I

RELATIVE AVERAGERECRUITING YIELDS

BY AFQT CATEGORY

Jacquelyn Hughes

William H. Sims

j U ;,< , IN!",

! V -

.,.,,-~

~I

i , -%. 'r I J- !l

;' ?, r, .FOR NAL ANALYSES......... -f-, :. E : 1626 Alt:x ria, Virgim 22 .20')268

AP~r~vE ) O& PUEDiC RGLEAS -E; DiST.RIBUTION UNUM IT[iBU

,,1 c -c 1L, c In5Z :a c N O 0 14 -87 -C 0 1

7htr;I Noi I~ irniitium n1,0sre hu~tpinian cftGNA at th3 t!m of-, no orsn t- pi-o of Onoc Dopatmont rd tho Nnivy.

REPORT DOCUMENTATION PAGE Form ApprovedOPM No. 0704-0188

Public reporti burden for this collection of infwmao is estimated to avrap I hor pe response. incuding the tima for veiewing insauctions. sa-rchi existing data sources gatheig ando-ilmiog the dam emded, and reviewing the collection of informatoi. Send tmmewts regarding this burden ortimahl or say other aswia of this collection of information, inudit sgget.-onfor teducg this burden, to Washington Headquarters Servwc, Diectoram for Informattion Operations and Reports, 1215 Jefferou Davis Higbway, Suite 1204, Arlington, VA 22202-4302. and tothe Offiwe of Information and Regulatory Affairs, Office of Mamgen and Budget, Washington, DC 20503.

1. AGENCY USE ONLY (Leave Blank) 2 REPORT DATE 3. REPORT TYPE AND DATES COVERED

May 1989 Final

4. TILE AND SUBTITLE 5. FUNDING NUMBERS

Relative Average Recruiting Yields by AFQT Category C N00014-87-C-O0l

PE - 65153M

6. AUTHOR(S)Jacquelyn Hughes, William H. Sims PR - C0031

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) S. PERFORMING ORGANIZATIONREPORT NUMBER

Center for Naval Analyses CRM 89-1144401 Ford Avenue

Alexandria, Virginia 22302-0268

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING AGENCY

Commanding General REPORT NUMBER

Marine Corps Combat Development Command (WF 13F)Studies and Analyses Branch

Quantico, Virginia 22134

11. SUPPLEMENTARY NOTES

12& DISTRIBUTION/AVALABILTrY STATEMENT 12b. DISTRIBUTION CODE

Approved for Public Release; Distribution Unlimited

13. ABSTRACT (Maximum 200 words)

-The number of recruits enlisted from a target population varies significantly by aptitude score. This memorandum calculates the ratio of recruits topopulation by AFQT category for the fiscal years 1980 to 1987.

14. SUBJECT TERMS 15. NUMBER OF PAGESSAFQT (armed forces qualificaiton test), Aptitude tests, Enlisted personnel, Mental ability, Military personnel, 18Military requirements, Performanc.human), Personnel selection, Qualificatons, Recruits, Standards, Statistical 16. PRICE CODEanalysis, Statistical data, Test scores

17. SECURITY C'_,ASSIFT0T.)N 18. SECURITY CLASSIFICATON 19. SECURITY CLASSIFICATION 20. LIMITATION OF ABSTRACTOF REPORT C OF THIS PAGE CPR OF ABSTRACT CPR SAR

NSN 7540-01-280-5500 Standard Form 298, (Rev. 2-8Prearibad by ANSI Std. 239-1829-01

ED CENTER FOR NAVAL ANALYSESA D,,sint of ud son Ins,.ule 4401 Ford Avenue * Post Office Box 16268 - Alexandria, Virginia 22302-0268 - (703) 824-2000

20 September 1989

MEMORANDUM FOR DISTRIBUTION LIST

Subj: Center for Naval Analyses Research Memorandum 89-114

Encl: (1) CNA Research Memorandum 89-114, Relative Average Recruiting

Yields by AFQT Category, by Jacquelyn Hughes and William H.

Sims, May 1989

1. Enclosure (1) is forwarded as a matter of possible interest.

2. The number of recruits enlisted from a target population varies

significantly by aptitude score. This memorandum calculates the ratio

of recruits to population by AFQT category for the fiscal years 1980 to

1987.

4ewis RICabeDirector

Manpower and Training Program

Distribution List:

Reverse page D(cs . or

IN W! )TIC 1 :.:i

6 J,, (-,h, . ' I

6 ., , , o

A~ L,(

Dist=.-.2 ..

