Asymmetric labour force participation decisions

Abstract

This article finds compelling evidence of asymmetric labour force participation decisions across demographic groups in response to changes in labour market conditions. This behaviour is consistent with predictions from the standard labour-leisure choice model and suggests that asymmetry in individual behaviour plays some role in the observed asymmetry of the aggregate unemployment rate. It is estimated that the weighted average difference in response to a one percentage point change in unemployment rates is sizeable when compared to the average monthly change in the aggregate labour force.

Keywords:

• labour supply
• labour force participation
• aggregate unemployment
• asymmetric effects

JEL Classification::

• J22
• J64

Acknowledgements

Christopher Bollinger, Sabrina Pabilonia, and Richard Rogerson offered valuable comments; research assistance was provided by Andrew Balthrop, Mark Carter, M. Laurel Graefe, Menbere Shiferaw and Nicole Baerg.

Notes

¹ Several statistical descriptions of the asymmetric time series pattern have also been proposed, including Hamilton (2005), Moshiri and Brown (2004), Rothman (1998), van Dijk and Franses (2000) and Montgomery et al. (1998).

² Exceptions include Blau (1978), who allows a differential impact of the unemployment rate on the individual labour force entry and exit decisions of married woman between two time periods, and Darby et al. (2001) who use aggregate time series data and allow for different participation rate responses of demographic groups to expansionary and contractionary phases of the business cycle.

³ Some have argued that a measure of flows into employment (e.g. see Hall (2005) and Carlsen et al. (2006)) would be a better measure of labour market strength. However, the high degree of correlation between these different flow measures and the unemployment rate reassures us of the usefulness of the unemployment rate. In addition, flow data for the US are not available on a disaggregated level. Other measures that might be considered are less desirable for this investigation. For example, the link between employment prospects and measures of demand such as Gross Domestic Product (GDP) or the employment to population ratio is less direct.

⁴ This specification is similar to that applied by Mocan and Bali (2005) to investigate the effect of movements in the nemployment rate on criminal activity. Another application of a similar specification can be found in Koop and Potter (1999).

⁵A person's labour supply decision in the current month is expected to respond to the previous month's observed unemployment rate. This lagged structure avoids perfect simultaneity between the observed unemployment rate and observed labour force participation decisions in the same month.

⁶ Other model specifications were estimated with no appreciable qualitative difference in any of the conclusions resented here. Specifically, a model with a double-symmetry threshold parameter (allowing for two different δs), a model with a continuous unemployment rate difference regressor, a model with a simple unemployment rate movement (up or down) indicator variable, and a model in which δ is measured as a percentage rather than level change in the unemployment rate were all explored.

⁷ The CPS is a monthly survey of about 50 000 households conducted by the United States Bureau of the Census for the Bureau of Labour Statistics. The CPS is the primary source of information on the labour force characteristics of the US population. The sample is selected to represent the civilian noninstitutional population, and respondents are interviewed to obtain information about the employment status of each member of the household 15 years of age and older.

⁸ Over the entire period 1980 to 2009 sample time period, the median change in unemployment rates from one year to the next is 1.5 percentage points.

⁹ Evidence of weaker labour market attachment of women and the less-educated can be found in Erosa et al. (2005) and Antecol and Bedard (2004). Also see Ammuermüller et al. (2006) for evidence in Germany consistent with labour market conditions having a stronger expected effect on labour force participation of women and less-skilled workers.

¹⁰ The absence of an income effect among nonparticipants was first discussed by Ben-Porath (1973, p. 702). Technically, the income effect only approaches zero in the limit as desired hours of work goes to zero (see Hausman (1983) and Heckman (2001)). As a result, the impact of a given decline in the probability of finding a job on labour market exits will be less than the impact on entrances of a probability increase of the same size.

¹¹ The difference in the number of people entering and exiting in response to a one percentage point change in the unemployment rate is calculated as Population.