4.3 Descriptive Data Analysis
4.3.5 Italy
Basic Sample Description
The Italy sample retains 2820 men, each with between 4 and 8 observations, the average being 6.4.27 Of these 2820 individuals, 1639 (58.1%) are initially observed in private sector employment and remain in the sample for an average of 6.4 years. There are 786 men (27.9%) who are first observed in the public sector and remain in the sample for 6.6 years on average. The other 395 individuals (14.0%) are initially unemployed and remain in the sample for an average of 5.9 years.
In terms of education, the Italy sample breaks down overall as 9.0% “high” education level, 51.6% “medium” and 39.4% “low”. The overall break down masks some substantial differences between the two sectors: the private sector comprises 6.6% high, 50.0% medium and 43.4% low educated workers, compared with 14.3%, 56.6% and 29.2% respectively for high, medium and low educated workers in the public sector. Thus the public sector attracts better educated workers, with more than double the proportion of high educated workers compared with the private sector, a greater proportion of medium educated also, with a much smaller proportion of low educated workers. Public sector workers are on average markedly older than private sector workers (43.3 years old compared to 37.8) and have more potential labour market experience (average of 24.5 years compared to 20.2).
Figure 4.4 illustrates the evolution of the Italian labour market over time, the top panel showing the unemployment rate for our sample period and the 4 years preceding it, the bottom panel showing the public sector share of total employment in our sample. The unemployment rate is largely stable at around 7.5% for the years preceding our sample period (which begins in 1994) before rising to just over 8.5% in 1994 and remaining around this level throughout the majority of the sample before declining back down towards 7.5% in the final years of the sample. The public share of total employment in the sample remains almost constant, declining slightly over the sample period from around 34% in 1994 to 32% in 2001.
27
There is some sample attrition, 19.2% after 4 years and 60.7% at 8 years, which we assume to be exogenous. Some of the attrition is a consequence of our sample construction rules which treat an individual as censored from the first time that they have a gap in their response history (this affects 11.3% of our sample).
6 7 8 9 10 11 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 year
Italy: unemployment rate (LFS)
27.5 30 32.5 35 37.5 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 year
Italy: public share of total employment (sample)
Figure 4.4: Italy: Unemployment rate, males, 1991-2001, and Sample public sector share
Differences in Earnings
Earnings levels. Table 4.7 describes public-private differences in earnings levels through a num- ber of simple regressions, where again the dependent variable is the log of current gross monthly earnings. There is a definite difference between the sectors in terms of their hours distribution for full-time workers, with the public sector working fewer hours on average: median 36 hours per week compared with 40 in the private sector; therefore we will tend to underestimate any positive public premium in hourly wage. There is also less of a spread in hours in the public sector, the standard deviation in weekly hours is lower by around fifty-five minutes.
The first column of Table 4.7 shows that the estimated raw public pay gap in the data is 8.6 log points (9.0%) in favour of the public sector. However, conditioning on education band and a quadratic in potential labour market experience (column 2) the estimated premium is not significantly different to zero. This changes when we allow the effect of the public sector to vary according to education band and experience (column 3), the public premium is then estimated to be statistically significant and 21.0 log points (23.4%), though falling with experi- ence. This specification suggests that the premium does not vary depending on the education band. Estimating the model with individual-level fixed effects controlling for a quadratic in ex-
perience (column 4) the public premimum is not significantly different from zero.28 Introducing interactions of the public sector dummy with the quadratic in experience (column 5), the public premium is still not statistically significant but is estimated to be around 5.4 log points (5.5%), with a large standard error resulting in a t-statistic of only 1.15.29 This latter fixed effects specification (column 5) suggests that returns to experience are greater in the private sector at all levels of experience, such that the small public premium traces a U-shape in experience, be- coming negative at 12 years of experience and remaining negative thereafter, with an estimated minimum of −4.4 log points at 34 years of experience.
