AD-A39 993 A STUDY OF THE EFFECTIVENESS OF THE RRMY'S NATIONAL 1/1ADVERTISING EXPENDITURES VOLUME I EXECUTIVE SUNMARY(U)AYER (N W) INC NEW YORK 31 AUG 81 USAREC-SR-81-1-YOL-i

UNCLASSIFIED MDRAD3-79-D-BS8i F/G 5/1 NL

Eu.....illllI

I,1 213 .2L! JL6l11111-2 L. A d

MICROCOPY RESOLUTON TEST CHARTNATIOL4 B F STANDARDS-1963-A

- .I

UNITED STATES ARMYRECRUITING COMMAND

.I, Study Report 81-1 AD

A Study of the Effectivenessof the Army's

_ National Advertising Expenditures

ExecLtwe Summary

N W AYER INCORPORATED:: ,AUGUST1981 DTIC

OkEECTE

Approved for Public Release 12

I' Distribution Unlimited A

Prepared for theUnited States Army Recruiting Command

Fort Sheridan, Illinois 60037I.'J- ," % , " 1 " ' • h" d

fing, 14 Q F I iU SECURITY CLASSIFICATION OF TWOS PACE (IS... D. ate fr

REKR DOM ENTAION AGEREAD 911STRUCTIONSREPOT DcUMEITAION AGEBEFORE COMPLETING FORM1. REPORT MUMNSIRQV ACESSIO NO S PiENTIS CATALOGC NUMBERN

14. TITLE (me u.Ju TYPC OF REPORT 11PEIOD COVERED

A STUDY Or TIMf EFFEC11VENCSS Or TnI AR!IYbSFil af-NATIONAL ADIVERTISnM: EXPENITURES 27 Sept 79 - 31 Aug 81 *

vOLMM I S. PLRFORMING OR*. REPORT HUMMER J-

7. AUTNOR(qJ 0. CONTRACT OR GRANT kumUzlie-()

N W AYER INCORPORATLO MDA9O3-79-D-OOO1

S. PIRPORMING ORGANIZATION NAME AND ADDRtESS 10. POOGGRAM ELEMENT, PNOJECT. TASRAREA IWORK NIT NUMBERS

N W Ayer Inc.1345 Avenue of the. AmericasNew_ York,_NY__10105____ _________

It. CONTROLLING OPPIC9 NAME AND ADDRESS 12. RE11PORT DATE

United States Army Recruiting Coan 13 81BE F AEFt. Sheridan, iL USIO AE

14. MONITORING AGENCY kML & AODRE19S(tUilerent go CGantroifij Office) IS. SECURITY CLASS. (of Chiaetape")

UNCLASSIFIED

N/A %i.DC SIFCAINONRONSCHEDUSLE NA

IS. DISTRIBUTION &.l ATEMENT (at lte Ro0m1,

Approved for public release; distribution unlimited.%

I7. DISTRIBUTION STATEMENT (at Shia abstract entered in Disch 20,9.11 wtn from, Report)

N/A

Is. SUPPLEMENTARY NOTES

Volume I of three volumes

I9. AMY WORDS (Conginuo on reverse side it necesary &-d Identify by block number)

ADVERTISING, ACCESSIONS,* CONTRACTS, ACCESSIONS AS CONTRACTED, ARMY,RECRUITING, EFECTIVENESS, COST, MODEL, BOX-JENKINS, ECONOMETRIC,* ASVAB,EXMS, CONVERSION, LINKAGE.

TO. JAd-TnACT EL.UmOse- on revre *it P& IU.aeefP 01.,7iutp pb~e br

V~. rh~yS ddertis I115 ~L~eI~e it ~jf, weI-4 slwtaJ 10 tie coa.t -0II 1N -i IV~e 111 .111.0401 so~ 1w;" pr(ICQ~I, . A.!ve.rL is illg, Wa.a reatit~ ti) ASVAII exanv'i, , and exaimb....

w&*rL tIwn rtI1.itIJCI to iC~C(!s.jujIS.

DD , 1~5 1473 tinI 1.89 1 Ut I t W 6 b 4161%01. 1 11

* * %

The U.S. Army Recruiting Cesmand (USAREC) has asked N W Ayer Incorpor-

.iI.

ated to study the effectiveness of the Army's national recruitmentadvertising. N W Ayer's Marketing Services Department undertook thistask in September, 1979, with guidance from USAREC's Program Analysisand Evaluation Division. In addition, their assistance in acquiring,providing, and checking data and data sources was essential.

