Cost Analysis Step by Step

CONNECTIONS TO WORK
Cost Analysis Step by Step

A How-to Guide for Planners and Providers
of Welfare-to-Work and Other Employment and
Training Programs
David H. Greenberg
Ute Appenzeller
Manpower Demonstration Research Corporation

Board of Directors
ROBERT REISCHAUER, Chairman RUDOLPH G. PENNER

Senior Fellow Senior Fellow
Brookings Institution Urban Institute
PAUL H. O’NEILL, Treasurer MARION O. SANDLER

Chairman and CEO Chairman and CEO
Alcoa Golden West Financial Corporation and
World Savings and Loan Association
ISABEL V. SAWHILL
Senior Fellow
MARY JO BANE Brookings Institution
Professor of Public Policy
John F. Kennedy School of Government ROBERT SOLOW
Harvard University Institute Professor
Massachusetts Institute of Technology
ANTONIA HERNANDEZ
President and General Counsel MITCHELL SVIRIDOFF
Mexican American Legal Defense and Professor Emeritus and Senior Fellow
Educational Fund Community Development Research Center
New School for Social Research
JAMES H. JOHNSON, JR.
E. Maynard Adams Professor of Business, WILLIAM JULIUS WILSON
Geography, and Sociology Malcolm Wiener Professor of Social Policy
Director, Urban Investment Strategies Center John F. Kennedy School of Government
University of North Carolina Harvard University
ANNA KONDRATAS
Senior Associate
Urban Institute
RICHARD J. MURNANE
Professor of Education JUDITH M. GUERON
Graduate School of Education President
Harvard University Manpower Demonstration Research Corporation
CONNECTIONS TO WORK

A How-to Guide for Planners and Providers
of Welfare-to-Work and Other Employment and
Training Programs
David H. Greenberg
Ute Appenzeller
October 1998

THE CONNECTIONS TO WORK PROJECT IS FUNDED BY
THE CHARLES STEWART MOTT FOUNDATION
AND THE ROCKEFELLER FOUNDATION. DEVELOPMENT AND
PUBLICATION OF THIS GUIDE WERE SUPPORTED
BY THE ROCKEFELLER FOUNDATION.
Dissemination of MDRC publications is also supported

by MDRC’s Public Policy Outreach funders: the Ford Foundation,
the Ambrose Monell Foundation, the Alcoa Foundation,
and the James Irvine Foundation.
The findings and conclusions presented in this guide

do not necessarily represent the official positions or policies
of the funders.
For information about MDRC, see our Web site: www.mdrc.org.

MDRC is Registered in the United States Patent Office.
Copyright © 1998 by the Manpower Demonstration Research Corporation. All rights reserved.
Contents
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .viii
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ix
Chapter 1 Who This Guide Is For and How It Can Be Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Potential Uses of Cost Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Cost Analyses of Programs Prior to Their Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Cost Analyses of Ongoing Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 How to Use This Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Chapter 2 Basics of Conducting Cost Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.1 The Issue of Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Alternative Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Choosing Among Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Definition of Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Gross Costs Versus Net Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Start-Up Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.4 Average Costs Versus Marginal Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.5 Joint Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.6 Regular Versus Special Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 The Time Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1 Role of Discounting in Conducting Cost Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.2 Adjusting for Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.3 Adjusting for Large Equipment Purchases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Chapter 3 Costing Out Ongoing Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.1 Costs per Case Versus Aggregate Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Data Needed for Cost Analyses of Ongoing E&T Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2.1 Records Program Agencies Should Keep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2.2 Data Needed from Outside the Program Agency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 Steps in Estimating the Gross Costs of Ongoing Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Step 1. Enumerate Program Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21
Step 2. Compute Unit Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22
a. The Basic Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
b. Determining Staff Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
c. Treatment of Overhead . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Step 3. Select the Analysis Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26
Step 4. Determine Participation Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
Step 5. Determine the Average Length of Participation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
Step 6. Compute Gross Cost per Case for Each Program Component . . . . . . . . . . . . . . . . . . . .28
Step 7. Compute Total Gross Cost per Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
3.4 Converting Gross Costs into Net Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4.1 Obtaining Comparison Groups for Cost-Benefit Analyses . . . . . . . . . . . . . . . . . . . . . . . . 31
iii
Contents iv
Chapter 4 A Real-World Cost Analysis of an Ongoing Program:

Implications of Program Design for Program Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33
4.1 JOBS Cost Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34
4.2 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .35
4.2.1 Effects of Program Models on Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.2 Unit Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2.3 Participation Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.4 Length of Stay in a Program Component . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.5 Support Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.6 Site-Specific Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
a. Agency Discretion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
b. Client Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
c. Welfare Grant Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
d. Labor Market Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Estimating the Costs of an Ongoing Program: An Illustration . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Chapter 5 Predicting Costs of Programs That Have Not Yet Been Implemented . . . . . . . . . . . . 49
5.1 Steps in Estimating the Gross Costs of Not-Yet-Implemented Programs . . . . . . . . . . . . . . . . . .50
Step 1. Develop a Flow Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .50
Step 2. Predict Total Program Referrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .50
Step 3. Predict the Assignment Rates for Each Program Component . . . . . . . . . . . . . . . . . . . . .50
Step 4. Predict the Participation Rate for Each Program Component . . . . . . . . . . . . . . . . . . . .51
Step 5. Predict the Average Length of Participation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51
Step 6. Estimate Unit Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51
Step 7. Compute Costs per Referred Case for Each Program Activity and Support Service . .51
Step 8. Compute Total Costs per Referred Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .52
Step 9. Compute Aggregate Program Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .52
5.2 Estimating the Costs of a Not-Yet-Implemented Program: An Illustration . . . . . . . . . . . . . . . . 52
5.3 Obtaining the Values Needed to Cost Out an Employment and Training Program . . . . . . . . . 63
5.3.1 Predicting the Total Number of Cases Referred (Step 2) . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.2 Predicting Assignment Rates (Step 3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
a. Length of Time to Reach Assignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
b. Predicting the Number of Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.3.3 Predicting Participation Rates (Step 4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3.4 Predicting the Average Length of Participation (Step 5) . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3.5 Estimating Unit Costs (Step 6) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
a. Using Existing Estimates of Unit Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
b. Program Unit Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.4 Changes in Program Costs Over Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.4.1 An Assessment of the Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
a. How Sensitive Are the Findings to the 1.2 Percent Assumption? . . . . . . . . . . . . . . . . . . . . . . 77
b. Effect of Policy Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.5 Predicting the Net Costs of a Not-Yet-Implemented Program . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Contents v
Chapter 6 Using Cost Information in Designing, Bidding, and Budgeting . . . . . . . . . . . . . . . . . 81

6.1 Sensitivity Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.2 Bids for Program Components by Community-Based Organizations . . . . . . . . . . . . . . . . . . . . 86
Step 1. Develop a Flow Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .88
Step 2. Predict Total Program Referrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .88
Step 3. Predict the Assignment Rate for the Program Component . . . . . . . . . . . . . . . . . . . . . . .88
Step 4. Predict the Participation Rate for the Program Component . . . . . . . . . . . . . . . . . . . . . .88
Step 5. Predict the Average Length of Participation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .88
Step 6. Estimate Unit Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89
6.3 Developing a Program Budget . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89
Appendix A: Selected Employment and Training Programs: Summary Descriptions

and Main Evaluation Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Appendix B: Assignment Rates for Employment and Training Activities . . . . . . . . . . . . . . . . . 103
Appendix C: Average Length of Stay in Employment and Training Activities . . . . . . . . . . . . . . 107
Appendix D: Participation Rates for Employment and Training Activities . . . . . . . . . . . . . . . . 111
Appendix E: Unit Costs for Employment and Training Activities . . . . . . . . . . . . . . . . . . . . . . . . 117
Appendix F: Program Effects on AFDC Receipt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .131
References and Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .135
Recent Publications on MDRC Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Tables and Figures

Table
1.1 Which Sections of the Guide Are for You? ............................................................................... 6
2.1 Examples of How E&T Effects Are Viewed from Alternative Perspectives .............................. 8
3.1 Minnesota Family Investment Program/STRIDE Time Sheet:
Staff Members’ Time Allocation to Program Components..................................................... 25
4.1 Gross Costs of the Labor Force Attachment and Human Capital
Development Approaches Within Two Years After Orientation ............................................. 35
4.2 Gross Costs of Riverside’s Labor Force Attachment and Human
Capital Development Approaches Within Two Years After Orientation................................. 39
4.3 Unit Costs of Riverside’s Labor Force Attachment and Human
Capital Development Approaches ........................................................................................... 40
4.4 Participation Rates for Employment and Training Activities in
Riverside Within Two Years After Orientation ........................................................................ 41
4.5 Average Length of Stay in Employment and Training Activities
in Riverside Within Two Years After Orientation .................................................................... 42
4.6 Support Service Costs of Welfare-to-Work–Related Activities Within
Two Years After Orientation..................................................................................................... 43
Contents vi
4.7 Welfare Status of AFDC Recipients at the Beginning of the Program.................................... 45

4.8 Estimating the Costs of an Ongoing Program: Two-Year Costs of Riverside’s
Labor Force Attachment Approach ......................................................................................... 47
5.1 Estimating the Costs of a Not-Yet-Implemented Program: Hypothetical
Example 1, Steps 1-5 ................................................................................................................ 55
Example 1, Step 6: Unit Cost.................................................................................................... 56
Example 1, Steps 7 and 8: Cost per Referred Case for Each Program Activity
and Support Service................................................................................................................. 57
Example 1, Step 9: Aggregate Costs ......................................................................................... 58
Example 2, Steps 1-5 ................................................................................................................ 59
Example 2, Step 6: Unit Cost.................................................................................................... 60
Example 2, Steps 7 and 8: Cost per Referred Case for Each Program Activity
and Support Service................................................................................................................. 61
Example 2, Step 9: Aggregate Costs ......................................................................................... 62
5.9 Effect of a Change in Assignment Rates on the Predicted Costs per Case
Referred to the Program .......................................................................................................... 63
6.1 Estimating the Costs of an Ongoing Program: Two-Year Costs of
Riverside’s Labor Force Attachment Approach ....................................................................... 83
6.2 Illustrations of Changes in Costs Resulting from Alternative Modifications
to the Riverside LFA Program ................................................................................................. 84
6.3 Estimating the Costs of an Ongoing Program: Serving a Harder-to-Serve
Caseload: A Hypothetical Example ......................................................................................... 85
6.4 Estimating the Costs of an Ongoing Program: Riverside’s Human Capital
Development Approach ........................................................................................................... 87
6.5 Program Budget Prototype...................................................................................................... 91
A.1 JTPA Services ......................................................................................................................... 101
C.1 National Evaluation of Welfare-to-Work Strategies: Average Length of
Stay in Employment and Training Activities Within Two Years After
Orientation ............................................................................................................................. 108
C.2 California’s Greater Avenues for Independence (GAIN) Program: Average Length
of Stay in Employment and Training Activities Two to Three Years After
Orientation ............................................................................................................................. 109
C.3 Florida’s Project Independence: Average Length of Stay in Employment
and Training Activities Within Two Years After Random Assignment................................. 110
Contents vii
D.1 National Evaluation of Welfare-to-Work Strategies: Participation Rates for

Employment and Training Activities Within Two Years After Orientation .......................... 112
D.2 California’s Greater Avenues for Independence (GAIN) Program: Participation
Rates for Employment and Training Activities Within Two to Three Years
After Orientation.................................................................................................................... 113
D.3 Minnesota Family Investment Program (MFIP): Participation Rates for
Employment and Training Activities, in Urban Counties, Within One Year
After Random Assignment .................................................................................................... 114
D.4 Florida’s Project Independence: Participation Rates for Employment and
Training Activities Within Two Years After Random Assignment ........................................ 115
E.1 National Evaluation of Welfare-to-Work Strategies: Unit Costs for
Employment and Training Activities (in 1993 Dollars)........................................................ 118
E.2 California’s Greater Avenues for Independence (GAIN) Program:
Unit Costs for Employment and Training Activities (in 1993 Dollars) ................................ 121
E.3 Florida’s Project Independence: Unit Costs for Employment and
Training Activities (in 1993 Dollars) ..................................................................................... 123
E.4 National Job Training Partnership Act (JTPA) Study: Unit Costs for
Employment and Training Activities (in 1988 Dollars)........................................................ 124
E.5 JOBSTART Demonstration: Unit Costs for Employment and Training
Activities (in 1986 Dollars).................................................................................................... 125
F.1 National Evaluation of Welfare-to-Work Strategies: Program Effects on AFDC
Receipt Within a Two-Year Follow-Up Period ...................................................................... 130
Figure
4.1 Intended Sequence of Activities in a Labor Force Attachment
Welfare-to-Work Program....................................................................................................... 37
4.2 Intended Sequence of Activities in a Human Capital Development
Welfare-to-Work Program....................................................................................................... 38
5.1 Flow Diagram for a Hypothetical Welfare-to-Work Program ................................................ 54
5.2 The Flow Through Riverside’s Labor Force Attachment (LFA) Program
Approach for 100 Typical LFA Sample Members Within Two Years
After Orientation...................................................................................................................... 66
5.3 The Flow Through Grand Rapids’ Labor Force Attachment (LFA) Program
Approach for 100 Typical LFA Sample Members Within Two Years
After Orientation...................................................................................................................... 67
B.1 National Evaluation of Welfare-to-Work Strategies: Assignment
Patterns Within a Two-Year Follow-Up Period, by Site: Labor Force
Attachment Approach ............................................................................................................ 104
B.2 National Evaluation of Welfare-to-Work Strategies: Assignment Patterns
Within a Two-Year Follow-Up Period, by Site: Human Capital
Development Approach ......................................................................................................... 105
Acknowledgments
This book could not have been completed without the help, knowledge, and experience
of many people. It was envisioned and funded by The Rockefeller Foundation, following
a Rockefeller-organized conference on innovative approaches to connecting low-income
people to employment, which pointed to a widespread need for systematic information
on estimating program costs. We are grateful for the strong support and valuable com-
ments of Julia Lopez, Director of the foundation’s Equal Opportunity division, and
Raymond Colmenar, Senior Research Associate at the foundation.
At MDRC, Fred Doolittle was closely involved with the project from the very begin-
ning, and his focus and vision had a major influence on the final product. David Butler
and Irene Robling offered detailed comments on the text from a practitioner’s point of
view. Amy Brown, Robert Ivry, Donna McGill, and James Riccio also reviewed the manu-
script. Mary Farrell instructed us about the cost analysis for the National Evaluation of
Welfare-to-Work Strategies and carefully reviewed drafts of the guide. Charles
Michalopoulos developed for us the Markov model used in chapter 5.
The appendices could not have been completed without the help of staff involved in
various MDRC projects: Veronica Fellerath, James Kemple, Ginger Knox, Laura Storto,
and Johanna Walter. Larry Orr at Abt Associates Inc. provided unpublished data from
the National JTPA Study.
Joan Johnson edited the tables. Christine Schwartz and Emily Danyluk fact-checked
the document. Alice Tufel edited the book and had many helpful suggestions for making
it more accessible. Judith Greissman provided further editorial support. Stephanie
Cowell and Patt Pontevolpe word-processed the text. Brad Petrie assisted in developing
the accompanying floppy disk, and Martey Dodoo and Irene Robling performed test
runs of the disk. Edward Rowe and his colleagues at Rowe & Ballantine designed the
book.
Finally, we are indebted to several practitioners. Detailed reviews were provided by
Rosella Stern, Executive Director of Washington Works, and Joan Zinser, Deputy
Director of the County of San Diego Health and Human Services Agency. Heather Hiles,
President of San Francisco Works, also commented on a draft.
The Authors
viii
Preface
The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 dramati-
cally increased states’ flexibility in designing income support, employment, and social
service programs for welfare recipients. As a result, many states are experimenting with
new program approaches and with contracting out service provision. Because of these
developments, counties and nongovernment service providers are increasingly called
upon to estimate the cost of existing and yet-to-be-implemented programs.
This guide provides a systematic approach to costing out employment and training
programs for low-income people. It is geared toward practitioners in both government
agencies and community-based organizations who are asked to assess the cost of pro-
grams or program components that are currently running or are still in the planning
stages. The guide helps states, localities, and service providers to calculate and present
cost estimates through a step-by-step approach. The steps can be directly applied by
using the spreadsheets provided on the accompanying disk.
This guide is the latest publication in the Connections to Work series, a project con-
ducted by MDRC with support from The Rockefeller Foundation and the Charles
Stewart Mott Foundation. The authors are David H. Greenberg, Professor of Economics
at the University of Maryland, Baltimore County, and Ute Appenzeller of MDRC. By
making this knowledge accessible to program providers and planners at a time of high
demand for programs that move people from welfare into jobs, and by illustrating the
approach with cost figures from real programs, the authors aim to provide a highly prac-
tical guide that can be used as an easy-to-access reference book.
Judith M. Gueron
President
ix
Chapter 1
Who This Guide Is For

