The name is absent

Appendix B3: Job Change and Rank Variables

This appendix provides details on the procedure used to minimize classification errors in the rank
variables and in the job change variable. 25% of the observations reporting no change in job are
associated with a change in rank ( 15% for a change up and 8% a change down). This shows inconsistency
in the information provided in the two variables on job change and rank affiliation. This inconsistency
results probably from errors in rank classification (reflected in the high percentage of demotion and the
high average wage growth associated with these demotions).

On the other hand, it may also be the case that some of the changes in rank affiliation between two
periods are “true” changes and the error is in the variable indicating no change in job situation. Note
that for some years, the information on job change is provided in two separate questions. A first question
asking whether the worker’s job situation has changed, another question asking about the type of job
change, increasing the chances of inconsistency in the answers provided. Moreover, errors in the no job
change information are suspected based on the fact that average wage growth for those who report no
change in job but for which there is a change in rank (up or down) is 3.28% which is substantially higher
than the average wage growth of the non job and non rank changers of 1.92%. Some of the reported
non job changes must in fact be job changes associated with a rank change.

To decide when to correct for a possibly false change in rank I used the information on wage growth.
Any observations about non job changers associated with a change in rank up (promotion) for which
workers are receiving a wage growth of more than 5% is treated as a real change in rank. If it is less
than 5% then it is considered as a no change in rank and the current rank is set equal to the previous
period rank. In the case of demotion, wage growth has to be less than 0 for the observation to be
considered as a change in rank.

The first two columns of Table B3 below reports average characteristics for job changers who experience
or not a rank change with the corrections done on the rank variables. Comparing with the last two
columns of Table B2.1 (before correction), one can see that average individual characteristics and
the differences among rank changers and non rank changers did not change much implying that the
correction did not change the informational content of the sample. Only average wage growth is now
higher for rank changers and lower for non rank changers. The remaining of the analysis will be done
based on these rank corrections.

For informational purpose, average characteristics for job changers by type of rank change are presented
in the next three columns. Also since the Gibbons and Waldman model focuses on rank changes, average
characteristics by rank changes (including job change) are provided in the last three columns.

Appendix Table B3: Average Individual Characteristics by Job and/or Rank Changes

40

<<   <   39   40   41   42   43