Subj: Center for Naval Analyses Research Memorandum 89-114Distribution ListSNDLA 1 ASSTSECNAV MRAA6 HQMC MPR & RA

Attn: Deputy Chief of Staff, Manpower (2 copies)Attn: Personnel Procurement (2 copies)Attn: Code MA (2 copies)

FF38 USNAAttn: Dr. Robert F. Lockman, Economics Department

FF42 NAVPGSCOLAttn: LibraryAttn: Code 64

FF44 NAVWARCOL (2 copies)Attn: Elll

FF67 NAVFITWEPSCOLFJA I COMNAVMILPERSCOMFJB 1 COMNAVCRUITCOMFKQ6D NAVPERSRANDCEN

Attn: Director, Manpower Systems (Code 61)Attn: Technical LibraryAttn: Technical Director (Code 01)

FTI CNETV12 CGMCCDC

CRM 89-114/May 1989

RELATIVE AVERAGE RECRUITINGYIELDS BY AFQT CATEGORY

Jacquelyn HughesWilliam H. Sims

A Division of = Hud-qn Institute

CENTER FOR NAVAL ANALYSES4401 Ford Avenue. Post Office Box 162b8 • Alexandria. Virginia 22302-0268

ABSTRACT

The number of recruits enlistedfrom a target population varies signifi-cantly by aptitude score. This memoran-dum calculates the ratio of recruits topopulation by AFQT category for the fis-cal years 1980 to 1987.

-iii-

EXECUTIVE SUMMARY

Each year the armed services recruit several hundred thousand newpersonnel. Enlistment standards for these personnel are established interms of mental aptitudes as measured by the Armed Forces QualificationTest (AFQT), which is constructed from the Armed Services VocationalAptitude Battery (ASVAB). Estimates of the costs of recruiting as afunction of fine-grained intervals of ASVAB scores are frequentlyneeded, as is the case for the congressionally directed joint-serviceJob Performance Measurement project. The goal of this project is to setcost-effective enlistment standards. Cost estimates are currentlyavailable only in terms of very large intervals ofASVAB scores.

This research memorandum develops a methodology for calculatingrelative recruit yields for intervals of AFQT scores. Recruit yield, inthis context, is defined as the proportion of a population that enlistswhen recruiting effort is undifferentiated. Although this approach doesnot directly provide estimates of recruiting costs for different AFQTcategories, it does provide useful information on the recruit yieldsfrom different mental groups of an undifferentiated recruiting effort.The procedure is illustrated by applying it to readily available datafor individual AFQT categories. If the method proves useful it caneasily be applied to smaller score intervals.

The fundamental assumption underlying the analysis is that therecruiting effort is undifferentiated--that is, that there is an apti-tude level at and above which recruiters of at least some services areindifferent to the aptitude level of the recruit. This aptitude levelmay vary from year to year.

The data used in this analysis are on male high-school-graduate(and senior) applicants and accessions from all military services withinthe Department of Defense during fiscal year (FY) 1980 through FY 1987.The 1980 ASVAB Reference Population data set was used to establish thetarget population base.

RESULTS

The assumption of an undifferentiated recruiting effort appears tohold, and the method is deemed appropriate for use in AFQT category IIIBand above for recent years and in AFQT category IVA in FY 1980. Resultscould probably be calculated in aptitude intervals as small as 10 per-centile points and might be extended to AFQT category IVB.

Estimates of relative average DOD recruiting yields for male highschool graduates by AFQT category are shown in table I and may be sum-marized as follows:

e Individuals with aptitude slightly above the mean (cate-gory liA) or slightly below the mean (category IIIB) ofthe population have about the same recruiting yield.

-V_

" Individuals with aptitude about one standard deviationabove the mean (category II) have a yield about 30 percentlower than that at the mean.

" Individuals with aptitude about one standard deviationbelow the mean (category IVA) have a 60-percent higheryield than that at the mean.

" The yield drops sharply at high aptitudes; the enlistmentrate for category I is about 70 percent lower than that atthe mean.

Table I. Relative average DOD recruiting yields by recruit aptitudelevel

AFQTpercentile Standard deviation

AFQT of mean percentile Relative averagecategory Range Mean from population meana recruiting yield

I 93-99 96.5 +1.6 0.31II 65-92 79.5 +1.0 0.70IlIA 50-64 57.0 +0.2 1.00IIIB 31-49 40.0 -0.3 1.05IVA 21-30 25.0 -0.8 1.61

a. (Mean AFQT percentile - 50.0)/28.9.

