American Economic Association

Lifetime Earnings and the Vietnam Era Draft Lottery: Evidence from Social Security
Administrative Records
Author(s): Joshua D. Angrist
Source: The American Economic Review, Vol. 80, No. 3 (Jun., 1990), pp. 313-336
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/2006669 .
Accessed: 01/06/2011 01:12

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=aea. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.

http://www.jstor.org
 Lifetime Earnings and the Vietnam Era Draft Lottery:
Evidence from Social Security
Administrative Records

By JOSHUA D. ANGRIST*

The randomlyassigned risk of inductiongenerated by the draft lottery is used to

construct estimates of the effect of veteran status on civilian earnings. These
estimates are not biased by thefact that certain types of men are more likely than
others to service in the military. Social Security administrative records indicate
that in the early 1980s, long after their service in Vietnam was ended, the
earnings of white veterans were approximately 15 percent less than the earnings
of comparablenonveterans.(JEL 824)

A central question in the debate over mili- schooling. Regarding the general position of
tary manpower policy is whether veterans veterans, a member of the Twentieth Cen-
are adequately compensated for their service. tury Fund's Task Force on Policies Toward
The political process clearly reflects the de- Veterans concludes that "Within any age
sire to compensate veterans: since World group, veterans have higher incomes, more
War II, millions of veterans have enjoyed education, and lower unemployment rates
benefits for medical care, education and than their nonveteran counterparts."'
training, housing, insurance, and job place- The goal of this paper is to measure the
ment. Recent legislation provides additional long-term labor market consequences of mil-
benefits for veterans of the Vietnam era. Yet, itary service during the Vietnam era. Previ-
academic research has not shown conclu- ous research comparing civilian earnings by
sively that Vietnam (or other) veterans are veteran status may be biased by the fact that
worse off economically than nonveterans. certain types of men are more likely to serve
Many studies find that Vietnam veterans earn in the armed forces than others. For exam-
less than nonveterans, but others find posi- ple, men with relatively few civilian opportu-
tive effects, or effects that vary with age and nities are probably more likely to enlist. Es-
timation strategies that do not control for
differences in civilian earnings potential will
incorrectly attribute lower civilian earnings
of veterans to military service. The research
*Department of Economics, Harvard University, reported here overcomes such statistical
Cambridge, MA 02138. Grateful thanks go to Warren problems by using the Vietnam era draft
Buckler, Cresston Smith, Ada Enis, and Bea Matsui for
their assistance in producing the Social Security data; to
Chester Bowie for his help in producing the SIPP data;
and to Mike Dove for providing DMDC administrative
records. Special thanks also go to David Card and 'The quote is from Michael Taussig (1974, p. 51).
Whitney Newey, from whose instruction and comments Legislation pertaining to veterans benefits is outlined in
I have benefited greatly, and to Alan Krueger and an Veterans Administration (1984) and in other annual
anonymous referee, whose careful reviews of an earlier reports of the Veterans Administration. Studies by
draft led to substantial improvement. Data collection Sherwin Rosen and Paul Taubman (1982), Saul Schwartz
for this project was funded by the Princeton Industrial (1986), and Jon Crane and David Wise (1987) find that
Relations Section. Funds for computation and financial Vietnam veterans earn less than nonveterans. Dennis
support of the author were provided by the Industrial DeTray (1982) and Mark Berger and Barry Hirsch
Relations Section, the Princeton Department of Eco- (1983) find some positive effects for different age and
nomics, the Sloan Foundation, and the Olin Founda- schooling classes, and Veterans Administration (1981a)
tion. researchefs find an overall positive effect.
313
314 THE AMERICAN ECONOMIC REVIEW JUNE 1990

lotteries to set up a naturalexperimentthat Section IV tests the hypothesisthat veter-

randomlyinfluencedwho servedin the mili- ans earn less than nonveteransbecausethey
tary.2 have less civilian labor market experience.
Section I describesthe Social Securityad- Results in this sectionsuggestthat the earn-
ministrativerecords used in the empirical ings loss to white veteransis equivalentto a
work and providesbackgroundon the draft loss of two years of civilian labor market
lotteries.In each lottery,priorityfor induc- experience.Section V reviews some of the
tion was determinedby a RandomSequence potential pitfalls in estimationbased on the
Number (RSN) from 1-365 that was as- draft lottery. Section VI offers conclusions
signed to birthdates in the cohort being and indicatesdirectionsfor futureresearch.
drafted. Men were called for induction by
RSN up to a ceiling determinedby the De- andData
I. Background
fense Department,andonly men with lottery
numbersbelow the ceiling could have been A. National RandomSelection3
drafted. Therefore,men with lottery num-
bers below the ceilingare referredto here as There were five draft lotteriesduringthe
"draft-eligible." Vietnam War period.The 1970 lottery cov-
The empiricalanalysisbeginsin SectionII ered 19- to 26-year-oldmenbornin 1944-50,
with estimatesof the effectof drafteligibility although most of the men drafted in 1970
on earnings.If draft eligibilityis correlated were born in 1950. Otherlotterieswere re-
with veteran status but uncorrelatedwith stricted to 19- and 20-year-olds.The 1971
other variables related to earnings, then lottery coveredmen born in 1951, the 1972
earningsdifferencesby draft-eligibility status lotterycoveredmen bornin 1952,and so on,
can be attributedto militaryservice.In Sec- through1975. However,no one was drafted
tion III, informationon the proportionsof after 1972, and congressionalconscription
draft-eligibleand draft-ineligiblemen who authorityexpiredin July1973.
actually served in the military is used to Draft lottery RSNs were randomly as-
convertestimatesof the effectof draft eligi- signed in a televised drawing held a few
bility into estimatesof the effect of military months before men reachingdraft age were
service. The assumptions underlying this to be called.4Draft-eligibilityceilings-RSN
procedureare those thatjustifyinstrumental 195 in 1970, RSN 125 in 1971, and RSN 95
variablesestimation;in principle,any func- in 1972-were announcedlater in the year,
tion of the RSN providesa legitimateinstru- once Defense Departmentmanpowerneeds
ment for veteranstatus. In the second part were known.As a consequenceof this delay,
of SectionIII, an instrumentalvariablesesti- many men with low numbersvolunteeredfor
mation strategyis developedwhich is more the military to avoid being drafted and to
efficientthan one basedsolely on draft-eligi- improve their terms of service (Angrist
bility status. Results in Section III indicate 1989b). There was even a behavioral re-
that white veterans earn approximately15 sponse to the lottery in enlistmentrates for
percentless than nonveteransas muchas ten the 1953 cohort, althoughno one born in
years after their dischargefrom the military. 1953 was drafted. In the analysis that fol-
lows, the "draft-eligibilityceiling" for men
born in 1953 is set at RSN 95, the highest
lotterynumbercalledin 1972.
2A candid assessment of the problems caused by
nonrandom selection for military service is given by
Crane and Wise (1987), who note they were unable to
use econometric sample selection models to generate
robust estimates of the effects of military service on 3This section draws on Curtis Tarr (1981) and the
civilian earnings. The first researchersto use the lottery Selective Service System (1986).
to solve the selection problem were Norman Hearst, 4Men born from 1944-49 were already of draft age
Tom Newman, and Stephen Hulley (1986), who present when the 1970 lottery was held on December 1, 1969.
lottery-based estimates of delayed effects of military For nonveterans in this group, subsequent liability for
service on mortality. service was determined by 1970 lottery numbers.
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTER Y 315

Only the initialselectionprocesswas based Social Security taxable employment.How-

on RSN order. Subsequentselection from ever, becauseSSA proceduresfor the collec-
the draft-eligible, nondeferred pool was tion of W-2 forms are relativelynew, W-2
based on a number of criteria.The most earningsdata are probablyless reliablethan
important screening criteria were the pre- the FICA data.
inductionphysicalexaminationand a mental The original CWHS data set does not
aptitude test. In 1970, for example,half of contain informationon date of birth. SSA
all registrantsfailed pre-inductionexamina- programmersmatched date of birth vari-
tions and 20 percent of those who passed ables to the CWHSin a specialextractcre-
were eliminatedby physicalinspectionscon- ated for this project.Lotterynumberswere
ducted at induction (SelectiveService Sys- then matchedto dates of birth,using tables
tem, 1971). Of course, the fact that armed published in the 1969-73 SemiannualRe-
forces selection criteriawere ultimatelynot ports of the Directorof SelectiveService.
randomdoes not meanthat the initial prior- The Internal Revenue Servicelimits dis-
ity for inductionwas not randomlyassigned closure of data collected for tax purposes.
by RSN. To adhere to these disclosurerequirements,
The year 1970 was the last time men over the SSA could releaseonly aggregatedata.
the age of 20 were drafted. In principle, The aggregate data set contains sample
nonveteransborn between 1944 and 1949 statistics for cells definedby year of earn-
continued to be at risk of inductionin the ings, year of birth,race,and five consecutive
1970 lottery, but the majorityof men who lottery numbers. Cell statistics include
ended up serving from these cohorts had means, variances, fraction with earnings
alreadyentered the militaryby the time of equal to the taxablemaximum,fractionwith
the 1970lotterydrawing.Veteransbornfrom earningsabove the taxablemaximum,frac-
1944-49 who managedto avoid serviceuntil tion with zero earnings,and numberof ob-
1970 may not constitute a representative servationsin each cell.
sample. Therefore,the analysis here is re-
stricted to men who turned 19 in the year II. The Effectof DraftEligibility
they were at risk of induction.This sample on Earnings
includes men who were born between 1950
and 1953. Figure 1 shows the historyof FICA tax-
able earnings for draft lottery participants
B. Social SecurityEarnings Data born between 1950 and 1953. For each co-
hort there are two lines drawn: one for
Earningsdata used in this studyaredrawn draft-eligiblemen, and one for men with
from the Social Security Administration's lottery numbers that exempted them from
(SSA) Continuous Work History Sample the draft.
(CWHS).The CWHSdata set, describedin The impactof drafteligibilityon the earn-
detail in the Appendix, is a one percent ings profilesis striking.Thereappearsto be
sample drawnfrom all possibleSocial Secu- no differencein earningsuntil the year of
rity numbers.The CWHSincludestwo earn- conscriptionrisk in the draftlottery.Subse-
ings series:the firstcontainsinformationon quently, the earningsof draft-eligiblewhite
the 1964-84 earningsof men in employment men born in 1950-52 fall below the earnings
coveredby FICA (SocialSecurity)up to the of draft-ineligiblewhite men born in 1950-
Social Security taxable maximum. It also 52. The earningsof draft-eligiblenonwhites
includes FICA taxable earningsfrom self- also fall below the earningsof other non-
employment.The secondseries,beginningin
1978, contains total compensationas re-
ported on Internal Revenue Service Form
5Eamings are in 1978 dollars. The deflator used for
W-2, excluding earningsfrom self-employ- all tabulations is the CPI on p. 313 of The Economic
ment. In principle,the W-2 earningsdata are Report of the President (Council of Economic Advisors,
neithercensorednor limitedto earningsfrom 1988).
316 THE AMERICAN ECONOMIC REVIEW JUNE 1990

3000 -

2000-
<.
D .
o 1000- *
0'

U - 2000 -

-3000
- 0.08 -0.04 0 0.04 0.08 0.12 0.16

PROBABILITY
RESIDUAL
Notes: The figure plots the history of FICA taxable earnings for the four cohorts born
1950-53. For each cohort, separate lines are drawn for draft-eligible and draft-ineligible
men. Plotted points show average real (1978) earnings of working men born in 1953,
real earnings + $3000 for men born in 1950, real earnings+ $2000 for men born in 1951,
and real earnings + $1000 for men born in 1952.

