Return on Investment in Higher Education–Evidence for Different Subjects, Degrees and Gender in Germany

Abstract

phone: +491753185282, fax.: +496971625283.

Full text

Return on Investment in Higher Education – Evidence
for Different Subjects, Degrees and Gender in Germany
Mark Wahrenburg
1
, Martin Weldi
2
Goethe University Frankfurt
This version: March 1, 2007
1
Chair of Banking and Finance, Johann Wolfgang Goethe University, Mertonstrasse 17, 60054 Frankfurt,
Germany. Email: wahrenburg@finance.uni-frankfurt.de.
2
PhD student at the Chair of Banking and Finance, Johann Wolfgang Goethe University, Mertonstrasse 17,
60054 Frankfurt, Germany. Email: martinweldi@web.de, phone: +491753185282, fax.: +496971625283.

II
Abstract:
This paper considers higher education as an investment decision and presents empirical evidence on
the private monetary rate of return on this investment. Employing German data, this is to the best of
our knowledge the first paper that simultaneously differentiates by subject, higher education degree
and gender and calculates the internal rate of return applying an expanded "Mincer-Earnings-
Equation" to estimate the earnings capacity. Employing a sample of 17,180 higher education
graduates derived from the German Labor Force Survey 2004, we show that the returns on
investment strongly differ between the different forms of higher education, with studies in the
subjects Art, Agriculture and partly Languages and Cultural Sciences not representing attractive
investments from an economic point of view. We find that the decision what to study is worth several
hundred thousand Euros. We further show that each gender has a return advantage in subjects where
it reaches a strong relative presence. Assuming private financing of the actual cost of study, the
overall order of the different forms of education remains stable, but the investment in some subjects
is no longer clearly attractive (e.g., Engineering studies). Comparing the return of an investment in
higher education and the cost of study, we find that more expensive subjects (apart from Medicine)
yield a lower return.
Keywords: Returns to Education, Human Capital, Higher Education Earnings Capacity.
JEL-Classification: I21, I28, J31.

1
1 Introduction
According to Human Capital Theory higher education represents an investment decision. Compared
to other investment alternatives, education must yield a higher rate of return in order to be pursued
from an economic point of view. Knowledge about the return on investment might help individuals to
make better informed schooling decisions by adding an economic perspective to it. Taking the return
on investment as a private decision criterion is gaining in importance, as we currently observe an
increasing private contribution to higher education cost in many countries fueled by tight government
budgets. Moreover, even if higher education is mainly publicly financed, knowing the private return
to different education alternatives allows to generate important insights for a prioritization with
regard to the allocation of public funds to certain areas of education or might help to explain the
demand for the different forms of education. Assessing the rate of return is of particular interest for
the German higher education market, on which we focus our analysis, as Germany is the largest
higher education market in Europe, is still characterized by a high public financing share in higher
education
3
, but has recently announced the introduction of tuition fees
4
.
In this paper, we follow a pure investment perspective and analyze the private monetary returns to
higher education under two financing regimes: First, a system with full public financing of higher
education and second, a system where the actual production cost of higher education are covered by
the students, thus reflecting the trend to increased private contribution. In line with existing literature,
we do not directly consider any externalities that might benefit the individual or the society as a
whole (e.g., a consumption value of studies or better health and lower crime).
A large set of empirical research shows that, overall, higher education represents an attractive private
investment (see Psacharopoulos (1994) and Blöndal et al. (2002) for international comparisons and
Lauer and Steiner (2000) and Ammermüller and Weber (2005), among others, for recent studies with
3
According to OECD (2004), Table B3.2b, p. 243, the public financing share in tertiary education reaches 92
percent.
4
Several universities introduce flat tuition fees of 500 Euro per semester in 2007.

2
German data). Each year of (higher) education yields a private return of between 7 and 19 percent
5
on an international scale and 6 and 10 percent in Germany depending on the data used and
methodology applied, which is above the return on alternative investments. However, the existing
studies do not simultaneously differentiate between three factors that have been shown to
significantly influence the income prospects of graduates: gender, degree and subject. In addition to
the income prospects, the cost of study largely differs between the different subjects and degrees. It is
therefore necessary to obtain a more detailed picture of the returns to education, which is performed
in this paper.
Most previous studies approximate the private rate of return through the coefficient for years of
schooling or degree-dummy variables in a "Mincer-Earnings-Equation". In this paper, we will apply
an expanded "Mincer-Earnings-Equation" following Murphy and Welch (1990) to empirically
estimate experience-income profiles of both higher education graduates and high school graduates
without further vocational education as the relevant control group. We use the generated wage
profiles as input parameters and calculate the return on investment through the internal rate of return
(IRR) and the net present value (NPV) of the cash flow streams as the relevant criteria for an
investment decision (see Blöndal et al. (2002), Psacharopoulos (1995)).
We base our analysis on a large dataset of about 24,000 higher education graduates and 1,400 high
school graduates who had the right to, but did not pursue higher education, derived from the German
Labor Force Survey 2004. This allows us to assess the monetary benefits of higher education and is
ideally suited for our research as it contains detailed information on both subject and degree of
graduates.
The aim of this paper is to analyze how the private return on investment in higher education
measured through the internal rate of return differs between the different gender-degree-subject-
subgroups. We test the hypothesis that the returns strongly differ between the subgroups with some
subjects having returns far below the average returns found in previous studies due to low earnings
5
See Blöndal et al. (2002).

3
prospects. In addition, we investigate how the picture changes if the actual cost of study must be
covered by the students and what, if any, the "unobservable" return would be to make certain types of
higher education a worthwhile investment.
This paper adds to the existing literature in the following ways. It is, to the best of our knowledge, the
first study for Germany that estimates the return on investment in higher education for the different
gender-degree-subject-subgroups by applying a classical investment approach and relying on
regression techniques to estimate the income prospects of both higher education graduates and high
school graduates without further vocational education as the control group. It analyzes the returns to
education in greater detail through simultaneous differentiation by gender, degree, and subject, as
well as inclusion of the actual cost of study. In addition, we use our calculations as input parameters
for a regression analysis to test the factors influencing the return on investment. Our analysis
therefore combines the advantages of the two approaches used to estimate returns to education in the
literature. The direct calculation through the internal rate of return of an investment project allows for
a better treatment of the actual cost occurred and the regression-based analysis of the income
prospects and the return on investment allows to consider various influencing factors.
Our main findings are the following. We show considerable variation in the rates of return to higher
education across the different subjects and degrees, with some forms of higher education not being
attractive from economic point of view. We further show that the decision what to study and what
type of institution to attend might be worth several hundred thousand Euros. Concerning the different
subjects, we confirm the overall order of subjects found by prior research with Medicine and Law,
Economics and Social Sciences yielding the highest private returns and studies in the subjects Art
and Agriculture as well as to a certain extent Languages and Cultural Studies yielding returns below
those of alternative investments.
When looking at gender- and degree-specific returns to investment, we find a differentiated picture
and can only partly confirm the generalist finding of prior research that women have a higher return
than men and that studies at a University of Applied Sciences yield a higher return compared to
University studies. We show that each gender reaches a return advantage in subjects where it shows a

4
strong relative presence. Moreover, graduates from a University of Applied Sciences reach a higher
relative return in subjects that are strongly represented at this type of institution.
When taking the cost of study into account assuming a private financing of higher education cost, the
overall order of the different subgroups remains stable, but the investment in some subjects is no
longer clearly attractive (e.g., Engineering studies). Comparing the IRR of an investment in higher
education and the cost of study, we find that apart from the subject Medicine the most expensive
subjects also yield the lowest IRR, which might give an indication for a misallocation of public funds
unless there are high social returns.
Regression analysis with NPV and IRR as dependent variables supports most of the findings derived
from our calculations. Gender and degree are no significant predictors of the return on investment,
but subject helps to explain a large part of its variation. The regression analysis confirms the subject-
specific return advantages for men and women, while the subject-specific advantages for the two
degrees have not proven to be statistically significant. Moreover, we find a significant negative
relationship between cost and NPV when excluding the subject Medicine.
The paper is outlined as follows. The next section briefly describes the methodology applied and
gives an overview of relevant previous research. The third section describes the dataset and estimates
detailed experience-earnings profiles both for higher education graduates and high school leavers
using OLS regression. In the fourth section we calculate the returns on investment in different forms
of higher education and validate our findings with regression analysis. Section 5 concludes.
2 Background and Methodology
2.1 Human Capital and Signaling Theory
Following Human Capital Theory (see Schultz (1961), Becker (1993) and Mincer (1974) for
pioneering work) education can be considered as an investment project. It requires resources that
have a cost in terms of opportunity cost through foregone earnings as well as direct cost, and
increases the productivity of the individuals taught. Assuming that individuals get paid based on their
productivity graduates with a higher education degree should yield a higher income than individuals

