than QLFS formal employment in manufacturing, electricity, gas and water, construction, transport, storage and communication and non-government services, but somewhat higher numbers in the mining, trade and finance sectors. The QES reports 86 per cent of the QLFS formal employment total in 2010, 90 per cent in 2019/20, but less than 75 per cent in the utilities, construction, transport and non-government services sector. Other than utilities, these are sectors in which small unincorporated enterprises are prominent.

The mining and quarrying sector is a special case. In 2010, the QES reports over 50 per cent more employment in this sector than the QLFS. Poor QLFS survey coverage of mining employees in institutional accommodation is one possible reason for this. This is also a sector in which QES employment over a three-month period should be expected to exceed employment as measured in the QLFS, due to the persistence of long-distance alternating work tours and home rest. QLFS coverage appears to have improved in more recent surveys – the 2019/20 gap in

10 The QES survey comprises approximately 20,000 enterprises and government entities, and so its effective coverage of employment is several orders of magnitude larger than the household-based QLFS.

(15)

employment is 12 per cent. In table 2 above and in the analysis which follows, we adjust QLFS mining employment upwards to align with the QES data.

Table 7: QES employment estimates by industry (excl agriculture and domestic service): 2010 to 2019/20 Employment (‘000s)¹ % growth

pa Ratio of QES : QLFS (formal) employment²

Sector 2010 2019/20 Increase/decrease 2010-

2019/20 2010 2019/20

Mining and quarrying 499 458 -41 -0.9% 1.53 1.12

Manufacturing 1,177 1,212 34 0.3% 0.72 0.79

Electricity, gas and water supply 56 61 4 0.8% 0.61 0.48

Construction 436 583 147 3.2% 0.55 0.66

Wholesale and retail trade 1,697 2,279 582 3.2% 0.83 1.05

Transport, storage and communication 381 498 118 3.0% 0.63 0.72

Financial intermediation, insurance, real

estate and business services 1,828 2,344 516 2.7% 1.13 1.03

Community, social and personal services

Government services 1,821 2,112 292 1.6% 1.01 0.95

Non-government services 433 663 231 4.7% 0.54 0.64

Total 8,328 10,212 1,884 2.2% 0.86 0.90

Quarterly data averaged over four quarters.

The ratio of QES : QLFS employment in mining is calculated here for QLFS employment before the adjustments made in table 2 above.

Table 8: QES estimates of mean monthly earnings by industry (excluding agriculture and domestic service): 2010 and 2019/20

Sector

Mean Monthly Earnings

(R)

Real Mean Monthly Earnings (R 2020 prices)

% increase in real earnings per

year 2010 2019/20 2010 2019/20 2010-2020

Mining and quarrying 12,282 26,800 20,135 27,391 3.4%

Manufacturing 11,292 20,328 18,511 20,776 1.3%

Electricity, gas and water supply 25,953 46,822 42,545 47,855 1.3%

Construction 9,633 19,367 15,791 19,793 2.5%

Wholesale and retail trade 8,485 15,365 13,910 15,703 1.3%

Transport, storage and communication 16,565 28,033 27,155 28,651 0.6%

Financial intermediation, insurance, real estate and

business services 14,647 27,494 24,011 28,100 1.7%

Community, social and personal services

Government services 15,597 29,945 25,569 30,605 2.0%

Non-government services 12,108 26,210 19,849 26,788 3.3%

All sectors 12,753 24,007 20,906 24,536 1.7%

Table 7 also shows somewhat higher QES than QLFS formal employment in 2010 or 2020 in three other sectors.

These differences might be explained by differences in the definition of “formal” employment or statistical error.

Adjustments other than in mining employment are not proposed below.

The QES provides estimates of “gross earnings” for the formal business sectors, derived from business payroll aggregates. Table 8 shows QES estimates of mean monthly earnings for 2010 and 2019/20. Mean earnings for the

(16)

industries covered by the QES have increased in real terms over the past decade by 1.7 per cent a year, overall, – and somewhat higher in mining, construction and services than in other sectors. This is strikingly in contrast to the mean earnings trends indicated by the QLFS formal employment data reported in tables 5 and 6 above. It is also readily apparent that mean monthly earnings reported in the QES are substantially higher than the QLFS estimates in all industries.

5. Earnings by industry: QLFS, QES and SNA estimates compared

For the formally employed, the QES mean monthly earnings levels in 2010 are 75 per cent higher than the QLFS average, and over 40 per cent higher in all industries. For 2019/20, the gap is wider: the monthly average (in 2020 prices) in the QES is double that of the 2018 QLFS formally employed, and over 50 per cent higher in all industries.