FIGURE 1. SOCIALSECURITYEARNINGSPROFILESBY DRAFT-ELIGIBILITY

STATUS

whites, but the gap appearsto narrowand However,the earningsof draft-eligiblewhite

become positive for some nonwhitecohorts men continuedto lag behindthe earningsof
in later years. The fact that earningsdo not draft-ineligiblewhitemen through1984.
differ by draft-eligibilitystatus before the The picture for nonwhites is less clear.
lotteries is a consequenceof the random The earningsof draft-eligiblenonwhitesborn
assignmentof drafteligibility.The only thing in 1950 and 1951 exceed those of draft-
that distinguishes draft-eligiblemen from ineligible nonwhites in some of the later
draft-ineligiblemen is the higher conscrip- years.On the otherhand,for nonwhitesborn
tion risk faced by eligible men after the in 1952, time-series variation in earnings
lottery. differencesby eligibilitystatus is similarto
Figure 2 presentsa magnifiedview of the that of whites. The generalimpressionfor
effect of draft eligibilityon earnings.This the three older cohortsof nonwhitesis that
figureplots the time-seriesof differencesin the earnings of draft-eligiblemen at least
earnings by draft-eligibilitystatus for each had caughtup with the earningsof draft-in-
cohort. As in Figure 1, Figure 2 shows no eligiblemen by 1984.
differenceduringthe yearsbeforethe yearof Earningsof white men born in 1953 do
conscriptionrisk, while in subsequentyears, not appear to differby draft-eligibilitysta-
the earningshistoriesdiverge.Figure2 also tus. The earningsof draft-eligiblenonwhites
shows that the loss of earningsto draft-eligi- born in 1953 generallyexceed the earnings
ble white men was largestduringthe period of nonwhites who were not draft-eligible.
they were most likely to be in the service. Differencesbetweenthe effectof draft eligi-
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 317

uJ

z uA
X LL
LL_

0- I-
0-
Z~~~~~~
____ ___ _ _ _- _ _ _ _ ,- _ _ -_ ,

Z 2
W 0- 4-4 I- 0B 1951

< ui

IJJ

6668 70 7274 76 78 8082 84 z 66 68 70 7274 76 78 808284

YEAR YEAR

rel(1978 dollars.
COHORT BORN 1950
BORN 1951
FIGUR 2T1.L-k 0IFREC INi EANIG BY DRF-LGBLT STATUS
BORN 1952
-- BORN 1953

Notes: The figure plots the difference in FICA taxable earnings by draft-eligibility
status for the four cohorts born 1950-53. Each tick on the vertical axis represents $500
real (1978) dollars.

FIGuRE 2. THE DIFFERENCEIN EARNINGSBy DRAFT-ELIGIBILITYSTATUS

bility on men born in 1953 and the effecton The effect of draft eligibilityon nonwhites'
the three older cohorts might be explained probability of being at the taxable maxi-
by the transitionto an All-VolunteerForce mum, although impreciselymeasured,also
in 1973. Men who volunteerfor the military appears to go in the same directionas the
are probablyless likely than drafteesto suf- effect of draft eligibilityon mean earnings.
fer a careerdisadvantagefrom their service. These resultsare worthnotingbecause,when
Estimatesof the effect of draft eligibility the effectof drafteligibilityon the probabil-
are reportedin Table 1 for both FICA earn- ity of being censoredhas the same sign as
ings and W-2 earnings.Standarderrorsasso- the effect on earnings,estimates tabulated
ciated with the estimates are reported in using censored data tend to underestimate
parentheses.The statisticsin Table 1 show the true effect.6
that the loss in FICA earningsto draft-eli-
gible white men is sometimes statistically
significantand amounts to 2-3 percent of
earnings.EstimatedW-2 earningslosses are 6The effect of censoringon estimatedtreatmentef-
similar,but tend to be largerand morevari- fects is discussedin the appendix.Angrist(1989c)also
able than the estimatedlosses in FICA earn- reportsestimatesof the effectof drafteligibilityon the
probabilityof havingno recordedearnings.Thesetabu-
ings. In contrast,differencesin earningsby lationsindicatethatdraft-eligiblewhitesweresomewhat
draft-eligibilitystatus for nonwhites rarely morelikelyto havehad FICAearningsduringthe years
exceed theirstandarderrors. in whichtheywerein the service,and thatdraft-eligible
Elsewhere(Angrist 1989c), I have shown nonwhitesare more likely to have had no earningsin
recent years. There is no statisticallysignificantevi-
that draft-eligiblewhite men are less likely dence for either race, however,of any lastingeffect of
to have earnings above the FICA taxable draft eligibilityon the probabilityof havingzero earn-
maximum than draft-ineligiblewhite men. ings.
318 THE A MERICAN ECONOMIC REVIEW JUNE 1990

TABLE 1-DRAFT-ELIGIBILITY TREATMENT EFFECTS FOR EARNINGS

Whites

FICA Taxable Earnings Total W-2 Compensation

Year 1950 1951 1952 1953 1950 1951 1052 1953
66 -21.8
(14.9)
67 -8.0 13.1
(18.2) (16.4)
68 -14.9 12.3 -8.9
(24.2) (19.5) (19.2)
69 -2.0 18.7 11.4 -4.0
(34.5) (26.4) (22.7) (18.3)
70 - 233.8 -44.8 -5.0 32.9
(39.7) (36.7) (29.3) (24.2)
71 - 325.9 - 298.2 - 29.4 27.6
(46.6) (41.7) (40.2) (30.3)
72 - 203.5 -197.4 -261.6 2.1
(55.4) (51.1) (46.8) (42.9)
73 - 226.6 - 228.8 - 357.7 - 56.5
(67.8) (61.6) (56.2) (54.8)
74 - 243.0 -155.4 -402.7 -15.0
(81.4) (75.3) (68.3) (68.1)
75 - 295.2 - 99.2 - 304.5 - 28.3
(94.4) (89.7) (85.0) (79.6)
76 - 314.2 - 86.8 - 370.7 -145.5
(106.6) (102.9) (98.2) (93.0)
77 - 262.6 - 274.2 - 396.9 - 85.5
(117.9) (112.2) (111.1) (107.1)
78 - 205.3 - 203.8 -467.1 - 65.3 1,059.3 233.2 175.3 -1,974.5
(132.7) (127.0) (127.3) (123.1) (2,159.3) (1,609.4) (1,567.9) (912.1)
79 - 263.6 - 60.5 - 236.8 89.2 -1,588.7 523.6 - 580.8 - 557.9
(160.5) (152.3) (153.9) (148.7) (1,575.6) (1,590.5) (736.7) (750.1)
80 - 339.1 - 267.9 - 312.1 - 93.8 -1,028.1 85.6 - 581.3 -428.7
(183.2) (175.3) (178.2) (170.7) (756.8) (599.8) (309.1) (341.5)
81 - 435.8 - 358.3 - 342.8 34.3 - 589.6 - 71.6 -440.5 - 109.5
(210.5) (203.6) (206.8) (199.0) (299.4) (423.4) (265.0) (245.2)
82 - 320.2 -117.3 -235.1 29.4 - 305.5 - 72.7 - 514.7 18.7
(235.8) (229.1) (232.3) (222.6) (345.4) (372.1) (296.5) (281.9)
83 - 349.5 - 314.0 -437.7 - 96.3 - 512.9 - 896.5 - 915.7 30.1
(261.6) (253.2) (257.5) (248.7) (441.2) (426.3) (395.2) (318.1)
84 - 484.3 - 398.4 -436.0 - 228.6 -1,143.3 - 809.1 - 767.2 - 164.2
(286.8) (279.2) (281.9) (272.2) (492.2) (380.9) (376.0) (366.0)

III. The Effectof MilitaryService where f,B is a cohort effect, at iS a period

on Earnings effect common to all cohorts, and uit is a
residual. The coefficient a is the effect of
A. Estimates Using Draft Eligibility military service on civilian earnings. If si is
correlated with the unobserved components
Estimates of the effects of military service of the earnings equation, then a will not be
are based on a simple linear model for earn- consistently estimated by Ordinary Least
ings. Denote the earnings of man i in cohort Squares (OLS). For example, correlation be-
c at time t by Yd, and let si be an indicator tween si and u,, may arise because the armed
of veteran status. Then we may write forces' eligibility criteria are correlated with
earnings, but not accounted for by the
(1) ycti==AC+ at + sia + uit econometrician, or because veterans are
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 319

TABLE1-CONTINUED

Nonwhites

FICA Taxable Earnings Total W-2 Compensation

Year 1950 1951 1952 1953 1950 1951 1952 1953

66 -11.8
(27.6)
67 12.9 -4.0
(34.2) (30.6)
68 - 29.5 -6.2 -12.0
(44.5) (37.3) (35.0)
69 -5.1 67.8 3.4 -42.4
(66.8) (53.4) (43.4) (36.4)
70 - 99.8 62.2 24.7 -9.0
(78.5) (75.7) (62.2) (44.9)
71 -164.8 - 144.3 -25.0 18.2
(92.7) (86.4) (85.1) (60.7)
72 -188.8 -156.7 -208.2 60.4
(113.6) (105.7) (104.2) (92.8)
73 -85.7 -134.8 -175.6 115.5
(137.7) (127.0) (129.0) (119.4)
74 -179.3 - 96.7 -181.4 216.5
(165.0) (160.1) (155.6) (145.1)
75 -190.3 - 236.1 -183.7 111.6
(189.3) (186.8) (185.8) (166.9)

