Content uploaded by Christian Grund
Author content
All content in this area was uploaded by Christian Grund on Nov 24, 2014
Content may be subject to copyright.
Discussion Paper 4/2004
Severance Payments for Dismissed Employees
Severance Payments for Dismissed Employees in
Germany
by
Christian Grund
March 2004
The Bonn Graduate School of Economics is
sponsored by the
Severance Payments for Dismissed Employees
in Germany
CHRISTIAN GRUND
March 2004
Abstract
This contribution investigates severance payments for dismissed employees in Germany.
Particularly, it responds to the following questions: Who receives severance payments? By
which characteristics is the level of severance payments determined? Is overcompensation to
be considered a relevant issue? Hereby, individual and collective dismissals are distinguished.
This is the first study on this issue using individual representative data – the German Socio-
Economic Panel – and multivariate methods. The results indicate that rather women, persons
with many years of tenure and working in big firms receive severance payments. There is a
huge variance in the size of the payments, which can only partly be explained by tenure, the
wage and citizenship. About one quarter of dismissed employees is better off in their
following careers independent of having received a severance payment.
Key words Severance Payments; Dismissals; Plant closings, Dismissal Protection
JEL Classification Codes M51, M52, J65, J32, J53
Author
Dr. Christian Grund
University of Bonn
Department of Economics
BWL II
Adenauerallee 24-42
D-53113 Bonn
Phone: ++49-228-739213
E-mail: christian.grund@uni-bonn.de
1
Severance Payments for Dismissed Employees
in Germany
Abstract
This contribution investigates severance payments for dismissed employees in Germany.
Particularly, the following questions are addressed: Who receives severance payments? By
which characteristics is the level of severance payments determined? Is overcompensation to
be considered a relevant issue? Hereby, individual and collective dismissals are distinguished.
This is the first study on this issue using individual representative data – the German Socio-
Economic Panel – and multivariate methods. The results indicate that rather women, persons
with many years of tenure and working in big firms receive severance payments. There is a
huge variance in the size of the payments, which can only partly be explained by tenure, the
wage and citizenship. About one quarter of dismissed employees is better off in their
following careers independent of having received a severance payment.
Key words Severance Payments; Dismissals; Plant closings, Dismissal Protection
JEL Classification Codes M51, M52, J65, J32, J53
1. Introduction
In 2000 the British Vodafone Group acquired the German Mannesmann AG, what resulted in
a redundancy of Mannesmann CEO Klaus Esser, inter alia. However, the financial
consequences for Klaus Esser were not too bad: he received a totalling severance payment of
over € 30,000,000. This ‘golden parachute’ was approved by the board of directors and
justified with Esser’s success in increasing the shareholder value of the Mannesmann AG (see
Financial Times Deutschland 2000).
Obviously, there are other rules to define whether ordinary employees other than top
managers receive severance payments in the case of dismissals and how to fix their size.
Apart from the above mentioned prominent case, astonishingly little is known about
severance payments for individually or collectively dismissed employees in Germany. The
legal framework is not very explicit with respect to this issue and there are only few and not
very updated studies, which concentrate on decisions of labour courts and severance
payments in connection with social plans in firms with mass layoffs (see Falke et al. 1981,
Falke 1983, Hemmer 1997a).
2
The aim of this study is therefore to analyse the relevance of severance payments in Germany.
Both, the probability and the size of severance payments are examined. We make use of the
German Socio-Economic Panel (GSOEP), a large representative data set, which has not been
consulted for this issue before.
This contribution does not focus on the effects of severance payments. See instead Burda
(1992) for a theoretical analysis or Lazear (1990), Addison and Grosso (1996), Addison et al.
(2000) and Hunt (2000) for empirical investigations of the effects on labour market outcomes.
Kodrzycki (1998) and Pencavel (2001) examine the effect of severance pay on individual
behaviour empirically.
Some contributions have already investigated the size of severance payments to dismissed
employees on an aggregated level. Within a wider context, several contributions analyse
adjustment costs – differentiated in hiring and firing costs. The results vary substantially
among countries and industries (see e.g. Burda 1991). For instance, Del Boca and Rota (1998)
estimate hiring costs of 2.0 to 2.6 months of labour costs and a huge range from 0.5 to 20
months of labour costs for firing costs in Italy. Hamermesh and Pfann (1996: 1280) and Hunt
(2000: 181) provide brief overviews of the relevant literature. The few studies dealing with
the size of individual severance payments in Germany are discussed in Section 2 below.
There are different theoretical economic approaches dealing with severance payments, which
are associated with this empirical study to some extent.
1
Fabel (2002) analyses the effect of
severance payments on the layoff decisions of firms and finds decreasing layoff rates for older
employees if severance payments increase with seniority. Suedekum and Ruehmann (2003)
focus on the effect of severance payments on the human capital accumulation of employed
workers. In general, investments in firm-specific human capital are strengthened by severance
payments. However, the penalty function of dismissals is weakened simultaneously. Goerke
1
The general seminal work dates back to Oi (1962), who characterised labour as a quasi-fixed factor and
examines – among other things – a rational order of dismissals at redundancies.
3
(2003) incorporates this point even more explicitly within an efficiency wage model assuming
that severance payments in the case of collective dismissals are also given to shirkers.
Galdon-Sanchez and Güell (2003) point out a double moral hazard problem in dismissal
conflicts within an efficiency wage framework. If firing costs are higher in the case of
redundancies compared to disciplinary dismissals, employers will always have an incentive to
claim each dismissal as a disciplinary one. On the other hand, workers have always an
incentive to claim unfair dismissal in order to receive a (higher) severance payment. If
information problems occur for a court, wrong decisions lead to the situation that some
redundant employees do not receive severance payments and some other fairly dismissed
employees by disciplinary reasons do receive severance payments.
Other authors provide bargaining models, where employees and employers ex ante bargain
about wages and the amount of severance payments. Hence, the determination of the size of
severance payments is in the spotlight of these contributions. For example, Booth (1995) und
Pita (1997) obtain a full insurance of the employees as a result. In this sense, full insurance
means constant marginal utility of the employees over the states. Fabel (1996) even finds
cases of efficient overcompensation of employees within his ex ante bargaining model with
labour cost minimising firms.
Malo (2000: 270f) points out that there is considerable empirical evidence of ex-post
bargaining about severance payments in most European countries as well. In this sense “ex
post” means that negotiations about severance payments begin subsequent to the perception of
a negative shock. He provides a model, which fitted especially the Spanish institutional
framework of individual dismissals. However, the German case is represented by ex-post
bargaining very often as well, because unions are responsible for wage bargaining industry
wide and works councils rather than unions can help employees to receive substantial
severance payments, when dismissals occur. In another paper Malo presents an ex-post
(Nash-) bargaining model representing collective dismissals in continental Europe (see Malo
2001). One of the important results is again an overcompensation of the dismissed employees.
Malo (2001: 84-86) and Fabel (1996: 592) try to find some evidence of overcompensation
4
that is in line with these theoretical considerations. Fabel argues on the basis of average
severance payments in Germany and Malo gives a numerical example, using aggregated
averages of severance payments and wages in Spain. A first direct empirical examination,
using individual data – missing so far –, is given in this contribution (subsection 3.4).
At least the ex-ante bargaining models implicitly presume an insurance function of severance
payments. Severance payments are supposed to act as insurance benefits against uncertain
employment conditions. In this sense, severance payments can be characterised as breach
penalties firms have to pay for breaking employment relationships (see Pita 1997). Hence, the
size of the severance payments has to be bargained ex ante. Another possible function of
severance payments is the welfare function. In this case, the underlying objective is the
reduction of financial disadvantages caused by the dismissal. It is therefore clearly future
based and ex post negotiations between employers and employees are necessary.
2
In the sense
of such a welfare function, severance payments can only be pushed through by legal
requirements or with employers having a sense of social responsibility. The latter possibility
is usually neglected in economic models. Relevance of this welfare function might result in
higher severance pay probabilities and sizes for persons faced by unemployment subsequent
to their dismissals and individuals with children under age or other obligations to pay
maintenance. The empirical relevance of the welfare function will be explicitly examined in
this study.
The paper proceeds as follows: Section 2 presents the legal framework with respect to
severance payments in Germany. Additionally, the results of earlier studies are summarised.
Section 3 comprises the empirical study on severance payments for dismissed employees
based on the German Socio-Economic Panel. We will respond in detail to the issues, who the
beneficiaries of severance payments are, which characteristics determine their levels, and
whether overcompensation is important. Individual and collective dismissals are kept separate
throughout the study. Section 4 concludes.
2
See Hemmer (1997a: 27-30) for a summary of different possible functions of severance payments.