Volume I Is the Executive Sumary and Is Intnded for the generalreader who wants an overview of the project's objectives, methods,and key findings. This summary highlights the marketing andfinancial aspects 6f the analysis.

Volume II is the Main Report and is intended for the reader who wantsto fully understand the details of the project: its inception, method-ology, data, results, validation, and economic Implications.

Voles III is comprised of Appendices intended for the specialist whowants to thoroughly analyze the methods and data used In the analysis.iA step by step description of bow the model was built is documented inthe Appendix entitled "Essential Elements of Analysis."

4At our request our methodology and conclusions have been reviewedU Lby Professor Martin X. Starr of the Graduate School of Business of

Columbia University. He Judged our statistical procedures soundand the conclusions acceptable on a statistical and analytic basis.

The findings in this report are not to be construed as an officiaDepartment of the Army position, unless so designated by other authorizeddocuments.

rer

k] It"aLI -....... *

lv - i t i O JJ

A STUDY OF THE EFFECTIV =ESS OF_' THE AMY'S NATIONAL ADVERTISING EXPENDITURES

VOL. IEXECUTIVE SUMMARY

1, OEJULTS ......... ..... o................. . .....55S**5*****s ********** . ..

R~~KE INDGSoo.......... .... ....... ............ ........................31

Advertising Effects Differ by Target Group ............. ... 3Effects of an Extra $1,000,000 of Advertising.............3

.. Relative Cost-Effectiveness of Advertising.......... ....... 3Effect of Terminating the GI Bill.........................6Effect of Chane In Total ObjectLves............. ........6Exams to Accessions Conversion Rate........................6Model Reliability and Validation...........................7

CONCLUSIONS..*...*...*** ... * 0e0eo.e .e e *.*......................... 7

1 VOL. IIMAIN REPORT

ITRODUCTIONSection I. MBI..... es..se....55ee .. e .sl

Section I. HISTORY OF THE APPRDACN.............................. .1-1Section III. GENERAL METHODS.................• • • • ............ 1-2

Selection of a Criterion. .............................. 1-2Zxperimentatlon or gonometrics.•........*.............1-3

* Section IV. OTHER STUDIES OF MILITARY RECRUITMENT ADVERTISING ...... l-4The GAO Study ....................................... 14

1. The Attitudes and Image knaysiso ............... 1-42. The Leads Analyais..............................1-4

The Morey and McCan Study.......... .... ee........1-5Section V. ASSUMPTIONS........... ......e.. .. e..............s. 1-5Section VI. SCOPE AND LIMITATIONS..... ...................... .. sos.1-6

CHAPTER, 2METHODOLOGYSection I. IDENTIFICAITON OF THE BASIC PROCEDURAL STRUCTURE ....... 2-1

The Process Flow...................................... .21Choice of a Two-Stage Model Building Procedure.........2-2

Section II. IDENTIFICATION OF ASSUMPTIONS AND HYPOTHESES ...... .... 2-5The Number of Ezam Models,.............................2-5The Form of and Variables in the Exm Models..........•2-5The Number of Accession Models....... ...9. .e.e...... 2-6The Form of and Variables in the Accession Models......2-6

Section III. INSPECTION OF THE DATA ON EXAM-TAING.............e2-6Section IV. ESTIMATION OF THE IMMEDIATE AND DELAYED EFFECTS OF

VARIABLES OTHER THAN ADVERTISING ON EXAM-TAKING .... 2-7Sec tion V. ESTIMATION OF THE CURRENT AND DELAYED EFFECTS OF

ADVETSING........ ..ee se .. ..s........ . .......... e se. 2-8Section 71. JOINT ESTIMATION OF THE EFFECTS OF THE ADVERTISING

AND EON-ADVERTISING VARIABLES .......... . ........... 2-8Section VII. LINKING EXAMS TO ACC CESSIONS.......................2-8

-t.

