You are on page 1of 9

Economics and Human Biology 7 (2009) 191–199

Contents lists available at ScienceDirect

Economics and Human Biology


journal homepage: http://www.elsevier.com/locate/ehb

The nonlinear link between height and wages in Germany, 1985–2004


Olaf Hübler
Leibniz Universität Hannover, Institut für Empirische Wirtschaftsforschung, Königsworther Platz 1, 30167 Hannover, Germany

A R T I C L E I N F O A B S T R A C T

Article history: Based on data of the German Socio-Economic Panel, this article investigates the
Received 18 June 2009 relationship between height and wages by gender. Unlike previous investigations, which
Accepted 18 June 2009 have been limited to an examination of linear effects, this one finds that height influences
on wages are curvilinear, and more so for men than for women. More specifically, it finds
JEL classification: that women who are shorter than average and men who are somewhat taller than average,
J31 but not among the tallest, enjoy significant wage advantages. Furthermore, using Blinder’s
J71
decomposition to determine two components of wage differences, we find that these
Keywords: differences can be partitioned into an endowment component and unexplained influences
Height (discrimination). There is a difference between the public and private sectors and between
Wages men and women as to the degree of the latter effect. This investigation supports the
Men hypothesis that short and very tall men employed in the private sector are disadvantaged
Women the most. The outcome for women is less robust than for men.
Nonlinearity ß 2009 Elsevier B.V. All rights reserved.
Discrimination
Private and public sector
Germany

1. Introduction Two contrasting explanations have been offered for the


fact that shorter employees are paid less than taller ones:
In recent empirical investigations, aspects of appear- that taller individuals are more productive and that shorter
ance have extended the traditional Mincer earnings workers are subject to discrimination. The former expla-
equation. In particular, friendship relations in school,1 nation is based on hypotheses (Persico et al., 2004; Case
beauty,2 weight,3 height,4 and body mass index5 have been and Paxson, 2008), on one of the following effects: taller
advanced as potential wage determinants. All analyses that people have a physical advantage, are healthier, better
incorporate height confirm that this is a statistically educated, have greater cognitive abilities or self-select into
significant determinant. high-paying occupations.
First, tall people have physical advantages in the
competition for resources. This used to be more relevant
E-mail address: huebler@ewifo.uni-hannover.de.
than it is today, because physical labour no longer
1
Galeotti and Mueller (2005). correlates with high earnings. In any case, taller people
2
Hamermesh and Biddle (1994), Biddle and Hamermesh (1998) and are more disease-resistant and therefore more productive
Mobius and Rosenblat (2006).
3
than shorter people (Waaler, 1984; Riley, 1994).6 In
Averett and Korenman (1996), Cawley (2004), Cawley and Danziger
(2005), Cawley et al. (2005), Heineck (2006) and Cawley et al. (in press).
4
Persico et al. (2004), Heineck (2005), Case and Paxson (2008) and
6
Komlos (2009). However, the height–health association is a controversial issue in the
5
Kennedy and Garcia (1994), Behrman and Rosenzweig (2004), Conley literature. Samaras and Elrick (2002) and Samaras et al. (2003, 2004)
and Glauber (2005), Brunello and D’Hombres (2007), Heineck (2006), claim that it is shortness that offers physical advantages. In contrast to
Grewe (2008), Johansson et al. (2009) and Villar and Quintana-Domeque Waaler’s (1984) results theirs show that shorter people live longer than
(2009). taller people.

1570-677X/$ – see front matter ß 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.ehb.2009.06.003
192 O. Hübler / Economics and Human Biology 7 (2009) 191–199

