Mechanisms and Extrapolation in the Abortion-Crime Controversy - Mechanism and Causality in Biology and Economics - page 194

providing evidence for a causal claim may wane as researchers become better able
to study the target population. But there is a simple reason why such a course of
events is unlikely to occur in relation to the impact in the USA of legalized abortion
in the 1970s upon crime rates in the 1990s, namely, that being born from an
unwanted pregnancy is unmeasured in the USA data. This variable can only be
accurately measured in rather unusual circumstances, as illustrated by the European
studies described in Sect. 1 . Moreover, the claim that being born unwanted increases
the likelihood of criminal activity later in life is a basic premise of the mechanism
underlying Donohue and Levitt's hypothesis. 11 Furthermore, that mechanism plays
an important role in reinforcing their arguments that the inverse correlation they find
between abortion rates and lagged crime rates reflects a causal impact rather than the
presence of some latent confounding factor. Finally, this important and apparently
inescapable role of extrapolation does not, in and of itself, demonstrate any grave
infirmity in Donohue and Levitt's overall argument. It is commonplace for
mechanisms to play an important role in causal inference in social science. And in
this case the extrapolation appears sufficient for the case at hand. First, a plausible
case can be made for extrapolating claim about positive causal relevance (i.e., that
being born unwanted makes a person more likely to be convicted of crimes later in
life). Secondly, Donohue and Levitt's quantitative extrapolation—that being born
unwanted doubles the chance of criminal conviction—need only be roughly accurate
for the purposes of their argument, and this rough accuracy is supported by the
compatibility of the results of their scale-up model and their estimates from national-
level statistical data. Of course, the purpose of this chapter is not to defend the
correctness of Donohue and Levitt's hypothesis. The point here is merely that
extrapolation plays an important role in that argument and, furthermore, that this
role of extrapolation is not a reason for thinking that they have failed to make a strong
case for their conclusion. If Donohue and Levitt's statistical arguments are basically
correct, then the extrapolation is one significant supporting plank in the overall
structure of a strong argument. Therefore, this case belies the objection that extrapo-
lation can be relevant as evidenced only in a context of massive uncertainty.
5 Conclusions
Let us recap the three interconnected philosophical themes relating to mechanisms
and extrapolation that are highlighted by the case study discussed here. The first of
these is that there are different types of causal claim that one might wish to
extrapolate and that extrapolations of more informative causal claims typically
11 For example, Levitt characterizes the hypothesis as resting on two premises: “(1) unwanted
children are more likely to commit crime, and (2) legalized abortion leads to a reduction in the
number of unwanted births” ( 2004 , pp. 181-182).