Recruiting Efficiency and Enlistment Objectives - Defense Technical ...

6 downloads 376 Views 3MB Size Report
CNRC Code ?0. CNRC Code 22. OP-13. jP-135. OP-135D. OP-91. OP-91D. OP- 913. OP-OB. OP-01B7. ASSTSECNAV MRA. DASN(M). USNA, Nimitz Library.
CRM 87-181 (Revised) / February 1988

RESEARCH MEMORANDUM

CD

RECRUITING EFFICIENCY AND ENLISTMENT OBJECTIVES: AN EMPIRICAL ANALYSIS Timothy W. Cooke

O3TIC ELECTE JUN 131988

A Division of

[,

V.*

Hudson Institute

-

_

CENTER FOR NAVAL ANALYSES 4401 Ford Avenue • Post Office Box 16268 • Alexandria, Virginia 22302-0268

I

"I2'--L•fON -' .•

10.8)

Total

Efficient (STAT = 0)

27

25

4

56

Low (STAT < 4.8)

14

10

17

41

High (STAT > 4.8)

15

8

44

67

Total

56

43

65

164

Contingency coefficient

e.464

X2 statistic = 44.

88a

a. The nuil hypothesis is that the contingency coefficient is zero. The x test statistic is used. It is compited as 2 SX

r Zi

k z

2

(Oo - Ei i

,

where 0. and Elg are, respectiiely, the observed and expected 1 number ofJobserva z ns in the ith row and Jth column, r is number of rows, and k the number of columns. Under the null hypothesis the expected frequencies for each cell is computed as the product of the two common marginal totals divided by the 2umber of observations. The null hypothesis is rejected if observed x is larger than expected for a given significance level.

-30-

II-|

remaining 65 observations with high inefficiency. (statistical) classification also has

The alternative

56 efficient observations,

although they are not the same ones. Of the 56 observations gauged ef'ficient with the DEft approach, 27 are also gauged efficient with the

statistical approach. Of the remaining 29 efficient in the DEft approach, 1A4 were classified with a low level ot Inefficiency In the statistical approach. For the 67 observations with high levels of inefficiency based on the statistical measure, 15 are labeled efficient with DEft, 414 of the 65 are Classified as having a high level of ir~efficiency with both approaches. In 19 of the 1614 cases (12 percent), the two methods disagree significantly with efficient performance by one

measure being labeled highly inefficient by the othier. Wh~en the DEft method classifies a district's performance as (highly) inefficient, it is unlikely that the statistical approach will find the district effi-

cient. This is not true to the same degree whei the statistical approach identifies a district as (highly) inefficient. The contingency coefficient given in table 8 is One measure of association between the two efficiency criteria of the table. The

coefficient is positive and significant. This suggests that the classifications based on the two measures are not independent and provides further evidence of the similarity In the way the two approaches gauge efficiency for the sample. CONSISTENCY OF CONTRACT OBJECTIVES With information on district goals or contract objectives, the DEA approach can be used to provide additional information to planners. By

comparing the contract objectives of similar districts, the consistency of the allocation of objectives across districts can be assessed. Typically, recruiting performance is evaluated on the basis Of goal attainment. Combining the relative measures of technical efficiency with observations of goal attainment provides additional information for more efficient allocation of contract objectives and resources across districts. This section and the next analyze three measures of perforrance. The Table 9 lists these measures for each district and fiscal year. first measure, labeled TE, is the DEA measure of technical efficiency. This measure is similar to the DEA measures discussed in the previous section in that the same output (total A- and CU-cell contracts) and set

?

1. This may be due to the "global" nature of the statistical measure, ~and the "local" nature of the flEA measure. DEA compares only similar districts; the statistical approach extrapolates across districts. 2. Data an contract objectives in 1981 were not available, and, therefore, 1981 is excluded from this analysis. Furthermore, since the data are not pooled across years, analysis of 1983 recruiting is included. -31-

T AB9LE TECHNICAL

~rPICENcy

(TE). COAL EFFICIENCY (CE). ANDOCOAL ATTAWNMET (CO'4AL)

1982 TE GE

*

1983

1984

RESULTS 1985

GOAL

TE

GE

COAL

TE

GE

GOAL

TE

GE

MOAL, s P

lei 102 103 104 119 161

6.6 96. 15.6 13.? 17 0 11.2

1.1 0.9 6.2 9.6 19.5 22.6

AT AT 60 AT AT AT

19.4 0.0 2.5 .3.6 2.2 8.1

6.6 0.0 6.8 1.6 9.5 11.2

AT AT AT AT AT AT

6.7 9.6 9.1 14.1 0.0 10.0

9.1 6.6 3.5 9.5 6.3 5.4

90 80 80 AT AT AT

10's 690 20.6 21.5 Is67 .34.4

4.3 9.9

4.4 9,f 6.4

K O G

.310 311 312

15.6 0.0 g.e 6.6 40.3 15'.46,16. 19.4 6.6

16.2 0.9 31.7 6.6 19.5

.0 9.0 6.0 0.6 ZA.4 t.0 9.0 9.0

1.3 9.0 0.0 9.9 16,5 6.0 0.9 06ý

AT AT AT AT AT AT 80 AT

0.0 0.6 O's 0.0 22.4 0.9 0.0 28.2

1.1 6.9 0.6 R.6 11.6 .0 6.0 9.2

AT AT AT BG 80 0.6 B 90 AT

0.6 0.0 6.9 0.a 24.1

14.6 6.6

AT AT AT AT AT AT AT AT

0. 0.0

1.!s 960 0.9 9.0 12.7 0.6 9.6 0.9

AT Fr AT AT AT BG go BC

36.5 6.4 .3.7 10.5 297 0.0

AT AT AT AT AT AT

19.3 0.0 0.0 9.7 3.0 6.6

24.0 0.0 16.5 14.6 21.4 0.0

AT AT AT AT AT AT

3.2 0.0 0.0 5.4 0.0 6.0

80G AT AT AT AT AT AT AT

9.9 6.6 6.0 6.1 6.2 9.0 6.0 0.0

0.0 3.9 6.6 5.3 10.2 6.6 6.8 9,9

AT AT AT AT AT AT AT AT

0.0 960 6.0 3.1 12.2 0,0 0.0 6.6

.31.

314 3116 348

466 409 417 418 4210 420.9 521 '424

6 4 9.6 o.0 16.3 21.1

0.0

00

0.'1 6.9 .0. 3. 0 2.6 6.6

AT 90 AT 80 AT 90

8.2 0.0 9.9 9.2 0.0 O.0

90. 6.0 6.9 960 26 6.6

GG K. 80 W 8M 80

0.6 3.2 9.6 G.6 7.3 6.6 6.6 6.6

ect AT 80 9o AT AT AT 9C

0.6 0.0 6.0 0.6 6. 0. 2.2 6l6

09. 9.6 0.0 .0. .6 .6 6. 3.6

80 AT 80 BC AT 9 8 AT

6.6 6.0 a,0

AT AT B0

6.0 6.6 90.

