Family Background, Occupational Choice and Intergenerational

Income Mobility - Germany and the United States compared

Veronika V. Eberharter

University of Innsbruck, Department of Economics* Veronika.Eberharter@uibk.ac.at

presented at SOEP 2006, 3-7 July in Berlin

Abstract:

Using GSOEP-PSIDl data we analyze the impact of human capital characteristics and family background variables on occupational choice and intergenerational income mobility in Germany and the United States. The results do not corroborate more pronounced traditional social role patterns in Germany and a higher social mobility in the American society. For both the countries the results of the multinomial model of occupational choice confirm the human capital hypothesis: occupational choice is gender specific, and more educated persons are more likely to perform a higher occupational status. Family background characteristics account for part of the intergenerational occupational success but they work differently in the countries. Concerning intergenerational income mobility the US sample show a higher intergenerational persistence of economic status and a more pronounced influence of family background characteristics.

Key words: occupational choice, intergenerational issues,

personal income distribution, mobility measurement JEL-Classifications: D90, J24, D31, J60

- 2 -

1 Introduction

The research on intergenerational economic and social mobility sets out to explain

the fact, that the children’s economic and social positions are correlated with that of

their parents. The parental position may affect the economic and social success of

the children’s generation directly, but more likely the intervening variables are

affected by and in turn affect economic and social success. Both, the social and

economic status are correlated with the occupational structure, which intersects with

other structures, such as industry, and differs according a variety of factors including

religion, locality or ethnic group. The occupational status is continuously graded in

regard to economic status, and overlaps on a number of variables like distribution of

income, educational attainment, measured intelligence, ect. (Blau and Duncan 1967).

The analysis on occupational choice adopts the logic of the neoclassical human

capital theory: persons choose occupations that maximize the discounted present

value of potential lifetime earnings, which entail the lowest training costs, and which

offer the lowest discounted present value of expected earnings forgone due to

unemployment (Becker 1964, Mincer 1974; Boskin 1974). Empirical studies on

intergenerational social mobility typically compare the occupational position a person

held when entering the labor market with the occupational position of her father or

her mother using mobility tables. Alternative approaches predict the occupational

choice by factors referring to educational level, previous job history and family

background (Harper and Haq 1997).

In contemporary research on income inequality the link between social classes and

income dynamics is widely discussed. The structuring impact of economic and social

stratification on contemporary inequalities is one of the emerging theoretical and

policy questions. Layte and Whelan (2002) proved that there is a link between social

class and income dynamics: social class, education and employment status are

traditional predictors of income and poverty dynamics. Intergenerational income

mobility is strongly related with the inequality of the social opportunities. The relative

- 3 -

income position overlaps on a number of variables like educational attainment,

measured intelligence or occupational position across generations.

The investigations of intergenerational income dynamics, too, are based on the

human capital approach: the parent’s investments in their children increase the

children’s human capital, which in turn affects the children’s wages and earnings,

and thus their relative income position. The studies considered in Becker and Tomes

(1986) support the seminal hypothesis of non-linearities in the intergenerational

income dynamics and report an intergenerational elasticity of log income or log

earnings of about 0.2 in various industrialized countries. Solon (1989) among others

showed that the high intergenerational mobility in part was due to sample selection

and transitory fluctuations in earnings, which pollute the parental long-run earnings.

Using better quality data, more representative samples and appropriate methods

reduces the bias and the degree of intergenerational elasticity of earnings rose

around 0.4 or even higher. Considering capital market constraints leads to a higher

correlation between the income of parents and children among low income families

than among high income families. High ability children tend to come from high

earnings families because high-income parents will be more able to relax the liquidity

constraints faced by their children. Consequently, all else equal, the children of high-

income parents will have higher levels of both education and income. The

mechanisms that may underpin these results are (i) that the relative investments in

children made by rich and poor parents might change or (ii) that the payoff to these

investment might change or (iii) that the returns to genetic or biologically transmitted

characteristics might change (Becker and Tomes 1979; Becker and Tomes 1986;

Solon 1992; Zimmerman 1992; Mulligan 1997; Solon 1999; Solon 2002).

The recent analysis on income mobility trends focuses on the question whether the

large rise of income inequality occurring over the last decades has been accompanied

by a decline in intergenerational mobility. The “convergence” of intergenerational

income mobility is a function not only of single-generation income correlation but

also of structural changes in the income distribution (Björklund and Jäntti 1997;

Couch and Dunn 1997; Dearden, Machin and Reed 1997; Corak and Heisz 1999;

- 4 -

Solon 1999; Fertig 2003/04; Solon 2004; Mayer and Lopoo 2005; Mazumdar 2005).

Various studies report empirical evidence of country differences in income and

earnings mobility patterns as well as distinct transition mechanisms for different age

cohorts at different points of the income distribution. Bottom-up mobility is more

likely than top-down mobility and the intergenerational persistence of economic and

social positions is increasing, which implies decreasing chances for persons at the

lower end of the income distribution. The degree of immobility at the top and at the

bottom of the distribution might be exaggerate, for upward mobility is not possible

for those born at the top, downward mobility is not possible for those born at the

bottom.

The intention of this paper is to analyze, how human capital variables and parental

background characteristics work on occupational choice and income dynamics in

Germany and the USA – two countries differing concerning social role patterns and

the permeability of the social system (Giele and Holst 1997; Dunstmann 2004). We

start from the hypothesis that the link between occupational choice, social

stratification, labor market behavior and income dynamics works differently

according to the family role setting in an economy. In more traditional societies the

link between the parental economic and social background is more expressed and

the family background characteristics are more important for the economic and social

status of the individual. According to the human capital approach we expect gender

differences concerning the occupational choice and income dynamics in both the

countries. Due to the stronger link between generations we suppose the influence of

family background characteristics on both occupational choice and the mobility of the

real equivalent household income to be more pronounced in Germany.

The paper proceeds as follows: section 2 presents the data and sample organization,

section 3 specifies the model specifications to analyze occupational choice and

income mobility, section 4 presents the empirical results, and section 5 concludes

with a discussion of stylized facts about the structuring determinant of the

intergenerational heritage of social and economic status and derives policy

implications and directions for further research.

- 5 -

2 Data Base and Sample Organization

The empirical analysis is based on nationally representative data from the German

Socio-Economic Panel (GSOEP) and the Panel Study of Income Dynamics (PSID),

which were made available by the Cross-National-Equivalent-File (CNEF) project at

the College of Human Ecology at Cornell University, Ithaka, N.Y.. The PSID started in

1980 and contains a nationally representative unbalanced panel of about 40,000

individuals in the United States. From 1997 on the PSID data are available biyearly.