These are far greater differences than might be explained by the narrower scope of QES employment – the differences are large even in industries in which QES coverage is near complete. Amongst factors that might account for these differences are the following:

• Reporting of “net” rather than gross earnings in response to QLFS questions, excluding employer-paid benefits and payroll deductions,

• Exclusion of bonus and overtime pay in reported QLFS earnings,

• Exclusion or under-estimation of earnings in kind,

• Exclusion from QLFS reported earnings of entrepreneurial earnings, share-incentive distributions or participation schemes and other once-off or extraordinary payments,

• Under-representation in the QLFS of higher income households,

• Under-reporting by household respondents and missing data in the QLFS/LMDSA datasets, and limitations of imputation procedures.

“Gross earnings” as reported in the QES is somewhat narrower than overall employment earnings, in that its sampling frame excludes small or unregistered enterprises and agricultural and household employment. Its coverage of “earnings,” however, is somewhat wider than compensation of employees in that in principle it includes remuneration in kind, participation in profit-sharing schemes and surplus distributions accruing to employers or business owners. These differences vary by industry, and there are perhaps inconsistencies in accounting treatments in the underlying survey data.

While the QLFS-based LMDSA earnings data provide useful evidence on the distribution of earnings in the overall economy and within the main industry groups, substantial adjustments are needed if these estimates are to be used for macroeconomic or aggregate estimation purposes. We seek, in the analysis which follows, to provide estimates of earnings distributions consistent with both overall employment in the economy and the national production accounts. A secondary objective is to provide earnings distributions by industry consistent with the potential contributor base of the UIF and other payroll-based social insurance arrangements.

Tables 9 and 10 illustrate the considerable variation in available estimates of earnings by industry for 2010 and 2019/20, shown here in annual aggregate rather than monthly mean terms, all in 2020 prices. The tables show both QES gross earnings aggregates and “compensation of employees” as reported in the national production accounts, alongside implied QLFS/LMDSA total annual earnings by industry aggregates (after upward adjustment

(17)

of agricultural and mining employment as discussed above). These are computed for QLFS formal non-agricultural (non-household) employment, and for formal and informal employment separated between “employees” and

“employers and the self-employed”. In broad coverage design, formal non-agricultural employment in the QLFS is similar to the QES formal employment series. Earnings of “employees” in the QLFS corresponds approximately with compensation of employees in the SNA production accounts.

There are substantial differences between the QLFS, QES and national accounts estimates of aggregate earnings in 2019/20.

• The QES reports total gross earnings that is one-fifth higher than compensation of employees in the national accounts, despite its exclusion of agriculture and domestic services. This difference has widened over the past decade – in 2010, QES gross earnings was 3.2 per cent higher than the SNA aggregate. In all except the (small) utilities sector, the QES suggests faster growth in earnings over the past decade than the national production accounts.

• QES earnings is markedly higher than the SNA compensation totals in the construction, trade, finance and services sectors, and is lower in mining, manufacturing and utilities. The difference is especially wide in the finance and business services sector, which suggests high contributions of mixed income and profit sharing to earnings in these activities.

• The QLFS estimates are substantially lower than both the QES and SNA aggregates, notably in mining (even after upward adjustment in QLFS employment), manufacturing and government services. Overall, the QLFS estimates of total earnings are about one-fifth lower than the SNA compensation estimate in both 2010 and the more recent period, and the gap between SNA compensation and QLFS employees’ earnings is over 50 per cent in both periods. The divergence is particularly large for government services.

• The QLFS estimate of earnings in agriculture is substantially higher than the SNA compensation aggregate.

This is in part an accounting outcome – in the national accounts, a disproportionately large share of agricultural output is attributed to the operating surplus or mixed income of farmers.

• The national accounts do not separate domestic household service from other private or non-government community, social and personal services. Taking this into account, the SNA compensation of employees in the non-government services sector is substantially lower than both the QLFS and QES estimates. Private health services are the dominant activity in this sector: reported earnings presumably include mixed income of own-account practitioners and non-profit institutions serving households, for SNA purposes.

Sectoral differences might arise from differences in industry classification, though it is striking that the QES shows substantially higher overall earnings than the SNA estimate, even in sectors such as construction and non- government services where the enterprise survey base is narrower than the overall industry. An exact correspondence should not be expected. Amongst other factors, there are differences between the QLFS, QES and SNA remuneration definitions, there may be industry misclassifications in household responses, and the QLFS relies on earnings estimates by household members whereas the QES and national accounts draw on data from business surveys and other sources. Nonetheless, these differences are surprisingly large, and the divergence between the QES and both the QLFS and SNA estimates of aggregate earnings have widened over the past decade.

