Do guns displace books? The impact of compulsory military service on educational attainment

Transcription

1 Unversty of Wollongong Research Onlne Faculty of Busness - Papers Faculty of Busness 2014 Do guns dsplace books? The mpact of compulsory mltary servce on educatonal attanment Thomas K. Bauer Rhensch-Westfalsches Insttut Fur Wrtschaftsforschung Stefan Bender Insttute for Employment Research Alfredo R. Paloyo Unversty of Wollongong, apaloyo@uow.edu.au Chrstoph M. Schmdt RWI Essen Publcaton Detals Bauer, T. K., Bender, S., Paloyo, A. R. & Schmdt, C. M. (2014). Do guns dsplace books? The mpact of compulsory mltary servce on educatonal attanment. Economcs Letters, 124 (3), Research Onlne s the open access nsttutonal repostory for the Unversty of Wollongong. For further nformaton contact the UOW Lbrary: research-pubs@uow.edu.au

2 Do guns dsplace books? The mpact of compulsory mltary servce on educatonal attanment Abstract To estmate the causal effect of conscrpton on the probablty to obtan a unversty degree, we use a regresson-dscontnuty desgn assocated wth the ntroducton of conscrpton n Germany. There s evdence that conscrpton ncreased the lkelhood of completng hgher educaton. Keywords educatonal, servce, mltary, attanment, compulsory, do, mpact, books, dsplace, guns Dscplnes Busness Publcaton Detals Bauer, T. K., Bender, S., Paloyo, A. R. & Schmdt, C. M. (2014). Do guns dsplace books? The mpact of compulsory mltary servce on educatonal attanment. Economcs Letters, 124 (3), Ths journal artcle s avalable at Research Onlne:

3 Do guns dsplace books? The mpact of compulsory mltary servce on educatonal attanment Authors Afflatons Thomas K. Bauer a Stefan Bender b Alfredo R. Paloyo c Chrstoph M. Schmdt d a Rhensch-Westfälsches Insttut für Wrtschaftsforschung (RWI), Ruhr-Unverstät Bochum (RUB), Forschungsnsttut zur Zukunft der Arbet (IZA) b Insttut für Arbetsmarkt- und Berufsforschung (IAB) c Unversty of Wollongong and RWI; correspondng author: alfredo@paloyo.net; d RWI, RUB, IZA, Center for Economc Polcy Research (CEPR) Date 11 May 2014 JEL Classfcaton I28, J24 Keywords regresson dscontnuty, conscrpton, career nterrupton, skll atrophy, TS2SLS ABSTRACT Conscrpton typcally drafts young men when they are at the heght of ther learnng ablty. It can depress the demand for hgher educaton snce skll atrophy and delayed entry nto the cvlan labor market reduce the returns to human-captal nvestments. To estmate the causal effect of conscrpton on the probablty to obtan a unversty degree, we use a regressondscontnuty desgn assocated wth the ntroducton of conscrpton n Germany. There s some evdence that conscrpton ncreased the lkelhood of completng hgher educaton.

4 1 INTRODUCTION Conscrpton nto the armed forces s stll actvely enforced n many countres. Typcally, men are called around the tme they are 18 years old for a medcal examnaton, after whch they can ether be declared ft to serve n the armed forces or exempted (temporarly or permanently). The selected ndvduals serve n the armed forces for a specfed perod of tme. Mltary servce can therefore be construed as a dsrupton of the human-captal-acquston process of young men. In Germany, natonal servce was compulsory untl June 30, 2011 for men ether n the form of mltary servce (Grundwehrdenst) or cvlan servce (Zvldenst) for conscentous objectors. When Germany was allowed to rearm tself after becomng a member of NATO n 1955, men born on or after July 1, 1937 had to undergo mental and physcal examnatons upon comng of age to determne whether they were ft to serve n the German armed forces (Bundeswehr). However, those born earler were exempted from CMS. Colloqually, they are called the Whte Cohort (weßer Jahrgang) because they nether served n World War II nor n the new German armed forces. We use the Whte Cohort as a control group to whch we can compare the educatonal outcomes of those facng a postve probablty of beng drafted. 2 DATA CONSTRUCTION AND DESCRIPTION We use two datasets for our emprcal analyss. The frst dataset ( Penson Data ) contans nformaton on the servce status of men and ther dates of brth, whch allows us to compute the probablty of servng n the armed forces as a functon of one s date of brth. The second data source ( IABS ) conssts s provded by the 1

