Working Paper Series
The Intergenerational Effects of Compulsory Schooling
MariannePage
UniversityofCalifornia,Davis
AnnHuffStevens
UniversityofCalifornia,Davis
PhilipOreopoulos
UniversityofToronto
August01,2005
Paper#05-19
Thestrongcorrelationbetweenparents’economicstatusandthatoftheir
childrenhasbeenwell-documented,butlittleisknownabouttheextentto
whichthisisacausalphenomenon.Thispaperattemptstoimproveour
understandingofthecausalprocessesthatcontributetointergenerational
immobilitybyexploitinghistoricalchangesincompulsoryschoolinglawsthat
affectedtheeducationalattainmentofparentswithoutaffectingtheirinnate
abilitiesorendowments.Weexaminetheinfluenceofparentalcompulsory
schoolingonchildren’sgradefor-ageusingthe1960,1970and1980U.S.
Censuses.Ourestimatesindicatethataone-yearincreaseintheeducationof
eitherparentreducestheprobabilitythatachildrepeatsagradebybetween
twotofourpercentagepoints.Among15to16yearoldslivingathome,we
alsoestimatethatparentalcompulsoryschoolingsignificantlylowersthe
likelihoodofdroppingout.Thesefindingssuggestthateducationpoliciesmay
beabletoreducepartoftheintergenerationaltransmissionofinequality.
DepartmentofEconomics
OneShieldsAvenue
Davis,CA95616
(530)752-0741
http://www.econ.ucdavis.edu/working_search.cfm
The Intergenerational Effects of Compulsory Schooling
Philip Oreopoulos
Department of Economics
University of Toronto
Marianne E. Page
Department of Economics
University of California, Davis
Ann Huff Stevens
Department of Economics
University of California, Davis
August 2005
Abstract: The strong correlation between parents’ economic status and that of their children has
been well-documented, but little is known about the extent to which this is a causal phenomenon.
This paper attempts to improve our understanding of the causal processes that contribute to
intergenerational immobility by exploiting historical changes in compulsory schooling laws that
affected the educational attainment of parents without affecting their innate abilities or
endowments. We examine the influence of parental compulsory schooling on children’s grade-
for-age using the 1960, 1970 and 1980 U.S. Censuses. Our estimates indicate that a one-year
increase in the education of either parent reduces the probability that a child repeats a grade by
between two to four percentage points. Among 15 to 16 year olds living at home, we also
estimate that parental compulsory schooling significantly lowers the likelihood of dropping out.
These findings suggest that education policies may be able to reduce part of the intergenerational
transmission of inequality.
We would like to thank Adriana Lleras-Muney, Joshua Angrist, and Daron Acemoglu for
making available the data from their work on compulsory schooling laws. Melanie Guldi and
Sami Kitmitto provided invaluable research assistance. We are grateful to seminar participants
at UC-Davis, the University of Toronto, the University of British Columbia, and Duke
University for their helpful comments. The National Science Foundation provided funding for
this project.
1
I. Introduction
A high degree of persistence in economic status exists across generations, but we know
very little about the causal processes that drive this phenomenon. For example, children who
grow up in more highly educated families have better labor market outcomes as adults than
children who grow up in less educated families, but we do not know whether this is because
education changes something about childhood experiences or because genetic or environmental
factors that contributed to the parents’ educational levels are shared by their children. This paper
attempts to improve our understanding of the causal processes that contribute to intergenerational
immobility by exploiting historical changes in compulsory schooling laws that affected the
educational attainment of parents without affecting their innate abilities. Quantifying the extent
to which children’s well-being can be improved by increasing their parents’ education also has
important implications for public policy: most discussions about the government’s role in
providing educational aid, for example, focus on the individual’s return to education and ignore
the possibility of social benefits. Knowing that there are intergenerational returns to increased
schooling would provide a further rationale for such programs.
Few studies have attempted to isolate the causal effect of education on the next
generation’s well-being.
1
This is at least partly due to the fact that it is difficult to find plausible
sources of identifying variation. It is also hard to find large, nationally representative datasets
that simultaneously provide information on parental characteristics and children’s outcomes.
Our use of compulsory schooling laws applied to Census data allow us to overcome both of these
1
By “causal” we mean the effect of an exogenous increase in education itself, rather than the simple correlation
between a child’s outcome and her parents’ education, which also reflects the effect of innate parental
characteristics. Changes in parental schooling levels may change many family background characteristics such as
income, choice of marriage partner, and decisions about children’s education. Our IV estimates will capture the
effect of all family background characteristics that are affected by education but will not reflect the effect of innate
parental characteristics such as intelligence.
2
problems. Several studies have already demonstrated a strong relationship between these laws
and individuals’ educational attainment,
2
and have used this relationship to identify the effects of
education on earnings [Acemoglu and Angrist, 2000], criminal activity [Lochner and Moretti,
2004], mortality [Lleras-Muney, 2005], and subjective measures of well-being [Oreopoulos,
2003], but this is the first study to estimate the intergenerational effects of the U.S. laws.
Our main analysis is based on a sample of children ages 7-15 taken from the 1960, 1970
and 1980 individual U.S. Census files. We examine the effects of parental education on
children’s human capital accumulation, and find that it has substantial and significant positive
effects. A one year increase in parents’ combined schooling reduces the probability that a child is
at the normal grade given her age by two to four percentage points. This effect is somewhat
larger than OLS estimates would suggest. We also find evidence that among teenagers still living
at home, parents’ educational attainment decreases children’s likelihood of dropping out of high
school.
II. Background Literature
A few previous studies have attempted to isolate the causal influence of parental
education by using variation within sibling or twin pairs [Behrman and Rosenzweig, 2002;
Currie and Moretti, 2003; Rosenzweig and Wolpin, 1994]. This enables them to control for
mother fixed effects, but may exacerbate biases if within pair differences in characteristics affect
outcomes independent of their effect on education.
3
Such estimates are also known to be more
prone to measurement error problems (Griliches, 1979). Currie and Moretti (2003) also isolate
2
See Angrist and Krueger (1991), Lleras-Muney (2005), and Goldin and Katz (2003). Harmon and Walker (2000)
and Oreopoulos (2003) find that a similar relationship exists in the United Kingdom and Canada.
3
Behrman, Rosenzweig, and Taubman (1994), for example, find birth weight differences between twins correlate
significantly with schooling differences and subsequent earnings. Recent work by Antonovics and Goldberger
(2005) questions the robustness of the Behrman and Rosenzweig’s (2002) results.
3
the effect of mothers’ education on birth outcomes using county level date on college openings
between 1940 and 1990 to capture differences in the availability of educational services among
different cohorts of women. Sacerdote (2002, 2004) compares outcomes of adoptees who were
randomly assigned to their adoptive parents. This approach is appealing because it is impossible
to ascribe the intergenerational link to a genetic inheritance, but an observed correlation could
still be driven by parental characteristics that are not genetically transmitted to children but that
nevertheless affect both individual and offspring’s human capital accumulation. Across these
studies, the magnitude and precision of the estimated effects of parental education varies, but
most find that parental education has at least a small impact on children’s outcomes.
4
Most existing studies have also relied on small, non-representative, datasets. The
Behrman and Rosenzweig study, for example, uses samples of 212 female twin pairs and 122
male twin pairs who participated in the Minnesota Twin Registry. Rosenzweig and Wolpin
(1994) use a sample of unusually young mothers from the NLSY. Likewise, Sacerdote’s
analyses focus on fewer than 200 adoptees per sample who were born in Britain (the first
sample) or adopted in Colorado (the second sample).
Recent papers by Chevalier (2003) and Black, Devereux and Salvanes (2005) also use
changes in compulsory schooling laws to identify the effect of parental education on the next
generation’s outcomes. Chevalier’s study uses a change in the compulsory schooling
requirement that took place in Britain in 1957. He finds large effects of mother’s education on
children’s educational attainment, but does not find statistically significant effects of father’s
education. Because identification is based on a single change in the minimum schooling law,
however, he is unable to disentangle compulsory schooling effects from cohort effects.
4
In a related paper, Bleakely and Chin (2004) show that children of immigrant parents with stronger English
language skills have better educational outcomes, but they do not explicitly estimate the causal effect of parental
education.