-vi-

CONTENTS

Page

Tables............................................................. ix

Introduction........................................................ 1

Data............................................................... 2

Assumptions......................................................... 2

Results............................................................. 6

Extension of the Results.............................................9

References......................................................... 11

Appendix A: Supplemental Data.....................................A-1

-Vii-

TABLES

Page

1 AFQT Categories ................................................ 2

2 DOD Accessions, Applicants, and Population of Male HighSchool Graduates by AFQT Category ............................ 3

3 Ratio of Applicants to Population by AFQT Category,All Services .................................................. 5

4 Ratio of DOD Accessions to Applicants by AFQT Category ...... 6

5 Ratio of DOD Accessions to Population by AFQT Category ...... 7

6 Relative Average DOD F3cruiting Yields by AFQT Category

and Year ...................................................... 7

7 Relative Average Recruiting Yield by RecruitAptitude Level ............................................... 8

-ix-

.. ...

INTRODUCTIOIN

All services are currently involved in a congressionally directedjoint-service project aimed at linking military enlistment standards toon-the-job performance and, through tnis linkage, setting cost-effectivestandards. As used here, the term enlistment standards generally refersto scores on the Armed Services Vocational Aptitude Battery (ASVAB) andthe Armed Forces Qualification Test (AFQT) score, derived from theASVAB.

To address the performance aspect of the project, extensive hands-on tests of job performance are being built and administered to re-cruits. Analyses of the resulting data will yield the relationshipbetween ASVAB scores and job performance. I

Performance, however, is only part of the issue. It is well knownthat recruits with higher aptitudes will perform better, but is the bet-ter performance worth the presumed extra cost of recruiting and retain-ing these recruits? In a pioneering effort to explore this issue, theRAND Corporation developed a cost-benefit model [1, 2] for use by thejoint-service project. Alternative models have also been examined [3].For maximum effectiveness these models require a well-defined relation-ship between the ASVAB scores of individual recruits and the costs ofrecruiting, training, maintaining, and retaining them.

This research memorandum addresses the issue of the relationshipbetween AFQT scores and the number of accessions for a given population.Previous efforts in this area [4,5,6] have tended to be cross sectionalor time-series econometric analyses of high-level aggregations of re-cruits (i.e., high school graduates or upper AFQT categories). The jobperformance project, on the other hand, needs recruiting costs delin-eated for very small intervals of AFQT scores. Although recruitingcosts are not directly estimated in this memorandum, calculating rela-tive yields of recruiting provides important information on the relativeyields from different AFQT groups.

A method that should enable average recruiting yields to be esti-mated in very small intervals of AFQT scores is proposed here. Theprocedure is illustrated by applying it to readily available data forindividual AFQT categories (table 1). If the method proves useful itcan easily be applied to smaller score intervals.

The fundamental assumption underlying the analysis is that eachyear the recruiting effort made on the youth population is undifferen-tiated above a certa n recruit aptitude level. Above that aptitudelevel the recruiters of at least some of the armed services are indif-ferent to the specific aptitude level of the recruit. It follows thatthe recruiting product (number of recruits per unit of target popula-tion) per AFQT interval is inversely proportional to the average re-cruiting effort (cost) for that AFQT interval. Analysis of data from

-1-

1980 shows that an undifferentiated effort can be assumed for male highschool graduates at or above AFQT category IVA. Further analysis couldprobably extend the coverage to include AFQT category IVB.

Table 1. AFQT categories

Range ofAFQT category percentile score

I 93-100II 65-92liA 50-64IIIB 31-49IVA 21-30IVB 16-20IVC 10-15V 0-9

DATA

The data used in this analysis are shown in table 2. The numbersof male high school graduates (HSG) and high school senior (HSSR) appli-cants and accessions from 1980 through 1987 were developed from dis-aggregated data supplied by the Defense Manpower IData Center (DMDC).Data on the number of male high school graduates1 age 17 to 21 in hepopulation was calculated from the 1980 ASVAB Reference Population dataset [7]. Note that these scores have been correctly renormed, whenappropriate.

ASSUMPTIONS

The fundamental problem is to identify an aptitude level at orabove which recruiters' effort to recruit individuals is undifferen-tiated. Recruiting effort may be considered to be of two kinds: theeffort to get individuals to apply for enlistment, and the effort toturn applicants into accessions. These two kinds of effort are exploredin turn.