Section . MEASURES OF RECRUITING PERFORMANCE ..................... 3-1. Section 11. MEASURES OF ADVERTISING EXPEDITURE ACTIVITY.. * ....... 3-i-1

Section Ill. USAREC-CONTROLJLABLE POLICY VARIABLES...................3-2Section IV. ON-CONTROLLABLE ENVIRONMENTAL INFLUENCES*.............. 3-2

CHAPTER 4RESULTSSection I. INSPECTION OF THE CATEGORY I-111A EXAM DATA ............ 4-1L Seasonality and Auto-Correlation.. ....... . ..... ..... 4-1

Adjustment For GI Bill Termination ............. ....... 4-1Section II. ESTIMATION OF THE IMMEDIATE AND DELATED EFFECTS OF

4VARIABLES OTHER THA ADVERTISING ON CAT 1-111A.X... .......... .................. ............... .4-3Section III. EFFECTS OF THE CURRENT AND DELAYED EFFECTS OF

ADVERTISING ON CAT 1-IU1A EXAMS .................... 4-3Section IV. JOINT ESTIMATION OF ADVERTISING AND NO-ADVERTISING

EFFECTS ON EA,-TA ..NG ......... ............. 4-5Final CAT I-111A Exam Model.. .............. .~~*9e~e*eeo4-5Final CAT II1-IV odxa l............... ............ 4-5

Section V. LINKING EXAMS T ACCESSIONS...* ............... ..... 4-7Accessions md Exam s 47

'. , Manpower Needs and Conversion Rtes...................4-8Impact of the Other System Variables on the Level

Final Linkage odels...................... ............. 4-9

CHATER5VALIDATIONSection I. GOODNESS OF FIT ....................................... S-1

Fitting Procedure. ............................ ...... .5-1Evaluation of it.......•.... ......................... 5-1

Section II. STRUCTURAL SENSITIVITY OF FITTED EXAM MODELS ........... 5-2Variable Exclusion... ......... .... ...... ... . ......... -2

-, Split-Hlf Analysis..... • • •. • • • • • ••.. . ..... *oS.... .5-3

CHAPTER 6ECONOMIC IMPLICATIONSSection I1. R EU ES..... .... ................................... 6-1

Implications of a 10Z Increase in Advertising..........6-3Implications of Increasing Advertising Expenditures

by $1,000,000... .................................. .6-3Advertising Costs Per Recruit......................Relative Cost-Efficiency Of Advertising ............... 6-4Increased Objectives........ ........ ....... ...... .6-5Effect of GI Bill TemLuation .......................... 6-6

'44,'

4IIa

, ...0" , . . . , , . .. ,.. . , -,.,€. , .. ,.,. , ., ,,.,.,..., ,"

VOLUNE IIIAPPENDICES

[i 5USSTA ILNS OF ANALYSIS APPENDIX

section A.1 How Does The Recruitiug Process Work? ........................ 1Section A.2 How Is The Market Set nted?.................................5Section A.3 What Key Factors Are In The Data Iase?........................6Section A.4 What Factors Are Not In The Data lase?.......................9Section A.5 What Time Frame Does The Analysis Cover?..................... 0Section A.6 Why And Now Was A Two-Stage Procedure Selected?..............11Section A.7 What Was The Exam Model-Building Strategy?...................12Section A.8 What Were the Key Findings Regarding Category I-IlIA Exams?..13

A.8.1 What Seasonality Was Found For Category I-IIIA Exams?........13A.8.2 Why And How Was Exam Data Adjusted To Remove The Effect

Of The GI Bll .? ...........- * ....... 15A.8.3 mow Do The Key Variables Other Than Advertising Impact

Exam Variation? ........................................... 18A.8.4 What Is The Pre-Advertising Model? .......................... 26A.8.5 What Effects Were Hypothesized For Each Of The Media? ..... 0030A.8.6 Now Was The Complete Category I-i1A Exam Model Specified?...42

Section A.9 Bow Did The Category XXXB-IV Analysis Compare To TheCategory .-.................................................. 44

A.9.1 What Seasonality Was Found For Category IIIB-IV Exams? ....... 45A.9.2 Category IIIB-IV Exam Model Cross Correlations With

Non-Advertising Factors ......... o ........................... 46- A.9.3 Category III5-IV Exams Pro-Advertising Model Specification...51

k,&.9.4 Category Ill-IV Exam Nodel Media Analysis - Cross

A.9.5 Category 111B-IV Exam Model Final Model Specification........61Section A.1O Bow Was The Pay Off (Accessions) Related To Exam Variation?..62

A1.1 What Are The Key Issues Regarding The Linkage BetweenExams And Accessions?,..... ... ..... ..................... 64

A.10.2 What Did We Observe In The Time Series Relating Exams

A.1O.3 What Factors Had A Major Impact On The Conversion Rate?......69- A.1O.4 How Do Variations In Manpower Needs Impact Conversion

Rate Variation?.. ............................................ 73A.10.5 Do Other Factors Impact The Conversion Rate?.................81A.1O.6 How Can The Final Linkage Models Best Be Specified?..........87

Section A. 11 Are The Models Robust?............ ........................... 92A.11.1 How Well Does The Two Stage Model Specification Track

The Actual Data? ............................................. 93A.11.2 Are The Exam Models Consistent Over Different Time Spans?...102A.11.3 Are The Estimated Advertising Coefficients Sensitive To

The Other Variables Included In The Model? .................. 1O5Section A.12 Step by Step Procedures For Using The Models................110

I UNTRANSFORID DATA SERIES APPENDICES

[ Data Appendix A. Performance Variables

Accessions by Date of Contract

|A-1 High School Degree/Category I-IlAA.2 Uitsh School Degree/Category IIIS-IVA.3 Non Degree/Category I-IIAA.4 Non-Degree/Category IIIB-IV

ASVAB Exams

A.5 Category I-IlIAA.6 Category IIIB-IV

Data Appendix B. Media Variables

Advertising Spending

B.1 Television3.2 RadioB.3 NewspaperB.4 OutdoorB.5 Direct Hail1.6 Local Advertising3.7 Regular MagazinesB.8 Special Magazines

4 ~

Media Deflators

t; 1.9 Spot Television3.10 Network Television3.11. Spot Radio3.12 Network Radio3.13 Newspapers3.14 Magazines

, 1.15 Outdoor

Data Appendix C. Policy Variables

C-1 Recruiter Accession ObjectivesC.2 E1 PayC.3 ecruiter Numbers

Data Appendix D. Environmental Variables

D.1 Youth (16-19) UnemploymentD.2 Civilian Minimum Wage

ADJUSTED OR TRANSFORMED DATA SERIES APPPENDICES

Data Appendix S. Performance Variables

Accessions by Date of Contract (the December 1976 GI Billdetermined spike removed from the data)

1.1 HihSchool Degree/Category I-lIIAE.2 High School Degree/Category IIIB-IV3.3 Non Degree/Category I-lIIA

ASVAB Exams (the December 1976 gI bill Determined spikeremoved from the dataM)

Z.4 Category 1-111AU .5 Cateogry 1115-IV

Data Appendix F. Media Variables

Advertising spending (deflated to constant Dec. 1978 dollars,,and adJusted to reflect net spending levels throughout)

F.1 TelevisionF.2 RadioF.3 Newspaper7.4 OutdoorF.5 Direct HailF.6 Local AdvertisingF.7 Regular Magazines7.S Special Magazines

Aggregated Advertising Spending (aggregation is over deflatednot dollars)

P.9 Total Media at Current periodF.10 Total Media Sum of Periods T-4 through T-11P.11 Television (lagged 1 period) + Local Advertising

+ Newspaper

F-12 Total Media - Sum of Current Period Through Period T-53Data Appendix 0. Policy Variables

G.1 Recruiter Accession Objectives - Twelve Month CenteredMoving AverageII

G.2 Recruiter Accession Objectives - Ratio of MonthlyObjectives to Moving Average

G.3 Relative Pay - the Ratio of El Pay to the CivilianMinimum Wage

..

F A STUDY OF THE EFFECTIVENESS OfpTHE ARMY'S NATIONAL ADVERTISING EXPEDITURES

VOLUME IF ~EEESUM &RY

LIST OF TABLES

TABLE 1. Eetiuated Avertising Coterexi............TABLE 2. Relative Cost-Efifectiveness of drtsn.........5C ~ ~~TARLE 3. Effect of Terminating the GI il.............6TABLE 4. Effect of 101 Increase In Recruiting Objectives ..............*6TABLE 5. COnversionRate: Long-Teru Averages (Monthly) ...............*7

LIST 01 FIGURS

FIGURE 1. The Models In Flow Chart Form ... ... ~*.**S*~*.55FIGURE 2. An EXtra $1,000,000 Buys 800 Qutality Accessions ........ e..s*4

A STUDY OF THE EFFECTIVENESS OFTHE ARMY'S NATIONAL ADVERTISING EXPENDITURES

LIST OF TABLES

I VOLUME II- MAIN REPORT

TABLE 4-1. Mental Category I-IlA Exam, Model (Pre-Advertising)e........4-4TABLE 4-2o Mental Category I-IIlA Exan Model....o.....................4-6TABLE 4-3. Mental Category IIIB-IV Exam Model...........................4-6TABLE 4-4. USDG/CATI1-11IA Accessions -Logarit~ms.............-1TABLE 4-5. BSDG/CAT IIIB-IV Accessions - Logarithus.....................4-1OTABLE 4-6. *HSDGICAT I-lIIA Accessions - Logarithms ... .................*4-1O

TABLE 5-1. Goodness of Ft.... ...................-

uTABLE 6-1. Incremental Accessions Due to Increasing Key Factors105leyond Their 1980 Levels ......................... 6-2

TABLE 6-2. Advertising Costs Per Recruit .................. 6-r ~~TABLE 6-3. Relative Cost Efficiencies .....................-

TABLE 6-4. Effective of Increasing Total Objectives by 1OZ ..............6-5TABLE 6-5. Effect of Terminating the GI Bill........... e..............6-6

3

A STUDY OF THE EFFECTIVENESS OFTHE ARMY'S NATIONAL ADVERTISING EXPENDITURES

I. LIST OF FIGURES

VOLUME II

HAIN REPORT

FIGURE 2-1. Process Ovriv..... 9............. ..... 2-3FIGR 2-2. The Two-Stage Pro.......... *. .. ..... . ...... .....-

FIGURE 4-1. Category I-IlA 9... ...................... .4-2FIGURE 4-2. The Linkage Strutu..... . .. .. . ......... ..... 4-7

FIGURE 6-1. Effects of an Additional $l.OtO( In Advertiqing ...... s.e...6-3

4

EXECUTIVE SUMMARY

OBJECTIVES

The principal objective of this modeling project was to measure adver-tising's impact on recruiting in terms of both quantity and quality ofrecruit@. The secondary objectives were to determine the importance ofthe other major factors and to analyze the impact of media alternatives.

RESULTS )

-The project has been successful in that these objectives have been met:advertising payout has been measured according to target group, theimportance of six other factors has been quantified, and although avail-

7. able data was limited, some indication of media differences was developed.

METHOD

a The methodology employed followed standard econometric practices. Alist of measurable variables was developed and data concerning thosevariables was compiled. The analysis covers the five-year period of1976-1980.-

Exam-taking by prospects was chosen as the criterion against which wedirectly measured advertising's effects because exam-taking occursfairly early in the recruiting process: after the first contact withthe recruit, but before the critical negotiating session between theprospect and the career counselor.

We next built models that explained the effect of the GI Bill, youthunemployment, seasonality, the minimum wage, starting pay, recruitingobjectives and the number of recruiters. Two exam-taking models werebuilt: one for the smarter than average prospect (Mental Category I-IliA) and a second for the less smart prospects (Mental Category -IIIB-IV).

Once most of the variations in exam-taking were explained by thesefactors, we anaylzed the effects of advertising on the residual exam-

i' taking. Each advertising medium (TV, radio, etc.) was examined one at atime.

The advertising and non-advertising effects were then jointly estimatedand validated. Stage I in Figure 1 summarizes these models in flow chartform. Stage II of the analyses involved linking exams to accessions.Accession records were reviewed, and each accession was allocated to themonth in which the enlistment contract was signed. This is effectivelya contract series, exclusive of those contracts that do not result in

afinal accession. To our knowledge, this is the first time that recruitingperformance has been measured using a criterion so far "upstream" in therecruiting process.

- - - - 2 -

'I

L ----STAGE I- -STAGE II--s-'-------. ... . I

Mot Controllable: I

I-season&ltit IGI BGill

I unemploysment Iuinirnm wae I

IK1EMS I I ACCESSIONSI I LINKAGEI 1(as conztracted)I

I ~CAT I-lIIA - III I IHDELSI I HSDGICAT IIIv-IV I I NHSD I

.- __I_______________: _I

s. starting pay II • recruiter forceI • advertising Itotal obJectivesl

Figure 1. The Models in Flow Chart Form

--; ': ' ' . . . . . ,.I. . . . . . . .. . . .

