Measuring worker flows

Abstract

The study explores differences between two measurement concepts of worker flows widely used in the literature referred to as the turnover and reallocation concepts. It is found that measuring worker flows by the turnover concept leads to substantially (about 5% of total employment) higher worker flow estimates and slightly increases age, size and industry group effects on firm level worker flows as well as differences between growing and declining firms relative to the reallocation concept.

Acknowledgements

This research was financed by a research grant from the Austrian National Bank (Jubelfondsprojekt No. 8889) in the framework of a project entitled ‘Job Flows, Worker Flows and Churning in Austria’. The authors would like to thank Helmut Mahringer and Hedwig Lutz for helpful comments.

Notes

¹ Further research addresses the effects of job and worker flows on wages and wage drift (e.g. Den Butter and Eppink, 2003), the effect of innovation on job creation and destruction (e.g. Vainiomaki and Laaksonen, 1999) and a number of further issues (see Davis and Haltiwanger, 1999 for a survey).

² As a consequence, differences between the two measuring concepts decrease with the frequency of measurement and the two concepts accord exactly when measured at a sufficiently high frequency (in the present case on a daily basis).

³ Note that the definition of job creation and destruction holds irrespective of how worker flows are measured.

⁴ Data from this source has been widely used in both macro- as well as micro-econometric studies of the Austrian Labour market (e.g. Winter-Ebmer, 1995; Kratena, 2000; Hofer et al., 2001).

⁵ Public services are excluded since there is no information on tenured public sector employees. Agriculture is excluded due to the high share of self-employment, on which there is no information.

⁶ Stiglbauer (2003) argues that the data are mostly enterprise level, because enterprises lack incentives to increase administrative reporting burdens by choosing to report at establishment level. Unfortunately, since data are provided anonymously, there is no way to correct these deficiencies.

⁷ This high share of short term employment spells which end in re-employment in the same firm is consistent with the high share of temporary layoffs in Austria found by Fischer and Pichelmann (1991).

⁸ This should, however, be no surprise. R ² values increase with the number of included variables and firm fixed effects increases the number of variables included by over 540 000. Information criteria, which correct for the inclusion of more dependent variables, such as the Schwartz or Akaike criteria, suggest that firm dummy variables do not always improve the model fit significantly.

⁹ Time dummies, however, explain slightly more of the variance in the turnover concept. This can be attributed to the higher seasonality of turnover measures.

¹⁰ While separation and hiring rates are highly correlated across measurement concepts (with correlation coefficients of 0.67 and 0.66, respectively), regressions of firm level separation (hiring) rates according to the turnover concept on separation (hiring) rates according to the reallocation concept, reconfirm this finding. A coefficient of 1.10 (1.05) is obtained which is significantly higher than 1 at all conventional significance levels in both cases.

¹¹ The specification of this regression is where Y_i is the dependent variable X_i the set of explanatory variables, α _j is an industry fixed effect and ζ _i an error term.

¹² These results are robust to omission of industry fixed effects.