Hysteresis and non-linearities in unemployment rates

Abstract

This study tests whether there is evidence of mean reversion in unemployment rates using the recently developed unit root test of Kapetanios et al. (2003). In this framework, the null hypothesis of a unit root process is tested against the alternative of a globally stationary exponential smooth transition autoregressive process. Applying the test to monthly data for Australia, Canada, Finland, Sweden and the USA, it is concluded that unemployment hysteresis finds less support when non-linearities are allowed for compared to the benchmark of using a standard Augmented Dickey–Fuller test.

Acknowledgements

Financial support from Jan Wallander's and Tom Hedelius’ foundation is gratefully acknowledged.

Notes

¹ One can think of other mechanisms than unions and laws that imply insider power over wages, like the possibility for insiders to harass outsiders who underbid the prevailing wage; see for instance Lindbeck and Snower (1988). This should, however, be possible both for workers in Europe and the USA, so the presence of unions and laws should still lead to more insider power in Europe.

² A substantial part of the literature relies on more structural models to investigate the issue of hysteresis though; see for instance Jacobson et al. (1997) and Taheri (2000). Furthermore, it is not uncommon that the definition of hysteresis is slightly different so that a unit root in the unemployment rate does not necessarily imply hysteresis; see for instance Assarsson and Jansson (1995) and Iregui and Otero (2003).

³ For more recent developments in this field, see for instance Song and Wu (1997), Carrion-i-Silvestre et al. (2001), Strazicich et al. (2001) and Österholm (2004).

⁴ See for instance Blanchard and Summers (1986).

⁵ The data for Australia, Canada, and the USA had been seasonally adjusted whereas the data for Finland and Sweden had not.

⁶ This finding potentially questions that of Koustas and Veloce (1996) who argue that Canadian unemployment is of ARFIMA type. As pointed out by Granger and Teräsvirta (1999), non-linear dynamics can be mistaken for long memory.