5
2. Severance pay in Germany – legal framework and prior evidence
As mentioned above, German law neither lays down general severance payments in the case
of individual dismissals nor minimum severance payments in the event of collective
dismissals in Germany as opposed to other European countries like France, Spain or the UK.
3
Obviously, an empirical examination of the receipt and size of severance payments is
particularly interesting for countries without explicit laws, which strictly define recipients and
size of severance pay. In spite of the absence of mandatory regulations, severance payments
in connection with dismissals can be observed in Germany as well. Different types of
severance payments can be distinguished (see e.g. Inhoffen and Müller-Dahl 1981). The first
type is based on the Protection Against Dismissal Act (“Kündigungsschutzgesetz”). Until
December 2003 it was applicable for firms with more than five employees and employees
with more than six month of tenure.
4
Severance payments can be the outcome of dismissal
protection claims if ordinary dismissals are socially unjustified or extraordinary dismissals are
causeless.
5
In these cases severance payments are arranged if one party makes the application
to cancel the employment relationship (although the dismissal was socially not justified),
because a further co-operation between the employer and the employee cannot be expected.
A court decision is not necessary for a severance payment. Frequently the parties come to an
agreement with the help of a court without an official decision and in many cases employees
and employers agree upon a certain amount of severance pay without using a court at all.
They may anticipate the costs and the uncertain outcome of claims and are both better off – in
terms of expected utility – with a mutual agreement. Additionally, many firms worry about a
3
See Küchle (1990),the OECD (1999) and Bertola et al. (1999) for an international comparison of legal
frameworks. See Eger (2002) for a more detailed description of the legal framework for dismissals in Germany
compared to the USA.
4
In December 2003 dismissal protection has been lowered for firms up to 10 employees. Since the empirical
study of this contribution covers the years 1991 to 2002, it is not faced by this reform. However, from October
1996 to January 1999 the threshold of 10 employees did also exist.
5
An ordinary dismissal has to be justified with a misconduct of the employee (e.g. repeated late arrivals), lack of
capability (e.g. repeated or long term illness) or redundancies of the firm (§1 Protection Against Dismissal Act).
An additional requirement for a socially justifiable dismissal is that there is no possibility to employ the worker
on a different position or plant in the firm. Misconduct of the employees is the reason for the majority of cases
(see Falke 1983: 19). See Kittner and Kohler (2000) for a detailed description and discussion of the German
legal position in comparison to the USA. Additionally, Emerson (1988: 808-811) and the OECD (1999) provide
a summary of international legal frameworks regarding individual and collective dismissals.
6
loss of reputation if conflicts concerning dismissals become public knowledge.
6
Actually, the
share of dismissal protection claims with respect to all dismissals was only about eight
percent at the beginning of the 1980s (see Falke 1983: 27).
7
This value increased up to 27
percent in 2001, when German labour courts decided on more than 250,000 dismissal
protection claims. Thereby, 4 of 5 court decision are in favour of the employee (see Jahn and
Schnabel 2003).
A different, second type of severance payment can occur in the context of operational changes
in connection with mass dismissals and a social plan.
8
The German Works Constitution Act
(“Betriebsverfassungsgesetz”, §§ 112-113) defines that a social plan, as a result of the
coordination of the interests of the management and the works council, shall allay the
economic disadvantages of the dismissed employees, in particular earnings decreases. The
German Work Constitution Act aims at firms with at least five employees. A social plan can
be conducted with a minimum number of six dismissed employees.
9
Obviously, the function
of a social plan for the firms is somewhat different. From the firms’ point of view personnel
costs are supposed to be reduced with the help of dismissals and dismissal protection, as well
as the necessity of social plans, acts as a restriction of the firms’ decision-making authority.
Severance payments from dismissal protection claims or settlements can be credited against
payments from social plans. Both types of severance payments are paid as a lump sum and
not on a weekly or monthly basis for example. The entitlement to possible unemployment
benefits is not affected by the severance pay in general.
The size of severance payments is not clearly determined by law, either. The Protection
Against Dismissal Act denominates only maximum limits at the amount of 12 monthly wages
in general and 15 or rather 18 monthly wages for older employees with many years of tenure.
6
That is why an increasing number of firms authorise some kind of “dismissal consults”, who teach managers
how to dismiss employees in order to avoid conflicts when dismissals become necessary (see Schrep 2003).
7
This fraction was somewhat higher, if the dismissals were justified with the behaviour or the person (0.1) than
in cases in which the dismissals were justified with operational reasons (0.04).
8
Heseler and Mückenberger (1999) provide a comprehensive overview of the management of redundancies in
Germany.
9
In detail, a social plan has to be conducted if 20 percent of the workforce or at least six employees are faced
(firms with less than 60 employees), 20 percent or at least 37 affected employees (firms with less than 250
employees), 15 percent or at least 60 affected employees (firms with less than 500 employees) and 10 percent or
at least 60 affected employees in firms with at least 500 employees.
7
But there is scope of discretion of the courts, which is endorsed by juridical commentaries on
the law, in order to establish an “equitable” severance payment in every single case (see
Ascheid et al. 2000: 631ff.). Hence, the size of severance payments should also be determined
by future prospects next to past employment characteristics. Apart from court decisions, it is
possible that employers and employees make a settlement, which can be coordinated with the
help of a judge, though. In this case, an empirical formula, such as a severance payment to an
amount of half a monthly wage per year of tenure, is argued to be well established.
Characteristics next to tenure and the previous wage that can play a role in court decisions
encompass the age
10
, the amount of social adverseness, the economic situation of the firm, the
expected or actual unemployment duration of the dismissed employees and maintenance
obligations (see Dorndorf at al. 2001: 602-604).
The size of severance payments in the case of collective dismissals using social plans is not
fixed either. Some kind of „severance pay formula“ is usually used to calculate the individual
payments for the affected employees. Frequently these payments are increasing with tenure,
previous wage and age. But other individual characteristics may be relevant as well.
Additionally, the jurisprudence indicates that big companies have to pay more on average
because of higher financial opportunities (Inhoffen and Müller-Dahl 1981: 1474).
Only little is known about the size of individual severance payments in Germany up to now.
On the basis of a firm sample Falke et al. (1981) find severance payments of DM 7,149 (€
3,655) on average. But the amounts are very unevenly distributed with few very high
payments. The median is about DM 2,700 (€ 1,380) in this early study. Data of labour courts
decisions in Germany at the end of the seventies show that severance payments are arranged
at DM 2,000 (€ 1,023) on average (median) at first instance (see Falke 1983). By dividing the
severance payments by monthly wages (in DM) and tenure (in years) it is possible to calculate
severance pay factors, which are slightly above 0.5 on average. This means that the labour
courts adjudge half a monthly wage per year of tenure to dismissed workers.
10
Usually, the amount of severance payments should increase with age. But it can also decrease for older
employees, who attain mandatory retirement (65 years) in the nearer future.
8
In a more recent study, Hemmer (1997a) examines the level of severance payments in the
context of social plans. The data contain over 100 firms, which are faced with a social plan.
The financial burden of the firms with social plans account for 4% of personnel expenses or
1.3% of total revenue in 1995. The mean severance payment per worker within the social
plans of his sample increased from DM 13,360 (€ 6,830) in 1985 to DM 19,900 (€ 10,175) in
1995 (see Hemmer 1997a: 113). Hemmer considers possible individual determinants like
wages and tenure for the size of severance payments. Unfortunately, he is not able to analyse
these features due to lack of data. The relevance of the different possible functions of
severance payments has not been examined until now, neither. Remember that higher
(probabilities of) severance payments are expected for persons with maintenance obligations
and those faced by unemployment subsequent to their dismissal, if a welfare function of
severance payments is empirically relevant.
3. Empirical Study
As it is shown in the previous section, only court decisions and the outcomes of social plans
have been analysed in very few studies so far and additionally most of the data are somewhat
outdated. Court decisions only account for a small fraction of dismissal protection claims (see
Kittner and Kohler 2000: 27). Up to 90 percent of the cases end up with a mutual agreement
between the employer and the employee without a court decision. Hence, most of the relevant
cases of severance payments have not been analysed before. The study presented in this
section will use an individual German data set and integrate all kind of severance payments.
There are hardly any comparable international studies either. To our knowledge, only one
study analysed determinants of received severance payments and the amount of severance pay
with individual data and multivariate methods so far. Kodrzycki (1998) matches a
Massachusetts displaced workers data set with information on severance benefit plans of 15
employers with mass layoffs, which result in some 2,000 observations. She investigates that
the probability of receiving a severance payment after a dismissal increases with job tenure,
9
tends to increase with age and is higher in firms of the manufacturing sector. The size of
severance pay – measured in the number of weekly wages
11
– also increased with tenure as
well as with former annual earnings, and is below average in the manufacturing sector (see
Kodrzycki 1998: 67). The shortcoming of this study is obviously that the data are not
representative and only employees of a few firms faced by collective dismissals are taken into
account.