and How It Can Be Used
This guide presents step-by-step instructions for conducting cost analyses of both
ongoing and proposed social programs for the poor. It focuses on one particular type
of program: employment and training (E&T) programs that are intended to improve
the performance of low-income people in the labor market and reduce their reliance on
welfare payments and unemployment compensation. But it also recognizes that these
programs can serve volunteers or be part of a welfare system that requires those receiv-
ing aid to participate in employment-oriented services. It is aimed at planners and
practitioners in government agencies that fund, operate, and contract out such programs
and in community-based or other nonprofit organizations that deliver the services.
1
Cost Analysis Step by Step 2
1.1 Introduction
Employment and training programs attempt to accomplish their objectives by providing
such services as job search assistance, training, education, subsidized employment in the
private sector, and community service jobs in the public sector. Long viewed as an
important tool for increasing employment, such programs have been an integral part of
welfare reform efforts since the 1960s. Starting in the mid 1970s, some states began to
mandate participation in these services as a condition of aid for many categories of
recipients. With the passage of major federal welfare legislation in 1996 — the Personal
Responsibility and Work Opportunity Reconciliation Act — the receipt of aid is linked
closely with participation in employment-oriented services, and welfare agencies are
now putting even greater emphasis on using E&T programs — which in a welfare set-
ting are often called “welfare-to-work programs” — to help recipients achieve economic
independence. By funding welfare through federal block grants to the states (Temporary
Assistance for Needy Families, or TANF), the 1996 law gives states great freedom in
designing E&T programs. Over time, this freedom is likely to result in important
changes in the program models used to serve welfare recipients. These changes will have
significant cost implications for states and localities, and it is important for program
providers and planners to understand these cost implications as states plan expansions
or modifications of their welfare-to-work initiatives.
This guide provides a systematic approach for “costing out” an E&T program. Cost
estimates can be obtained either for the program as a whole or for specific program
activities and support services. These estimates can be used for planning new E&T pro-
grams and for assessing and modifying existing programs. The approach developed in
the guide can also be helpful in formulating Requests for Proposals and in responding to
them. Although the scope of the guide is limited to E&T programs, the general frame-
work can be readily adopted to obtaining cost estimates for other social service pro-
grams.
The approach presented here has proven useful to practitioners and researchers
in the field of social service programs. However, to the best of our knowledge, nothing
similar to this guide exists. This guide provides practical guidelines for:
■ measuring program costs;
■ presenting cost information in the most useful manner possible;
■ dealing with issues and problems encountered in measuring costs.
Section 1.2 describes the various ways in which cost analyses of E&T programs are
used. Section 1.3 provides recommendations on how the guide can best be used for con-
ducting and interpreting cost analyses, and presents an overview of how the remainder
of the guide is organized.
Chapter 1: Who This Guide Is For and How It Can Be Used 3
1.2 Potential Uses of Cost Analyses

