Dennis Essers. Institute of Development Management and Policy (IOB) University of Antwerp

Transcription

1 South African labour market transitions during the global financial and economic crisis: Micro-level evidence from the NIDS panel and matched QLFS cross-sections Dennis Essers Institute of Development Management and Policy (IOB) University of Antwerp Presentation at the UNU-WIDER Conference on Inclusive Growth in Africa: Measurement, Causes and Consequences Helsinki, 21 September 2013, Parallel Session 4.1: Labour Mobility

2 Contents Introduction Data description: NIDS and matched QLFS Transition matrices and mobility measures Empirical model set-up Results and discussion Concluding remarks 21/09/2013 2

3 Introduction Macro-level impacts of global crisis on developing and EM economies: private capital flows, trade, remittances, etc. (IMF 2009, 2010; ODI 2010; World Bank 2009) South Africa was well-integrated into the world economy and did not escape the crisis; entered recession in 2008Q4, driven by decline in manufacturing, mining, wholesale/retail trade and financial/real estate/business services Recovery has been anaemic and punctuated by renewed global economic slowdown 21/09/2013 3

4 Annualised growth of (seasonally-adjusted) quarterly GDP at constant prices (%) /09/2013 4

5 Introduction (2) Adverse macro-economic trajectory has not been without consequences for South Africans (e.g., Ngandu et al. 2010) Focus here on labour market transitions: Official (QLFS) figures indicate net employment loss of 1 million individuals over 2008Q4-2010Q3 and rise in unemployment rates over Labour market status is critical determinant of household and individual wellbeing (World Bank 2012), also in SA (Leibbrandt et al. 2012) (Pre-crisis) high and structural unemployment and segmented labour markets described as SA s Achilles heel (Kingdon & Knight 2009) Economic recessions tend to have heterogeneous impacts on workers (e.g., Kydland 1984; Cho & Newhouse 2011; Hoynes et al. 2012) Complement to earlier crisis impact studies, which use repeated crosssections of QLFS (Leung et al. 2009; Verick 2010, 2012) Research question: which individual, household-level and jobspecific characteristics are associated with staying employed, or not, in SA during the height and aftermath of the global crisis? 21/09/2013 5

6 Evolution of narrow and broad unemployment rates (QLFS), annual averages (%) Narrow and broad unemployment, overall Broad unemployment, by gender Narrow Broad Male Female Cross-sectional data only provide a netted-out picture of changes in SA labour markets To evaluate gross changes we need longitudinal datasets 21/09/2013 6

7 Data description: NIDS National Income Dynamics Study (NIDS) is SA s first nationally representative, multi-purpose, individual-level panel data survey 2 NIDS waves : panel of 21,098 individuals appearing both in wave 1 (Jan2008-Dec2008) and wave 2 (May2010-Sep2011) Analysis restricted to adults aged in 2008 (cf. Cichello et al. 2012) 6 mutually exclusive labour market statuses: Regular wage employment Self-employed Casual and other employment Searching unemployed Discouraged unemployed Not economically active (NEA) Problems with NIDS (SALDRU 2012): Some misclassification between different categories of the non-employed during wave 2 fieldwork Between-waves attrition rates are especially high for better-off Whites, which reduces reliability of estimates for this group 21/09/2013 7

8 Data description: QLFS Quarterly Labour Force Survey (QLFS) is SA s official, nationally representative survey on labour market activity since 2008Q1 Designed as rotating panel of dwellings (+/- 30,000); each quarter 25% of dwellings is replaced; only household identifiers are generally maintained Matching of individuals from quarter t to quarter t+1 using household ID, age, gender, race, education, marital status (Ranchod & Dinkelman 2008): 760,847 matched obs over 2008Q1-2012Q4, average matching of 68.8% IPW techniques to correct for non-random matching on observables Analysis restricted to adults aged in quarter t 5 mutually exclusive labour market statuses: Formal sector employment Informal sector employment Searching unemployed Discouraged unemployed Not economically active (NEA) Problems with matched QLFS: Non-random matching on unobservables False matches 21/09/2013 8

9 Transition matrices: NIDS Transition matrix for labour market status, /11: row proportions (%) Labourmarket status in Reg. wage employment Labour market status in 2010/ Casual and Reg. wage Selfemployment search. disc. Unemployed, Unemployed, other NEA employment employment Selfemployment Casual and other employ Unemployed, search Unemployed, disc NEA Overall mobility, M total = M upward + M downward + M within non-employment + M within employment 51.4% = 12.6% % % + 6.6% 21/09/2013 9

10 Transition matrices: QLFS Transition matrices for labour market status, 2008Q1-2012Q4: row proportions (%) Labour market status in quarter t+1 Formal sector employment Informal sector employment Unemployed, search. Unemployed, disc. NEA Labourmarket status in quarter t Formal sector employment Informal sector employment Unemployed, search Unemployed, disc NEA Overall mobility, M total = M upward + M downward + M within non-employment + M within employment 2008: 21.0% = 4.8% + 4.0% + 8.9% + 3.3% 2009: 19.4% = 3.6% + 3.5% + 9.6% + 2.7% 2010: 19.0% = 3.4% + 3.0% % + 2.4% 21/09/ : 18.7% = 3.2% + 2.9% % + 2.3% 2012: 18.2% = 3.3% + 3.0% + 9.6% + 2.4% 10

11 Model set-up Simple (survey-weighted) binary probit models: Pr(y=1 X, Z) = Φ(X β + Z δ) Two kinds of probits: 1) NIDS: y equals 1 if individual in regular wage employment in 2008 and again in 2010/11; 0 if no longer in regular wage employment in 2010/11 2) QLFS: y equals 1 if individual in formal sector employment in quarter t and again in quarter t+1; 0 if no longer in formal sector employment in quarter t+1; quarter-to-quarter transitions are pooled per year over X is vector of individual and household-level demographic and location variables: age cohort, education, race, household size, rural/urban, province dummies, etc. Z is vector of job-specific variables: occupation and industry types, union membership, contract type/duration, etc. All estimations separate for men and women 21/09/