general tall people also possess more authority than private attorneys are better able to attract and retain
shorter ones and are more willing to take risks (Dohmen clients than are their shorter colleagues (Biddle and
et al., 2005). This could provide further advantages for Hamermesh, 1998). Tall men – Sunder (2006) – have an
them and may be reflected in higher wages. advantage when it comes to attracting mates. Further-
Second, one can argue that the correlation between more, in today’s society tall persons are presumed to be the
height and wages is spurious but height is not the cause of dominant ones; and when it comes to colleagues, tall is
economic success. For example, good nutrition in the early preferred to short (Rubenstein et al., 1998). This argument
stages of development of children promotes growth and leads to the hypothesis: that to favour tall persons over
fitness. Well-educated and wealthy parents are more likely short persons constitutes discrimination, i.e., that society
to be well informed about good nutrition than less retains an established height prejudice long after the
privileged adults, and they are also more able to afford physical advantages of tallness have lost their relevance in
goods and services, such as extra tuition, that may help the marketplace.
their children succeed later in life. Moreover, if a tall The purpose of this paper is to reconsider the relation-
generation is successful thanks to specific abilities and if ship between height and wages for men and for women
the next generation inherits both the above-average height using German data. The analysis is focussed on nonlinear
and the above-average abilities, this would induce a links.
statistical correlation between height and wages.7
Third, height during adolescence influences one’s 2. Data and descriptive results
chances of accumulating skills at this early stage of
development. Teenagers who are stigmatised because they The empirical analysis is based on representative
are short may have problems developing interpersonal micro-data collected from 1985 to 2004 by the German
skills and a positive self-image, and they may even find Socio-Economic Panel (GSOEP).8 This investigation is
themselves excluded from group activities. They tend to be modelled on that of Schultz (2002) who considers only
less easygoing and more risk-averse than their taller full-time and part-time employees aged 25–55. For
counterparts. Moreover, the phenomenon of the self- technical details on the GSOEP see Haisken-DeNew and
fulfilling prophecy may take effect. Children who stigma- Frick (2003). The panel, formulated in 1984, had a sample
tized because they are short tend in one of two opposing size of 5921 households and 12,290 individuals. In 1991
directions; either they choose not to invest in long-term the sample was supplemented by data from the former
efforts, because it seems futile to try, or they work all the East Germany. In 2004 the sample contained 11,796
harder, in order to achieve parity with taller children. In households and 22,019 individuals. Height was among the
contrast, tall high-school teenagers are associated with data collected in the 2002 and 2004 surveys. As height is
fewer symptoms of depression (Rees et al., in press) and time-invariant for people between the ages of 25 and 55, it
tend to think of themselves as leaders, a tendency that is possible to link this information to the preceding and
persists after they enter the labour force. They also develop subsequent waves. However, many of the 2002 and 2004
better interpersonal skills, which will be important for height data are slightly different from each other, and so
success in their working lives. Kuhn and Weinberger the 2 years’ average is used. The mean difference of height
(2005) emphasize that high-school leadership is connected (x) between 2002 and 2004 is Dx̄ ¼ 0:0146 cm
with stamina, energy, the work ethic, persistence and (sx = 2.2214 cm) for men and women together. Those
motivation – all character traits that are highly valued in people who were not in the age group of 25–55 in one or
the labour market making it easier for taller individuals to both of these 2 years are excluded from the following
reach positions in the upper echelons of corporate analysis (sample 1). Because the height differences may be
hierarchy. due to misreporting, a sensitivity test is used, restricting
Fourth, Case and Paxson (2006, 2008) argue that height the sample to subjects whose height reports of 2002 and
is positively associated with cognitive ability, which is 2004 are within 2 cm of each other (sample 2). In this case
rewarded in the labour market, and that taller individuals the number of observations decreases (less so for men than
are more likely than shorter ones to select higher paying for women) but the results are hardly affected. For men,
occupations, that require more advanced verbal and the hypothesis of the same average height in the two
numerical skills and greater intelligence than do lower samples cannot be rejected. Also for women we find nearly
paying occupations. the same average height in the two samples. The difference
Other researchers argue that height can also be is less than 1 mm.
interpreted as a distinct element of a person’s appearance, The simple correlation coefficient between height and
and therefore that height is directly correlated with wages per month based on a linear relationship is only
earnings. One’s general appearance, of which one’s height rw = +0.070 for women and rm = +0.132 for men. This weak
is an element, can affect confidence and communication, correlation may be due to a nonlinear relationship.
thereby influencing productivity. For example, taller Preliminary analysis supporting this hypothesis appears
in Fig. 1(a) and (b).
The graphs depict the expected logarithms of real
7
Garcia and Quintana-Domeque (2006) argue in a similar way when
gross monthly earnings if only a height variable is used
they discuss the associations between body mass index and labour
market outcomes. A third factor may be the reason that BMI and wages
8
correlate. http://www.diw.de/english/soep/26636.html.
O. Hübler / Economics and Human Biology 7 (2009) 191–199 193

Fig. 1. Height effects on (log)wages per month estimated with a fractional polynomial height function. (a) Men. (b) Women.

as the determinant of wages. Height is grouped into nine mean height of 166 cm.9 Part-time work might be
5-cm intervals. For each of these intervals the means responsible. Therefore, we split the sample into full- and
of height and wages per month are calculated based part-time workers. The wage effects of the former (dashed
on all individuals in GSOEP aged 25–55 in the period 1985– line) have nearly the same course as those for the
2004. In other words, the average height (in 2002 and entire sample (thick solid lines). For female part-
2004) is compared to all the wages recorded for the time workers the nonlinear relationship is somewhat
individual in the survey. The methodological starting point stronger (thin solid line). Part-time work for men can be
is a general fractional polynomial (Royston and Altman, neglected because only 3% of male employees do not work
1994). The order of the polynomial is then determined full-time.
with the aim of an optimal nonlinear fit between
the polynomial and the data. Fig. 1(a) demonstrates
9
an obvious nonlinear second-order relationship (confi- The average real gross wage per month within the optimal height
dence interval). Men reach the maximum height–wage interval for males (185–190 cm) is 2072 euro. 1861 and 1734 euros are
the analogous figures for the interval 175–180 and 200–205 cm,
effect roughly 10 cm above the average height, 178 cm. respectively. For women we have determined 1177 euro for the height
However, for women (Fig. 1(b)) a weaker curvilinear interval 165–170 cm, 1016 euro for 150–155 cm and 1085 euro for 180–
relationship is found with maximum at 3 cm above the 185 cm.
194 O. Hübler / Economics and Human Biology 7 (2009) 191–199

Table 1
Estimates of height coefficients – dependent variable: log wages per hour.

