Abstract
This article decomposes the rise in the cross-sectional variance of male annual earnings in Sweden between 1991 and 1999 into its persistent and transitory components. The results show that the persistent component accounts for basically all of the increase in earnings dispersion. This implies that the answer to the 1990s trend reversal in Swedish earnings inequality is to be found in explanations that focus on persistent changes in the labour market, such as changes in the price of skills.
Acknowledgements
I am grateful to Per-Anders Edin, Meredith Beechey, Patrik Hesselius, Bertil Holmlund, Stephen Jenkins, Mårten Palme and an anonymous referee. Financial support from the Jan Wallander and Tom Hedelius Foundation is gratefully acknowledged.
Notes
1 See Edin and Holmlund (1995) and Palme and Wright (Citation1998) for an account of changes in Swedish pay differentials between 1968 and 1991. See for instance Katz and Autor (Citation1999) for an international comparison of pay differentials.
2 A random walk specification is also used, among others, in Dickens (Citation2000) and Moffitt and Gottschalk (Citation2002).
3 The age intervals are chosen based on graphs of how the longer-lag auto-covariances, which mainly reflect persistent inequality, vary over the life cycle; these graphs are available on request.
4 See for instance Haider (Citation2001) and Baker and Solon (Citation2003) for a further discussion on this subject.
5 Wald tests are used throughout the analysis.
6 In fact, the initial variances change every two years, corresponding to the cohort estimates in . For example, the initial variance for cohort 1951/52 is a direct estimate of the variance of the transitory component for individuals aged 40 years in 1991. For 1993, I use the initial variance for cohort 1953/54, whose members are 40 in this year.