6.6 6.6 6.6

AT AT AT

5219 547 559

691 6.6 6.6

6.6 2.2 6.9 5.4 279 6.6 6.6 6.6

7310 7,11

6d, .0 00 6.6

6.6 669 6.0

AT AT AT

6.0 6.0 0.9

-0.0 0.6 6.6

AT AT AT

732 7331 734 '14f

6.6 06 445%7 0.6 9.9

08 .6 9.3 6 6 6.6

AT AT AT AT AT

9.7 8.9 16.4 6.9 6.0

11.3 0.6 6.3 6.6 6.0

AT AT AT AT AT

2.6 12.4 7.9 906 6.6

9.6 3.9 6.5 6.6 6.0

AT 9G 00G AT AT

6.6 164 8.9 6.6 6.6

6.6 6.6 6.6 9.6 6.6

00G 90 AT AT AT

.6

16.6 0.0 6.6 4.9 8 29.7

6. P 6 0 6.6 3.2 13.9

AT AT AT AT AT

11.6 0,9 6.6 6.3 8.4

6.6 6.6 6.8 6.6 e.g

AT AT AT AT AT

89.7 6.0 6.6 6.6 6.6

1.9 6.6 6.6 0.6 6.6

BC AT 80 AT AT

6.6 6.6 6.6 6.6 6.6

2.2 6.6 6.6 6.0 3.2

AT AT AT AT AT

.0, 3. R 529.6 520.0 563.8

747

*

2

837

8.3A R84

-32-

0.0 .0. 6.6

of inputs is specified. There is, however, one Important difference between the measures. The TE measures discussed in this section are computed for each year separately. The DRA measures In the previous section were computed for the pooled (by year) sample. Thus, the efficiency of a district was gauged relative to all other districts operating in the same year and in other years as well as to itself operating in a different year. The TE measures In table 9 gauge the performance of a district relative to only the performance or other, districts operating in that same year. The second performance measure listed in table 9, labeled GE, Is a DEA measure or goal efficiency. This measure differs from the TE measure in that the output specified is the active-duty contract objective. Thus, GE is the ratio of the maximum potential contract objective to the actual contract objective. The maximum potential contract objective is determined by the contract objectives of other similar districts. The third variable in table 9, labeled COAL, is a binary variable that compares actual production to the goal. If the district attained or exceeded the contract objective, then GOAL a AT. Otherwise, GOAL a BG. The contract objectives and recruiting resources (e.g., number of recruiters) are allocated among areas based on past production levels and the estimated relationship between past production levels, the level of resources, and indicators of the economic and demographic conditions. The net result should be that two districts operating in similar recruiting markets with the same number and type of recruiters should have similar contract objectives. Consistency of objectives depends on the appropriate appraisal of the recruiting environment. The ability to predict changes in the environment due to business cycle swings is also required. The goal efficiency measures (GE) in table 9 are used to examine the consistency of the allocation or the contract objectives across districts. The measure gauges a district's actual contract objective relative to the maximum potential contract objective determined by the other districts in that year. If the measure is 0.0, the district's objective is consistent in that It is at least as large as other similar districts' objectives. On the other hand, if the measure is greater than 0.0, the contract objective is low relative to the contract objectives of other districts operating in a similar recruiting environment with similar resources. Note that a value of 0.0 of the GE measure does not necessarily mean that the district's contract objective was set efficiently, only that relative to the other similar districts, that district's objective was the maximum objective specified. If each district's contract objective is determined solely by the characteristics of the recruiting environment included in the analysis, then the GE measure should be zero for all districts. An ordinary least squares regression of contract objectives on the five

-33-

-__ ......

characteristics of the environment for each fiscal year explained a minimum of 93 percent of the variation in the contract objectives. The .. fact that the variables used to compute GEeplainjnuofO-the-cross .... section variation in contract objectives suggests that the characterization-of the recruiting -environment -used- to-compute the -GE ie-as-ures-it

....

similar to that used to set contract objectives.

Tables 10 and 11 summarize the GE measures given in table 9. Table 10 gives the number of inefficient observations by fiscal year and, area. The most apparent change is the decrease in the number of districts whose goals were gauged inefficient between 1982 and 1985. In 1982, over half of the districts (22) had contract objectives that were low relative to similar districts operating that 'year. Three years later, most districts' contract objectives were "in line" with other similar districts; only 9 districts had lower goals relative to other districts in that year. In-each year, districts with inefficient goals were found within each area, although area 7 has the smallest proportion overall. Area 1 has the largest incidence of inconsistent (low) objec-

tives. TABLE 10 NUMBER OF OBSERVATIONS WITH INEFFICIENT CONTRACT OBJECTIVES BY YEAR AND AREA Area (number of obs.)

1982

1983

1984

1985

1

5

5

5

4

5

2

3

2

4 (6)

5

4

3

1

5

3

4

2

0

1

2

3

0

3

0

1

2

22

17

17

9

(6) 3 (8)

(8) 7

(8) 8

(5) Total

-34-

A __

-1

TABLE 11 DISTRICT ACTUAL AND EFFICIENT CONTRACT OBJECTIVES BY -YEAR ....... (Where Different)

1982

1983

1984

1985

District

Actual

Efficient

Actual

Efficient

101

2,809

2,857

2,226

103 104 119 161

2,988 3,678 2,896 2,596

3,173 4,031 3,461 3,183

2,312 3,003 2,246 2,100

310

1,951

2,267

1,759

312

2,081

2,741

314

2,124

2,538

315 316 348

2,471 1,878

2,590 2,141

406

2,184

2,981

409 417

2,963 2,662

3,153 2,760

1,849

2,154

418 420

2,858 2,245

3,158 2,912

2,219 1,657

2,543 2,012

2,261 1,843

2,329 1,891

524

2,255

2,305

1,883

1,956

1,697

1,751

527

1,984

2,091

1,538

1,620

528

2,050

2,622

1,872

2,063

1,610

1,719

1,560

1,736

--

542

--

732 733

--

--

.... --

734

1,954

2,136

836

3,646

3,668

839 840

2,576 3,368

2,658 3,836

2,555

2,875

Average [Average difference] No.

--

[320]

inefficient

Average TE

Actual

Efficient

Actual

Efficient

2,411

1,951

2,129

2, 132

2,224

2,465 3,051 2,459 2,335

2,352 2,678 2,023 1,893

2,434 2,932 2,150 1,995

2,996 2,138 2,066

3,128 2,348 2,198

1,782

1,680

1,698

1,711

1,737

--

--

1,707

1,989 --

--

--

--

1,569

1,742 ---.

--

......---

--

1,722

2,135 --

--

--

--

1,771

1,766 --

--

--

--

1,943

2,129 £187]

22

--

....

1,905

1,907

....

--

--

--

--...

.... 2,049

--

--

1,967

1,833

....

--...

--

1,344 1,586

1,473 1,648

.... ....

1,430

1,552

--

2,962

3,018

--

1,907

--

3,155

3,224

--

--.--

--

2,851

2,942

2,315

2,429

2,009

--

[114] 9

5.6

''

2,106

--...

--

5.3

-35-

..fw-n ltM nwat~t SSflm .t

---

..

-

17

10.0

1,953

--

1,536

4

--

1,733

1,407

-A

_

--

[102]

17

13.4

-

--

.

For the districts classified as having inefficient contract objec-. tives, table 11 lists the actual ob ective and the efficient or maximum objective determined by the method. Also included are the average of these variables and the average of their difference (maximum-contract objective minus actual contract objective). In 1982, the average difference is about 320 contracts. In 1985, when contract objectives were set more consistently across districts, the difference between the maximum and actual objectives for the 9 inefficient districts is, on average, 120 contracts.

1. Recall that the GE measure is one minus the ratio of the maximum objective to actual objective. -36-

--

EVALUATING OBJECTIVES

Despite the fact that efficient contract objectives are imperfectly known and goals are sometimes inconsistently set across districts, the achievement of contract objectives plays an important role in evaluating district performance. The ability to attain the contract objective depends on whether contract objectives Fire consistent with resources allocated to a district and recruiting conditions. A consistent allocation of contract objectives is obtained only if the market conditions facing each district are correctly evaluated. Districts not achieving their contract objective may nevertheless be as productive as similar districts that achieved their goal. Conversely, some districts that achieve their contract objectives may not

be as productive as similar districts that did not.