(1) Men (2) Women (3) Men (4) Women (5) Women
*** *** **
HEIGHT 0.0025 0.0010 0.7718 1.0505 0.0709**
(0.0007) (0.0009) (0.2620) (0.4189) (0.0309)
HEIGHT2 0.0043*** 0.0061** 0.0002**
(0.0015) (0.0025) (0.0001)
HEIGHT3/103 0.0081*** 0.0118
(0.0028) (0.0092)

N 22,836 13,918 22,836 13,918 13,918


R2 0.436 0.365 0.436 0.366 0.366
F_1 66.69*** 4.79** 32.33*** 9.03*** 10.81***
F_2 15.12*** 11.15*** 16.85***
F_3 5.53**

Note: The standard errors in parentheses are estimated by the cluster approach, which takes into account intragroup correlation, but the observations are
independent across groups (individuals). Wages per hour are measured in real terms and euros. Control variables in specifications (1)–(5): SCHOOLING,
TENURE, TENURE2, EXPERIENCE (actual experience), EXPERIENCE2, NBULA (dummy; =1, if new German federal state (Bundesland)), 20 TIME DUMMIES (for
1985–2004). F tests of joint significance of the height variables (H0: height effects on wages are zero; H1: height effects are linear (columns (1) and (2)), of third-
order (columns (3) and (4)) or of second-order (column (5)) are presented in the last but two line (F_1), and the F tests of nonlinear height effects (H0: height
effects on wages are linear; H1: height effects are of third-order (columns (3) and (4)) or of second-order (column (5)) are presented in the penultimate line (F_2).
F_3 presents the F test statistic where the second-order approach (H0) is tested against the third-order approach (H1). Source: GSOEP 1985–2004.
*
Significance at the 0.10 level.
**
Significance at the 0.05 level.
***
Significance at the 0.01 level.

Next, we investigate the height effects on (log) wages hypothesis. For men, all height effects are significant
per hour and other wage determinants are incorporated. (Table 1, column (3)), while for women, two of the three
Nonlinear relations are analysed in greater detail using an coefficients are (Table 1, column (4)). Second- and fourth-
extended Mincerian earnings function. order height functions were also determined. (With one
exception these are not presented in the tables.) F tests at
the bottom of Table 1, column (3) show that the height
3. Econometric results effects are jointly significant (F_1 = 32.33) and that the
third-order height approach is preferred to the linear one
3.1. Linear or nonlinear height wage premium?
for men (F_2 = 15.12). For women the results are less
We first estimate wage functions, using pooled data obvious. The third- and the second-order approach
from 1985 to 2004 stratified by gender, and height as an (Table 1, columns (4) and (5)) are superior to the linear
explanatory variable. The individual height is time- model (F_2 = 11.15 and 16.85). However, the decision
invariant, but wages vary between 1985 and 2004. In all between the second- and the third-order model is
tables, the dependent variable is the log of real gross wages ambiguous. At a = 0.01 the F test (F_3 = 5.53) cannot reject
per hour. Control variables comprise not only schooling, the null hypothesis (H0: second-order height function; H1:
tenure and experience but also time dummies and a third-order height function) while at a = 0.05 we should
regional dummy. Time dummies are incorporated because prefer the third-order height function. The maximal height
individual invariant time effects of height on wages are effect on wages for men is at 191.5 cm (based on Table 1,
likely. The regional dummy is necessary because without column (3)) and for women at 159.9 (based on Table 1,
any causality both wages and heights are greater in column (4)).11
Western Germany than in Eastern Germany.10 As a first robustness check we run additional regres-
Among men, a 10-cm height increase is associated with a sions with a part-time dummy variable as regressor and
2.5% wage increase (Table 1, column (1)). Among women, separate estimates for full-time and part-time workers are
the estimated height coefficient is smaller (Table 1, column
(2)). All of the control variables have the expected effect on
wages, and the coefficients are significantly different from 11
If H0 is preferred at test F_3 in Table 1, i.e. if we decide for the second-
zero. These results conform with those of other empirical order height function for women, the maximal height effect on wages for
investigations. women is at 168.2 cm (based on Table 1, column (5)). A comparison of
If a third-order polynomial of height is used as the these results with those in Fig. 1(a) and 1(b) demonstrates that the height
with maximal wage effects increases for men and decreases for women if
independent variable, the results confirm the nonlinear log wages per month are replaced by log wages per hour and control
variables are incorporated. In Fig. 1(a) and 1(b) the maximum is roughly
reached at 188 cm for men and at 169 cm for women. Indirect gender-
10
Komlos and Kriwy (2003) find that West Germans are taller than East specific wage effects are a possible explanation of these differences. If
Germans and that there is a tendency for East German males, but not of these influences are suppressed as in Fig. 1 the wage–height function is
females, to catch up with their West German counterparts. Hence, we shifted. For example, a linear schooling (S)  gender (SEX = 1, if man;
analyse the differences between the two German regions in more detail. SEX = 0 otherwise) relationship (S = c0 + c1SEX + u) could induce this
The appendix, Table A.1, Fig. A.1, presents the nonlinear height–wage result. We observe for men a shift to the left (from Table 1, column (3) to
relationship separated for gender and the two regions analogously to Fig. 1(a)) and for women to the right. Therefore we conclude that c0 > 0
column (3) and (5) in Table 1. and c1 < 0.
O. Hübler / Economics and Human Biology 7 (2009) 191–199 195