(18)

These differences indicate wide margins of error in the official employment and earnings’ series, signalling a need for caution in interpreting both the distributional evidence and trends in earnings.¹¹

Table 9: Estimated annual earnings by industry: QLFS, QES and SNA Compensation of Employees 2010 (in 2020 prices)

R billion (2020 prices)

QLFS Estimated Annual Earnings 2010

QES Gross Earnings

2010 (Formal non-

agricultural enterprises)

SNA Compensation

of Employees 2010 Formal non-

agricultural Employment

Formal and Informal Employment

Employees

Employers

& Self-

employed Total

1. Agriculture 26.4 14.4 40.8 32.0

2. Mining and quarrying 78.3 77.9 0.5 78.4 120.6 141.6

3. Manufacturing 202.6 182.7 39.0 221.7 261.5 337.2

4. Electricity, gas and water 16.6 16.4 0.4 16.8 28.8 36.2

5. Construction 93.4 68.3 55.5 123.7 82.7 72.6

6. Wholesale and retail trade 239.0 189.4 124.1 313.4 283.2 242.7

7. Transport and communication 105.2 84.7 41.0 125.8 124.0 122.4

8. Finance and business services 276.5 232.4 55.6 287.9 526.7 321.8

9. Community and personal services

Government services 317.1 318.3 0.4 318.7 558.6 569.4

Non-government services 133.9 110.0 43.0 154.0 103.0

}

^148.2

Domestic (household) service 31.0 0.9 31.9

Total 1,462.8 1,338.6 374.7 1,713.3 2,089.2 2024.1

Table 10: Estimated annual earnings by industry: QLFS 2018, QES and SNA Compensation of Employees 2019/20 (in 2020 prices)

R billion (2020 prices)

QLFS Estimated Annual Earnings 2018 earnings distributions (2020 prices)

applied to 2019/20 employment

QES Gross Earnings

2019/20 (Formal non-

agricultural enterprises)

SNA Compensation

of Employees 2019/20 Formal non-

agricultural Employment

Formal and Informal Employment Employees

Employers

& Self-

employed Total

1. Agriculture 37.2 11.3 48.5 33.4

2. Mining and quarrying 62.3 62.9 0.2 63.1 150.7 168.2

3. Manufacturing 201.1 180.4 38.8 219.2 302.1 388.7

4. Electricity, gas and water 33.5 33.8 0.3 34.1 35.0 51.6

5. Construction 100.3 87.4 51.0 138.4 138.5 88.6

6. Wholesale and retail trade 253.2 225.4 117.2 342.6 429.4 296.1

7. Transport and communication 102.5 104.9 23.4 128.4 171.4 153.4

8. Finance and business services 394.2 328.1 90.3 418.4 790.5 403.6

Government services 376.2 376.6 0.6 377.2 775.8 741.4

Non-government services 179.3 127.6 86.2 213.9 213.2

}

^172.9

Domestic (household) service 43.8 0.7 44.6

Total 1,702.6 1,608.2 420.1 2,028.3 3,006.6 2,498.2

11 Utilisation of SARS and UIF payroll and earnings data would provide useful further perspectives on both aggregate earnings and distributional trends. See Kerr (2020).

(19)

6. Estimated 2010 and 2019/20 earnings distributions, consistent with adjusted SNA earnings by industry estimates

We proceed as follows in constructing a distribution of earnings within the main industry categories consistent with agriculture- and mining-adjusted QLFS employment numbers and the national production accounts.

We first construct adjusted SNA earnings estimates by adding estimates of earned “mixed income” of the household and unincorporated enterprises sector to the SNA compensation of employees estimates, by industry.

The adjustments to the industry employee earnings estimates are based on the proportions of 2010 and 2018 total QLFS earnings in each industry attributable to “employers” and the “self-employed” rather than

“employees”.¹² We make one change to these calculated adjustments: for agriculture, we retain the 2010 adjustments ratio in 2019/20 as the QLFS reported result in the later period yields an implausibly low level of aggregate earnings. These adjustments are shown in tables 11 and 12. We note that the earnings of domestic (household) employees are recorded in the national accounts as compensation of employees in the personal service sector, and we take this QLFS earnings estimate directly into the adjusted SNA estimates.

These adjustments raise the share of earnings from work in value added at basic prices in 2019/20 to approximately 65 per cent, whereas the share of compensation of employees in GDP (excluding mixed income) amounts to about 55 per cent.

