Rees and Sabia 325
sleep, both of which have been found to be associated with migraine headache Kelman and Rains 2005; Scher et al. 2005.
5
Finally, we include indicators of drunkenness to capture any “hangover effects” that might have been misdiagnosed
by parents as migraine headache, as well as school and grade effects to control for community and school characteristics that could potentially be correlated with mi-
graine triggers such as noise, weather, or even florescent lighting.
6
After estimating Equation 2, we used nearest neighbor matching without replace- ment to assign migraineurs with a nonmigraineur whose estimated propensity score
was within 0.0015. In a further effort to ensure common support, we dropped mig- raineurs whose estimated propensity score was higher than the maximum or less
than the minimum estimated propensity score of nonmigraineurs, and then dropped the 10 percent of migraineurs whose propensity score was furthest from the propen-
sity score of their match. This procedure produced a matched sample that appears to be well-balanced across observables.
7
The second column of Table 2 presents the PSM estimates. When the male and females samples are combined, the PSM estimates are smaller in absolute magnitude
than the OLS estimates, but nonetheless provide evidence of a negative relationship between migraine headache and educational attainment. Specifically, migraine head-
ache is associated with a 0.071 decrease in high school GPA and a 0.045 decrease in the probability of college attendance. The estimated effect of migraine headache
on high school graduation is not statistically significant at conventional levels. When the sample is divided by gender, the PSM estimates become less precise, but suggest
that the effect of migraine headache on male educational attainment may be larger than its effect on female educational attainment. For instance, migraine headache is
associated with a 0.077 decrease in the probability of attending college for males, but a statistically insignificant 0.024 decrease for females.
VI. Sibling Comparisons
Propensity score matching will generate an unbiased estimate of the effect of migraine headache provided that the matching variables adequately capture
the influence of confounders. However, we cannot rule out the possibility that dif-
5. There also is evidence that migraine headache sufferers are more likely to exhibit neurotic tendencies than nonsufferers Rasmussen 1992; Kentle 1997; Cao et al. 2002, as well as evidence that the association
between tension headache and neuroticism is at least as strong as the association between migraine head- ache and neuroticism Rasmussen 1992; Kentle 1997; Breslau and Rasmussen 2001; Cao et al. 2002; Zwart
et al. 2003. In an effort to explore whether neuroticism could be driving our results, we used PSM to estimate the relationship between nonmigraine headache and educational attainment. None of the outcomes
under study were related to nonmigraine headache, suggesting that the negative relationship between mi- graine headache and educational attainment is not attributable to neuroticism.
6. Estimates of Equation 2 indicated that a number of the variables in X
i
and Z
i
were associated with the probability of migraine headache. Respondents from more highly educated and richer households were less
likely to have suffered from migraine headache; those with greater symptoms of depressive symptomotol- ogy higher CES-D scores, greater cognitive ability, more sleep problems, hypertension, and more frequent
drunkenness were also more likely to have suffered from migraine headache. 7. Appendix Table A1 shows mean values of the variables in X
i
and Z
i
by migraine status for the matched sample. Migraineurs and nonmigraineurs appear to be similar with regard to the observables.
326 The Journal of Human Resources
ficult-to-measure family-level variables are driving the results presented thus far. For instance, although we controlled for family income, we do not observe how much
support or encouragement the respondent received at home. In order to address the issue of family-level unobservables, we restrict our sample
to siblings raised in the same family, j, and estimate the following equation:
E ⳱ Ⳮ ⬘X Ⳮ Migraine Ⳮ Ⳮε , 3
ij 1
ij 2
ij j
ij
where
j
is a vector of family fixed effects and the vector X
i
includes controls for age at Wave III, gender, PPVT score, BMI, height, and whether the respondent
had an older sibling. The advantage of this estimation strategy is that only within- family variation is used to estimate the effect of migraine headache on educational
attainment. All factors common to both siblings are controlled for by the vector
j
, eliminating the need to observe and measure factors having to do with the home
environment. The third column of Table 2 shows estimates of the relationship between migraine
headache and educational attainment controlling for family fixed effects FFE. The results suggest that migraine headache can lead to substantial reductions in educa-
tional attainment, although there is evidence that its impact varies by gender and the specific outcome under study.
When the combined sample of males and females is examined, migraine headache is associated with statistically significant decreases in high school GPA and the
probability of college attendance but not high school graduation; when the sample is restricted to brothers, FFE estimates suggest that migraine headache is associated
with significant decreases in high school GPA and the probability of high school graduation but not college attendance; and when the sample is restricted to sisters,
FFE estimates suggest that migraine headache is associated with a significant de- crease in the probability of college attendance but the estimated effects of migraine
headache on high school GPA and the probability of high school graduation are not statistically significant.
8
Thus, controlling for family fixed effects does not produce consistently significant estimates of the relationship between migraine headache and educational attainment,
although the fixed effects estimates are, with only one exception, negative. Moreover, they are generally larger than the OLS estimates. We view this pattern of results as
evidence that family-level unobservables are not driving the relationship between migraine headache and educational attainment.
8. These samples are restricted to siblings with different migraine histories. When grades are the dependent variable, the combined sample is composed of 214 siblings from 105 families 70 of whom were twins,
the male sample is composed of 63 brothers from 31 families 26 of whom were twins, and the female sample is composed of 86 sisters from 42 families 23 of whom were twins. When high school completion
or college attendance is on the left-hand side, the combined sample is composed of 280 siblings from 137 families 85 of whom were twins, the male sample is composed of 73 brothers from 36 families 30 of
whom were twins, and the female sample is composed of 104 sisters from 51 families 29 of whom were twins. The hypothesis that the estimated effect of migraine headache is equal across these samples is never
rejected.
Rees and Sabia 327
VII. School Absences, Difficulty Paying Attention in Class, and Difficulty Completing Homework as