5
that did not pursue higher education. Education should continue as long as there is a positive
difference between the marginal benefit and the marginal cost of education.
Some researchers dispute the productivity enhancing effect of higher education that is the
fundamental assumption beyond Human Capital Theory. According to the signaling hypothesis (see
Spence (1973)) education serves as a signal for higher quality, but it is the inherent ability that
determines the productivity of individuals. Potential employers take higher education as a positive
signal for the productivity and motivation of individuals. While at least a certain productivity
enhancing effect of higher education appears undeniable (for medicine and engineering graduates,
e.g., the skills obtained at university represent a prerequisite to do their work) and could not be
rejected by empirical work, we do not intend to contribute to solving the puzzle. We are interested in
the question what the yield to higher education as an investment is and do not distinguish whether
this yield is paid due to enhanced productivity or a positive signal associated with education.
2.2 Methodology
Considering (higher) education as an investment, we follow classical investment theory suggesting
that the Net Present Value (NPV) or the Internal Rate of Return (IRR) of the cash flow streams
associated with higher education represent the relevant criteria in order to establish investment
priorities.
6
The NPV is the present value of the difference between the benefits of higher education
and its cost. The IRR represents the discount rate that equates the present value of additional income
compared to those who had the right to, but did not pursue higher education (control group) to the
present value of cost (opportunity cost through foregone earnings and, under a private financing
scheme, direct cost of study). If this rate of return is higher than an adequate market interest rate at
which the individual can borrow, education represents a worthwhile investment for the individual. If
the private IRR is below the relevant interest rate, we can quantify an "unobservable" return
component that would be necessary to make education a worthwhile investment.
6
See also Psacharopoulos (1995) for a discussion of different methodological approaches.

6
The following figure highlights the stylized cash flows from an investment in higher education
following Psacharopoulos (1995):
Cash
flow/
income
0
55
Age
20
D T
Cost
Benefits
Higher Education
Graduates
Higher School
Leavers
Cash
flow/
income
0
55
Age
20
D T
Cost
Benefits
Higher Education
Graduates
Higher School
Leavers
We confine our analysis to the private monetary return to different forms of higher education and
analyze two higher education financing regimes: A system with full public financing of higher
education production cost and a system where the actual production cost of higher education are
covered by the students, thus reflecting the trend to increased private contribution
7
. In line with
existing literature, we do not directly consider any externalities that might benefit the individual or
the society as a whole (e.g., the consumption value of studies or better health and lower crime).
We calculate the private monetary return on investment through the following formula for the public
financing scheme (1) and the private financing scheme (2).

==
+
+
=
+

D
d
d
CG
s
T
t
Dt
CG
s
HE
sdj
r
w
r
ww
11
,,
)1()1(
(1)

==
+
+
+
=
+

D
d
d
HE
dj
CG
s
T
t
Dt
CG
s
HE
sdj
r
Cw
r
ww
1
,
1
,,
)1()1(
(2)
7
For arguments in favor of private contribution to higher education cost see among others Barr (1993, 2004),
Chapman (1997, 2005) and Friedman (1962).

7
Where:
studies,educationhigherofDuration
Graduates,EducationHigherofLifeWorking
ddegreeandjsubjectforStudyofCost
earnings),(foregonesgenderwithgroupcontrolfor theIncomeNet
s,genderd,degreej,subjectforgraduateseducationhigherofIncomeNet
rate),(discountReturnofRateInternal
,
,,
=
=
=
=
=
=
D
T
C
w
w
r
HE
dj
CG
s
HE
sdj
For ease of calculation, previous research mainly estimates the returns to education through a
"Mincer-Earnings Equation" with the natural logarithm of net income as dependent variable and
years of schooling and experience as independent variables. The coefficient for years of schooling
approximates the rate of return of one additional year of schooling. This approach assumes a linear
return to all years of schooling, measures the return to education only through income differentials of
different forms of education and does not allow to analyze the impact of a private coverage of the
production cost of study. Heckman et al. (2005) shows employing US data that the conditions under
which the derived schooling coefficient equals the marginal internal rate of return (i.e., separability
of experience and schooling and neglecting direct cost) have not been fulfilled in recent years. An
estimation of rates of return through "Mincer-type" Earnings Equations might also be subject to both
an endogeneity bias due to, e.g., omitted ability and a selectivity bias due to neglecting the decision
whether a person works or not (see, e.g., Card (1999) and Heckman et al. (2003) for a review of
studies analyzing these issues). As there is a broad set of literature available and many of the
contributions do not find significant estimation biases, we do not cover those issues in the scope of
our research.
In our study, we use an expanded "Mincer-Earnings-Equation" following Murphy and Welch (1990)
to estimate earnings profiles of higher education graduates working full-time and our control group
as input parameters for the IRR-calculation, but do not directly derive return figures from a "Mincer-
Earnings-Equation".

8
2.3 Relevant empirical literature
There is a large set of literature available empirically assessing the private rate of return to higher
education. In the following, we want to give an overview of relevant previous empirical studies
differentiating by degree, subject or gender both for Germany and on an international scale
8
. Overall,
the studies highlight that higher education is an attractive private investment. Each year of (higher)
education yields a private return of between 7 and 19 percent
9
on an international scale and 6 and 10
percent in Germany depending on the data and methodology applied.
Most German studies approximate the private rate of return through the coefficient for years of
schooling or for a degree-dummy variable estimated in a "Mincer-Earnings-Equation". Two recent
studies are outlined in the following.
10
Lauer and Steiner (2000) differentiate by level of education
and higher education degree (University of Applied Sciences and University) as well as gender using
data from the Socio-Economic-Panel (SOEP). They find an overall return to year of education of 10
percent for women and 8 percent for men. Robustness checks show that the rates of return are
slightly decreasing over time and robust with regard to extended specifications to account for a
possible endogeneity bias. When estimating a duration of study adjusted annual return from higher
education they find an excess return for both male and female graduates from a University of Applied
Sciences compared to University graduates, the difference being higher for men (about one
percentage point) than for women (less than 0.5 percentage points) in the period 1984-1997
11
. Their
analysis suggests that the rate of return decreases with the duration of education. Ammermüller and
Weber (2005) also consider higher education subject when estimating the rate of return through a
"Mincer-Earnings Equation" applying two different datasets, but do not differentiate between the two
higher education degrees. Overall, they also find a return per year of education between 8 and 10
8
In addition to the studies described in this section, various other studies estimating the return to higher
education exist. However, the covered studies represent those with the most relevant findings for our
research question.
9
See Blöndal et al. (2002).
10
For a review of studies conducted before 2000 we refer the reader to Asplund and Pereira (1999), Chapter 6.
11
Bellmann, Reinberg and Tessaring (1994) also find in a prior study using 1987 data and taking individuals
with vocational training as the control group an excess return, which is, however, much higher (almost 5
percentage points).

9
percent for West-Germany with SOEP data, with women in general showing higher returns.
12
When
differentiating by level of education with SOEP 2002 data, Ammermüller and Weber find that
obtaining a higher education degree yields an annual return of 9.7 percent for men and 10.4 percent
for women. Concerning the impact of the subject chosen, they show with data derived from the Labor
Force Survey 2000 that returns are highest for higher education graduates in the subjects Medicine,
Economics/Law and lowest for Agriculture-, Art- and Music-majors.
13
The annual rates of return
range from 3.5 percent for female Agriculture graduates to 12 percent for male Law graduates.
Moreover, their results indicate that each gender reaches high relative returns compared to the other
gender in subjects where it has a strong presence (engineering for men and studies to become a
teacher for women). Both Lauer and Steiner and Ammermüller and Weber do, however, only take
opportunity cost into account. Moreover, as the income variable in the SOEP reflects gross income,
their results mix private and social returns and they consider a broad control group (individuals with
no degree, lower or upper secondary education), which might potentially lead to upward biased
returns to higher education.
Apart from the various "Mincer-based" studies, there is hardly any prior research for Germany that
calculates the internal rate of return as the discount rate that equates an income stream from higher
education to a stream of cost associated with it. Ederer and Schuller (1999) calculate the rate of
return for different subjects, but do not differentiate by gender and only analyze University graduates.
They find that there is a strong inter- and inner-subject variation in the labor force participation and
dropout-adjusted returns to education. The private rates of return reflecting opportunity cost range
from 8.5 percent for Medicine to -1.6 percent for Languages and Cultural Studies. In addition to the
private returns, Ederer and Schuller (1999) also estimate fiscal returns taking account for the
different production cost of study. Medicine, Art, and Languages and Cultural Studies have the
lowest fiscal returns due to the high cost of study. Their results, however, can only serve as a rough
indication of the internal rate of return since the relevant information relies on profiles derived from
12
Estimations with Labor Force Survey (Mikrozensus) data yield to similar results.
13
The overall results for the relative order of the different subjects are in line with research for other countries
(Blundell et al. (2000) for UK, Rumberger and Thomas (1993) for US), although slight differences occur.

