Intergenerational transmission of educational attainment: The role of household assets

Intergenerational transmission of educational attainment: The role of household assets

Economics of Education Review 33 (2013) 112–123 Contents lists available at SciVerse ScienceDirect Economics of Education Review journal homepage: w...

235KB Sizes 0 Downloads 0 Views

Economics of Education Review 33 (2013) 112–123

Contents lists available at SciVerse ScienceDirect

Economics of Education Review journal homepage: www.elsevier.com/locate/econedurev

Intergenerational transmission of educational attainment: The role of household assets Jin Huang * School of Social Work, Saint Louis University, USA

A R T I C L E I N F O

A B S T R A C T

Article history: Received 3 May 2012 Received in revised form 21 September 2012 Accepted 27 September 2012

High intergenerational persistence of educational attainment is an indicator of educational inequality and a barrier to equal opportunities in the labor market and beyond. This study uses data from the Panel Study of Income Dynamics to generate a sample of two cohorts of children (’84 and ’94 cohorts), and it examines whether intergenerational transmission of educational attainment varies by household economic resources, especially household assets. Results show that, among male children in the ’94 cohort, household assets increase the strength of the association between parents’ and children’s years of schooling. Also, household assets are found to interact with parental education to affect educational attainment, as measured by college completion, among female children in the ’94 cohort. Research and policy implications are discussed. ß 2012 Elsevier Ltd. All rights reserved.

JEL classification: I25 Keywords: Economic impact Human capital Educational economics

Education is a primary determinant of long-term economic success and a key mechanism of social mobility. It is capable of lifting children out of disadvantage and improving their chances for success as adults (Baum & Ma, 2007; Haveman & Wolfe, 1995; Kane, 2004). Children’s education outcomes are affected by aspects of family background, such as parental education and household economic resources (Bjo¨rklund & Salvanes, 2010; Haveman & Wolfe, 1995). It is well known that children whose parents have more schooling tend to have better educational outcomes. In addition, there is a strong intergenerational association between the level of parental schooling and the level of the child’s schooling (Hertz et al., 2007). The intergenerational transmission of educational attainment from parents to children is considered an indicator of educational mobility; high intergenerational persistence of educational attainment serves as a barrier to opportunities in the labor market and beyond (Black & Devereux, 2010).

* Correspondence address: 3550 Lindell Blvd., St. Louis, MO 63103, USA. Tel.: +1 314 977 2750; fax: +1 314 977 2731. E-mail address: [email protected] 0272-7757/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.econedurev.2012.09.013

The intergenerational transmission of educational attainment has been investigated extensively in the theoretical and empirical literature. For example, Hertz et al. (2007) estimate the correlation between the level of parents’ schooling and the level of children’s schooling in a sample of 42 countries. The correlation coefficients are about .60 in South America, about .40 in Western Europe, and .46 in the United States. The lowest coefficients are from Nordic countries. Vast differences across countries in the association between parents’ and children’s educational attainment suggest that this association may vary by geography and institutional settings (e.g., the educational system). The differences also imply that intergenerational transmission of educational attainment is not solely a process of genetic inheritance; educational mobility is affected by external conditions (e.g., return on educational investment, family investment in child development, and public policy) other than biological factors. The association between parents’ and children’s educational attainment may vary by other family background factors (e.g., economic resources), as well. However, few studies have examined heterogeneity and patterns in the intergenerational transmission of educational attainment. Doing so could enable an important advance in identifying the

J. Huang / Economics of Education Review 33 (2013) 112–123

113

1. Background

and college graduation. The matrices can show parents’ educational attainment (e.g., college completion) on one axis and children’s attainment on the other (Bauer & Riphahn, 2004). Empirical studies also have examined whether the association between parents’ and children’s schooling changes over time (Hertz et al., 2007), as the time trend may represent another type of heterogeneity in intergenerational transmission. The findings are fairly mixed, however. For example, Hertz et al. (2007) analyze data from the 1967, 1972, and 1977 birth cohorts, suggesting that, in the United States, the coefficients of the correlation between parents’ and children’s schooling have increased slightly over time but the regression coefficients have fallen. The strong association of parents’ and children’s level of schooling does not necessarily reflect a causal relationship between parents’ and children’s education, however. The literature considers two main channels for the transmission of educational attainment between generations. On the one hand, abilities may be transmitted from parents to their biological children by the inheritance of genes (nature). Genetic factors are found to explain about half of the variation in adult IQ (Sacerdote, 2011), and some portion of the association between parents’ and children’s schooling may be due to shared genetics. On the other hand, the transmission may work through parents’ investment in child development (nurture). Parents with higher education may value education more; compared with parents who have less education, they may spend more time and economic resources to improve their children’s educational achievement. The research design and the PSID data used in this study cannot identify the ‘‘nurture’’ effects or separate them from genetic effects; that is, this study does not differentiate between genetic and behavioral determinants of the intergenerational transmission of educational attainment. Rather, it focuses on patterns in the association between parents’ and children’s schooling and on variations in that pattern across levels of household economic resources.

1.1. Intergenerational transmission of educational attainment as a measure of intergenerational mobility

1.2. Do assets increase or decrease intergenerational transmission of educational attainment?

In the United States, rising economic inequality over the last few decades has been closely related to declines in social mobility across generations (Anger, 2011). The literature on intergenerational social mobility focuses predominantly on income and education. Income mobility is commonly measured by assessing the correlations of earnings across generations and by examining intergenerational earnings elasticity. In the United States, the estimated earnings elasticity is about .50 to .60 (Black & Devereux, 2010). Similarly, intergenerational transmission of educational attainment can be measured by the correlation between the parent’s and the child’s level of schooling as well as by the use of regression to estimate the effect of parental education on the child’s educational attainment. In addition, transition matrices provide a strategy for measuring educational attainment with such discrete variables as high school graduation, college entry,

Household economic resources—namely, assets and income—play important roles in children’s educational outcomes and intergenerational transmission of educational attainment. As reviewed in Elliott et al. (2011), studies generally find that household assets are positively associated with children’s academic achievement and postsecondary education. Household assets and parental education may affect children’s educational outcomes through similar channels. One possibility is that the positive association between household assets and children’s educational attainment could be due to genetic endowments that affect both parents’ ability to accumulate assets and children’s educational attainment. Parents with higher ability accumulate more assets and have children who are able to obtain more schooling. A second possibility is that households with assets have financial resources to invest in children’s

causal mechanisms that underlie intergenerational trends in educational attainment (Bauer & Riphahn, 2004). Household assets also are an important determinant of children’s developmental outcomes. Extensive research shows that household assets are positively associated with children’s educational attainment (Conley, 2001; Elliott, Destin, & Friedline, 2011; Kim & Sherraden, 2011; Nam & Huang, 2009, 2011; Orr, 2003). Although children’s educational outcomes are affected by both parental education and assets, the two might interact in their relationships with those outcomes. In other words, intergenerational transmission of educational attainment from parents to children may vary by level of household economic resources (e.g., assets and income). Few studies examine the association between household assets and intergenerational transmission of educational attainment. One hypothesis is that constrained household assets may increase the likelihood of intergenerational transmission of educational attainment, because a lack of financial resources could limit investments in child development by parents, even highly educated parents (Kane, 2004). However, household assets may reduce intergenerational persistence of educational attainment if household economic resources have greater marginal effects on child development among disadvantaged children than among their nondisadvantaged counterparts (Duncan, Ziol-Guest, & Kalil, 2010; Huang, 2011; Huang, Guo, Kim, & Sherraden, 2010). This study examines the role of household economic resources (especially household assets) in the intergenerational transmission of educational attainment and aims to understand the potential variations by gender in the effects of such resources. It uses data from the Panel Study of Income Dynamics (PSID) to generate a sample of two cohorts of children. Separate analyses are conducted for females and males, because previous research finds that children’s gender is associated with differences in educational attainment (e.g., Bailey & Dynarski, 2011).

114

J. Huang / Economics of Education Review 33 (2013) 112–123

human capital (e.g., academic and cognitive development) and to improve children’s life chances (Conley, 2001; Glick & Sahn, 2009; Mayer, 1997; Shapiro, 2004). Compared with counterparts from wealthy households, low-wealth children often live in physical environments that offer less stimulation and fewer resources (e.g., educational toys and books) for learning. In contrast, children from wealthy households are more likely to receive high-quality education and to have resources to pay for postsecondary education (Nam & Huang, 2009). Furthermore, as asset holding can shape positive psychological well-being (Schreiner & Sherraden, 2007; Sherraden, 1991), children from households owning assets may be more likely than other children to develop future orientation, self-esteem, and self-efficacy, all of which are important to achieving academic success. It is not unexpected that household assets and parental education may have a confounded association with children’s educational attainment, as assets and education are highly correlated and may affect educational outcomes in similar ways. Beyond this possibility, an interesting question is whether the association between parents’ and children’s educational attainment differs by the level of household assets, or whether household assets and parental education interact in influencing children’s educational outcomes. The answer has important implications for policies to promote educational mobility. If household assets strengthen the association between parents’ and children’s educational attainment, policy should address the interaction of low financial resources and low parental education to prevent children in such households from inheriting their parents’ disadvantage. However, if household assets lessen the transmission of low educational attainment from parents to children, the asset-building approach may have double effects, easing the limitations imposed on educational attainment by both liquidity constraints and low parental education. Nonetheless, the direction of the interaction between household assets and parental education, if such an interaction exists, is unclear. One hypothesis is that the strength of the association between parents’ and children’s educational attainment increases along with the level of household assets. Such an increase would indicate that there is a positive interaction between household assets and parental education. Thus, children of higher educated parents have the ability to obtain more schooling, yet a lack of financial resources could limit their parents’ investment in child development (Kane, 2004) and consequently could constrain the opportunity of these children to reach their full educational potential (Gaviria, 2002). An alternative hypothesis is that household assets can lessen the strength of the association between parents’ and children’s educational attainment. Previous studies suggest that household economic resources have greater marginal effects on child development for disadvantaged children than for others (Duncan et al., 2010; Huang, 2011; Huang et al., 2010). In other words, children whose parents have lower education benefit more from the growth of household assets. In this hypothesis, household assets interact negatively with parental education. Similar to the second hypothesis, the third one posits that the strength of the

association between parents’ and children’s educational attainment diminishes as the level of household assets grows because household assets may show a greater association with parental education as they increase. This changing association between parental education and household assets would also result in a negative interaction effect on a child’s educational attainment. The current study tests these hypotheses using the PSID data, and investigates the influence of household economic resources on the intergenerational transmission of educational attainment. 2. Methods 2.1. Data and sample The PSID is a longitudinal survey, and the resulting data are publicly available from the Institute for Social Research at the University of Michigan. The PSID collects demographic information and socioeconomic characteristics (e.g., educational attainment and household income) from a nationally representative sample of individuals and their families. It collected data annually from 1968 to 1997 and biennially after 1997. The PSID measured household assets in 1984, 1989, 1994, and 1999. Since 1999, the assets data have been collected biennially. Taking advantage of the longitudinal data provided by the PSID, the study creates two cohorts: (1) Black and White children who were 13–20 years old and living in the parental household in 1984 (the ’84 cohort) and (2) Black and White children who were 13–20 years old and living in the parental household in 1994 (the ’94 cohort). The age range (8 years) is chosen to ensure an adequate sample size. Ideally, more cohorts (e.g., more than three cohorts) should be used to better examine the trend of the association between household economic resources and intergenerational transmission of educational attainment. However, the study cannot use younger cohorts because data on educational attainment are not available for cohorts younger than the 1994 group. The study cannot use cohorts older than the 1984 group because the PSID started collecting the information on household assets in 1984. The final sample is composed of 2466 adolescents and young adults. Of these, 1335 are from the ’84 cohort and 1131 are from the ’94 cohort. 2.2. Measures 2.2.1. Dependent variable The dependent variable captures the educational attainment of individuals in the two cohorts. It is measured in 1996 for the ’84 cohort and in 2007 for the ’94 cohort. The study uses three indicators of educational attainment: a continuous measure of the number of years of schooling (possible values range from 1 to 17), a dichotomous measure of college entry (college entry = 1, 0 otherwise), and a dichotomous measure of 4-year college completion (completion = 1, 0 otherwise). 2.2.2. Independent variables Major independent variables are mothers’ educational attainment and household economic resources, which are

J. Huang / Economics of Education Review 33 (2013) 112–123

drawn from the PSID household survey. A mother’s educational attainment is also measured by three indicators: the number of years of schooling, whether she entered college (college entry = 1, 0 otherwise), and whether she completed a degree from a 4-year college (college completion = 1, 0 otherwise). To maximize the size of the sample, mothers’ rather than fathers’ educational attainment is used as the independent variable because the response rate on fathers’ educational attainment is lower. Previous studies show that estimates of the association between parents’ and children’s educational attainment are only slightly affected by whether parental attainment is measured as the mother’s education or the father’s (e.g., Hertz et al., 2007). In addition, the literature suggests that more educated mothers may be more likely to affect parental time allocation and parental productivity in childenhancing activities; their effect on children’s educational attainment may thus be greater than that of fathers (Black & Devereux, 2010). As measured in this study, household economic resources include assets and income. Household assets are measured in 1984 for the older cohort and in 1994 for the younger cohort. The study uses two asset measures: net worth and financial assets. Defined as the total amount of household wealth, net worth is created by the PSID and is calculated as the sum of values of all assets (including business holdings, home equity, real estate, cars, amount of money in saving and checking accounts, stocks, mutual funds, investment trust, and other assets) net of all liabilities. Financial assets exclude values of home equity, real estate, business holdings, and cars from the measure of net worth. The variable measuring household income is created by averaging the 3 years of household income (i.e., 1982–1984 for the ’84 cohort and 1992–1994 for the ’94 cohort). Using the average allows the study to account for income fluctuation (Solon, 1992). 2.2.3. Control variables The study includes two groups of control variables. The first group is composed of children’s characteristics, including age, gender (male = 1, female = 0), race (Black = 1, White = 0), and the birth order to the mother (measured as the firstborn, second, and third or above). The second group of controls is composed of householders’ and household’s characteristics measured in 1984 and 1994, respectively, for the two cohorts. Householders’ characteristics include age, gender (male = 1, female = 0), marital status (married = 1, not married = 0), and employment status (employed = 1, not employed = 0). The analysis controls for two household characteristics: household size and number of children. These are measured in 1984 in the ’84 cohort and in 1994 for the ’94 cohort. 2.3. Analysis As discussed above, intergenerational transmission of educational attainment is generally measured by the correlation of parents’ and children’s schooling, by transition matrices, and by a regression coefficient of parental education on child education (Black & Devereux, 2010). The study first reports on the correlation between

115

parents’ and children’s schooling. It then reports on transition matrices that track college entry and college completion by the levels of household economic resources. Since household economic resources (income and assets) are measured with continuous variables, they are categorized into low-, mid-, and high-level groups in the descriptive analyses. The cutoff lines for mid-level and high-level household resources are drawn at the 33rd and 67th percentiles on the distributions of the each resource variable (income and assets) to create three groups of the same size for each variable. Although the cutoff lines are somewhat arbitrary, they allow the study to explore the potential heterogeneity in intergenerational transmission of educational attainment across different levels of economic resources. This categorization results in groups that are similar to those constructed using the cutoff lines applied by Krueger (2012). He suggests using a band around the median household income: households are grouped as those with incomes that fall within 50% of the household median income, those with incomes above that band, and those with incomes below the band. If the association between parents’ and children’s educational attainment varies by a household economic resource, results from measures of intergenerational transmission should differ across the three groups for that resource. To further examine the relationship between intergenerational transmission of educational attainment and household economic resources in multivariate analyses, the study takes a strategy similar to that of Bauer and Riphahn (2004), running an ordinary least squares (OLS) regression for the continuous measure of schooling years and a logit regression for the dichotomous measures of educational attainment (college entry and 4-year college completion). The model includes a term for the interaction of mother’s education and household economic resources: Y ¼ b0 þ b1 ME þ b2 HE þ b3 ðMEHEÞ þ b4 X þ e;

(1)

where Y indicates child’s education; ME denotes mother’s education; HE is household economic resources measured by assets or income; X is a vector of control variables; and e indicates random errors. Following the convention used in the literature to address skewness and to obtain a semielasticity explanation (Conley, 2001; Nam & Huang, 2009; Orr, 2003), the logarithms of income and assets are created for regression analyses. In Eq. (1), the coefficient of the interaction terms (b3) is the parameter of interest because a statistically significant estimate of b3 indicates that the association between mothers’ and children’s educational attainment varies by household economic resources. The study runs additional analyses to check the robustness of the findings. First, the continuous measures of household economic resources in Eq. (1) are replaced by the three-level categorical measures created for descriptive analyses. Second, the measure of mothers’ schooling years is used to replace the dichotomous measures of mothers’ college entry and completion in logit regressions. Third, the study uses conditional models to analyze college

116

J. Huang / Economics of Education Review 33 (2013) 112–123

entry and completion. Only those who finished high school or entered college are included in the conditional models for college entry or college completion. All these analyses are conducted separately for males and females in each cohort. This approach allows the analyses to control both for gender differences and cohort differences. The Wald test is used to examine whether gender differences and cohort differences, if found, are statistically significant. 3. Results 3.1. Sample characteristics Sample characteristics by cohort and child’s gender are reported in Table 1. Children’s educational attainment is measured at about age 29 for both cohorts. Across the two cohorts, the mean number for children’s schooling years is around 13–14 years. More than half of children entered college, but less than 30% graduated from a 4-year college.

Women in the ’94 cohort are the exception; 39% of them graduated from a 4-year college. The mean number of mothers’ schooling years is about 1–1.5 years lower than the mean for their children’s years. The difference indicates an intergenerational improvement in educational attainment. The likelihood that mothers in the ’94 cohort entered college is 43%, higher than that for mothers in the ’84 cohort (25%). Similarly, the probability of 4-year college completion nearly doubles for mothers between the two cohorts. Overall the distribution of educational attainment in the sample is consistent with that of the general population. For instance, the descriptive statistics reported in Table 1 are similar to figures from the 2007 Current Population Survey: about 63% of 29-year-old females and 54% of 29-year-old males reported college experiences in the 2007 survey. In general, householder characteristics are similar across the four columns in Table 1, although householders in the ’94 cohort are less likely than their counterparts to be married. In addition, Table 1 reports mean and median values of household assets and income.

Table 1 Sample characteristics (N = 2466). Variables

1984 cohort

1994 cohort

Female

Male

Female

Male

(1)

(2)

(3)

(4)

13.5 53.4 27.1 28.6 13.6

13.3 49.1 25.4 28.5 12.2

14.1 64.9 39.3 28.9 16.3

13.6 56.6 25.8 29.3 16.1

33.6 24.5 41.9

34.6 23.6 41.9

39.3 36.7 24.0

45.5 32.9 21.5

Mothers’ characteristics Educational attainment Schooling years (mean) College entry (%) 4-Year college graduation (%)

12.0 24.5 8.8

12.2 25.9 10.8

13.0 43.0 20.5

13.0 41.4 18.8

Household characteristics Householder’s gender (male, %) Householder’s age (mean) Householder’s employment (yes, %) Householder’s marital status (yes, %) Household size (mean) Number of children (mean)

76.5 43.6 82.8 74.5 4.1 1.8

82.1 44.1 84.6 77.3 4.2 1.9

71.7 42.3 87.2 70.3 4.0 2.0

78.2 43.3 89.3 75.1 4.0 1.8

Household economic resources Three-year average income (mean) 33th percentile 50th percentile (median) 67th percentile Financial assets (mean) 33th percentile 50th percentile (median) 67th percentile Net worth (mean) 33th percentile 50th percentile (median) 67th percentile

38,736 21,331 32,420 43,483 144,060 4500 14,400 33,000 188,944 21,000 47,600 85,513

35,611 22,509 29,262 39,612 66,342 5900 14,400 33,000 109,400 27,637 51,500 88,000

50,089 32,415 45,901 63,066 87,792 4000 16,600 46,600 132,195 15,500 48,100 98,000

59,098 37,986 49,029 64,484 77,017 8300 24,700 54,000 125,766 30,000 56,050 127,973

N

703

632

606

525

Children’s characteristics Educational attainment Schooling years (mean) College entry (%) 4-Year college graduation (%) Age when educational attainment measured Black (%) Birth order (%) First child Second child Third child and above

J. Huang / Economics of Education Review 33 (2013) 112–123

3.2. Correlation of mothers’ and children’s years of schooling Table 2 presents estimates of correlations between mothers’ and children’s years of schooling. In addition, the estimates are broken down by cohort and child’s gender across the levels of household economic resources. First, the overall correlation between mothers’ and children’s years of schooling is .43 among females in the ’84 cohort and .49 among females in the ’94 cohort. It is .47 among males in the ’84 cohort and .50 among males in the ’94 cohort (Panel A of Table 2). These correlations are consistent with the findings in previous studies (e.g., .46 in Hertz et al., 2007). They suggest that intergenerational persistence of educational attainment increases slightly between the two cohorts. Panel B reports correlations of mothers’ and children’s schooling by household income. The correlation coefficients range from .27 to .66, and these correlations differ across income groups within each column. One pattern identified in Columns 2–4 is that the correlation coefficient for the mid-level income group is the lowest. This may imply that there is a nonlinear association between income and intergenerational transmission of educational attainment; it also may suggest that the level of educational mobility (i.e., the difference between the mother’s and child’s attainment) is higher in the mid-level income group than in the other income groups. In Panel C, estimates for the ’84 cohort parallel those from Panel B; the correlation between mothers’ and children’s schooling years is smallest for the mid-level financial assets group (.23 for females and .32 for males). However, the magnitude of the correlation for the midlevel group is dramatically larger in the ’94 cohort (.57 for females and .50 for males). The difference between the cohorts seems to indicate that the middle financial-assets group’s educational mobility has decreased over time. In addition, results in both Panel C (financial assets) and Panel D (net worth) indicate that the association between household assets and intergenerational transmission of education varies by child’s gender in the ’94 cohort. The magnitudes of coefficients for females in low-level

117

resource (financial assets and net worth) groups are greater than those for females in the high-level groups, but the opposite is true among males: the magnitudes of the coefficients for the low-level groups are smaller than those for the high-level groups. 3.3. Transition matrices of college entry Table 3 presents estimates of the probability that sampled children enter college. Estimates are conditional on mothers’ college entry and displayed for each cohort and gender. For instance, the first two columns (Columns 1 and 2) present estimates of the likelihood that females in the ’84 cohort entered college. Column 1 shows estimates for the female children of mothers who did not enter college, and Column 2 shows estimates for the female children of mothers who did enter college. The results suggest that the children of mothers who enter college are clearly more likely to enter college than are the children of mothers who never enter college; the differences in the probability of children’s college entry are reported in Columns 3, 6, 9, and 12. For example, in the ’94 cohort, the probability of college entry among male children of mothers with college entry is 40 percentage points higher than that among the male children of mothers without college entry (Panel A, Column 12). Regardless of a mother’s educational attainment, her child’s probability of college entry generally increases with the levels of household income, financial assets, and net worth (the row difference). From a child development perspective, the column differences and row differences in Table 3 are consistent with previous findings that both parental education and household wealth are positively related to children’s educational attainment. However, a social inequality perspective suggests that family background, measured here in terms of parental education and household economic resources, also contributes to educational inequality. As mentioned above, transition matrices are employed to investigate whether intergenerational transmission of educational attainment varies with household economic

Table 2 Correlation of schooling years between children and mothers (N = 2466). Correlation

1984 cohort

1994 cohort

Female

Male

Female

(1)

(2)

(3)

(4)

Panel A: correlation of schooling years .43 Panel B: correlation by level of income Low-level .35 Mid-level .40 High-level .34 Panel C: correlation by level of financial assets Low-level .46 Mid-level .23 High-level .44 Panel D: correlation by level of net worth (with home equity) Low-level .41 Mid-level .35 High-level .38

.47

.49

.50

.39 .27 .48

.66 .34 .37

.48 .45 .46

.35 .32 .51

.46 .57 .30

.22 .50 .55

.35 .40 .46

.47 .56 .23

.41 .34 .53

N

632

606

525

703

Male

J. Huang / Economics of Education Review 33 (2013) 112–123

118 Table 3 Transition matrices of college entry. Transition matrices

1984 cohort

1994 cohort

Female

Male

Female

Male

NC

CE

Dif.

NC

CE

Dif.

NC

CE

Dif.

NC

CE

Dif.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

.75

.29

.39

.78

.39

.48

.84

.36

.38

.78

.40

.50 .77 .84

.23 .35 .13

.19 .41 .62

.65 .56 .93

.44 .15 .31

.38 .64 .55

.71 .85 .92

.33 .21 .37

.24 .38 .67

.66 .77 .88

.42 .39 .21

.65 .68 .87

.34 .16 .16

.21 .37 .60

.63 .76 .83

.42 .39 .23

.35 .52 .64

.75 .91 .86

.40 .39 .22

.28 .39 .51

.60 .81 .82

.32 .42 .31

.51 .71 .91

.30 .15 .30

.20 .34 .62

.54 .75 .87

.34 .41 .25

.35 .45 .74

.71 .92 .87

.36 .47 .13

.26 .41 .50

.62 .76 .85

.36 .35 .35

Panel A: whole sample .46 Panel B: by level of income Low-level .27 Mid-level .42 High-level .71 Panel C: by level of financial assets Low-level .29 Mid-level .52 High-level .61 Panel D: by level of net worth Low-level .21 Mid-level .56 High-level .61

Notes: NC = mothers do not enter college; CE = mothers enter college; Dif. = difference between estimates of NC and CE.

3.4. Transition matrices of college completion

resources; column differences are compared across rows for one gender in a specific cohort. If the column difference increases across rows, this indicates that intergenerational transmission of educational attainment rises with household economic resources, and vice versa. For example, in Panel C of Table 3, the difference between Columns 5 and 4 decreases from .42, to .39 and then to .23 across the three levels of financial assets (reported in Column 6). This shows that intergenerational transmission of educational attainment declines as household financial assets increase. Although this example does not reflect the general pattern found in Table 3, a similar finding is that, across the panels and cohorts, the low-level resource group generally has the highest intergenerational association. Another finding in Table 3 is that intergenerational transmission is low for female children from the ’94 cohort living in households with high-level net worth. That finding is consistent with the findings in Table 2.

Table 4 presents estimates of the probability that sampled children complete a 4-year college degree. These estimates are conditional on mothers’ college completion. Two findings are particularly noteworthy. First, the differences in male children’s college completion rates are similar for the two cohorts. Among male children in the ’84 cohort, the difference in the probability of 4-year college completion between the children of mothers who completed college and those of mothers who did not is .44; the difference is.43 for males in the ’94 cohort (see Columns 6 and 12). These estimates indicate that there is hardly any change from one cohort to the other in intergenerational transmission among male children. In contrast, the difference among female children in the two cohorts decreases by 8 percentage points (from .50 to .42; see Columns 3 and 9). This suggests that

Table 4 Transition matrices of college graduation. Transition matrices

1984 cohort

1994 cohort

Female

Male

Female

Male

ND

CD

Dif.

ND

CD

Dif.

ND

CD

Dif.

ND

CD

Dif.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

.73

.50

.20

.68

.44

.29

.71

.42

.16

.63

.43

.31 .65 .84

.20 .47 .47

.07 .20 .35

.45 .52 .77

.38 .32 .42

.16 .37 .58

.38 .68 .79

.22 .31 .21

.05 .13 .36

.39 .52 .72

.34 .39 .36

1.00 .54 .77

.86 .30 .45

.09 .21 .32

.89 .35 .78

.80 .14 .46

.15 .33 .46

.61 .75 .73

.46 .42 .27

.07 .11 .30

.30 .49 .78

.23 .38 .48

.55 .92 .71

.33 .73 .41

.10 .19 .30

.55 .46 .76

.45 .27 .46

.13 .29 .53

.72 .67 .73

.59 .38 .40

.06 .16 .25

.49 .55 .69

.43 .39 .44

Panel A: whole sample .23 Panel B: by level of income Low-level .11 Mid-level .18 High-level .37 Panel C: by level of financial assets Low-level .14 Mid-level .24 High-level .32 Panel D: by level of net worth Low-level .12 Mid-level .19 High-level .30

Notes: ND = mothers do not have 4-year college degree; CD = mothers have 4-year college degree; Dif. = difference between estimates of ND and CD.

J. Huang / Economics of Education Review 33 (2013) 112–123

educational mobility, measured by intergenerational transmission of college graduation, improved among women. Comparison of the results in Columns 7 and 1 suggests that this improvement is mainly due to increases in college graduation rates among the female children of mothers who lack 4-year college degrees, especially among those with mid- or high-level assets and income. The second finding worthy of note is the result for female children in the ’84 cohort who have low levels of financial assets and whose mothers hold a college degree; the estimated probability of college completion is very high (100%) among these women. The estimate may be due to the fact that a very low proportion of sampled mothers have a 4-year college degree. This finding may suggest that the estimates for women in this group (low-level, Column 2 of Panel C) are less reliable than those for other groups. The same concern may be raised for the estimated probability of college completion among male children in the ’84 cohort who have low-levels of financial assets and whose mothers hold college degrees (Column 5 of Panel C). Estimates for the ’94 cohort suggest that intergenerational transmission of 4-year college completion does not

119

vary by household income (Columns 9 and 12, Panel B). Although intergenerational persistence of college completion among male children rises with financial assets (Column 12, Panel C), intergenerational persistence among female children decreases as financial assets and net worth grow (Column 9, Panels C and D). 3.5. Regression results Tables 5–7 present results from regressions that examine children’s years of schooling, college entry, and 4-year college completion. Each cell in Columns 1, 2, 4, and 5 represents a separate regression model. The table reports regression coefficients of parental education and the terms for interactions with household economic resources. Gender differences are reported in Columns 3 and 6. Cohort differences are presented in Columns 7 and 8. 3.5.1. Schooling years Table 5 reports results from an OLS regression that examines children’s years of schooling. None of the estimated terms for the interaction of parental education with household income is statistically significant (Columns

Table 5 OLS regression results of schooling years. Variables

1984 cohort

Panel A: income Mother’s schooling years Log(income)  mother’s schooling years Panel B: financial assets Mother’s schooling years Log(financial assets)  mother’s schooling years Panel C: net worth Mother’s schooling years Log(net worth)  mother’s schooling years

1994 cohort

Cohort difference

Female

Male

Male–female

Female

Male

Male–female

Female–female

Male–male

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

.257 .049

.486 .006

.742 .055

.037 .026

.541 .088

.578 .062

.294 .023

1.027 .094

.229*** .009

.390*** .005

.161 .004

.422*** .007

.144 .032**

.278* .039**

.193** .016*

.246 .027

.153** .016**

.183* .024**

.030 .008

.379*** .001

.142 .028***

.237* .029***

.226** .017*

.041 .004

* p < .1. ** p < .05. *** p < .01.

Table 6 Logit regression results of college entry. Variables

Panel A: income Mother’s college entry Log(income)  mother’s college entry Panel B: financial assets Mother’s college entry Log (financial assets)  mother’s college entry Panel C: net worth Mother’s college entry Log (financial assets)  mother’s college entry ** p < .05. *** p < .01.

1984 cohort

1994 cohort

Cohort difference

Female

Male

Male–female

Female

Male

Male–female

Female–female

Male–male

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

.156 .062

4.750 .328

4.594 .390

1.062 .227

2.143 .337

1.081 .110

1.218 .165

6.893 .665

.634 .050

2.335** .083

1.701 .133

1.762*** .021

1.471*** .017

.291 .038

1.128 .071

.863 .100

.038 .102

2.082 .048

2.044 .150

1.713*** .014

.975 .063

.738 .077

1.675 .116

1.107 .111

120

J. Huang / Economics of Education Review 33 (2013) 112–123

Table 7 Logit regression results of college graduation. Variables

Panel A: income Mother’s college graduation Log(income)  mother’s college graduation Panel B: financial assets Mother’s college graduation Log(financial assets)  mother’s college graduation Panel C: net worth Mother’s college graduation Log(net worth)  mother’s college graduation

1984 cohort

1994 cohort

Cohort difference

Female

Male

Male–female

Female

Male

Male–female

Female–female

Male–male

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

8.843 .683

9.468 .877

5.395 .402

11.401* .877

6.006 .475

6.020 .596

2.558 .194

.625 .194 2.623** .095

3.178** .138

4.066** .216

3.042 .122

.555 .043

3.448*** .214***

2.615** .057

.833 .157

.825 .119

.563 .081

1.024 .094

4.020** .246***

3.858** .164

.162 .082

.046 .030

.816 .044

* p < .1. ** p < .05. *** p < .01.

3.5.2. College entry and college completion Table 6 presents results from a logit regression that examines college entry. None of the terms for the interactions between mothers’ college entry and household economic resources (including income and) is statistically significant. Estimates in Table 6 provide no empirical support for the position that intergenerational transmission of college entry varies by household economic resources. Table 7 presents results for children’s college completion, but only two of the interaction terms are statistically significant. Among female children in the ’94 cohort, household assets (either financial assets or net worth) lessen intergenerational persistence of college completion to a statistically significant degree (Column 4: Panel B b = .214, p < .01; Panel C b = .246, p < .01).

3.6. Magnitudes of the interactive associations of household assets Figs. 1 and 2 display the magnitudes of associations between each of two household asset measures (financial assets, net worth) and intergenerational transmission of educational attainment in the ’94 cohort. Based on the analyses in Table 5 (Column 5, Panel B for financial assets and Panel C for net worth), Fig. 1 presents predicted differences in schooling years by household assets (financial assets, net worth) for a typical male child. The median values of control variables are used to define a typical child. The typical child is a 29-year-old White individual who is the second child to his/her mother; the individual lives in a household headed by a male who was 43 years old, married, and employed in 1994; the individual’s household included four members (two of whom were children). The difference in schooling years is calculated by subtracting the predicted schooling years of a typical

[(Fig._1)TD$IG]

1.8 Financial Assets Net Worth

1.6

Difference in Schooling Years

1, 2, 4, and 5, Panel A). The results indicate that the association between children’s and mothers’ years of schooling does not vary by the level of household income, gender, or cohort. In the ’84 cohort, the association between mothers’ and children’s years of schooling differs by net worth. For both female and male children, the association between children’s and mothers’ schooling years increases significantly as net worth grows (Column 1: b = .016, p < .05; Column 2: b = .024, p < .05); that is, as net worth increases, so also does the intergenerational persistence of educational attainment, measured by schooling years. However, in the ’94 cohort, household assets (both financial assets and net worth) increase the magnitude of the association between parents’ and children’s schooling years among male children (Column 5: Panel B b = .032, p < .05; Panel C b = .028, p < .01) but not among their female counterparts; this gender difference is statistically significant (Column 6: Panel B b = .039, p < .05; Panel C b = .029, p < .01). In terms of cohort differences, intergenerational transmission of years of schooling among female children is estimated to vary to a statistically significant degree between the two by level of financial assets (Column 7: b = .016, p < .1) and net worth (Column 8: .017, p < .1).

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0

10000

20000

30000

40000

50000

Asset Value

Fig. 1. Predicted difference in schooling years by household assets for a typical male child in the 1994 cohort. The figure is based on the analyses in Table 5 (Column 5 in Panel B, for finanacial assets, and in Panel C for net worth). Difference in predicted schooling years is calculated by subtracting the predicted schooling years of a typical male child in the 1994 cohort whose mother has 12 years of schooling from that of a child in that cohort whose mother has 16 years of schooling.

[(Fig._2)TD$IG]

J. Huang / Economics of Education Review 33 (2013) 112–123

Difference in the Probability of College Completion

1.0 Financial Assets Net Worth 0.8

0.6

0.4

0.2

121

transmission of educational attainment only remain statistically significant for male children. Also, household assets (financial assets and net worth) improve educational mobility for female children in the ’94 cohort if educational attainment is measured by college completion. Supplementary analyses conducted to test the robustness of the findings generate results that are similar to those in the main analyses. The interactive association between household assets and parental education weakens if categorical asset measures are used in analyses. 4. Discussion and conclusion

0.0 0

10000

20000

30000

40000

50000

Asset Value

Fig. 2. Predicted probability of college completion by household assets for a typical female child in the 1994 cohort. The figure is based on the analyses in Table 7 (Column 4 in Panel B, for financial assets, and in Panel C for net worth). The difference in predicted probability of college completion is calculated by subtracting the predicted probability for a typical female child in the 1994 cohort whose mother does not complete college from that of a child in the same cohort whose mother completes college education.

male child whose mother has 12 years of schooling from that of a typical male child whose mother has 16 years of schooling; a greater difference indicates a stronger intergenerational transmission of educational attainment. Fig. 1 shows that the estimated difference in schooling years grows as financial assets and net worth respectively increase. If the value of financial assets increases from zero to $50,000, the difference in schooling grows by about .4 years. The marginal effect of household assets on intergenerational transmission of educational attainment seems greater in the first $10,000. Based on the analyses in Table 7 (Column 4, Panel B for financial assets and Panel C for net worth), Fig. 2 shows difference in the predicted probabilities of college completion by household assets for a typical female child in the ’94 cohort. The difference in the probability of 4-year college completion is calculated by subtracting the predicted probability of completion for a typical female child whose mother did not complete college from that of a typical female child whose mother has a college degree. Fig. 2 indicates that the difference in the probability of college completion declines as the values of financial assets and net worth respectively increase. If asset values increase from zero to $50,000, the difference in the probability of completion declines by about 10 percentage points. Household assets weaken the intergenerational transmission of educational attainment (measured by college completion) for female children in the ’94 cohort. In sum, regression analyses provide no evidence for the position that the association between parents’ and children’s educational attainment varies by level of household income. If attainment is measured by the years of schooling, results suggest that household net worth strengthens the association between mothers’ and children’s educational attainment among both females and males in the ’84 cohort. However, in the ’94 cohort, the association of household assets (financial assets, net worth) with intergenerational

4.1. Patterns of intergenerational transmission The study examines heterogeneity in the intergenerational transmission of educational attainment from parents to children in two cohorts, and it focuses on the roles played by household economic resources. Separate analyses are conducted for females and males. Multivariate analyses do not identify a consistent relationship between the measured household economic resources and intergenerational transmission of educational attainment across children’s gender and cohorts. Although the relationship between household assets and intergenerational transmission may change over time and may vary by child’s gender, this inconsistency suggests that we possess a rather limited understanding of the dynamics of household economic resources, parental education, and children’s educational attainment. An improved conceptualization is needed to explain these different patterns of intergenerational transmission. 4.2. Gender, household assets, and intergenerational association Results show that household assets have opposite associations with intergenerational transmission of educational attainment for female and male children in the ’94 cohort. Household assets increase intergenerational persistence of educational attainment for male children but reduce it for female children. The study proposes three hypotheses regarding heterogeneity (by household economic resources) in the intergenerational transmission of educational attainment. The first hypothesis posits that household assets increase the strength of the association between parents’ and children’s educational attainment because capable children living in low-wealth families are more likely to be limited by liquidity constraints than are counterparts in higher-wealth families. The second holds that household assets decrease the strength of the association between parents’ and children’s educational attainment because disadvantaged children benefit more from financial resources than do counterparts who are not disadvantaged. The third hypothesis assumes that parental educational attainment and household economic resources interact to affect children’s educational attainment negatively. The assumption is motivated by the observation that the association between household assets and parental education grows as household assets increase.

122

J. Huang / Economics of Education Review 33 (2013) 112–123

The findings regarding children’s gender do not support the third hypothesis. If this hypothesis is true, the pattern of intergenerational transmission should not differ by children’s gender. Since the association between parents’ and children’s educational attainment is found to vary by children’s gender, the study’s results do not seem to support the first two hypotheses. Nonetheless, descriptive findings do support these hypotheses. For instance, results in Panel B of Table 4 suggest that the first hypothesis (that rising assets increase strength of the association between parents’ and children’s attainment by easing liquidity constraints) seems to apply to the ’84 cohort. If household income increases from the low level to the medium level, the probability of college graduation increases more among children whose mothers completed college than among those whose mothers did not. However, greater marginal effects for children whose mothers did not enter college are observed in the ’84 cohort (Panel C of Table 3) with respect to college entry. If financial assets increase from the low to medium level, the probability of college entry improves more among children whose mothers did not enter college than among those whose mothers did so. Both hypotheses could prove true in the long process of child development. Liquidity is more likely to constrain postsecondary education than early attainment, whereas the marginal effects of financial resources (in terms of ‘‘nurture’’ effects) may be greater for disadvantaged children in early childhood than in adolescence (Huang et al., 2010). Descriptive results in Tables 3 and 4 also explain why the relationship between household assets and intergenerational transmission of educational attainment differs by child’s gender in the ’94 cohort. Regarding college entry (Table 3), the improved educational mobility for female children is mainly related to the increased probability of college entry among children in the mid-level groups whose mothers also entered college and among those in the high-level groups whose mothers did not enter college. This is also true for the college graduation rate (Table 4). In contrast, male children in the high-level group whose mothers did not enter college have a decreased probability of college entry and a decreased probability of college graduation in the two cohorts. The combined effect of household wealth and child’s gender on educational mobility is consistent with the findings in Bailey and Dynarski (2011), who also suggest that the change in educational mobility is largely driven by women. A possible explanation of the gender difference is that households with female children, especially those with financial resources, respond more readily than those with male children to increased return on education. 4.3. Can asset building reduce intergenerational education inequality? Although not a general pattern, the relationship between household assets and intergenerational transmission of educational attainment among female children in the ’94 cohort suggests that asset building has the potential to reduce intergenerational inequality in educational attainment. As estimates in Tables 3 and 4 show,

current distributions of family background (including parental education and household economic resources) increase that inequality. If household wealth can reduce intergenerational transmission of educational attainment, as it does for sampled female children in the ’94 cohort, then asset building, especially progressive asset building with disadvantaged families, may generate double effects against educational inequality. The key to the success of the asset-building approach in promoting intergenerational educational mobility is to understand the underlying causes of the differential effects that household assets have on intergenerational transmission of educational attainment by child’s gender. It would be interesting to examine factors and policies that may affect women and men differently. As suggested by Bailey and Dynarski (2011), some of these factors could be gender difference in classroom interaction, returns on education, marriage, and labor market opportunities. 4.4. Limitations The study has several limitations. First, as is known, the process to achieve educational success is complex, and many factors that affect child educational outcomes, such as school system, educational policy, and labor market returns on education are not included in the analysis. It is not clear how these factors interact with household economic resources and intergenerational transmission of educational attainment; a more complete model should be developed in future research. Second, as discussed above, the study cannot separate ‘‘nurture’’ effects from ‘‘nature’’ effects, nor is it focused on the causal mechanism that connects household assets, parental education, and child’s educational outcomes. Rather, the study describes the role of household economic resources in the pattern of intergenerational transmission of educational attainment. Third, although the sample includes both Black and White children, the association between household economic resources and intergenerational transmission of educational attainment may vary by race, and this possibility should be investigated in the future. In addition, a possible limitation of the current design may be that the effects of age, periods, and cohorts are confounded. In summary, the study shows that the association between household assets and intergenerational transmission of educational attainment has changed over time. It also indicates that the association varies by children’s gender. Progressive and inclusive asset-building policies may have the potential to promote educational mobility, but the achievement of this mobility depends to some extent on an accurate understanding of how household assets interact with other determinants of children’s education. References Anger, S. (2011, May). The intergenerational transmission of cognitive and noncognitive skills during adolescence and young adulthood. Discussion Paper 5749. Bonn, Germany: Institute for the Study of Labor. Bailey, M. J., & Dynarski, S.M. (2011, December). Gains and gaps: Changing inequality in U.S. college entry and completion. Working Paper 17633. Cambridge, MA: National Bureau of Economic Research.