Earnings dispersion. In each sector the standard deviation of log earnings is almost identical, 0.293 in the private sector as compared with 0.296 in the public sector. The 90:10 percentile ratios in raw earnings suggest however that the private sector distribution is more spread, 2.058 (private) against 1.944 (public), with corresponding figures conditional on age and education being 1.897 (private) and 1.844 (public) respectively. These figures indicate that the extent of wage compression is similar in each sector though still greater in the public sector.
Earnings mobility. The analysis of earnings level and dispersion suggest that in Italy the public and private sectors differ in terms of their earnings levels while the overall compression of earnings is similar for each. Moreover, the sectors differ with respect to earnings mobility, as illustrated in the top panel of Table 4.8 which shows the transition matrices between the quintiles of the unconditional raw earnings distributions from one year to the next for the public and private sectors respectively.30 In these matrices we can see that there is greater persistence in rank in all of the earnings quintiles for those individuals continuously employed in the public sector compared with those continuously employed in the private sector, though in the higher earnings quintiles the level of persistence is very similar. This pattern is also exhibited in the transition matrices for ranks of log earnings conditional on education and a quadratic in (potential) labour market experience (lower panel of the Table) and contributes to a picture of generally greater earnings persistence in the public sector.
Given that the earnings distributions are very similarly spread in each sector, the greater 28
The reported fixed effects regressions use the within-estimator. First differenced estimates for column 4 produces very similar results.
29
Estimating column 5 using first-differences produces similar results though with the public premium 13.3 log points and significant.
30
frequency of transitions between earnings quintiles in the private sector suggests greater earnings mobility in the private sector in terms of the actual earnings received. Moreover, computing the 1-lag auto-covariance of normalised log earnings for each individual, having conditioned on education and a quadratic in (potential) labour market experience, we find greater auto- covariance of earnings for individuals employed in the public sector for successive periods. The mean auto-covariance for these individuals is 0.882, compared to 0.736 for individuals employed in the private sector for successive periods.
Differences in Job Mobility
The transition matrix below shows the changes in employment sector between one wave and the next, with rows referring to sector in yeart-1, columns to sector in year t:
Private Public Unemp. Private 95.2 1.9 2.7 Public 4.8 94.0 1.0 Unemp. 21.9 4.7 73.3
Movement directly from private sector employment in year t-1 to public sector employment in yeartis very rare: only 1.9% of individuals initially in private sector employment are observed the following year employed in the public sector, though movements in the opposite direction are more frequent (4.8%). The annual rates of transition into unemployment are small for each sector, though the private sector rate (2.7%) is more than double the public (1.0%) suggesting greater job security in the public sector. This greater security of public sector employment is also reflected in the probability of ever being observed out of employment depending on the sector the individual is first observed in: of those initially observed in the public sector, only 3.7% are ever observed unemployed during the course of the sample, whereas for those first observed in the private sector, 11.5% are subsequently observed unemployed.
Of the unemployed in year t-1, 73.3% remain unemployed in year t. 21.9% gain employ- ment in the private sector, while just 4.7% move into public sector employment. Of the 612 individuals (21.7% of the sample) ever observed unemployed during their sample observations, 51.3% record being unemployed for three or more of their consecutive interviews. This “long- term” unemployed status has a great effect on the annual re-employment probability, reducing
it to 12.8% compared with a 69.2% 1-year re-employment rate for the “short-term” unemployed. Thus unemployment persistence overall is high, with almost three-quarters of the unemployed remaining so in the next interview, and there is some evidence that this is particularly the case for the “long-term” unemployed.
In addition to differing re-employment probabilities according to “long-term” unemployment status, there is some evidence of public sector attachment: an unemployed individual whose most recent employment was in the public sector has a 15.8% chance of gaining a public sector job in the next period, compared with only a 3.3% chance for those unemployed whose most recently observed employment was in the private sector. The overall re-employment rate for those whose most recent employment was in the private sector is however slightly higher (33.6%) than it is for those most recently employed in the public sector (28.6%).