With regard to effort of the first kind, it is reasonable to assume

a priori that an undifferentiated effort extends into fairly low apti-tude levels. Before individuals apply for enlistment and subsequent

1. High school seniors were selected on the basis of high school statusat time of testing with ASVAB (survey question 4 in reference 7).2. The sample size is 1,762 cases. The distribution of aptitudes seemsreasonably stable (see appendix A).

-2-

,-4.& r, r-.7%~ r- a r- C1 M-4 t r-.rl - In .D r 4 - It0 r-U) U

[-444ce~ -'(44

00 ON C) It It. a% a% Ln 0C14 -4 0 l

-4- . - 0 r- I . 0 z

U)~~~ ~ ~ 0- 4100 1(1-4 r-4

5.a 0Ll '0 ' Cu0

0 0

cor- e r - trn-a~- M %S %0 -)u4

a' - r --4Cc U - r-: - 4 4) P.

0 -4 , 1,0,-4-4:)10Me0 .400 -4 r-4 .) .

0 It 000r C M4r n "r4 %r 04 m~ -4I m

a% -e ,- ,~'r- - 4) (A.-4

cz ~ c.-o -%Dr- -4 -40 r- Cn co Ja'

Su 0 4- 00C'-4 (n)

4J r 0 CA*M n: 00 - ,C 1 e 4C : *: 04 (2)

.,4 0~ %0-4 00 0 -Tr- -4 M r- -4 -410 m 914

'--4 m 4Cl) OC.ALr 4 - - rr 0 -

-44Cf0-

10 -4cO'0. c',aM 040Mr- 0% -4r wC' X,4

r-4*U .C uC1.4 -4Cnr-001 4 0 ca0r-44 ).

4J)- 0 tk ) )-40 CJ-) c b

co 4)4CD-

., 40 1 -4a 40 r-.a'C 4C14 %0 ouv400- O 0-4 0 -4T0 0 m0%D- c q0 r '4)(7% Cu

ca %0 C T4C -4 r- %rST' V) 4~ 0 4)-~4a

Cu~~~a CC. O ~ .CJ ~ A u

0 01-4 Q.-4%0 1 o n0 - 0r 0 1 -C1 40 0 n% awA.C) Lr. - C 0 C, ) W 'C0C ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . 0 C1 M- M 1 1 ' 0L)%o: i0ma 0 nL .1

-4 U) 0 U

0 E- 0 1:-1 4: c N 4:a1 go 11- ~ w0)

-4 W (1) 1-4 1-4 1-~' 1-4 -41- > -4 - - > U " -44 MC

E4 ci Cu .0 Q

-3-

subsequent ASVAB testing, the reiruiter will generally have only crudeinformation about their ability. This information might include gradesit high school, scores on a screening test, and personal impressions.Based on such information, it is difficult to reliably predict whether aprospect is in one or the other of any two adjacent AFQT categories. Anindividual estimated to be in category IVA might really be in IIIB andwould be lost as a recruit if not encouraged to apply. Recruiters haveequal difficulty judging whether an individual falls into one or theother of the adjacent categories IIIB and liA. On the other hand, mostrecruiters probably learn quickly to identify persons with very lowaptitude, perhaps category V, and do not waste further effort on them.The difficulty lies in judging where the break occurs.

Data on the ratio of applicants to population by AFQT categoryshould help quantify the break point. It is reasonable to expect thatthe ratio of applicants to population in any year should increase asthe aptitude (and thus options) of the potential applicant decreases.On the other hand, if recruiters discourage individuals below a certainaptitude from applying, then the trend should reverse. In table 3, theactual tabulated ratios are presented, and the outcome is as expected.For example, in 1987 a ratio of 2.9 percent in category I increases to8.1 percent in category IIIB, reverses to 7.9 percent in category IVA,and then declines smoothly to 1.8 percent in category V.

Similar trends characterize the other years. Fiscal year 1980 isparticularly worthy, in that the trend holds through category IVB/C andonly reverses in category V. FY 1980 was unusual because a misnormedASVAB [8, 9] was used from 1976 through 1980 so that the information therecruiters had about the aptitude of individuals both before and afterASVAB testing was inaccurate. The misnormed ASVAB overestimated apti-tudes so that many individuals who were thought to be in category IlIIAwere actually in IIIB, many thought to be in IIIB were actually in IVA,and so on. Prior experience as well as screening tests tied to the mis-normed ASVAB would, of course, yield the same misleading expectations.As a result many who would have been discouraged from applying were infact encouraged. The data in table 3 suggest that only applicantsactually in category V were actively discouraged from applying.

Categories in table 3 that show declining ratios are assumed todenote aptitude levels at which some applicants were discouraged fromapplying; hence, they are excluded from our final analysis.

1. A small percentage of applicants (5 to 15 percent) will have beentested earlier on the version of ASVAB used in high schools. For theseindividuals, recruiters will have good aptitude information.2. Generally limits are set on the number of category IVA applicantsthat may be accepted for enlistment. No category V applicants areaccepted.

-4-

Table 3. Ratio of applicantsa to population by AFQT category, allservices

Fiscal yearAFQT

category 1980 1981 1982 1983 1984 1985 1986 1987

I .0 4 2b .022 .027 .028 .023 .022 .027 .029II .043 .059 .068 .070 .056 .052 .057 .056liA .065 .072 .082 .086 .070 .069 .074 .075IIIB .061 .072 .082 .082 .072 .085 .082 .081IVA .099 .119 .121 .102 .092 .091 .0 7 8c .0 7 9cIVB/C .112 .122 .1 0 7c .0 7 8c .0 7 2c .0 5 9c .0 4 7c .0 4 6cV .0 7 3c .0 8 0c .04 9c .0 2 7c .0 2 3c .0 2 8c .0 2 0c .0 18c

a. HSG/HSSR only.b. Probable anomalous result due to poorly determined AFQT conversion

table for ASVAB forms 6 and 7 in the upper percentiles [8].c. The declining ratio is taken as an indication that individuals in

that category were not actively recruited.

The second kind of recruiting effort is convincing an applicant whohas tested on ASVAB to enlist. Here the general expectation is that theratio of accessions to applicants should increase as aptitude decreases(less options available to applicant) up to the point at which the apti-tude is below the minimum standards or is in a category in which only alimited number of recruits are accepted. The ratios, as shown intable 4, generally conform to expectations.

For example, in 1980 the ratio of 0.49 for AFQT category I in-creases slowly to 0.58 in category IVA before decreasing slightly to0.51 in category IVB/C and to zero in category V. This result is inter-preted to indicate that in 1980 an undifferentiated recluiting effortwas expended on persons in AFQT category IVA or higher.

The data for more recent years support the assumption of an undif-ferentiated effort for category IIIB and above. This result does notimply that all services actively seek category IIIB personnel--only thatsome do. Categories for which the assumption does not seem to hold arenoted in table 4 and are excluded from further analysis.

1. The actual extent of the misnorming was such that an undifferentiatedeffort may well have extended to category IVB. Individuals thought tobe in category IIIB (31st percentile) were actually in category IVB withtrue percentile scores of 17.

-5-

Table 4. Ratio of DOD accessionsa 'b to applicants by AFQT category

Fiscal yearAFQT

category 980 1981 1982 1983 1984 1985 1986 1987

I .49 .56 .46 .50 .67 .81 .59 .63II .52 .55 .51 .55 .74 .67 .67 .68IliA .55 .61 .55 .60 .83 .72 .76 .76IIIB .57 .61 .56 .62 .79 .75 .89 .75IVA .58 5 3C 46c

4 1c 48c 4 5c 37c 32c

IVB/C .5 1c 26c 1 6c 0 8c 0 1c 0 1c 0 0c 00cV .00c .00c .00c .00c .00c .00c .00c .00c

a. Male HSG/HSSR only.b. All accessions in a given year do not necessarily come from that

year's applicants because some applicants are placed in delayed entryprograms.

c. The declining ratio is taken as an indication that individuals inthat category are not actively recruited.

RESULTS

The recruiting product (that is, the ratio of accessions to popula-tion, taken from table 2) produced by the undifferentiated effort issummarized in table 5. For example, the data indicate that 2.1 percentof the population of male high school graduates age 17 to 21 in AFQTcategory I were enlisted in 1980. This may be contrasted with a3.6 percent rate for category IIIA in that year.

In 1987 the percentage by AFQT category ranged from 1.8 to6.0 percent. At first glance, this seems like a very small fractionand might lead one to conclude that recruiters do not actively recruitbut function as "ordertakers." If this were the case, it might imply agoal-driven outcome and weaken the assumptions on which this analysis isbased. However, the percentage of population accessed is more properlymultiplied by a factor of 5 because the population base is a five-yearage group while accessions must be produced each year. Viewed in thislight recruiters are expected to enlist about 9 percent of category Imale HSGs and 30 percent of category IIIB from each graduating class.With this perspective, the process is reasonably seen as more nearlysupply-limited and less goal-driven.

1. The actual population of youth has declined somewhat between 1980 and1987. Because absolute numbers are not central to the discussion, thepopulation figures for 1980 are used throughout the period.

-6-

Table 5. Ratio of DOD accessionsa to populationb by AFQT category

Fiscal yearAFQT

category 1980 1981 1982 1983 1984 1985 1986 1987

I .021 .012 .012 .014 .016 .018 .016 .018II .023 .032 .035 .038 .042 .035 .038 .038liA .036 .044 .045 .052 .058 .050 .056 .057IIIB .035 .043 .046 .051 .057 .064 .073 .060IVA .058 - - - - - - -

a. Male HSG/HSSR only.b. Male HSG only.

The relative average recruiting yield is taken as the inverse ofthe recruiting product in table 5 scaled to the yield in AFQT categoryliA. The resulting estimates of relative yields are shown in table 6.For example, the data indicate that in 1980, the relative yield forcategory I male HSG was 0.58, compared to category liA male HSG. Putdifferently, if recruiters were attempting to enlist recruits from apopulation of 1,000 category I male HSGs, they would expect to access21. From a population of 1,000 category liA male HSGs, they wouldexpect to access 36. Therefore, the relative yield for category I is21/36, or 0.58. The estimates of relative yields are generally stableacross years and are averaged in the last column of table 6. Theseaverages are taken as the best estimates of relative average DODrecruiting yields by AFQT category.

Table 6. Relative average DOD recruiting yieldsa by AFQT categoryand year

Fiscal yearAFQT

category 1980 1981 1982 1983 1984 1985 1986 1987 Mean

I .5 8b .27 .27 .27 .28 .36 .29 .32 .31II .64 .72 .78 .73 .72 .70 .68 .67 .70liA 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00IIIB 0.97 .98 1.02 .98 .98 1.28 1.30 1.05 1.05IVA 1.61 - -- -- -- -- -- -- 1.61

a. Male HSG/HSSR only.b. Probable anomalous result due to poorly determined AFQT conversion

table for ASVAB forms 6 and 7 in the upper percentiles [8].

-7-

The estimates of relative average recruiting yields are compared tothe relative aptitudes of the target population groups in table 7. Thedata indicate, for example, that the mean AFQT percentile in AFQT cate-gory I is 96.5, which is 1.6 standard deviations above the mean of thepopulation. Male HSGs in this category are estimated to have a popula-tion yield that is only 31 percent of those in category liA.

Table 7. Relative average recruiting yield by recruit aptitude level

AFQTpercentile Standard deviation

AFQT of mean percentile Relative averagecategory Range Mean from population mean recruiting yield

I 93-99 96.5 +1.6 .31II 65-92 79.5 +1.0 .70liA 50-64 57.0 +0.2 1.00IIIB 31-49 40.0 -0.3 1.05IVA 21-30 25.0 -0.8 1.61

a. (Mean AFQT percentile - 50.0)/28.9.

Viewed in the context of table 7, the results do not seem unreason-able. Individuals slightly above or slightly below mean aptitude arejust as likely to enlist. The recruiting yield at about one standarddeviation above the mean aptitude is 30 percent lower than that at themean, whereas the yield in category IVA, which is nearly one standarddeviation below the mean, is 60 percent higher.

The results in tables 5 and 6 are expressed in recruits per popula-tion. Another way of interpreting the findings is to calculate the sizeof the population necessary to recruit a given number of accessions inan AFQT category. For example, from table 5 it can be estimated that 57recruits would be expected from a population of 1,000 category liA maleHSGs. To get the same number of accessions from category I, the popula-tion needs to be 3,167, or 3.17 times larger. This approach provides arough estimate of the relative difficulty of recruiting in the highestAFQT category.

It must be stressed that this analysis is based on the assumptionof undifferentiated recruiter yield--an assumption that seems reasonableand is supported by the analysis. However, the analysis also indicatesa substantial benefit could be gained if some way could be found to dif-ferentiate recruiter efforts more accurately and thereby enhance theability of the services to recruit upper-mental-group individuals.