The GSOEP started in 1984 and contains with a representative sample of about

29,000 German individuals, and includes households in the former East Germany

since 1990. Both the surveys track the socioeconomic variables of a given household,

each household member is asked detailed questions about age, gender, marital

status, educational level, labor market participation, working hours, employment

status, occupational position, income situation, as well as household size and

composition (Burkhauser et. al. 2001). Both the data bases do not provide a

sufficient long time horizon to observe the parents’ and the children’s generation at

identical life cycle situations, but cover a sufficiently long period to observe the

socioeconomic characteristics of the parents living with their children and to link

these data with the children’s individual and household socioeconomic characteristics

when becoming members of other family units. Both the data bases do not allow

identifying parents-mother-children relations exactly. For this analysis we define

“mothers” and “fathers” as adults, whose marital status is “married” or “living with

partner” and who are living in households with persons with the marital status

“child”.

The German sample in the underlying analysis includes children born between 1964

and 1969 and co-resident with their parents in 1984. The US sample includes

children born between 1961 and 1966 and co-resident with their parents in 1981. We

consider children aged 15 to 20 years to avoid the overrepresentation of children

staying at home until a late age (Kolodinsky and Shirey 2001). The children are at

- 6 -

least 29 years old when we observe their occupational status and their relative

income position in the period 1998-2002 (Germany) or 1996-2001 (USA). We exclude

persons in full-time education, because their labor supply differs from the rest of

population. In the parental generation we consider persons up to 60years to avoid a

too large bias of retired persons. The selection process leads to a sample of 1,613

German women and men and 2,142 US women and men out of the children’s

generation.

3 Methodology

(a) The Multinomial Logit Model of Occupational Choice

According to the human capital approach individuals are assumed to be rational and

to have preferences over a set of alternative occupations. They will choose the

occupation that offers the highest discounted present value of potential future

earnings. The preference for the occupation j compared to any other occupation

maximizes the utility of an individual i. The utility of an occupation depends on a set

of individual characteristics and can be approximated by the linear relation

( )ij i i j ju u X X ß ε= = + , (1)

where i indexes the individuals (i=1,…,N), j indexes M+1 alternative occupations

(j=0,1,…M), ßj is a 1xK vector of (unknown) parameters and Xi is the i-th observation

on the Kx1 vector of explanatory variables. The disturbances jε indicate the random

error associated with occupation j, and are assumed to be independently and

identically distributed as a log Weibull distribution. In cases where the outcome

categories “can plausibly be assumed to be distinct and weighted independently in

the eyes of a decision maker” a multinomial logit model can be safely used (Mc

Fadden 1973).

- 7 -

We introduce the response variable Y for occupational choice which takes the value 1

if ij iku u j k> ∀ ≠ , and 0 else. We are interested in how ceteris paribus changes in

the elements Xi affect the probabilities of occupational choice

0

e( )e

i j

i k

X ß

i MX ß

k

P Y j X

=

= =

∑, (2)

where i indexes the individuals, and j indexes the alternative occupational choices,

j=0, 1,…,M, k=0,…M with j+1 nominal outcomes. Because the probabilities must

sum to unity, ( 0 )iP Y X= is determined once we know the probabilities for

j=1,2,3,…M and we need M parameter vectors to determine M+1 outcomes

1

e( )1 e

i j

i k

X ß

i MX ß

k

P Y j X

=

= =+∑

(3)

with 00,1,2,..., ; 0,..., ;ß =0j M k M= = (Mc Fadden 1973; Maddala 1983). The odds

ratio /j kP P does not depend on the other choices, which follows from the

independence of disturbances in the model. The log-odds ratios that an individual i

will prefer occupation j over occupation k can be written as the natural logarithm of

the relation of the probabilities of the occupation j and the reference category k

ln ( )ji j k

k

PX ß ß

P⎛ ⎞

= −⎜ ⎟⎝ ⎠

. (4)

Since we have fully specified the density of Y given Xi, the estimation of the model is

best carried out by maximum likelihood. The resulting estimates are unbiased,

consistent, asymptotically normal, and asymptotically efficient. The log-likelihood of

the model is given by

- 8 -

1 0 1 0

1

eln ( ) ln P ln1 e

i j

i k

X ßN M N M

ij j ij MX ßi j i j

k

L ß n Y j n= = = =

=

⎛ ⎞⎜ ⎟⎜ ⎟⎡ ⎤= = =⎣ ⎦ ⎜ ⎟+⎜ ⎟⎝ ⎠

∑ ∑ ∑ ∑∑

(5)

where the nij characterize the sum of frequency weights of the observations that

belong to the cell corresponding to Y=j in the subpopulation i 1

0i

ij

if Y jn

otherwise

∈⎧⎪= ⎨⎪⎩

,

so that for each individual i one and only one of the nij is 1.

Table 1: occupational categories

Y=j ISCO88 - occupational group

(0) 9 elementary occupations

(1)

8 plant and machine operators and assemblers

7 craft and related workers

6 agricultural and fishery workers

(2) 5 personal service

4 trade service

(3) 3 associate professionals

2 professionals

1 managers, senior officials

The empirical specification of the dependent variable in the multinomial logit model

indicates four distinct occupational groups. We rearrange the ISCO88 (international

standard of occupations) occupational categories according the skill requirements

(Goldthorpe 1987; Erikson and Goldthorpe 1992a; Erikson and Goldthorpe 1992b;

Goldthorpe 2000) into 0=elementary occupations, 1=working level occupations,

2=personal and trade services and 3=professional and managerial occupations.

According to the human capital theory, individuals will more likely choose the

occupation that offers the highest discounted present value of economic and social

status. There is a distinctive ranking of the four occupational dimensions:

occupations in the higher-numbered categories offer higher prestige and social

- 9 -

status. This is particularly true for countries, where economic and social hierarchies

are salient. (Table 1)

The explanatory variables in Xi contain a set of individual human capital variables as

well as and socioeconomic status characteristics that are expected to affect

occupational choice. These variables are the same for all alternatives, but their

effects on the individual’s probability to be in a given occupation are allowed to differ

for each alternative. (Table 2)

Table 2: Explanatory variables

Xi Description GENDERc gender of the individual: 1 male, 2 female EDUc Educational attainment is measured in school years. In the case of

missing values the years of education are set equal to the amount reported in the next year, for it is possible to increase the number of schooling but impossible to decrease it.

