III. IMPUESTO SOBRE HIDROCARBUROS
1. LA SENTENCIA DEL TRIBUNAL DE JUSTICIA DE LA UNIÓN EUROPEA DE
Although America’s time in the war lasted less than two years, there was still con- siderable variation in the probability of military service across cohorts. Figure 3.2 shows the share of WWI veterans by cohort, based on the 1900-1930 linked sample. I divide the population into two age groups to allow for a clearer presentation of the trends in military service. Figure 3.2A focuses on the older men born from December of 1883 to November of 1888, while Figure 3.2B looks at the younger generation of men born from December of 1894 to November of 1899.18 This essentially provides a closeup view of the first “hump” in Figure 3.1, but uses the more precise birth month variable in the linked sample to define cohorts. In general, the likelihood of military service rises from under 10 percent for men born before the middle of 1886 to a peak of around 50 percent for the 1895 and 1896 cohorts, before dropping thereafter.19
On closer examination, Figure 3.2 reveals four discontinuous jumps in the proba- bility of being a WWI veteran that coincide with the changes in registration regimes or phases within a registration regime. The two larger discontinuities, which will pro- vide the identifying variation in this chapter, are demarcated by the solid vertical lines labelled RDO and RDY. RDO, drawn between the May and June cohorts of 1886 in
Figure 3.2A, indicates the upper age limit of the first registration. Cohorts born just after this threshold came under the first registration, while those who were slightly older fell under the third registration that had wider age bands. RDY, positioned
between the May and June cohorts of 1897 in Figure 3.2B, marks the transition from Phase A to Phase B of the second registration. Men born just after this cutoff came
18These two age groups do not comprise the full linked sample. However, they correspond to the
specific cohorts that will be used in the main analysis below.
19The general increase in the share of veterans after the 1886 cohort is likely to be the result
of two factors. First, deferment on the basis of marital status and dependent children was more applicable to older men. Second, younger men were less likely to be important figures in their respective industries, and thus had a higher chance of being drafted since taking them out of the civilian labor force would not compromise industry performance significantly.
under Phase B of the second registration or the third registration, but those born earlier were subject to either the first registration or Phase A of the second registra- tion instead. To facilitate the exposition below, I will sometimes collectively refer to these two thresholds as RDj, where j = {O, Y }, particularly when discussing issues
that apply to both cutoffs. In addition to RDO and RDY, there are also two smaller
discontinuities in the likelihood of military service as indicated by the vertical dashed lines in Figure 3.2B. The line to the left of RDY denotes the change from the first
registration to Phase A of the second registration, while the line on the right of RDY
represents the shift from Phase B of the second registration to the third registration. As precision is important for the main analysis, these smaller jumps will not be used as part of the identification strategy.
What explains the discrete jumps in the likelihood of military service when reg- istration regimes switch around RDO and RDY?20 Neither the economics nor the
history literature has provided explicit reasons for these discontinuities.21 They are
not simply the result of rapidly declining demands for manpower at the end of a war, as that would have predicted continuous and steep declines rather than distinct breaks in the probability of serving in the military between specific cohorts.22 I argue
20While they are not part of the main identification strategy, one might also be interested in
understanding the reasons for the two smaller discontinuous jumps in Figure 3.2B indicated by the vertical dashed lines. The dip in the likelihood of WWI service after the dashed line on the left of RDY likely reflects an amendment made to the draft bill for the second registration. The amendment
provided that those under the second registration “were to be placed at the bottom of the list of those liable to military service in the several classes to which they were assigned... draft authorities were directed not to order into military service any registrant of the June, 1918, registration until those registrants of the first registration... who were available for general military service, had been ordered to report for duty” (United States War Office, 1919). The drop in the probability of military service after the dashed line on the right of RDY, in contrast, may simply reflect the fact that the
third registration was conducted later than Phase B of the second registration, so there was even less time to mobilize cohorts from the third registration compared to those under Phase B of the second registration.
21Although Fetter (2013) and Larsen et al. (2015) use the break at RD
Y to study how WWI
military service affected homeownership and assortative mating respectively, they do not discuss any reasons for the presence or the magnitude of the discontinuity.
22Focusing on WWII, Bound and Turner (2002) document that among white men born between
that these jumps arise from the interaction between the age range of each registration regime (or phase within a registration regime) and the unexpected timing of the war’s end. Several pieces of historical evidence may be used to substantiate the latter claim. The Allies anticipated that the war would continue well into 1919 (Chambers, 1987). General John J. Pershing, the commander in charge of the American Expeditionary Forces (AEF) in France, believed that the US would play a key role in defeating the Germans, but in 1919 rather than 1918. He expected the Western Front to remain in a stalemate in 1918, with both sides battle-weary (Zieger, 2000). It was in ex- pectation of this longer battle that the third registration was called for. Manpower estimates indicated that two million more American troops would be required from October of 1918 to June of 1919 (United States War Office, 1919). In fact, US Pres- ident Woodrow Wilson “agreed on July 18, 1918 to triple the size of the AEF by the spring of 1919 to nearly 3.2 million men and to have an American army of 5 million men in France by the time the war was expected to end in late 1919” (Chambers, 1987). Given this context, the rapid advances made by the Allies in the second half of 1918, as well as Germany’s decision to seek armistice well before the end of the same year, were unanticipated.23,24 The timing and suddenness of the war’s end, in turn,
affected the likelihood of military service particularly for those coming under Phase B of the second registration and the third registration. The time interval between these registrations and the armistice on November 11, 1918, at approximately two to three months, was relatively short to allow for a full mobilization of the registrants.25 In
share fell dramatically for subsequent cohorts as WWII drew to a close. However, while the decline was rapid, it did not occur at a specific cohort – there was no discrete shift (Fetter, 2013).
23The turn of events was, in part, due to the tactical nous of General Ferdinand Foch, the supreme
Allied commander. Foch was able to identify and counter-attack German weak points (Zieger, 2000). The AEF also played its part in putting the Germans on the defensive, particularly at Aisne-Marne, considered to be Foch’s first and perhaps most crucial counterstroke (Zieger, 2000).
24German Chancellor Prince Max asked US President Woodrow Wilson for armistice on October
4, 1918. By then, the national lottery for the third registration had already been conducted from September 30, 1918, to October 1, 1918.
fact, most persons affected by the third draft were still in the midst of being classified for service when the war ended (Chambers, 1987). Furthermore, the 1918 influenza pandemic led many states to suspend physical examinations during October of 1918 (United States War Office, 1919), slowing down the rate of induction and ultimately the probability of military service.
Do the discrete changes in the probability of military service constitute a valid identification strategy? The success of this approach hinges on whether being born just before or after the thresholds RDO and RDY is as good as random. If so, changes
in the outcomes of interest at these cutoffs can reasonably be attributed to differences in the likelihood of military service. As discussed in the preceding paragraph, the ex- istence and position of the discontinuities are arbitrary functions of the registration age limits and the timing of the war’s end. That all four discontinuities in Figure 3.2 match the age limits of the three registrations down to the month of birth strongly suggests that the discontinuous jumps in the likelihood of military service are exoge- nously driven. In the next section, I further substantiate this claim by comparing the household characteristics during childhood of those born just before and after the thresholds RDO and RDY.