76 -105.3 - 333.7 - 308.9 -46.4

(214.7) (215.4) (216.5) (199.3)
77 112.4 - 206.8 -251.1 153.5
(238.5) (240.4) (248.5) (233.5)
78 163.6 -108.6 -424.9 381.9 -1,145.0 2,978.2 -4,676.2 -482.7
(272.6) (269.2) (279.4) (275.7) (2,395.6) (2,869.6) (1,393.1) (2,206.0)
79 187.0 -210.3 -391.7 312.0 4,005.4 1,545.0 - 494.7 -1,043.3
(317.2) (323.0) (324.8) (326.3) (2,721.2) (2,191.1) (2,683.8) (1,660.2)
80 203.2 4.8 -212.6 344.0 790.2 376.4 - 292.7 288.6
(363.1) (368.4) (372.5) (370.3) (648.1) (533.6) (440.9) (416.4)
81 534.5 313.2 - 305.8 717.8 802.5 415.9 - 272.3 784.4
(413.5) (419.1) (429.1) (433.7) (524.6) (745.1) (492.8) (503.1)
82 285.1 175.4 - 262.5 810.4 326.0 - 244.3 -160.2 675.1
(461.2) (471.6) (476.7) (486.3) (608.9) (647.8) (590.0) (564.1)
83 96.0 419.5 -177.3 543.6 315.4 254.3 - 53.6 462.3
(512.6) (538.1) (531.5) (523.2) (720.0) (767.5) (643.4) (638.9)
84 - 76.8 -223.1 -123.3 641.3 - 287.4 -718.6 - 288.0 827.3
(548.2) (562.8) (568.5) (568.2) (804.0) (771.5) (721.0) (716.8)

Notes: Standard errors in parentheses.

The table shows the difference in earnings by lottery-determined draft-eligibility status.
Eligibility ceilings are RSN 195 for men born in 1950, RSN 125 for men born in 1951,
and RSN 95 for men born in 1952 and 1953.
Earnings data are from the Social Security Administration CWHS, described in the text
and the Appendix.

self-selected on the basis of unobserved onal to the error term, uir. For example, one
characteristics. such instrumentis a dummy variable, di,
The draft lottery facilitatesestimationof that equals one if the ith individual was
(1) because functionsof randomlyassigned draft eligible. Suppose that attentionis re-
lottery numbers provide instrumentalvari- stricted to a single cohort. Then, use of d1
ables that are correlatedwith si, but orthog- and a constant as instrumentalvariables
320 THE AMERICAN ECONOMIC REVIEW JUNE 1990

leads to the following estimator for a., eligibility on veteran status for whites born
1950-52 range from 0.10 to 0.16. Thus, a
(2) ?&= (ye - -n)/( pe _
pn) rule of thumb for conversion of draft-eligi-
bility treatment effects into estimates of the
where is the proportion of the - cohort effects of military service is to multiply by
A

actually entering the military, is mean 1/0.15 = 6 2/3.

earnings, and superscript e and superscript Wald estimates of the effect of military
n denote the draft-eligible and draft-ineligi- service for selected cohorts and years are
ble samples. Note that the numerator of (2) presented in Table 3. The sample is re-
consists of estimates of the effect of draft stricted to the subset of whites born 1950-52
eligibility plotted in Figure 2. because the results in Table 2 suggest that
Intuitively, equation (2) simply adjusts this is the group for whom draft eligibility is
earnings differences by draft-eligibility status most likely to be a useful instrument. Earn-
for the fact that not all draft-eligible men ings variables are for 1981-84 because the
actually served in the military, while some impact of military service in these years rep-
men who were not draft eligible voluntarily resents a long-term effect. Furthermore, as a
enlisted for service. The justification for esti- practical matter, both FICA and W-2 earn-
mation of the effects of military service in ings data are likely to be more reliable in
this manner is clear: it is assumed that noth- recent years-the FICA data because of in-
ing other than differences in the probability creased employment coverage and the W-2
of being a veteran is responsible for differ- data because of improvements in data collec-
ences in earnings by draft-eligibility status. tion procedures.
This formula may also be recognized as an Table 3 reports three sets of estimated
application of Abraham Wald's (1940) draft-eligibility effects for use in the numera-
grouping method, where the data have been tor of the Wald estimator. Column (1) pre-
grouped by draft-eligibility status. Applica- sents estimates for FICA earnings and col-
tions of this formula will therefore be re- umn (3) presents estimates for W-2 earnings;
ferred to here as "Wald estimates." the figures in both of these columns are
In addition to draft-eligibility treatment copied directly from Table 1. In addition,
effects, implementation of the Wald estima- column (2) reports estimates for an earnings
tor requiresestimatesof pe and An. These series constructed by applying a simple
estimates are tabulated from a special ver- non-parametric correction for censoring to
sion of the 1984 Survey of Income and Pro- the FICA earnings data. The correction pro-
gram Participation (SIPP). The SIPP data cedure is described in detail in Section 6 of
used here were matched to an indicator of the Appendix. Briefly, data are adjusted for
draft-eligibility status from information on censoring by using the fraction with recorded
birthdates included in the Census Bureau's earnings at the taxable maximum, combined
in-house version of the SIPP file. Additional with mean earnings above the taxable maxi-
details on the SIPP data are provided in mum estimated from Current Population
Section 7 of the Appendix. Surveys, to estimate population mean earn-
In the upper panel of Table 2, the columns ings from censored mean earnings. Note that
labeled pe and An show probabilitiesof the effects estimated using the adjusted data
veteran status tabulated using the SIPP. Be- are usually bracketed by the effects esti-
cause of the small number of observations mated using the unadjusted FICA and W-2
available for single-cohort statistics, each data. Therefore, only Wald estimates con-
SIPP estimate is actually the average for structed from the adjusted data are reported
three consecutive cohorts. For example, SIPP in the table.
estimates assigned to men born in 1951 are Wald estimates for adjusted FICA earn-
based on data for men born in 1950, 1951, ings, reported in column (5) of Table 3,
and 1952. The last column of Table 2, la- indicate that white veterans suffered an an-
beled pe _
Apn, shows the differencein the nual earnings loss of roughly $2000 constant
probability of military service by draft-eligi- (1978) dollars of $3,500 current dollars. This
bility status. Estimates of the effect of draft is approximately 15 percent of annual W-2
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 321

TABLE2-VETERAN STATUSAND DRAFT ELIGIBILITY

Whites

Data Set Cohort Sample P(Veteran) pe pn pe - -n

SIPP(84)a 1950 351 0.2673 0.3527 0.1933 0.1594

(0.0140) (0.0325) (0.0233) (0.0400)
1951 359 0.1973 0.2831 0.1468 0.1362
(0.0127) (0.0390) (0.0180) (0.0429)
1952 336 0.1554 0.2310 0.1257 0.1053
(0.0114) (0.0473) (0.0146) (0.0495)
1953 390 0.1298 0.1581 0.1153 0.0427
(0.0106) (0.0339) (0.0152) (0.0372)
DMDC/CWHSb 1950 16119 0.0633 0.0936 0.0279 0.0657
(0.0019) (0.0032) (0.0019) (0.0037)
1951 16768 0.1176 0.2071 0.0708 0.1362
(0.0025) (0.0053) (0.0024) (0.0059)
1952 17703 0.1515 0.2683 0.1102 0.1581
(0.0027) (0.0065) (0.0027) (0.0071)
1953 17749 0.1343 0.1548 0.1268 0.0280
(0.0026) (0.0053) (0.0029) (0.0060)

Nonwhites

Data Set Cohort Sample P(Veteran) pe pn pe_ ph

SIPP (84)a 1950 70 0.1625 0.1957 0.1354 0.0603

(0.0292) (0.0699) (0.0491) (0.0854)
1951 63 0.1703 0.2014 0.1514 0.0500
(0.0292) (0.0827) (0.0448) (0.0940)
1952 52 0.1332 0.1449 0.1287 0.0161
(0.0275) (0.1040) (0.0373) (0.1105)
1953 55 0.1749 0.1980 0.1612 0.0367
(0.0305) (0.0865) (0.0470) (0.0984)
DMDC/CWHSb 1950 5447 0.0417 0.0548 0.0271 0.0276
(0.0027) (0.0042) (0.0032) (0.0053)
1951 5258 0.0794 0.1173 0.0599 0.0574
(0.0037) (0.0076) (0.0040) (0.0086)
1952 5493 0.0953 0.1439 0.0794 0.0644
(0.0040) (0.0095) (0.0042) (0.0104)
1953 5303 0.0925 0.0984 0.0904 0.0080
(0.0040) (0.0079) (0.0046) (0.0092)

Notes: Standard errors in parentheses. pe is the probability of being a veteran

conditional on being draft eligible; p" is the probability of being a veteran conditional
on being ineligible.
aWave I, Panel I of the 1984 Survey of Income and Program Participation.
Probabilities are for service in the Vietnam era. Estimates are weighted by the SIPP
sampling weight and smoothed over 3 cohorts.
bDefense Manpower Data Center Administrative Records' information on acces-
sions, from 1970-73, combined with information on cohort size from the Social
Security Administration Continuous Work History Sample.

compensation for white men between 1981 statistically significant at conventional

and 1984. The similarity of coefficient esti- levels.7
mates across cohorts and years suggests that
the Wald estimates provide a robust measure
of the impact of military service. Taken indi- 7The'asymptoticstandarderror of the Wald esti-
vidually, however, few of the estimates are mates is derived from the limiting distributionof
322 THE AMERICAN ECONOMIC RE VIEW JUNE 1990

TABLE3-WALD ESTIMATES

Draft-Eligibility Effects in Current $

FICA Adjusted FICA Total W-2 Service Effect
Earnings Earnings Earnings pe _ pn in 1978$
Cohort Year (1) (2) (3) (4) (5)

1950 1981 - 435.8 -487.8 - 589.6 0.159 -2,195.8

(210.5) (237.6) (299.4) (0.040) (1,069.5)
1982 - 320.2 -396.1 - 305.5 -1,678.3
(235.8) (281.7) (345.4) (1,193.6)
1983 - 349.5 -450.1 -512.9 -1,795.6
(261.6) (302.0) (441.2) (1,204.8)
1984 -484.3 - 638.7 -1,143.3 -2,517.7
(286.8) (336.5) (492.2) (1,326.5)
1951 1981 - 358.3 -428.7 - 71.6 0.136 -2,261.3
(203.6) (224.5) (423.4) (0.043) (1,184.2)
1982 -117.3 - 278.5 - 72.7 -1,386.6
(229.1) (264.1) (372.1) (1,312.1)
1983 -314.0 -452.2 - 896.5 -2,181.8
(253.2) (289.2) (426.3) (1,395.3)
1984 - 398.4 - 573.3 - 809.1 - 2,647.9
(279.2) (331.1) (380.9) (1,529.2)
1952 1981 - 342.8 - 392.6 -440.5 0.105 - 2,502.3
(206.8) (228.6) (265.0) (0.050) (1,556.7)
1982 -235.1 - 255.2 - 514.7 -1,626.5
(232.3) (264.5) (296.5) (1,685.8)
1983 -437.7 - 500.0 - 915.7 - 3,103.5
(257.5) (294.7) (395.2) (1,829.2)
1984 -436.0 - 560.0 - 767.2 - 3,323.8
(281.9) (330.1) (376.0) (1,959.3)

Notes: Standard errors in parentheses.