1ZY FINDINGS

The key findings of the study are summarized below. In arriving at thesefindings we assumed that omitted variables were not dominating influences,that past (1976-1980) relationships will persist, and that all factorswill stay relatively close to their observed ranges.

Advertising Effects Differ by Target Group.

Advertising effects differ by mental group. The number of recruitsfrom each mental group that access as a result of advertising varies,as does the time lag between their exposure to advertising and their

Cactual accession.

Some mental category 1-11iA prospects are affected by TV, newspaper and localadvertising almost immediately (in the month of exposure or one month later),while others demonstrate a delayed response, generated by all measured mediaused in the campaign.

f7 Mental category 1113-IV prospects are affected by all media in thecampaign up to five months after Azposure.

Effects Of An Extra $1,000,000 Of Advertising.

Advertising effects on CAT IIIB, NHSDG contracts were assumed to be- limited because we are offered more of such prospects than we wish to

accept.* Therefore, the effect of an extra $1.0 4 in advertising fundswas calculated considering the three remaining groups: I-IlA SSDG,II3-IV HSDG, and I-i11A NHSDG. Figure 2 shows what an extra $1,000,000in media expenditures would have bought in 1980, and Table 1 shows thecorresponding cost per recruit.

Relative Cost-Effectiveness Of Advertising.

Advertising Is a more cost-effective way of increasing quality accessionsthan increasing pay or increasing the number of recruiters. This conclusionfollows from an economic analysis of what would have happened if we had in-

-creased each factor In the model, one at a time, to 1OZ above its FY 80 level,while holding all other factors at their actual FY 80 levels. The amount andcost of each 101 change, as well as the resulting incremental accessions,are seen in Table 2.

Although a 10% increase in starting pay would net twice as many recruits asincrementing advertising by the same percentage, it would be much more costlythan increasing advertising. An increase In advertising is 10 times ascost-effective as an increase In starting pay, and is twice as efficient as

'increasing the recruiter force (for detail, see Chapter 6 In the main report).

*Our objective was to measure the effects of advertising on quality acces-*sions; therefore, we spent little time analyzing the IIIB-IV NSHDG group.

This group was considered to be more constrained by the demands of the Armythan by the supply of candidates, because accession variation reflects ad-ministrative pressures as much as the response to marketing efforts.

-. -I' -. . - . . -- I . -. - -

I-~l434

0300

Fisure 2. An Extra $1,000,000 Buys 800 Quality Accessions

STable 1. Istleated AvringCost Per Recruit

Adveost per

Masrket Sejpent Recruit

1CATT -TI ,ESDG $ 3,300MY RI SDG $ .1500AYRSDG o.r CAT I-TZ1A, MSDG $ 1,250

'Ni

tQ. I

Table 2. Rulative Cost-Effectiveness of Advertising& (Impact of 10 Changes)

Incremental Accessions

"' [ CAT I-IIIA,

Amount of Chaue CAT 1-i11A any NHSDG or any________ cost (_) _S HSDG _ s_ _

Ij Unemployment 1.7 points - 2,400 4,200 5,200

iNiumm Wage +31i/hr. - -2,300 -4,400 -5,400

Starting lay $44/mo. $72* 2,300 4,400 5,400

"-crulter Force 533 bodies $16** 2,200 5,842 6,740

Total Objectives 1,100/uo. - -1,800 -5,200 -5,000

Advertising - $ 3 900 2,000 2,400

(in 1980 dollars)

* Starting Pay Budget - $720.OMN/year**Recruiters each cost $30,000/year.

q

aI.,

I

~~1

Effect of Terminating the G.I. 3111.

The G. 1. Bill was an Important recruiting tool, the termination of whichhad a significantly negative Impact on accessions, as Illustrated inTable 3.

The Impact of restoring the 0.. B111 cannot be projected, however, sinceits termination coincided with major policy changes and budget cuts thatalso effected accessions. In addition, the progra's 30-year history anduniversal awareness further complicate any attempts at predicting theeffects of restoration.