J. Huang / Economics of Education Review 33 (2013) 112–123 Bauer, P., & Riphahn, R. (2004, October). Heterogeneity in the intergenerational transmission of educational attainment: Evidence from Switzerland on natives and second generation immigrants. Discussion Paper 1354. Bonn, Germany: Institute for the Study of Labor. Baum, S., & Ma, J. (2007). Education pays 2007: The benefits of higher education for individuals and society (report). Washington, DC: College Board. Bjo¨rklund, A., & Salvanes, K. (2010). Education and family background: Mechanisms and policies. In Hanushek, E. A., Machin, S., & Woessmann, L. (Eds.), Handbook of the economics of education (Vol. 3, pp. 201–247). Amsterdam: North-Holland Black, S. E, & Devereux, P. J. (2010, April). Recent development in intergenerational mobility. Discussion Paper 4866. Bonn, Germany: Institute for the Study of Labor. Conley, D. (2001). Capital for college: Parental assets and postsecondary schooling. Sociology of Education, 74(1), 59–72 http://dx.doi.org/ 10.2307/2673145. Duncan, G. J., Ziol-Guest, K. M., & Kalil, A. (2010). Early childhood poverty and adult attainment, behavior, and health. Child Development, 81(1), 306– 325 http://dx.doi.org/10.1111/j.1467-8624.2009.01396.x. Elliott, W., III, Destin, M., & Friedline, T. (2011). Taking stock of ten years of research on the relationship between assets and children’s educational outcomes: Implications for theory, policy and intervention. Children and Youth Services Review, 33(11), 2312–2328 http://dx.doi.org/10.1016/ j.childyouth.2011.08.001. Gaviria, A. (2002). Intergenerational mobility, sibling inequality and borrowing constraints. Economics of Education Review, 21(4), 331–340 http://dx.doi.org/10.1016/S0272-7757(01)00031-0. Glick, P., & Sahn, D. E. (2009). Cognitive skills among children in Senegal: Disentangling the roles of schooling and family background. Economics of Education Review, 28(2), 178–188 http://dx.doi.org/10.1016/j.econedurev.2007.12.003. Haveman, R., & Wolfe, B. (1995). The determinants of children’s attainments: A review of methods and findings. Journal of Economic Literature, 33(4), 1829–1878. Hertz, T., Jayasundera, T., Piraino, P., Selcuk, S., Smith, N., & Verashchagina, A. (2007). The inheritance of educational inequality: International comparisons and fifty-year trends. The B.E. Journal of Economic Analysis & Policy7(2) http://dx.doi.org/10.2202/19351682.1775 (article 10). Huang, J. (2011). Asset effects for children with disabilities: Analysis of educational and health outcomes. Doctoral dissertation. St. Louis, MO:

.

123

Washington University in St. Louis. Retrieved from http://openscholarship.wustl.edu/etd/161/ Huang, J., Guo, B., Kim, Y., & Sherraden, M. (2010). Parental income, assets, borrowing constraints and children’s post-secondary education. Children and Youth Services Review, 32(4), 585–594 http://dx.doi.org/ 10.1016/j.childyouth.2009.12.005. Kane, T. J. (2004). College-going and inequality. In K. M. Neckerman (Ed.), Social inequality (pp. 319–353). New York: Sage. Kim, Y., & Sherraden, M. (2011). Do parental assets matter for children’s educational attainment? Evidence from mediation tests. Children and Youth Services Review, 33(6), 969–979 http://dx.doi.org/10.1016/j.childyouth.2011.01.003. Krueger, A. (2012, January). The rise and consequences of inequality in the United States. Washington, DC: Center for American Progress. Retrieved from http://www.americanprogress.org/events/2012/01/12/17181/therise-and-consequences-of-inequality/. Mayer, S. E. (1997). What money can’t buy: Family income and children’s life chances. Cambridge, MA: Harvard University Press. Nam, Y., & Huang, J. (2009). Equal opportunity for all? Parental economic resources and children’s educational attainment. Children and Youth Services Review, 31(6), 625–634 http://dx.doi.org/10.1016/j.childyouth.2008.12.002. Nam, Y., & Huang, J. (2011). Changing roles of parental economic resources in children’s educational attainment. Social Work Research, 35(4), 203–213 http://dx.doi.org/10.1093/swr/35.4.203. Orr, A. J. (2003). Black-White differences in achievement: The importance of wealth. Sociology of Education, 76(4), 281–304 http://dx.doi.org/10.2307/ 1519867. Sacerdote, B. (2011). Nature and nurture effects on children’s outcomes: What have we learned from studies of twins and adoptees? In Benhabib, J., Jackson, M. O., & Bisin, A. (Eds.), Handbook of social economics (Vol. 1B, pp. 1–30). Amsterdam: North-Holland Schreiner, M., & Sherraden, M. (2007). Can the poor save? Saving and asset building in individual development accounts New Brunswick: Transaction Publishers. Shapiro, T. M. (2004). The hidden cost of being African American: How wealth perpetuates inequality. New York: Oxford University Press. Sherraden, M. (1991). Assets and the poor: A new American welfare policy. Armonk, NY: ME Sharpe. Solon, G. (1992). Intergenerational income mobility in the United States. American Economic Review, 82(3), 393–408.

.