Columns (1) and (3) are taken from Table 1.
Column (2) reports draft-eligibility treatment effects on earnings adjusted for censoring
at the FICA taxable maximum. The adjustment procedure is described in the Ap-
pendix. Column (4) reports SIPP estimates of the effect of draft eligibility on veteran
status, taken from Table 2. Column (5) reports estimates of the effect of military service
on civilian earnings is implied by columns (2) and (4).

B. EfficientInstrumental five consecutive lottery numbers. Consider

VariablesEstimates the following grouped version of (1), where
Yctjis mean earnings for members of cohort
The Wald estimator is based solely on c at time t with lottery numbers in group j,
earnings differences by draft-eligibility sta- and is the fraction of cohort c with
tus. A more efficient estimator exploits all lottery numbers in group j who served:
the information on RSNs in the aggregate -
data by fitting earnings model (1) to obser- (3) Yctj= 3c+ St+ Pa + uCtj
vations on mean earnings for each group of
Intuitively, estimation of (3) simply general-
izes Wald's method to grouped data with
rn e -
yn)l(
e - -"). The standard error is therefore more than two groups.
equal to 1/( e - n) times the standard error of the Generalized Least Squares (GLS) esti-
numerator because the numerator has a nondegenerate mates of (3) may easily be shown to have
limiting distribution, while (pe - p) converges to a
constant. The same standard error formulas arise from
an instrumental variables interpretation
application of conventional Instrumental Variables for- (Angrist, 1988). In this case, the instrument
mulas. set includes dummy variables that indicate
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 323

groups of five consecutive lottery numbers at risk, an estimate of total cohort size is
for each race, cohort, and year of earnings. simply 100 times the CWHS cohort size. For
There are 73 dummy variables for a particu- example, to estimate the probability of being
lar race, cohort, and year; the first indicates a veteran conditional on being draft eligible,
men with lottery numbers 1-5 and the 73rd the number of draft-eligible men in the
indicates men with lottery numbers 360-365. DMDC data is divided by 100 times the
Furthermore, the quadratic form minimized number of men in the CWHS with lottery
by the GLS estimator is an overidentifica- numbers below the induction ceiling. Stan-
tion test statistic associated with the use of dard errors for these estimates are computed
dummy variables as instruments. This statis- by applying the usual formula for a binomial
tic tests the exclusion of lottery number proportion.9
group dummies from equation (1). It may For comparison with the SIPP estimates,
also be viewed as a measure of the goodness- DMDC/CWHS estimates of pe and pn are
of-fit of the cell means to equation (1).8 reported in the lower panel of Table 2. These
In principle, implementation of the esti- figures show that, with the exception of the
mation strategy based on (3) is straightfor- 1950 cohort, the SIPP and DMDC/CWHS
ward-the estimates are simply coefficients procedures give reasonably similar estimates
from GLS regressions of mean Social Secu- of _- p". Inaccuracy of the DMDC/
rity earnings on estimates of fi.. The SIPP CWHS estimates for 1950 is a consequence
sample is too small to allow accurate estima- of the fact that DMDC administrative
tion of a full set of P , however. Thus, a records are unavailable before July 1970.
second set of probabilities was estimated Therefore, despite the limitations of the SIPP
from a combination of Defense Manpower data, the SIPP must be used to construct
Data Center (DMDC) administrativerecords probabilities for the 1950 cohort. The SIPP
and CWHS data on cohort size. Detailed sample is too small to allow estimation of a
descriptions of the DMDC administrative complete set of p for all lottery number
records may be found in Angrist (1989b). cells in 1950. Consequently, SIPP estimates
Briefly, the DMDC data show the total num- for 1950 are computed for only two cells,
ber of new entrants to the military by race, defined by draft-eligibility status, and CWHS
cohort, and lottery number from July 1970 earnings data for men born in 1950 are also
through December 1973. grouped by draft eligibility. Thus, for each
DMDC and CWHS administrativerecords race and year, the sample used to estimate
are used to estimate 1bj by first counting the equation (3) includes 73 cell means for each
number of entrants to the military by race, of the three cohorts born from 1951-53, plus
cohort, and lottery number interval. These two cell means for the 1950 cohort.
numbers are the numerator of the c . Esti- A graphical version of equation (3) is de-
mates of overall cohort size, to be used in picted in Figure 3, which shows the relation-
the denominator of Pj3, are derived from the ship between probabilities of veteran status
CWHS. Recall that the CWHS is a one (PCJ)and mean W-2 compensationin 1978
percent sample, so that if the CWHS sam- dollars (jctJ) between 1981 and 1984. Plot-
pling frame is identified with the population ted in the figure are the average (over four
years of earnings) residuals from a regression

8
A general reference on overidentification testing is
Whitney Newey (1985). See also Angrist (1988), where 4The formula used is j '(1 - ')/n,], where p is the
GLS on grouped data is shown to be the minimum estimated proportion of servers and nC is the number in
variance linear combination of all the Wald estimators the CWHS cohort. For example, 5749 draft eligible
that can be computed from any division of grouped white men in the CWHS were born in 1951, and DMDC
observations into linearly independent pairs. The overi- administrative records show that 119,062 draft-eligible
dentification test statistic for dummy variable instru- white men born in 1951 served between July 1970 and
ments is also shown to be the same as the Wald statistic December 1973. pe is therefore 119,062/574,900 = 0.21,
for equality of alternative Wald estimates. with estimated variance equal to (0.21 * 0.79)/5749.
324 THE AMERICAN ECONOMIC REVIEW JUNE 1990

a
(I)~~~~~~~~~~~U
16000 a 14500

-J 14000 0g 12500 -

a 12000 0500

a) i2(-i) 5 *

z #,-/ 500-

CO 1950 _ 6500.
w
4000- / i 2500 'A

2000 ,f, ~ 11111I1III

I

66r68e70
72i74 76 78
80t82h84 66 68 70 72 74 7678 8082 84
YEAR Z YEAR
1953 _- _ _ _ -_ /
DRAFT
ELIGIBLE INELIGIBLE
COHORT1950 --.

1951 1990
1952
1953

Notes: The figure plots mean W-2 compensation in 1981-84 against probabilities of
veteran status by cohort and groups of five consecutive lottery numbers for white men
born 1950-53. Plotted points consist of the average residuals (over four year-s of
earnings) from regressions on period and cohort effects. The slope of the least squares
regression line drawn through the points is - 2,384 with a standard error of 778, and is
an estimate of a in the equation

Yctj==fc + at + P a + iu4t

FIGURE 3. EARNINGS AND THEPROBABILITYOF VETERANSTATUSBY LOTTERYNUMBER

of earningsand probabilitieson period and ties of veteranstatus.This heteroscedasticity

cohort effects.10Thus, the slope of the ordi- also appearsin comparisons(not shownhere)
nary least squares regression line drawn of cell variancesby draft-eligibilitystatus;
through the points correspondsto an esti- draft-eligiblemen have somewhatless vari-
mate of a. This slope is equal to -2,384 able earnings.
dollars,with a standarderrorof 778 dollars. As pointed out earlier,estimationof (3) is
An interesting feature of the figure is the the same as instrumentalvariablesestima-
apparent heteroscedasticityof the earnings tion of (1) using dummyvariablesas instru-
residuals.Dispersion around the regression ments.However,inferencefor the case where
line is reducedfor cells with high probabili- the estimation strategyis implementedby
regressing CWHS mean earnings on
DMDC/CWHS or SIPP probabilities is
complicatedby the use of datafrommultiple
l?There are 221 points plotted in the figure: 4 years samples.Assumingthat the samplesused to
of earnings times 3 cohorts with 73 cells plus 4 years of
earnings times one cohort with 2 cells (men born in calculatemean earningsand the sampleused
1950) = 884, divided by four to compute the average to calculateA3are independent,the optimal
over years. Two-SampleInstrumentalVariables(TSIV)
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 325