EMPc Employment status individual: 1 full-time, 2 part-time OCC-HHp occupation hh-head in the parental household: 0 elementary, 1

working, 2 trade, services, 3 professional, managerial HUMCAPp Human capital stock: sum of the years of education of the household-

head and her partner in the parental household HSTATUSp Relative income situation of the parental household: 1 long-rum real

equivalent post-government household income < median, 0 long-run real equivalent post-government household income > median

SIZEp Size of the parental household in 1984

The gender variable takes the value 1 for men and the value 2 for women. We

suppose that gender is likely to have different effects on the occupational choice.

Only in a world of equal opportunities and in absence of discrimination and

segregation we would expect to find no gender differences, and differences in

personal tastes should be the only factor to affect occupational choice, ceteris

paribus. The children’s human capital is captured by the years of education in 1996

(USA) or 1998 (Germany). In the case of missing values the years of education are

set equal to the amount reported in the next year, for it is possible to increase

educational attainment but impossible to decrease it. We include the employment

- 10 -

level as a dummy variable which takes the value 1 if the individual is full-time

employed and 0 if the individual is part-time employed.

In traditional societies we expect that family background characteristics may exert a

differential effect on social skills, human capital investment and tastes as well as

occupational choice of sons and daughters through sex-typing. Due to a stronger

link between generations we suppose the influence of family background to be more

pronounced in Germany. We introduce the family background characteristics

economic status, human capital stock, occupational status and household size. We

consider the household income after taxes and transfers as a measure of the

parents´ economic status rather than wages or earnings. The income variables in

the databases refer to the prior calendar year, we use these variables referring to the

wave of the interview, but questioned in the following wave. To exclude transitory

income shocks and cross-section measurement errors we use the average post-

government income in the period 1984 to 1988 (Germany) or 1981 to 1985 (USA).

We employ the CPI (2001=100) and the OECD-equivalence scale to calculate the real

equivalent post-government household income. The dummy variable “economic

status” refers to the relative income position of the parental households and holds

the value 1, if the parental household has a real equivalent post-government income

below the median and 0 otherwise. The human capital stock of the parental

household is considered as the sum of years of schooling of the household head and

her partner in the initial observation years 1981 (USA) or 1984 (Germany). In

countries with traditional role models labor-market know-how and professional

connections are inherited from the parental generation, and especially from the

father. We include the household-head’s occupational position and classify it in the

same manner as the dependent variable as a series of dummies. To analyze whether

the household size in the parental household interfere with equal chances of the

children of we include the number of persons in parental household in 1981 (USA) or

1984 (Germany) as covariates.

- 11 -

(b) Intergenerational income mobility

One way of measuring how economic (dis)advantages are transmitted across

generations is to estimate the correlations in economic outcomes between parents

and children. The most common approach to estimate intergenerational income

mobility is to apply ordinary least squares (OLS) to the regression of a logarithmic

measure of the children´s income variable on a logarithmic measure of the parental

income variable. There is a systematic heterogeneity in the income growth rates over

the life cycle. The measurement error in log current income as a proxy for log

lifetime income is mean-reverting early in the life-cycle and mean-departing later on.

For this reason the intergenerational elasticities might be lower when the children’s

income is measured early in the life-cycle. To account for this pattern we have to

control both children´s and parent´s age and age square in order to correct for life-

cycle differences

2 2

0 1 2 3 4 5ci pi pi pi ci ci iy ß ß y Age Age Age Ageβ β β β ε= + + + + + + , (6)

where i denotes a parents-child pair and piy is the log of the parental long-run real

equivalent household income, ciy represents the log of the children’s long-run real

equivalent household income, Agepi and Ageci represent the average age of parents

and children in the respective observation period. The random component iε is

assumed to be distributed as ),0( 2σN . The constant term 0β represents the change

in income common to the children’s generation, the coefficient 1β is the elasticity of

the child’s income level with respect to the parent’s income level. The larger 1β the

more likely an individual as an adult will inhabit the same economic position as her

parents, which implies a greater persistence of the intergenerational income levels.

Small values of 1β confer substantial advantages to the children of the well off. The

relationship between the intergenerational income elasticity and the intergenerational

income correlation is given with 1pi

ci

y

y

ßσ

ρσ

⎡ ⎤= ⎢ ⎥

⎢ ⎥⎣ ⎦, where σ is the standard deviation of

the log income of parents and children. However, ρ is equal to 1β if the degree of

- 12 -

inequality of the cross sectional income distributions does not change across

generations, that is if yci and ypi have equal variances (Mulligan 1997).

The intergenerational relation in economic status is transmitted to children mainly by

institutional and parental background characteristics affecting their education and

occupational status as well as their unobserved characteristics directly and indirectly.

The extension of equation (6) includes a set of parent and child control variables

(Zp,Zc) characteristics

2 2

0 1 2 3 4 5ci pi pi pi ci ci c c p p iy ß ß y Age Age Age Age Z Zβ β β β δ δ ε= + + + + + + + + . (7)

The parental control variables Zp account for the fact that the association between

parent and child income might not really be due to income per se, but to the effect

of parental education, occupational position or labor market behavior on the

children’s income. The control variables Zc account for the individual characteristics

of the children, which in part could express indirect effects on parental income: the

higher the income of the parents the higher their investments in the education of the

children, which in turn causes a higher income of the children. At the other side an

unusually intelligent or talented child may gain a higher education or income. To the

extent that including these variables lowers (raises) the focus coefficient ß1, we can

say that these other effects “account for” the raw intergenerational income

correlation, ß1 from equation (6).

The intergenerational elasticity is an average mobility measure that throws not

important light on the probabilities of economic success conditional to the economic

background of the parents. Transition matrices show the intergenerational income

persistence at different points in the income distribution, accounting for the fact, that

the linear functional form assumed in equation (6) may be incorrect. Transition

matrices allow measuring the degree of mobility, to analyze the determinants of

mobility as well as to see whether the observed change in mobility is found at the

top or at the bottom of the distribution. The movement from one income position

and the factors that influence them are a key issue from a welfare point of view and

- 13 -

are modelled with the conditional probability of having an income status in year t

given the income status in year t-1 (Heckman 1981).