Using this informa-I

tion on the relation between performance and goal attainment may be valuable in allocating changes in district objectives. IF

To illustrate the use of comparing the two performance measures, the observations are divided into four groups based on goal performance and the technical efficiency measure. Table 12 gives the number of districts classified into each of four groups by year. In 1985, there are 2 districts with production above goal, but less production than other similar districts. They would have been candidates for goal increases. On the other hand, 1~4 districts were as productive as similar districts, but did not achieve goal. Their failure to meet goal is not associated with differences in productivity between them and those efficient districts that attained goal. The nine districts that were as productive and did not achieve their objectives deserve closer scrutiny.

-37-

TABLE 12 COMPARISON OF GOAL PERFORMANCE AND TECH~NICAL EFFICIENCY Number efficient

*Below

Number inefficient

Average inefficiency

1982 Attained goal (39) Below goal (2)

24 1

15 1

6.5 7.8

1983 Attained goal (40)

23

17

3.7

Below goal (1) 1984 Attained goal (22) goal (19)

1

0

0.0

16 11

6 8

3.2 4.1

1985 Attained goal (18) Below goal (23)

16 14l

2

1.8

9

5.8

-38-

I

CONCLUSIONS b

This research has examined the relative performance of different recruiting districts and the consistency of recruiting objectives during fiscal years 1981 through 1985. The results can be summarized as follow$: * The consistency of enlistment contract goals assigned to districts improved remarkably between FY '1982 and FY 1985. The number of districts with goals lower than similar districts (as defined by economic and demographic characteristics of the districts) fell by nearly 60 per-A cent. The average amount by which goals were lower fell by 64 percent. The only significant inconsistency is that area 1 districts are more likely to have lower goals than similar districts in other areas. This may reflect a relative difficulty of recruiting in the Northeast that is not captured in this analysis. 9 The improvement in consistency of goaling over this period is reflected in more similar performance among similar districts. Fiscal years 1984 and 1985 show an improvement in average technical efficiency of about 5 percentage points relative to 1981 and 1982. This may be associated in part with a desired shift in the qualityI composition of enlistment contracts between 1982 and 1984. * Districts in area 1 have been less effective in producing A- and CU-cell recruits than similar districts in other areas. * A cross-classification of districts by technical efficiency and goal attainment can be used to identify where goals should be redistributed, if necessary. Those districts achieving their objectives but not producing as many enlistments as similar districts should be considered for larger goals. On the other hand, districts performing as well as other similar districts but not achieving their objectives are candidates for smaller goals. Districts that are neither performing as well as other districts nor ach ieving their objectives deserve special attention. * Two different approaches to the measurement of Navy Recruiting District production potential, DEA and statistical, produced measurements that generally agree on the ranking of districts with respect to technically efficient production of A- and CU-cell enlistment contracts. .. 39-

When the DEA procedure classifies a district as (highly) inefficient, it is likely that the statistical approach will find a qualitatively similar result; however, when the statistical method identifies a district as technically inefficient, it is less likely that the DEA approach will concur. All analykis in this paper is based on a particular characterization of the recruiting environment. A number of potentially important factors have been omitted, including advertising and interservice competition, among others. Other research along these lines would examine the allocative efficiency of the recruit mix produced by different districts. Such analysis is complicated by the necessity to assign relative values to recruit types. Nevertheless, allocative efficiency may be more important than technical efficiency.

-40-

. 2

REFERENCES [1]

A. Charnes, W.W. Cooper, and R. Rhodes. "Measuring the Efficiency of Decision Making Units." European Journal of Operational Research vol. 2, no. 6, (1978): pp. 429-444

[2)

R. Fare, S. Grosskopf, and C.A. K. Lovell. The measurement of Efficiency of Production. Boston: Kluwer-Nijhof Publishing, Kluwer Academic Publishers, 1985

[3]

The University of Massachusetts, Department of Industrial Engineering and Operations Research, A Bibliography of Data Envelopment Analysis (1978-1986), mimeograph, by Lawrence M.

Seiford, Jan 1987 [4]

Dennis J. Aigner and Peter Schmidt, eds. "Specification and Estimation of Frontier Production, Profit and Cost Functions." Journal of Econometrics v. 13(1), (1980): pp. 1-138

[51

CNA Research Memorandum 86-110, Econometric Time Series Analysis of Navy Recruiting: 1977-1985, by Timothy W. Cooke, May 1986 (27860110)

[6]

CNA Study 1168, Enlisted Supply: Past, Present, and Future, by Lawrence Goldberg, Sep 1982 (07116801)

[71

CNA Study 1073, Recruiters, Quotas and the Number of Enlistments, by Christopher Jehn and William F. Shughart, USN, Dec 1976 (07107300)

[8)

United States Military Academy, Working Paper, Recruiting Goals, Enlistment Supply, and Enlistments in the U.S. Army, by Thomas V. Daula and D. Alton Smith, Dec 1984

£91

The RAND Cor-ration, R-3065-MIL, Recruiter Incentives and Enlistment Suppl"

)y J.N. Dertouzos, May 1985

[10]

CNA Report 121, Educational Quality Requirements for Marine Corps Enlisted Personnel, by Laurie J. May, Jul 1986

£111

U.S.

Army Research Institute for the Behavioral and Social WP 86-18, setting Battalion Recruiting Missions, by A. Nelson, C.E. Phillips, and E.J. Schmitz, Apr 1986

Sciences, MPPRG,

1. The number in parentheses is a CNA internal control number. -41-

[12] Arie Y. Lewin and Richard C. Morey. "Measuring the Relative Efficiency and Output Potential of Public Sector Organizations: An Application of' Data Envelopment Analysis." Journal of Policy and znformation systems 5(4) (1981): pp. 217-285 (13] Nancy M. Ostrove et al. youth Attitude Tradking Studyz Arlington, VA: Defense Manpower Data Center, Jun 1987

Pall 1986,

[14] The RiMJD Corporation, R-3353-FMP, The Enlistment Bonus Experiment, by J. Michael Polich, James N. Dertouzos, and S. James Press, Apr 1986

(151 CNA Research Memorandum 87-58, Indicators of Navy Recruiting Success, by Timothy W. Cooke and Robert F. Lookman,

May 1987

(16] CNA Research Memorandum 86-269, Indiviual Incentives In Navy-i Recruiting, by Timothy W. Cooke, Dec 1986 (17] D.J. Aigner, K. Lovell, and P. Schmidt. "Formulation and Estimation of Stochastic Frontier Production Functions Models." Journal of Econometrics 5(1): pp. 21-38 [18] James Jondrow et al. "On the Estimation of Technical Inefficiency in the Stochastic Frontier Production Function Model." Journal of Econometrics 19(2): pp. 233-238 (191 W.H. Greene.

LIZDEP Manual, mimeograph,

-42-

1986

-

-.

- -.

-

-

- -

-

-

-

0

J

APPENDIX A THE RECRUITING DATA SET

0'

N'

S

b

,

.-

.'.

S

0 if I.

S

I '... I

r.

A

.

9

APPENDIX A RECRUITING DATA SET This appendix contains the recruiting data set used. Tables A-1 through A-5 list all variables described in table 1 of~ the text by district and year (one table for each year 1981 through 1985). The variables are measured in the Units specified in the text. Note that data on contract goals for FY 1981 were not available, and the GOAL variable is set to aero for that year.