- may be
estimator has a simple form that able maximum using the procedure de-
brieflydescribedas follows.Let denotethe scribedin section6 of the Appendix.
vector of Jctj' p denote the vector of The results in Table 4 show that white
and iu(O)denote the vectorof uct., where 0 veteransborn from 1950-52 sufferedan an-
in parenthesesrepresentsthe dependenceof nual earnings loss of between $1,500 and
residualson the parametervector. Also, let $2,100 constant(1978)dollars.These results
V( ) denote the covariancematrix of the generally are similar in magnitudeto the
argument.Then the optimalTSIV estimator Wald estimatesreportedin Table 3. Also, as
chooses 0 to minimize in Table 3, regressionestimatesfor adjusted
FICA earningstend to be bracketedby the
results for unadjustedFICA and W-2 earn-
@)'[V( y)+ a2V(fp)] u(e) ings. Although many of the estimates for
individualcohortsin Model 1 are not signif-
which is also the GLS minjimandfor (3).11 icant, the combinedestimatesfor whites in
The minimizedvalue of m(O) is an overi- Model 2 are substantiallylargerthan twice
dentificationtest statisticfor the validityof their standarderrors.In contrast,resultsfor
dummyvariablesas instruments.If some of nonwhitesshow no evidenceof a statistically
these dummy variablesare correlatedwith significantearningsloss to veterans.
the regressionerror, then m(O) should be The overidentificationtest statistics re-
largerelativeto a chi-squaredistributionwith ported in Table 4 take on values less than
degrees of freedom equal to the difference their degreesof freedom,suggestingthat the
between the numberof instrumentsand the residualsin equation(1) are not correlated
numberof estimatedparameters. with lottery-basedinstruments.12It should
Table 4 presents two sets of TSIV esti- be noted,however,thatlow valuesof the test
mates of equation(3) for 1981-84 earnings statistics may indicate low power in a test
in 1978 dollars.Model 1 allowsthe effect of with so many degrees of freedom.On the
veteranstatus on earningsto varyby cohort, other hand, without a particularalternative
while Model 2 restrictsestimatedserviceef- hypothesisin mind, it seems naturalto re-
fects to be the same across cohorts. Note port the omnibusgoodness-of-fittest.
that, as in Table 3, the heading "adjusted Subtractingthe test statisticfor Model 1
FICA earnings" refers to FICA-taxable from the test statistic for Model 2 gives a
earnings adjusted for censoringat the tax- chi-squaretest for the restrictionof equal
treatmenteffects across cohorts.The set of
restrictionsimposed by equal treatmentef-
fects has three degreesof freedom.None of
"1See Angrist (1989d) for details. Estimates of a for the chi-square statistics for Model 2 are
use in the formula for (D were computed by weighted larger than the correspondingstatistics for
least squares using the inverse of the sampling variance Model 1 by as muchas three,indicatingthat
-
of the as weights. Estimates of V(y) and V(p) are
discussed in Section 5 of the Appendix. Note that the
TSIV estimator may also be motivated as an application
of Optimal Minimum Distance (OMD) techniques such
as those described by Gary Chamberlain (1982). Ignor-
ing period and cohort effects, OMD estimates for the 12Degrees of freedom for the overidentification tests
current problem are tabulated by choosing a and p to are calculated as follows. For each race, the data consist
minimize of four years of earnings for three cohorts with 73
lottery number cells each. The fourth cohort, men born
q(a,P) = in 1950, has four years of earnings with 2 lottery num-
ber cells each. This gives a total of 884 cells or, equiva-
lently, 884 categorical instruments. Model 1 includes 4
[y - PC' V(y) 0 1 y_pal
cohort dummies, 3 year dummies, and 4 treatment
LP-P p Lo V(p')] L-P p effects. 884 minus 11 parameters gives 873 degrees of
freedom. Model 2 has 3 fewer parameters than model 1
By concentrating out the estimate of p, it is easy to and consequently the chi-square statistic for model 2
show that q(a, p) = m(0). has 876.degrees of freedom.
326 THE AMERICAN ECONOMIC REVIEW JUNE 1990

VARIABLESESTIMATES
TABLE 4-TWO-STAGE INSTRUMENTAL

Whites

FICA Adjusted FICA Total W-2

Cohort Taxable Earnings Earnings Compensation

Model 1
1950 -1709.2 - 2093.7 -1895.0
(946.8) (1108.8) (1333.1)
1951 -1457.1 - 1983.7 - 2431.4
(959.3) (1036.1) (1152.1)
1952 -1724.0 - 1943.0 - 2058.7
(863.1) (927.2) (1001.9)
1953 1223.8 900.7 -488.6
(3232.1) (3505.3) (3936.0)
X2(873) 578.3 630.3 569.5

Model 2
1950-53 -1562.9 -1920.4 - 2094.5
(521.8) (575.9) (646.3)
X2(876) 579.1 631.0 569.7

Nonwhites

FICA Adjusted FICA Total W-2

Cohort Taxable Earnings Earnings Compensation

Model 1
1950 3893.7 3891.9 5711.8
(5358.5) (6244.5) (7206.0)
1951 - 891.3 - 333.4 2609.0
(4397.1) (4664.2) (4894.6)
1952 - 3182.9 - 3457.7 - 3068.0
(3997.4) (4195.2) (4229.2)
1953 - 5928.3 - 8571.4 - 6325.8
(10296.3) (10697.1) (11410.6)
X2(873) 616.7 681.7 693.6

Model 2
1950-53 -643.3 - 999.7 366.7
(2407.5) (2602.5) (2734.2)
X2(876) 618.4 683.4 695.6

Notes: Standard errors in parentheses.

The table shows estimates of the effect of military service on average 1981-84 earnings
in 1978 dollars. The estimation method is optimally weighted Two-Sample Instrumental
Variables, described in the text. FICA and W-2 earnings are from the Social Security
CWHS. The adjusted FICA series is described in the Appendix.

the estimated treatment effects are not statis- is a poor substitute for lost civilian labor
tically different across cohorts. marketexperience.As evidencefor this hy-
pothesis, Grilichesand Mason (1972) show
IV. MilitaryServiceandLoss of Labor that the longer they werein the military,the
MarketExperience less veteransearn relativeto nonveterans.A
test of the loss-of-experiencehypothesis is
The simplest explanation for a veteran developed here using the functional form
earnings penalty is that military experience commonlyemployedin empiricalstudies of
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTER Y 327

human capital. The earningsfunctionmoti- schooling is independentof veteran status.

vated by the theory of human capital is Then using dummyinstrumentalvariablesto
loglinear in years of schooling and log- group equation (4b) by cohort, year, and
quadraticin years of labor market exper- lottery number, average log earnings for
ience. This functional form puts testable members of cohort c at time t in lottery-
restrictions on the time-seriesof earnings numbercell j are
differencesby veteranstatus.'3
Adapting the human capital earnings
function for the problemat hand, the earn- (5) Yctj= 8t + floxot + YXct
ings of individual i in cohort c at time t
may be written - [,3l - y12] A-

(4a) Yca= 8t + wi,o + /Oxic, - [2-yl ] Cxct ) + Uct ,

+ yx 2 +U, where 8Anow includes the period mean of

'4
it.
where yci now denotes log earnings,8, is a A generalizationof model (5) allows the
time-varyingintercept, BoL, y, and So are linear experienceterm to vary with veteran
parameters.xict is the civilianlabor market statusby letting the slope for individuali be
experienceof man i in cohort c at time t, Pi3= f8 + Pls,. In this case, mean cell earn-
taken here to be equal to [t - (c + 18) - w, - ings are characterizedby
sil], where wi is the deviationof i's school-
2
ing from the samplemeanlevel of schooling, (6) Yctj= 8a+ f3OXc + yx
and 1 is years of military experiencefor
veterans.As before, si is a dummyvariable - [30l-yl2+I3ll]pC
that indicatesmilitaryservice.
To focus on parametersthat can be esti- - - +
mated using Social Securitydata, equation [2yl fil]( P-xCt) iCt

(4) is rewrittenas
Models (5) and (6) both have the follow-
ing reduced form in terms of unrestricted
(4b) Ycti= Sit regressioncoefficients:

+/30(X -sil) + y(XCt- Sil)2 (7) Yctj= a+ /oxct + yx t

- (2ywixct -2ylwisi) + ui, + 7TPcj + 72( Pc*jxcx) + uct

where xct = t - (c + 18) and 8t = St+ W(8O

Note that the reducedform veteraneffectis
- 1h)+ yw7. Now, as in the previousanaly-
cty 1 + 'T2Xct Thus, these modelsparame-
terize a time-varyingveteran status coeffi-
sis, assume that schoolingdoes not vary by cient as a linear function of labor market
lottery number.Assume also that schooling
is independent of cohort-this seems rea-
sonable for the small cohort range consid-
ered here. Finally,to focus solely on the loss Averaging over c, t, and j eliminates (d yw,x, -
aywisi) because wi is orthogonal to x,, and s, by
of labor market experience, assume that assumption. Using the fact that E(s,jc, j) = E(s2lc, j)
= pcj, (4) simplifies to (5). Note that (5), (6), and (7) are
not estimable if allowance need be made for cohort as
well as period effects. Qualitatively similar estimates to
those reported below were obtained when t was
13See Mincer (1974) for theoreticaljustification of the dropped in favor of cohort effects, although the good-
human capital earnings function. A recent survey of the ness-of-fit test leads to rejection of models without
human capital literature is Willis (1986). period effects.
328 THE AMERICAN ECONOMIC REVIEW JUNE 1990

TABLE5-EARNINGS-FUNCTION MODELSFOR THE VETERANEFFECT,

WHITESBORN 1950-52

Model (6): Model (7):

Model (5): Loss of Experience, Unrestricted
Loss of Experience Reduced Growth Rate Reduced Form
Parameter (1) (2) (3)

Experience Slope, ,80 0.1022 0.1016 0.1016

(0.007) (0.007) (0.007)
Experience Squared, y - 0.0027 - 0.0025 - 0.0025
(0.0003) (0.0003) (0.0003)
Veteran Effect on Slope, 3,8 - 0.0035
(0.0023)
Veteran Loss of Experience, / 2.08 1.84
(0.38) (0.43)
s7T= - [fol- y12 + 3l] -0.189
(0.052)
,g2 = - [2yl - l] 0.006
(0.004)
Age at Which Reduced Form 50.1
Veteran Effect (7r1+ 7r2xc,) = 0 (15.9)
X2(dof) 1.41(1) 813.57(1247)

Notes: Standard errors in parentheses.

The table reports estimates of experience-earnings profiles that include parameters for the effect of veteran status.
Estimates are of equations (5), (6) and (7) in the text. The estimating sample includes FICA taxable earnings from
1975-84 for men born 1950, 1976-84 earnings for men born in 1951, and 1977-84 earnings for men born 1952. The
estimation method is optimally weighted Two-Sample Instrumental Variables for a nonlinear model in columns (1)
and (2), and for a linear model in column (3).

