• Goal of this segment: Write a report on your findings

I Individual report!

I Main focus: Question/Topic you focused on in the research segment

• What should always be covered:

I Motivation: Why did you pick your topic? Why is it interesting? (Refer back to magazine article and empirical evidence.)I Analysis of Existing Research: What is the state of research on the topic?

Description of research methods, datasets etc.

I Discussion: What is the main takeaway from the literature? Compare and

contrast findings. What is your take on the topic? What is missing in the

existing literature?

Summary of what was presented and worked on throughout the moduleHow Did the Increase in Economic Inequality between 1970 and 1990 Affect Children’s

Educational Attainment?

Author(s): Susan E. Mayer

Source: American Journal of Sociology , Vol. 107, No. 1 (July 2001), pp. 1-32

Published by: The University of Chicago Press

Stable URL: https://www.jstor.org/stable/10.1086/323149

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide

range of content in a trusted digital archive. We use information technology and tools to increase productivity and

facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access

to American Journal of Sociology

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

How Did the Increase in Economic

Inequality between 1970 and 1990 Affect

Children’s Educational Attainment?1

Susan E. Mayer

University of Chicago

This study estimates the effect of changes in economic inequality

between 1970 and 1990 on children’s educational attainment. Data

on individual children from the Panel Study of Income Dynamics

is combined with other data on state characteristics. Growing up in

a state with widespread economic inequality increases educational

attainment for high-income children and lowers it for low-income

children. Most of the effect is due to factors unassociated with family

income or economic segregation in the state. These other factors

include state spending for schooling and the increase in the returns

to schooling over this period.

Disparities in hourly wages, annual earnings, and household income have

all increased over the past generation in the United States. A considerable

amount of research has tried to determine why income inequality grew

over this period (Morris, Bernhardt, and Handcock 1994; Morris and

Western 1999). Much less research has been done on the consequences of

inequality than on its causes. This article estimates the effect of changes

in income inequality on mean educational attainment and on the disparity

in educational attainment between rich and poor children. I also separate

the effect of income inequality that is due to the nonlinear effect of a

1

This research was funded by the Russell Sage Foundation as part of a larger collaboration with Christopher Jencks and Paul Jargowsky. David Knutson, Lenard Lopoo, and Gigi Yuen-Gee Liu provided exemplary research assistance. I am grateful to

participants in the Harris School Faculty Seminar, a MacArthur Network on Inequality

and Economic Performance seminar at Massachusetts Institute of Technology, and the

Harvard Summer Institute on Inequality and Social Policy for providing useful comments. I especially acknowledge useful comments and criticisms from David Ellwood,

Peter Gottschalk, Christopher Jencks, Erzo Luttmer, Abhijit Banerjee, Duncan Snidel,

and anonymous reviewers. Direct correspondence to Susan Mayer, Harris School, 1155

East 60th Street, Chicago, Illinois 60637.

䉷 2001 by The University of Chicago. All rights reserved.

0002-9602/2001/10701-0001$02.50

AJS Volume 107 Number 1 (July 2001): 1–32

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

1

American Journal of Sociology

family’s own income from other effects of inequality. If growing income

inequality contributes to inequality in educational attainment between

children from rich and poor families, inequality in one generation will be

perpetuated in the next generation.

This article focuses entirely on changes in the overall dispersion of

household income. Inequality of household income increased between

1970 and 1980 and increased even more between 1980 and 1990 (Karoly

1993; U.S. Bureau of the Census 1998). Inequality among families with

children also grew (Lichter and Eggebeen 1993). Changes in inequality

between groups, such as blacks and whites or men and women may or

may not parallel changes in the level of overall economic inequality, and

their effect on educational attainment may differ from the effect of overall

economic inequality.

The proportion of 25–29 year olds who had graduated high school or

earned a GED increased from 75.4% in 1970 to 85.4% in 1980, then did

not change between 1980 and 1990 (U.S. Department of Education 1998,

table 8), when inequality increased.2 The percentage of 25–29 year olds

who had enrolled in college declined from 44% in 1970 to 52% in 1980,

and the percentage who had graduated college increased from 16.4% in

1970 to 22.5% in 1980. Neither college enrollment nor college graduation

increased between 1980 and 1990 (U.S. Department of Health and Human

Services 1998).

Ellwood and Kane (1999) show that between the early 1980s and 1992

the proportion of children in the poorest income quartile who went on to

some postsecondary schooling increased from 57% to 60%. But the proportion of children in the top income quartile getting some postsecondary

schooling increased from 81% to 90%. Thus the increase in college entrance rates was greater among affluent than among low-income children,

suggesting that the growth in inequality may have had different effects

on children from different family backgrounds.

HYPOTHESES ABOUT THE EFFECT OF INEQUALITY

Income inequality can affect educational attainment in several ways. The

first is through the incentives provided by higher returns to schooling.

The second is through the declining utility of family income. Income

inequality can also affect educational attainment through processes that

are independent of a family’s own income. Social scientists have identified

at least three such processes. They involve changes in subjective feelings

2

The percentage of 16- to 24-year-old high school graduates who had a GED increased

from 15.8% in 1980 to 17.8% in 1990.

2

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

of relative deprivation or gratification, changes in the political processes

that shape educational opportunities and costs, and changes in economic

segregation.

Change in Incentives

Part of the growth in inequality in the United States was due to increased

returns to schooling (Murphy and Welch 1992; Juhn, Murphy, and Pierce

1993).3 Because higher returns increase the incentive for children to stay

in school, we would expect educational attainment to increase when economic inequality increases.4

The Declining Utility of Family Income

If the relationship between educational attainment and parental income

is linear, then when the rich gain a dollar and the poor lose a dollar, the

educational attainment of the rich will increase by exactly as much as

the educational attainment of the poor decreases, leaving the mean unchanged. However, suppose that a 1% increase in income generates the

same absolute increment in educational attainment, regardless of whether

income is initially high or low. If the relationship between parental income

and children’s schooling takes this semilogarithmic form, and all else is

equal, a costless redistribution of income from richer to poorer households

will increase children’s mean educational attainment, because shifting a

dollar from the rich to the poor increases the education of poor children

by a larger percentage than it decreases the education of rich children.5

3

Rising returns to schooling is not the main source of inequality growth. The withineducation group variance of income rose almost as fast as the between-group variance

of income (Juhn et al. 1993; Karoly 1993), and educational attainment accounts for

only 15%–20% of the variance in income initially.

4

Welch (1999) notes that the proportion of men working full-time, year around, with

at least one year of college increased greatly after 1980 when the returns to schooling

also increased. Welch takes this as evidence that men responded to the increase in

returns to schooling by getting more schooling and that the increase in schooling is

therefore a benefit of the rise in inequality. However, the test of the response to the

rise in the return to schooling is not the increase in schooling among workers but

rather the change in schooling for all members of young cohorts.

5

Mean educational attainment might not increase when the rich get richer, and even

if it does, the increase might not be efficient. If the rich “overinvest” in schooling when

they get richer and the poor “underinvest,” the mean level of educational attainment

might stay the same, but the efficiency of the investment would decline. In this article,

I assume that no one overinvests in schooling.

3

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

Relative Deprivation and Gratification

Social comparison theory assumes that individuals evaluate themselves

relative to others. Relative deprivation theory holds that people compare

themselves to others who are more advantaged than themselves (Merton

and Kitt 1950; Davis 1959; Runciman 1966; Williams 1975).6 Imagine two

families with the same income. Family A lives in a wealthy area, while

family B lives in a poor area. Assuming their choice of where to live is

entirely exogenous and that other families in the same area are their only

reference group, relative deprivation theory predicts that members of

family A will feel more deprived than members of family B. Feelings of

relative deprivation can lead to isolation and alienation from the norms

and values of the majority. If children feel relatively deprived, they may

be less inclined to study or stay in school. Relative deprivation can also

make parents feel stressed and alienated, lowering their expectations for

their children or reducing the quality of their parenting (McLoyd 1990).

Relative deprivation theories assume that children or parents compare

themselves to others who are better off, while largely ignoring those who

are worse off. If parents all compare themselves to the richest people in

society, for example, they will feel poorer whenever the rich get richer.

Of course, people also compare themselves to others who are worse off.

Sociologists refer to this as “relative gratification” (Davis 1959). If either

children or their parents mostly compare themselves to the poorest people

in society rather than to the richest, increases in inequality will make

most people feel richer because the distance between them and the people

at the bottom of the distribution will grow. If people mostly compare

themselves to some real or imagined national average, increases in inequality will make the rich feel richer and the poor feel poorer. How this

will affect either educational attainment or other outcomes is

unpredictable.

Relative deprivation usually cannot be directly observed, so it is often

inferred from its behavioral manifestations. A large social-psychological

research literature uses experimental evidence to document the importance of interpersonal comparisons, in general, and relative deprivation,

6

An important distinction is between individual relative deprivation, in which an

individual compares his or her personal situation to the situation of other individuals,

and group relative deprivation, in which a person compares his or her relevant group’s

situation with the situation of another group. Growing inequality can affect both sorts

of relative deprivation, but I mainly emphasize individual comparisons, not group

comparisons. Individual comparisons are more likely to lead to isolation and stress,

while group comparisons are more likely to lead to collective action (Gurr 1970; Smith,

Spears, and Hamstra 1999).

4

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

in particular.7 Most of the sociological research on the importance of

interpersonal comparisons to educational outcomes has been done in the

context of neighborhood and school effects. In one of the earliest of such

efforts, Davis (1966) argued that reference groups have both a comparison

function and a normative function (as “sources and reinforcers of standards”). Although the latter could make living in an affluent area an

advantage for low-income children, the former could make it a disadvantage because it fosters academic competition and relative deprivation.

Davis showed that the more academically selective a college was, the

lower any given student could expect his grades to be. As a result, students

who chose selective colleges were less likely to choose careers that required

graduate training. Other studies find that attending high school with highachieving classmates reduces educational attainment, while attending

school with high-SES classmates increases it. Because SES and achievement are correlated, attending a high-SES school has little net effect on

educational attainment. (See Jencks and Mayer [1990] for a review of

these studies.) These studies do not prove that children feel relatively

deprived when they must compete with higher achieving students, but

they demonstrate that more advantaged classmates or neighbors can be

both an advantage and a disadvantage. When a child’s own family income

stays the same but inequality increases, the child will be exposed both to

more advantaged and more disadvantaged children. This could have either positive or negative effects on the child’s educational attainment.

Changes in Political Behavior

Changes in inequality can affect political behavior. Some research suggests

that increases in economic inequality may decrease voters’ willingness to

support redistributive policies at the national level (Perotti 1996; Alesina

and Rodrik 1994). This could also happen at the state or local level. For

example, high levels of inequality could encourage the rich to enroll their

children in private schools, making them less interested in supporting

public schools. But high levels of inequality could also increase voters’

willingness to support redistributive policies if they fear political instability (Piven and Cloward 1993) or think that poverty contributes to

crime.8 Other research suggests that redistributive spending reduces economic inequality at the national level (Gustafsson and Johansson 1999).

7

For compilations of recent research and summaries of older research see Ellemers,

Spears, and Doosje (1997) and Suls (1991).

8

Relative deprivation is an important explanation for the relationship between crime

and economic inequality (Nettler 1984; Messner and Tardiff 1986; Rosenfeld 1986) and

for collective action more generally (Gurr 1970).

5

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

Redistributive policies can thus be both a cause and an effect of economic

inequality. Disentangling cause from effect using cross-national data (as

these studies do) is difficult, because samples are generally very small, lag

structures are uncertain, and exogenous shocks that constitute national

experiments are rare. Still it is clear that changes in economic inequality

can in principle affect political support for redistributive policies, which

can thereby affect spending on schools.9

Nationwide, per pupil expenditures for elementary and secondary

schooling have increased since 1970, and spending has become more equal

across school districts in many states (Murray, Evans, and Schwab 1998;

Card and Payne 1998). Research on the effect of per pupil school expenditures on children’s school performance is equivocal (Elliott 1998; Hanushek 1996; Hedges, Laine, and Greenwald 1992), but most evidence

suggests that higher expenditures improve children’s test scores by at least

a little, which could in turn increase their educational attainment.

Research also shows that higher college tuition reduces enrollment and

graduation. (Kane [1999] summarizes this research.) Although real tuition

rates at state four-year colleges and universities have increased since 1980,

so has both state and federal financial aid for college students (U.S. Department of Education 1998). The number of two-year community colleges, which have low tuition, has also increased. However, the extent to

which state differences in economic inequality affect the cost of attending

a public college in the state is unknown.

Economic Segregation

The effect of economic inequality depends to some extent on the geographical proximity of the rich to the poor. This assumption is built into

conventional measures of inequality, which describe the dispersion of income among all households in some geographic area, such as a nation,

state, or neighborhood. Durlauf (1996) argues that as inequality increases,

the rich and poor have less in common and therefore segregate more

geographically. According to this argument, the degree of economic in-

9

Given a progressive tax rate, anything that makes the rich richer would increase the

revenue available to pay for schools. Over the long run, voters can decide to change

the progressivity of taxes. On average, state and local taxes are slightly progressive.

For example, in 30 large cities in 30 different states in 1995, a family of four with an

income of $25,000 paid an average of 8.2% of their income in state and local taxes,

while a family of the same size with an income of $75,000 paid on average 9.6% of

their income in taxes (U.S. Bureau of the Census 1998). Some economists assume that

less redistributive tax policies lead to more economic growth and hence that inequality

(within broad limits) is associated with economic growth. Greater economic growth

could lead to greater tax revenues, even if the tax rate declines.

6

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

equality at, say, the city level may affect the degree of economic segregation within the city. This will in turn affect the degree of economic

inequality within neighborhoods. Wilson (1987) argues that economic segregation causes economic inequality, rather than the other way around.

This happens because economic and racial segregation reduce the quality

of inner-city schools and other institutions and leads to a spatial mismatch

between jobs and low-skilled workers. Both Durlauf and Wilson argue

that economic segregation hurts children’s life chances.

A pernicious potential consequence of economic inequality is to recreate

economic inequality in the next generation. This could happen if inequality benefits advantaged children and hurts disadvantaged children.

For example, if economic inequality is associated with increased economic

segregation, this could hurt low-income children’s educational attainment

while increasing high-income children’s educational attainment. Because

educational attainment is associated with future earnings, this could exacerbate inequality in the next generation.

The remainder of this article assesses these possibilities empirically. I

first estimate the effect of economic inequality on children’s educational

attainment. Second, I test the hypothesis that the effect of inequality is

different for rich and poor children. Third, I assess how much of the effect

of inequality is due to the increase in returns to schooling. Fourth, I assess

how much of the effect of inequality is due to the nonlinear relationship

between parental income and children’s educational attainment. Finally,

I test the hypotheses that the effects of inequality are the result of changes

in (1) the level of economic segregation in states, (2) state per pupil expenditures on primary and secondary schooling, and (3) the cost of attending college at state institutions. Both economic segregation and the

political behavior that affects school resources can be behavioral manifestations of either relative deprivation or relative gratification. Because

it is impossible in my data to measure feelings of relative deprivation or

gratification directly, it is impossible to completely isolate their effects.

DATA AND METHODS

Most research on the social consequences of economic inequality estimates

a model in which inequality in a geographic area, such as a nation or

state, predicts an aggregate-level outcome, such as the mortality rate or

the crime rate for the geographic area. Most research on educational

attainment, in contrast, uses individual-level data to predict years of

schooling from measures of family background and characteristics of the

labor market. Some research combines individual-level data with aggregate data to predict, say, the effect of neighborhood social composition

7

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

on children’s educational attainment, holding constant their family background. These models of “neighborhood effects” or “school effects” have

many well-known estimation problems (Tienda 1991; Manski 1993; Jencks

and Mayer 1990). Nonetheless, such models do solve some of the problems

associated with using exclusively aggregate data. In this article, I estimate

models similar to those used to estimate neighborhood effects. I estimate

the effect of state-level inequality on individual children’s educational

attainment, often holding the children’s family characteristics constant.

Level of Aggregation

Economic inequality may have different effects at different levels of aggregation. Theory provides little guidance as to which geographic unit is

most relevant for the relationship between inequality and educational

attainment. If school financing plays a crucial role in educational attainment, the relevant units of aggregation are the political jurisdictions that

fund public schools and universities. Nationwide, states and local school

districts typically provide roughly equal amounts of money for elementary

and secondary schooling, with relatively little money coming from the

federal government. State funds tend to equalize district funding, so the

state pays a greater share for poor districts and a smaller share for rich

districts (Murray et al. 1998). It follows that decisions about the degree

of inequality in school district funding are primarily made at the state

level. If household income inequality affects educational outcomes primarily by affecting taxpayers’ inclination to fund public education, the

state may be the right level of aggregation.

On the other hand, school districts also share in funding decisions and

make significant decisions about policies that affect educational outcomes.

Income inequality within school districts might affect both voters’ inclination to pay taxes for schools and other school policies. However, parents

often choose their school district partly on the basis of who lives there.

If the same parental characteristics that cause parents to choose one district over another also affect their children’s educational outcomes, and

if these parental characteristics are not measured accurately, the estimated

effect of school district inequality on educational attainment could be

biased. This form of selection bias should be less important for estimating

the effect of state-level characteristics on educational attainment because

parents are less likely to move to a different state than to move to a

different school district in order to improve their children’s educational

prospects.

Theories about the effect of income inequality that involve interpersonal

comparisons are ambiguous about what is the most relevant geographic

unit because it is not clear how individuals select the people to whom

8

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

they compare themselves. If children compare themselves to the people

they see on television, the nation as a whole is probably the relevant

comparison group. If the nation is the relevant comparison group, the

only way to study the impact of inequality would be to make crossnational comparisons or comparisons across time. If children compare

themselves mainly to their neighbors and classmates, inequality in a relatively small geographic area, such as a neighborhood or school attendance area, may be more relevant than inequality in a larger unit. But,

as I have noted, selection problems are likely to be more serious at the

level of the neighborhood or school district. In this article, I use states as

aggregation units. States may not be the only or even most salient unit

of aggregation, but there are sound theoretical reasons to expect states to

be important. As we will see, they are empirically important.

Data

The educational attainment of a state’s residents affects inequality, and

vice versa. This is not likely to be a problem in the short run for high

school graduation because while economic inequality in a state can affect

the probability that a teenager will graduate from high school, it takes

some time for the high school graduation rate to affect the dispersion of

household incomes in the state. Using economic inequality in a state to

predict the educational attainment of the adults in the state would pose

a more serious problem because the direction of causality is unclear. This

problem is exacerbated by the fact that many adults no longer live in the

state where they were raised, and the distribution of income in a state

may affect the kinds of migrants who settle there.

To solve these problems, I use data from the Panel Study of Income

Dynamics (PSID) to estimate the effect of state economic inequality measured during adolescence on children’s eventual educational attainment.

I estimate separate models for a child’s chances of completing high school,

entering college, completing four years of college, and for years of completed schooling, because income inequality is likely to have different

effects on enrollment choices at different ages and at different levels of

schooling. My PSID sample includes all respondents who were in the data

set when they were 12–14 years old and when the educational outcome

of interest was measured. I count respondents as having graduated high

school if they reported that they had completed 12 or more years of

schooling when they were 20 years old. The analysis of high school graduation therefore includes all respondents who were in the sample at ages

12–14 and at age 20 (N p 3,504). I count respondents as having attended

college if they reported completing one or more years of college by the

time they were 23 years old. I count them as having graduated college if

9

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

they report having completed at least 16 years of schooling by the time

they were 23 years old. For these outcomes, therefore, respondents had

to be in the PSID sample when they were ages 12–14 and at age 23

(N p 3,240).10

Most of my measures of state characteristics come from the 1970 1%

Public Use Microdata Sample (PUMS) of census data and from the 1980

and 1990 5% PUMS. Values for a few other state characteristics come

from published sources, as described in the appendix. The appendix also

provides more detail on the sample and the variables used in this article.

The Measure of Inequality

I use the Gini coefficient as a measure of inequality mainly because of its

familiarity.11 The correlation between a state’s Gini coefficient and two

other commonly used measures of economic inequality (the standard deviation of log income and the ratio of the ninetieth to the tenth percentile)

are very high (0.925 and 0.963). The correlation between inequality in the

top half of a state’s income distribution (the 90 to 50 ratio) and inequality

in the bottom half of its distribution (its 50 to 10 ratio) is also high (0.721).

The correlation between each of these alternative measures of state income

inequality with individual’s educational attainment is negative but small

(none differs from zero by more than 0.061). Thus it is not clear that any

one measure of economic inequality is likely to be better than another at

predicting educational attainment.

I use census data to calculate the Gini coefficient for each state in 1970,

1980, and 1990. I then use linear interpolation to create a Gini coefficient

10

I measure high school graduation two years after students’ normal date of graduation

and college enrollment up to five years after the normal enrollment age. However, I

measure college graduation only about a year after the normal college graduation date.

If inequality affects the timing of school transitions, these different “grace” periods

could be a problem. For example, if rising inequality delays but does not reduce college

graduation, my estimates would be upwardly biased estimates of the effect of inequality

on college completion. Because I require children to be in the sample from adolescence

on, increasing the age at which I measure educational outcomes reduces the sample

size considerably. However, I did experiment with measuring college completion and

educational attainment at age 25 (three years after the normal age of college graduation), and the point estimates were similar to those in the models reported here.

11

The Gini coefficient is the proportion of the total area below the 45 degree line that

lies above the Lorenz curve, which plots the cumulative percentage of households

against the cumulative percentage of income received by them. See Firebaugh (1999)

for a comparison of inequality measures across countries. See Atkinson (1970, 1983)

for a discussion of statistical differences among inequality measures.

10

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

for each state in each intercensus year for which I have PSID data.12 I

assign children the level of inequality in their state when they were 14

years old. The national Gini coefficient increased from 0.361 in 1970 to

0.368 in 1980 then to 0.381 in 1990. The Gini coefficient also varies across

states. In 1970, the lowest state Gini coefficient was 0.320, and the highest

was 0.427. In 1990, the lowest was 0.337, and the highest was 0.438. The

standard deviation of the Gini coefficient across states was 0.027 in 1970,

0.018 in 1980, and 0.022 in 1990.

RESULTS

The Effect of Inequality on Mean Educational Attainment

If the degree of economic inequality in a state were a random accident,

we could compare educational attainment in states with high and low

levels of inequality, assume that all else was more or less equal across

states, and treat observed differences in educational outcomes as a byproduct of economic inequality. The effect of inequality (Gini) in state s

and year t on the educational attainment (E) of individual i would then

be given by the value of bg in the following model:

Eist p b0 ⫹ bgGinist14 ⫹ e ist ,

(1)

where e ist is the usual random error term and t14 indicates that it is

measured at age 14.

Model 1 of table 1 shows the results of equation (1) for all four outcomes.

In all models in table 1, the t-statistics are corrected for the fact that

individuals are clustered in states and years. For high school graduation,

enrolling in college, and graduating college, the results are from a probit

model. The cell entries are partial derivatives evaluated at the mean of

the distribution. The results for years of schooling are estimated with an

OLS model. Cell entries are unstandardized regression coefficients. Using

model 1 for the entire sample, the effect of the Gini coefficient is negative

for all outcomes, but it is not statistically significant at the 0.05 level for

any outcome.

States vary in many ways besides their level of economic inequality.

Some of these differences are associated with both economic inequality

and with educational attainment. In this article, I try to estimate what

would happen to educational attainment as a result of an exogenous

12

Nationwide inequality increased somewhat more rapidly in the later part of the

1970s than in the earlier part of the decade. During the early 1980s, the increase in

inequality was nearly linear. Any deviation from the linear trend is a source of measurement error in the inequality measure, and thus it probably biases the coefficient

of the Gini coefficient toward zero.

11

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

TABLE 1

Effect of Gini Coefficient on Education Outcomes

Model and Outcome

Model 1, no controls:

High school graduate (probit partial derivative) . . . . . . . .

Enrolled in college (probit partial derivative) . . . . . . . . . . .

Graduated college (probit partial derivative) . . . . . . . . . . . .

Years of schooling (OLS unstandardized coefficient) . . .

Model 2, controls region and year dummies, %black,

%Hispanic, mean income, and unemployment rate:

High school graduate (probit partial derivative) . . . . . . . .

Enrolled in college (probit partial derivative) . . . . . . . . . . .

Graduated college (probit partial derivative) . . . . . . . . . . . .

Years of schooling (OLS unstandardized coefficient) . . .

Model 3, adds returns to schooling:

High school graduate (probit partial derivative) . . . . . . . .

Enrolled in college (probit partial derivative) . . . . . . . . . . .

Graduated college (probit partial derivative) . . . . . . . . . . . .

Years of schooling (OLS unstandardized coefficient) . . .

Full

Sample

High

Income

Low

Income

⫺.619

(⫺1.480)

⫺.347

(⫺.560)

⫺.559

(⫺1.130)

⫺4.579

(⫺1.668)

.118

(.220)

2.252

(2.560)

.881

(1.180)

6.128

(1.579)

.680

(.970)

⫺.226

(⫺.330)

⫺.424

(⫺1.060)

⫺2.994

(⫺.974)

.222

(.270)

3.404

(2.900)

1.024

(1.150)

10.565

(2.160)

.908

(.940)

4.763

(2.930)

2.715

(2.010)

19.769

(2.978)

⫺1.417

(⫺.990)

⫺.190

(⫺.150)

⫺1.639

(⫺2.320)

⫺8.750

(⫺1.386)

.070

(.090)

2.892

(2.440)

.711

(.770)

8.637

(1.760)

.885

(.910)

4.090

(2.460)

2.346

(1.660)

18.604

(2.739)

⫺1.564

(⫺1.100)

⫺.213

(⫺.160)

⫺1.734

(⫺2.330)

⫺9.586

(⫺1.518)

Note.—t-statistics are in parentheses and are corrected for clustering in states and years.

change in economic inequality. An exogenous increase in inequality might

result from polarization of the job distribution due to industrial restructuring or from a technological innovation that changed the skill needs of

employers and therefore changed the wage premium for some skills. In

response to such changes, states might differ in how much inequality

increased depending on the skill distribution in the state, the available

mechanisms for increasing high-premium skills, the generosity of the

state’s social programs, the “culture” of the state, and many other factors.

To estimate the effect of an exogenous change in inequality, one must

control all the exogenous determinants of inequality that also affect educational attainment.

To control potentially relevant omitted variables, I first include dummy

12

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

variables for the Northeast, South, and Midwest to control characteristics

of the region that remain unchanged over the period of observation. An

alternative strategy would be to control state fixed effects. Such a model

would be equivalent to estimating the within-state effect of a change in

inequality. This strategy has the advantage of controlling all invariant

characteristics of states. However, it has three important disadvantages.

First, it can magnify measurement error in independent variables, including the measure of inequality, which would downwardly bias the

estimated effects. Second, if the lag structure of the model is not correctly

specified, this too can result in downwardly biased estimates of inequality.

Third, including state fixed effects greatly reduces the degrees of freedom

available to estimate the model, which in turn increases the standard

errors of the estimates. Nonetheless, I test the sensitivity of my model to

inclusion of state rather than region fixed effects and report those results

below.

I control year fixed effects to account for the secular national trend in

educational attainment. With both region and year fixed effects, all variation in inequality derives from a combination of changes in inequality

within states over time and differences in equality among states in the

same region.

I also control a set of exogenous state-level determinants of inequality

that can change over time. Past and present racial and ethnic discrimination means that racial and ethnic diversity can affect economic inequality and educational attainment. I control the percentage of state

residents who were African-American and the percentage who were Hispanic in the year a child was 14 years old.13

A state’s average household income is negatively correlated with economic inequality, and mean household income could obviously affect children’s educational attainment. The correlation between mean household

income and the state Gini coefficient was ⫺0.724 in 1970, ⫺0.425 in 1980,

and ⫺0.559 in 1990.14

13

In principle, inequality could affect the racial composition of the state, and vice

versa. But in practice, inequality cannot have much effect on a state’s racial composition, because the interyear correlation for both percentage black and percentage

Hispanic are about 0.98.

14

The negative correlation could reflect a negative effect of inequality on mean income

rather than the other way around. However, empirical research on the relationship

between economic inequality and economic growth at the national level is inconclusive.

(See Forbes [2000] for a recent review of this evidence.) In addition, it is not clear that

the (unknown) factors that generate a correlation between inequality and subsequent

economic growth at the national level would also apply to the economies of U.S. states.

Because inequality and mean income are highly correlated, and because it seems likely

that state income levels affect inequality more than vice versa in the United States, I

control state mean income.

13

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

I also control the state unemployment rate measured at the same time

as inequality is measured. Fluctuations in the unemployment rate are

mainly attributable to short-term fluctuations in the business cycle and

do not contribute much to the level of inequality in the state. The correlation between the state unemployment rate and the Gini coefficient is

only ⫺0.047. However, among states with the same mean income, those

with high levels of unemployment are likely to have more inequality

because unemployment reduces the income of some state residents. Research is inconclusive about the effect of local unemployment rates on

educational attainment (Betts and McFarland 1995; Kane 1994; Manski

and Wise 1983; Grubb 1988).

With fixed effects and control variables, the model becomes

Eist p b0 ⫹ bgGinist14 ⫹ bx Xst14

⫹ gr ⫹ gt ⫹ e ist ,

where gr is a set of dummies for the regions and gt is a set of dummies

for the year in which inequality is measured. In this model, X represents

exogenous state characteristics that may have changed over time, including racial composition, the unemployment rate, and mean income.

The results from this model are shown in model 2 in the first column

of table 1. With these controls, the effect of the Gini coefficient is positive

for all outcomes and statistically significant at the 0.05 level for enrolling

in college and years of schooling.

I next try to separate the effect of the incentive provided by greater

returns to schooling from other effects of inequality. Imagine two states

where inequality grows by the same amount and mean income does not

change. In state A, inequality within education groups increases, perhaps

because employers increase the pay of workers with noncognitive skills

not usually learned in school. In state B, inequality grows because the

returns to schooling increase. In both states, the rich, on average, get

richer and the poor, on average, get poorer. If the effect of inequality on

educational attainment were entirely due to the incentive effect of increased returns to schooling, educational attainment would increase in

state B but not in state A.

My measure of returns to schooling is the average effect of an extra

year of schooling on log wages in a given state and year, estimated for

workers ages 18–65. I estimate the effect of state returns to schooling (Rst)

when a child was 14 years old on his or her eventual educational attainment. I use returns when a child was age 14 rather than returns at a later

age for two reasons. First, the decision about how much schooling to get

is intertwined with decisions about what to study: a student who does

not expect to attend college often makes decisions about what to study

in high school that, in turn, make college attendance very difficult. Second,

I assume that the rate of return to schooling often affects individual

14

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

enrollment decisions indirectly, by affecting the way “significant others”

value education. These indirect influences are likely to mean that current

attitudes reflect past as well as current returns. Thus the model becomes

Eist p b0 ⫹ bgGinist14 ⫹ br Rst14 ⫹ bx Xst14

⫹ gr ⫹ gt ⫹ e ist.

The results for this model are shown in model 3 in the first column of

table 1. States with high returns to schooling also have higher levels of

inequality, so the effect of the Gini coefficient in these models is smaller

than in models that omit returns to schooling. These results suggest that

a one standard deviation increase in a state’s Gini coefficient (i.e., a change

of 0.02) increases children’s chances of enrolling in college by 0.058 (13.4%

of the mean) and increases their overall educational attainment by 0.173

years. These results suggest that the positive effect of inequality on educational outcomes is only partly due to the incentive provided by the

rising returns to schooling. If the return to schooling were an entirely

exogenous cause of the increase in inequality, this model would provide

the best estimate of the effect of inequality on educational attainment.15

The Effect of Inequality on Rich and Poor Children

The last two columns in table 1 show the results of these three models

separately for high- and low-income children. “High-income” children are

those in the top half of the income distribution. “Low-income” children

are those in the bottom half of the income distribution. Dividing the

sample at the midpoint allows all variables to interact with household

income in a way that is easy to interpret and preserves enough high- and

low-income cases for a meaningful analysis. Other divisions of the sample,

such as quartiles, provide qualitatively similar results but with larger

standard errors. A model that interacts household income with the Gini

coefficient and all other relevant variables is difficult to interpret and also

results in very large standard errors. Dividing the sample in half is instructive even though it may not capture all the nuances of the effect of

inequality at different parts of the income distribution.

Columns 2 and 3 for model 3 in table 1 show that living in a highinequality state improves all educational outcomes for high-income children and hurts all educational outcomes for low-income children. The

15

It is possible that schooling opportunities in a state affect the returns to schooling.

Absent migration, states with less postsecondary schooling opportunities will have a

lower supply of highly educated workers, raising the wage premium and creating more

inequality. This would also presumably increase the incentive to go to college. Over

time, states would then reach an equilibrium in which they provided the “right” supply

of college graduates to meet the state’s demand for college graduates. If this happens,

there would be little variation across states in returns to schooling.

15

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

positive effects of the Gini coefficient on college enrollment and years of

schooling are statistically significant at the 0.05 level for high-income

children. Only the negative effect of inequality on graduating college is

statistically significant at this level for low-income children. Thus the

overall positive effect of the Gini coefficient on enrolling in college and

on years of schooling in column 1 is entirely due to the positive effect for

high-income children. Inequality has no overall effect on college entrance

because the positive effect of inequality for high-income children is offset

by the negative effect for low-income children.

The Processes through which Inequality Affects Educational

Attainment

I first separate the effect of inequality on educational attainment that is

due to the nonlinearity of families’ own income from the other effects of

inequality. These estimates answer the question, How would children’s

educational attainment differ if their families had the same income (averaged over several years) and they lived in states with the same mean

income and returns to schooling, but different levels of economic

inequality?

To estimate the importance of the declining utility of family income,

we first must determine the approximate functional form of the relationship between parental income and children’s educational attainment. Table 2 shows that any of three nonlinear functions of household income

averaged over three years is a better predictor of educational outcomes

than a linear model in which an extra dollar always has the same effect,

regardless of a family’s initial income. But the differences among the three

nonlinear forms are relatively small. Thus while the effect of family income is probably nonlinear, these data do not allow one to choose among

the nonlinear specifications. I use log income as a measure of nonlinearity

because of its theoretical appeal and ease of interpretation. To determine

how much of the effect of inequality on educational attainment is due to

the nonlinear relationship between family income and children’s educational attainment, I estimate the following equation:

Eist p b0 ⫹ bY lnYist12⫺14 ⫹ bgGinist14 ⫹ bx Xst14

⫹ gr ⫹ gt ⫹ e ist.

In this equation, bY captures the effect of family income when children

were 12–14 years old, while bg captures the effects of inequality that

operate independently of a household’s own income.

Comparing model 4 in table 3 with model 3 in table 1 for all children

shows that controlling log household income decreases the positive effect

of the Gini coefficient on all outcomes. But the reductions are modest,

and the remaining effects of the Gini coefficient are still positive. Con16

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

TABLE 2

Effect of Selected Functional Forms of Household Income on Measures of

Educational Attainment

Dependent Variable

Linear

Logarithm

Cube Root

Logarithm Plus

Extreme Deciles

High school graduate (x2) . . .

Entered college (x2) . . . . . . . . . .

College graduate (x2) . . . . . . . .

Years of schooling (R2) . . . . . .

174.52

386.94

252.06

.127

214.69

401.35

273.21

.129

218.26

417.50

278.61

.136

219.28

425.81

290.99

.137

Source.—PSID data described in the appendix.

Note.—The model with “logarithm plus extreme deciles” predicts educational attainment from log

household income, a variable equal to one if the child was in the richest income decile and another

variable equal to one if the child was in the poorest income decile.

trolling parental income reduces the effect of the Gini coefficient on the

educational outcomes of both high- and low-income children. But most

of the positive effect of inequality on high-income children and most of

the negative effect of inequality on low-income children does not operate

through the effect of families’ own income.

Most models of neighborhood or school effects begin with an individuallevel model of an outcome such as high school graduation then add an

aggregate-level variable for the neighborhood or school social composition. For example, such a model might predict high school graduation

from family background characteristics such as parental education and

income and a school or neighborhood-level variable such as mean family

income (Crane 1991; Evans, Oates, and Schwab 1992; Mayer 1991). In

such models, family background variables are intended to control for the

fact that parents select the schools their children attend and the neighborhoods in which they live. To see if parental selection across states with

different levels of inequality is a problem, I control two other measures

of family background, namely, whether the child is black and parental

education. Model 5 in table 3 shows that for the whole sample controlling

these variables somewhat reduces the effect of the Gini coefficient on the

measures of educational attainment. With these background measures

controlled, omitted family background variables are less likely to be a

source of bias, since other family background characteristics that affect

state of residence are likely to be correlated with parents’ race and

education.

The results in model 5 in table 3 suggest that increasing the Gini coefficient by 0.02 would increase college enrollment by 0.047 or about 10%

of the mean through processes unrelated to children’s own income. This

increase accrues entirely to high-income children. A 0.02 increase in the

Gini coefficient would increase schooling by 0.322 years for high-income

17

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

TABLE 3

The Effect of the Gini Coefficient in Models Separating the Effect of

Parents’ Income from Other Effects of Inequality

Model and Controls

Model 4, controls variables in model 3 plus log household income:

High school graduate (probit partial derivative) . . . . . . . .

Enrolled in college (probit partial derivative) . . . . . . . . . . .

Graduated college (probit partial derivative) . . . . . . . . . . . .

Years of schooling (OLS unstandardized coefficient) . . .

Model 5, adds child’s race, parents’ education:

High school graduate (probit partial derivative) . . . . . . . .

Enrolled in college (probit partial derivative) . . . . . . . . . . .

Graduated college (probit partial derivative) . . . . . . . . . . . .

Years of schooling (OLS unstandardized coefficient) . . .

Full

Sample

High

Income

Low

Income

.051

(.060)

2.636

(2.170)

.213

(.240)

7.157

(1.572)

.600

(.640)

3.560

(2.170)

1.998

(1.420)

17.057

(2.712)

⫺1.372

(⫺.940)

⫺.028

(⫺.020)

⫺1.513

(⫺2.190)

⫺7.970

(⫺1.275)

⫺.281

(⫺.350)

2.344

(1.850)

.077

(.080)

5.435

(1.184)

.292

(.320)

3.150

(1.890)

1.957

(1.340)

16.124

(2.499)

⫺1.816

(⫺1.200)

⫺.243

(⫺.180)

⫺1.586

(⫺2.170)

⫺10.161

(⫺1.617)

Note.—t-statistics are in parentheses and are corrected for clustering in states and years.

children and decrease schooling by 0.203 years for low-income children

through processes unrelated to family income.

Next I test the hypothesis that economic segregation between school

districts accounts for the effect of economic inequality. If we divide a state

into mutually exclusive geographic areas such as school districts, we can

decompose the total variance of household income for the state (jts2 ) into

two additive components: a between-area component (jbs2 ) and a within2

area component (jws

). This yields the identity

2

jts2 p jbs2 ⫹ jws

.

The ratio of the between-area variance to the total variance (jbs2 /jts2 ) is

a measure of economic segregation (Jargowsky 1996). In the absence of

economic segregation, all areas have the same mean income and

jbs2 /jts2 p 0. With complete economic segregation, there is no income var18

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

iation within geographic areas and jbs2 /jts2 p 1.16 Thus my measure of

economic segregation is the percentage of a state’s total income variance

that is between school districts in the state.17

Comparing model 6 in table 4 to model 5 in table 3 shows that controlling economic segregation hardly changes the effect of the Gini coefficient on any measure of schooling. This is because, contrary to what

one might have expected, economic segregation between school districts

has a small and statistically insignificant (at 0.05) effect on all these measures of educational attainment, so controlling it does not appreciably

alter the effect of the Gini coefficient. (See tables A2–A4 in the appendix.)

To see if the effect of inequality operates through state support for

schooling, I estimate a model that controls state per pupil expenditures

on primary and secondary schooling, college tuition at the state’s flagship

institution, and the amount of grant money available for college tuition

per state resident ages 15–24 years old.

Model 7 in table 4 shows that, for the whole sample, controlling these

variables reduces the positive effect of inequality on enrolling in college

to close to zero. It leaves a negative, statistically significant residual effect

of inequality on graduating college and a negative but statistically insignificant effect on years of schooling. Controlling these variables reduces

the positive effect of inequality on high-income children’s educational

attainment and increases its negative effect on low-income children’s educational attainment. Greater state spending on schooling is associated

with higher college enrollment and graduation, and states with more inequality spend more on schooling, thus reducing potential negative effects

of inequality.

The full regression results from this model are shown in tables A2–A4

in the appendix. High tuition reduces a child’s chances of enrolling in

and graduating from college. Thus it also reduces years of schooling. This

16

There are many other possible measures of economic segregation (White 1987; James

1986). The most commonly used measures are the “exposure index,” which gives the

average probability that members of one group live in the same neighborhood as

members of another group, and the “index of dissimilarity,” which gives the percentage

of residents with a particular characteristic who would have to move for the group

to be equally represented in all neighborhoods. These measures were developed to

estimate racial segregation, so they require classifying people into discrete categories.

Some research has measured economic segregation with such measures (Massey and

Eggers 1990), but because income is continuous, by breaking income into discrete

categories, this approach throws away potentially valuable information (Jargowsky

1996).

17

I have replicated these results using the variance of state income between census

tracts as the measure of segregation, and the results are very similar to the results

using school districts. I show the results using school districts because they are relevant

political jurisdictions for schooling outcomes.

19

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

TABLE 4

The Effect of the Gini Coefficient in Models Controlling State

Education Variables

Model and Controls

Model 6, controls variables in model 5 plus economic

segregation:

High school graduate (probit partial derivative) . . . . . . . .

Enrolled in college (probit partial derivative) . . . . . . . . . . .

Graduated college (probit partial derivative) . . . . . . . . . . . .

Years of schooling (OLS unstandardized coefficient) . . .

Model 7, adds college tuition, grants for college tuition,

and per pupil expenditure for elementary and secondary schooling:

High school graduate (probit partial derivative) . . . . . . . .

Enrolled in college (probit partial derivative) . . . . . . . . . . .

Graduated college (probit partial derivative) . . . . . . . . . . . .

Years of schooling (OLS unstandardized coefficient) . . .

Full

Sample

High

Income

Low

Income

⫺.450

(⫺.500)

2.504

(1.810)

.063

(.060)

5.465

(1.114)

⫺.104

(⫺.110)

3.323

(1.820)

1.852

(1.190)

12.378

(1.847)

⫺1.394

(⫺.820)

⫺.327

(⫺.210)

⫺1.474

(⫺1.730)

⫺8.216

(⫺1.185)

⫺.375

(⫺.430)

.193

(.130)

⫺2.404

(⫺1.920)

⫺2.709

(⫺.483)

⫺.247

(⫺.260)

.908

(.420)

⫺1.049

(⫺.540)

3.545

(.420)

⫺1.145

(⫺.750)

⫺1.876

(⫺1.260)

⫺2.156

(⫺3.150)

⫺14.395

(⫺2.110)

Note.—t-statistics are in parentheses and are corrected for clustering in states and years.

is the case whether the child comes from a high- or low-income family.

On the other hand, more state grant money for college increases college

graduation but has a small and statistically insignificant effect on enrolling

in college and on years of schooling. This is because grants for tuition

increase college attendance for high- but not low-income students. This

is consistent with other evidence on the effect of the costs of attending

college (Kane 1999). Per pupil spending on elementary and secondary

schooling increases college completion. It also increases college attendance

for low-income but not high-income children.

Sensitivity Analysis

To see if these results are robust to alternative models, I first substituted

state dummy variables for region dummy variables. Second, I estimated

a two-stage least-squares model that instruments the Gini coefficient. This

model is intended to control unobserved heterogeneity in state charac20

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

teristics that could bias the estimated effect of the Gini coefficient. In the

interest of brevity, I report only the estimates for years of schooling.

When I substitute state fixed effects in model 3 for the whole sample,

the Gini coefficient on years of schooling is 12.646, which is similar to

the 8.637 shown in table 1, but the t-statistic for the state fixed-effects

model is only 0.761. Similarly, for children above the median income, the

coefficient for the Gini coefficient is 23.809 (t p 1.099 ), and for children

below the median income, it is ⫺4.074 (t p ⫺0.066). Again, the point

estimates are similar to the estimates from models with region dummy

variables, but the t-statistics are very small. In all cases, estimates from

the region fixed-effects model were within the 95% confidence interval of

the state fixed-effects estimates.

Economic inequality has increased partly because rising returns to skill

have increased wages more in some industries than in others. I use this

fact to create an instrument for predicting changes in economic inequality

that are arguably determined by national economic forces and thus exogenous with respect to other changes at the state level. The reasoning

for the instrument is that if a technological “shock” raised wages in some

industries and not others, and if the state’s industrial mix could not respond quickly, the industrial mix when the shock occurred should produce

a state-level change in inequality over the short to medium run when all

else is equal. The industrial mix at the time of the shock would presumably

affect educational attainment at that time, but it would affect subsequent

changes in educational attainment only through its effect on inequality,

including its effect on returns to schooling. The appendix describes my

measure of industrial mix.

When I re-estimate model 3 for the whole sample as a two-sided least

squares model with this instrument, the effect of the Gini coefficient on

years of schooling is 3.769, which is about half the OLS estimate. However,

the sampling error of this estimate is very large (11.28), so this test is

inconclusive.

CONCLUSIONS

My results suggest five conclusions. First, the growth in inequality since

1970 probably did not have much affect on high school graduation.

Second, the growth in inequality since 1970 increased overall years of

schooling mainly by increasing college entrance rates. Model 3 in table 3

suggests that a one standard deviation increase in the Gini coefficient

results in a 0.058 increase in a student’s probability of going to college

(13.4% of the mean) and an additional 0.173 years of schooling.

Third, the growth in income inequality contributed to an increase in

21

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

inequality in educational attainment between rich and poor children. A

0.02 increase in the Gini coefficient is associated with a reduction of 0.192

years in low-income children’s schooling and an increase of 0.372 years

in high-income children’s schooling. This is not primarily because the

growth in inequality increased the income of rich but not poor children’s

families. The differential effect of inequality on high- and low-income

children persists even when their families’ incomes are controlled.

Fourth, the effect of inequality is only partly due to the nonlinear

relationship between parental income and children’s outcomes and the

incentive provided by increasing returns to schooling. Fifth, an increase

in per pupil expenditures at the elementary and secondary level and lower

college costs are positively associated with state inequality, and both raise

educational attainment. Greater inequality is also associated with greater

economic segregation, but this does not appear to affect children’s educational attainment.

The results in this article present a problem. Growing income inequality

raised mean educational attainment but also exacerbated disparities in

educational attainment between rich and poor children. This is likely to

contribute to economic inequality in the next generation. These findings

suggest that it is important to find ways to reduce the potentially harmful

effects of inequality on low-income children. The results in this article

suggest that higher spending on elementary and secondary schooling and

lower college tuition increase the educational attainment of low-income

children and by doing so reduce the gap in schooling between high- and

low-income children. But these efforts were not enough to prevent inequality from hurting low-income children’s educational attainment.

This article has at least three potential limitations that invite further

research. First, it focuses on inequality at the level of the state. Inequality

in smaller geographic areas could either be more or less important than

inequality at the state level (though the absence of an effect of segregation

makes this somewhat unlikely). Inequality at the national level may also

be more important than inequality in a state. Research that assesses the

importance of inequality at different levels of aggregation would be very

useful. Second, I may not have controlled all state characteristics that

contributed both to changes in inequality and in educational attainment.

This leaves the possibility that the estimates suffer from omitted variable

bias. Unfortunately, the number of state characteristics that can be included is severely limited by the number of states and the small number

of PSID families. A third closely related problem is that these estimates

probably do not include all the important processes through which economic inequality affects educational attainment. We need better theories

about the social consequences of economic inequality in general and about

its consequences for educational attainment in particular. Without strong

22

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

theories, identifying the process through which inequality operates from

empirical data involves substantial risks of both type 1 and type 2 error.

Nonetheless, this article has demonstrated the importance of understanding the consequences of economic inequality for both advantaged and

disadvantaged children.

APPENDIX

Description of the Data and Variables

PSID Data

I use data from the 1993 wave of the PSID. The high school graduation

sample includes respondents ages 20–37 years old in 1993 who were not

missing data on any variable. The sample for college enrollment, college

graduation, and years of schooling includes respondents who were ages

23–37 years old in 1993 who were not missing data. Observations are

weighted to account for the PSID sample design.

PSID variables were constructed by pooling across the 26 currently

available waves of the PSID family file (years 1968–93). Variables are

assigned to respondents based on their age. For example, I average family

income when children were ages 12–14 years. Thus it was averaged over

1985–87 for children born in 1973 and over 1990–92 for children born in

1978. Following is a description of variables created with PSID data.

Their means, standard deviations, and correlation are in table A1.

High school graduate is a dummy variable equal to “1” if the individual

had either earned a GED or had 12 years of schooling by age 20, “0”

otherwise.

Enrolled in college is a dummy variable equal to “1” if the individual

had completed at least 13 years of schooling by age 23, “0” otherwise.

Completed college is a dummy variable equal to “1” if the individual

had completed at least 16 years of schooling by age 23, “0” otherwise.

Log family income is cash income averaged over the three years when

the child was age 12–14. Income values are in 1998 dollars using the CPIU-X1 price adjustment. I use the natural logarithm of the averaged value.

Parental education is the highest year of schooling completed by the

mother when the child was age 14. If this was missing, I use the mother’s

education when the child was age 13 and so on until age 11. If all of

these values were missing, then I assigned the father’s education when

the child was age 14.

Black is a dummy variable equal to “1” if the child was African-American, “0” otherwise.

23

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

Census Data

Most of the state-level variables used in this analysis come from the 1970,

1980, and 1990 Public Use Microdata Sample (PUMS) of the U.S. Census.

In 1980 and 1990, I used the 5% samples. In 1970, I used the 1% sample

because that is what is available.

Mean household income was computed by summing the components

of income for each person in a household. Using components of person’s

income rather than person’s total income increases the detail available at

the upper tail of the distribution by avoiding Census Bureau top-coding

of total income. To reduce problems of comparability over time that arise

from changes in the Census Bureau’s top-codes for income components,

I created uniform income components and top-codes for all years. Variables are top-coded by reassigning values above the lowest ninety-ninth

percentile of positive values among the years to the median of all values

across years that lie above that lowest ninety-ninth percentile. The same

was done for negative values using the highest first percentile as the cutoff.

I sum the resulting components to get household income. All measures

of income are adjusted to 1998 dollars using the CPI-U-X1. I use this

income measure to calculate state-level measures of income and income

inequality. Persons in group quarters were excluded from all calculations.

Percentage black and percentage Hispanic are estimated using 1970,

1980, and 1990 PUMS data. I use linear interpolation to assign values

for the state in the year when the child was 14 years old.

Returns to schooling for individual i in state s and year y is estimated

for workers ages 18–65 using the following model:

lnWisy p b0 ⫹ bs Sis ⫹ e is ,

where W is the hourly wage and S is years of schooling. In this model,

bs is the percentage increase in wages due to an additional year of schooling. I experimented with 12 different measures of returns to schooling,

using different age groups and different functional forms, and estimating

separate models for men and women. I use the measure that increased

R2 the most when added to the model of the effect of inequality. This

measure also corresponds best to economic theory about the functional

form of returns to schooling and produces an estimated return to schooling

that is consistent with previous research (Winship and Korenman 1999;

Mayer and Knutson 1999; Ceci 1991).

Economic segregation is estimated by calculating the total variance of

household income for each state using the income measure described

above. I calculate mean income for each census tract in the state using

the STF4 and STF5 census files. I weight each tract mean by the population of the tract. The variance of the weighted means is the variance

24

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

of household income between census tracts. To get the within-tract variance, I subtract the between-tract variance from the total variance of

household income in the state.

Industrial mix of states in 1970 is measured by assigning each worker

between the ages of 25 and 64 in a state the national mean earnings of

workers in the same three-digit industry. I then calculate the dispersion

of these means separately for each state, weighting each industry by the

percentage of the state’s workers in the industry. Thus for each state, I

calculate the national mean wage in 1970 of workers in individual i’s

industry. I then calculate a new variable that assigns each worker in 1970

the average 1980 wage for his or her industry. I repeat this using average

wages for 1990. I calculate the standard deviations of the 1970, 1980, and

1990 measures in each state. This measure, which I call “industrial mix,”

is the amount of interindustry income inequality we would expect to find

in the state if the industrial mix had not changed between 1970 and 1990.

The R2 when I regress the change in the Gini coefficient on the change

in industry mix is 0.365.

Other State-Level Variables

Elementary and secondary public school expenditures per capita is the

state’s total expenditure for elementary and secondary public schools divided by the state population of children ages 5–17 calculated for the year

when the individual was age 14. State population data for 5–17 year olds

are from the U.S. Census Bureau’s web page. Elementary and secondary

public school expenditures are from U.S. Bureau of the Census, Statistical

Abstract of the United States annual volumes for 1970–90 (91st to 110th

eds.).

College tuition is the in-state tuition for the state’s flagship university.

The data are from the Higher Education General Information Survey

(U.S. National Center for Education Statistics, various years.)

Grants for college is the total price-adjusted need-based grant expenditures for a state divided by the number of state residents ages 15–24.

The data are from the Higher Education General Information Survey

(U.S. National Center for Education Statistics, various years.)

25

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

TABLE A1

Correlations among Variables

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

26

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

1. High school graduate . . . . . . . . . . . . . . 1.000

2. Enrolled in college . . . . . . . . . . . . . . . . .

.022 1.000

3. Graduated college . . . . . . . . . . . . . . . . .

.031

.547 1.000

4. Years of schooling . . . . . . . . . . . . . . . . .

.041

.787

.731 1.000

5. Mean household income/$1,000 . . . .

.055

.101

.068

.106 1.000

6. %black in state . . . . . . . . . . . . . . . . . . . . ⫺.044 ⫺.089 ⫺.056 ⫺.080 ⫺.320 1.000

7. %Hispanic in state . . . . . . . . . . . . . . . .

.085

.075

.042

.055

.342 ⫺.119 1.000

8. Child’s race is black . . . . . . . . . . . . . . . ⫺.114 ⫺.116 ⫺.137 ⫺.144 ⫺.124

.330 ⫺.005 1.000

9. State Gini coefficient . . . . . . . . . . . . . . . ⫺.030 ⫺.013 ⫺.022 ⫺.039 ⫺.570

.170

.188

.156 1.000

10. State returns to schooling . . . . . . . . .

.030

.036

.030

.048 ⫺.003

.502

.334

.137

.447 1.000

11. Log household income . . . . . . . . . . . .

.256

.333

.269

.358

.286 ⫺.208

.120 ⫺.374 ⫺.184 ⫺.023 1.000

12. Parent’s education . . . . . . . . . . . . . . . .

.271

.347

.276

.364

.240 ⫺.213

.092 ⫺.229 ⫺.141

.060

.425 1.000

13. Expenditures on schools/$1,000 . . .

.064

.133

.103

.125

.706 ⫺.263

.217 ⫺.133 ⫺.118

.058

.182

.251 1.000

14. College tuition/$1,000 . . . . . . . . . . . . .

.004 ⫺.032 ⫺.023 ⫺.010

.351 ⫺.091 ⫺.302 ⫺.068 ⫺.433 ⫺.046

.066

.042 .317 1.000

15. Grants for college . . . . . . . . . . . . . . . .

.024

.083

.112

.103

.255 ⫺.118

.302 ⫺.048

.044

.241

.118

.049 .410 .207 1.000

16. Segregation

.042

.084

.103

.102

.243 ⫺.018

.312 ⫺.017 ⫺.206

.186

.187

.114 .401 .363 .213 1.000

Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.832

.427

.182 12.906 36.728 11.460 4.971

.155

.402

.062 10.759 11.431 2.934 1.914 .075 .070

SD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.374

.494

.386 2.121 4.526 7.744 6.170

.362

.018

.011

.657 2.661 .780 .772 .079 .033

Source.—See data description in the appendix.

Note.—Means and SDs are based on the 3,504 cases in the sample for models predicting high school graduation. The sample for college outcomes is 3,240 cases

so the means differ slightly.

TABLE A2

Predictors of School Outcomes for the Full Sample

Probit Partial Derivative

Variable

Gini coefficient . . . . . . . . . . . . . . . .

State mean income/$1,000 . . . .

State %black . . . . . . . . . . . . . . . . . . .

State %Hispanic . . . . . . . . . . . . . . .

State unemployment rate . . . . .

Returns to schooling . . . . . . . . . .

Log parental income . . . . . . . . . .

Child is black . . . . . . . . . . . . . . . . . .

Parent’s years of schooling . . .

Segregation between school

districts . . . . . . . . . . . . . . . . . . . . . .

Per pupil expenditures on

schools/$1,000 . . . . . . . . . . . . . . .

College tuition/$1,000 . . . . . . . . .

Grants for college tuition . . . . .

OLS Coefficient

High School

Graduate

Enrolled in

College

Graduated

College

Years of

Schooling

⫺.375

(⫺.430)

⫺.009

(1.600)

.003

(1.470)

⫺.002

(⫺.830)

⫺.005

(⫺1.770)

1.553

(.940)

.087

(6.230)

⫺.011

(⫺.490)

.025

(7.610)

.193

(.130)

⫺.012

(⫺1.360)

.004

(1.470)

⫺.005

(⫺1.310)

.000

(.060)

2.998

(1.040)

.208

(8.390)

.066

(1.640)

.056

(7.810)

⫺2.404

(⫺1.920)

⫺.015

(⫺2.300)

.002

(1.280)

⫺.001

(⫺.280)

⫺.003

(⫺.720)

1.954

(1.010)

.099

(5.920)

⫺.047

(⫺1.980)

.034

(7.260)

⫺2.709

(⫺.483)

⫺.045

(⫺1.473)

.021

(1.774)

⫺.014

(⫺1.023)

⫺.012

(⫺.614)

9.973

(.866)

.771

(10.392)

.000

(.003)

.202

(10.350)

.267

(.680)

.484

(.790)

.572

(1.390)

1.841

(.733)

.025

(.870)

⫺.026

(⫺1.140)

⫺.134

(⫺.890)

.068

(1.400)

⫺.134

(⫺3.910)

.059

(.260)

.054

(1.770)

⫺.072

(⫺3.300)

.307

(2.280)

.265

(1.532)

⫺.392

(⫺3.069)

.331

(.433)

Note.—All models also control region and year dummy variables, as described in the text.

27

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

TABLE A3

Predictors of School Outcomes for High-Income Sample

Probit Partial Derivative

Variable

Gini coefficient . . . . . . . . . . . . . . . .

State mean income/$1,000 . . . .

State %black . . . . . . . . . . . . . . . . . . .

State %Hispanic . . . . . . . . . . . . . . .

State unemployment rate . . . . .

Returns to schooling . . . . . . . . . .

Log parental income . . . . . . . . . .

Child is black . . . . . . . . . . . . . . . . . .

Parent’s years of schooling . . .

Segregation between school

districts . . . . . . . . . . . . . . . . . . . . . .

Per pupil expenditures on

schools/$1,000 . . . . . . . . . . . . . . .

College tuition/$1,000 . . . . . . . . .

Grants for college tuition . . . . .

OLS Coefficient

High School

Graduate

Enrolled in

College

Graduated

College

Years of

Schooling

⫺.247

(⫺.260)

⫺.013

(⫺2.470)

.002

(1.100)

.001

(.320)

⫺.001

(⫺.320)

.430

(.250)

.054

(2.010)

.029

(1.580)

.017

(5.280)

.908

(.420)

⫺.005

(⫺.410)

⫺.002

(⫺.570)

⫺.008

(⫺1.820)

.003

(.440)

5.064

(1.330)

.268

(5.510)

⫺.061

(⫺.960)

.064

(6.760)

⫺1.049

(⫺.540)

⫺.006

(⫺.550)

⫺.002

(⫺.670)

⫺.003

(⫺.760)

⫺.004

(⫺.750)

3.028

(1.000)

.157

(4.950)

⫺.132

(⫺2.640)

.055

(7.750)

3.545

(.420)

⫺.016

(⫺.354)

⫺.005

(⫺.312)

⫺.018

(⫺1.095)

.004

(.163)

7.167

(.436)

1.051

(6.379)

⫺.456

(⫺2.504)

.219

(8.790)

.743

(1.920)

.767

(.940)

.745

(1.130)

3.888

(1.086)

.049

(1.820)

⫺.024

(⫺1.070)

⫺.003

(⫺.020)

⫺.006

(⫺.100)

⫺.135

(⫺3.020)

.646

(2.240)

.014

(.300)

⫺.066

(⫺1.930)

.859

(4.160)

.066

(.289)

⫺.251

(⫺1.445)

2.532

(2.230)

Note.—All models also control region and year dummy variables, as described in the text.

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

TABLE A4

Predictors of School Outcomes for Low-Income Sample

Probit Partial Derivative

Variable

Gini coefficient . . . . . . . . . . . . . . . .

State mean income/$1,000 . . . .

State %black . . . . . . . . . . . . . . . . . . .

State %Hispanic . . . . . . . . . . . . . . .

State unemployment rate . . . . .

Returns to schooling . . . . . . . . . .

Log parental income . . . . . . . . . .

Child is black . . . . . . . . . . . . . . . . . .

Parent’s years of schooling . . .

Segregation between school

districts . . . . . . . . . . . . . . . . . . . . . .

Per pupil expenditures on

schools/$1,000 . . . . . . . . . . . . . . .

College tuition/$1,000 . . . . . . . . .

Grants for college tuition . . . . .

OLS Coefficient

High School

Graduate

Enrolled in

College

Graduated

College

Years of

Schooling

⫺1.145

(⫺.750)

.007

(.600)

.002

(.650)

⫺.008

(⫺1.640)

⫺.014

(⫺2.050)

5.518

(1.570)

.053

(1.600)

⫺.032

(⫺.810)

.041

(6.310)

⫺1.876

(⫺1.260)

⫺.021

(⫺2.210)

.008

(2.220)

.000

(.100)

⫺.004

(⫺.710)

⫺.740

(⫺.250)

.016

(.490)

.093

(2.500)

.039

(5.330)

⫺2.156

(⫺3.150)

⫺.014

(⫺3.270)

.002

(1.430)

⫺.002

(⫺.960)

⫺.002

(⫺.850)

.710

(.490)

.007

(.470)

⫺.002

(⫺.140)

.009

(3.090)

⫺14.395

(⫺2.110)

⫺.070

(⫺1.710)

.033

(2.088)

⫺.016

(⫺.848)

⫺.036

(⫺1.402)

18.820

(1.336)

.093

(.660)

.205

(1.463)

.181

(7.001)

⫺.778

(⫺.880)

.207

(.270)

.007

(.020)

⫺1.810

(⫺.610)

⫺.061

(⫺1.140)

⫺.000

(⫺.610)

⫺.273

(⫺.820)

.173

(2.780)

⫺.078

(⫺1.910)

⫺.606

(⫺2.300)

.067

(2.530)

⫺.053

(⫺2.720)

⫺.246

(⫺1.840)

.445

(1.564)

⫺.548

(⫺2.928)

⫺3.170

(⫺2.854)

Note.—All models also control region and year dummy variables, as described in the text.

REFERENCES

Alesina, Alberto, and Dani Rodrik. 1994. “Distributive Politics and Economic Growth.”

Quarterly Journal of Economics 109:465–90.

Atkinson, Anthony B. 1970. “On the Measurement of Inequality.” Journal of Economic

Theory 2:244–63.

———. 1983. The Economics of Inequality. Oxford: Oxford University Press.

Betts, Julian, and L. McFarland. 1995. “Safe Ports in a Storm: The Impact of Labor

Market Conditions on Community College Enrollment.” Journal of Human

Resources 30 (4): 741–65.

Card, David, and Abigail Payne. 1998. “School Finance Reform, the Distribution of

29

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

School Spending, and the Distribution of SAT Scores.” Working Paper no. 6766.

National Bureau of Economic Research, Cambridge, Mass.

Ceci, Stephen J. 1991. “How Much Does Schooling Influence General Intelligence and

Its Cognitive Components? A Reassessment of the Evidence.” Developmental

Psychology 27 (5): 703–22.

Crane, Jonathon. 1991. “The Epidemic Theory of Ghettos and Neighborhood Effects

on Dropping Out and Teenage Childbearing.” American Journal of Sociology 96 (5):

1226–59.

Davis, James A. 1959. “A Formal Interpretation of the Theory of Relative Deprivation.”

Sociometry 22:280–96.

———. 1966. “The Campus as a Frog Pond: An Application of the Theory of Relative

Deprivation to Career Decisions of Young Men.” American Sociological Review 52:

511–18.

Durlauf, Steven. 1996. “A Theory of Persistent Income Inequality.” Journal of

Economic Growth 1:75–93.

Ellemers, Naomi, Russell Spears, and Bertjan Doosje, eds. 1997. Social Identity

Context, Commitment, Content. New York: Blackwell.

Elliott, Marta. 1998. “School Finance and Opportunities to Learn: Does Money Well

Spent Enhance Students’ Achievement?” Sociology of Education 71 (3): 223–45.

Ellwood, David A., and Thomas J. Kane. 1999. “Who Is Getting A College Education?

Family Background and the Growing Gaps in Enrollment.” Unpublished

manuscript. Harvard University, Kennedy School of Government.

Evans, William, William Oates, and R. M. Schwab. 1992. “Measuring Peer Group

Effects: A Study of Teenage Behavior.” Journal of Political Economy 100:966–91.

Firebaugh, Glenn. 1999. “Empirics of World Income Inequality.” American Journal of

Sociology 104 (6): 1597–1630.

Forbes, Kristin. 2000.“A Reassessment of the Relationship between Inequality and

Growth.” American Economic Review 90 (4): 869–87.

Grubb, N. W. 1988. “Vocationalizing Higher Education: The Causes of Enrollment

and Completion in Public Two-Year Colleges 1970–1980.” Economics of Education

Review 7 (3): 301–19.

Gurr, Ted R. 1970. Why Men Rebel. Princeton, N.J.: Princeton University Press.

Gustafsson, Bjorn, and Mats Johansson. 1999. “In Search of Smoking Guns: What

Makes Income Inequality Vary over Time in Different Countries?” American

Sociological Review 64 (4): 585–605.

Hanushek, Eric A. 1996. “School Resources and Student Performance.” Pp. 585–605

in Does Money Matter? edited by Gary Burtless. Washington, D.C.: Brookings

Institution Press.

Hedges, Larry V., Richard Laine, and Rob Greenwald. 1992. “Does Money Matter?

Meta-Analysis of Studies of the Effects of Differential School Inputs on Student

Outcomes.” Educational Researcher 23 (3): 5–14.

James, Franklin. 1986. “A New Generalized ‘Exposure-Based’ Segregation Index:

Demonstration in Denver and Houston.” Sociological Methods and Research 14 (3):

301–16.

Jargowsky, Paul A. 1996. “Take the Money and Run: Economic Segregation in U.S.

Metropolitan Areas.” American Sociological Review 61:984–98.

Jencks, Christopher, and Susan E. Mayer. 1990. “The Social Consequences of Growing

Up in a Poor Neighborhood.” Pp. 111–86 in Inner-City Poverty in the United States,

edited by Laurence Lynn and Michael McGeary. Washington, D.C.: National

Academy Press.

Juhn, Chinhui, Kevin Murphy, and Brooks Pierce. 1993. “Wage Inequality and the

Rise in Returns to Skill.” Journal of Political Economy 101 (3): 410–42.

Kane, Thomas. 1994. “College Entry by Blacks since 1970: The Role of College Costs,

30

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Educational Attainment

Family Background and the Returns to Education.” Journal of Political Economy

102 (5): 878–911.

———. 1999. The Price of Admission: Rethinking How Americans Pay for College.

Washington, D.C.: Brookings Institution Press.

Karoly, Lynn. 1993. “The Trend in Inequality among Families, Individuals, and

Workers in the United States: A Twenty-Five Year Perspective.” Pp. 19–97 in Uneven

Tides: Rising Inequality in America, edited by Sheldon Danziger and Peter

Gottschalk. New York: Russell Sage Foundation Press.

Lichter, Daniel, and David Eggebeen. 1993. “Rich Kids, Poor Kids: Changing Income

Inequality among American Children.” Social Forces 71 (3): 301–76.

Manski, Charles F. 1993. “Identification of Endogenous Social Effects: The Reflection

Problem.” Review of Economic Studies 60 (3): 531–42.

Manski, Charles, and David Wise. 1983. College Choice in America. Cambridge, Mass.:

Harvard University Press.

Massey, Douglas, and Mitchell Eggers. 1990. “The Ecology of Inequality: Minorities

and the Concentration of Poverty.” American Journal of Sociology 95:1153–88.

Mayer, Susan E., 1991. “How Much Does A High School’s Racial and Socioeconomic

Mix Affect Graduation and Teen Fertility Rates?” Pp. 79–102 in The Urban

Underclass, edited by Christopher Jencks and Paul Peterson, Washington, D.C.:

Brookings Institution Press.

Mayer, Susan E., and David Knutson. 1999. “Does the Timing of Schooling Effect

How Much Children Learn?” Pp. 79–102 in Earning and Learning: How Schools

Matter, edited by Susan E. Mayer and Paul E. Peterson. Washington, D.C.:

Brookings Institution Press.

McLoyd, Vonnie. 1990. “The Impact of Economic Hardship on Black Families and

Children: Psychological Distress, Parenting and Socioemotional Development.”

Child Development 61:311–46.

Merton, Robert, and Alice Kitt. 1950. “Contributions to the Theory of Reference Group

Behavior.” Pp. 40–105 in Studies in the Scope and Method of “The American

Soldier,” edited by Robert Merton and Paul Lazarsfeld. Glencoe, Ill.: Free Press.

Messner, Steven, and Kenneth Tardiff. 1986. “Economic Inequality and the Levels of

Homicide: An Analysis of Urban Neighborhoods.” Criminology 24:297–316.

Morris, Martina, Annette Bernhardt, and Mark Handcock. 1994. “Economic

Inequality: New Methods for New Trends.” American Sociological Review 59 (2):

205–19.

Morris, Martina, and Bruce Western. 1999. “Inequality in Earnings at the Close of

the Twentieth Century.” American Sociological Review 25:623–57.

Murphy, Kevin, and Finish Welch. 1992. “The Structure of Wages.” Quarterly Journal

of Economics 107:215–36.

Murray, Shelia, William Evans, and Robert Schwab. 1998. “Education-Finance Reform

and the Distribution of Education Resources.” American Economic Review 88 (4):

789–812.

Nettler, Gwynn. 1994. Explaining Crime, 3d ed. New York: McGraw-Hill.

Perotti, Roberto. 1996. “Growth, Income Distribution, and Democracy: What the Data

Say.” Journal of Economic Growth 1:149–87.

Piven, Francis Fox, and Richard Cloward. 1993. Regulating the Poor: The Functions

of Public Welfare. New York: Vintage Books.

Rosenfeld, Richard. 1986. “Urban Crime Rates: Effects of Inequality, Welfare

Dependency, Region and Race.” Pp. 116–30 in The Social Ecology of Crime, edited

by James Byrne and Robert Sampson. New York: Springer-Verlag.

Runciman, W. G. 1966. Relative Deprivation and Social Justice: A Study of Attitudes

to Social Inequality in Twentieth Century England. Berkeley: University of

California Press.

Smith, Heather, Russell Spears, and Intse Hamstra. 1999. “Social Identity and the

31

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

American Journal of Sociology

Context of Relative Deprivation.” Pp. 205–29 in Social Identity: Context,

Commitment and Content, edited by Naomi Ellemers, Russell Spears, and Bertjan

Doosje. Oxford: Blackwell Publishers.

Suls, Jerry, ed. 1991. Social Comparison: Contemporary Theory and Research. Hillsdale,

N.J.: Lawrence Erlbaum.

Tienda, Marta. 1991. “Poor People and Poor Places: Deciphering Neighborhood Effects

on Poverty Outcomes.” Pp. 224–62 in Macro-Micro Linkages in Sociology, edited

by J. Haber. Newberry, Calif.: Sage.

U. S. Bureau of the Census. 1998. Statistical Abstract of the United States. Washington,

D.C.: Government Printing Office.

U.S. Department of Education, National Center for Educational Statistics. 1998. Digest

of Educational Statistics. Washington, D.C.: Government Printing Office.

U.S. Department of Health and Human Services. 1998. Trends in the Well-Being of

America’s Children and Youth. Washington, D.C.: Government Printing Office.

Welch, Finis. 1999. “In Defense of Inequality.” American Economic Review 89 (2):

1–17.

White, Michael. 1987. American Neighborhoods and Residential Differentiation. New

York: Russell Sage Foundation.

Williams, Robin Jr. 1975. “Relative Deprivation.” Pp. 355–78 in The Idea of Social

Structure: Papers in Honor of Robert K. Merton, edited by Lewis A. Coser. New

York: Harcourt Brace Jovanovich.

Wilson, William Julius. 1987. The Truly Disadvantaged. Chicago: University of Chicago

Press.

Winship, Christopher, and Sanders Korenman. 1999. “Economic Success and the

Evolution of Schooling.” Pp. 49–78 in Earning and Learning: How Schools Matter,

edited by Susan E. Mayer and Paul E. Peterson. Washington, D.C.: Brookings

Institution Press.

32

This content downloaded from

130.113.111.210 on Mon, 27 Jan 2020 16:35:44 UTC

All use subject to https://about.jstor.org/terms

Final Report (15%)

• Goal of this segment: Write a report on your findings

I Individual report!

I Main focus: Question/Topic you focused on in the research segment

• What should always be covered:

I Motivation: Why did you pick your topic? Why is it interesting? (Refer

back to magazine article and empirical evidence.)

I

I

Analysis of Existing Research: What is the state of research on the topic?

Description of research methods, datasets etc.

Discussion: What is the main takeaway from the literature? Compare and

contrast findings. What is your take on the topic? What is missing in the

existing literature?

• Deadline for the first report: Feb 25, 2018 at 6pm

Expectations for Final Report

• Intended to show the individual student’s grasp of the subject

æ Independent work, copying will be penalized

• Summary of what was presented and worked on throughout the module

• Grade will depend on clarity of text and quality of content

• Length: 3-4 pages. Use of tables and figures is encouraged, but they should

be placed in the appendix (which won’t count toward the 3-4 pages of main

text)

The Manchester School Vol 83 No. 4

doi: 10.1111/manc.12064

423–450

July 2015

DIRECT AND INDIRECT INFLUENCES OF PARENTAL

BACKGROUND ON CHILDREN’S EARNINGS: A

COMPARISON ACROSS COUNTRIES AND GENDERS*

by

MICHELE RAITANO

Department of Economics and Law, Sapienza University of Rome

and

FRANCESCO VONA

OFCE Sciences-Po and Skema Business School, Nice

The association between parental occupation and children’s earnings in

eight EU countries is compared using the European Union Survey on

Income and Living Conditions (EU-SILC) data set, analysing: (i)

residual background correlations (RBCs) on earnings, controlling for

children’s education and occupation, and (ii) patterns by gender, controlling for selection into employment. Findings on cross-country differences

confirm well-known differences in intergenerational income inequality.

RBCs are statistically significant irrespective of gender in the UK, Spain

and Italy, for men in France and Ireland, for women in Denmark and not

significant in Germany and Finland. Not controlling for selection delivers downward biased estimates of RBCs, highlighting the effect of family

background on employability.

1

INTRODUCTION

The analysis of intergenerational inequality and social mobility has

attracted increased attention by applied economists in last few decades. The

recent availability of new data for several countries has allowed comparisons of country differences in terms of intergenerational income inequality,

usually measured by the intergenerational elasticity β that is estimated

regressing children’s log income (when adult) on parental income

(Bjorklund and Jäntti, 2009; Blanden, 2013). For European countries,

empirical studies generally converge on the following country taxonomy:1

Nordic countries are the most mobile, followed by Germany and by other

central European countries with France in the worst position of these,

whereas Anglo-Saxon and Southern European countries are the least

mobile of all European countries.

* Manuscript received 9.11.12; final version received 4.4.14.

1

For a review of the several studies computing intergenerational income inequality in different

countries, see Solon (2002), Corak (2006, 2013), d’Addio (2007), Bjorklund and Jäntti

(2009) and Blanden (2013).

© 2014 The University of Manchester and John Wiley & Sons Ltd

423

424

The Manchester School

While the β elasticities allow the quantification of the size of

intergenerational income inequality, they do not provide insights into which

transmission mechanisms explain the association between parental and children’s earnings. This issue is particularly relevant for cross-country analyses.

By way of example, the transmission of human capital is likely to be relatively

more important in countries with a heterogeneous schooling system such as

the UK; in turn, the influence of family background on employability and

labour market outcomes is usually perceived as more important in Southern

countries such as Italy and Spain. The importance of these mechanisms is also

likely to vary across genders in that female participation heavily depends on

welfare policies and on optimal household decisions (Gornick and Meyers,

2003; Raaum et al., 2008). Except for a few studies,2 estimates of

intergenerational income elasticities have usually been carried out for men

only to get rid of participation constraints, which are particularly cumbersome to address for women.

This paper contributes to the active strand of literature on cross-country

differences in intergenerational inequality by investigating the role of different

transmission mechanisms and of their heterogeneous influence across genders.

More specifically, we compare eight EU countries (Germany, France, Spain,

Italy, the UK, Ireland, Denmark and Finland) for which we expect a large

variation in the importance of these mechanisms, reflecting differences in

welfare policies, culture, labour markets and educational systems.

Taking advantage of the rich information contained in the 2005 module

on intergenerational inequality of the European Union Survey on Income

and Living Conditions (EU-SILC), we can distinguish the influence of family

background at various stages of the children’s life cycle. To fully exploit the

possibility offered by the EU-SILC, we develop a conceptual framework that

stresses the role of human capital accumulation on children’s prospects, as

postulated by the standard approach (Becker and Tomes, 1979, 1986), but

also focuses on labour market outcomes, such as participation, employability

and occupational sorting. This approach allows us to derive a direct ‘residual’

correlation between family background and children’s earnings (RBC henceforth), obtained once controlling for the indirect influence of background on

education and occupational sorting.

The novel result of this paper descends in a straightforward manner

from the simple distinction between direct and indirect transmission mechanisms. In particular, we show that the differences between ‘mobile’ and

‘immobile’ countries are mainly captured by direct RBC and less by the

influence of family background on educational and occupational attainments. Moreover, we show that RBC is underestimated when the decision to

2

Among these, see Chadwick and Solon (2002) and Hirvonen (2008), respectively, referring to the

USA and Sweden, and Jäntti et al. (2006) and Raaum et al. (2008), who compare Nordic

countries, the UK and the USA.

© 2014 The University of Manchester and John Wiley & Sons Ltd

Direct and Indirect Influences of Parental Background

425

participate in the labour force is not explicitly accounted for. Especially for

women, participation is influenced by family background, and modelling it

using a Heckman two-step procedure could help to obtain unbiased RBCs.

The EU-SILC is the only data set that consents to analyse the

intergenerational inequality in EU countries by means of a homogenous

survey. As a main drawback, however, the EU-SILC does not contain information on parental income. We hence use parental occupation as a proxy of

family background. In making this choice, we follow rich sociological literature pointing to parental occupation as the best predictor of children’s success.

The main claim of this literature is that parental occupation is not only a good

proxy for unobservable aspects of human capital (Willis, 1986), but it also

captures the individual’s position in the social scale, the capacity to influence

economic decisions or becoming part of certain social networks (Erikson and

Goldthorpe, 1992; Ganzeboom and Treiman, 1996). Furthermore, as noted by

Ermisch et al. (2006), parental occupation is also likely to be an adequate

measure of people’s permanent socio-economic status, as the position of

individuals in the occupational hierarchy is relatively stable over time.

The rest of the paper is organized as follows. Section 2 introduces our

conceptual framework. Section 3 illustrates in detail the data set and our

proxy of family background. Section 4 shows preliminary evidence that

indirect transmission mechanisms cannot account for cross-country differences in the effects of parental background on child earnings. Sections 5 and

6 directly address the issues of cross-country and cross-gender differences in

RBCs respectively. Section 7 briefly concludes.

2

CONCEPTUAL FRAMEWORK

Traditional economists’ view on intergenerational inequality focuses on the

key role played by family background in the accumulation of human capital.

In the literature, the association between parental characteristics and children’s human …

Purchase answer to see full

attachment

#### Why Choose Us

- 100% non-plagiarized Papers
- 24/7 /365 Service Available
- Affordable Prices
- Any Paper, Urgency, and Subject
- Will complete your papers in 6 hours
- On-time Delivery
- Money-back and Privacy guarantees
- Unlimited Amendments upon request
- Satisfaction guarantee

#### How it Works

- Click on the “Place Order” tab at the top menu or “Order Now” icon at the bottom and a new page will appear with an order form to be filled.
- Fill in your paper’s requirements in the "
**PAPER DETAILS**" section. - Fill in your paper’s academic level, deadline, and the required number of pages from the drop-down menus.
- Click “
**CREATE ACCOUNT & SIGN IN**” to enter your registration details and get an account with us for record-keeping and then, click on “PROCEED TO CHECKOUT” at the bottom of the page. - From there, the payment sections will show, follow the guided payment process and your order will be available for our writing team to work on it.