3.1 Data
This study is based on data from the German Socio-Economic Panel (GSOEP), a yearly
representative sample survey of people living in Germany.
12
The sample of this study is
restricted to persons, who were affected by an individual or collective dismissal (including
plant closings) within the observation period 1991 to 2002. Information for the reason of a
separation is only given in the GSOEP from 1991 on. The sample includes only individuals,
who were full-time employees before the dismissal (blue and white collars). Thus, building
the data set requires matching information of the individuals of two subsequent years. We
have information about the severance payments and the employment status of the present
years in addition to the information of the individuals’ jobs in the previous year. It yields a
sample size of 2,534 dismissed employees. In the years 1991 to 1998 and 2001 to 2002 the
questionnaire of the GSOEP distinguishes between individual and collective dismissals.
Hence, the observations of the years 1999 and 2000 are missing in each separated analysis,
which lead to 1,452 persons faced by individual dismissals and 769 by collective dismissals in
the data set. The fraction of the latter group decreases from 40 to 30 % during the observation
period. A severance payment is received by 27 percent of all individuals in the sample,
whereby this fraction is somewhat higher for collectively dismissed employees (0.33) than
individually dismissed employees (0.26) (see Table 1 and 2). Hence, more than two of three
11
The form of severance payments varies among employers in Massachusetts. Some are disbursed weekly and
others as a lump sum (see Kodrzycki 1998: 44). Hence, the system differs slightly from severance payments in
Germany, where generally a lump sum is paid.
12
The data is available from the German Institute of Economic Research (DIW), Berlin.
10
dismissed employees in Germany have not received a severance payment until the date of the
survey.
13
This study has got the following objectives: First, the receipt of severance payments is
examined. In a second step we analyse the subgroup of employees who received severance
payments and ask for the determinants of their size. An integral part is the analysis of the
relevance of a welfare function of severance payments. Additionally, we have a closer look on
the variance of the payments and on the relevance of overcompensation.
In order to investigate these issues, we make use of several other variables of the GSOEP as
well. At first, there are individual characteristics as sex, age, years of schooling, employment
status, citizenship, tenure, marital status and children in the household, for example.
Secondly, job based characteristics such as firm size and branch of industry are taken into
account as well.
There are only insignificant differences in the descriptive statistics of the persons faced by
individual and collective dismissals with respect to most of the variables (see Table 1). Two
thirds in the sample are male; nearly half of them have at least one child under age in the
household and about 0.6 are blue collar workers. More than 80% are German, and half of
them are from East Germany. Considerable differences can be pointed out as to observed
unemployment spells subsequent to the dismissal and for years of tenure. Individually
dismissed employees are more likely to be faced by unemployment (0.57 as compared to 0.4
for collective dismissals).
14
Tenure is much higher for collectively dismissed persons (10
years compared to 6 years), even though the difference as to the average age is less than two
years.
13
However it is important to take into account that the sample include 10 percent of observations with tenure not
greater than half a year. These are employees, who are dismissed during their probation period and are not
covered by the Protection Against Dismissal Act. Excluding these observations has no qualitative effect on the
following results.
14
This difference is highly significant based on a T-Test. Hence, it supports the “lemon-story” of Gibbons and
Katz (1991). They derive theoretically and find empirical support for the hypothesis that individually dismissed
employees have lower productivity and have therefore worse future labour market outcomes in terms of
unemployment spells and wages.
Table 1: Descriptive statistics of individual and collective dismissals
WHOLE SAMPLE INDIVIDUAL DISMISSALS COLLECTIVE DISMISSALS
All
(n=2534)
Persons
with
severance
payments
(n=689)
Persons
without
severance
payments
(n=1845)
All
(n=1452)
Persons
with
severance
payments
(n=379)
Persons
without
severance
payments
(n=1073)
All
(n=769)
Persons
with
severance
payments
(n=253)
Persons
without
severance
payments
(n=516)
Severance Payment 0.272 1 0 0.261 1 0 0.329 1 0
Sex (male) 0.673 0.617 0.694 0.674 0.588 0.704 0.651 0.628 0.663
Age (years) 39.48 42.19 38.46 38.90 41.92 37.84 40.77 42.34 39.99
Years of schooling 11.41 11.57 11.35 11.31 11.58 11.21 11.67 11.57 11.72
Marital status (single) 0.258 0.155 0.297 0.280 0.177 0.316 0.203 0.119 0.244
Child in household 0.422 0.415 0.425 0.419 0.422 0.418 0.446 0.431 0.453
Unemployed 0.521 0.496 0.530 0.565 0.562 0.566 0.402 0.387 0.409
Blue collar worker 0.616 0.557 0.638 0.624 0.528 0.658 0.572 0.593 0.562
Tenure (years) 7.30 12.16 5.49 6.19 11.12 4.46 9.94 13.98 7.96
Firm size:
1 - 5 employees
6 - 19 employees
20 - 199 employees
200 - 1999 employees
≥2000 employees
0.116
0.247
0.347
0.189
0.101
0.025
0.122
0.351
0.328
0.174
0.150
0.293
0.345
0.138
0.074
0.112
0.249
0.343
0.191
0.105
0.016
0.116
0.338
0.351
0.179
0.145
0.296
0.345
0.134
0.079
0.111
0.207
0.356
0.212
0.114
0.036
0.099
0.375
0.316
0.174
0.147
0.260
0.347
0.161
0.085
Region (West Germany) 0.519 0.507 0.523 0.521 0.517 0.522 0.503 0.478 0.516
Citizenship (German) 0.836 0.830 0.838 0.820 0.815 0.822 0.850 0.846 0.853
12
3.2 Who receives severance payments?
As mentioned above, there are no mandatory severance payments in connection with
dismissals in Germany and within this sample less than one third of the dismissed employees
receive a severance payment. Hence, the question arises, whether there are any individual or
job based determinants for the receipt of severance payments. A first possibility to analyse
this issue is to look at the descriptive statistics of possible determinants, separated for persons
with and without severance payments (see Table 1). Additionally, Table 2 presents the
percentages of persons with severance payments in the corresponding subgroups of the
sample.
Several variables are inspected for both, the answer of the questions who gets severance
payments and which characteristics determine the size of severance payments (subsection
4.3). Tenure, age and the previous wage are mentioned in the Protection Against Dismissal
Act and are well known factors of social plan practise as well. Hence, we can expect a
positive correlation between these variables and severance payments. Additionally, firm size
is obviously important, since the Protection Against Dismissal Act and the Works
Constitution Act are not valid for very small firms. Apart from that, the economic situation of
the firms regularly influences court decisions (see Dorndorf et al. 2001: 603), which may
enlarge severance payments for huge firms, if they have better financial opportunities. If firms
account for the individual situation of the dismissed employees, underage children in the
household and an actual unemployment status following the dismissal should enlarge the
probability and the size of severance payments. Significant effects of these variables would
confirm the relevance of a welfare function of severance payments as discussed above. There
might be also differences between sex, citizenship and occupational status due to some kind
of statistical discrimination. If groups of employees differ in the information degree
concerning dismissal protection – and therefore in the probability to file a suit –, firms can try
to dismiss the uniformed groups (e.g. foreign employees) without (or with lower) severance
13
payments. Last but not least, regional and industry distinctions as well as the year of the
dismissal are taken into account.
The univariate results indicate that a receipt of severance payments is less likely for males,
younger employees, singles, employees with less schooling and tenure, blue collars and
employees working in small firms. There are significant differences in the means of these
variables between the persons with and without severance payments.
15
The most noticeable
differences occur especially as to the firm size. Only 6 percent of dismissed workers in firms
with not more than five employees, but nearly half in larger firms with more than 200
employees get severance payments. Additionally, differences between the corresponding
percentages of men (0.25) and women (0.32) as well as of singles (0.16) and other marital
status (0.31) are remarkably high as well (see Table 2). Table 1 also shows a considerable
influence of tenure. The average tenure of persons with severance payment (12.2 years)
comes up to more than twice the amount of persons without severance payments (5.5 years).
No significant differences can be found as to the citizenship, the region of the workplace, the
unemployment status subsequent to the dismissal and the fact that there are children in the
household of the dismissed persons. The results for individual compared to collective
dismissals are similar, but not identical. The most striking result for both groups is that the
probability of severance payments increases with the firm size. However, the results for sex
and blue collars not differ significantly for collective dismissals.