5 Insttute for Employment Research (IAB). 1 Except for some specal groups, the dataset contans the total German populaton that was ganfully employed for at least one day for the perod The followng observatons were removed from both datasets: females, East Germans (defned as havng at least one employment spell n East Germany), persons who had an employment spell outsde West Germany at any pont (ncludng foregners and ethnc German mmgrants), persons born before January 1, 1934, persons born after December 31, 1940, and professonal solders. We excluded those people wthout a tertary-system-entry certfcate (Hochschulrefe). Snce these people were vrtually excluded from enterng unversty, ther decson to seek out tertary educaton s most lkely unaffected by the ntroducton of conscrpton. For the Penson Data, these sample restrctons leave us wth 15,835 men born around the threshold date who, f they performed CMS, served for not more than 12 months. For the IABS, we end up wth 320,620 men. Fgure 1 plots the share of conscrpted men by month of brth. Fgure 2 shows the percentage of men who obtaned a unversty degree over the sample cohorts. In general, one can observe that the completon rates for tertary educaton are ncreasng over tme, whch generally corresponds to the easng of the access to hgher educaton over the same perod. About a year after the threshold date, the share of men wth a unversty degree s much hgher than the correspondng year before the threshold date. The overall mean for each sde of the cutoff pont clearly ndcates that the non- Whte Cohorts generally have a hgher educatonal attanment than the Whte Cohorts. 1 See Bender, Haas, and Klose (2000) for detals. A new verson of the IABS s avalable for researchers va the Research Data Center of the German Employment Agency n the Insttute for Employment Research (RDC IAB). See 2

6 [FIGURE 1] [FIGURE 2] 3 ESTIMATION STRATEGY AND RESULTS The Whte Cohort was exempted from conscrpton. Let N be equal to 1 f ndvdual does not belong to the Whte Cohort and 0 f he does. The treatment status s represented by M, whch s equal to 1 f ndvdual was eventually called to serve n the mltary and 0 otherwse. Let B represent the threshold value at whch pont the condtonal probablty of recevng treatment jumps. The Whte Cohort s exempted from conscrpton so that E [ M B < B] = 0 ; for those born on or after B, the probablty to be drafted s a functon of a vector of ndvdual characterstcs x,.e., E[ M B B] = f ( x ). We mplement a two-sample two-stage least-squares (TS2SLS) approach (Angrst and Krueger 1992; Inoue and Solon 2010). The reason s that we can observe B n both datasets but only M n the Penson Data and only Y, an ndcator of havng a unversty degree, n the IABS. 2 The sequental system of equatons s M Y 2 1 = α + f ( B = α + g( B ) + δn ) + τmˆ ε 1 + ν, 2 where the second subscrpt ndcates whch dataset s used (1 = Penson Data; 2 = IABS). The varable M ˆ 2 denotes the cross-sample ftted values, whch are computed 2 In Bauer et al. (2012), ndvduals were matched from both datasets. However, for ths paper, we obtaned the whole sample of men from the IABS, not all of whom are avalable n our orgnal dataset. Therefore, we had to rely on TS2SLS to retreve the requred estmates, as explaned n a prevous footnote. 3

7 by usng the parameter estmates obtaned from the frst stage based on the Penson Data but usng the observatons n the IABS. We apply the restrcton f B ) = g( B ) so that the system can be estmated va two-sample two-stage least-squares. ( 1 2 Table 1 presents the frst-stage results usng the Penson Data; Table 2 shows the results of a regresson of a bnary varable ndcatng completon of a unversty degree on the nstrument based on the IABS. All regressons nclude the control functon f B ) wth varyng degrees of the polynomal order. Table 1 shows that the ( probablty of beng drafted s postvely related to one s date of brth. For the cohort of men born n 1937, we recover an estmate of percentage ponts. [TABLE 1] Table 2 presents the reduced-form estmates based on the IABS. Here, we hghlght the fact that, wthout nstrumentng conscrpton, we obtan a statstcally sgnfcant relatonshp for two subsamples (Cohorts and 1937) between one s date of brth, agan as captured by the bnary nstrumental varable for beng part of the Whte Cohort or not, and the probablty of havng a unversty degree. [TABLE 2] Table 3 shows results usng the TS2SLS. Our estmates ndcate that conscrpton had a postve mpact on the probablty of obtanng a unversty degree. Although the statstcal sgnfcance of our estmates depend on whch cohorts we use, the estmate s sgnfcant for precsely that subsample for whch the valdty of the RD estmate s most credble. We calculate that conscrpton rased the probablty of havng a unversty degree by about 15 percentage ponts. That beng sad, the fact 4

