A Quarterly Forecasting Model for Employment Ekkehard Ernst and - - PowerPoint PPT Presentation

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

A Quarterly Forecasting Model for Employment

Ekkehard Ernst and Christian Viegelahn

Employment Trends Unit, International Labour Organization

Project LINK Annual Meeting, New York October 24, 2012

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Employment forecasts at the ILO

Annual projections presented in the ILO Global Employment Trends report

Medium-term unemployment/employment outlook by region Based on GDP forecast of the IMF Only annual information is available Country forecasts possible, but updates are done only every 6 months

Increased demands for short-term country-specific forecasts

No quarterly model is currently available

Need for a quarterly model at the ILO

⇒ Start with models for G7 countries, here: USA ⇒ Consider hiring expectations, uncertainty and labour market mismatch

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Use of vector autoregressive model

Start from the following benchmark VAR:

dGDPt = α1 +

L

• j=1

β1,jdGDPt−j +

L

• j=1

γ1,jdGCFt−j +

L

• j=1

δ1,jdEMPt−j + ǫ1,t dGCFt = α2 +

L

• j=1

β2,jdGDPt−j +

L

• j=1

γ2,jdGCFt−j +

L

• j=1

δ2,jdEMPt−j + ǫ2,t dEMPt = α3 +

L

• j=1

β3,jdGDPt−j +

L

• j=1

γ3,jdGCFt−j +

L

• j=1

δ3,jdEMPt−j + ǫ3,t

dGDP: real GDP growth dGCF: real investment growth dEMP: employment growth L: optimally chosen lag order (AIC)

Extensions of benchmark model

⇒ Add indicators for hiring expectations, uncertainty, labour market mismatch

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Indicator for hiring expectations

Manpower hiring index

Temp work agency Manpower undertakes survey among 66,000 hiring managers in 42 different countries Manpower hiring index is published in the Manpower Employment Outlook Survey on a quarterly basis Length of national, sectoral, sub-national time series for G7 countries ranges from 1973-2012 to 2003-2012 Survey question: How do you anticipate total employment at your location to change in the next quarter as compared to the current quarter? Index takes percentage of employers anticipating an increase in hiring activity and substracts percentage of employers anticipating a decrease in hiring activity

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Hiring expectations and employment growth in the US

Manpower hiring index and non-farm employment growth (rescaled), USA

Source: ILO calculations based on ManpowerGroup and U.S. Bureau of Labor Statistics. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Indicator for uncertainty

Economic policy uncertainty index

Index of Baker et al. (2012), based on three components 1 Index of search results on economic policy uncertainty from newspapers 2 Number of tax code provisions set to expire in future years 3 Disagreement of economic forecasters on predictions

Implied volatility indicator

Indicator based on Black and Scholes (1973) option pricing model, applied to employers and the labour market Interpretation of the Manpower hiring index as an option price and labour productivity as the underlying stock

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Uncertainty and employment growth in the US

Political uncertainty index, adjusted volatility indicator and non-farm employment growth (rescaled), USA

Source: ILO calculations based on U.S. Bureau of Labor Statistics, ManpowerGroup, Baker et al. (2012) and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Indicator for labour market mismatch

There is evidence for an outward shift of the US Beveridge curve in 2009-2011:

Source: ILO calculations based on data from OECD.

Labour market mismatch indicator

Indicator calculated on the basis of data on job openings and layoffs in 16 sectors Calculated as 1 − 16

i=1 min(sh openingit, sh layoffit)

sh openingit: openings in sector i as share of total openings sh layoffit: layoffs in sector i as share of total layoffs Illustrates how “well” in terms of sectors laid-off workers fit to open positions

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Mismatch and employment growth in the US

Labour market mismatch indicator and non-farm employment growth (rescaled), USA

Source: ILO calculations based on U.S. Bureau of Labor Statistics. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US non-farm employment

Benchmark specification vs. specification with hiring expectations

Source: ILO estimates based on U.S. Bureau of Labor Statistics, ManpowerGroup and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US employment by sector (2012Q3 = 100)

Specification with hiring expectations by sector

Source: ILO estimates based on U.S. Bureau of Labor Statistics, ManpowerGroup and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US employment by sector (2012Q3 = 100)

Specification with hiring expectations by sector

Source: ILO estimates based on U.S. Bureau of Labor Statistics, ManpowerGroup and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US employment by sector (2012Q3 = 100)

Specification with hiring expectations by sector

Source: ILO estimates based on U.S. Bureau of Labor Statistics, ManpowerGroup and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US non-farm employment

Directly forecasting non-farm employment vs. “bottom-up” approach: benchmark specification and specification with hiring expectations

Source: ILO estimates based on U.S. Bureau of Labor Statistics, ManpowerGroup and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Gains in forecast precision when including hiring expectations

Percentage gains in RMSE, non-farm employment growth, USA

Note: Based on pseudo-out-of sample analysis with increasing window.

Gains in forecast precision when forecasting US sectoral employment:

In 12 out of 15 sectors, there are large gains in forecast precision Only for information services and education and health services, sectoral hiring indices do not work well - likely because in these cases hiring indices that are used do not cover exactly these sectors For utilies employment, the hiring index is neutral to forecast performance

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US non-farm employment

Benchmark specifications vs specification with economic policy uncertainty

Source: ILO estimates based on U.S. Bureau of Labor Statistics, ManpowerGroup, Baker et al. (2012) and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US non-farm employment

Benchmark specifications vs specification with implied volatility

Source: ILO estimates based on U.S. Bureau of Labor Statistics, ManpowerGroup and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Forecasts of US non-farm employment

Benchmark specifications vs specification with labour market mismatch

Source: ILO estimates based on U.S. Bureau of Labor Statistics and OECD. Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Gains in forecast precision when including uncertainty indicators or labour market mismatch

For both uncertainty indicators, there are losses in forecast precision when comparing to the benchmark specification; however, this is only true for US employment forecasts Possible reasons:

Uncertainty does not matter for employment decisions of employers Uncertainty has only started to matter recently and it is therefore difficult to capture it in a time series model

For labour market mismatch, time series is too short to undertake a pseudo-out-of sample analysis

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment

Introduction Methodology Hiring expectations Uncertainty Mismatch Results Conclusions

Conclusions

US employment:

Most models predict a continuation of the current upward trends in US employment Especially construction and services are driving employment gains, while manufacturing employment is stagnating There are some significant downward risks for US employment due to labour market mismatch and uncertainty

Modeling:

The inclusion of hiring expectations considerably improves forecasts of employment in the US

Ekkehard Ernst and Christian Viegelahn Employment Trends Unit, International Labour Organization A Quarterly Forecasting Model for Employment