Table 2
Estimates of log wages per hour (in EUR) with height-interval dummies as determinants.

ln w (1) Men (2) Women

Coef. Std. Err. Coef. Std. Err.

HEIGHT  150 0.1297*** 0.0304


150 < HEIGHT  155 0.0488*** 0.0156
155 < HEIGHT  160 0.0056 0.0082
160 < HEIGHT  165 0.0017 0.0073
Height  165 0.0099 0.0125
165 < HEIGHT  170 0.0352*** 0.0073
170 < HEIGHT  175 0.0054*** 0.0059 0.0082 0.0088
175 < HEIGHT  180 0.0056 0.0127
***
180 < HEIGHT  185 0.0261 0.0061 0.0377 0.0412
185 < HEIGHT  190 0.0383*** 0.0068 0.0918 0.0727
190 < HEIGHT  195 0.0559*** 0.0124
195 < HEIGHT  200 0.0095 0.0210
HEIGHT > 200 0.0843 0.0663

N 22,836 13,918
R2 0.436 0.364

Note: The standard errors are estimated by the cluster approach. The same specification as in Table 1, columns (1) and (2) is used, where height is
substituted by height-interval dummies. The reference group is always the height interval with the average height of the sample under consideration.
Source: GSOEP 1985–2004.
*
Significance at the 0.10 level.
**
Significance at the 0.05 level.
***
Significance at the 0.01 level.

Fig. 2. Adjusted height effects on (log)wages per hour estimated by LOWESS and based on estimates of Table 2.

executed. (The results are not presented in the tables.) The into the wage function; the estimates are stratified by
former check demonstrates that the height coefficients do gender. The results in Table 2 are similar to the ones with
not change compared with those in Table 1. Separate the estimates in which height is modelled as a third-order
estimates for full-time and part-time workers show that polynomial. Based on these estimates Fig. 2 demonstrates
the full-time worker estimates are nearly the same as those the nonlinear course of height effects on log wages per
for the entire sample (full-time and part-time workers). hour.12 Table 2 supports the hypothesis that those men
The estimates for part-time workers differ. However, in earn most whose height is well above the mean, whereas
this case all height effects are insignificant due to the small
sample.
As a second robustness check, height is modelled as a 12
We smooth the estimated height effects on wages per hour in Table 2
spline function; i.e., height is divided into 5-cm intervals, by means of the lowess procedure, an outlier robust technique (Cleveland,
and dummy variables for these intervals are incorporated 1979).
196 O. Hübler / Economics and Human Biology 7 (2009) 191–199

among women the maximum is reached just below the women, the opposite is the case for tenure and experience
mean. – see Appendix, Table A.2.
Unobserved productivity differences or discrimination
3.2. Endowment and discrimination effects are not negligible, especially for men. Differences amount-
ing to 16.5% in the log wage per hour of the men in groups
We next decompose the effect into an endowment and A1 and A2 are explained by the weighted differences in
a discrimination effect. For decomposition A the sample is coefficients (D). This figure for women is smaller, 9.4%, but
divided into two groups (A1 and A2) by height. The first not negligible.
group (A1) comprises those whose height is within 10 cm
of the ideal (from an economic standpoint) height: of 3.3. Differences between the private and public sectors
191.5 cm for men and 159.9 for women (Table 1, columns
(3) and (4)). The second group (A2) comprises those If height discrimination in the workplace,13 affecting
employees whose heights are outside the interval of the wages, exists (Table 3 lends support to this hypothesis),
first group (i.e., who display the low-wage effects employers and customers may be responsible. This
experienced by those who are much shorter or much hypothesis is tested via a procedure, suggested in
taller than those who display the largest wage effects). another context by Biddle and Hamermesh (1998)
Blinder’s (1973) decomposition procedure is as follows: who distinguish between employees and the self-
employed and between the public and private sectors.
This study leaves the self-employed out of consideration
  