To examine the parents’ and the children’s relative positions in the age-adjusted

income distributions we use a first-stage OLS regression of the children’s and

parents’ log long-run real equivalent household income on age and age squared. We

then split the residuals from these two regressions into five equal segments and

create a parent-child income transition matrix. Each element mki of a matrix indicates

the probability (in percent) that a child belongs to the kth quintile of the distribution

for children, given that her parents belong to the ith quintile of the parental

distribution. The entries sum to 1 along the columns. The more independent the

children’s and the parents’ income, the greater the likelihood that the elements of

this transition matrix should be close to 0.2 percent. The greater the elements of the

transition matrix differ from 0.2, the greater the intergenerational similarity in

relative age-adjusted income position. Then we run the regression including a set of

individual and family background control variables and again create parent-child

income transition matrices to evaluate the influence of family background on

intergenerational income mobility.

4 Empirical Results

(a) Summary Statistics

The tables 3 and 4 contrast the mean and percentage distribution of the individual

and parental characteristics with respect to gender. The summary statistics indicate

gender differences and traditional role patterns in both the countries. Women are

more likely engaged in the “trade & service” occupations or in the “professional &

managerial” category, whereas “elementary” and “working”-occupations are

predominated by men. Full-time employment among women is higher in United

States than in Germany. In Germany the proportion of full-time employed men

doubles the proportion of full-time employed women. In the United States women’s

- 14 -

annual working hours make about two third of the annual working hours of men.

This fact may be due to the different economic conditions and institutional labor

market settings, but also to traditional role models. In both the countries the gender

differences in educational attainment are not significant.

Table 3: Individual characteristics, children Germany 1998 USA 1996 men women men women age 31.3 34.4 31.2 31.1 school years 12.1 11.7 12.8 12.9 occupational status (in %) 0 elementary 1 worker 2 trade, service 3 professional, managerial

22.5 36.4 14.8 26.3

6.0 6.8

26.5 60.7

22.4 22.3 24.7 28.6

6.6 4.9

24.3 64.3

annual working hours 1,805 917 1,804 1,236 full-time employed (in %) 70.4 31.5 65.5 43.6 income status 1 equ.hh-income < median 40.6

48.6

40.0

53.4

ln hh-income mean 9.11 8.51 9.73 9.47 N 1,613 2,142 Source: GSOEP-PSID, own calculations

Concerning family background characteristics the descriptive statistics indicate

country similarities. In both the countries the household-head and her partner

combine about 20 years of schooling. The (male) household-heads are mainly

engaged in “elementary” and “working” occupations, the (female) partners are

mainly working in the “trade & service” or in the “professional & managerial”

categories. In Germany the proportion of full-employed (male) household-heads is

higher than in the United States, the (female) partners less likely perform full-time

jobs than in the United States. Compared to the children’s generation gender

differences are less expressed. In both the countries women more likely have

parents with a real equivalent household income below the median.

- 15 -

Table 4: Family background characteristics

German sample US sample men women men women age hh-head and partner 45.7 45.2 45.1 46.3 age hh-head and partner at birth of the child

29.5 29.2 29.2 30.4

age hh-head and partner at birth of the child 1 lt 25 2 25 to lt 30 3 30 to lt 35 4 35 to highest

18.8 33.7 28.9 18.5

18.4 39.0 23.2 19.5

27.5 25.9 24.0 22.6

26.0 22.5 21.4 30.2

human capital stock (average school years of hh-head and partner)

20.3 20.3 21.9 22.0

occupational status hh-head 0 elementary 1 worker 2 trade, service 3 professional, managerial

27.3 38.9 13.6 20.2

26.4 41.3 12.7 19.6

28.9 23.7 19.2 28.1

26.3 24.5 19.7 29.5

occupational status partner 0 elementary 1 worker 2 trade, service 3 professional, managerial

9.8 19.4 41.9 28.9

12.6 16.6 44.5 26.3

11.0 10.3 34.7 44.0

9.9

13.6 37.4 39.1

working hours hh-head (in % 1968 1936 working hours partner (in %) 841 775 hh-head full-time employed (in %)

80.6 78.7 78.4 68.6

partner full-time employed (in %) 22.7 22.1 26.6 27.4 income status (in %) 1 income < median 36.3 46.6

43.1

49.8

ln hh-income mean 9.97 9.96 9.23 9.17 average family labor income household size 2.9 2.9 2.7 3.3 N 1,613 2,142 Source: GSOEP-PSID, own calculations

Table 5 compares the occupational status of women and men out of the children’s

generation in 1996 (USA) or 1998 (Germany) with the occupational position the

household-head in the parental household held in 1981 (USA) or 1984 (Germany). In

Germany, “trade and service”-occupations (47.6 percent) and “professional &

managerial” occupations (61.5 percent) reveal the highest intergenerational

immobility. The results reveal gender differences: German men engaged in “working”

or “professional & managerial” occupations more likely inherit their occupational

- 16 -

status from the (male) household-head. Women traditionally are engaged in the

“trade & service” category or in “professional and managerial” occupations.

Table 5: Intergenerational occupational transitions

(a) GERMANY occupational status hh-head 1984 occupational status

children 1998 1 2 3 4 TOTAL

1 .346 .171 .124 .075 .186 2 .271 .418 .133 .098 .265 3 .223 .225 .476 .213 .257 4 .160 .185 .267 .615 .291

TOTAL 1 1 1 1 1

occupational status hh-head 1984

1 2 3 4occupational status

children 1998 male female male female male female male female

1 .496 .075 .242 .041 .143 .095 .116 .0252 .347 .134 .562 .155 .206 .024 .116 .0763 .074 .493 .118 .423 .429 .548 .147 .2914 .083 .299 .079 .381 .222 .333 .621 .608

TOTAL 1 1 1 1 1 1 1 1

(b) USA occupational status hh-head 1981 occupational status

children 1996 1 2 3 4 TOTAL

1 .287 .153 .112 .076 .149 2 .171 .292 .114 .085 .163 3 .196 .175 .346 .179 .209 4 .356 .381 .429 .660 .479

TOTAL 1 1 1 1 1

occupational status hh-head 1981

1 2 3 4occupational status

children 1996 male female male female male female male female

1 .451 .100 .228 .064 .187 .030 .127 .0232 .225 .111 .452 .101 .167 .055 .146 .0213 .117 .264 .103 .260 .346 .346 .173 .1864 .207 .526 .218 .575 .300 .570 .554 .769

TOTAL 1 1 1 1 1 1 1 1Occupational status: 0 elementary, 1 worker, 2 trade & services, 3 professional & managerial; Source: GSOEP-PSID, own calculations

Concerning these occupations the results show a high intergenerational mobility

between the (male) household head in the parental household and their female

children. In the United States the intergenerational occupational immobility of men

- 17 -

working in the “elementary” and “professional and managerial” categories is more

expressed, but there are less gender differences than in Germany.