10
empirical datapoints (initial and average earnings) and not on econometric estimates, and the return
for the subjects represents the aggregated unweigthed average of all studies belonging to a subject.
In addition to these studies, several international studies provide insights on the returns to education
for the different gender and degrees both for Germany and other countries. Psacharopoulos (1994),
building on his prior studies conducted in the 1970s, is a comprehensive meta paper combining return
to education figures for a large set of countries. He shows that over all countries considered, women
have a higher rate of return per year of education than men. Concerning the different subjects,
Engineering, Medicine, and Law and Economics have overall the highest private rates of return.
Blöndal et al. (2002) calculate and compare internal rates of return to higher education for various
OECD-countries differentiating by gender. Their calculations are based on a broad definition of the
private rate of return that takes labor force participation, cost of study (tuition) and subsidies to
students in addition to opportunity cost and income differentials into account. They find a rate of
return to higher education of 9.1
14
percent for German men with SOEP data, which is below the
average for the country set of 11.6 percent, but above the return on alternative investments. In
contrast to other studies, they find higher rates of return for men than for women (8.4 percent), which
can, however, be explained by the inclusion of labor force participation rates and a benefit from
higher relative labor force participation for men
15
. Instead of estimating experience-income profiles,
however, their calculations are based on average empirical earnings.
Having in mind that the results of the above mentioned studies are difficult to compare due to
differences in methodology, variable specification, data and observation period, the studies indicate
that the rate of return to higher education differs between the different degrees of higher education
and subjects as well as between the two gender. However, to the best of our knowledge, no study
simultaneously differentiates between all three factors and takes actual cost of study into account. It
remains therefore open whether the prior findings hold if we analyze the returns to higher education
14
2.7 percent of which are attributable to public student support.
15
Without adjusting for unemployment risk, the difference between men and women narrows to 0.2 percentage
points.

11
(investment) for the different degree-subject-gender-subgroups, both under the current public
financing scheme and under the assumption of private coverage of higher education cost. Moreover,
existing studies are primarily based on "Mincer-Earnings-Equations" or rely on empirical averages to
calculate an internal rate of return, but do not apply econometric techniques to estimate income
profiles of higher education graduates and the respective control group as input parameters of an IRR
calculation, which shall be performed in this paper.
3 Earnings Capacity Estimates
3.1 Data and Descriptive Statistics
We use the German Labor Force Survey 2004 ("Mikrozensus"), the official representative statistics
of the population and the labor market in Germany, for our analysis of the earnings capacity of both
higher education graduates and the control group of high school graduates without further
(vocational) education. The Labor Force Survey involves every year 1% of all households
(continuous household sample survey) who have the same probability of selection (random sample).
It is ideally suited for our research as it contains a sufficiently large number of graduates and
information on their subject and degree to derive detailed experience-earnings profiles.
16
We analyze
cross-sectional data as representative longitudinal data is not available for Germany.
The income variable in the Labor Force Survey comprises monthly net income. Net income is the
relevant income to assess the private benefits of higher education as it abstracts from taxation. As the
income variable consists of income classes, we use the mid-point of class for our analysis
17
.
Our calculations are based on a scientific-use file of the Labor Force Survey 2004. It is a factually
anonymized 70 percent-sample of the original Labor Force Survey, which was drawn as a
systematically random selection from the original data by the Federal Statistical Office. The scientific
16
The other potential source with information on higher education degree and subject of graduates, the Socio
Economic Panel (SOEP), has a too small sample size of higher education graduates with information on the
subject (only approximately 2000 graduates).
17
As the income distribution is right-censored at values above 18,000 Euro, we approximate the highest class
(that represents less than 0.3% of the higher education graduates) with a value of 22,000 Euro assuming the
same distance than in the previous class.

12
use-file consists of information on 499,849 individuals. For the purpose of our analysis, we derive
two subsamples from the scientific-use file: one for higher education graduates (Panel A) and one for
our control group of high school graduates without further (vocational) education (Panel B).
In order to derive Panel A we exclude all individuals who do not have a higher education degree
(graduation from University or University of Applied Sciences) (-459,871). This leaves us with
39,978 higher education graduates in the scientific-use file. In addition, individuals younger than 25
years or older than 55 years are excluded from the dataset to avoid a selection bias due to early and
partial retirement.
18
We also do not take into account individuals who did not indicate their income or
subject. Furthermore, we drop graduates who have not been interviewed at their main residence to
avoid double counting. After controlling for labor force participation, Panel A consists of 17,180
higher education graduates working full-time, on whom we will base our income estimation.
To get the control group subsample (Panel B), we exclude all individuals who have no high school
degree and those who have a vocational degree, leaving us with individuals that only have a high
school degree. In analogy to Panel A, we do not consider people older than 55, who did not indicate
their income or have not been interviewed at their main residence. Furthermore, we exclude
individuals who are currently attending an education institution and do not take into account those
who perform their military or alternative national service as they are likely to attend college
afterwards. After correcting for individuals whose major income source is not income from work,
this leaves us with 1,828 high school graduates who are eligible for higher or vocational education
but do not obtain a further degree, 1,416 of them work full-time (Panel B).
Table 2 presents the descriptive statistics for our selected subsamples. For the full-time-working
higher education graduates (Panel A), the average monthly net income is 2,656 Euro. 34 percent of
the full-time working higher education graduates are female. Concerning study information, Table 2
also shows the subject
19
and degree frequencies. Engineering is the most represented subject (almost
18
We follow Fitzenberger and Reize (2002), who also apply this selection.
19
We follow the official subject classification of the Statistisches Bundesamt. The subject Sport, however, is
part of other due to a low number of observations.

13
30 percent), followed by Law, Economics and Social Studies, and Languages and Cultural studies,
subjects in which almost one quarter graduated in. Most graduates hold a university degree (60
percent) and 40 percent have a degree from a University of Applied Sciences. When turning to the
subsample of high school graduates without any further (vocational) education (Panel B), we see a
lower average net monthly income (1,753 Euro) and a slightly higher level of potential work
experience, which reflects the fact that the control group starts working at a lower age. The share of
females is slightly lower (32 percent).
As we are interested in the earnings capacity of both higher education graduates and the control
group, table 3 provides a more detailed picture of the income variable differentiating by subject,
degree and gender. For Panel A, the table shows that the average net income differs largely between
the different subjects. Graduates in Medicine earn on average the highest income, followed by Law,
Economics and Social Studies. Art majors have the lowest average net monthly income of less than
2000 Euro. Graduates from a University earn on average a premium of 143 Euro compared to
graduates from a University of Applied Sciences and males have an average monthly net income
premium of around 900 Euro, even when abstracting for labor force participation by considering full-
time workers only. In our control group (Panel B), males have an average income premium of around
600 Euro. The analysis of the income variable gives an indication that the overall average income
difference between Panel A and Panel B of 900 Euro decreases significantly when looking at
subgroups. Art and Agriculture majors have only an average income premium of around 300 Euro
compared to the control group.
3.2 Estimation of Experience-Income Profiles
Instead of taking the empirical averages, we estimate "Mincer-type" earnings equations for our
subsamples of higher education graduates and the control group with the natural logarithm of income
as the dependent variable following common practice among economists to predict the mean monthly
net income per year of experience. Based on the estimated coefficients, we are able to derive detailed
experience-earnings profiles. In line with the literature, we estimate profiles for full-time employees
only.

14
Following the literature, we perform an ordinary least squares (OLS)-regression
20
(see Mincer
(1974), Becker (1993) for fundamental research). While the original specification by Jacob Mincer
uses a quadratic function of work experience in the earnings function, Murphy and Welch (1990)
showed in their paper that the quadratic specification leads to significantly biased estimates of the
earnings profile, overestimating earnings at low levels of experience and underestimating earnings at
high levels of experience. According to their analysis, a cubic or even quartic specification fits the
"real" earnings data, measured by the estimated means per experience cohort, much better. Since our
data supports their finding, we also apply higher order polynomials in experience in our analysis.
21
Higher Education Graduates
The descriptive statistics in section 3.1 have shown that average net monthly income of higher
education graduates differs by subject, degree and gender. An analysis by experience cohort indicates
that there might be differences in earnings growth for the different subgroups. We will therefore
differentiate by subject, degree and gender when estimating the earnings capacity of higher education
graduates using regression analysis, which leaves us with 32 subgroups (eight subjects, two degrees
and two sexes). Following Murphy and Welch (1990) we take the average income for each subgroup
(further differentiated by year of experience) as dependent variable in the regression analysis
22
and
introduce dummy variables for each subgroup (gender, degree and subject) as independent variables.
To test the hypothesis that the slope of the age-earnings profiles differs for the different subgroups,
we also introduce interaction terms for experience and degree, and experience and sex into the
regression.
20
As the dependent variable is censored, one might be worried to apply OLS regression. An estimation with
tobit regression, however, does not show any qualitative improvements.
21
In order to determine the optimal functional form in terms of degree of experience, we estimated a regression
with dummies for every year of experience. When comparing the results of this regression with regressions
with different degrees of experience, we found that a cubic specification fits the empirical data much better
compared to the quadratic form. In the fourth order, we found improvements for the control group, but only
marginal improvements for the higher education subsample.
22
We decide to use average income instead of individual values, as estimates with log of individual data tend to
underestimate the empirical mean profiles.