A-1

TABLE A-1 1981 RECRUITING DATA SET

DISTRICT ACCON 101

99

POP

UR

130730 i

URBPCT

.4

a7.9

PAY 110.9a 104.91

ATTN

GOAL

2698

00

127.14 111.29

4112 3150 3664

100.0 99.0

114.3a 109.64

2808 2763

0 0

9.4 7.8 5.8 6.1 9.0 6.4 6.2 7.0

67.3 83.2 82.3 68.5 87.3 54.1 69.8 92.3

115.43 119.82 137.00 120.74 118.90 125.86 123.30 123.30

2047 1771 2609 2233 2140 2887 1864 2926 8

0 0 0 0 0 0 0

95039 122708 121630 136218 107

8.3 6.4 9.4 87 a.6

79.4 88.1 82.5 6. 74.7

122.5a 111.74 99.4a 105.71 1.922

1957 3128 2805 2850

0 0 0 02

151

212367

12.2

83.3

95.90

4597

0

528 529 542 559

1427 1034 858 527 957 1271 1049 735

145 82 65 815 68 91 98 a2 62

203232 106395 93096 99729 151986 140208 92236 110485

8.4 8.6 9.0 5.6 5.8 5.8 9.5 7.7

92.0 58.2 45.8 58.7 60.3 42.2 63.5 68.6

92.94 102.24 107.55 116.64 108.37 107.55 100.19 97.44

2712 2195 1945 1840 1812 2135 1967 1633

0 0 0 0 0 0 0 0

725 730 731 732 733 734 746 747

762 683 661 678 774 714 508 758

71 60 68 65 68 71 63 67

90242 65835 98978 89490 102120 88943 64487 66408

5.3 6.2 4.9 4.7 5.5 8.0 6.6 8.6

68.9 56.1 81.3 86.5 53.3 69.7 80.5 31.2

93.04 110.52 94.88 97.84 107.35 102.55 102.24 124.32

1820 1629 1794 1769 1575 1675 1313 1644

0 0 0 0 0 0 0 0

1424

157

220985

6.3

98.1

101.04

3785

0

200242 120548 192504

6.1 7.8 7.6

a4.7 78.7 100.0

103 98

129011 137682

7.3 7.2

937 786 1475 A12 946 1281 623 692

75 60 119 81 79 V 62 2

73873 60774 110717 71923 93307 107254 67148 685703

406 409 417 418

1021 1419 1522 1563 18

67 91 95 101 7

422

2368

521 524 526

102 103 104

2054 1439 1571

155 104 136

119 161

1292 130a

310 311 312 313 314 315 316 348

836

a37 838 839 840

SThe

1444

RTR

1076 1820 1209 1342

94 183 91 119

127126 246403 123070 159766

8.1 7.4 8.5 6.9

61.4 91.6 70.3 88.8

99.73 99.32 98.82 109-79

2559 5570 2746 3755

0 0

0 0 0 0

contract goal data for FY81 were not available and

the value of GOAL is

zero for all distriots in that year.

A-2

TABLE A-2 1982 RECRUITING DATA SET PAY

ATTN

GOAL

87.8 84.6 76.6 100.0 100.0 99.0

117.55 111.12 134.67 117.88 121.15 116.13

2858 4352 2954 3974 2970 2666

2809 4131 2988 3678 2896 2596

12.3 9.5 7.2 7.3 11.3 8.4 7.8 8.3

67.4 63.3 82.3 68.7 57.3 54.2 69.9 92.2

122.35 126.68 145.21 127.98 126.03 133.40 130.69 130.69

2156 2124 2237 2547 2199 2899 2012 3229

1951 1908 2081 2254 2124 2471 1878 2818

93312 120527 118421 133425 107219 207353

10.1 7.7 12.2 11.1 12.0 14.5

79.4 88.0 82.3 74.6 68.8 83.1

131.29 119.69 106.56 113.23 122.64 102.73

2549 3006 3108 3056 2424 5105

2184 2963 2662 2858 2245 4660

120 76 61 68 77 85 77 61

198299 104283 91555 98007 149895 137470 90205 108540

10.3 9.8 10.7 6.9 7.3 7.3 11.0 10.1

92.0 58.1 45.7 58.6 60.4 42.3 63.5 68.5

99.56 109.51 115.20 124.94 116.08 115.20 107.33 104.37

3326 2473 1858 2036 2165 2778 2479 1895

3373 2255 1847 1984 2050 2706 2160 1742

860 827 767 805 870 ?49 666 1004

66 59 70 63 71 67 65 69

90520 65702 98438 89694 101323 87957 63405 65044

6.7 7.2 5.6 6.2 6.2 9.1 7.3 10.7

69.1 56.4 81.3 86.7 53.3 69.9 80.5 31.4

98.62 117.15 100.57 103.71 113.79 108.69 108.37 131.78

1885 1693 1824 1737 1940 1955 1640 1965

1771 1629 1719 1644 1806 1954 1558 1734

1561 1352 2161 1388 1503

126 92 153 89 116

217731 126627 245169 122049 160866

8.0 9.8 9.8 10.8 9.6

98.1 P1.4 91.5 70.4 88.7

104.80 103.45 103.03 102.51 113.88

3692 2917 5223 2917 365,1

3646 2736 4787 2576 3368

UR

DISTRICT ACCON

RTR

POP

101 102 103 104 119 161

1495 2096 1492 1776 1447 1351

102 151 100 140 95 88

128046 196209 117654 186740 125624 134106

7.0 7.8 8.9 8.0 8.7 8.1

310 311 312 313 314 315 316 348

1043 1009 1687 1196 993 1590 811 798

77 63 118 83 75 90 66 52

72964 59945 110802 71092 92370 105977 66174 65898

406 409 417 418 420 422

1378 1627 1895 1664 1281 2645

77 97 85 90 77 149

521 524 526 527 528 529 542 559

1494 1245 900 1021 1157 1328 1383 959

725 730 731 732 733 734 746 7-17 836 837 838 839 840

URBPCT

t-3

I•r!f

A1vu f.

•-.&

M•

IW, A °AI

o-

u Am•• a



MIN.-. W,.

w,.

•Nuam

-.

,ra

..-,•-

-

.

-..

-,

.A.

..

...

.

..

..

..

..

TABLE A-3 1983 RECRUITING DATA SIT

I

I•

____A-

ATTN

GOAL

118.81 109.47 132.87 116.13 119.38 114.41

2322 3401 2424 3237 2389 2187

2220

67.4 63.5 82.4 66.9 57.4 54.3 70.0 92.0

120.73 125.00 143.29 126.29 124.36 131.83 128.96 128.96

1815 1708 2678 1872 1736 2*34 1516 1854

1759 1638 2522 1838 1707 2177 1531 1815

11.6 7.1 13.8 11.9 16.2 15.1

79.5 87.9 82.3 74.6 68.6 83.0

129.83 118.36 108.38 111.98 121.28 101.59

1914 2437 2068 2357 1750 3743

1722 2360 1849 2219 1657 3652

193186 101735 89346 95815 147198 133920 87737 106116

11.8 11.9 13.9 8.1 8.6 8.0 11.4 11.5

92.0 58.0 45.6 58.7 60.5 42.5 63.6 68.5

98.46 108.30 113.93 123.56 114.79 113.93 106.14 103.22

3083 2012 1364 1568 1961 1829 1686 1605

2997 1883 1316 1538 1872 1762 1610 1591

69 62 68 70 83 75 66 66

90191 65245 98241 90324 99952 86267 62757 63003

7.2 8.9 5.9 9.9 9.5 11.9 9.1 12.4

69.1 56.6 81.4 86.8 53.5 70.0 80.6 31.6

97.31 115.59 99.23 102.33 112.28 107.25 106.93 130.03

1696 1468 1620 1685 1918 1818 1258 1583

1635 1440 1612 1560 1835 1766 1211 1508

1809 1544

131 92

213627 124907

9.5 10.7

98.0 61.4

103.07 101.73

3336 2276

3303 2199

2310 1520 1797

157 95 108

242555 120331 160734

10.5 11.1 10.7

91.5 70.6 88.7

101.32 100.81 112.00

3985 2429 2948

3694 2258 2818

DISTRICT ACCON

RTR

POP

UR

URBPCT

101 102 103 104 119 161

1460 2075 1724 1924 1632 1414

99 152 103 127 95 89

124990 191191 114852 180149 121712 130241

7.1 7.4 10.8 8.5 8.9 8.2

87.8 84.6 76.5 100.0 100.0 99.0

310 311 312 313 314 315 316 348

1262 1079 1767 1165 1123 1600 838 908

86 68 122 84 79 89 68 55

71611 58576 110184 69629 90592 104305 64658 65464

13.8 10.3 8.4 7.6 12.7 9.6 6.8 9.5

406 409 417 418 420 422

1303 1605 1655 1700 1363 2850

81 95 85 101 76 152

91153 117997 114658 129969 103920 201120

521 524 526 527 528 529 542 559

1925 1482 953 1018 1384 1307 1232 1151

138 84 64 70 87 83 72 64

725 730 731 732 733 734 746 747

1090 942 975 1063 1174 1057 760 1029

836 837 838 839 840

PAY

3301 2312 3003 2246 2100

TABLE A-4 1984 RECRUITING DATA SET DISTRICT ACCON 101 102 103

1214 1638 1421

RTR 87 124

POP.

UR

URBPCT

PAY

ATTN

GOA..

5.4 5.3 7.9

87.7 84.5 7,.2

-104 572

100 110

121869 186460 111220

114.89 108.60 131.62

173260

1888 1981 2745 2797 22-19 -- 2352

7.4

100.0

119 161

1395 1155

i11521

.83 79

118202 126156

2683

7-.2 6.2

100--..0 99.0

. 118.40 113.50

310 311 312

1138 1011 1648

77 68 112

313 314 315

69770 57488 109523

10.5 7.2 6.3

925 950 1312

72 71 88

67.8 63.8 82.3

68144 88301 102047

118.62 122.82 140.79

682 755

65 65

69.2 57.5 64.7

1680 1575 2462

316 348

6.0 9.3 6.9

1726 1594 2556

63734 64317

124.09 122.19 129.34

5.3 7.2

1677 1564 1992

70.8 91.8

1712 15692038

126.71 126.71

1295 1477

1534 1407

406 409

1248 1315

417

77 85

1434

88889 115488

76

79.5 87.9

126053 100575

104.46

1905 2111

90 75

82.3

1965 2064

1437 1360

10.3

128.70 117.33

418 420

110540

9.0 5.4 9.3 12.7

2074

2150

111.01 120.23

1934

422

74.6 68.4

145

194550

2137 1899

2261 1843

11.5

82.9

100.71

3278

3772

521

1649

129

187192

9.0

92.0

97.60

2753

3136

524 526 527 528 529 542 559

1136 894 852 1175 1093 992 906

71 64 62 80 67 61 56

99013 86885 93525 143650 130356 85315 103156

9.4 10.9 6.0 6.6 6.2 8.1 7.9

57.7 45.4 58.4 60.5 42.6 63.8 68.3

107.36 112.94 122.48 113.7P 112.64 105.21 102.32

1733 1483 1298 1852 1562 1506 1421

1697 1558 1387 1833 1502 1505 1451

725

836

730 731 732

62

89386

740 738 947

55 56 66

5.7

64377 97360 89349

69.6

95.61

76

113.58 97.50 100.55

1307

963

56.8 81.4 86.5

1352

733 -

7.0 4.3 7.7

943

7.7

1080 1123 1344

734

98414

1107 1040 1408

74

53.7

84422

110.32

1570

668 909

69.9

1586

746 747

9.9

57 65

62187 60968

105.38

7.6 9.9

1419

1430

80.7 31.7

105.07 127.76

1086 1501

1086 1450

836

1617

122

837

208158

1271

7.2

76

122812

97.9

102.94

8.4

2702

61.4

2962

101.61

1960

838 839 840

2089 1489 1758

145 92 106

238108 118641 159876

8.5 9.0 7.2

A-5

91.4 70.5 88.7

101.20 100.68 111.86

2678

2045&-.. 2023-.... 1914 1893

3411 2173 2868

1939

3678 2130 2844

-

..

TABLE A-5 1985 RECRUITING DATA SET DISTRICT ACCON

RTR

101 102 103 104 119 161

1096 1455 1.3221504 1210 932

88 118 -100 124 91 86

121018 184460 110106 168700 116868 123856

5.0 4.3 7.0 6.8 6.1 5.4

87.5 84.3 75.8-100.0 100.0 98.9

310 .311 312 313 314 315 316 348

1169 904 1796 662 979 1345 709 678

82 71 119, 77 75 90 66 62

67953 57603 109889 67410 .86133 100110 64913 62547

9.4 6.6 5.8 6.0 8.6 5.9 5.4 6.8

68.2 118.91 64.2 123.12 82.1 141.13 69.4 124.39 .. 57.5 -122.49 .5,2 129.65 71.7 127.02 91.5 127.02

406 409 417 418 420 422

1180 1185 1482 1390 1350 1848

81 86 82 91 78 140

87639 114471 107234 123044 98056 187817

7.9 4.8 9.4 8.3 10.9 10.3

79.4 87.9 82.3 74.5 68.2 82.8

521 524 526 527 528

1637 1264 906 873 1351

133 85 63 59 80

18.1772 97049 84566 92227 140802

8.6 8.8 10.1 5.4 6.0

128476

6.8

ý542 559

935 989

66 64

83747 100423

725 730 731 732 733 734 746 747

909 817 756 894 1042 961 690 1038

67' 61 64 63 79 71 58 70

836 837 838 839 840

1942 1327 1933 1504 1746

529

1180

72

139 83 164 92 114

POP

UR

URBPCT

PAY

ATTN

GOAL ..

114.02 107.78 130 ,63114.34 117.51 112.65

17M2 2409 2012544 1756 1566

2132 2982 . -2-565--2996 2138 2066

1791 1565 2884 1730 1741 2186 1411 1571

1711 1605 2864-... 1692 1733 2261 1565 1736

129.24 117.82 104.90 111.47 120.73 101.13

1700 1962 2199 2134 1953 2864

2188 2172 2284 2440 2049 3646

92.0 57.3 45.2 58.1 60.6

98.01 107.81 113.41 122.99 114.27

2727 2044 1519 1428 2083

3005 2022 1630 1509 2076

7.7 7.2

64,.1 68.2

105.65 102.75

1466 1628

1748 1602

89085 64087 97311 87568 97880 83305 62967 59640

5.7 7.3 4.8 8.0 7.4 10.6 7.8 9.6

70.2 57.1 81.5 86.1 53.9 69.8 80.8 31.8

95.84 113.85 97.74 100.79 110.59 105.64 105.32 128.07

1563 1320 1424 1484 1770 1790 1269 1696

1560 1282 1375 1506 1876 1762 1253 1687

203907 119835 234676 116966 159147

6.4 8.2 7.8 8.4 6.7

97.9 61.4 91.3 70.4 88.6

102.55 101.22 100.81 100.30 111.44

3210 2160 3738 2353 2852

3155 2153 3713 2332 2851

A-6

42.7

113.41

1709

1855

.

-

--

-

-

.-

wnwrnw.

-.

.

..

,

wn

1.,

7 an rn

r

an Err WrS

-. -.-

-.

-

0

.4

-

APPENDIX B DEA LINEAR PROGRAMMING PROBLEMS

*

77.

*1 I

4

"I

I Ii

S

S \ * 14

*

I

S p

.'