The resulting attribution of 2010 mixed income to aggregate earned income is R492.4 billion in 2020 prices, or approximately 79 per cent of the total gross operating surplus / mixed income of households and non-profit institutions serving households.¹³ For 2019/20, the attribution of mixed income to earnings is R542.6 billion, also 79 per cent of GOS/mixed income.

For 2010, the resulting adjusted SNA earned income aggregate of R2,516.5 billion (in 2020 prices) is 20 per cent higher than the QES gross remuneration total, despite a substantially higher QES earnings estimate in the finance and business services sector. The overall result seems plausible, taking into account the exclusion of agriculture, domestic service and small and informal enterprises from QES coverage. But the differences within several industries are large, notably in manufacturing, construction, transport and private services, raising questions about the reliability of these sectoral estimates.

12 Useful advice by Professor Rob Davies contributed to the adoption of this adjustment procedure.

13 In current prices, gross operating surplus/mixed income of the household sector was R377.3 billion in 2010, equivalent to approximately R620 billion in 2020 prices. In 2019, it was R664.0 billion, or R685 billion in 2020 prices.

(20)

Table 11: Adjusted SNA Earnings by Industry, based on QLFS 2010 distribution of earnings between employers and employers/self-employed

QLFS Distribution of Earnings – 2010

SNA Earned Income - 2010 (R billion 2020 prices)

Employees

Employers &

Self- employed

Compensation

of Employees Earnings in

mixed income Aggregate earnings

1. Agriculture 64.8% 35.2% 32.0 17.4 49.3

2. Mining and quarrying 99.4% 0.6% 141.6 0.9 142.5

3. Manufacturing 82.4% 17.6% 337.2 72.0 409.2

4. Electricity, gas and water 97.5% 2.5% 36.2 0.9 37.1

5. Construction 55.2% 44.8% 72.6 59.0 131.6

6. Wholesale and retail trade 60.4% 39.6% 242.7 159.0 401.7

7. Transport and communication 67.4% 32.6% 122.4 59.3 181.7

8. Finance and business services 80.7% 19.3% 321.8 77.0 398.8

Government services 99.9% 0.1% 569.4 0.7 570.1

Non-government services 72.1% 27.9%

}

^147.3 ^45.3 ^162.5

Domestic (household) service 97.3% 2.7% 0.9 31.9

Total 2,024.1 492.4 2,516.5

Table 12: Adjusted SNA Earnings by Industry, based on QLFS 2018 distribution of earnings between employers and employers/self-employed

QLFS Distribution of

Earnings – 2018 SNA Earned Income – 2019/20 (R billion 2020 prices)

Employees

Employers &

Self- employed

Compensation

of Employees Earnings in

mixed income Aggregate earnings

1. Agriculture¹ 65.0% 35.0% 33.4 18.0 51.4

2. Mining and quarrying 99.7% 0.3% 168.2 0.5 168.7

3. Manufacturing 82.3% 17.7% 388.7 83.5 472.2

4. Electricity, gas and water 99.1% 0.9% 51.6 0.5 52.1

5. Construction 63.2% 36.8% 88.6 51.6 140.2

6. Wholesale and retail trade 65.8% 34.2% 296.1 153.9 450.1

7. Transport and communication 81.8% 18.2% 153.4 34.2 187.6

8. Finance and business services 78.4% 21.6% 403.6 111.1 514.7

Government services 99.8% 0.2% 741.4 1.1 742.9

Non-government services 59.7% 40.3%

}

^172.9 ^87.2 ^216.4

Domestic (household) service 98.4% 1.6% 0.7 44.6

Total 2,498.2 542.6 3,040.8

The (rounded) 2010 breakdown between employees and employers/self-employed in Agriculture as the QLFS 2018 ratio of 76.6% : 23.4% yields a level of SNA earnings lower than the QLFS aggregate.

For 2019/20, the adjusted SNA earnings aggregate of R3,040.8 billion is just 1.1 per cent higher than the QES remuneration total. This difference is too small to accommodate the earnings of over 6 million workers in sectors or enterprises excluded from QES coverage. The main divergence is in the finance and business services sector, in which QES earnings is over 50 per cent higher than the adjusted SNA aggregate. In government services, the

(21)

QES estimate is 4.4 per cent higher than the SNA estimate. In manufacturing and utilities, in contrast, the adjusted SNA estimates are substantially higher than the QES estimates. These differences are difficult to explain, and raise questions about both sectoral classification and whether the national accounts have under-estimated earnings and income growth over the past decade.¹⁴

The adjusted average employment levels and aggregate earnings by industry for 2010 and 2019/20 (in 2020 prices) are shown in table 13, together with the upward adjustment factors required to raise QLFS earnings to the estimated SNA aggregates.