-8-

EXTENSION OF THE RESULTS

The results presented in table 7 are given in smaller intervals

than has generally been possible and should be useful as is. However,estimates for even smaller intervals of aptitude would be most useful to

the job performance project. These could easily be done. The procedure

described here is only limited by the sample size available in the ref-

erence population base. The existing base (approximately 2,000 cases)

should support calculations in intervals of 10 percentile points.Applicant data from the years 1976 through 1980 would be particularly

appropriate for addressing yields of low aptitude personnel. Individualservices will, of course, have different aptitude ranges where theessential assumptions hold and may have different relative yields. Itmay also be possible to use these results to calculate recruiting costs

for different aptitude mixes, although this requires making some assump-

tions about the effectiveness of recruiters.

Another potentially valuable extension of this analysis would be touse these results to calculate recruiting costs for different aptitude

mixes. This requires making some assumptions about the effectiveness ofrecruiters, but may provide more useful information than estimates basedon the traditional econometric approaches. For example, if recruiterswere able to differentiate among individuals with different aptitudes, arough appoximation of the relative cost of recruiting an individual in

AFQT category I versus category liA would be the reciprocal of therecruiting yield of 0.31 shown in table 6. This gives an estimated rel-ative cost of 3.23, which is likely to be a slight underestimate of thetrue cost because it assumes that recruiters are able to differentiateamong applicants of different abilities, and this is unlikely to alwaysbe true. Further refinements could be made by outlining more explicitly

the way in which recruiters operate, and how that relates to the various

yields estimated in this paper. Additional research along these lineswould make a significant improvement in the estimation of the relativecosts of recruiting different aptitude mixes.

-9-

REFERENCES

[1] Rand Corporation, R-2874-MRAL, Recruit Aptitudes and ArmyJob Performance, by D. Armor, R. Fernandez, K. Bers, andD. Schwarzbach, Sep 1982

[2] Rand Corporation, R-3067-MIL, Setting Enlistment Standards andMatching Recruits to Jobs Using Job Performance Criteria, byR. Fernandez, Jan 1985

[3] CNA Research Memorandum 87-54, Alternative Modeling Approaches forSetting Cost-Effect jve Qualification Standards, by L. May,May 1987 (27870054)

[4] CNA Study 1117, The Supply of Marine Corps Recruits - A MicroApproach, by W. Cralley, Sep 1979 (07111700)

[5] Rand Corporation, R-3065-MIL, Recruiter Incentives and EnlistmentSupply, by J. Dertouzos, May 1985

[6] CNA Report 121, Educational Quality Requirements for Marine CorpsEnlisted Personnel, by L. May, Jul 1986 (94012100)

[7] Office of the Secretary of Defense (MRA&L), Profile of AmericanYouth: 1980 Nationwide Administration of the Armed ServicesVocational Aptitude Battery, Mar 1982

[8] CNA Study 1152, A Reexamination of the Normalization of the ArmedServices Vocational Aptitude Battery (ASVAB) Forms 6, 7, 6E, and7E, by W. Sims and A. Truss, Apr 1980 (07115200)

[9] Office of the Secretary of Defense (Directorate for AccessionPolicy), Technical Memorandum 80-1, Renorming ASVAB 6/7 atArmed Forces Examining and Entrance Stations, by M. Maier andF. Grafton, Aug 1980

1. The numbers in parentheses are internal CNA control numbers.

-i-

APPENDIX A

SUPPLEMIENTAL DATA

APPENDIX A

SUPPLEMENTAL DATA

Table A-i tabulates the distribution of male high school graduatesages 17 to 21 in the reference population by AFQT category. The samplesize of 1,762 cases is modest but probably satisfactory, and it corre-sponds to the traditional age grouping used by manpower planners. TheAFQT distribution of the larger sample, age 17 to 23, is shown to besimilar, increasing confidence in the stability of the results.

Table A-i. Distribution of male high schoolgraduates in populationa by AFQT category

Percentage Number of cases

AFQT Age Age Age Age

category 17-21 17-23 17-21 17-23

I 10.5 13.1 133 283II 35.4 37.3 514 933liA 15.5 14.3 257 414IIIB 18.9 16.9 362 555IVA 7.2 6.9 172 267

IVB/C 8.0 7.3 185 289

V 4.5 4.2 139 200

Total 100.0 100.0 1,762 2,941

a. 1980 Reference Population.

A-I