The persistent labour-market effects of the financial crisis

Abstract

This article estimates the effects of the financial crisis on the Swedish labour market. Using an unobserved components model and an external forecast, we estimate a future path for the Nonaccelerating Inflation Rate of Unemployment (NAIRU). Judging by this analysis, the labour market will be in equilibrium again in 2013. Linking the NAIRU to other labour-market variables through an estimated vector error correction model and population projections, it is found that this new equilibrium is associated with a smaller equilibrium labour force and lower equilibrium employment.

Acknowledgments

We are grateful to the seminar participants at the National Institute of Economic Research for their valuable comments on this article.

Notes

¹ The fact that business cycle fluctuations can cause permanent effects on, for example, unemployment and labour-force participation is well established in the academic literature. Explanations to why macroeconomic shocks can have long-lasting effects on the labour market include decay of human capital, discouragement, habituation and the behaviour of unions in the wage bargaining process; see, for example, Pissarides (1992), Clark et al. (2001) and Blanchard and Summers (1986).

² In this article, the terms NAIRU and equilibrium unemployment rate are used synonymously.

³ Note that the constant term in Equation 8(8) is restricted such that a constant term is allowed for in the cointegrating relationship, but no drift in the system.

⁴ Population predictions are well known for having extremely small prediction errors, and any uncertainty in our calculations stemming from these predictions is dwarfed by estimation uncertainty in the other steps. For a discussion regarding the use of population data in macroeconomics, see, for example, Lindh (2004).

⁵ This is 10 years less than when the unobserved components model was estimated. The reason a shorter sample is employed is that we are concerned that the behaviour of women during the 1970s is not representative of that today.

⁶ Lag length in the ADF–tests and VARs was established using the Akaike information criterion.

⁷ If H ₀ : r = 1 had been rejected, it would imply that both series are stationary; see, for example, Taylor and Sarno (1998).