15
Our regression function to estimate the mean net monthly income for full-time workers dependent on
work experience, gender, degree and subject therefore takes the following form:
uITedumDegreeSexconstw
sd
x
x
x
j
jjds
H
E
)
)
++++++=

=
,
3
1
**
ˆ
*
ˆ
*
ˆ
.
ˆ
ln

With:
.
error term
ˆ
s,genderandexp.andd,degreeandexp.ofspolynomialdifferentforn termsInteractio
,experienceoforderth-xst variableindependenfor thetsCoefficien
,experienceoforderth-for xst variableIndependen
s,sexd,degreej,subjectforst variableindependenfortsCoefficien
ˆ
j,subjectforabledummy varitIndependen
constant,Regression.
income,netmeanoflogarithmNaturalln
,
,
,
=
=
=
=
=
=
=
=
u
IT
e
dum
const
w
s
d
x
x
s
dj
j
H
E


)
)
Table 4 gives an overview of the results from the above specified regression. All standard-errors are
heteroscedasticity-robust. The coefficients for the independent variables (subject, degree, gender and
experience) are highly significant. Moreover, the introduced interaction terms for both experience
and degree, and experience and gender are highly significant as well
23
, which indicates that the
experience-income profiles for the different degree and gender values differ by their slope and
justifies a closer investigation of the subgroups. The coefficients for the subject dummy variables
indicate that Medicine graduates earn on average the highest income followed by Law, Economics
and Social Sciences graduates. Graduates with a subject Art or Agriculture earn the lowest income,
which reflects the findings from the empirical mean analysis and is in line with prior research by,
e.g., Ammermüller and Weber (2005). The regression analysis also shows that University graduates
earn c.p. a premium compared to graduates from a University of Applied Sciences after some years
of experience. This result is in line with the Human Capital Theory that postulates that longer
23
We also tested interaction terms for subject and experience. However, we did not consider them in the final
regression as the coefficients have to a large part not been significant, indicating that for the different
subjects only a level effect can be observed.

16
education should yield a higher outcome as each additional year of schooling yields a positive return
and supports the finding of Lauer and Steiner (2000). Moreover, males earn more than females and
the coefficients of the interaction terms indicate that the income gap between males and females
increases with experience. This reflects the finding of Lauer and Steiner (2000) and Fitzenberger and
Reize (2002).
Since the highly significant interaction terms indicate different slopes for the experience-earnings
profiles by degree and gender, we also estimate regressions for four degree-gender-subgroups with a
dummy variable for subject. Table 4 also shows the result of the four subgroup regressions. Overall,
the results are in line with the regression analysis on the whole sample, but yield a more detailed
picture. In all subgroup regressions, most subject variables become highly significant.
Figures 1.1-1.4 show the experience-income profiles for the different subgroups of higher education
graduates derived from the results of the subgroup regression analysis. The profiles are concave until
a work experience of around 25 years, indicating that earnings increase over time, but at a decreasing
rate, which is consistent with the existing literature (Becker (1993), Mincer (1974), Murphy and
Welch (1990)). Our estimated profiles also show, however, that in our data net income increases
again after 25 years of experience until an experience of 30 years, which could be explained by major
promotions at this experience level. In line with prior research for Germany, the estimated profiles
are much steeper for men than for women, as well as for University graduates compared to graduates
from a University of Applied Sciences. The estimated profiles have an implied cumulative annual
experience-specific growth rate of 3.2 percent for male University graduates, 1.8 percent for female
University graduates, 2.7 percent for men graduating from a University of Applied Sciences and only
1.1 percent for women graduating from a University of Applied Sciences. The predicted annual net
income at a work experience of 30 years differs significantly between the subjects dependent on the
gender and degree of graduates. Male University graduates show a variation in expected average net
income of almost 2000 Euro, while Female University graduates have a variation of less than 1000
Euro.

17
High School Graduates Without Further Vocational Education (Control Group)
To predict the average net monthly income of the control group, we differentiate by gender. Our
regression function takes the following form
24
:
uITeSexconstw
s
x
x
x
s
C
G
)
)
++++=

=
4
1
**
ˆ
.
ˆ
ln

The right hand side of table 4 shows the results from the Control Group regression. All standard-
errors are heteroscedasticity-robust. The coefficients for the independent variables (experience and
gender) and the interaction terms for the different polynomials of experience and gender are highly
significant
2
5
. Figure 1.5 shows the derived experience-income profile both for men and for women.
As for the subsample of higher education graduates, the profiles are concave until a work experience
of around 25 years, increase again until an experience of 30 years, and decrease slightly afterwards,
which is in line with the functional form of the profiles derived from the original "Mincer-Earnings-
Equation"-specification for high years of experience. The estimated profiles are only slightly steeper
for men than for women, with an implied annual growth rate of 2.8 percent for men and 2.7 percent
for women.
4 Private Return on Investment
4.1 Calculations
We use the estimated earnings capacity to empirically calculate the IRR and NPV of an investment
into different forms of higher education. For our calculations, we assume an annual (constant) wage
growth, e.g., due to technological progress, since the estimated income per year of experience for the
different groups is derived from cross-sectional data and wages do not remain constant over time. We
therefore adjust the estimated average earnings per year of experience with a wage growth factor
24
In comparison to the regressions for the higher education subsample, we consider a forth order polynomial in
experience-specification. We apply this specification since it adds additional explanatory power and can be
explained by the longer potential work experience until the age of 55 for the control group compared to
higher education graduates.
25
For the control group, regressions with gender-subgroups yield the same result and can be directly
constructed from the full regression displayed in table 4.

18
g
of 2 percent, which reflects the average annual long-term real growth rate of German gross wages
for employees
2
6
. For a detailed description of all assumptions underlying our Return on Investment-
calculations, we refer the reader to Appendix A.
Table 5 shows the results of the Internal Rate of Return- and Net Present Value-calculations for the
two financing regimes considered.
<Table 5>
Public Financing of Cost of Study
When first looking at the traditional public financing scheme displayed at the left hand side (i.e. with
opportunity cost through foregone earnings as the only cost considered), we find that both the IRR
and the NPV differ strongly between the different forms of higher education. Our calculations show
in a detailed picture that most, but by far not all forms of higher education, as implicitly suggested by
undifferentiated prior research, are a worthwhile investment in an economic sense yielding an IRR
above alternative investments. On the one hand, Medicine and Law, Economics and Social Sciences
have the highest private returns, being above 11 percent for all degrees and gender, followed by
Mathematics and Natural Sciences showing a return on investment of above eight percent for all
subgroups. On the other hand, studies in the subjects Art and Agriculture as well as to a certain extent
Languages and Cultural Studies do yield a rate of return below long-term government bonds, for
some subgroups being even negative. The overall order of subjects is in line with the findings by
Ammermüller and Weber (2005) and international research. We do, however, find a much wider
range in the returns when considering the different subgroups, ranging from a high of 13.6 percent
for female Medicine students to a low of -16 percent for Male Art students. When analyzing the NPV
assuming a discount rate of 4 percent, we find that for the subgroup of Male University students it
differs between 243 thousand Euro for Law, Economics and Social Science graduates to a negative
26
The figure is derived from nominal gross monthly wages for employees in Germany for the period 1976-
2005 published by the statistical authority (Statistisches Bundesamt), adjusted for inflation. An overall
adjustment seems justified since several empirical studies found that the German wage structure stayed
fairly stable in the past. Fitzenberger and Kurz (2003) find that earnings grew quite uniformly in Germany
and that between and within groups' ratios remained constant over time. Abraham and Houseman (1995)
also find great stability in wage dispersion in Germany in the 1980s.

19
106 thousand Euro for Art students. Assuming you are an average male student who wants to study at
a University, your choice of subject might therefore be, c.p., a 340 thousand Euro question. For
Female University of Applied Sciences graduates, e.g., the variation is less, but the NPV of the
higher education investment ranges still from 100 thousand Euro to -10 thousand Euro.
When looking at gender-specific returns to investment, we find a differentiated picture after taking
account for subject and degree. For some subjects we confirm the finding of prior studies for
Germany that show an overall return premium for women compared to men (see Lauer and Steiner
(2000) and Ammermüller and Weber (2005)). The premium is largest for graduates in the subject
Languages and Cultural Sciences (more than 4 percentage points for University graduates and more
than 10 percentage points for graduates from a University of Applied Sciences). On the other hand,
however, we show that male graduates in the subject Engineering (both from a University and from a
University of Applied Sciences) as well as Law, Economics and Social Sciences graduates from a
University yield a higher return compared to women. It appears that both gender have a return
advantage in subjects where they reach a strong relative presence (men in Engineering and Law,
Economics and Social Sciences and women in Languages and Cultural Sciences and Art) indicating
that the two gender choose the subjects where they have a competitive advantage compared to the
other gender, a point first mentioned by Ammermüller and Weber (2005).
Concerning degree-specific returns to education, we cannot support the generalist finding of prior
studies that the duration of studies has a major impact on the rate of return and that shorter studies,
e.g. at a University of Applied Sciences, yield a higher return (see Lauer and Steiner (2000)). In our
sample, this is only true for studies in the subjects Engineering and partly Mathematics and Natural
Sciences and Economics and Social Studies. It appears that graduation from a University of Applied
Sciences yields a higher relative return in subjects that can be considered as their major competence
areas (measured by the number of students enrolled).