ln w̄1  ln w̄2 ¼ ðx̄1  x̄2 Þ0 b1 þ ðb1  b2 Þ0 x̄2 þ ðb0;1  b0;2 Þ because of the small number of observations available in
¼: E þ D ¼: E þ C þ U ¼: R (1) our sample. From an economic viewpoint wage differ-
ences between employees of group A1 and A2 (see
where the index 1 denotes group A1 and 2 is group A2. Table 3) can be expected in the private sector if members
ln w̄k is the average of estimated (log)wages per hour in of one group have more authority, are less risk-averse or
group k = 1; 2, b0 is the estimated constant term, b* is the if employers see that (other things being equal)
vector of estimated coefficients of regressors and x̄ is the customers prefer to interact with moderately tall sales-
vector of average regressors. These are the control men and with moderately short saleswomen. Then these
variables in columns (1) and (2), Table 1. E, the endowment favoured employees will generate higher earnings not
component, is based on differences in the averages of the only for themselves but also for their employers. In the
two groups’ wage determinants. D is split into the effect C, public sector this reasoning is largely irrelevant, because
induced by the differences in the coefficients of wage the profit motive is absent.
determinants, and the unexplained effect U, which is If the height effect on wages is more important in the
measured by the difference between the intercept of group private sector than in the public sector, this can be
1 and the intercept of group 2. Although Blinder denotes D interpreted as evidence of discrimination on the part of
as discrimination effect we should emphasise that this customers or of unobserved productivity. To test this,
term may also be due to omitted variables, especially to separate regressions are run for the private and public
unobserved productivity differences. sector. Again, the specification from Table 1, columns (1)
The results in Table 3 indicate that the wage differences and (2) is used. Table 4 presents the estimates, where
can be largely explained by differences in the endowment height is measured by a dummy variable HEIGHT_A14
(E). More precisely, men who benefit from a positive height using decomposition A in Table 3. Tests about the
effect on their wages are better educated, but have less job difference in the coefficients on HEIGHT_A between the
experience and less tenure than other men. Among private and the public sector reject evidently the null
hypothesis for men as well as for women. Additionally,
the Biddle–Hamermesh test finds significant height
Table 3 effects for men in the private sector whereas the height
Blinder decomposition (as %) of log wage differences between height coefficient observed in the public sector is very small
groups.
and insignificant. This finding is compatible with the
Men Women hypothesis that discrimination takes place in the private
Explained variation sector.
100  (E/R) 83.5 90.6 The sign for the height coefficient of men in the public
sector is negative but insignificant. Perhaps such an effect
Unexplained variation
100  (D/R) 16.5 9.4 exists because this group includes men in the low-effect

Note: The decomposition is based on the first specification in Table 1,


columns (1) and (2) without height; R = E + D; E = endowment component
13
of log wage differences between group 1 and 2 (see Eq. (1)); Rosenberg (in press) finds that there exists discrimination against
D = unexplained component due to unobserved productivity differences short men and tall women in employment in the United States though
or discrimination (see Eq. (1)). A positive sign of E/R indicates advantages existing federal law already prohibits against height discrimination.
14
in the endowment component to employees within group 1 (A1: height The modelling of height by the dummy HEIGHT_A (=1, if member of
with high-wage effects) compared with those in group 2 (A2: height with group A1; =0, if member of group A2) means a nonlinear specification. The
low-wage effects). A positive sign of D/R indicates that the employees in choice of HEIGHT_A assumes implicitly – as consequence of the results in
group 2 suffer discrimination relative to those in group 1. Source: GSOEP Table 1 – that short and very tall male employees (A2) have smaller height
1985–2004. effects on wages than those from A1 members.
O. Hübler / Economics and Human Biology 7 (2009) 191–199 197

