4.3 Descriptive Data Analysis
4.3.2 Germany
Basic Sample Description
For Germany the constructed sample retains 3026 men, who each have between 4 and 8 consec- utive observations, 6.8 being the average.9 Of the 3026 men, 2208 (73.0%) are initially observed in private sector employment, and remain in the sample for an average of 6.8 years. 612 indi- viduals (20.2%) are initially observed in the public sector and are retained in the sample for 6.9 years on average. The remaining 206 men (6.8%) are initially unemployed and remain in the sample for an average of 6.3 years.
The ECHP includes a standardised education measure – the ISCED classification10 – coded into 3 categories: “high” is ISCED levels 5-7 and corresponds to all classes of tertiary level education, “medium” is ISCED level 3, corresponding to upper-secondary (post-compulsory)
5
In addition to those reporting themselves to be unemployed, the unemployment category includes: working unpaid in a family enterprise, in education or training (though having been in the labour market at some point), and doing housework, looking after children or other persons.
6
The ECHP variable pi211mg, which is in the national currency of each country.
7
We do not drop theseobservationsonly replace their earnings as missing. Therefore the individuals concerned still convey information to the sample and contribute to the modelling of the labour market dynamics.
8
Results using the UK sample in the ECHP, which is itself taken from the BHPS, concur with those found by Postel-Vinay and Turon (2007) when using a larger sample available in the British Household Panel Survey.
9
There is some sample attrition, 14.4% after 4 years and 47.3% at 8 years, which we assume to be exogenous. Some of the attrition is a consequence of our sample construction rules that treat individuals as censored from the first time they have a gap in their response history (this affects 6.5% of the individuals in our sample).
10
education, and “low” is ISCED levels 0-2, corresponding to levels of education up to the end of secondary schooling. In terms of education, the German sample breaks down overall as 27.2% high, 61.4% medium, and 11.3% low11. This masks some differences between the job sectors, the breakdown for the private sector being 25.4% high, 62.0% medium and 12.6% low, as compared with the public sector: 35.0% high, 59.8% medium and just 5.2% low. The public sector therefore attracting markedly more high educated and fewer low educated individuals. In terms of age and labour market experience, the public sector workers are on average a little older than those in the private sector (41.7 years old versus 39.4) and on average have slightly more experience12 (21.9 years versus 20.6). 4 5 6 7 8 9 10 11 12 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 year
Germany: unemployment rate (LFS)
15 17.5 20 22.5 25 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 year
Germany: public share of total employment (sample)
Figure 4.1: Germany: Unemployment rate, males, 1991-2001, and Sample public sector share
To illustrate the evolution of the German labour market over time, Figure 4.1 shows the male unemployment rate13for the years 1991-2001 (top panel), thus covering the years preceding our sample as well as the sample years themselves, and the public sector share of total employment for our sample (bottom panel). Both are reasonably stable over the time of our sample, the unemployment rate rising in the early nineties before levelling out at around 9% from 1994
11
See appendix C.1 for a tabulation of the education breakdown for each country.
12
‘Labour market experience’ or more accurately ‘potential labour market experience’ is defined as current age minus the age when the individual first entered the labour market.
13
onwards. The public sector share of total employment falls slowly from around 23% at the start of the sample to just under 20% by 2001.
Differences in Earnings
Earnings levels. We now illustrate public-private differences in earnings through a number of simple regressions, see Table 4.1. For each country we will be looking at (log) current gross monthly earnings. Differences in monthly work hours for full-time workers could lead us to understate any positive public premium in hourly wages and differences in cross sectional wage variances. However, in Germany the public and private sectors have very similar hours distri- butions – median weekly work hours is 40 for each – though with the public sector exhibiting less variance in hours (standard deviation of weekly hours is smaller by around one hour and a quarter).
The first column of Table 4.1 shows that the raw public pay gap in our sample is 5.2 log points (around 5.3%) in favour of the public sector. However this positive premium appears to be driven to a large extent by selection. When controlling for a quadratic in potential labour market experience and education band (column 2) the public premium falls to 1.4 log points and is statistically insignificant. Moreover, allowing the effects of education and experience to differ between the sectors (column 3), results in a statistically significant negative public premium of 18.8 log points (17.1%). This negative public premium is reduced for the medium educated compared with the high or low educated and also is reduced in experience. Allowing for fixed individual effects14 in a specification conditioning on just the quadratic in experience (column 4) reduces the size of the negative public premium and it is not significant. However introducing the interactions of the public sector dummy with education and the quadratic in experience in the fixed effects model (column 5) finds the public premium to be−13.2 log points (−12.4%) and significant. This is in line with the findings of Dustmann and van Soest (1998) who consistently find a significant negative public premium, reducing in age (experience), and robust to various modelling assumptions. Believing the final (column 5) fixed effects specification, we conclude that the returns to experience are greater in the public sector, such that the public premium traces an inverted U-shape in experience with a maximum estimated at 25 years. For levels of
14
The reported fixed-effect regressions use the within-estimator. First differenced estimates are very similar for Germany.
experience between 16 and 34 years there is in fact a small positive public premium.