experience.Excludingthe time-varyinginter- earningsfunctionsare commonlyfit in logs,

cept, model (6) containsfour structuralpa- the dependentvariableis takento be the log
rameters;80, 81, y, and 1, and four reduced of mean earnings for each cell. The log of
form parameters;fLo,y, 7T,and 7T2.Model the mean is not the same as the mean of the
(5) imposes one testable restrictionon the log, but the CWHSdata set does not contain
reducedformby settingfB= 0.15 the mean of log earnings.If earnings are
Table 5 showsresultsfromnonlinearGLS approximatelylognormallydistributed,use
estimation of (5) and (6), and results from of the log of the meanwill providea reason-
LinearGLS estimationof (7), using data on able approximation.In practice,estimatesof
the real FICA earningsof white men born
from 1950 to 1952. The weighting matrix
used in estimationwas derivedin a manner
earnings by cohort c and suppose that each time-series
similarto the weightingmatrixused to con- is of length T. Then the second term in the optimal
struct the estimates in Table 4.16 Because weighting matrix has the following block corresponding
to the time series of earnings for lottery number cell j
of cohort c:
'5A third model is derived by letting both y and ,B
vary with veteran status. This model leads to a reduced (l/ncj) fc
form similar to (7), with the only modification being the
addition of a linear term of the form T3( cj * x2,). In X I eTeT' ] [Pcj Pcj)]}rf C,
the empirical work, however, no evidence was found
that such a term belongs in the earnings function re- where eT is a vector of T l's. In practice, lc is
duced form. replaced by weighted least squares estimates (weights
16The only modification arises from the fact that for are the inverse sampling variances of Yctj) of the re-
equations (5)-(7), reduced form treatment effects ap- duced form equation, (7). Estimates of models for the
pearing in the weighting matrix are time varying. Let log of earnings replace V(5) with delta-method esti-
Fc denote the vector of act for the time-series of mates of the covariance matrix of log(.P).
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTER Y 329

models in levels resulted in inferences simi- V. Caveats

lar to those arising from estimates of models
in logs. The consistency of lottery-based estimates
The sample used to estimate equations of the effects of military service turns on two
(5)-(7) begins in the fifth year after the key assumptions. First, earnings model (1)
lottery in which members of the cohort were must be an accurate representation of the
drafted. For example, the sample begins in impact of military service. Second, functions
1975 for men born in 1950. This allows for of the draft lottery must be valid instru-
three years of service and one year of read- ments for s, in the earnings regressions.
justment to civilian life. Median length of Three models leading to the failure of these
service of Vietnam era veterans was 37 assumptions are discussed briefly below. The
months (Veterans Administration 1981b, p. first two, incorporating treatment effect het-
16). Evidence from Table 4 suggests that erogeneity and missing covariates, merely re-
veteran effects are essentially zero for the sult in a reinterpretation of the estimates.
1953 cohort, so it was excluded from the The third model, allowing for earnings-mod-
estimation. ifying draft-avoidance behavior, calls into
Estimates of models (5), (6), and (7) are question the assumption that earnings dif-
presented in columns (1), (2), and (3) of ferences by lottery number can be attributed
Table 5. The chi-square statistic in column solely to veteran status.
(3) is an overidentification test statistic for
the overall goodness-of-fit of the reduced A. TreatmentEffect Heterogeneity
form and the chi-square statistic in column
(1) is for the restriction fB = 0. Restricting The effect of veteran status can only be
f, to be zero does not affect the overall fit. estimated for those who served in the armed
The effect of veteran status on earnings forces. If veterans are more or less likely to
growth, although negative, is not statistically benefit from military service than the rest of
significant. The sum of the test statistics in the population, then the estimated veteran
columns (1) and (3) is an overidentification status coefficient does not characterize the
test statistic for the overall goodness-of-fit of impact of veteran status on a random sam-
model (5). This statistic also takes on a value ple. This problem is formalized in the ran-
less than its degrees of freedom, indicating dom coefficients model for treatment effect
that the simplest loss-of-experience model is heterogeneity:
not at odds with the data.
The loss of experience estimated using Ycti= 13c+ St + sia5i+ uit
model (5) is approximately 2 years. This is
somewhat low relative to the median length ai= ao + eie
of service, suggesting that military experi-
ence may be a partial substitute for civilian where ai is the effect of military service on
experience. Table 5 also shows the age at the earnings of person i, with population
which the reduced form veteran effect finally mean equal to a0. If ei is uncorrelated with
reaches zero. This occurs when xct = - 7TJ1 the instruments, an instrumental variables
sT2 so that act = 0. The reduced form esti- estimator of the random coefficients model
mates in column (3) imply that the loss of identifies a0 + E(,ilsi = 1), and not a0
earnings to veterans decays to zero around (James Heckman and Richard Robb, 1985).
age 50 with a standard error17of 16. Related to the problem of treatment effect
heterogeneity is the fact that not all Vietnam
era accessions to the military were induced
by the draft. A substantial fraction of enlist-
17
Delta method standard errors are given by the ments were made by "true volunteers" who
square root of (a1/7T22) - (2'g127Tl/T) +
(?2712/X24), would have volunteered for service in the
where a1, cF2,and aJ12 are theelementsof the covariance
matrix of the estimated 7T, and 7T2. a12 is estimated to
absence of a draft (Tarr, 1981). Suppose that
be - 0.00019, and the square roots of a1 and a2 appear the impact of military service on the earn-
in Table 5. ings of true volunteers differs from the im-
330 THE AMERICAN ECONOMIC REVIEW JUNE 1990

pact on the earnings of draftees and men some other reason than through an effect on
who enlisted because of the draft. Then us- the probability of military service. For exam-
ing functions of the draft lottery as instru- ple, it is sometimes argued that during the
ments will only identify the effect of military Vietnam era, students went to college to
service for the latter group. To see this, let fi avoid the draft and that educational stan-
be an indicator for the veteran status of dards were reduced to avoid having to flunk
draftees and draft-motivated volunteers and students out of school. Lawrence Baskir and
let gi be an indicator of veteran status for William Strauss (1978) claim that Vietnam
true volunteers. Write era college enrollment was 6-7 percent
higher than normal because of the draft.
aisi = affi + aggi If draft-avoidance behavior is correlated
with lottery numbers and with variables re-
for the treatment effect experienced by i. lated to earnings besides veteran status, then
Note that functions of draft lottery numbers lottery-based instruments will be correlated
will only be correlated with fi. Conse- with the regression error in equation (1).
quently, af is identified by lottery-based IV Such correlation will bias estimates of the
estimators, but aggi becomes part of the effects of military service constructed using
regression error. the lottery. But in a previous study using
micro data (Angrist, 1989a), specification
B. The Absence of Covariates tests provided no evidence of a relationship
between lottery numbers and characteristics
An additional problem arises because So- other than veteran status. The overidentifi-
cial Security data contain no information on cation test statistics reported here also show
covariates other than race and age. This may no evidence of omitted variables bias. Fi-
be of concern if the impact of veteran status nally, even if having a low lottery number is
is primarily through its effect on covariates. correlated with a tendency to stay in school,
For example, veterans might have a higher the fact that earnings rise with schooling
level of educational attainment because of implies that lottery-based estimates of the
financial aid available through the GI Bill. effect of veteran status will tend to underes-
In the absence of data on education, esti- timate the true effect.
mated veteran effects confound the "pure"
effect of military service with its effect on VI. Conclusions
education. Formally, the need for covariates
may be represented by replacing at with xiA Estimates based on the draft lottery indi-
in (1). In this case, instrumental variables cate that as much as ten years after their
estimates identify discharge from service, white veterans who
served at the close of the Vietnam era earned
a= [E(xilsi =1) -E(xilsi = O)]8 + a. substantially less than nonveterans. The an-
nual earnings loss to white veterans is on the
In many applications, however, it may be a order of $3,500 current dollars, or roughly
that is actually the parameterof interest. For 15 percent of yearly wage and salary earn-
example, the fact that veteran status influ- ings in the early 1980s. In contrast, the esti-
ences civilian earnings primarily through its mated veteran effects for nonwhites are not
influence on third variables might be of little statistically significant.
importance for issues related to veterans In light of the results reported here, some
compensation. more conventional estimates of the effect of
Vietnam era veteran status do not appear to
C. Earnings-ModifyingDraft be too far off the mark. Rosen and Taubman
AvoidanceBehavior (1982) report estimates close to these, finding
a 19 percent annual earnings loss to Vietnam
Perhaps the most serious problem arises if era veterans. Crane and Wise (1987) find an
the risk of induction affected earnings for 11 percent reduction in 1979 weekly earn-
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 331

ings. On the other hand, Mark Berger and may provide a useful tool for research on
Barry Hirsch (1983) find essentially no effect changing educational attainment in the 1960s
on 1977 weekly earnings, and Angrist (1989a) and 1970s.
reports OLS coefficients of zero when using Finally, there remains the question of rec-
the National Longitudinal Survey (NLS) to onciling the loss of earnings to Vietnam era
estimate the effects of veteran status on veterans with the apparent benefits of mili-
hourly wages. In contrast to the OLS esti- tary service to veterans of World War II and
mates, lottery-based estimates from the NLS other eras (Rosen and Taubman, 1982;
indicate that white veterans had lower hourly Berger and Hirsch, 1983). Elsewhere, Alan
wages in 1981 than their nonveteran coun- Krueger and I have argued that the need for
terparts. Similar results from the SIPP are reconciliation is, at least in part, illusory
reported in Angrist (1989c). Thus, lottery- (Angrist and Krueger, 1989). Although OLS
based estimates from a variety of sources regressions usually show that the effect of
provide conclusive evidence that white Viet- World War II veteran status is large, posi-
nam veterans were disadvantaged by their tive, and significant, these results may actu-
service.18 ally be a consequence of selection bias. By
This paper also proposes a simple expla- exploiting the fact that World War II vet-
nation for the loss of earnings to white veter- eran status is also correlated with exact date
ans: they earn less because their military of birth, we have implemented an instrumen-
experience is only a partial substitute for the tal variables estimation strategy similar in
civilian labor market experience lost while in spirit to the one used here. The results of
the armed forces. Goodness-of-fit tests sug- this procedure indicate that the true impact
gest that for whites, the time-series of vet- of World War II veteran status on earnings
eran status coefficients is consistent with this is no larger than zero and may well be
hypothesis. Experience-earningsprofiles esti- negative.
mated using Social Security data imply that
white veterans suffered an earnings reduc- APPENDIX: DATA SOURCES AND METHODS
tion equivalent to the loss of two years civil-
1. CWHS Data Collection9
ian labor market experience. The Social Security Administration maintains the
The analysis reported here leads naturally earnings histories of covered employees in a data base
to further research on a number of topics. known as the Summary Earnings Record (SER). Ap-
One of these is the question of alternatives proximately one year after the SER has been updated
to the loss-of-experience explanation for the with the latest year's earnings, a one percent sample of
earnings histories is extracted. The sampling frame con-
reduction in white veterans' earnings. Vet- sists of all issued Social Security numbers, and the
eran status may be a screening device, as sample is stratified using the regional information coded
suggested by DeTray (1982), or there may be in the numbers.
cohort size effects such as those discussed by Prior to 1978, the FICA taxable earnings of employ-
ees were reported to the SSA by employers on a quar-
Finis Welch (1979). Because the Social Secu-
terly basis. Self-employed workers report their earnings
rity data include information on variances, annually on schedule SE of Internal Revenue Service
testable implications of these theories might (IRS) Form 1040, which is forwarded to the SSA by the
also include restrictions on second, as well as IRS. Since 1978, employers have no longer been re-
first, moments. Another question for future quired to make quarterly reports. Instead they file IRS
form W-2 with the SSA, as well as with the IRS, on an
research is whether draft eligibility affected annual basis.
educational and career plans independently After 1978, all earnings, including those above the
of its effect on military service. The lottery FICA taxable maximum, are to be reported to the SSA
on form W-2. Furthermore, all employers are required