8 that both the statstcal sgnfcance and the magntude of the effect change over the subsamples does not make the resultng estmates more compellng. [TABLE 3] 4 DISCUSSION AND CONCLUSION Smlar to prevous country-specfc studes (Card and Lemeux 2001; Maurn and Xenogan 2007), we fnd uncover evdence ndcatng that the ntroducton of conscrpton ncreased the demand for post-secondary nvestments n human captal. We estmate the effect of ths ntroducton to be 1.7 percentage ponts. We attrbute ths to the fact that enrollment n an educatonal nsttuton s an effectve method to avod the draft. Note, however, that ths s not the effect of mltary servce but rather the reduced-form mpact of beng draft-elgble. The estmates range from about 9 to 15 percentage ponts, wth the hgher bound exhbtng the hghest level of statstcal sgnfcance wthn our range of estmates. Snce our estmate s much hgher and less robust than the reported effects n other studes, we emphasze cauton n generalzng ths estmated effect to the rest of the populaton, especally consderng the sample excluson crtera that were used as well as the fact that the dentfcaton n the RD desgn rests on dfferences n a small neghborhood. We post two reasons to explan our fndngs. The frst reason can be ndrectly attrbuted to skll atrophy. Human captal n partcular, that part of one s stock that s not used n the armed forces but s rewarded n the cvlan sector tends to deprecate whle on actve duty. Replacng lost human captal requres addtonal educaton on the part of the former conscrpt. 5

9 Second, post-conscrpton human-captal acquston allows men to catch up wth the rest of ther cohort n terms of earnngs. It s well known that the experence earnngs profles are dssmlar by educatonal attanment. Whle, on average, earnngs ncrease as a functon of labor-market experence, the slope of ths functon s much hgher for unversty graduates than for non-graduates. Furthermore, the age earnngs profle of a unversty graduate would exhbt a longer flat regon before reflectng a postve slope, ndcatng the fact that the person has spent more years n the school system before enterng the cvlan labor market. Ths flat regon would be even longer for conscrpts, whch could nduce them to resort to unversty tranng after ther servce n the hopes of quckly matchng the earnngs of ther non-conscrpted peers. REFERENCES Angrst, Joshua D. and Alan B. Krueger (1992). The effect of age at school entry on educatonal attanment: An applcaton of nstrumental varables wth moments from two samples. Journal of the Amercan Statstcal Assocaton 87(418): Bauer, Thomas K., Stefan Bender, Alfredo R. Paloyo and Chrstoph M. Schmdt (2012). Evaluatng the labor-market effects of compulsory mltary servce European Economc Revew 56(4): Bender, Stefan, Anette Haas and Chrstoph Klose (2000). The IAB Employment Subsample : Opportuntes for analyss provded by the anonymzed subsample. Forschungsnsttut zur Zukunft der Arbet (IZA) Bonn: IZA Dscusson Paper No

10 Card, Davd and Thomas Lemeux (2001). Gong to college to avod the draft: the unntended legacy of the VetnamWar. Amercan Economc Revew 91(2): Inoue, Atsush and Gary Solon (2010). Two-sample nstrumental varables estmators. Revew of Economc Studes 92(3): Maurn, Erc and Theodora Xenogan (2007). Demand for educaton and labor market outcomes: Lessons from the abolton of compulsory conscrpton n France. Journal of Human Resources 42(4): Murphy, Kevn M. and Robert H. Topel (1985). Estmaton and nference n two-step econometrc models. Journal of Busness and Economc Statstcs 3(4):

11 Tables TABLE 1 Frst-stage estmates (Penson Data) Cohort All Non-Whte Cohort *** [0.0204] *** [0.0148] *** [0.0120] *** [0.0165] Polynomal degree J F-statstc Adjusted R Observatons 15,835 11,099 6,668 2,171 NOTES. The dependent varable s havng served n the armed forces. The regressons nclude a constant term. Bracketed numbers are the standard errors. * p < 0.10, ** p < 0.05, *** p < SOURCE. Authors own calculatons. 8

12 TABLE 2 Reduced-form estmates (IABS) Cohort All Non-Whte Cohort [0.0075] * [0.0071] [0.0069] ** [0.0080] Polynomal degree J F-statstc Adjusted R Observatons 320, , ,121 43,197 NOTES. The dependent varable s havng a unversty degree. The regressons nclude a constant term. Bracketed numbers are the standard errors. * p < 0.10, ** p < 0.05, *** p < SOURCE. Authors own calculatons. 9

13 TABLE 3 Impact of CMS on unversty educaton Cohort All Non-Whte Cohort * ** [0.0748] [0.0573] [0.0577] [0.0679] Polynomal degree J NOTES. Bracketed numbers are Murphy Topel (1985) standard errors. The regressons nclude a constant term. * p < 0.10, ** p < 0.05, *** p < SOURCE. Authors own calculatons. 10

14 Fgures FIGURE 1 Probablty to be drafted for mltary servce by month of brth.4 Probablty to be drafted NOTE. Dashed lnes represent quadratc fts over Whte Cohort and non-whte Cohort observatons. 1/34 7/34 1/35 7/35 1/36 7/36 1/37 7/37 1/38 7/38 1/39 7/39 1/40 7/40 Month and year of brth SOURCE. Authors own llustraton based on Penson Data. 11

15 FIGURE 2 Share of men wth a unversty degree by date of brth Share of men wth a unversty degree Dfference between brthday and threshold date n days NOTE. Dashed lnes represent quadratc fts over Whte Cohort and non-whte Cohort observatons. SOURCE. Authors own llustraton based on IABS. 12