Additionally, there might be industry effects that explain differences in receiving severance
payments. Table 3 shows that dismissals are more common in the sectors farming,
manufacturing and construction. The share of persons in these industries within the sample is
much higher than within the whole German workforce. In contrary, dismissals are not so
widespread in the service industries. The fraction of persons with severance payments differs
considerably between the industries as well. Especially people who are employed in the
manufacturing industry have high probabilities of severance payments in the case of
dismissals (0.39). Opposed to that, the fraction is very low for the sectors
farming/forestry/fishery and construction (0.15 each).
15
This is tested with a usual T-Test on identical averages.
14
Table 2: Percentage of persons with severance payments in subgroups
Whole sample Individual dismissals Collective dismissals
Number
of obser-
vations
Percentage
of persons
with
severance
payments
Number
of obser-
vations
Percentage
of persons
with
severance
payments
Number
of obser-
vations
Percentage of
persons with
severance
payments
Whole sample 2534 0.272 1452 0.261 769 0.329
Men
Women
1705
829
0.249
0.318
978
474
0.228
0.329
501
268
0.317
0.351
Single
Others
655
1879
0.163
0.310
406
1046
0.165
0.298
156
613
0.192
0.364
Child in household
No child in household
1070
1464
0.267
0.275
609
843
0.263
0.260
343
426
0.318
0.338
Unemployed
Not unemployed
1319
1215
0.259
0.286
820
632
0.260
0.263
309
460
0.317
0.337
Blue collars
White collars
1562
972
0.246
0.314
906
546
0.221
0.328
440
329
0.341
0.313
Firm size:
1 - 5 employees
6 - 19 employees
20 - 199 employees
200 - 1999 employees
≥2000 employees
294
625
879
480
256
0.058
0.134
0.275
0.471
0.469
162
362
498
277
153
0.037
0.122
0.257
0.480
0.444
85
159
274
163
88
0.106
0.157
0.347
0.491
0.500
West Germany
East Germany
1314
1220
0.266
0.279
756
696
0.259
0.263
387
382
0.313
0.346
German
Foreigner
2118
410
0.270
0.281
1191
261
0.259
0.268
654
115
0.327
0.339
Table 3: Severance payments in different industries
Share of employees
in the whole
German workforce
in 2002
(N=27,2 mill.)
Share of
dismissed
persons in the
sample
(n=2534)
Fraction of persons
in industry in sample
with severance
payments
(n=689)
Farming, Forestry, Fishery 0.013 0.032 0.150
Manufacturing 0.232 0.364 0.393
Construction 0.058 0.208 0.146
Service industries 0.697 0.397 0.237
Σ 1 1 0.272
15
Another possible, and plausible, determinant that influences the probability of receiving
severance payments is the general economic situation. Indeed, the correlation between the
yearly fraction of dismissed persons, who received severance payments, and the growth rate
of the German GDP is considerably high during the observation period 1991 to 1998 (0.736;
p=0.010; see Figure 1 for an illustration and Table A in the appendix). The correlation is even
more obvious for individual dismissals (0.775; p=0.008) than for collective dismissals (0.414;
p=0.066). Hence, we observe a higher probability of severance payments in years of high
economic growth. Explanations for this finding are speculative. From an economic point of
view, firms could be forced by (the anticipation of) court decisions, which may take the firms’
economic situation into account, when judges decide on severance payments in dismissal
protection claims. Hereby, the firms’ economic situation is highly correlated with the general
economic trend. Ichino et al. (2003) report that local labour market conditions indeed
influence court decisions at dismissal protection claims in Italy. This explanation cannot be
separated from possible social motives of firms, whose ability to pay is higher in
economically good years.
Figure 1: Fraction of dismissed persons with severance payments and ∆GDP in
Germany (1991-2002)
Note: No value for ∆GDP in 1991 because of the German re-unification in 1990.
0.1
0.2
0.3
0.4
0.1
0.05
0.025
0.075
Fraction of dismised employees
with severance payments ( )
∆GDP ( )
Year1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
16
But the results up to this point might change, if we use multivariate tools and control for
different possible determinants simultaneously. By making use of a binary probit approach we
examine the determinants of receiving severance payments. The results are listed in Table 4.
The dichotomous dependent variable has got a one in the case of persons with a severance
payment and a zero otherwise.
The results confirm that firm size increases the probability of severance payments
considerably. Hence, especially small firms with less than six employees, which are not faced
by the Protection Against Dismissal Act, do not compensate dismissed persons.
Unfortunately, it is not possible to control for the existence of works councils in the firms of
the dismissed employees. Partly, the firm size effect might rather be a works council’s effect,
since larger firms are more probable to have a works council and works councils act towards
severance payments in negotiations with the management.
Additionally, the probability of severance payments increases with tenure and is significantly
higher for women. The positive effect of tenure is in line with anticipated (or actual) court
decisions. The longer the relationship endures the more difficult is a dismissal to be socially
justifiable. The higher probability of severance payments for women – particularly after
individual dismissals – can at least partly be explained by the fact that dismissals caused by
misconduct of the employees are more common among males (0.67 of all dismissals)
compared to females (0.52, see Falke 1983: 24). It is a reasonable policy for firms to abstain
from severance payments for persons dismissed because of misconduct in order to keep the
threatening penalty of detected misconduct or shirking as high as possible.
The results for the industry dummies confirm the descriptive statistics. The probability for
severance payments is particularly high for employees of the manufacturing sector and very
low in the farming and construction sector. Possibly, differences between the power of the
unions of the single industries are responsible for this result. There is information for
individual union membership in the GSOEP in the years 1993 and 1998. Including a union
17
membership dummy in the regression for the sub sample of these years, no effect can be
found. But it may also be the case that powerful unions cause higher severance probabilities
in these industries for both members and non-members. The evidence is at least in line with
this consideration. The net union density
16
, defined as employed union members divided by
all employees, is particularly high in the industries with a high fraction of severance
payments. For example, the union density of the early 1990s amounts to 0.39 in chemistry,
0.43 in mining and even more in some other sub-sectors of manufacturing compared to only
0.17 in the construction sector (see Fitzenberger et al. 1999: 258f).
In regression (2) we take the consideration into account that the probability to receive a
severance payment might not be independent from the expected amount of the payment. If a
dismissed employee only expects to receive a marginal payment, he will probably not exert
much effort to get it. Therefore, we include the expected estimated severance payment as an
additional variable in the regression (see also regression 1 of Table 6 in the subsequent
subsection). Indeed, the estimated amount of the payment is highly correlated with the
probability to receive a payment. The results with respect to the other variables are not
affected by this modified specification, though.
In spite of the very different legal situation for individual and collective dismissals, there are
no significant differences with respect to the probability of payments (regression 3). However,
we can observe considerable differences between the influencing factors of severance
payments with individual and collective dismissals respectively (regression 4 and 5). For
individually dismissed persons there are higher probabilities of severance payments for
German citizens as compared to foreigners, for West German persons as compared to East
Germans and for employees with many years of schooling. Possibly, employers anticipate that
the probability of dismissal protection claims is lower for foreigners and less educated
employees, what could make it rational from their point of view to pay severance payments
16
Sometimes a gross union density, defined as all union members over all employees is mentioned as well. Since
there are several non-employed union members like students, unemployed and retired persons, this measure
overestimates the influence of the unions especially for shrinking industries like mining. Here, the gross union
density amounts to 0.7.
18
for German rather than for foreign employees. The same explanation may hold for differences
between West- and East Germany as well. Since employees (and employers) were not
perfectly informed about dismissal protection of employees in the first years after the German
re-unification in 1990, risks of dismissal protection claims might have been much lower for
East German firms.
Age, marital status, children in the household and unemployment spells subsequent to the
dismissal do not affect the severance pay probability significantly. Hence, neither the future
job nor individual social criteria are important determinants of receiving severance pay.
Apparently the employers do not care very much about the individual situations and future
career opportunities of the dismissed employees. There hardly seems to be any relevance of a
welfare function of severance payments as mentioned in the introduction above. The results
for the year dummies are not listed. In principal the descriptive results are confirmed. The
probability of severance payments differs across years cyclically.
So far we have pointed out the important determinants for the receipt of severance payments.
The probability of severance payments depends on both individual characteristics, like in
particular tenure but also sex and citizenship, and firm characteristics, like firm size and
industry. But nothing is said about the size of severance payments until now. The amount of
the payments might differ between different groups of employees as well, which would have
important monetary consequences for the affected employees as well as for the economic
situations of the firms. Thus, in the next section we analyse the subgroup of dismissed persons
with severance payments in more detail. We want to examine the determinants of the size of
severance payments. It seems interesting whether high probabilities of severance payments
come along with high sizes or, in contrary, whether probability and size act as substitutes.