18Lottery-based estimates from the SIPP and NLS

suggest that nonwhite veterans may have higher monthly 19This section draws on Warren Buckler and Cresston
and hourly earnings than comparable nonveterans. The Smith (1984), U.S. Department of Health and Human
estimates for nonwhites are too imprecise, however, to Services (1987), and personal correspondence and con-
be viewed as conclusive. versationsiwith Buckler and Smith.
332 THE AMERICAN ECONOMIC REVIEW JUNE 1990

to file W-2s with the SSA regardless of whether their between 33 and 36 percent for blacks. To evaluate these
employees are engaged in FICA taxable employment. figures, note that the author's tabulations show that
In practice, however, many employers do not report the roughly 10 percent of white men in these cohorts have
earnings of those engaged in non-FICA-taxable employ- zero recorded wage and salary earnings in the late
ment. A further shortcoming of the W-2 series is the 1970s' Current Population Surveys. Suppose that of the
poor quality of the data during the first years of annual 90 percent who work, 12 percent are in the uncovered
reporting. sector so that only 79 percent of the cohort may be
expected to have positive FICA wage and salary earn-
2. Coverage and Truncation of the CWHS Earnings ings. Adding an estimated 5 percent who only have
Series FICA self-employment earnings implies that 16 percent
FICA coverage includes most wage and salary and should have zero FICA earnings of any type. Thus,
self-employment earnings. For the sample period used 14-18 percent of CWHS earnings being zero for whites
here, the most important coverage exceptions are the between 1973 and 1980 does not seem unreasonable.
majority of federal civilian employees, some state and In recent years, the fraction with zero earnings ap-
local government employees, some agricultural and do- pears to be too high to be accounted for by labor force
mestic workers, and the employees of some nonprofit participation or employment in the uncovered sector.
organizations. A view of coverage by industry in 1981 is This is probably because of the long delay in filing and
given in Robert Meyer (1985, Table 2.1). Meyer's fig- recording Social Security taxable earnings. In their anal-
ures for state, local, and civilian federal government ysis of Social Security data, Card and Sullivan (1988)
employees imply that by 1981, roughly 58 percent of all also note the problems caused by filing delay. Problems
civilian government workers were covered. of undercoverage and filing delay may be especially
FICA taxable maximums are reported in Appendix severe for nonwhites. The 15 percent of CPS nonwhites
Table Al. The combined effects of limited coverage and with zero wage and salary earnings is not large enough
censoring at the taxable maximum are conveniently to explain the approximately 34 percent of nonwhites
summarized by the percentage of all earnings that are with zero FICA earnings in the CWHS.
reported to the SSA. These statistics, reported in De- The fraction with FICA earnings at the taxable max-
partment of Health and Human Services (1987, Table imum is more variable than the fraction with zero
30), show that after 1981 over 90 percent of wage and earnings, ranging from 3 to 15 percent for whites and
salary earnings and over 75 percent of self-employment between 2 to 10 percent for blacks. The FICA earnings
earnings were reported to the SSA. of men with multiple employers can exceed the taxable
The W-2 earnings series excludes earnings from maximum because reported earnings are censored by
self-employment. Unpublished estimates indicate that source. The fraction of men with FICA earnings above
roughly 6.4 percent of men born between 1944 and 1953 the taxable maximum is around 1-2 percent. The W-2
with nonzero earnings in 1984 had self-employment and FICA earnings series show roughly equal fractions
earnings only (figures from private correspondence with at or above the FICA taxable maximum, suggesting that
SSA employees). Other differences between FICA and both variables are drawn from the same underlying
W-2 earnings coverage are described by Mary Millea distribution. But problems with early years of the W-2
and Beth Kilss (1980). series are clearly reflected in the sample statistics. For
example, the standard deviation of whites' W-2 earnings
3. Matching Dates of Birth to the CWHS in 1978 is six times the mean and does not fall below
In the CWHS, information on race and sex is ob- the mean until 1981. Another disturbing feature of the
tained from a computerized record of applications for a W-2 series is that nominal earnings appear to fall from
Social Security number called the NUMIDENT file. 1978-80. The W-2 series also has a substantially higher
For this project, dates of birth were also matched from fraction of zeros than the FICA series does. However,
the NUMIDENT file to the CWHS. Draft lottery num- part of this difference is caused by the inclusion of
bers were then matched to the dates of birth using self-employment earnings in the FICA series. Also, gen-
lottery number tables in Selective Service System Semi- erally there are some individuals with FICA taxable
annual Reports for 1969-73. A small number of indi- earnings but no federally taxable compensation (Millea
viduals were discarded from the final data set because and Kilss, 1980).
there was no information on either their sex, race, or
exact date of birth. 5. Covariance Estimates for Social Security Earnings
and for p
4. CWHS Descriptive Statistics Information on second moments in the aggregated
Descriptive statistics for the CWHS cohorts studied CWHS is restricted to variances. Therefore, off-diagonal
here are reported in Table Al. The combined statistics elements of the covariance matrix of Social Security
for four cohorts were constructed by computing earnings must be estimated. Recall that y denotes the
weighted averages of cohort means. Unless otherwise vector of Yctj, where c indexes cohort, t indexes year of
noted, statistics in the table refer to men with positive earnings, and j indexes
- lottery number cells. The co-
earnings. Sample sizes decline over time due to attrition variance matrix of is block diagonal, with nonzero
from mortality. elements for the covariance between Yc,j and ickj' and
The descriptive statistics indicate that after 1972 the zeros everywhere else. Thus, there is only correlation
fraction of men in the sample with zero FICA earnings between elements of the time-series of earnings for a
varies roughly between 15 and 22 percent for whites and particular cohort and lottery number group. The in-
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 333

TABLEAl-DESCRIPTIVE STATISTICSFOR MEN BORN 1950-53

TaXable FICA FICA at FICA W-2 W-2 at W-2

Race Year N Maximum EarningS Limita,b Zeros EarningSc Limita Zerosa

White 69 68,407 7,800 1,473 0.003 0.321

(1,457)
70 68,339 7,800 1,977 0.010 0.241
(1,825)
71 68,244 7,800 2,581 0.030 0.194
(2,233)
72 68,154 9,000 3,614 0.040 0.157
(2,802)
73 68,053 10,800 4,738 0.046 0.141
(3,460)
74 67,966 13,200 5,727 0.034 0.145
(4,170)
75 67,882 14,100 6,459 0.053 0.169
(4,843)
76 67,794 15,300 7,698 0.078 0.167
(5,548)
77 67,691 16,500 8,974 0.107 0.163
(6,206)
78 67,598 17,700 10,441 0.149 0.165 15,435 0.167 0.244
(7,050) (91,032)
79 67,503 22,900 12,388 0.092 0.166 14,786 0.097 0.261
(8,455) (65,359)
80 67,413 25,900 13,769 0.092 0.175 14,561 0.096 0.272
(9,678) (28,180)
81 67,316 29,700 15,641 0.089 0.183 16,363 0.092 0.272
(11,129) (15,295)
82 67,265 32,400 16,743 0.092 0.200 17,907 0.093 0.282
(12,371) (16,298)
83 67,190 35,700 18,046 0.089 0.210 19,595 0.089 0.275
(13,621) (22,470)
84 67,114 37,800 19,717 0.103 0.219 21,595 0.101 0.281
(14,883) (20,856)
Nonwhite 69 21,514 7,800 1,306 0.002 0.464
(1,471)
70 21,501 7,800 1,711 0.007 0.417
(1,865)
71 21,475 7,800 2,197 0.020 0.388
(2,275)
72 21,443 9,000 3,072 0.025 0.352
(2,922)
73 21,412 10,800 3,992 0.029 0.330
(3,609)
74 21,380 13,200 4,802 0.021 0.336
(4,359)
75 21,339 14,100 5,404 0.031 0.359
(5,015)
76 21,310 15,300 6,453 0.049 0.350
(5,836)
77 21,275 16,500 7,510 0.071 0.341
(6,616)
78 21,243 17,700 8,751 0.096 0.338 13,439 0.116 0.391
(7,541) (63,722)
79 21,200 22,900 10,262 0.061 0.329 13,581 0.067 0.393
(8,938) (63,244)
80 21,167 25,900 11,405 0.063 0.331 11,716 0.067 0.397
(10,147) (13,895)
81 21,130 29,700 12,986 0.063 0.335 13,421 0.063 0.388
(11,628) (14,644)
82 21,109 32,400 14,045 0.067 0.354 14,983 0.065 0.408
(12,849) (16,112)
83 21,077 35,700 15,101 0.067 0.358 16,271 0.065 0.396
(14,167) (19,089)
84 21,042 37,800 16,391 0.077 0.353 17,905 0.073 0.387
(15,237) (20,784)

Notes: Statistics are from the Social Security Adminiatration CWHS. Standard deviations of earings in parentheses. Amounts are in current
dollars. Sample statistics are weighted averages of cells for each race, year of birth, and five consecutive lottery numbers.
aFractions at limit are fractions of nonzero earnings at or above FICA taxable maximum. Fractions zero are fractions of all nondecedents
with zero earings.
bFICA earings are wage and salary and self-employment earnings in Social Security taxable employment. FICA taxable earings are
censored at the taxable maximum except for those with multiple sources. Multiple sources are censored by source.
cW-2 earnings are total W-2 form wage and salary compensation, not censored at the Social Security taxable maximum. W-2 earings do not
include earings from self-employment.
334 THE AMERICAN ECONOMIC REVIEW JUNE 1990

tertemporal correlation structure is assumed constant no parametric distributional assumptions, the adjust-
across lottery number groups so that the 73 cells avail- ment is only approximate because the CPS estimates of
able for each race and cohort may be used to estimate ,Lj do not vary by lottery number. However, the adjust-
correlation matrices for all race-cohort combinations. ment does incorporate variation in p by lottery num-
Correlations are converted to covariances using the ber.
within cell variances available from the CWHS data. Data on earnings above the FICA taxable maximum
This procedure is also applied to the adjusted FICA (,ul.) are taken from the Mare-Winship March CPS
series described in Section 6 of the Appendix, with the Uniform files. Cohort was determined on the basis of
modification that adjusted standard errors (diagonal age in 1985; ages 34-35 were assigned 1950, ages 33-34
elements of Q) are used to convert correlations to were assigned 1951, ages 32-33 were assigned 1952, and
covariances. ages 31-32 were assigned 1953. The CPS reports wage
The covariance matrix of p is also block diagonal, and salary earnings in the year preceding the survey
with elements equal to the variance of PCj in every year. To compute the adjusted 1981-84 earnings series
element of the block corresponding to the time series used in Tables 3 and 4, data from the CPSs for 1982-85
for cohort c and lottery cell j. The variance of Pcj is were used to compute separate a I by cohort and race
estimated using the standard formula for an estimated for each year. Additional details and summary statistics
proportion. The sample size in the formula is taken to for the CPS data are reported in Appendix A of Angrist
be the size of the SIPP cohort for those born in 1950 (1989c). It should be noted that CPS earnings data are
and the size of the CWHS cohort for those born from also censored-at $75,000 for 1981-83 earnings and at
1951 to 53. $100,000 for 1984 earnings.
Standard errors for the adjusted series are calculated
6. FICA Earnings Adjusted for Censoring as follows. Let m = [m m 'm ]' denote the vector of
For economy of notation, in this section all cells are sample moments corresponding to /,uCp', and gLand let
indexed by j. The relationship between the expectation ij ( =i, j=1,2,3) denote the corresponding blocks of the
of censored earnings and the expectation of uncensored covariance matrix of m. The covariance matrix of m is
earnings for cell j is given by assumed to be given by

