ability workers will get higher wage increases than low ability workers and the same ordering
will hold for wage increases at all experience levels. The model generates serial correlation in
promotions for the same reasons. If ηz and (η" — η') are both sufficiently large then high ability
workers are promoted to job 2 more quickly and also spend less time on job 2 before being
promoted to job 3. Moreover, since those who receive larger wage increases are also those who
are promoted to job 2 earlier in their careers, wage increases predict promotions.
The model gives predictions consistent with the fact that wage increases predict promotion.
A large wage increase indicates an increase in expected innate ability which means that on
average effective ability will grow more quickly in the future so that the worker will need less
time to reach the target level of expected effective ability needed for promotion.
Finally, wage increases upon promotion explain a fraction of the difference between average
wages across levels because, on average, some of the workers at higher job levels are more
experienced. The difference between average wages at different levels is given by the average
experience or effective ability accumulated. This difference is bigger than the average wage
increase at promotion which captures the value of only one year of experience.
The model with perfect information predicts that average wage increases at promotion are
higher than average wage increases that would occur if workers remains in their current job
levels. This is because increases in effective ability for those who get promoted are valued
in part at the rate of the current job level (cj) and in part at the higher rate of the next
job level (cj). For the same reason, however, the model predicts that average wage increases
after promotion are higher than the average increases at promotion as increases in effective
ability are entirely valued at the higher job level. This conflicts with the empirical findings
which shows that wage increases at promotion are higher than wage increases before and after
promotion. Moreover, the monotonicity of the effective ability accumulation function precludes
the possibility of real wage decreases.
When information on innate ability is imperfect (but symmetric in that workers and firms
have the same information about ability), workers and firms start with the initial belief p0 that
a given worker is of innate ability θH and with 1 — p0 that he is θL . Learning takes place at
the end of each period when the realization of a worker’s output for that period is revealed.
Learning occurs gradually because of the productivity shock εijt , which introduces noise into
the output produced.
To be precise, each period a worker’s output provides a noisy signal, zit, about his effective
ability where: