sented, and the 45,000 standard deviation of household income indicates substantial income dispersion.
IV. Results
A. The Basic Result
As a preliminary test of interactions between family income and MAOA status, Table 2 reports mean levels of educational attainment separated by MAOA status and whether
the child came from a home with an income above or below the sample median.
16
The fi rst column of Table 2 shows that among children from homes with incomes below the median, those with positive MAOA status are approximately six percentage
points more likely to attend college, ten percentage points more likely to graduate from college, and on average complete around 0.2 years more schooling than children
without positive MAOA status. In contrast, the second column shows that among children from homes with incomes above the median, MAOA status is virtually unas-
sociated with college attendance and graduation, and that in fact children with positive MAOA status complete an average of 0.27 fewer years of schooling than those without
positive MAOA status. We could equivalently view the summary statistics by row, in which case they suggest that the association between educational attainment and com-
ing from a home with income above the median is larger for children without positive MAOA status. These simple mean differences suggests that educational outcomes may
depend interactively on economic background and MAOA status.
Next, I estimate a series of models that include a large set of control variables and
16. The sample median annual income was 38,000.
Table 2 Mean Educational Outcomes by Income and MAOA Status
Household Income Below Median
Household Income Above Median
MAOA = 0 College attendance:
0.58 College attendance:
0.84 College graduation:
0.25 College graduation:
0.52 Years of schooling:
13.12 Years of schooling:
14.89 Observations:
220 Observations:
169 MAOA = 1
College attendance: 0.64
College attendance: 0.85
College graduation: 0.35
College graduation: 0.52
Years of schooling: 13.32
Years of schooling: 14.62
Observations: 246
Observations: 299
Notes: Sample includes males only. Median sample income was is 38,000 in 1994 dollars.
the results are reported in Table 3. Because previous work has found substantial con- cavities in the relationship between family income and educational attainment Loken,
Mogstad, and Wiswall 2012; Dahl and Lochner 2012, I enter family income in logs and estimate models of the following form, where each individual student is denoted
with subscript i
17
1 Educational Attainment
i
= β + 
1
Log Income
i
+ ′
X
i
 + ε
i
if MAOA
i
= 0,1 where X
i
is a vector of controls that includes school dummies,
18
number of siblings, birth order, racial and ethnic group indicators,
19
language spoken at home, and sets of indicators for child age by month and year of birth and parental age by year of
birth.
20
The inclusion of school fi xed- effects in these models is noteworthy because access to higher quality schools is a potentially crucial mechanism by which income
infl uences educational success. Selection bias of this nature is eliminated here by re- stricting the comparison to students in the same school.
Equation 1 was estimated via OLS. For the binary dependent variables indicating college attendance and college graduation a linear probability model was used but the
results are similar when using a logit model or other common limited dependent vari- able estimation methods. The results parallel the mean differences found in Table 2
in suggesting that the association between family income and educational outcomes depends on MAOA status. For all of the outcomes studied, the estimated effect of
childhood income on educational attainment is much smaller for students with positive MAOA status than it is for students without positive MAOA status. P- values from sig-
nifi cance tests for cross- model differences in the log income coeffi cients are reported at the bottom of Table 3, and in all three cases the differences across MAOA groups
are statistically signifi cant at beyond the 5 percent level. Taking the models predicting college graduation as an example, the results indicate that for students without positive
MAOA status, a doubling of gross family income is associated with an 11.7 percent- age point increase in the likelihood of college graduation. However, the size of this
effect is only 1.7 percentage points for students with positive MAOA status. Results for college attendance and years of education similarly reveal large differences in the
strength of the income- education association across the two MAOA groups.
B. Robustness Checks