Table 4 plausible reason is that short male managers are over-


Estimates of log wages per hour in EUR with height effect in the private
represented in the public sector and tend to discriminate
and public sectors.
in favour of shorter workers. The right hand estimate
Private sector Public sector of Table 4 says us that women of group A1 (short but not
Men Women Men Women too short) earn more on average than those from A2
HEIGHT_A 0.0563*** 0.0054 0.0053 0.0316*
(very short or tall) if we control for the major wage
(0.0106) (0.0169) (0.0067) (0.0175) determinants.
N 17,208 8445 5394 5290
R2 0.461 0.325 0.424 0.392
4. Discussion
Note: Standard errors estimated by the cluster approach are in
parentheses. Control variables are the same as in Table 1, columns (1)
and (2). HEIGHT_A is a dummy (=1, if height is within 10 cm of the height
In contrast with previous studies, which concluded
with the strongest wage effects (191.5 cm for men and 159.9 cm for that there is a positive linear relationship between wages
women) – see decomposition A in Table 3). Source: GSOEP 1985–2004. and height, this paper provides evidence for the pre-
*
Significance at the 0.10 level. valence of nonlinearities. The maximum height effect is
**
Significance at the 0.05 level.
*** experienced by moderately tall men at 191 cm and
Significance at the 0.01 level.
moderately short women at 160 cm. The results of the
latter are sensitive to the modelling of height effects. This
pattern is due not only to endowment differences but also
height range who have moved here from the private sector, to unobserved factors related to productivity and dis-
calculating that they will suffer less discrimination in the crimination. 16% of the difference in wages can be
public sector. This self-selection process would make for explained by discrimination among men and 9% among
more type-A2 managers and workers in the public sector women. Discrimination against short and very tall men in
than in the private sector. Type-A2 managers favour type- the private sector and against tall and very short women in
A2 subordinates. Indeed, in the sample the average height the public sector is confirmed. The coefficient on
in the private sector (mean(HEIGHT)_private = 174.39) HEIGHT_A = 0.0316 is significant at the 10% level. This
exceeds that in the public sector (mean(HEIGHT)_pub- means that members of group A1 (moderately short
lic = 172.34). The same phenomenon is observed among women) earn more than those in group A2 (very short and
employees in managerial positions – MPOS (mean tall women). This result is interpreted as discrimination
(HEIGHT)_private_MPOS = 178.36; mean(HEIGHT)_public_ against members of group A2 because the estimation
MPOS = 175.51). controls for the major wage determinants (endowment)
The estimates in Table 4 support the hypothesis of and we cannot see that there exist unobserved produc-
discrimination against not only short but also very tall tivity advantages of group A1.
men employed in the private sector. In addition, height is The theoretical approaches presented in the litera-
associated with productivity effects, a fact that becomes ture, based as they are on productivity arguments, fail to
apparent when the private sector is split. Research into explain neither this evidence of nonlinear relationship
individual industries (results not reported here) reveals nor the difference in the effect for men and women. For
that impressive height–wage premiums are paid in example, Case and Paxson (2008) presume in their
sectors such as agriculture, stone quarrying, sand-and- analytical model that cognitive ability is a linear
clay industry, recycling and computing, while extremely function of the unobserved individual’s endowment
low premiums are recorded in the service sector. A and that height is a linear function of the endowment,
plausible interpretation of these findings has yet to be making for a linear relationship between height and
provided. However, the correlation between the height wages.
coefficient and height based on private-industry data15 is Saturation is the most obvious explanation for non-
positive (rm = + 0.16 for men). This outcome demon- linearity. The height effect has a ceiling. This is partially
strates that male employees in industries in which the incorporated by the log modelling of wages. Nevertheless,
height effect on wages is important are taller than there remain additional curvilinear effects. Why is the
average. Thus, the wage premium could be interpreted as gender difference for saturation so considerable? Fewer
an incentive for self-selection: moderately tall men women are in managerial positions in which the projection
choose to work in those industries in which their height of authority, power and risk-aversion are important. With
will be a financial advantage. Nevertheless, there remains women beauty compensates height as a factor for
a puzzle why some industries offer a premium and others employment. Furthermore, social norms and discrimina-
do not. tion are responsible; two supplements should be men-
In conformity with Sections 3.1 and 3.2, the results for tioned.
women differ markedly from those for men. In the private
sector the height effects are insignificant among women
15
but large and significant among men. The null hypothesis This means the regression in Table 4 for men in the private sector is
‘‘no discrimination between A1 and A2’’ cannot be estimated separately for 62 private industries (i = 1, . . ., 62) using the
GSOEP data. Eight industries with less than 100 employees in the sample
rejected for women. In the public sector the HEIGHT_A are excluded. Then the 54 resulting coefficients on HEIGHT_A are
coefficient for women is positive, significant at the 0.10 correlated with the average HEIGHT_A, i.e. the ratio of male employees,
level and absolutely larger than in the private sector. A which belong to height group A1, to all male employees in industry i.
198 O. Hübler / Economics and Human Biology 7 (2009) 191–199