The required upward adjustments to the QLFS earnings distributions are sizeable in most sectors, and represent about a 50 per cent increase to the implied aggregate earnings in the economy derived from the QLFS in both 2010 and 2019/20. There is considerable variation by sector. The adjustments are extraordinarily high for mining and manufacturing and government services. Material differences remain between these industry earnings aggregates and the QES remuneration totals summarized in table 11 – notably in the manufacturing and the finance and business services sectors. But the gaps have narrowed in most sectors, suggesting that the adjustments summarized above move in the right direction.

We turn now to the distribution of earnings across the employment totals reported in the QLFS.

We have examined two alternative adjusted earnings distributions by industry for 2010 and 2018, aligned with the adjusted SNA earnings aggregates set out in tables 11, 12 and 13, which include mixed income shares attributed to industries. These provide distributions of gross earnings from work consistent with the national production accounts. By construction, the implied adjustments to mean earnings levels by industry in the two alternatives are the same.

14 The national accounts were last rebased and benchmarked in 2014 (for years up to 2013). This revision raised the estimated GDP for 2010 by 2.8 per cent. The next rebasing was scheduled for 2020, but has been deferred. The 2018 supply and use tables have also been deferred, and the March 2020 GDP estimates did not include the usual adjustments to the 2018 and 2019 years based on independent annual estimates. Statistics SA has indicated that rebased and benchmarked GDP estimates will be published in the second half of 2021.

(22)

Table 13: Adjusted Employment and Annual Earnings by Industry: 2010 and 2019/20 QLFS Employment

(adjusted)¹ (000s)

SNA Aggregate Earnings (adjusted)² (R billion 2020 prices)

QLFS Earnings Uprating Required to align

with SNA 2010 2019/20

Growth

pa 2010 2019//20 Growth

pa 2010 2019//20

1. Agriculture 812 887 1.0% 49.3 51.4 0.4% 1.21 1.06

2. Mining and quarrying 502 466 -0.8% 142.5 168.8 1.8% 1.82 2.68

3. Manufacturing 1,860 1,768 -0.5% 409.2 472.2 1.6% 1.85 2.15

4. Electricity, gas and water 95 130 3.4% 37.1 52.1 3.7% 2.22 1.53

5. Construction 1,117 1,360 2.2% 131.5 140.2 0.7% 1.07 1.01

6. Wholesale and retail trade 3,111 3,401 1.0% 401.4 450.0 1.2% 1.28 1.31 7. Transport and communication 823 1,003 2.2% 181.7 187.5 0.3% 1.45 1.46 8. Finance and business services 1,770 2,553 4.0% 398.5 514.7 2.8% 1.39 1.23 9. Community and personal services

Government services 1,838 2,292 2.4% 570.1 742.8 2.9% 1.79 1.97

Non-government services 1,064 1,459 3.5% 162.3 216.4 3.2% 1.06 1.01 Domestic (household) service 1,253 1,312 0.5% 31.9 44.6 3.7% 1.00 1.00

Total 14,246 16,631 1.7% 2,515.5 3,040.6 2.1% 1.47 1.50

QLFS employment adjusted upwards for agriculture in 2010 and mining in 2010 and 2019/20.

SNA compensation of employees adjusted to include mixed income attributable to employers and the self- employed in unincorporated enterprises.

In appendix A, we set out the results of pro-rating earnings distributions by industry proportionately, so that the 90^th-10^th percentile ratios remain unchanged within each industry and the resulting shift in the overall earnings distribution is an outcome of the relative upward adjustments in each sector.

We prefer our second adjusted earnings alternative, in which the underlying assumption is that under-reporting in the QLFS rises, in relative terms, as earnings rises. Proportionate increases are computed for the log of earnings within each industry, calibrated to yield the require gross earnings uprating per industry. The results of these calculations are set out in tables 14 and 15, for 2010 and 2018 respectively.

(23)

Table 14: Mean earnings and selected earnings quantiles by industry, 2010 – QLFS total employment with log- earnings uprated proportionately to adjusted SNA sectoral gross earnings estimates

Ratio of 90^th - 10^th percentile

Sector 10th 25th 50th 75th 90th 99th

1. Agriculture 5,063 1,321 1,901 2,472 3,373 9,778 53,192 7

3. Manufacturing 18,361 2,810 4,688 8,980 18,859 41,361 164,725 15

5. Construction 9,840 1,380 2,589 4,443 8,581 19,493 103,062 14

6. Wholesale and retail trade 10,768 1,561 2,958 5,022 10,164 21,994 92,281 14 7. Transport and communication 18,450