19
Table 4: Binary probit regressions on received severance payments
Whole sample
(1) (2) (3)
Individual
dismissals
(4)
Collective
dismissals
(5)
Sex (male)
-0.213***
(3.08)
-0.312***
(4.29)
-0.240***
(3.32)
-0.266***
(2.86)
-0.210*
(1.73)
Age
-0.002
(0.50)
-0.005
(1.14)
-0.002
(0.59)
0.003
(0.64)
-0.011
(1.54)
Years of Schooling
0.029*
(1.85)
0.010
(0.61)
0.029*
(1.76)
0.038*
(1.76)
0.009
(0.33)
Marital status (single)
-0.240**
(2.50)
-0.232**
(2.41)
-0.223**
(2.16)
-0.102
(0.78)
-0.415**
(2.35)
Child in household
0.024
(0.35)
0.005
(0.07)
0.059
(0.78)
0.125
(1.29)
-0.054
(0.44)
Unemployed
0.067
(1.11)
0.088
(1.44)
0.098
(1.50)
0.088
(1.06)
0.091
(0.83)
Blue collar worker
-0.059
(0.74)
0.013
(0.16)
-0.028
(0.28)
-0.175
(1.63)
0.187
(1.39)
Tenure
0.040***
(10.3)
0.023***
(4.33)
0.042***
(10.1)
0.046***
(8.26)
0.038***
(5.73)
Firm size (base: 20-199):
1 - 5 employees
6 - 19 employees
200 - 1999 employees
≥2000 employees
-1.024***
(7.16)
-0.351***
(4.19)
0.302***
(3.83)
0.349***
(3.59)
-0.974***
(6.80)
-0.327***
(3.89)
0.282***
(3.56)
0.304***
(3.10)
-1.087***
(6.98)
-0.416***
(4.56)
0.308***
(3.71)
0.304***
(3.00)
-1.284***
(5.65)
-0.459***
(4.00)
0.384***
(3.60)
0.265**
(2.02)
-0.872***
(3.88)
-0.381**
(2.45)
0.166
(1.22)
0.391**
(2.34)
Industries (base: services):
Farming/Forestry/Fishing
Manufacturing
Construction
-0.597***
(2.97)
0.242***
(3.35)
-0.129
(1.34)
-0.568***
(2.83)
0.231***
(3.20)
-0.179*
(1.83)
-0.585***
(2.84)
0.222***
(2.91)
-0.172*
(1.66)
-0.921***
(3.18)
0.211**
(2.11)
-0.076
(0.58)
-0.138
(0.44)
0.280**
(2.28)
-0.249
(1.30)
Region (West Germany)
0.095
(1.30)
-0.069
(0.85)
0.113
(1.46)
0.194*
(1.93)
-0.006
(0.05)
Citizenship (German)
0.187*
(1.95)
0.160*
(1.66)
0.211**
(2.08)
0.216*
(1.70)
0.195
(1.12)
Collective Dismissal
___ ___ -0.026
(0.38)
___ ___
Estimated Severance Pay
$
___ 0.308***
(4.67)
____ ___ ___
Intercept
-1.491***
(4.76)
-3.412***
(6.58)
-1.539***
(4.67)
-1.985***
(4.70)
-0.724
(1.30)
Year Dummies Yes Yes Yes Yes Yes
Observations 2534 2534 2221 1452 769
Log-Likelihood -1180.46 -1169.75 -1049.09 -631.21 -399.31
Pseudo R² 0.204 0.211 0.209 0.243 0.180
Notes: Absolute t-statistics in parentheses. *, ** and *** indicate significance at the 0.10, 0.05 and
0.01 level.
$
: Estimated amount of severance pay in accordance with model 1 of Table 6 (= 6.519 +
0.0558 • Tenure + 0.00063 • Monthly Gross Wage.
20
3.3 By which characteristics is the size of severance payments determined?
As already mentioned above, there are some obvious determinants for the size of severance
payments. Very often tenure and the wage are used to determine severance payments. But
other characteristics might be important as well. For example higher severance payments for
persons with children in the household or faced by unemployment may hint to a relevance of
a welfare function of employers with social motives. Higher payments from big companies
may occur due to higher ability to pay.
Table 5 indicates that the size of severance payments is higher for men, persons with no child
in the household and other marital status than single, white collars, persons employed in big
firms, and in West Germany. But we can observe higher former wages and partly higher
tenure for these groups, too. Therefore, it is useful to calculate so called severance pay factors
(see Hemmer 1997a: 146), which are defined by:
years)(in Tenure €)(in gemonthly wa Gross
€)(in payment Severance
factorpay Severance
⋅
= .
The wages and severance payments of all years are deflated with the German consumer price
index with the base of the year 2002. A severance pay factor of 0.5 means a severance
payment of half a monthly wage per year of tenure. Model (1) of Table 6 shows that tenure
and the previous wage are indeed very important factors to determine the amount of severance
payments. Looking at the mean severance pay factors of the variables, we find a different
picture compared to the basic size of severance payments. Indeed, there are higher severance
pay factors for white collars and in West Germany. These results can be explained with the
worse economic situation of East-German firms and with a better relationship of white
collared employees to the management respectively. There are no considerable differences for
the other variables.
21
Table 5: Average amount of severance pay in subgroups
Whole Sample Collective Dismissals Individual Dismissals
Obser-
vations
Amount of
severance pay
Severance
pay factor
Obser-
vations
Amount of
severance pay
Severance
pay factor
Obser-
vations
Amount of
severance pay
Severance
pay factor
Whole sample 689 9243 0.480 253 10408 0.451 379 8617 0.503
Men
Women
425
264
10933
6522
0.473
0.492
159
94
11855
7959
0.436
0.475
223
156
10555
5847
0.508
0.495
Single
Other marital status
107
582
5999
9840
0.535
0.471
30
223
7652
10778
0.365
0.462
67
312
5398
9308
0.637
0.474
Child in household
No child in household
286
403
7931
10175
0.490
0.474
109
144
8595
11780
0.460
0.443
160
219
7731
9265
0.519
0.490
Unemployed
Not unemployed
342
347
8316
10157
0.454
0.506
98
155
9084
11245
0.451
0.451
213
166
8120
9255
0.462
0.554
Blue collars
White collars
384
305
7661
11235
0.396
0.587
150
103
8204
13617
0.366
0.574
200
179
7152
10254
0.415
0.600
Firm size:
1 - 5 employees
6 - 19 employees
20 - 199 employees
200 - 1999 employees
≥2000 employees
17
84
242
226
120
5792
4478
6491
11648
15588
0.234
0.418
0.462
0.464
0.628
9
25
95
80
44
7897
5643
7168
15069
12148
0.250
0.426
0.415
0.461
0.565
6
44
128
133
68
4225
4455
5642
9803
14977
0.209
0.381
0.489
0.479
0.678
Industries:
Farming/Forestry/Fishing
Manufacturing
Construction
Service industries
12
362
77
238
3502
10104
9176
8245
0.245
0.446
0.533
0.598
7
148
21
77
4086
11107
9992
9751
0.214
0.419
0.371
0.554
5
184
45
145
2685
9053
9706
7930
0.287
0.465
0.581
0.533
West Germany
East Germany
349
340
12057
6355
0.568
0.390
121
132
15238
5980
0.585
0.328
196
183
10463
6640
0.577
0.423
German
Foreigner
572
117
8982
10521
0.472
0.520
214
39
10010
12588
0.430
0.564
309
70
8552
8905
0.501
0.509
Note: Severance pay factor = amount of severance pay / (gross monthly wage
*
tenure).
22
Whereas the average amount of severance pay is higher for persons faced by collective
dismissals (€ 10,400) compared with individual dismissals (€ 8,600), things change as to the
average severance pay factor, which is higher for individual dismissals (0.5 versus 0.45).
Next to the comparison of the averages, the analysis of the distribution of severance payments
is useful. We can observe a huge variance as to the amount of severance payments as well as
to the severance pay factor (see Figure 2 and 3 as well as Table B and Figure A and B in the
appendix). Nearly a quarter of severance payments do not exceed 2,000 €. However, more
than 10 percent of severance payments in the sample come at least to € 20,000. This huge
span is not only caused by individual differences in the previous wage and/or tenure.
The distribution of the severance pay factors reveals considerable inequality as well. 10
percent of the dismissed persons receive a factor of at least 0.92, whereas the bottom 10% of
the distribution has to be content with a factor of at most 0.11. The inequality of severance
pay factors is even higher for the subgroup of collectively dismissed employees. The
corresponding values for the 10
th
and the 90
th
percentile are 0.09 and 0.96 respectively.
