Data and Methods

\label{data-and-methods}
To determine the long-run effects of veterans’ transitions into the labor force during weak and strong labor markets, we require a dataset covering a long period of time—only such a dataset will allow us to follow veterans for multiple decades while also including a wide variety of economic conditions. Fortunately the Health and Retirement Study (HRS) provides very long-term information on a sample of veterans who served during several different periods of time. While the recent conflicts have involved large numbers of service members, veterans are far more predominant among older men. About 20 percent of men aged 50-65 are veterans and among men over 50, veteran status is more common than college completion.11Authors’ calculations from the American Community Survey 2014. This data source indicates roughly 6.3 million male veterans aged 50-65 and nearly 15 million male veterans aged 50-plus in 2014 and these numbers accord well with various figures available from va.gov. In contrast, as of 2014 there were some 2.6 million Post-9/11 era veterans (see va.gov). Despite the recent conflicts, most veterans today are approaching or have passed retirement age; about one-third of male veterans are over 70 years of age and about three-quarters are over 50. Therefore, it is not surprising that veterans are well represented in the HRS, which has information on a total of 7,820 veterans.
For this study we focus on a subsample of veterans, those that i) served four years or less; ii) enlisted before the elimination of the draft and establishment of an all-volunteer-force (AVF); iii) left active service before the age of 25; iv) are younger than 65 years old (i.e. pre-retirement age). There are about 2,800 veterans in the HRS that fit our sampling criteria. Figure 2 shows the distribution of this subset of veterans in the HRS by the year they exited active service. It also shows the national unemployment rate that prevailed at the time they exited. Figure 2 shows that there is large variation in the unemployment rate even among veterans leaving the military in the same era, and that our sample includes veterans from several different eras.
For this study, we leverage on the following datasets: i) RAND HRS Flat Files, which contains information on veteran status, the year enlisted in active service and the year exited from active survey, ii) the RAND HRS Data File, which contains variables harmonized across waves regarding employment status, work expectations, and several measures of household wealth; ii) the respondent cross-year SSA summary earnings file, which provides information on reported earnings from federal tax returns, 1951 through 2013, for those HRS respondents have given their consent to access this information; the cross-year prospective social security wealth measures of pre-retirees, available for 1992, 1998, 2004 and 2010 (datasets contributed by Kapinos et al.); and the pension wealth data files, from 1992 to 2010 (contributed by Gustman, Steinmeier and Tabatabai).
Figure : Number of Veterans in the HRS, by year exited Active Service, and Unemployment Rate
[CHART]
NOTES: Authors’ Calculations based on HRS files and Bureau of Labor Statistics data.
Table 1 shows the sample descriptive statistics, calculated on all the observations from the RAND HRS Data File used in the analysis (a total of 12,081 person-year observations). The average age in our analysis sample is 58.43 years. The average age at enrollment is 19.61 years, and the average age at exit from active service is 22.11 years. The average veteran in our sample exited active service 36.76 years ago, with a prevailing unemployment rate of 4.97%. The majority of veterans are male (98%), white (87%), with a high school degree or GED (40%), married (80%) and lives in a coupled household (84%). Table 1A provides similar information, but divides the sample into roughly equal groups depending on the unemployment rate at the time of exit from the military. Table 1A indicates that the two groups are reasonably similar on many measures, such as age, but that those who exit in periods of high unemployment hive slightly higher levels of education and slightly higher earnings than veterans who exited during periods of lower unemployment.
Table
Descriptive Statistics