20
When comparing the IRR or NPV of an investment in higher education and the cost of study
27
, we
cannot observe a positive relationship between input in terms of cost of study and output measured
through the return on investment for different education alternatives. Figure 2 even gives an
indication for a negative relationship and shows that expensive subjects
28
also yield a low IRR/NPV.
As a consequence, under a public financing scheme the government invests high amounts of money
per student in certain subjects that do not yield high private returns and thus also provide low tax
revenues as a major component of monetary social returns. In order to still justify the governmental
sponsorship from an investment perspective this would require high other social returns or it gives an
indication for some public misallocation of funds.
Private Financing of Cost of Study
Switching from a public to a private financing system leads to a reduction in the internal rates of
return for all subgroups, but does not change the relative order of the different subgroups. The higher
the cost of study, the higher is the decrease in the IRR. The reduction in IRR is therefore less severe
for an education at a University of Applied Sciences due to a shorter duration of studies. Neither the
degree-specific findings derived under the public financing scheme, however, nor the overall order of
the different subjects change. Medicine and Law, Economics and Social Studies are still very
attractive investments although the return on the "expensive" subject Medicine decreases much more
than for the "cheap" subject Law, Economics and Social Studies, both having a similar return before
taking cost of study into account. The investment case for other subjects that have yielded a decent
private monetary return under a public financing system becomes, however, less clear. When
considering an IRR of 6 percent as an adequate discount rate (including a risk premium) for
investments in Human Capital, Engineering as well as Mathematics and Natural Sciences studies (at
a University) become borderline investment cases and Languages and Cultural Studies are only
worthwhile investments if studied by women at a University.
27
For a detailed description of the cost of study for the different degrees and subjects see Appendix B.
28
Excluding the subject Medicine.

21
Since we have confined our analysis to the private monetary returns to an investment in higher
education, studies that yield an IRR below that of alternative investments would require a high non-
monetary or "unobservable" return in order to become attractive from a personal point of view. For
Art, Agriculture and to a certain extent Languages and Cultural Studies this would mean that the
"unobservable" return component would have to close a return gap of 5 to more than 20 percentage
points. Since such an "unobservable return"-component appears to be high, certain subjects can
hardly be studied following a pure investment perspective.
The findings presented in this section are robust with regard to changes in the assumed annual wage
growth rate and different specifications of the wage equation for the control group.
4.2 IRR-/NPV-Regression Analysis
To validate the findings from the NPV- and IRR-calculations presented in section 4.1, we perform a
regression analysis with the NPV and IRR of the different forms of education presented in table 5 as
the respective dependent variables and gender, degree, subject and cost as independent variables.
The regression function takes the following form (example for NPV as dependent variable):
uITCostdumDegreeSexconstNPV
j
jjds
)
)
++++++=

**
ˆ
*
ˆ
*
ˆ
.

Table 6 gives an overview of the results from the above specified regression function both for NPV-
and IRR-regressions. All standard-errors are heteroscedasticity-robust. When looking at the NPV-
regressions on the left hand side of table 6, we can confirm most findings derived from the
calculations in the previous section. Both gender and degree are no significant predictors of the NPV,
which highlights that no gender or degree shows an overall return advantage. Many dummy variables
for the different subjects, however, are significant (Column I). We introduce interaction terms for
gender and subject (Column II) and confirm an advantage for women in the subjects Languages and
Cultural Studies and Art. The analysis, however, does not support the return premium for University
of Applied Sciences graduates in some subjects discussed in the previous section since interaction
terms for degree and subject are not statistically significant. Regression III confirms our hypothesis

22
that more expensive studies yield a lower return. We find a negative relationship between cost and
NPV when excluding Medicine. When accounting for subject, degree and sex, however, cost is no
longer statistically significant (Column IV).
The results are robust with regard to the discount rate applied and a consideration of cost of study
(NPV of private financing scheme) in the NPV-calculation. The IRR-regressions on the right hand
side also show similar results. The cost variable (Column III), however, is not significant, which can
be explained by the existence of large negative values for some forms of education
29
. In addition to
the results from the NPV-regression, regression II confirms a relative return advantage for males in
the subjects Law, Economics and Social Studies and Engineering.
5 Conclusion
In this paper we consider higher education as a private investment decision and present empirical
evidence on the private monetary rate of return on this investment. Employing German data, we are
to the best of our knowledge the first study that simultaneously differentiates by subject, higher
education degree and gender and calculates the internal rate of return and net present value as the
relevant criteria applying an expanded "Mincer-Earnings-Equation" to estimate the earnings capacity.
While a large set of undifferentiated studies showed that higher education yields a return above the
return on alternative investments, we find considerable variation in the rates of return to higher
education across the different subjects and degrees, with some forms of higher education not being
attractive from economic point of view. We show that the decision what to study and what type of
institution to attend might be worth several hundred thousand Euros.
Concerning the different subjects, we confirm the overall order of subjects found by prior research
for Germany with Medicine and Law, Economics and Social Sciences yielding the highest private
returns (above 11 percent for all degrees and gender), followed by Mathematics and Natural Sciences
(above 8 percent). Studies in the subjects Art and Agriculture as well as to a certain extent Languages
29
Applying an adjusted IRR figure confirms the negative relationship found in the NPV-regression.

23
and Cultural Studies, however, appear to be an unattractive investment yielding a rate of return below
long-term government bonds.
When looking at gender- and degree-specific returns to investment, our calculations show a
differentiated picture. Each gender reaches a return advantage in subjects where it shows a strong
relative presence (men in Engineering and Law, Economics and Social Sciences and women in
Languages and Cultural Sciences and Art) indicating that the two gender choose subjects where they
have a competitive advantage, a point first mentioned by Ammermüller and Weber (2005).
Moreover, graduates from a University of Applied Sciences yield a higher relative return in subjects
that are strongly represented at this type of institution (e.g., Engineering and partly Economics and
Social Studies).
Comparing the IRR of an investment in higher education and the cost of study, we find that apart
from the subject medicine expensive subjects also yield low returns, which gives an indication of
potential misallocation of public funds unless there are high social returns. When taking the cost of
study into account assuming a private financing of higher education cost, the overall order of the
different subgroups remains stable, but the investment case for some subjects is no longer clear (e.g.,
for Engineering studies).
Regression analysis with NPV and IRR as dependent variables confirms most of the findings derived
from our calculations. The presented findings therefore provide important insights in the
attractiveness of different forms of higher education. Before deriving an action plan from the results,
however, it is important to keep in mind that our findings are static and that changes in demand and
supply for different forms of higher education are likely to trigger changes in the rates of return to
different forms of education.

24
Literature
Abraham, K. and Houseman, S. (1995), Earnings inequality in Germany, in: Freeman, R., Katz,
L. (eds.) Differences and Changes in Wage Structures, University of Chicago Press, NBER,
p. 371-403.
Ammermüller, A. and Weber, A. M. (2005), Educational Attainment and Returns to Education in
Germany – An Analysis by Subject of Degree, Gender and Region, ZEW Discussion Paper
no. 05-17.
Asplund, R. and P. Pereira (1999), Returns to human capital in Europe: a literature review,
ETLA, Helsinki.
Barr, N. (1993), Alternative funding sources for higher education, Economic Journal, 103, p.718-
728.
Barr, N. (2004), Higher Education Funding, Oxford Review of Economic Policy, vol. 20, no. 2,
p. 264-283.
Becker, G. S. (1993), Human Capital: A Theoretical and Empirical Analysis with Special
Reference to Education, Chicago University Press, 1
st
edition 1975.
Bellmann, L., A. Reinberg and M. Tessaring (1994), Bildungsexpansion,
Qualifikationsstrukturen und Einkommensverteilung, in: R. Luedeke (ed.), Bildung,
Bildungsfinanzierung und Einkommensverteilung II, Berlin.
Blöndal, S., S. Field and N. Girouard (2002), Investment in Human Capital Through Post-
Compulsory Education and Training: Selected Efficiency and Equity Aspects, OECD
Economics Department Working Papers, no. 333, OECD Publishing.
Blundell, R., L. Dearden, A. Goodman and H. Reed (2000), The Returns to Higher Education in
Britain: Evidence from a British Cohort, The Economic Journal, vol. 110, p. 82-99.
Card, D. (1999), The Causal Effect of Education on Earnings, in: O. Ashenfelter and D. Card
(eds.) Handbook of Labor Economics, vol. 5, New York, North-Holland.
Chapman, B. (1997), Conceptual Issues and the Australian Experience with Income Contingent
Charges for Higher Education, The Economic Journal, vol. 107, no. 442, May, p.738-751.
Chapman, B. (2005), Income Contingent Loans for Higher Education: International Reform,
Centre for Economic Policy Research, Discussion paper no. 491.
Ederer, P. and P. Schuller (1999), Geschäftsbericht Deutschland AG, Schäffer Poeschel,
Stuttgart.
Fitzenberger, B. and Reize, F. (2002), Quantilsregressionen der westdeutschen Verdienste: Ein
Vergleich zwischen der Gehalts- und Lohnstrukturerhebung und der IAB-
Beschäftigtenstichprobe, ZEW Discussion Paper no. 02-79.
Fitzenberger, B. and Kurz, C. (2003), New insights on earnings trends across skill groups and
industries in West Germany, Empirical Economics, vol. 28, p. 479-514.
Friedman, M. (1962), Capitalism and Freedom, Chicago: The University of Chicago Press, 1962.
Heckman, J, L. Lochner and P. Todd (2003), Fifty Years of Mincer Earnings Regressions, NBER
Working Paper No. 9732.
Heckman, J, L. Lochner and P. Todd (2005), Earnings Functions, Rates of Return, and Treatment
Effects: the Mincer Equation and Beyond, NBER Working Paper No. 11544.
Hetmeier, H.-W. (1998), Methodik der Berechnung der Ausgaben und des Personals der
Hochschulen für Forschung und experimentelle Entwicklung ab dem Berichtsjahr 1995,
Wirtschaft und Statistik, vol. 2, pp. 153.
Lauer, C. and Steiner, V. (2000), Returns to Education in West Germany. An Empirical

25
Assessment, ZEW Discussion Paper no. 00-04.
Luedeke, R. and Beckmann, K. (1998), Social Cost of Higher Education: Production and
Financing. The Case of Germany (1994), Passauer Diskussionsbeiträge V-98-8, University
Passau.
Mincer, J. (1974), Schooling, Experience, and Earnings, New York, (National Bureau of
Economic Research).
Murphy, K. and Welch, F. (1990), Empirical Age-Earnings Profiles, Journal of Labor
Economics, vol. 8, p. 202-229.
OECD (2004), Education at a Glance.
Psacharopoulos, G. (1994), Returns to Investment in Education: A Global Update, World
Development 22 (9), p. 1325-1343.
Psacharopoulos, G. (1995), The Profitability of Investment in Education: Concepts and Methods,
The World Bank, Washington, DC.
Rumberger, R. and S. Thomas (1993), The Economic Returns to College Major, Quality and
Performance: A Multilevel Analysis of Recent Graduates, Economics of Education Review,
vol. 12, p. 1-19.
Schultz, T. (1961), Investment in Human Capital, American Economic Review, vol. 51 (1), p. 1-
17.
Spence, M. (1973), Job Market Signaling, The Quarterly Journal of Economics, vol. 87, p. 355-374.

26
Tables and Figures
Figure 1: Experience-Income Profiles
1.1 Higher Education Graduates – University, Male
Work Experience
500
1.000
1
.500
2.000
2.500
3
.000
3.500
4
.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Medicine
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
E
ngineering
Languages and
Cultural Studies
Agriculture
Art
Work Experience
500
1.000
1
.500
2.000
2.500
3
.000
3.500
4
.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Medicine
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
E
ngineering
Languages and
Cultural Studies
Agriculture
Art
500
1.000
1
.500
2.000
2.500
3
.000
3.500
4
.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Medicine
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
E
ngineering
Languages and
Cultural Studies
Agriculture
Art
MedicineMedicine
Law, Econ. and
Social Studies
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Mathematics and
Natural Sciences
E
ngineering
E
ngineering
Languages and
Cultural Studies
Languages and
Cultural Studies
AgricultureAgriculture
ArtArt
1.2 Higher Education Graduates – University, Female
Work Experience
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Medicine
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Engineering
Languages and
Cultural Studies
Agriculture
Art
Work Experience
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Medicine
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Engineering
Languages and
Cultural Studies
Agriculture
Art
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Medicine
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Engineering
Languages and
Cultural Studies
Agriculture
Art
MedicineMedicine
Law, Econ. and
Social Studies
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Mathematics and
Natural Sciences
EngineeringEngineering
Languages and
Cultural Studies
Languages and
Cultural Studies
AgricultureAgriculture
ArtArt

27
1.3 Higher Education Graduates – University of Applied Sciences, Male*
Work Experience
500
1.000
1
.500
2.000
2.500
3.000
3.500
4
.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Engineering
Languages and
Cultural Studies
Agriculture
Art
* Fields-of-study with less than 90 observations are not displayed
Work Experience
500
1.000
1
.500
2.000
2.500
3.000
3.500
4
.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Engineering
Languages and
Cultural Studies
Agriculture
Art
* Fields-of-study with less than 90 observations are not displayed
500
1.000
1
.500
2.000
2.500
3.000
3.500
4
.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Engineering
Languages and
Cultural Studies
Agriculture
Art
500
1.000
1
.500
2.000
2.500
3.000
3.500
4
.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Engineering
Languages and
Cultural Studies
Agriculture
Art
Law, Econ. and
Social Studies
Law, Econ. and
Social Studies
Mathematics and
Natural Sciences
Mathematics and
Natural Sciences
EngineeringEngineering
Languages and
Cultural Studies
Languages and
Cultural Studies
AgricultureAgriculture
ArtArt
* Fields-of-study with less than 90 observations are not displayed
1.4 Higher Education Graduates – University of Applied Sciences, Female*
Work Experience
Art
Law, Econ. and
Social Studies
Engineering
Languages and
Cultural Sciences
500
1.000
1.500
2.000
2.500
3.000
3.500
4.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
* Fields-of-study with less than 90 observations are not displayed
Work Experience
Art
Law, Econ. and
Social Studies
Engineering
Languages and
Cultural Sciences
500
1.000
1.500
2.000
2.500
3.000
3.500
4.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Work Experience
Art
Law, Econ. and
Social Studies
Engineering
Languages and
Cultural Sciences
Work Experience
Art
Law, Econ. and
Social Studies
Engineering
Languages and
Cultural Sciences
ArtArt
Law, Econ. and
Social Studies
Law, Econ. and
Social Studies
EngineeringEngineering
Languages and
Cultural Sciences
Languages and
Cultural Sciences
500
1.000
1.500
2.000
2.500
3.000
3.500
4.000
4.500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
* Fields-of-study with less than 90 observations are not displayed

28
1.5 High School Graduates Without Further Vocational Education (Control Group)
Work Experience
500
1,000
1
,500
2,000
2,500
3,000
3,500
4
,000
4,500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
Male
Female
Work Experience
500
1,000
1
,500
2,000
2,500
3,000
3,500
4
,000
4,500
12345678910 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
MaleMale
Female

29
Figure 2: NPV/IRR from Investments in Higher Education compared to Cost of Study
2.1 NPV vs. Cost of Study*
-150
-100
-50
0
50
100
150
200
250
300
0 10 20 30 40 50 60
Cost
in tsd. Euro
NPV
in tsd. Euro
-150
-100
-50
0
50
100
150
200
250
300
0 10 20 30 40 50 60
Cost
in tsd. Euro
NPV
in tsd. Euro
2.2 IRR vs. Cost of Study*
-20%
-15%
-10%
-5%
0%
5%
10%
15%
0 10 20 30 40 50 60
Cost
in tsd. Euro
IRR
* Without subject Medicine
-20%
-15%
-10%
-5%
0%
5%
10%
15%
0 10 20 30 40 50 60
Cost
in tsd. Euro
IRR
* Without subject Medicine