A4

APPENDIX B DEA LINEAR PROGRAMMING PROBLEMS This appendix provides a further description of the DEA approach to the measurement of productive efficiency. The piecewise linear reference technology specification and linear programming problem are included. The piecewise linear reference technology or frontier is constructed from the observed inputs and output of the recruiting districts in the sample. We assume that there are n inputs, denoted by x = (xl, x2, ... , xn), one output, denoted by y, 1 and1 k observations (or recruiting districts) of x and y, i.e., (x , y ), i = 1, ... , k.1 Denote the matrix of observed inputs by X, where X is of dimension (k, n), k observations and n inputs. Denote the k dimensional vector of observed outputs by Y. The construction of the piecewise linear techrology requires the following restrictions: (i)

(ii)

(iii)

yi > O, i

k Z i=1

1, ... , k, i.e., a positive output is produced by each activity (recruiting district)

> 0, j = 1,

n,

..

J

n = x >O, i = 1,

k,

i.e., each input is utilized by at least one activity i.e., each activity uses at least one input.

Finally, let z = [l1, ... , zk] be a (1 x k) row vector representing the intensity (or activity) parameters for each of the k observations. Following Fare, Grosskopf, and Lovell [2], the technology formed by the data (X and Y) can be written as: T =

y): z . Y = y >

.

X

=

k

zi = 1}

(B-1)

i=1

1. For this application, x would consist of the inputs and environmental variables given on the right-hand side of equation 1. The output y is ACCON, the total A- and CU-cell contracts. B-1

A

Computation of the degree of efficiency relative to the frontier is accomplished via linear programming (LP) techniques. The measu e og technical efficiency relative to T for a given observation (x6, y ) is defined as: DEA 0 = max to: (xO, 0y

0

(B-2)

)cT}

and is computed for that observation, l em:

DEA 0

s.t.

" __ -

maximize 8 0, z Y e ey 0

z z

.XX

0

I z E~z.-

1

(B-3)

i=1

0, zi

by solving the following LP prob-

, i

1,

...

,

k

.

B-2

S-APPENDIX

C DESCRIPTION AND LISTING OF DEA COMPARISON DISTRICTS

i.M ',i

L"'..

A-"

p

4,)F

.4

APPENDIX C DESCRIPTION AND LISTING OF DEA COMPARISON DISTRICTS The solution to the linear programming problem indicates to which distzicts on the reference frontier an inefficient observation is compared. The optimal basic variables, i.e., the zi 9 0 in the optimal solution of each LP problem, determines a "best practiae" subset of For recruiting districts to which the district's performance is gauged. each district, the efficient subset,*in which that district's efficiency is gauged, is given by the set of xi in the optimal basic solution. Tables C-1 through C-4 provide the computed efficiency measure and set of comparison observations for each district by year. The efficiency measure listed is the solution to the LP problem and gives the ratio of efficient to actual contracts rather than percent of actual contracts lost due to inefficiency as was presented in the text. The comparison For districts are identified by year of operation and district number. example, district 101 in 1981 is gauged inefficient with a computed efficiency meab're of 1.043. This district is gauged Inefficient relative to five "similar" districts. Only one district (102) was compared in FY 1981, one in 1984 (840), and three In 1985 (districts 312, 725, and 839). The comparison year-district observations are efficient observations (see table 3 of the text) that compose the facet of the efficient frontier~to which district 81-101 is compared. These districts that the z identify together generate the efficient facet or comparison group oý the reference technology to which the district is compared.

C-1

TABLE

C-i

DISTRICT COMPARISON DISTRICTS FOR 1981 MEASURES DISTRICT

DEA MEASURE

COMPARISON DISTRICTS (YEAR-DISTRICT) ----------------------------------------------

101 102 103 104 119 161

1.043 1.000 1.226 1.290 1.279 1.220

81-102 81-102 82-315 81-102 84-409 84-409

84-840

85-312

85-725

85-839

82-417 82-422 84-840 84-840

82-422 84-840 85-312 85-312

85-312 85-312 85-725 85-725

85-528

310 311 312 313 314 315 316 348

1.289 1 .000 1.179 1.100 1.455 1.057 1.307 1.000

82-417 81-311 81-102 85-311 82-315 82-315 81-348 81-348

82-542

84-311

85-310

84-731 85-312 82-417 82-417 84-311

85-312 85-313 84-311 85-315 84-316

85-725 85-315 84-529 85-828 84-730

85-725 85-725 85-529 84-731

406 409 417 418

82-348 8-1-409 81-417 81-102 82-417 81-422

82-417 84-840

84-311 85-312

85-527 85-725

85-839

81-529 84-529

82-417 85-315

85-312 85-527

85-528 85-528

422

1.189 1.008 1.000 1.085 1.221 1.000

521 524 526 527 528 529 542 559

1.000 1.098 1.000 1.233 1.384 1.000 1.000 1.183

81-521 81-526 81-526 81-529 81-102 81-529 81-542 84-529

81-529

81-542

81-725

85-837

84-529 81-529

84-731 84-731

85-315 85-528

85-527 85-725

85-72

84-559

84-725

84-837

725

1.000

81-725

730

1.000

81-730

731 732 733 734 746

1. 145 1.000 1.000 1.590 1.000

81-102 81-732 81-733 82-417 81-746

81-725

84-731

82-542

84-311

84-542

84-746

85-725

836 837 838 839 84C

1.244 1.177 1.000 1.181 1.334

81-521 81-422 81-838 81-422 81-102

81-725 81-529

85-836 81-725

85-837

84-837 84-840

85-725 85-312

85-837 85-725

420

747

1.000

85-839 85-839

81-747

C-2

85-839 85-839

85-74 85-83

85-839 85-725

TABLE

C-2

DISTRICT COMPARISON DISTRICTS FOR 1982 MEASURNS

DISTRICT

DEA MEASURE

COMPARISON DISTRICTS (YEAR-DISTRICT) -------------------------------------------------------

101 102 103 104 119 161

1.098 1.028 1.206 1.181 1.203 1.192

82-417 81-102 82-315 81-102 82-417 82-417

84-409 82-422 82-417 82-422 84-840 84-840

84-840 85-312 82-422 84-840 85-312 85-312

85-312 85-312 85-528 85-528

310 311 312 313 314 315 316 348

1.187 1.000 1.071 1.016 1.416 1.000 1.285 1.000

82-417 82-311 82-417 84-311 82-417 82-315 82-348 82-348

84-311

85-310

85-312

85-312 85-310 84-529

85-312 84-559

85-747 84-747

82-417

84-311

84-730

85-527

406 409 417 418 420 422

1.095 1.025 1.000 1.106 1.213 1.000

82-348 82-417 82-417 82-315 82-417 82-422

82-417 84-409

84-311 84-840

88-527 85-312

85-839

82-417 84-529

82-422 84-559

85-312 84-747

85-528

521 524 526 527 528 529 542 559

1.315 1.091 1.000 1.106 1.193 1.000 1.000 1.108

81-422 82-417 82-526 82-315 82-417 82-529 82-542 82-417

82-417 82-542

84-731 84-529

85-836 85-529

82-417 84-529

84-529 84-731

85-527 85-528

85-747

84-529

84-559

725 730 731 732 733 734 746 747

1.133 1.008 1.305 1.182 1.208 1.599 1.143 1.000

82-417 82-417 82-417 82-417 81-730 82-417 81-746 82-747

82-542 84-311 84-409 84-311 84-311 84-311 84-311

84-542 84-730 84-731 84-542 84-529 84-542 84-746

84-725 85-527 85-725 84-731 85-315 84-730 85-725

84-837 85-747 85-839 85-725 85-725 84-746

836 837 838 839 840

1.253 1.089 1.064 1.142 1.339

82-417 81-529 81-102 82-417 82-417

82-422 82-422 81-422 82-422 82-422

84-840 82-542 82-422 82-542 84-840

85-838 85-837

C-3

85-312

85-837 85-312

85-839

85-747

85-747

85-839 85-839

85-725

85-747

TABLE

C-3

DISTRICT COMPARISON DISTRICTS FOR 1984 MEASURES D:STRICT

DEA MEASURE

COMPARISON DISTRICTS (YEAR--DISTRICT)