First, most men and women prefer the company of the mean. Most men prefer women shorter than them-
those close to the statistical norm, and similar to selves. Sunder (2006) has found that girls who are slightly
themselves. This preference helps to explain the inverse shorter than average have an advantage over other girls in
U-shaped relationship between individual height and the dating game. He argues that this phenomenon may be
wages. Those who are far from the statistical norm tend explained with Nettle’s (2002a) finding that girls with
to be the objects of discrimination but the optimum for below-average height have an earlier onset of menstrua-
men is above the mean. Furthermore, the U-shaped pattern tion, while girls with late sexual maturity will be taller
is reinforced by the fact that health problems are because they are growing longer.
associated with both ends of the height spectrum (Nettle, Not only does height plays a key role in the private
2002a). realm of intimate, marital, and social relationships, but it
Second, there is a gender-specific asymmetric deviation plays a key role in the labour market as well. In the private
from the mean. Women but not men consider above- sector, discrimination takes place in favour of moderately
average height to be a positive characteristic in a mate, so tall men. In the public sector, we observe that moderately
moderately taller boys but not girls have an advantage in short women earn most on average. If this nonlinear
the dating game. The fact that women find good posture relationship between height and wages is due to dis-
attractive correlates with the finding that they prefer men crimination that persists the labour market is not in an
who possess confidence and physical strength. Evolution- efficient equilibrium, and if productivity is not a function of
ary psychology provides an explanation for this prefer- height, then one can conclude that height discrimination
ence; a woman whose mate is endowed with superior has negative economic consequences. If someday one finds
physical traits has a better chance of survival than others, less height discrimination in the marketplace, it may mean
and therefore she does as well. In many cultures tall men that the instincts that underlie personal and professional
are granted a certain social status, which is extended to judgments and choices have undergone an overdue
their mates. Although women are no longer as economic- evolution.
ally dependent on their mates as they used to be, men who
are taller than average are still considered more attractive
than others. Short men are less likely to be married or live Acknowledgements
in a permanent relationship than their taller counterparts
(Herpin, 2005). I thank Christian Dustmann, Knut Gerlach, Robert A.
The curvilinear phenomenon for women is more Hart, Dominik Hübler, John Komlos and three referees for
difficult to explain and less obvious as the empirical many very helpful suggestions and comments.
investigations have shown. Nettle (2002a,b) has found a
similar U-shaped relationship between reproductive suc-
cess and stature: specifically, a significant quadratic
relationship between a woman’s height and the total Appendix A
number of children she bears, whereby the maximal
number of children is correlated with a height well below See Fig. A.1, Tables A.1 and A.2.

Fig. A.1. Adjusted height effects on (log)wages per hour in Eastern and Western Germany estimated by LOWESS stratified by gender and region based on
estimates of Table A.1.
O. Hübler / Economics and Human Biology 7 (2009) 191–199 199