30
Table 1
Description of Variables
Variable Description
Earnings Capacity Regression Analysis
Dependent variable:
Ln mean income
N
atural logarithm of the average monthly net income of full-time
workers (in Euro as of March 2004)
Independent variables:
Exp Work experience of higher education graduates in years
Expsq Squared experience (Exp * Exp) in years
Exp3 3rd order of work experience (Exp * Exp * Exp) in years
Sex Dummy variable equal to 1 if gender is female.
Degree Dummy variable equal to 1 if highest degree is a University degree.
Subject
Law / Econ. / Social St.
Dummy variable equal to 1 if subject is Law, Econ. and Social St.
Math. / Natural sciences
Dummy variable equal to 1 if subject is Math. and Natural Sciences.
Medicine Dummy variable equal to 1 if subject is Medicine.
Agriculture Dummy variable equal to 1 if subject is Agriculture.
Engineering Dummy variable equal to 1 if subject is Engineering.
Art Dummy variable equal to 1 if subject is Art.
Other studies Dummy variable equal to 1 if subject is Other studies.
Interaction terms:
InterExpSex Interaction term for Exp and Sex (Exp * Sex)
InterExpsqSex Interaction term for Expsq and Sex (Expsq * Sex)
InterExp3Sex Interaction term for Exp3 and Sex (Exp3 * Sex)
InterExpDegree Interaction term for Exp and Degree (Exp * Degree)
InterExpsqDegree Interaction term Expsq and Degree (Expsq * Degree)
InterExp3Degree Interaction term Exp3 and Degree (Exp3 * Degree)
Reference categories (Omitted variables) for dummy variables:
Subject Languages / Cultural Studies
Degree University of Applied Sciences
Additional variables NPV/IRR-Regression Analysis
Dependent variables:
NPV
Net Present Value of the investment in higher education (in tsd. Euro)
IRR Internal Rate of Return of the investment in higher education (in
percent)
Independent variables:
Cost Cost of Study (in thousands of Euro)
Interaction terms:
IntersexLCS Interaction term for Sex and subject Languages / Cultural St.
IntersexLES Interaction term for Sex and subject Law / Econ. / Social St.
IntersexEng Interaction term for Sex and subject Engineering
IntersexArt Interaction term for Sex and subject Art

31
Table 2
Descriptive Statistics for Higher Education Graduates
This table presents descriptive statistics for the subsample of 17,180 higher education
g
raduates working full-time (Panel A) and the control group of 1,416 high school graduates
without further vocational education (Panel B) from the Labor Force Survey 2004. Panel A is
calculated after adjusting for individuals who did not state their subject or income, were
younger than 25 or older than 55, were not surveyed at their main residence and are not
working full-time. Panel B excludes individuals who did not state their income, were younger
than 17 or older than 55, were not surveyed at their main residence, currently attend an
e
ducation institution or perform their military/alternative national service and are not working
full-time.
Variables Mean
Standard
Error
Min Max N
Panel A: Higher Education Graduates who work full-time
Monthly Net Income 2656 15 225
22000 17180
Exp 16 0.062 1 31 17180
Sex 0.34 0.004 0 1 17180
Degree
0.60 0.004 0 1 17180
Subject
Languages and cultural studies 0.22 0.003 0 1 17180
Law / Econ. / Social studies
0.25 0.003 0 1 17180
Math. / Natural sciences
0.09 0.002 0 1 17180
Medicine
0.04 0.002 0 1 17180
Agriculture
0.02 0.001 0 1 17180
Engineering
0.29 0.003 0 1 17180
Art
0.04 0.001 0 1 17180
Other studies
0.06 0.002 0 1 17180
Panel B: High School Graduates without vocational education who work full-time
Monthly Net Income 1753 35 225
22000 1416
Exp 17 0.249 1 36 1416
Sex 0.32 0.012 0 1 1416

32
Table 3
Descriptive Statistics for Monthly Net Income of Full-time Workers
This table presents the means, standard errors, 25%-, 50%-, 75%-percentiles and number of
observations of the income variable (net monthly income measured in Euro) for several
subgroups derived from the scientific-use file of the Mikrozensus 2004. We distinguish by subject,
degree and gender for the subsample of higher education graduates (Panel A) and gender for the
c
ontrol group (Panel B).
Subgroups Mean
Standard
Error
Lower
Quartile Median
Upper
Quartile
N
Panel A: Higher Education Graduates who work full-time
Total
2,656 15 1,600 2,150 3,050 17,180
Split by Sex
Male 2,962 20 1,850 2,450 3,400 11,362
Female 2,059 18 1,400 1,850 2,450 5,818
17,180
Split by Degree
University 2,714 20 1,600 2,450 3,050 10,279
University of Applied Sciences 2,571 21 1,600 2,150 3,050 6,901
17,180
Split by Subject
Languages and cultural studies 2,316 18 1,600 2,150 2,750 3,713
Law / Econ. / Social studies
3,012 39 1,850 2,450 3,400 4,335
Math. / Natural sciences
2,761 51 1,850 2,450 3,050 1,519
Medicine
3,318 97 1,850 2,750 3,800 708
Agriculture
2,118 73 1,400 1,850 2,450 410
Engineering 2,724
25 1,850 2,450 3,400 4,921
Art
1,969 58 1,200 1,600 2,450 607
Other studies
2,034 45 1,400 1,850 2,450 967
17,180
Panel B: High School Graduates without further vocational education who work full-time
Total
1,753 35 1,000 1,400 2,150 1,416
Split by Sex
Male 1,943 49 1,200 1,600 2,450 961
Female 1,353 34 800 1,200 1,600 455
1,416

33
Table 4
Final Results from Income-Regression Analysis
T
his table reports coefficient estimates from OLS regressions relating the natural logarithm of average net monthly income (dependent variable) to subject,
degree, gender and experience variables (independent variables). It shows regressions on Panel A (Higher Education Graduates) and Panel B (High School
Graduates without further vocational training as the Control Group). Standard errors (reported in parentheses) are heteroscedasticity-robust. Languages and
Cultural Studies is the omitted subject, male the omitted sex and University of Applied Sciences the omitted degree variable. For a detailed description of the
v
ariables see Table 1.
D
ependent variable: Natural logarithm of monthly mean net income
Higher Education Graduates (Panel A)
Full regression Subgroup regressions
V
ariables
University, Male
University,
Female
Univ. of Applied
Science, Male
Univ. of Applied
Science, Female
C
ontrol
Group
(Panel B)
Constant 7.129*** 6.993*** 7.132*** 6.948*** 7.218*** 6.629***
(0.012) (0.014) (0.014) (0.018) (0.019) (0.016)
Exp 0.104*** 0.125*** 0.074*** 0.107*** 0.036*** 0.161***
(0.003) (0.003) (0.004) (0.003) (0.005) (0.010)
Expsq -0.005*** -0.006*** -0.003*** -0.005*** -0.001*** -0.011***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.001)
Exp3 0.000*** 0.000*** 0.000*** 0.000*** 0.000** 0.000***
(
0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Exp4 0.000***
(
0.000)
L
aw / Econ. / Social Studies 0.226*** 0.353*** 0.099*** 0.365*** 0.168***
(0.003) (0.003) (0.005) (0.015) (0.008)
Math. / Natural Sciences 0.108*** 0.150*** -0.003 0.355*** 0.158***
(0.005) (0.004) (0.011) (0.018) (0.042)
Medicine 0.27*** 0.335*** 0.213*** 0.423*** 0.166
(0.010) (0.011) (0.016) (0.150) (0.125)
Agriculture -0.181*** -0.044** -0.203*** -0.055** -0.298***
(0.014) (0.022) (0.034) (0.022) (0.043)
Engineering 0.077*** 0.115*** -0.070*** 0.264*** 0.053***
(0.004) (0.004) (0.013) (0.014) (0.012)
Art -0.177*** -0.212*** -0.190*** 0.031 -0.076**
(0.012) (0.016) (0.021) (0.030) (0.033)
Sex 0.04*** -0.386***
(0.015) (0.023)
InterExpSex -0.06*** 0.086***
(0.004) (0.014)
InterExpsqSex 0.003*** -0.011***
(0.000) (0.002)
InterExp3Sex 0.000*** 0.000***
(0.000) (0.000)
Degree -0.079***
(0.014)
InterExpDegree 0.024***
(0.003)
InterExpsqDegree -0.001***
(0.000)
InterExp3Degree 0.000*
(0.000)
N 17180 6351 3928 5011 1890 1416
N classes 914 238 237 209 230 72
Prob > F 0 0 0 0 0 0
R
2
75.7% 78.3% 55.6% 71.9% 34.1% 83.5%
*** Significant at 0 to 1 percent level, ** Significant at 1 to 5 percent level, * Significant at 5 to 10 percent level, others: Significant at above 10 percent level

