Abstract
This paper provides comprehensive and detailed empirical regression analyses of the
sources of wage persistence. Exploring a rich matched employer-employee data set and
the estimation of a dynamic panel wage equation with high-dimensional fixed effects,
our empirical results show that permanent unobserved heterogeneity plays a key role in
driving wage dynamics. The decomposition of the omitted variable bias indicates that the
most important source of bias is the persistence of worker characteristics, followed by the
heterogeneity of firms’ wage policy and last by the job-match quality. We highlight the
importance of the incidental parameter problem, which induces a severe downward bias
in the autoregressive parameter estimate, through both an in-depth Monte Carlo study
and an empirical analysis. Using three alternative bias correction methods (the split-panel
Jackknife (Dhaene and Jochmans, 2015), an analytical expression (Hahn and Kuersteiner,
2002), and a residual based bootstrap approach (Everaert and Pozzi, 2007, Gonçalves
and Kaffo, 2015)), we observe that up to one-third of the reduction of the autoregressive
parameter estimates induced by the control of permanent heterogeneity (high dimensional
fixed effects) may not be justified.
sources of wage persistence. Exploring a rich matched employer-employee data set and
the estimation of a dynamic panel wage equation with high-dimensional fixed effects,
our empirical results show that permanent unobserved heterogeneity plays a key role in
driving wage dynamics. The decomposition of the omitted variable bias indicates that the
most important source of bias is the persistence of worker characteristics, followed by the
heterogeneity of firms’ wage policy and last by the job-match quality. We highlight the
importance of the incidental parameter problem, which induces a severe downward bias
in the autoregressive parameter estimate, through both an in-depth Monte Carlo study
and an empirical analysis. Using three alternative bias correction methods (the split-panel
Jackknife (Dhaene and Jochmans, 2015), an analytical expression (Hahn and Kuersteiner,
2002), and a residual based bootstrap approach (Everaert and Pozzi, 2007, Gonçalves
and Kaffo, 2015)), we observe that up to one-third of the reduction of the autoregressive
parameter estimates induced by the control of permanent heterogeneity (high dimensional
fixed effects) may not be justified.
Original language | English |
---|---|
Publisher | IZA. Institute of Labor Economics |
Number of pages | 38 |
Publication status | Published - Oct 2021 |
Publication series
Name | IZA Discussion Papers |
---|---|
Publisher | IZA Institute of Labor Economics |
No. | 14798 |
ISSN (Electronic) | 2365-9793 |
Keywords
- Wage persistence
- High-dimensional fixed effects
- Match effects
- Incidental parameter problem