(b) Multinomial logit estimates for occupational choice

The tables 6 and 7 present the results of the human capital and family background

multinomial logits, the estimated coefficients and their t-ratios of the 6 unique and

distinct comparative occupational choices, indicating ij iku u j k> ∀ ≠ . The t-ratios

are the ratios of the estimated coefficients to their estimated asymptotic standard

errors, and are assumed to be asymptotically distributed N(0,1) under the null

hypothesis that the associated coefficients are zero.

In both the countries the results of the multinomial logit model support the human

capital hypothesis. The coefficients on the gender variable are numerically larger

than the coefficients of the other variables and have a negative sign. Non-zero

coefficients of the gender variable indicate a differential access to certain

occupations, depending on one’s sex. The results imply that other things equal being

male makes it more likely to be in any occupational group higher relative to any

other occupational group. In the case of “professional and managerial” occupations

this effect is less pronounced in the United States. To interpret the multinomial logit

results as a measure of gender discrimination should be made with caution: gender

may affect the occupational choice because of the discrimination by employers, but

gender may also affect the preferences for different occupations. Occupational choice

is the outcome not only of individual preferences, but is influenced by institutional

labor market settings as well as of employment chances and discrimination. The

existing laws concerning parental leave and family work reconciliation as well as

traditional role patterns may inhibit women from accessing occupations even if they

want.

- 18 -

The preponderance of positive and significant coefficients of education demonstrates

that each additional year of education makes it more likely to be in a higher

occupational position than to perform a lower numbered occupational category.

Presumably this is exactly that what one expects: education is an important

qualification and enables one to move up the job scale. The results are congruent

with the findings of Schmidt and Strauss (1975), that each additional year of

education gives knowledge and also enhances learning efficiency and thus enables

persons to move “up” the occupational ladder and increases the probability to reach

a higher occupational status. In Germany the employment status does not work

significantly on occupational choice. In the United States full-time employment

significantly raises the probability to choose “professional and managerial”

occupations.

Table 6: Multinomial logit results (t-ratios in parentheses), Germany

DEPENDENT VARIABLES estimated

coefficients (t-ratios) for independent

variable

( )1 0ln P P

( )2 0ln P P

( )3 0ln P P

( )2 1ln P P

( )3 1ln P P

( )3 2ln P P

CONSTANT 1.202* (5.408)

1.342* (4.584)

3.501* (2.057)

1.194* (4.258)

3.260* (1.995)

3.220* (2.817)

GENDER -3.070* (-7.775)

-2.572* (-7.871)

-1.802* (-6.071)

-2.680* (-7.844)

-1.143* (-2.054)

-2.024 (-4.888)

EDUc 0.648* (6.412)

0.552* (6.180)

0.391* (5.572)

0.555* (6.051)

0.390* (5.577)

0.388* (5.402)

EMPc 0.322 (0.888)

-0.318 (-0.970)

0.241 (0.882)

-0.278 (-0.808)

0.182 (0.658)

0.149 (0.529)

OCC-HHp

worker 2.038* (4.262)

1.362* (2.991)

0.536 (1.398)

1.383* (2.881)

0.485 (1.233)

0.233 (0.561)

OCC-HHp

trade&service 0.991* (2.246)

1.513* (3874)

0.191 (0.588)

1.580* (3.960)

0.200 (0.610)

0.189 (0.572)

OCC-HHp

prof.&manag. 0.314

(0.568) 0.152

(0.296) 0.999* (2.694)

-0.010 (-0.018)

1.000* (2.677)

0.958* (2.575)

HUMCAPp 0.131* (2.104)

0.153* (2.636)

-0.073 (-1.570)

0.148* (2.490)

0.085* (1.842)

0.097* (1.703)

STATUSp -0.628 (-1.349)

-0.799* (-2.794)

0.152 (0.421)

-1.195* (-2.366)

0.133 (0.362)

0.190 (0.511)

HH-SIZEp 0.154 (1.121)

0.336* (2.609)

0.265* (2.172)

0.383* (2.830)

0.282* (2.282)

0.252* (2.051)

LL -395.913

2χ 189.590

N 1,384

- 19 -

NOTE: *indicates significance at the 5percent level in a two-tailed test (p<0.05) SOURCE: GSOEP-PSID, own calculations

Table 7: Multinomial logit results (t-ratios in parentheses), United States

DEPENDENT VARIABLES estimated

coefficients (t-ratios) for independent

variable

( )1 0ln P P

( )2 0ln P P

( )3 0ln P P

( )2 1ln P P

( )3 1ln P P

( )3 2ln P P

CONSTANT 2.516* (6.687)

1.518* (5.336)

3.139* (5.186)

2.139* (4.190)

3.106* (5.145)

2.922* (4.776)

GENDER -2.790* (-6.694

-2.272* (-6.049)

-0.639* (-5.769)

-2.297* (4.984)

-0.648* (5.813)

-0.604* (-5.390)

EDUc 0.494* (3.878)

0.454* (4.877)

0.287* (2.485)

0.452* (2.462)

0.281* (2.309)

0.275* (7.970)

EMPc 0.148 (0.874)

-0.605* (-3.541)

0.353* (3.101)

-0.626* (-3.554)

0.354* (3.094)

0.352* (3.050)

OCC-HHp

worker 1.714* (8.559)

0.922* (4.714)

0.282* (1.847)

0.984* (4.937)

0.264 (1.741)

0.239 (1.549)

OCC-HHp

trade&service 0.907* (4.503)

1.686* (3.845)

0.124 (0.852)

1.675* (4.716)

0.123 (0.847)

0.117 (0.792)

OCC-HHp

prof.&manag. 0.494* (1.973)

0.568* (2.546)

1.038* (5.209)

0.555* (2.463)

1.034* (5.185)

1.030* (4.142)

HUMCAPp 0.028 (0.132)

0.018 (0.800)

0.017 (0.907)

0.031 (0.305)

0.013* (0.711)

0.012 (0.620)

STATUSp -0.311* (-2.102)

-0.275* (-2.026)

-0.308* (-2.662)

-0.272* (-1.960)

-0.304* (-2.629)

-0.320* (-2.734)

HH-SIZEp 0.106* (2.244)

0.058 (1.301)

0.020 (0.615)

0.066 (1.442)

0.023 (0.621)

0.029 (0.757)