12 NIDS probit estimates for regular wage employment transitions, /11 (baseline variables): average marginal effects (1a) Male (1b) Female Omitted: age Age Age * * Age Omitted: no education Primary education ** Secondary education *** Tertiary education *** Omitted: Black/African Coloured Asian/Indian White Married ** Household size *** Rural *** Province dummies Yes Yes Observations 1,122 1,199 21/09/

13 NIDS probit estimates for regular wage employment transitions, /11 (extra job variables): average marginal effects (2a) (2b) (3a) (3b) (4a) (4b) (5a) (5b) (6a) (6b) Male Female Male Female Male Female Male Female Male Female Baseline regressors (not shown) Omitted: elementary occupation Semi-skilled ** Manag./professional ** Omitted: agriculture, hunting, forestry, fishing Mining and quarrying *** Manufacturing Utilities *** Construction *** Wholesale and retail trade ** Transport, storage and communication Fin. intermed., real estate and bus. services Community, social and personal services Union member *** Written contract * Omitted: limited contract duration Unspecified contract duration Permanent contract ** Province dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 1,096 1, ,092 1,179 1,110 1,192 1,117 1,190 21/09/

14 QLFS probit estimates for formal sector employment transitions, 2008Q1-2102Q4 (baseline variables): average marginal effects (1a) (1b) Male Female Omitted: age Age * * *** *** *** *** *** * *** ** Age *** *** *** *** *** *** *** *** *** *** Age *** *** *** *** *** *** *** *** *** *** Omitted: no education Primary education ** * ** ** Secondary education *** *** *** ** *** *** *** *** ** ** Tertiary education *** *** *** *** *** *** *** *** *** *** Other education * ** *** *** *** * Omitted: Black/African Coloured *** ** *** ** *** Asian/Indian *** *** * * White * *** *** *** *** *** Married *** *** *** *** *** Household size *** *** *** *** *** *** ** *** *** *** Rural * *** ** * *** Observations 12,063 12,441 12,438 11,561 12,564 9,100 9,789 9,779 9,358 10,079 21/09/

15 QLFS probit estimates for formal sector employment transitions, 2008Q1-2102Q4 (extra job variables): average marginal effects (2a) (2b) Male Female Omitted: age Age *** *** *** *** *** *** ** Age ** ** *** *** *** *** *** *** *** *** Age *** *** *** *** *** *** *** *** *** *** Omitted: no education Primary education * * * Secondary education *** *** *** ** *** *** *** ** Tertiary education *** *** *** *** *** *** *** *** *** *** Other education * *** *** *** *** Omitted: Black/African Coloured *** * *** *** ** Asian/Indian *** *** White ** *** *** *** *** *** Married *** *** *** *** *** Household size *** *** *** *** *** *** ** *** *** *** Rural * *** *** ** Omitted: agriculture, hunting, forestry, fishing Mining and quarrying *** ** ** *** *** *** *** Manufacturing ** *** *** Utilities ** *** Construction *** *** *** *** *** Wholesale/retail trade * *** *** *** Transport et al * ** ** *** *** *** ** Fin. intermed. et al * *** Comm. et al. services ** *** *** * ** Observations 12,062 12,436 12,436 11,557 12,560 9,097 9,786 9,774 9,260 10,078

16 Main findings Considerable mobility in SA labour markets over (cf. other periods: Banerjee et al. 2008; Ranchod & Dinkelman 2008) NIDS and QLFS suggest that likelihood of continued employment differs significantly between particular types of workers in SA: Lower for younger workers, workers with less than secondary education and males employed in construction and trade Higher for trade union members and those with written and/or permanent contracts Evolution of transitions between labour market states over 2008Q1-2012Q4: Overall mobility gradually decreased; rise in unemployment rates mostly due to reduced inflows into employment (cf. Verick 2012) Time variation in economic significance of some demographic and job-specific exploratory variables; e.g., diminishing strength of buffering effect of higher education Not straightforward to distinguish trends that can be connected to the broader evolution of SA s economy over the course of the crisis 21/09/

17 Avenues for further research A study of other labour market transitions: e.g., factors that help or hinder the unemployed in SA in finding a job during the difficult economic climate of (cf. Posel et al for NIDS) Use of more detailed information on occupations/job tasks and specific subsectors of employment Extension of the NIDS panel with a third wave 21/09/

18 Thank you for your attention Mail: Presentation at the UNU-WIDER Conference on Inclusive Growth in Africa: Measurement, Causes and Consequences Helsinki, 21 September 2013, Parallel Session 4.1: Labour Mobility

19 Matching algorithm for QLFS (cf. Ranchod & Dinkelman 2008) 1) Append all QLFS cross-sections (quarters), sort on household identifier and quarter, and drop households that appear only once 2) For each quarter and within the same household, drop individuals that have same race, gender and ages differing by at most 1 year 3) Match remaining individuals across quarter t and quarter t+1 using household identifier, gender, race and age t = age t+1 4) Match also individuals across quarter t and quarter t+1 using household identifier, gender, race and age t +1 = age t+1 5) Keep only the individuals matched in Steps 3 and 4 to form expanded match panel 6) Impose extra consistency requirements on expanded match panel to form strict match panels, by dropping: Individuals whose level of education differs between quarter t and quarter t+1 Individuals whose marital status changes from married / divorced / widowed in quarter t to never married in quarter t+1 21/09/