First, we must point out that a computer at home without access to Internet may not be directly useful in learning tasks, but may
nevertheless improve some cognitive skill that is itself useful. In fact, this is the argument put forth by Biagi and Loi (2013) to explain why
gaming was the only computer activity for which there was a correlation with higher test grades.
A second explanation is that computer ownership is not the “cause” of the better test scores, but it is correlated to some other hidden
factor, which is itself the “cause” of the improved scores. Fairlie and London (2012) raise the possibility that computer ownership correlates
with unmeasurable variables such as academic motivation (which they call omitted variable bias) e students and their families who are
more “educationally motivated” are also the ones that most likely purchase computers for school use.We must point out that our computer
ownership variable do not necessarily refer to “computer for school use” e Prova Brasil asks whether the family owns a computer, not
whether the student has access to a computer for school use.
But the results in Section 3.3 disconfirm two of those hypothesis: that families that own computers are likely richer than families in the
same SEC that do not own them or have parents with higher levels of education, and it is these omitted variables that explain the better
scores.
Fig. 3 shows that the head of families that own a computer at home do not have, in general, higher levels of education than the head of
families without computers. The figure also shows that families with computers at home are richer than families without computers, in the
same SEC. But the results in Fig. 4 show that the wealth-adjusted effect size is similar to the non wealth-adjusted effect size, that is, even
when adjusting for the differences in wealth within a SEC, the effect size of owning a computer remains generally the same.
Thus, this paper shows that: