On the Estimation of Fixed Worker and Employer Effects

IZA Logo

IZA Seminar

Place: Schaumburg-Lippe-Str. 9, 53113 Bonn

Date: 12.04.2005, 12:15 - 13:30


Presentation by 

Erling Barth (Institute for Social Research, Oslo)


We introduce a way of computing the fixed worker and firm effects in large data sets that simplifies the calculations required to solve the set of normal equations. The effects of time varying covariates is estimated using within-job variation only, while the fixed worker and employer effects are calculated on the set of job-movers only using a simple block iterative algorithm. Finally, the effects of firm- and worker specific covariates, such as industry or education, are estimated from the fixed effects. We use the estimated models to revisit several of the long-debated issues in labour economics, such as the tenure effects on wages, the gender wage gap, the firm-size effect, inter industry wage differentials and the city-wage premium, and provide an estimate of the bias in the standard OLS regression model. The bias is decomposed into the part arising from worker heterogeneity versus the part arising employer heterogeneity for each of the variables.

For more information, please contact seminar@iza.org