Table A.1 Conley, D., Glauber, R., 2005. Gender, body mass and economic status,
Estimates of height coefficients in nonlinear models for log wages per new evidence from PSID, advances in Health Economics and Health
hour, stratified by gender, Eastern and Western Germany. Services research. The Economics of Obesity 17, 253–278.
Dohmen, T., Falk, A., Huffman, D., Sunde, U., Schupp, J., Wagner, G.G., 2005.
(1) Men (2) Women (3) Men (4) Women Individual Risk Attitudes: New evidence from a Large, representative,
Experimentally-Validated Survey. IZA Discussion Paper 1730.
East East West West Galeotti, A., Mueller, G., 2005. Friendship Relations in the School Class and
HEIGHT 0.6560*** 0.0872** 1.2465*** 0.0885*** Adult Economic Attainment. IZA Discussion Paper No. 1682.
(0.2035) (0.0351) (0.2359) (0.0215) Garcia, J., Quintana-Domeque, 2006. Obesity, Employment and Wages in
Europe. Economics Working Papers 897, Department of Economics
HEIGHT2 0.0038*** 0.0003** 0.0070*** 0.0003***
and Business, Universitat Pompeu Fabra.
(0.0012) (0.0001) (0.0013) (0.0001)
Grewe, J., 2008. Obesity and labor market outcomes in Denmark. Eco-
HEIGHT3/103 0.0720*** 0.0130*** nomics and Human Biology 6, 350–362.
(0.0228) (0.0025) Haisken-DeNew, J., Frick, J., 2003. Desktop Companion to the German
Socio-Economic Panel Study. DIW, Berlin.
N 4529 4513 18,307 8465
Hamermesh, D.S., Biddle, J.E., 1994. Beauty and the labor market. Amer-
R2 0.366 0.411 0.350 0.294
ican Economic Review 84, 1174–1194.
Note: The standard errors in parentheses are estimated by the cluster Hastie, T.J., Tibshirani, R.J., 1997. Generalized Additive Models. Chapman
& Hall, London.
approach. Control variables are the same as in Table 1. Source: GSOEP
Heineck, G., 2005. Up in the skies? The relationship between body height
1991–2004.
* and earnings in Germany. Labour 19, 469–489.
Significance at the 0.10 level.
**
Heineck, G., 2006. Height and weight in Germany, Evidence from the
Significance at the 0.05 level. German Socio-Economic Panel, 2002. Economics and Human Biology
***
Significance at the 0.01 level. 4, 359–382.
Herpin, N., 2005. Love, careers, and height in France, 2001. Economics and
Human Biology 3, 420–449.
Table A.2 Johansson, E., Böckerman, P., Kiiskinen, U., Heliövaara, M., 2009. Obesity
Descriptive statistics – mean and standard deviation in parentheses. and labour market success in Finland: the difference between
having a high BMI and being fat. Economics and Human Biology 7,
Men Women 36–45.
Kennedy, E., Garcia, M., 1994. Body mass index and economic productiv-
Group A1 A2 A1 A2 ity. European Journal of Clinical Nutrition 48 (Suppl.), S45–S53.
Komlos, J., Kriwy, P., 2003. The biological standard of living in the two
SCHOOLING 12.88 11.73 12.11 11.64
Germanies. German Economic Review 4, 459–495.
(2.77) (2.66) (2.56) (2.68)
Komlos, J., 2009. Recent Trends in Height by Gender and Ethnicity in the
EXPERIENCE 14.56 18.31 16.16 15.55 US in Relation to Levels of Income. NBER Working Paper No. 14635.
(8.00) (9.57) (8.09) (9.12) Kuhn, P., Weinberger, C., 2005. Leadership skills and wages. Journal of
Labor Economics 23, 395–436.
TENURE 9.17 10.86 8.79 8.23 Mobius, M.M., Rosenblat, T.S., 2006. Why beauty matters. American
(8.07) (9.08) (7.94) (7.63) Economic Review 96, 222–235.
Nettle, D., 2002a. Women’s height, reproductive success and the evolu-
Note: Age 25–55. Definition of group A1 and A2 see Table 3. The variables tion of sexual dimorphism in modern humans. Proceedings of the
SCHOOLING, EXPERIENCE and TENURE are measured in years. Source: Royal Society B 269, 1919–1923.
GSOEP 1985–2004. Nettle, D., 2002b. Height and reproductive success in a cohort of British
men. Nature 13, 473–491.
Persico, N., Postlewaite, A., Silverman, D., 2004. The effect of adolescent
References experience on labor market outcomes: the case of height. Journal of
Political Economy 112, 1019–1053.
Averett, S., Korenman, S., 1996. The economic reality of the beauty myth. Rees, D.I., Sabia, J.J., Argys, L.M., in press. A head above the rest and
Journal of Human Resources 31, 304–330. adolescent psychological well-being. Economics and Human Biology.
Behrman, J.R., Rosenzweig, M.R., 2004. The returns to birthweight. Review Riley, J.C., 1994. Height, nutrition, and mortality risk reconsidered. Journal
of Economics and Statistics 86, 586–601. of Interdisciplinary History 24, 465–492.
Biddle, J.E., Hamermesh, D.S., 1998. Beauty, productivity, and discrimina- Rosenberg, I.B., in press. Height discrimination in employment. Utah Law
tion: lawyers’ looks and lucre. Journal of Labor Economics 16, 172– Review.
201. Royston, P., Altman, D.G., 1994. Regression using fractional polynomials
Blinder, A., 1973. Wage discrimination: reduced form and structural of continuous covariates: parsimonious parametric modelling.
estimates. Journal of Human Resources 18, 436–455. Applied Statistics 43, 429–467.
Brunello, G., D’Hombres, B., 2007. Does body weight affect wages? Rubenstein, M., Wissman, M., Meyers, C., 1998. Don’t Want No Short,
Evidence from Europe. Economics and Human Biology 5, 1–19. Short Man: The Study of Height, Power, and Mate Selection. Miami
Case, A., Paxson, C., 2006. Stature and Status: Height, Ability, and Labor University.
Market Outcome. NBER Working Paper No. W12466. Samaras, T.T., Elrick, H., Storms, L.H., 2003. Is height related to longevity?
Case, A., Paxson, C., 2008. Stature and status: height, ability, and labor Life Science 72, 1781–1802.
market outcome. Journal of Political Economy 116, 491–532. Samaras, T.T., Elrick, H., Storms, L.H., 2004. Is short height really a risk
Cawley, J., 2004. The impact of obesity on wages. Journal of Human factor for coronary heart decease and stroke mortality? A review.
Resources 39, 451–471. Medical Science Monitor 10, 63–76.
Cawley, J., Grabka, M.M., Lillard, D.R., 2005. A comparison of the relation- Samaras, T.T., Elrick, H., 2002. Less is better. Journal of the National
ship between obesity and earnings in the U.S. and Germany. Schmol- Medical Association 94, 88–99.
lers Jahrbuch 125, 119–129. Schultz, T.P., 2002. Wage gains associated with height as a form of health
Cawley, J., Danziger, S., 2005. Morbid obesity and the transition from human capital. American Economic Review 92. Papers and Proceed-
welfare to work. Journal of Policy and Management 24, 727–743. ings, 349–353.
Cawley, J., Han, E., Norton, E.C., in press. Obesity and labor market out- Sunder, M., 2006. Physical stature and intelligence as predictors of baby
comes among legal immigrants to the United States from developing boomer’s very first dates. Journal of Biosocial Science 38, 821–833.
countries. Economics and Human Biology. Villar, J.G., Quintana-Domeque, C., 2009. Income and body mass index in
Cleveland, W.S., 1979. Robust locally weighted regression and smoothing Europe. Economics and Human Biology 7, 73–83.
scatter plots. Journal of the American Statistical Association 74, 829– Waaler, H.T., 1984. Height, weight, and mortality: the Norwegian experi-
836. ence. Acta Medica Scandinavica 679, 1–59.

You might also like