101 102 103 104 119 161

1.090 1.012 1.206 1.151 1.035 1.141

84-409 81-102 82-315 81-102 82-417 82-417

84-731 85-102 82-417 82-422 84-409 84-409

85-102 85-312 84-409 84-840 85-527 84-731

85-528 88-528 85-312 88-312 85-528 85-527

310 311 312 313 314 315 316 348

1.043 1.000 1.056 1.000 1.328 1.073 1.000 1.299

82-417 84-311 82-417 84-311 82-417 82-315 84-316 82-348

82-542

84-311

85-310

84-409 84-316 84-529 84-311

85-312 84-731 84-730 84-529

85-527 85-312 85-527 85-315

82-417

84-311

85-527

406 409 417 418 420 422

1.126 1.000 1.079 1.180 1.090 1.068

82-417 84-409 82-417 81-102 82-417 81-102

84-311

85-312

85-527

84-542 81-529 84-559 81-422

84-559 82-417 84-730 82-422

84-725 85-312 84-747 85-839

521 524 526 527 528 529 542 559

1.154 1.047 1.037 1.129 1.178 1.000 1.000 1.000

81-422 81-526 81-526 82-315 82-315 84-529 84-542 84-559

81-725 82-542 82-542 82-417 82-417

85-725 84-529 84-529 84-311 82-422

85-836 84-542 85-526 84-529 85-528

84-837 85-747 85-527 85-529

725 730 731 732 733 734 746 747

1.000 1.000 1.000 1.086 1.222 1.249 1.000 1.000

84-725 84-730 84-731 82-417 82-417 82-542 84-746 84-747

84-542 85-315 84-746

84-725 85-529 85-725

84-746 85-725

85-747

636

1.128

1.000 1.010

82-417

84-731

84-840

85-836

837 838

81-422

82-422

85-8368

839 840

1.025 1.000

82-422

82-542

85-837

84-837 84-802 82-417 84-840

C-4

85-836 85-836 85-527 85-528

85-313 85-747 85-725

a5-741

84-731 85-528

85-83

85-839

88-32.

85-741

85-83$

TABLE

C-4

DISTRICT COMPARISON DISTRICTS FOR 1985 MEASURES DISTRICT

DEA MEASURE

COMPARISON DISTRICTS (YEAR-DISTRICT)

101 102 103 104 119 161

1.150 1.000 1.236 1.246 1.184 1.395

81-102 85-102 82-315 81-102 82-417 84-409

84-409

84-731

85-102

85-312

82-417 82-422 84-409 84-731

84-409 84-840 84-840 84-840

85-312 85-312 85-312 85-528

85-527

310 311 312 313 314 315 316 348

1.000 1.000 1.000 1.000 1.315 1.000 1.020 1.269

85-310 85-311 85-312 85-313 82-315 85-315 84-311 81-348

82-417

84-311

84-529

85-527

85-74

84-318 84-311

84-731 84-316

85-312 84--731

406

1.156

82-417

84-311

85-312

85-527

409 417 418

1.008

84-731 85-725 84-409

85-102 85-839 84-840

85-312

1.154

84-409 82-417 82-417

85-312

85-528

85-83

420 422

1.154 1.176

82-417 81-102

84-311 81-422

84-730 82-422

85-527 85-839

85-747

521

1.151

81-102

81-422

81-725

85-836

524

1.000

85-524

526 527 528 529 542 559

1.000 1.000 1.000 1.000 1.099 1.050

85-526 85-527 85-528 85-529 82-417 82-417

82-542 84-529

84-311 84-542

84-542 84-559

725 730

1.000 1.000

85-725 85-730

731

1.106

81-102

84-731

85-312

85-725

732 733

1.062 1.126

82-417 81-529

84-542 85-315

84-725 85-529

84-746 85-725

734

1.208

82-417

82-542

84-746

85-725

746 747

1.004 1.000

82-542 85-747

84-746

85-310

836 837 838 839 840

1.000 1.000 1.047 1.000 1.003

85-836 85-837 81-102 85-839 81-102

81-422

81-838

85-836

84-409

84-840

85-312

1.032

C-5

85-839 85-836

i

85-725 84-725

85-7 84-73

85-747

85-831

85-725

-

--

--

--.-

---------------

l

ma-e~r n

APPENDIX D -.

PRODUCTION FUNCTION ESTIMATION RLULTS. FOR THE STATISTICAL APPROACH

NRA

Ie •

,.,-..Y"

APPENDIX D PIHODUCTION FUNCTION ESTIMATION RESULTS FOR THE STATISTICAL APPROACH

The statistical characteristics or the parametric, stochastic estimation and prediction of enlistment contracts are important for judging the confidence that may be placed in the predictions.

Table 0-1

presents the major statistics and parameter estimates associated with the linear specification. the two-stage procedure. sense.

The

R2

The fit

is 0.83 in the OLS first stag* of is thus reasonably good in an average

Maximum likelihood estimation of the parameters of the frontier

converged quickly with only the c~nstant term and population coefficient is 0.95, so that the standard X = changing. The estimate of deviation of inefficiency componenV is about the same as that of the random noise component.

Average technical inefficiency is estimated

from the formula in [17, p. 234].

It represents a loss of slightly less

than one percent of enlistment contracts.

The formulas in the text are

used to evaluate the most likely value (mode) of the inefficiency component for each observation based on the results of this regression. AN ALTERNATIVE STOCHASTIC PROCEDURE Schmidt (D-1] suggests using the results of a one-way fixed effects model of cross-section, time-series variation in observed production to impute stochastic measures of technical inefficiency.

In this approach,

the regression equation Is the same as equation 1 in the main text except for the inclusion of separate intercept terms for each of the districts. The estimation procedure is equivalent to OLS (as implemented in LIMDEP), but with district-specific fixed effects. These estimated fixed effects can be Interpreted as average differences in productivity, other things being equal,

assuming that the environmental variables have

the same effect on output in all districts and at all times.

D- 1

TABLE 0-1 SUiMMARY STATISTICS FOR STATISTICAL APPROACH (Dependent Variable: A- and CU-Cell Contraota)a Variable

Coefficient

t-ratto

-1,133.88

-5.35

1.00

POP

4.94

5.87

115.20

RTR

7.61

6.01

87.26

UR

54.33

7.69

7.87

625.96

4.18

1.13

79.90

0.67

0.73

0.95

1.88

--

CONSTANT

PAY URBPCT LAMBDA (X)

o2-

194.83

2a u

Mean oa

V

Average technical

inefficiency log-likelihood Mean of dependent variable Number of observations

a.

8.07 -1,079.7

1,208.8

1614

FY 1983 observations excluded from the sample.

D-2

vpriable

The results of a one-way fixed-effects estimation of the production relation are given in table D-2. Schmidt [D-1] suggests measuring technical inefficiency as the dtffcrence between the maximum estimated-

fixed effect and each-of -thie others.