34
Table 5
I
RR and NPV of Different Private Higher Education Investment Alternatives
This table presents the Internal Rate of Return (IRR) and the Net Present Value (NPV) associated with
private investments in different forms of higher education assuming full-time work. We differentiate for
gender, degree and subject and consider two different financing regimes: full public financing and full
private financing of actual higher education production cost. Our NPV calculations assume a discount
rate of 4%. For a detailed description of the underlying assumptions see Appendix A.
in thousand Euro, percent
Public Financing Private Financing
Only Opportunity Cost
Opportunity Cost and
Cost of Study
Subgroups NPV (4%) IRR NPV (4%) IRR
University, Male
Languages and Cultural studies 2.7 4.2% -19.5 3.0%
Law / Econ. / Social studies 242.8 13.2% 225.7 11.6%
Math. / Natural sciences 94.5 8.4% 61.8 6.4%
Medicine 228.6 12.8% 182.5 9.4%
Agriculture -21.7 2.6% -68.4 0.6%
Engineering 71.8 7.5% 37.7 5.5%
Art -105.8 -16.0% -146.4 -16.3%
University, Female
Languages and Cultural studies 68.7 8.6% 46.4 6.6%
Law / Econ. / Social studies 118.8 11.0% 101.7 9.2%
Math. / Natural sciences 67.4 8.5% 34.8 5.8%
Medicine 182.8 13.6% 136.7 9.2%
Agriculture -19.7 2.1% -66.3 -0.6%
Engineering 36.4 6.7% 2.3 4.1%
Art -14.3 2.7% -54.9 0.2%
University of Applied Sciences, Male*
Languages and Cultural studies -70.3 -9.4% -87.8 -9.9%
Law / Econ. / Social studies 147.3 12.1% 133.2 10.5%
Math. / Natural sciences 140.5 11.8% 124.3 10.0%
Agriculture -96.9 N/A -117.5 N/A
Engineering 79.3 9.0% 57.4 7.0%
Art -54.6 -3.0% -76.8 -4.1%
University of Applied sciences, Female*
Languages and Cultural studies 21.1 6.4% 3.7 4.3%
Law / Econ. / Social studies 100.2 12.4% 86.1 10.1%
Engineering 44.6 8.5% 22.7 5.8%
Art -10.6 2.5% -32.8 0.3%
* Subgroups with less than 90 observations are not displayed

35
Table 6
Results from NPV/IRR-Regression Analysis
This table reports coefficient estimates from OLS regressions relating the NPV and IRR of the different forms of higher education presented in
T
able 5 as dependent variables to gender, subject, degree and cost variables (independent variables). Standard errors (reported in parentheses)
are heteroscedasticity-robust. Law, Economics and Social Studies is the omitted subject, male the omitted sex and University of Applied Sciences
the omitted degree variable. The NPV figures are calculated with a discount rate of 4%. Regressions including the cost variable do not consider the
subject Medicine. For a detailed description of the variables see Table 1.
D
ependent variable: Net Present Value Dependent variable:Internal Rate of Return
NPV-Regressions IRR-Regressions
V
ariables I II III IV I II III IV
Constant 144.74*** 162.47*** 127.68** 172.90*** 0.099*** 0.125*** 0.107** 0.137**
(32.43) (26.13) (47.22) (47.42) (0.022) (0.013) (0.025) (0.046)
Sex -2.71 -43.22** -40.62* 0.040* 0.066* 0.077*
(20.43) (19.62) (21.58) (0.022) (0.033) (0.040)
Degree 17.80 22.87 38.01 0.006 0.002 0.016
(
22.97) (17.74) (34.55) (0.026) (0.025) (0.043)
L
anguages / Cultural St. -146.72*** -207.69*** -201.16*** -0.097** -0.135*** -0.135***
(43.00) (37.12) (32.45) (0.039) (0.044) (0.043)
Math. / Natural Sciences -54.89 -62.49* -50.29 -0.020 -0.053 -0.045
(
43.97) (34.31) (39.72) (0.030) (0.033) (0.051)
Medicine 44.53 42.00 0.008 -0.028
(
41.80) (27.52) (0.026) (0.033)
A
griculture -201.83*** -209.42*** -183.93** -0.134*** -0.166*** -0.146
(39.93) (41.40) (83.31) (0.040) (0.045) (0.099)
Engineering -94.27** -94.27*** -78.95* -0.042* -0.043*** -0.029
(34.04) (25.09) (37.90) (0.023) (0.009) (0.051)
Art -198.61*** -254.10*** -233.31*** -0.156*** -0.194*** -0.180**
(41.79) (45.42) (55.49) (0.043) (0.047) (0.063)
IntersexLCS 121.94*** 119.34***
(33.28) (30.78)
IntersexLES -0.075** -0.086*
(0.034) (0.042)
IntersexEng -0.073* -0.084*
(0.035) (0.042)
IntersexArt 110.99** 108.38**
(42.61) (40.80)
Cost -3.29** -1.19 -0.002 -0.001
(1.33) (2.90) (0.001) (0.004)
N 24 24 22 22 24 24 22 22
Prob > F 0 0 0.022 0 0.006 0 0.146 0
R
2
83.1% 91.6% 17.3% 89.1% 68.9% 74.0% 9.8% 73.0%
*** Significant at 0 to 1 percent level, ** Significant at 1 to 5 percent level, * Significant at 5 to 10 percent
level, others: Significant at above 10 percent level

36
Appendix A: List of underlying assumptions
•
Internal rate of return (and net present value) calculated as the relevant decision criteria for
the private investment in different forms of higher education
•
Only consideration of monetary effects
•
High school graduates without further (vocational) education considered as the control group
to calculate the opportunity cost of higher education
•
Entry into higher education assumed to be at the age of 20 (begin of working life for control
group)
•
Begin of working life assumed to be at the age of 26 for University graduates and at the age
of 25 for graduates from a University of Applied Sciences (no working activity while
pursuing higher education assumed)
•
Degree-adjusted duration of studies
D
of 6 years to obtain a University degree and 4.5 year to
graduate from a University of Applied Sciences
•
Working life assumed to last until 55 years
•
Annual expected average income for full-time workers
eskj
w
,,,
per year of experience,
gender, degree and subject estimated with OLS-regression analysis for both higher education
graduates and high school graduates without further post-secondary education
o average net monthly income (following Murphy and Welch (1990)) taken as
dependent variable and experience as well as gender-, degree- and subject dummy
variables as independent variables
o Income classes approximated by mid-point of class
o Slopes of income profiles invariant to subject, but differing by gender and degree
o Labor supply decisions assumed to be exogenous in the determination of the wage
o No adjustment for differences in labor force participation between higher education
graduates and high school graduates without further post-secondary education
•
Scenario calculation calculates the private return on investment assuming full private recovery of
education cost in addition to current taxation

37
•
Constant annual wage growth rate
g
of 2% assumed to transform cross-sectional wage into wage
profiles; robustness checks with growth rate of 1.5 and 2.5% performed
Appendix B: Cost of Study
We use detailed cost data per subject and degree and information on the number of students from the
Statistisches Bundesamt ("Fachserie 11") to assess the annual cost of study per student.
Building on the methodology suggested by Luedeke and Beckmann (1998) to determine the costs of
production in the higher education sector, we calculate the monetary cost for provision of higher
education in subject
j
and degree k using the following components:
Net production cost ("Grundmittel") = wages for university personnel + upkeep of buildings + other
expenditure – assorted fees and receipts ("Verwaltungseinnahmen") – research grants
("Drittmittel")
30
.
We adjust the published figures for the cost of central departments and the share of research.
We exclude research expenses that are included in the cost base, since (fundamental) research serves
to a large extent public purposes and is not directly attributable to teaching, by applying the official
research coefficients introduced by the Statistisches Bundesamt in 1995 (see Hetmeier (1998), pp.
153).
We follow Luedeke and Beckmann (1998) and allocate the expenses to the central departments
among the different subjects according to the number of students in the respective subject. For the
subject medicine, we also present a scenario calculation where we exclude the cost for central
departments of university hospitals that render medical studies expensive and are not directly related
to obtaining education. To get the annual cost of study
kj
c
,
, we divide the total cost in the respective
subject and degree by the number of students.
31
The following table presents the annual and total cost of study per student.
30
In contrast to Luedeke and Beckmann, however, we do not take into account imputed cost (e.g.,
depreciation).
31
Doing so, we do not consider the cost of dropouts.

38
Table 7
Cost of study
This table presents the annual and total cost of study measured through net production cost per student in
thousands of Euro as of 2003, which is calculated using wages for university personnel, upkeep of
buildings and other current expenditure adjusted for assorted fees and receipts and research grants for
the different subjects and degrees. We also exclude research cost by applying the official R&D
coefficients from the Statistisches Bundesamt and divide the cost for the central departments proportional
to the number of students in the subjects. We assume a duration of study of 6 years for University
degrees and 4.5 years for a degree at a University of Applied Sciences.
in thousands of Euro
Subgroups
Number of
students
Annual Cost
of study
Annual Cost
of study per
student
Total cost
of study
per student
University
Languages and Cultural studies 435,002
1,678,325 3.9 23.1
Law / Econ. / Social studies
386,502
1,148,470 3.0 17.8
Math. / Natural sciences
292,554
1,653,869 5.7 33.9
Medicine* 94 225 2,974,022 31.6 189.4
Without central departments of hospitals 94,225 752,023 8.0 47.9
Agriculture
22,121
178,649 8.1 48.5
Engineering
134,228
793,231 5.9 35.5
Art
66,035
463,915 7.0 42.2
University of Applied Sciences
Languages and Cultural studies 12,433 50,057 4.0 18.1
Law / Econ. / Social studies
211,793
688,716 3.3 14.6
Math. / Natural sciences
64,494
242,332 3.8 16.9
Agriculture
17,031
80,888 4.7 21.4
Engineering
183,643
930,002 5.1 22.8
Art
17,933 91,999 5.1 23.1
* Only human medicine including central departments of hospitals