This script is what underlies the following paper. It uses autoregressions in R to model time series trends of aggregate macroeconomic variables. The DataOUT folder contains the output of this script, which is pulled into "Writing/Forecasting paper..tex". Slides. YouTube presentation.
Horn, A.J. & Donaldson, A.R. 2021. Labour Market Effects of the Great Lockdown in South Africa: Earnings and Employment During 2020–2022. (SALDRU Working Paper 279). Cape Town: Southern Africa Labour and Development Research Unit, School of Economics, University of Cape Town. Available: http://opensaldru.uct.ac.za/handle/11090/1007
This paper quantifies the impact of the covid-19 economic shock on aggregate earnings and employment by industry in South Africa. We construct pre-covid-19 counterfactual forecasts for the 2020 Q2 – 2022 Q4 period and compare these with reported earnings and employment levels up to 2021 Q1. We find that total compensation of employees in 2020 Q2 was 9% below forecast while employment was 14% lower than the counterfactual. Between 2020 Q2 and 2021 Q1, aggregate earnings recovered more than three times as quickly as employment, indicating a rise in inequality. We calculate possible recovery paths of earnings and employment to 2022 Q4. We outline implications for the Unemployment Insurance Fund and suggest ways in which the employment recovery might be accelerated.
One needs to run this script before the earnings_industry_<year>_x.R scripts (in "LMDSA 2018/Scripts" and "QLFS 2010/Scripts"), as it creates the employment_industry table object in the R session and holds it in memory. The employment_industry table is only used to create earnings_industry.tex and annual_earnings.html, which are output tables showing estimates of aggregate earnings according to the base QLFS employment levels.