part-time employment, those who change position during a year, and those with multiple
jobs (127,403 observations). For similar reasons, we dropped individuals who worked for
less than 12 months (63,347 observations). In this way we isolate the on-the-job aspect
of the wage insurance contract, leaving the consideration of changes in the occupational
status to future work. Finally, we kept only individuals with non-zero recorded earnings in
all years (148 observations lost) and eliminated some outliers (808 observations)17.
Our measure of earnings covers remuneration for regular and overtime pay plus non-
wage compensation. We compute net earnings using the Italian tax code for the various
years and deflate them using the CPI.18 For workers with intermittent participation we
treat two strings of successive observations separated-in-time as if they pertained to two
different individuals.
Workers in the resulting sample are on average 41 years old in 1991; production workers
account for 64 percent of the sample, 35 percent are clericals and about 2 percent managers.
Males are 74 percent of our sample and those living in the South 15 percent. Finally, net
earnings in 1991 are roughly 25 million lire on average (at 1991 prices and exchange rates,
roughly $17,300), with a median of 23 million ($15,600).19
6 Estimation of the stochastic structure of firm performance
and workers’ earnings
6.1 Firm performance
As a measure of the idiosyncratic shock to firm performance (ε) we use unexplained variation
in the logarithm of value added at 1991 prices (deflated by the CPI). Value added is the
closest measure to the theoretical concept of firm performance y in the model described in
1'An observation is classified as an outlier if (a) real earnings are below 1 million lire (500 Euro), (2)
real earnings are below 6 million and the growth rate is below -200 percent, or (c) real earnings exceed 100
million lire and the growth rate is greater than 200 percent.
18Results are very similar if we use gross income as a measure of earnings.
19These descriptive statistics can be compared to a representative sample of the Italian population of
private sector workers drawn from the 1991 Bank of Italy SHIW. We find that demographic characteristics
in the SHIW are very similar to those in the INPS (in particular, the proportion of males, production
workers, clericals and managers) and average age is the same in the two samples.
18