IiTable 3. Effect of Terminating the G1 Bill

(by Market Segment)

Market Segment Effect on Accessions

CAT I-lilk, USD0 - down 11,000/yearAny USD0 - down 22,000/yearCAT I-IIUA, URSDt or

any USDG - down 27,000/year

Effect of Change in Total Objectives.The model shows us that increased "pressure" via increased quantity objectives

produces more exams, but fewer quality accessions, and has historically,therefore, been costly and counter-productive.

Table 4. Effect of 102 Increase in Recruiting Objectives

Market Segment 2 of change in quality accessions

CAT I-IIlA, USDG - 51Any UN -101Any USDG or CAT I-11lA, NISDO - 51

Exams to Accessions Conversion Rate.

In the course of the linkage analysis, we noted a fairly stable exams-to-accessions conversion rate, as Illustrated below. An increase in this con-version rate may well be the most cost-effective way to Improve quantityand quality simultaneously. As detailed analysis Is outside the scope ofthis contract, we suggest additional study.

A= 006 i '' . -. . ". .'."

I. (J -7

Table 5. Conversion Rate: Long-Term Averages(monthly)

Knowne Estimated* Estimated*- US Status RS Status Contracts "Conversion"

CAT I-I1A 7,000 USDG 50Z 60Z 2,620 632

USDG 50Z 40Z 1,570 56Z

CAT 111B-IV 13,800 RSDG 34Z 40Z 3,300 60Z

USDG 661 60Z 3,000 N/A

*Note that RSDG status cannot be determined prior to accession because

data systams collect "years of education completed."

Model Reliability and Validation.

The model passed all of the standard tests with good results. We areconfident that the major process dynamics have been correctly described.Statistical tests of the model validity have been successful: 85-90Zof variances have been explained, there is no auto-correlation inexams residuals, and all "t" values are in excess of 2.0.

.. In addition, variables have been systinatically excluded to double-checktheir impact, and a split-half analysis has been performed to measure the

* stability across time. The results show that this model is both reliableand stable and can be used for forecasting purposes.

CONCLUSIONS

All factors studied had statistically significant impacts. The impact of adver-tising is delayed and varies by mental category. Advertising was found tobe the ost cost-effective tool studied (within the range of the data).

i We also found that the effects of restoring the G Bill are not projectable,and that Increasing total quantity objectives puts pressure on recruitersthat is counter-productive in terms of quality. Finally, we found that themodel's reliability is good.

~t1*

.~ 0 f4~t4%4~ *0 V ~ *~ a~' ~1s~

14s. ' Ji'Y'Va. ~ .~a *t. ~

't r~'*~~% ~ t *60

' 1. w!!0 1 ~ ~.W' 4 A

' ~ P1 '

~ '

'7 ) 4' 4~'*'

-. 'r% tr44,

- .4, 4. ~.1*, rM?.4.i)t&%'4144.A '4

'Sr."

,,tI,. ; 4 w; 4 ~ -~

t~,Xtoa. ~sjA~~~ *''4'fl4' 4 'C '4

'"'4 A

46 $$.4. 7 4 tth4 ~ L ~j4~S4 C'

"4'

- 4 4 ~ ,r$9 $yt.Y2IHj rii~1i~~:'!r.4. Sfl44S 4fr4~A~'rJj

4 '-4 A

'$4. rt' 4

~" ~d? ci,? 44 '~"f~'~'h.a)$$A

4-. 44I~

~ +'~~4*anaa~wadbau(. a.hflScA..tt .4. .~At 4aL*aSL.z a.' 4

~ __~ ',t1S

4. fl'4 K4 1S 44~44 J -* s..~ - a44 US A

A4A(t~ 4.F4' ~.-,Z

~44*44"

WW7~P~,rt4)~ ~ 4, 4

'44 in44k~Yf 'Al' 4 ~ 4/V ~"'

~ P 4~T ~f, 4 t~ 4 Y$ '.~ ;'~ '4 -

-~ - **4, ~

~4. ~i - 4 S~ 4 L~tV 4$4 -~' b ~ ~ta'% ~$L~4 A3.,,,$EtM '~ b V .~ A 4%~V. ' 4

~ i-.k*% 149.44 4~tC 4~*U2 /r~tt. '7

'4" ~'

ii

LA'** &LsItI

fl~ 4K4

Is .ar 2 . - ~ "