Figure 2: Distribution of severance payments
0
5000
10000
15000
20000
25000
30000
35000
40000
0
,05
0,
15
0,
25
0,35
0,
45
0,55
0,
65
0,75
0,85
0,
95
F(severance payments)
Amount of severance pay in
€
23
Figure 3: Distribution of severance pay factors
0
0,2
0,4
0,6
0,8
1
1,2
1,4
0
,05
0
,15
0
,25
0,
3
5
0,
4
5
0,
5
5
0,
6
5
0
,75
0
,85
0
,95
F(Severance pay factors)
Severance pay factors
Note: F(·) = Distribution function of severance payments (figure 2) respectively severance pay
factors (figure 3).
Despite the observed huge variance in the data, there are some significant determinants of the
amount of severance pay, what is shown by the regressions. Model (1) of Table 6 points out
the particular relevance of tenure and the previous wage. The full set of variables like in
subsection 3.2 is used in addition to the gross monthly wage in the previous job within a
simple OLS approach to explain the log of the amount of severance payments in model (2) of
Table 6. Making use of the same variables again, a comparison to the results of subsection 3.2
is possible. It turns out that tenure, wage in the previous job, age, region and firm size are the
significant factors within this specification. The other variables, as well as the year and
industry dummies, have no significant effect. Although the coefficient children under age in
the household has a positive sign, a significant “social bonus” cannot be confirmed in general.
The coefficient for an unemployment spell subsequent to the dismissal is even negative.
Hence, we cannot find clear empirical evidence for a welfare function of severance payments.
The most important determinant for severance payments is the wage in the previous job. The
average previous wage in the sample is € 1,987. Hence, the coefficient of 0.0006 means that
on average a doubling of the wage doubles the severance payment as well.
24
Additionally, the amount of severance payments increases with the size of the firm. The
received payments of employees from big firms with more than 2000 employees are 50
percent higher than in medium sized firms with 20 to 200 employees. This is particularly true
for individual dismissals. The result for region differs considerably between the estimations
for individual and collective dismissals as well (see models (4) and (5)). Significant
differences between West- and East Germany cannot be observed in the case of individual
dismissals. However, severance payments in the context of collective dismissals are 50
percent higher in West Germany as compared to East Germany. Subsequent to the German re-
unification many East-German firms were bankrupt and did not have the financial
opportunities to afford higher severance payments even if they would have wanted to.
Probably, the workers receiving severance payments are not a random selection of all
dismissed workers. In order to take into account a possible selection bias a Heckman selection
model (see Heckman 1976) is used, where the amount of severance payments and the
probability to get a payment are estimated in one common approach. The results of the Probit
(Selection) model coincide with the outcomes of the binary probit model in the previous
subsection. It turns out that indeed a sample selection occurs. Nevertheless tenure and the
previous wage are confirmed as the most important determinants, and the results for age
remains significant as well. However, the results for firm size change and the coefficient for
citizenship becomes significantly negative. Hence, taking the selection into account, the
amount of severance payments is smaller for foreign dismissed employees.
Recapitulating, the average severance payment in the German sample amounts to € 9,200 in
prizes of the year 2002. However, a huge variance can be observed, which can still be found,
when we look at severance pay factors defined as the individual severance payment divided
by the previous gross monthly wage and tenure. In fact, the previous wage and tenure are the
most important determinants of the size of severance payments. But age and citizenship are
relevant factors as well. In order to discuss the effects of severance payments for the
individuals, the reemployment rates have already been mentioned. If an employee, who
received a severance payment, finds a job immediately after her dismissal and possibly even
increases her wage, she is better off compared to a further employment in her previous firm.
This leads directly to the question of the relevance of overcompensation.
25
Table 6: Determinants of the amount of severance payments (Dependent variable: Ln(severance payment))
Whole Sample
Individual
Dismissals
Collective
Dismissals
Heckman Selection Model
(1) OLS (2) OLS (3) OLS (4) OLS (5) OLS Ln(severance pay) Prob(severance pay)
Sex (male) -0.090 (1.22) -0.094 (1.21) -0.054 (0.61) -0.118 (0.76) 0.073 (0.85) -0.186*** (2.75)
Age 0.012*** (2.82) 0.013*** (2.84) 0.019*** (3.36) 0.004 (0.47) 0.013*** (2.64) 0.002 (0.45)
Years of Schooling 0.005 (0.30) -0.010 (0.53) -0.023 (1.11) 0.005 (0.13) -0.023 (1.18) 0.035** (2.24)
Marital status (single) -0.163 (1.50) -0.109 (0.94) 0.045 (0.33) -0.363 (1.61) 0.026 (0.21) -0.223** (2.37)
Child in household 0.079 (1.04) 0.104 (1.30) 0.188* (1.93) 0.062 (0.44) 0.052 (0.60) 0.016 (0.23)
Unemployed -0.025 (0.40) -0.016 (0.24) 0.002 (0.02) 0.018 (0.14) -0.075 (1.02) 0.065 (1.09)
Blue collar worker -0.066 (0.80) -0.088 (1.01) -0.097 (0.95) -0.121 (0.74) -0.031 (0.32) -0.066 (0.85)
Tenure 0.0558*** (17.4) 0.046*** (11.5) 0.047*** (10.9) 0.045*** (8.93) 0.050*** (5.91) 0.024*** (4.74) 0.042*** (10.9)
Monthly Gross Wage 0.0006*** (19.6) 0.0005*** (11.4) 0.0005*** (10.7) 0.0005*** (9.36) 0.0005*** (4.97) 0.0006*** (12.5) ----
Firm size (base: 20-199):
1 - 5 employees
6 - 19 employees
200 - 1999 employees
≥2000 employees
-0.509** (2.43)
-0.158 (1.51)
0.200*** (2.61)
0.471*** (5.08)
-0.473** (2.08)
-0.106 (0.93)
0.226*** (2.78)
0.470*** (4.81)
-0.256 (0.77)
-0.010 (0.74)
0.230** (2.40)
0.599*** (5.15)
-0.629* (1.81)
-0.138 (0.62)
0.219 (1.47)
0.282 (1.59)
0.258 (1.12)
0.126 (1.10)
-0.035 (0.38)
0.207* (1.83)
-1.010*** (7.12)
-0.353*** (4.27)
----
0.261*** (3.36)
0.309*** (3.25)
Industry (base: services):
Farming/Forestry/Fishing
Manufacturing
Construction
0.109 (0.45)
0.069 (0.91)
-0.045 (0.40)
0.097 (0.39)
0.046 (0.58)
-0.037 (0.30)
-0.077 (0.22)
0.070 (0.74)
0.052 (0.38)
0.212 (0.55)
0.031 (0.21)
-0.130 (0.51)
0.506* (1.90)
-0.099 (1.13)
0.027 (0.22)
-0.623*** (3.12)
0.233*** (3.28)
-0.142 (1.50)
Region (West Germany) 0.161* (1.78) 0.153 (1.57) -0.011 (0.10) 0.436** (2.18) 0.034 (0.35) 0.075 (1.06)
Citizenship (German) -0.126 (1.23) -0.131 (1.21) -0.192 (1.51) 0.037 (0.18) -0.295** (2.51) 0.151 (1.61)
Collective dismissal
---- 0.025 (0.36) ---- ---- ---- ----
Intercept 6.519*** (79.2) 6.339*** (19.1) 6.328*** (18.0) 6.313*** (15.0) 6.385*** (9.88) 8.184*** (20.0) -1.514*** (4.93)
Year dummies No Yes Yes Yes Yes Yes Yes
Observations 689 689 632 379 253 689 2534
R²
adj.
= 0.494 R²
adj.
= 0.543 R²
adj.
= 0.534 R²
adj.
= 0.580 R²
adj.
= 0.456
ρ = -0.888*** (29.50)
λ = -0.998*** (11.79)
Notes: Absolute t-statistics in parentheses. *, ** and *** indicate significance at the 0.10, 0.05 and 0.01 level.
26
3.4 Is overcompensation to be considered a relevant issue?
In order to answer this question it is first of all necessary to define and create a measure for
overcompensation. One can speak of an overcompensated dismissed employee, if the present
value of future wages in addition to the severance payment exceeds the hypothetical present
value of the wage profile in the dismissing firm (see e.g. Fabel 1996). This definition cannot
be used with the data of the GSOEP. Therefore, we will speak of overcompensation, if a
person received a severance payment
and has a higher wage in her new job after the
dismissal. Certainly, this simple measure is imperfect, because on the one hand
overcompensation may also occur in cases with high severance payments and minor wage
reductions or moderate durations of unemployment.
17
On the other hand, wage increases may
be transitory. But because of the relevance of increasing wage profiles the latter argument
does probably not fit for many cases. Hence, the applied proxy for overcompensation might
be interpreted as a lower bound for actual overcompensation.