Earnings dispersion. The standard deviation of log earnings in the public sector (0.324) is smaller than in the private sector (0.358), while the 90:10 percentile ratio of raw earnings are 2.323 and 2.522 respectively. Conditional on age and education these ratios are 2.129 (public) and 2.394 (private). These figures indicate a greater degree of wage compression in the public sector.
Earnings mobility. The regression results and the analysis of the earnings distributions show that the public and private sectors differ in terms of their earnings levels and the cross- sectional distribution of earnings. Moreover, there are differences in terms of earnings mobility, as illustrated in the upper panel of Table 4.2 which shows the transition matrices between quintiles of the unconditional log earnings distribution from one year to the next for the public and private sectors respectively.15 The matrices show that there is greater persistence in earnings rank amongst those individuals continuously employed in the public sector than amongst those continuously employed in the private sector. To further describe the persistence in earnings levels, the lower panel of the Table contains the corresponding transition matrices for the rank of earnings residuals after conditioning on education and a quadratic in (potential) labour market experience. Again we see the greater level of persistence of earnings rank in the public sector.
Bearing in mind that the earnings distribution is more spread in the private than the public sector, transitions between earnings quintiles in the private sector represent larger changes in earnings than similar transitions in the public sector, further underscoring the greater mobility of earnings in the private sector. In addition, we computed the 1-lag auto-covariance of normalised log earnings for each individual, having conditioned on education and a quadratic in (potential) labour market experience.16 For individuals employed in the private sector in consecutive periods the mean auto-covariance is 0.811 whereas the corresponding figure for the public sector it is higher at 0.944, again illustrating the greater persistence in earnings in the public sector.
15
The transitions relate to individuals who are employed in the same sector in yeart-1and yeart. It is possible to look at longer lags in our sample (up to 7 lags) however the numbers of individuals who are continuously employed in either sector at longer lags is not sufficiently large to allow robust inference.
16
We constructed normalised log earnings by first regressing log earnings (yit for individualiat datet) on the
covariates from the column 2 specification in Table 4.1, thus obtaining a predictor of mean earningsybit. We then
regressed the squared residuals from this latter regression on that same set of covariates to obtain a predictor of earnings varianceσb2
it. We then constructed earnings disturbances as (yit−byit)/bσit. Following this we calculated
the one-period auto-covariance between these disturbances, and compare the average for those in the public sector for both periods with those in the private sector for both 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 94.1 1.2 4.5 Public 6.6 91.1 2.1 Unemp. 36.8 5.2 57.8
It is clear that movement directly from the private sector to the public sector is very uncom- mon: only 1.2% of individuals initially employed in the private sector move to the public sector the next year; however movements in the opposite direction are more frequent (6.6%). The annual transition rate into unemployment from the private sector (4.5%) is more than double the corresponding figure for the public sector (2.1%), suggesting more job losses from the private sector.17 This greater security of employment in the public sector is also reflected in the differing probabilities of ever being observed in unemployment depending on observed initial sector. Of those first observed employed in the private sector 20.5% are recorded as unemployed at least once in their following observations, compared with only 7.8% for those initially employed in the public sector.
Of those unemployed, 57.8% remain unemployed in the next year, 36.8% gaining employment in the private sector, while just 5.2% of those unemployed in year t-1 are in public sector employment in year t. Of the 707 individuals (23.4% of the sample) ever observed unemployed during the time-span of the sample, 23.9% report being unemployed for 3 or more consecutive interviews during the sample. The 1-year re-employment probability for these “long-term” unemployed is considerably worse at 13.5% than it is for the “short-term” unemployed who have a 70.4% chance of finding a job in the following year. Thus unemployment persistence is quite high overall and there is some evidence that it is concentrated on the “long-term unemployed”. As well as differential re-employment rates according to whether the individual is a “long- term unemployed” type, there is also evidence of attachment to the public sector: an unemployed individual who has most recently been employed in the public sector is much more likely to
17
These are not pure job loss rates as voluntary quits are included in movements from employment to unem- ployment.
find employment in the public sector in the next year than an individual who’s most recent employment was in the private sector. Of the unemployed who were most recently in the public sector 8.3% return to public sector employment the next year, whereas for those unemployed who were most recently in the private sector, only 4.3% find public sector employment in the next year.