An Empirical Justification for the Use of Draft Lottery Numbers as a Random Treatment in Political Science Research
In a series of papers in the 1990s, Joshua Angrist and Alan Krueger (1991, 1992) sparked interest in the use of the draft lotteries held with the United States in the late 1960s and early 1970s as an instrument for economic outcomes. Over the last few years, there has been growing use of the draft lottery instrument within political science to study political attitudes and behaviors. The lottery is indeed a potentially powerful design because, if conducted correctly, it should provide true randomization for the “treatment” of military service (or behavioral reactions to the threat of such service). However, the first draft lottery conducted in 1969 was not conducted in a truly random manner. Because the Selective Service officials employed faulty procedures, the randomization failed. As a result, those men with birthdates at the end of the year were more likely to have low draft numbers than citizens with birthdates earlier in the year (meaning that they were more likely to be called to service). Previous research suggests that people born at the end of the year are different on key demographic markers than those born earlier in the year. Given the nature of the randomization failure in 1969, the use of draft lottery numbers could confound the effect of the draft with established quarter-of-birth effects. In practice, though, there are small and largely statistically insignificant differences on politically relevant variables between those individuals born early in the year and those born later in the year. Thus, researchers can treat the 1969 draft lottery numbers as if they were assigned at random. However, to account for unmeasured differences based on quarter of birth, I suggest that when using draft numbers as instruments in analyses, researchers should include robustness tests which include measures for the respondents’ quarter or month of birth.
Keywords not supplied
Document ID Number