Indeed, a substantial fraction can be assigned to the category of overcompensated employees
even with this definition. More than one fourth of dismissed persons with severance payments
are reemployed and experience a nominal wage increase (see Table 7). This fraction drops by
4 points focussing on real wage increases. Based on the fraction of reemployed persons, even
more than half get a higher wage in addition to the severance payment. Persons with
severance payments are not worse off in their subsequent careers compared to dismissed
employees without severance payments. Hence, once again a particular relevance of a welfare
function of severance payments cannot be confirmed.
The reemployment rate is much higher in cases of collective dismissals. While more than 50
percent of the affected persons are reemployed in the next year, only slightly more than one
third of individually dismissed employees have a new job. However, the distribution of wage
17
Unemployed persons receive unemployment benefits amounting to 60 percent of their net monthly wage of the
last year, if they have been employed subject to social insurance contribution at least 12 months within the last
three years. This fraction increases to 0.67 if the person has at least one child. The duration of the claim –
depending on the duration of previous employment and the age – was between 6 and 32 months during the
observation period. Additionally, a reduced tax rate for severance payments can enlarge the relevance of
overcompensation.
27
increases within the subgroup of reemployed persons with or without severance payments is
very similar between individual and collective dismissals.
Table 7: Consequences of dismissals – reemployment rates and wage increases
Whole
sample
Individual
dismissals
Collective
dismissals
Number of observations (n) 2534 1452 769
Reemployed persons (share of n) 0.429 0.361 0.580
Persons with nominal wage increases (share of n) 0.261 0.213 0.374
Persons with real wage increases (share of n) 0.221 0.178 0.320
Persons with severance payments (sp) 689 379 253
Reemployed persons (share of sp) 0.419 0.359 0.526
Persons with real wage increases (share of sp) 0.270 0.223 0.338
Persons with wage increases (share of sp) 0.232 0.170 0.314
Overcompensated employees can be characterized by comparing the descriptive statistics of
Table C (see Appendix) with those of not overcompensated employees. Apparently,
overcompensated workers tend to be well educated white-collar workers, German citizens
from East-Germany with no unemployment spell subsequent to their dismissal. To sum up,
we can state that overcompensation is indeed a relevant phenomenon. Although in many cases
severance payments are required to finance times of non-employment, a lot of persons are
better off with their severance pay in addition to higher wages.
Finally, some hints due to shortcomings of the data have to be given. First, a legally required
period of notice is usually part of the employment contract after a common six-month
probation period. This period of notice starts with one month and is increasing with tenure up
to seven months for employees with tenure of more than 20 years. Sometimes dismissed
employees do not have to work the whole remaining time, but do get their wage until the
expiration of the contract. Hence, this continued pay after dismissals without duty to work
corresponds to severance payments. This effect cannot be captured with the data.
Additionally, not every court decision may have been made by the time of the survey of the
GSOEP. Although 80 percent of dismissal protection claims are finished within six months
and almost all after twelve months (see Franke 1996: 100), some dismissal protection claims
28
were probably not concluded at the date of the particular survey.
18
In sum, slightly more than
27 percent of dismissed employees may benefit from severance payments and the size of the
payments may be slightly underestimated as well. For a more detailed analysis it would be
helpful to have additional information on the reason of each dismissal and on whether the
employees took legal action.
4. Conclusion
In this empirical study we analysed severance payments for dismissed employees in Germany
in the 1990s. In particular we responded to the questions: 1.) “Who gets severance
payments?“ 2.) “How much do recipients get?“ and 3.) “Is overcompensation relevant?“ It
turned out that approximately one third of dismissed employees receive a severance payment
from their former employers. Tenure and firm size are the most important determinants with
respect to the receipt of a severance bonus. Additionally there are industry and business cycle
effects, though. The size of severance payments is slightly higher for collectively (€ 10,400)
than individually dismissed persons (€ 8,600). The most important determinants for the size
are tenure and the previous wage. However, age and citizenship do matter as well. About one
quarter of dismissed employees are better off in the future in the sense that they receive higher
wages in addition to a severance bonus. Although law in Germany lays down a welfare
function of severance payments, there is hardly any evidence for adequate empirical
relevance.
The huge variances in the results are worth mentioning. Hence, there is enormous uncertainty
for both, employees and employers about the bonuses or respectively the costs of dismissals
at least since severance payments are not fixed in an ex ante bargaining. Due to the
unspecified legal situation it seems to be beneficial for employees to insist on severance
payments and threaten with a suit in order to increase the chance of a substantial severance
18
Note that the survey is conducted at the mid of each year and the individuals provide information about the
whole last calendar year.
29
payment.
19
Taking this behaviour into account, employers might be better off by fixing a
(generous) amount of severance payment as an explicit part of the employment contract. This
might even reduce total wage costs, because risk premiums of risk averse employees can be
economised.
Focussing on collective dismissals it is stated by Hemmer (1997c: 130) that two of three firms
wish the legal framework of social plans to be improved. Apart from that, Hemmer (1997b:
102) points out that firms more and more often make use of alternative procedures of the
adaptation of staffing levels instead of dismissals, such as early retirement plans, training
programs, assistance to become self-employed, outplacement and so on.
There is much discussion about modifications of the German severance pay system. Some
authors propose a more explicit orientation of the size of severance payments on the economic
situation of the dismissed employees, which is in line with the German Work Constitution Act
(§112). In detail, Hemmer (1997c: 132f) suggests that the severance payment should increase
with future employment status. This might be an improvement in terms of equity, but neglects
harmful incentives for omitted job search.
An increasing number of politicians as well as researchers discuss the possibility of
introducing mandatory severance payments for dismissed employees in Germany for
macroeconomic reasons as well (see e.g. IZA 2002, Jahn 2002, Handelsblatt 2003). It is often
argued that the uncertain legal situation leads to a reduction of recruitments. Concrete
suggestions include an annulment of dismissal protection within the first years of tenure in
favour of a mandatory severance pay of e.g. one monthly wage per year of tenure. Intuitively,
such a kind of modification of the legal situation concerning severance payments seems to be
reasonable. However, Malo and Perez (2003) recently address the problem of moving from an
unknown severance pay situation to a known severance pay one theoretically and find
ambiguous effects on severance pay and expecting firing costs. The examination of this issue
is an exciting task for further both empirical and theoretical research.
19
This is an integral part of the theoretic model of Galdon-Sanchez and Güell (2003) as well.
30
References
Ascheid, R.; Preis, U.; Schmidt, I. (2000): Großkommentar zum Kündigungsrecht, München.
Addison, J.; J. Grosso (1996): Job Security Provisions and Employment: Revised Estimates.
In: Industrial Relations (35), 585-603.
Addison, J; P. Teixeira; J. Grosso (2000): The Effects of Dismissals Protection on
Employment: More on a Vexed Theme. In: Southern Economic Journal (67), 105-122.
Bertola, G.; T. Boeri; S. Cazes (1999): Employment Protection and Labour Market
Adjustment in OECD Countries: Evolvine Institutionsand Variable Enforcement.
International Labour Office, Employment and Training Papers 48.
Booth, A. L. (1994): Layoffs with Payoffs: A Bargaining Model of Union Wage and
Severance Pay Determination. In: Economica (62), 551-564.
Burda, M. (1991): Monopolistic Competition, Costs of Adjustment, and the Behavior of
European Manufacturing Employment. In: European Economic Review (35), 61-79.
Burda, M. (1992): A Note on Firing Costs and Severance Benefits in Equilibrium
Unemployment. In: Scandinavian Journal of Economics (94), 479-489.
Del Boca, A.; P. Rota (1998): How Much Does Hiring and Firing Cost? Survey Evidence
from a Sample of Italian Firms. In: Labour (12), 427-449.
Dorndorf, E. et al. (2001): Heidelberger Kommentar zum Kündigungsschutzgesetz, 4.
Auflage, Heidelberg.
Eger, Thomas (2002): Opportunistic Termination of Employment Contracts and Legal
Protection Against Dismissal in Germany and the USA. German Working Papers in
Law and Economics 11/2002.
Emerson, Michael (1988): Regulation or Deregulation of the Labour Market. In: European
Economic Review (32), 775-817.
Fabel, O. (1996): Abfindungszahlungen an entlassene Arbeitnehmer. In: Zeitschrift für
Betriebswirtschaft (66), 587-606.
Fabel, O. (2002): Severance Pay Rules and Structural Layoff Decisions. In: Schmalenbach
Business Review (54 of zfbf), 118-135.
Falke, J. (1983): Kündigungspraxis und Kündigungsschutz. In: Ellermann-Witt et al. (eds.):
Beiträge zur sozialwissenschaftlichen Forschung, Opladen, 13-43.
Falke, J. et al. (1981): Kündigungspraxis und Kündigungsschutz in der Bundesrepublik
Deutschland, Reihe „Forschungsberichte“ hrsg. vom Bundesminister für Arbeit und
Sozialordnung, Bonn.
Financial Times Deutschland (2000): Mannesmann/Vodafone: Esser bekommt 60,5 Mio. DM
Abfindung (11.2.2000, URL: http://www.ftd.de/tm/tk/1047263.html)
Fitzenberger, B.; I. Haggeney; M. Ernst (1999): Wer ist noch Mitglied in Gewerkschaften –
eine Paneldatenanalyse für Westdeutschland. Zeitschrift für Wirtschafts- und
Sozialwissenschaften (119), 223-263.
Franke, K. (1996): Effizienz gesetzlicher Kündigungsschutzbestimmungen. Wiesbaden.
Galdon-Sanchez, J. E. ; M. Güell (2003) : Dismissal Conflicts and Unemployment (47), 323-
335.
Gibbons, R.; L. F. Katz (1991): Layoffs and Lemons. Journal of Labor Economics (9), 351-
380.
Goerke, L. (2003): Redundancy Pay and Collective Dismissals. In: FinanzArchiv (59), 68-90.
Hamermesh, D.; G. Pfann (1996): Adjustment Costs in Factor Demand. In: Journal of
Economic Literature (34), 1264-1292.
31
Handelsblatt (2003): Schröder will im Bündnis für Arbeit über Tabuthemen reden
(21.01.2003), 1.
Heckman, J. (1976): The Common Structure of Statistical Models of Truncation, Sample
Selection, and Limited Dependent Variables and a Simple Estimator for Such Models.
In: The Annals of Economic and Social Measurement (5), 475-492.
Hemmer, E. (1997a): Sozialpläne und Personalanpassungsmaßnahmen – Eine empirische
Untersuchung, Köln.
Hemmer, E. (1997b): Kritik der Unternehmen an Sozialplanregelungen. Der Arbeitgeber (49),
102-105.
Hemmer, E. (1997c): Sozialplanregelungen verbessern. In: Der Arbeitgeber (49), 130-135.
Heseler, H.; U. Mückenberger (1999): The Management of Redundancies in Europe: The
Case of Germany. In: Labour (13), 183-235.
Hunt, J. (2000): Firing Costs, Employment Fluctuations and Average Employment: An
Examination of Germany. In: Economica (67), 177-202.
Ichino, A.; Polo, M.; Rettore, E. (2003): Are Judges Biased by Labor Market Conditions? In:
European Economic Review (47), 913-944.
Inhoffen, A.O.; Müller-Dahl, F.P. (1981): Rückstellungen für Abfindungen aus Sozialplan
und Aufhebungsvertrag? In: Der Betrieb (34), 1473-1477.
Jahn, E. J. (2002): Zur ökonomischen Theorie des Kündigungsschutzes, Berlin.
Jahn, E. J.; Schnabel, C. (2003): Bestandsschutz durch Abfindungen: Höhere Rechtssicherheit
und Effizienz. In: Wirtschaftsdienst 83/4, 219-223.
IZA (2002): Bessere Rahmenbedingungen für mehr Beschäftigung. IZA Compact.
Sonderausgabe, November 2002.
Kittner, M.; T.C. Kohler (2000): Kündigungsschutz in Deutschland und den USA. In:
Betriebsberater (55), supplement to issue 13/2000.
Kodrzycki, Y. K. (1998): The Effects of Employer-Provided Severance Benefits on
Reemployment Outcomes. In: New England Economic Review (November/December
1998), 41-68.
Küchle, H. (1990): Kündigungsschutzvorschriften im internationalen Vergleich. In: WSI-
Mitteilungen, 407-414.
Lazear, E. P. (1990): Job Security and Employment. In: Quarterly Journal of Economics
(105), 699-725.
Malo, M. A. (2000): A Simple Model of Severance Pay Determination: The Case of
Individual Dismissals in Spain. In: Labour (14), 269-290.
Malo, M. A. (2001): European Labour Law and Severance Pay Determination in Collective
Redundancies. In: European Journal of Law and Economics (12), 73-90.
Malo, M. A.; J. Perez (2003): Individual Dismissals in Europe and the United States: A
Model on the Influences of the Legal Framework on Firing Costs. In: European Journal
of Law and Economics (15), 47-63.
OECD (1999): Employment Outlook, Paris.
Oi, W. (1962): Labor as Quasi-Fixed Factor. In: Journal of Political Economy (70), 538-555.
Pencavel, J. (2001): The Response of Employees to Severance Incentives – The University of
California´s Faculty, 1991-94. In: The Journal of Human Resources (36), 58-84.
Pita, C. (1997): Breach Penalties in Labour Contracts: Advance Notice and Severance Pay. In:
Labour (11), 469-495.
Schrep, B. (2003): Geräuschlos gefeuert. In: Der Spiegel (2.6.2003), 50-52.
Suedekum, J.; P. Ruehmann (2003): Severance Payments and Firm-specific Human Capital.
In. Labour (17), 47-62.
Statistisches Bundesamt (1999): Statistisches Jahrbuch für die Bundesrepublik Deutschland,
Wiesbaden.
32
Appendix
Table A: Fraction of dismissed persons with severance payments and ∆GDP in the 1990s
(corresponds to Figure 1)
Year
Fraction of dismissed persons, who
received severance payments
∆GDP
1991 0,343 ---*
1992 0,386 0,074
1993 0,351 0,025
1994 0,433 0,049
1995 0,341 0,038
1996 0,239 0,018
1997 0,215 0,021
1998 0,255 0,031
1999 0,206 0,026
2000 0,145 0,026
2001 0,136 0,02
2002 0,177 0,018
Note: *: Value for ∆GDP is missing in 1991 because of the German re-unification in 1990.
Table B: Distribution of severance payments and severance pay factors
(corresponds to Figure 2, 3, A and B)
Amount of severance payment Severance pay factor
Percentile
Whole
sample
Individual
dismissals
Collective
dismissals
Whole
sample
Individual
dismissals
Collective
dismissals
0.1
1,089 1,105 1,077 0.108 0.119 0.087
0.2
1,764 1,764 1,842 0.165 0.183 0.133
0.3
2,582 2,396 2,774 0.219 0.242 0.176
0.4
3,348 3,218 3,859 0.276 0.302 0.252
0.5
4,663 4,356 5,155 0.347 0.375 0.292
0.6
6,140 5,881 6,755 0.428 0.449 0.396
0.7
8,446 7,982 9,247 0.534 0.543 0.534
0.8
12,947 11,762 14,860 0.667 0.688 0.642
0.9
22,190 19,365 27,854 0.917 0.917 0.955
Note: Severance pay factor = amount of severance pay / (gross monthly wage
*
tenure)
33
Figure A: Distribution of severance payments
Figure B: Distribution of severance pay factors
Note: F(·) = Distribution function of severance payments (figure A) respectively severance pay
factors (figure B).
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
0
,05
0,
15
0,
25
0,
35
0,45
0,55
0,
65
0,
75
0,
85
0,95
F(severance payments)
Amount of severance pa
y
Individual dismissals
Collective dismissals
0
0,2
0,4
0,6
0,8
1
1,2
1,4
0
,05
0,
1
5
0,
2
5
0,
3
5
0
,45
0
,55
0,
6
5
0,
7
5
0,
8
5
0
,95
F(severance pay factors)
severance pay factors
Individual dismissals
Collective dismissals
34
Table C: Descriptive statistics and overcompensation
All
(n=2534)
Persons
without
severance
payments
(n=1845)
Persons
with
severance
payments
(n=689)
Persons with
severance
payments and
wage increases
(n=186)
Severance Payment 0.272 0 1 1
Sex (male) 0.673 0.694 0.617 0.667
Age (years) 39.48 38.46 42.19 39.12
Years of schooling 11.41 11.35 11.57 12.25
Marital status (single) 0.258 0.297 0.155 0.183
Child in household 0.422 0.425 0.415 0.505
Unemployed 0.521 0.530 0.496 0.231
Blue collar worker 0.616 0.638 0.557 0.473
Tenure (years) 7.303 5.488 12.16 10.25
Firm size:
1 - 5 employees
6 - 19 employees
20 - 199 employees
200 - 1999 employees
≥2000 employees
0.116
0.247
0.347
0.189
0.101
0.150
0.293
0.345
0.138
0.074
0.005
0.122
0.351
0.328
0.174
0.043
0.140
0.398
0.269
0.151
Industries:
Farming/Forestry/Fishing
Manufacturing
Construction
Service industries
0.032
0.364
0.208
0.397
0.037
0.304
0.243
0.416
0.017
0.525
0.112
0.345
0.038
0.500
0.075
0.387
Region (West Germany) 0.519 0.523 0.507 0.419
Citizenship (German) 0.836 0.838 0.830 0.930