-=C
+ Pj
(11 112 0

712 122 0
where y? is the mean of uncensored earnings, [j is the 0 0 233
mean of censored earnings, [ is the mean of earnings
above the taxable maximum, Lj is the taxable maxi-
mum, and pl is the fraction with earnings at or above The delta-method covariance matrix for the vector of
the taxable maximum.20 adjusted earnings, ml + m2(m3 - L), is
The FICA earnings series is adjusted for censoring
by applying this formula using estimates of 4 tabulated = 11 +2112(m3- L)
from March Current Population Surveys (CPS) for each
year, race, and cohort. Although this procedure involves
+ 22 (m3-L)2+
- m2233.

Only the diagonal elements of Q are estimated using the

above formula. Estimates of diagonal elements of F.,
20This formula ignores the fact that the censored and 122 are available from the CWHS cell statistics,
earnings of men with multiple employers may be above while diagonal elements of 233 are estimated from the
the taxable maximum. The formula may be used to CPS micro data. An estimator for diagonal elements of
analyze the bias in treatment effects estimated from
212 is easily shown to be (Angrist, 1989c)
censored data. Suppose that cell j is for a sample of
draft-eligible men and that cell k is for a sample of
draft-ineligible men. The draft-eligibility treatment ef- 012j [ i( L- Kji)I Inj-
fect estimated from the difference between CWHS cen-
sored mean earnings in cells j and k is the sample A simplified procedure, described in Section 5 of this
analogue of Appendix, is used to estimate the nonzero off-diagonal
elements of Q directly from the adjusted cell means.
-c - = ( -p?k) +(Pj p k)( L- -IL'.)
7. Matching Draft-Eligibility Status to the SIPP
The Survey of Income and Program Participation is
+ Pk( lk 1'- a Census Bureau longitudinal survey of approximately
20,000 households in the civilian noninstitutional popu-
Assuming that the effect of draft eligibility on both pi lation (U.S. Department of Commerce, 1985). Data for
and tl is of the same sign as the effect on o,u this the first wave of the first SIPP panel were collected from
expression shows that treatment effects estimated from four rotation groups in 1983 and 1984.
the censored data differ from the true treatment effect The SIPP public-use tapes contain year and month
by terms that are opposite in sign from the true effect. of birth. A Census Bureau "in-house" version of the
VOL. 80 NO. 3 ANGRIST: LIFETIME EARNINGS AND VIETNAM ERA DRAFT LOTTERY 335

SIPP contains information on day of birth, which is not (CWHS): Description and Contents," in
released to the public. At the author's request, this Statistical Use of AdministrativeRecords:
information was used to match a dummy variable for
draft-eligibility status to the public use version of SIPP Recent Research and Present Prospects
Panel I. Draft eligibility was determined by the official Volume I, Department of the Treasury,
RSN ceiling for men born from 1944-52 and by RSN Internal Revenue Service, Statistics of In-
95 for men born in 1953. Vietnam era veteran status is come Division, March 1984.
coded from the SIPP variablesVETSTAT, which records
veteran status, and from U-SRVDTE, which records the
Card, David and Sullivan,Daniel, "Measuring
period of service. the Effect of Subsidized Training Pro-
grams on Movements In and Out of Em-
REFERENCES ployment," Econometrica, January 1988,
56, 497-530.
Angrist, Joshua D., "Grouped Data Estima- Chamberlain,Gary, "Multivariate Regression
tion and Testing in Simple Labor Supply Models for Panel Data," Journal of Econo-
Models," Industrial Relations Section metrics, January 1982, 18, 5-46.
Working Paper, no. 234, Princeton Uni- Crane, Jon R. and Wise, David A., "Military
versity, July 1988. Service and the Civilian Earnings of
, (1989a),"Using the Draft Lottery to Youths," Chapter 6 in Public Sector Pay-
Measure the Effects of Military Service on rolls, D. A. Wise, ed., Chicago: University
Civilian Labor Market Outcomes," in Re- of Chicago Press, 1987.
search in Labor Economics, 10, Ron DeTray, Dennis N., "Veteran Status as a
Ehrenberg, ed., Greenwich: JAI Press, Screening Device," American Economic
1989. Review, March 1982, 72, 133-42.
, (1989b),"Selection for Military Ser- Griliches,Z. and Mason,William,"Education,
vice in the Vietnam Era," Industrial Rela- Income and Ability," Journal of Political
tions Section Working Paper, no. 250, Economy, May/June 1972, 80, Part 2:
Princeton University, April 1989. S74-S103.
, (1989c),"Lifetime Earnings and the Hearst, Norman,Newman,Tom B. and Hulley,
Vietnam Era Draft Lottery: Evidence from Stephen B., "Delayed Effects of the Mil-
Social Security Administrative Records," tary Draft on Mortality: A Randomized
Industrial Relations Section Working Pa- Natural Experiment," New England Jour-
per, no. 251, Princeton University, April nal of Medicine, March 6, 1986, 314,
1989. 620-24.
, (1989d),"Instrumental Variables Re- Heckman,James J. and Robb,Richard,"Alter-
gression with Moments from Two-Sam- native Methods for Evaluating the Impact
ples," Chapter 3 in EconometricAnalysis of Interventions," in J. Heckman and B.
of the Vietnam Era Draft Lottery, unpub- Singer, eds., LongitudinalAnalysis of La-
lished Princeton University doctoral dis- bor Market Data, New York: Cambridge
sertation, October 1989. University Press, 1985.
and Krueger,Alan, "Why Do World Meyer, Robert J., Social Security, 3rd ed.,
War II Veterans Earn More?" National Homewood, IL: Richard D. Irwin, 1985.
Bureau of Economic Research Working Millea, Mary T. and Kilss, Beth, "Exploration
Paper, no. 2991, May 1989. of Differences Between Linked Social Se-
Baskir, LawrenceM. and Strauss, William A., curity and Internal Revenue Service Wage
Chance and Circumstance:The Draft, The Data for 1972," in Studies from Intera-
War and the Vietnam Generation, New gency Data Linkages, Report No. 11:
York: Alfred A. Knopf, 1978. Measuring the Impact on Family and Per-
Berger, Mark C. and Hirsch, BanryT., "The sonal Income Statistics of Reporting Dif-
Civilian Earnings Experience of Vietnam- ferences Between the Current Population
Era Veterans," Journal of Human Re- and Administrative Sources, U.S. Depart-
sources, Fall 1983, 18, 455-79. ment of Health, Education and Welfare,
Buckler, Warren and Smith, Cresston, "The Social Security Administration, March
Continuous Work History Sample 1980.
336 THE AMERICAN ECONOMIC REVIEW JUNE 1990

Mincer, Jacob, Schooling, Experience and felter and R. Layard, eds., The Handbook
Earnings, New York: National Bureau of of Labor Economics, Amsterdam: North-
Economic Research, 1974. Holland, 1986.
Newey, WhitneyK., "Generalized Method of Councilof EconomicAdvisors,Economic Report
Moments Specification Testing," Journal of the President, February 1988, Washing-
of Econometrics, September 1985, 29, ton: USGPO, 1988.
229-56. SelectiveServiceSystem,Officeof PublicAffairs,
Rosen, Sherwinand Taubman,Paul, "Changes "A Short History of the Selective Service
in Life Cycle Earnings: What Do Social System," Washington: USGPO, 1986.
Security Data Show?" Journal of Human , Semiannual Report of the Director of
Resources, Summer 1982, 17, 321-38. Selective Service for the Period July 1 to
Schwartz, Saul, "The Relative Earnings of December 31, 1970, Washington: USGPO,
Vietnam and Korean-Era Veterans," In- 1971.
dustrial and Labor Relations Review, July (1969-1973), Semiannual Report of
1986, 39, 564-72. the Director of Selective Service, selected
Tarr,CurtisW., By the Numbers: The Reform issues, Washington: USGPO, 1969-73.
of the Selective Service System 1970-72, U.S. Departmentof Healthand HumanServices,
Washington: National Defense University Social Security Administration, Social Se-
Press, 1981. curity Bulletin Annual Statistical Supple-
Taussig,MichaelK., "The Rationale for Veter- ment, Washington: USGPO, 1987.
ans Benefits," in Those Who Served: Re- U.S. Departmentof Commerce,Bureau of the
port of the 20th CenturyFund Task Force Census, Survey of Income and Program
on Policies Toward Veterans, New York: Participation: Wave I Rectangular File,
The 20th Century Fund, 1974. Washington: USGPO, 1985.
Wald,Abraham,"The Fitting of Straight Lines Veterans Administration(1981a), Reports and
If Both Variables Are Subject to Error," Statistics Service, Research Monograph
Annals of Mathematical Statistics, Septem- 15: Educational and Income Characteris-
ber 1940, 11, 284-300. tics of Veterans, March 1979, Washington:
Welch, Finis, "The Effect of Cohort Size on USGPO, August 1981.
Earnings: The Baby Boom Babies Finan- , (1981b),Office of Reports and Statis-
cial Bust," Journal of Political Economy, tics, Data on Vietnam Era Veterans,
October 1979, Part 2, 87, S69-S97. Washington: USGPO, September 1981.
Willis, Robert J., "Wage Determinants: A , (1984), Administrator of Veterans
Survey and Reinterpretation of Human Affairs, Annual Report 1983, Washington:
Capital Earnings Functions," in 0. Ashen- USGPO, 1984.