Parental Education as an Alternative Measure of Economic Background

education are –1.052 and –1.487, respectively. The broad similarity of these two sets of estimates suggests that, while selection bias from nonrandom MAOA variation is important in principle, it may be relatively small in practice, at least in the context of the particular sample and environmental measures used here. The robustness of the basic result to sibling fi xed- effects specifi cations that use plausibly exogenous genetic variation to identify the interaction terms is important suggestive evidence that those interactions are causal in nature. However, a number of considerations make it inappropriate to assign the interaction a fully causal in- terpretation. One outstanding issue is whether the interactive effects are specifi c to family income, as opposed to socioeconomic background more generally, and this is discussed in the next subsection. Additional reasons for caution are discussed in Sec- tions V and VI below.

D. Parental Education as an Alternative Measure of Economic Background

As noted, a causal estimate of the interaction between childhood income and MAOA status requires exogenous variation in both of these variables, but with respect to in- come all of the models estimated above utilize simple cross- sectional variation. This leaves open the possibility that the observed interactions are actually between MAOA status and something associated with household income, rather than household income itself. Indeed, some researchers have questioned whether income is causally related to educational outcomes at all, or if this commonly observed association is wholly due to omitted variable bias. Recent work in this area suggests that household income does indeed exert a causal effect on educational outcomes. For instance: Dahl and Lochner 2012 use discontinuous changes in the U.S. Earned Income Tax Credit schedule to instrument for income and fi nd that a 1,000 increase in income improves student test scores by approximately 6 percent of a standard deviation; Milligan and Stabile 2011 use income variation stemming from changes in the Canadian Child Tax Benefi t and fi nd that a 1,000 increase in income decreases the chances that a child is diagnosed with a learning disability by 2.8 percentage points and positively affects child test scores; and Duncan, Morris, and Rodrigues 2011 use income variation induced by 16 “welfare- to- work” experiments that randomly assigned participants to benefi ts packages and employment programs, and in their preferred specifi cation fi nd that a 1,000 increase in income improved student achievement by 6 percent of a standard deviation. Despite this evidence, family income as measured in the present study still may simply refl ect more general socioeconomic status or unobserved parental character- istics, and it would be informative to know the extent to which the interactive effects are present with respect to other markers of socioeconomic background. A natural alternative, which is available in the Add Health data, is maternal education. 27 To investigate the relative importance of family income and maternal education more closely, Table 6 reports results from two sets of models: Panel A of Table 6 shows estimates from models that include indicators of whether the respondent’s 27. Paternal education is also available but has a large number of missing values and is more prone to mis- reporting because the mothers of Add Health participants typically completed the parental questionnaire and reported their education directly, while paternal education was reported by the adolescents. Thompson 283 Table 6 Maternal Education as an Alternative Measure of Economic Background College Attendance College Graduation Years of Education MAOA = 0 1 MAOA = 1 2 MAOA = 0 3 MAOA = 1 4 MAOA = 0 5 MAOA = 1 6 Panel A: Maternal Education Only Mom High School Graduate 0.163 0.071 –0.029 –0.042 0.797 0.752 0.125 0.070 0.131 0.089 0.643 0.481 Mom College Graduate 0.334 0.264 0.328 0.231 2.206 1.874 0.129 0.080 0.116 0.089 0.586 0.461 Observations 397 543 397 543 356 474 Panel B: Both Income and Maternal Education Log Income 0.091 0.032 0.069 0.014 0.748 0.224 0.047 0.020 0.043 0.025 0.278 0.141 Mom High School Graduate 0.382 0.019 0.062 –0.101 1.180 0.244 0.119 0.085 0.127 0.095 0.623 0.513 Mom College Graduate 0.424 0.219 0.357 0.195 1.821 1.322 0.125 0.101 0.119 0.097 0.617 0.488 Observations 362 495 362 495 328 432 Notes: All regressions are estimated for males only and control for birth order, number of siblings, race, language spoken in the home, parent and child ages, and school dum- mies. Standard errors, clustered at the school level, are displayed in parenthesis. , and indicate statistical signifi cance at the 10, 5 and 1 levels, respectively. mother completed high school and whether she completed college or beyond but do not include family income the omitted maternal education category is less than high school, while Panel B reports estimates from models that contain both family income and maternal education indicators. The results show that when maternal education is entered without family income, its association with child educational success is stron- ger among children without positive MAOA status than among children with positive MAOA status. For example, among those without positive MAOA status, having a mother who completed high school as opposed to dropping out of high school is as- sociated with a 16.3 percentage point increases in the probability of attending college, and having a mother who completed college is associated with a 33.4 percentage point increase in the probability of attending college. However, the analogous associations for children with positive MAOA status are considerably weaker, only 7.1 and 26.4 percentage points, respectively. When household income is added to these specifi ca- tions in Panel B of Table 6, the association between child educational outcomes and both maternal education and household income are much stronger for children without positive MAOA status than those with positive MAOA status. 28 In short, Table 6 shows that the results from the baseline models that used house- hold income also apply to another reasonable measure of family background, maternal education. Discerning precisely which aspects of socioeconomic background interact with MAOA status to determine educational outcomes will require additional work using research designs that rely on well measured and plausibly exogenous varia- tion in family background measures, and unfortunately such designs are not possible with the data used in this study. Given this, at present household income is probably best viewed as an imperfect index of general socioeconomic background, as opposed to having a well- identifi ed interactive effect in its own right. The fact that maternal education displays interactive effects that are broadly similar to those of household income reinforces this view.

V. Further Issues