LL -503.664

2χ 171.543

N 1,832 NOTE: *indicates significance at the 5percent level in a two-tailed test (p<0.05) SOURCE: GSOEP-PSID, own calculations

With regard to the parental background characteristics the results show that the

parental human capital stock has a significant and positive impact on occupational

choice in Germany: the higher the parental human capital stock the higher the

probability that children are to be found in the higher occupational status. In the

United States the parental educational attainment does not affect occupational

choice significantly. This fact might originate from the small variation in the parental

human capital stock. Concerning the effect of the parental economic status

differently on the children’s occupational choice we find country differences:

Germans whose parents have a real equivalent post-government household income

below the median less likely prefer “trade & service”-occupations over the

- 20 -

“elementary” and the “working” category. In the United States the coefficients of all

log odds combinations are significant: children whose parents are economically well

off will prefer the higher ranked occupation over the lower ranked ones. In both the

countries the empirical results partly confirm that persons follow their parents´

occupation. We find positive and significant coefficients for the influence of fathers

engaged in “working” or “trade & service” occupations on the children’s occupational

choice. In the United States the coefficients of the household head’s “professional

and managerial” occupations are significant for all of the log odds combinations. At

the other hand, in Germany the father’s occupation is influential only for choice of

high numbered occupations: Germans whose fathers are engaged in “trade &

service” occupations prefer “professional & managerial” occupations over all other

categories.

The family background variable “household-size” has a positive or a negative

potential impact on the prospects of the children’s economic and social success. A

larger number of persons in the household may have positive net effects on

household accumulation among the low strata, but may have opposite effects among

the upper strata. This implies that households belonging to the lower social strata

may have an economic incentive to have a larger number of children as a rational

household strategy. In the German sample the household size interferes significantly

with the occupational success of an individual in most of the comparisons. In the

United States the number of persons in a household does not work significantly in

most of the cases.

(c) Intergenerational income mobility

Table 8 reports the estimates of equation (6) and equation (7) for both the samples.

The regression of the log parental real equivalent post-government household

income on the children’s log real equivalent post-government household income

- 21 -

leads to an intergenerational elasticity of 0.241 for the German sample and an

intergenerational elasticity of 0.440 for the US sample.

Table 8: The elasticity of the children’s income with respect to parental income GERMANY Equation (1) Equation (2) ßp SE t-

statistic N ßp SE t-

statistic N

children 0.241 0.063 4.365 811 0.223 0.074 3.647 762 male 0.305 0.068 4.301 478 0.205 0.082 2.809 447 female 0.248 0.124 1.996 333 0.270 0.144 2.490 315

USA Equation (1) Equation (2) ßp SE t-

statistic N ßp SE t-

statistic N

children 0.440 0.046 4.666 1090 0.313 0,111 3.315 928 male 0.332 0.079 3.810 414 0.314 0.106 4.995 354 female 0.459 0.058 4.271 676 0.423 0.119 3.674 574

The inclusion of a set of control variables, including parental age and educational

attainment and family structure in the specification of the augmented regression (7)

is an attempt to identify the effect of changes in the household income for otherwise

identical individuals. We see how the coefficients would alter if one moved towards a

fixed effects specification that would wash out child and family characteristics not

previously controlled. The results reveal lower intergenerational correlations for both

the samples than indicated by the regression of equation (6), but we find gender and

country differences. For German women the intergenerational elasticity raised from

0.248 to 0.270 suggesting that women are less dependent on their family

background. Possible explanations are that these women benefit from various

welfare-state programs, as publicly provided child care or the improved parental

leave. Additionally, educational attainment and labor force participation of women

increased substantially. These changes may explain, in part, why women are more

independent of their family background than men. Another explanation could be that

labor market segregation and discrimination diminishes the importance of family

background for women. A final explanation could be that a women’s own family is

more important for her income status than her family background. In the United

States the intergenerational income elasticity slightly declines for men and women

- 22 -

revealing a higher dependence on their family background than in Germany. These

findings are consistent with Behrman and Taubman (1990) about the correlation of

the children’s earnings and the parents’ income. However, the results do not

corroborate with a more traditional German society and a higher intergenerational

mobility in the United States.

The intergenerational transition matrices of the age-adjusted log income variables for

children and parents show a noticeable persistence of the income positions of

parents and children in both the countries. (Table 9) The probability that a child ends

up in an income quintile different from the one occupied by her parent tends to be

monotonically decreasing the farther away that quintile is from that of their parents.

Children are most likely to fall into an income quintile exactly like that of their

parents and are very unlikely to end up in dramatically different one. In Germany

46.9 percent of the parental households in the lowest age-adjusted income quintile

have children whose income places them in that same quintile in the children’s age-

adjusted income distribution. However, many children are able to escape their

parents’ economic position. Over 30 percent of the parents in the lowest quintile

have children whose income places them in one of the three highest income

quintiles, but only 4.7 percent of children with parents in the lowest income quintile

perform into the highest quintile. A similar degree of persistence is evident at the

upper tail of the parental income distribution. 46.5 percent of high-income parents

have children who end up in the top quintile of the children’s age-adjusted income

distribution, and almost 70 percent have children whose income places them in the

two top quintiles. Only 5.7 percent of the children of high-income parents fall to the

lowest quintile. In the United States it seems to be more immobility of the income

status not only at the lower end (52.4 percent) but also in the middle of the income

distribution. The likelihood ratio χ2 –test confirms the persistence evident in the

table: we can strongly reject the hypothesis that the entries in the adjusted wealth

position transition matrix are equal to each other at any standard statistical level

(p<.001).

- 23 -

The higher probability for the stayers at both the very top and very bottom of the

parental income distribution represents the non-linearity in the mobility process, and

has something to do with the floors and ceilings in the design of transition matrices.

Table 9: Intergenerational transition matrix of age-adjusted log income positions (a) GERMANY Parental-age-adjusted log income quintiles (1984-1988)

Child-age-adjusted log income quintiles

(1998-2002)

1 2 3 4 5 TOTAL

1 .469 .282 .208 .157 .057 .225 2 .209 .220 .153 .184 .120 .175 3 .173 .176 .230 .218 .127 .186 4 .102 .208 .259 .232 .231 .210 5 .047 .114 .150 .208 .465 .204

TOTAL 1 1 1 1 1 1

Parental-age-adjusted log income quintiles (1984-1988)

1 2 3 4 5Child-age-adjusted log

income quintiles

(1998-2002)

male fem. male fem. male fem. male fem. male fem.