'In--this -case, the- measurement of

technical inefficiency does -not depend on any---distributional;-- assumptions, regarding the skewness of the error term. Another advantage of this approach is the ready availability of standard statistical tests for the significance of the resulting technical inefficiency estimates. r4

The maximum fixed effect is about 220 recruits. The standard errors for the fixed effects are large (reflecting only-four observations per district). They range from &bout 270 to 350. Only 6 districts have estimated fixed effects below -200 recruits per year, and they are predominantly (4 of 6) in area 3. This is surprising, because the four area 3 districts (310-313) were generally judged efficient in the other approaches. Perhaps this is evidence that even though they are efficient relative to the best practice of other districts, these districts had significant excess capacity and could have been goaled higher. Districts 161 and 420 are Judged inefficient in all approaches, and closer inspection may yield fruitful analytical insights.

I

D-3

MUM.

..

.

TABLE D-2 FIXED EFFECTS MODEL FOR CONTRACTS (A-cell plus CU-cell) Analysis of Variance Source Within Between Total

Variation

D.F.

O.1192E÷08 0.132OE,08 0.2512E+08

123 40 163

,Mean

quare

Proportion

O.9688E+05O.3301E+06 O.1541E+06

o.474351 0.525649

F test for homogeneity across individuals F(4O,123) = 3.41

Significance = 0.000

Ordinary least squares estimates Dependent variable Number of observations Mean of dependent variable Std. dev. of dep. variable Std. error of regression R-squared Adjusted R-squared Variable

POP ONBOARD URATE RELMILPY URBPCT Sigma

Coefficient

1.56 1O.69 2916.24 118.91 -219.37 142.33

Contract 164.00 1208.81 392.57 142.33 0.90 0.87

Std. error

T-ratio (sig.lvl.)

0.95

1.63 (0.10)

--a192.5 .a 29.96 7.86

a 15.15 (0.00) -- a -7.32 (0.00) 18.11 (0.00)

a..LIMDEP did not produce reliable estimates of confidence for these parameters.

D-~4

-_

-A

A

TABLE D-2 (Continued) F test for homogeneity across individuals (Estimated coefficients) F(40,118) z 2.22 Significance = 0.001 Estimated individual effects District 101 102 .103 104 119 161

310 311 312 313 314 315 S316 ,348 406 409 417 418 420 422 521 524 526 527 528 529 542 559 725 730 731 732 733 734 746 747 836 837 838 839 84o

Coeffieient -91.00 -160.38 -178.63 -90.33 64.00 -339.65 -283.93 -254.79

Std. error 316.97 316.60 316.29 324.59 307.92 291.20 284.14 272.20 287.10 293.28 325.42 340.55 336.62 344.53 334.02 319.84 311.22 292.37 297.63 319.29 314.05 335.92 322.49 323.09 323.20 316.21 306.65 287.36 290.83 304.13 310.30 32.38 320.88 314.20 324.66 314.20 302.53 285ý04 289.71 303.92 298.71

-366.14 -463.52 -68.o5 -78.58 -79.70 7.0O0 218.40 -52.52 -34.84 -162.27 -251.62 -167.43 -82.13 39.79 -31.59 -180.28 102.88 -47.25 -143.45 -86.46 -152.05 -138.12 49.53 - 129.13 -70.34 - 147.85 -10.80 -86.92 -65.83 -61.60 --144.86 133.58 -7.55

D-5

T-ratio -0.29 -0.51 -0.56 -0.28 0.28 -1.17 -1.00 -0.94 -1.28 -1.58 -0.21 -0.23 -0.24 0.02 o.65 -0.16 -0.11 -0.56 -0.85 -0.52 -0.26 0.12 -0.10 -0.56 0.32 -0.15 -0.47 -0.30 -0.52 -0.45 0.16 -0.40 -0.22 -o.47 -0.03 -0.28 -0.22 -0.22 -0.50 -0.44 -0.03

REFERENCE [1D-]

Peter Schmidt. "Frontier Production Functions." Reviews vol. 4, no. 2 (1985-1986): pP. 312-315

RCnonmetzic

-D

NI

I

-

SECURTY CLASSIFICATION OF THIS PAGE

REPORT DOCUMENTATION PAGE Is. REPORT SECURITY CLASSIFICATION

lb. RESTRICTIVE MARKINGS

UNCLASSIFIED 2a.

_-_-

UROITY CLASSIFICATION AUTHORITY

3. DISTRIBUTION I AVAILABILITY OP REPORT

2b. OECLASSIFICATION I DOWNGRAOING SCHEDULE

APPROVED)FOR PUBLIC R91 EAB

4. PERFORMING ORGANIZATION REPORT NUMIER(S)

S. MONITORING ORGANIZATION REPORT NUMIIR(S)

DISTErIUTION L

TNLMrnID.

CLaM 87-181 (Revised) 6-. NAME OF PERFORMING ORGANIZATION

6Sb. OFFICE SYMIOL. 7a. NAME OF MONITORING OR.ANIZATION

Center for Naval Analyses

Office of the Chiefof Naval Operntions (OP-01)

6C. ADDRESS (City, Staft, nWZIP Code)

MbADDRESS W. tsite, mndtI Code)

4401 Ford Avenue

Navy.

Alexandria, Virginia 22302-0268

Wa

81. NAME OF FUNDING I ORGANIZATION

Sb. OFFICE SYMIIOL

Office of Naval Research

ONR

ot

nt

. D.C. 203W-2000

9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER

N00014-87-C-0001

Sc. ADDRESS (CIty, Stat, NWD Code) 800 North Quincy Street Arlington, Virginia 22217

PROGRAM ELEMENT NO.

PROJECT NO.

65154N

R0148

WORK UNIT ACCESSION NO.

TASK NO.

11. TITLE (InludeSaerty clamflation)

Recruiting Efficiency and Enlistment Objectives: An Empirical Analysis 12. PERSONAL AUTHOR(S)

Pa~tricia g. Bvr&es Timodthy W. Cooker 13a. TYPE OF REPORT

Final

13b. TIME COVERED FROM

4. DATE OF REPORT (YONMWnth

[

Le

TO

S. PAGE COUNT

1

72

16. SUPPLEMENTARY NOTATION

17.

FIELD 05

COSATI CODES GROUP

1I. SUIJECT TERMS (Cotn•e anrewvrse if eessasy and idenliy bybbock number)

SUIB-GROUP

09

Enlistments, Linear Programming, Manpower, Military Requirements,

I-Naval Personnel, Navy Recruiting Districts, Recruiters, Recruiting, IRecruiting Quotas, Statistical Analysism

'

19. ABSTRACT (Contrnue on reverne if nenssaay and de*ntify by bdagk number)

SWith increasing accession requirements and a more competitive recruiting environment, the size of the enlisted recruiting force has grown substantially. The Center for Naval Analyses was asked to provide analyticl sulport to Commander, Navy Recruiting Command (CNRC), concerning ways to improve the return on this recruiting investment by exaining the geographic allocation of recruiters, gfoals, and incentives. This reseec memorandum reports on the empi~rical analysis of three questions: (1) How do recruiting dist icts c€ompare in terms of pt production? (2)Ae ~orphic differences in enlistment goals consistent with measured differences in recruwtingswar ket condirLions? 3)Hwcan the results be used to better evaluate district recruit production.
l

' ,

121. ABSTRACT SECURITY CLASSIFICATION

2p0.DISTRIBUTION /AVAILABILITY OF ABSTRACT

--

by"

IDO

h 1Z2b. TELEPHONE (AicludeArew. Code) I22 OFFICE SYMBOL

I22.. NAME OP RESPONSIBLE INDIVIDUAL

Lcd. Cashbaugh FORM 1473.804 MAR

NR..... 8.3 APR

editionl may be used until exihausted. All Other eiotions are Obsolete.,

SECURITY CLASSIFICATION OF THIS PAGE