1 .450 .488 .217 .344 .196 .220 .089 .224 .045 .0702 .171 .248 .217 .224 .172 .133 .212 .156 .083 .1623 .194 .152 .167 .184 .233 .227 .226 .211 .140 .1134 .132 .072 .258 .160 .233 .287 .247 .218 .242 .2185 .054 .040 .142 .088 .166 .133 .226 .190 .490 .437

TOTAL 1 1 1 1 1 1 1 1 1 1

(b) USA Parental-age-adjusted log income quintiles (1981-1985)

Child-age-adjusted log income quintiles

(1996-2001)

1 2 3 4 5 TOTAL

1 .541 .351 .161 .068 .025 .216 2 .223 .323 .239 .173 .115 .204 3 .135 .176 .272 .239 .159 .192 4 .077 .106 .223 .289 .241 .193 5 .024 .045 .106 .232 .460 .195

TOTAL 1 1 1 1 1 1

Parental-age-adjusted log income quintiles (1981-1985)

1 2 3 4 5Child-age-adjusted log

income quintiles

(1996-2001)

male fem. male fem. male fem. male fem. male fem.

1 . 498 .566 .306 .381 .139 .178 .058 .077 .019 .0312 .242 .203 .306 .334 .210 .260 .167 .179 .107 .1223 .155 .115 .205 .156 .302 .249 .223 .253 .144 .1734 .093 .062 .130 .090 .226 .221 .302 .276 .248 .234

- 24 -

5 .012 .035 .054 .039 .123 .093 .250 .215 .482 .440TOTAL 1 1 1 1 1 1 1 1 1 1

SOURCE: GSOEP-PSID, own calculations. Note: The likelihood ratio χ2 statistics is significant (p<.001) in all comparisons. These results corroborate when examining percentiles: children born in households

at the fifth percentile have a tendency to move up even further in the income

distribution, children born in households at the third percentile have the tendency to

move even lower in the income distribution. The percentiles in general reveal that

the probability of upward mobility is lower for children from low income backgrounds,

while the probability of upward mobility is higher for sons from high income

backgrounds. In both the countries the results show gender differences: women

experience higher status immobility than men in all income quintiles except at the

top of the relative income distribution. However, the results do not support the

hypothesis of a very mobile American society.

Table 10: Intergenerational transition matrix of age-adjusted log income positions considering family background characteristics (a) GERMANY Parental-age-adjusted log income quintiles (1984-1988)

Child-age-adjusted log income quintiles

(1998-2002)

1 2 3 4 5 TOTAL

1 .279 .261 .220 .121 .128 .201 2 .162 .232 .232 .103 .244 .198 3 .132 .319 .159 .227 .167 .198 4 .147 .159 .207 .318 .179 .201 5 .279 .029 .183 .227 .282 .201

TOTAL 1 1 1 1 1 1

(b) USA Parental-age-adjusted log income quintiles (1981-1985)

Child-age-adjusted log income quintiles

(1996-2001)

1 2 3 4 5 TOTAL

1 .339 .186 .216 .228 .154 .224 2 .258 .254 .176 .228 .138 .208 3 .145 .169 .189 .316 .185 .199 4 .129 .203 .176 .158 .277 .189 5 .129 .186 .243 .070 .246 .180

TOTAL 1 1 1 1 1 1SOURCE: GSOEP-PSID, own calculations. Note: The likelihood ratio χ2 statistics is significant (p<.001) in all comparisons.

- 25 -

The transition matrices considering the family background characteristics reveal

higher intergenerational income mobility in both the countries. In Germany persons

in the third quintile perform the highest intergenerational income mobility. In the

United States the persistence of income positions at the lower end of the income

distribution is more pronounced than in Germany. (Table 10)

5 Conclusion

In this paper we employed data from the GSOEP-PSID Equivalent File 1980-2002 to

analyze the influence of human capital variables and family background

characteristics on occupational choice and income dynamics of a 29 to 34 years

cohort in Germany and the United States. In both the countries gender significantly

affects the occupational distribution: men are engaged in “elementary and working”

occupations, and women are mainly engaged in “trade & service” and “professional &

managerial” jobs. The multinomial logit model reveals that being male augments the

probability to be in a higher occupational position. In both the countries educational

attainment increases the probability of advancing the occupational ladder: more

educated individuals have a higher probability of advancing into occupations that

offer a higher social status. This implies that the better access to human capital

facilitates the prospects for further accumulation and thus for climbing up the social

ladder.

Family background characteristics affect the occupational choice both directly

through genetic endowment, social connections, and wealth, and indirectly through

educational attainment. In Germany the economic status of the parents significantly

works on the occupational choice of the children in few cases, whereas the parental

economic position significantly influences the occupational choice in the United

States. The results corroborate with the evidence of occupational inheritance stated

by Lentz and Laband (1989) and Laband and Lentz (1992): the parental human

capital stock as well as their occupational position may shape the taste and

- 26 -

perception of what is an appropriate educational and professional career for the

child. In addition, knowledge transmission may contribute to occupational

persistence across generations. In both the countries the results support traditional

role patterns in many cases: the occupational status of the parents significantly

determines the occupational preferences of their children. Moreover, children whose

father is in a “professional & managerial” occupation have access to valuable social

connections and more chances for economic success. In the United States individuals

with different family backgrounds do not face equal options. There is some evidence

that individuals inherit their social status and their position in the occupational

distribution.

The results on intergenerational income mobility, too, document a substantial

amount of churning in the economic positions across generations. The age-adjusted

elasticity of children’s income with respect to parent’s income is 0.241 in Germany

and amount to 0.440 in the United States, indicating that intergenerational income

mobility is less pronounced in the United States. In the United States family

background variables affect intergenerational income mobility onto a higher extent

than in Germany. The transition matrices demonstrate that much of the

intergenerational income persistence arises from what occurs in the tails of the

income distribution. Children with a very low or very high parental income

background rarely end with an income substantially different from their parents´

relative income position.

The results on occupational choice and income mobility do not support the

hypothesis of more traditional role models in Germany and a higher social and

economic mobility in the United States. The economic and social policy implications

are quite clear: If the parental economic and social position prevents high ability

children from entering higher education, this will have negative implications on

productivity and national social welfare (Sen 1981; Goldthorpe 2000). Therefore, in

both the countries social policy is called upon to improve the access to educational

attainment to facilitate the prospects of further climbing up the occupational ladder,

to enforce upward intergenerational income mobility, and thus to raise the

- 27 -

permeability of the social structure. Additionally, social policy is required to enforce

gender equality chances, as well as family work reconciliation strategies within the

families.

References

Becker, Gary S., 1964. Human Capital. National Bureau of Economic Research, New

York. Becker, Gary S. and Nigel Thomes, 1979. An Equilibrium Theory of the Distribution of

Income and Intergenerational Mobility. Journal of Political Economy 87,6:1153-1189.

Becker, Gary S. and Nigel Tomes, 1986. Child endowments and the Quantity and Quality of Children. Journal of Political Economy 84: S1143-S1162.

Becker, Gary S., 1988. Family Behavior and Macro Behavior. American Economic Review 78: 1-13.

Behrman and Taubman, 1990. The intergenerational correlation between children´s adult earnings and their parents´ income: Results from the Michigan Panel Survey of Income Dynamics. Review of Income and Wealth 36,2:115-127.

Bjoerklund, A. and Jaenti, M. (1997). ‘Intergenerational mobility in Sweden compared with the United States’, American Economic Review, 87: 1021–1028.

Blau, Peter M. and Otis D. Duncan, 1967. The American Occupational Structure. New York.

Boskin, M.J., 1974. A Conditional Logit Model of Occupational Choice. Journal of Political Economy 82: 389-398.

Burkhauser, R.V., Butrica, B.A., Daly, M.C. and D.R. Lillard, 2001. The Cross-National Equivalent File: A product of cross-national research. In: Becker, Irene, Ott, Notburga and Gabriele Rolf (eds.) Soziale Sicherung in einer dynamischen Gesellschaft, Campus Verlag, Frankfurt, Germany: 354-376.

Couch, T.A. and T.A. Dunn, 1997. Intergenerational Correlation in Labour Market Status, a Comparison of the United States and Germany, Journal of Human Ressources 32,4: 210-232.

Corak, Miles and Andrew Heisz, 1999. The intergenerational earnings and income mobility of Canadian men: Evidence from the Longitudinal Income Tax Data, Journal of Human Ressources 34: 505-533

Dearden, Lorraine, Machin, S. and H. Reed, 1997, Intergenerational Mobility in Britain, Economic Journal 107: 47-64.

Dunstmann, Christian, 2004. Parental background, secondary school, track choice, and wages. Oxford Economic Papers 56: 209-230.

Erikson R. and J.H. Goldthorpe, 1992a. The Constant Flux: A Study of Class Mobility in Industrial Societies. Clarendon Press, Oxford.

Erikson R. and J.H. Goldthorpe, 1992b. Intergenerational Inequality: A Sociological Perspective. Journal of Economic Perspectives 16: 31-44.

Fertig, Angela, 2003/04. Trends in Intergenerational Earnings Mobility in the U.S.. Journal of Income Distribution 12:108-140.

- 28 -

Giele, J.Z. and E. Holst, 1997. Dynamics of Women´s Labor Force Participation in the United States and West Germany, 1983 to 1990. In: Proceedings of the 1996 Second International Conference of the German Socio-Economic Panel Study Users (edited by Th.A. Dunn and J. Schwarze), DIW-Vierteljahreshefte 1: 55 – 61.

Goldthorpe, John H., 1987. Social Mobility and Class Structure in Modern Britain. London.

Goldthorpe, John H., 2000. Social Class and the Differentiation of Employment Contracts. In: Goldthorpe, J.H. (ed.), On Sociology. Numbers, Narratives, and the Integration of Research and Theory. Oxford University Press, Oxford: 161-181.

Harper, Barry and Mohammed Haq, 1997. Occupational Attainment of Men in Britain. Oxford Economic Papers 49,4: 638-651.

Heckman, J.J., 1981. Statistical models for discrete panel data. In: Manski, C.F. and D. McFadden (eds.), Structural Analysis of Discrete Data with Econometric Applications. MIT Press, Cambridge, pp. 114-178.

Kolodinsky, Jane and Lee Shirey, 2000. The impact of living with an elder parent on adult daughter´s labor supply and hours of work. Journal of Family and Economic Issues 21,2: 149-174.

Laband, David N. and Lentz, Bernard F. (1992). ‘Self-recruitment in the legal profession’, Journal of Labor Economics 10: 182–201.

Lentz, Bernard F. and David N. Laband. 1989. "Why So Many Children of Doctors Become Doctors: Nepotism vs. Human Capital Transfers." Journal of Human Resources 24,3: 396-413.

Layte, R. and C.T. Whelan, 2002, Cumulative disadvantage or individualisation? A comparative analysis of poverty risk and incidence, European Societies 4,2:209-233.

Maddala, G.S., 1983. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge University Press, Cambridge.

Mayer, Susan E. and Leonard M. Lopoo, 2005. Has the Intergenerational Transmission of Economic Status Changed? The Journal of Human Resources 40,1: 169-185.

Mazumder, B., 2005. Fortunate Sons: New Estimates of Intergenerational Mobility in the U.S. Using Social Security Earnings Data. Review of Economics and Statistics 87,2: 235-255.

McFadden, Daniel, 1973. Conditional Logit Analysis of Qualitative Choice Behavior. In: Zarembka, P.(ed.), Frontiers of econometrics. Academic Press, New York: 105-142.

Mincer, J., 1974. Schooling, Experience and Earnings. NBER, New York. Mulligan, C.B., 1997, Parental Priorities and Economic Inequality, The University of

Chicago Press, Chicago. Sen, A.K., 1981, Poverty and Famines, an Essay on Entitlement and Deprivation,

Oxford 1981. Solon, Gary, 1989. Biases in the Estimation of Intergenerational Earnings

Correlations. Review of Economics and Statistics 71,1: 172-174 Solon, Gary, 1992. Intergenerational Income Mobility in the United States. American

Economic Review 82,3, 326-329.

- 29 -

Solon, Gary, 1999. Intergenerational Mobility in the Labor Market. In: Ashenfelter O. and D. Card (eds.), Handbook of Labor Economics, North Holland, Amsterdam.

Solon, Gary, 2002. Cross-Country Differences in Intergenerational Earnings Mobility. Journal of Economic Perspectives 16: 59-66.

Solon, Gary, 2004. A Model of Intergenerational Mobility Variation over Time and Place. In: Corak, M. (ed.), Generational Income Mobility in North America and Europe, Cambridge University Press, Cambridge, UK: 38-47.

Zimmerman, David J., 1992, Regression towards mediocrity in economic stature, American Economic Review 82: 409-429.