Cost analyses of social service programs can be usefully conducted at two different
points in time: prior to implementing a new program and after a program is already
running (even if the program has been in operation for a number of years). As discussed
next, each type of analysis can serve multiple purposes, although they differ in impor-
tant respects.
1.2.1 Cost Analyses of Programs Prior to Their Implementation

Before an agency implements any new program, it must have some idea of the
overall program cost in order to have sufficient funds allotted to operate the program.
Moreover, it needs information on the components of those costs to determine how
many new personnel will need to be hired, the cost of new equipment that will have to
be purchased, the rental cost of additional office space that may be required, and so
forth. Sometimes the agency sponsoring a new E&T program plans to contract out part
of the operation of the program to other institutions, and cost information is needed
by both parties to negotiate a contract. Thus, programs need to be costed out before
they are implemented.
The usefulness of conducting cost analyses for programs that are still at the plan-
ning stage goes well beyond budgeting and resource planning. Such analyses are also
useful in making decisions about whether a proposed program should be implemented
in the first place. If the potential costs of the program are deemed too high, then the
program may not be adopted. Even if a decision is made to go ahead with the program,
several possible variants of it may exist and the costs of these competing program
variants are likely to differ. In such cases, their relative cost is likely to be an important
consideration in deciding among them.
As important as costs are, however, they should never be the only factor in making
decisions about E&T programs. The extent to which program goals are met — that is,
the extent to which the program engenders benefits such as increases in earnings or
decreases in dependency on welfare payments — must also be taken into account. For
example, a higher-cost program might have advantages over a lower-cost program if it
also results in larger benefits.
1.2.2 Cost Analyses of Ongoing Programs

Cost analyses of social programs are conducted much more easily once a program
has been implemented. Cost analyses of not-yet-implemented programs must typically
be based on projections or informed guesses about such issues as the number of people
who will actually receive services and the length of time participants will receive those
services. Hence, they are inherently subject to considerably more uncertainty than
analyses of ongoing programs, for which much information is already available.
Cost analyses of ongoing programs can serve a number of purposes, although, as

will be seen later, the exact nature of the analysis will vary depending on the particular
purpose it is intended to serve. The first type of use is to monitor an ongoing program,
including both program activities that are performed in-house by the agency sponsoring
the E&T program and those that have been contracted out. Do the costs of each pro-
gram component appear reasonable? Are cost expectations being met? Was the pre-
implementation cost analysis correct? If not, why not? Answering these questions is part
of a learning process. The answers are also essential in determining whether the condi-
tions in performance-based contracts that contain cost criteria are being met. In addi-
tion, cost analyses can be used to make various adjustments. For example, if the pro-
gram is much more expensive than initially anticipated, it might be necessary to scale
back the services provided or accept fewer participants. If it is less expensive, somewhat
richer services might be provided or eligibility criteria might be relaxed.
Cost analyses can also be used to conduct evaluations of ongoing programs — that
is, techniques used to assess whether existing programs are working and whether they
can be improved. Program evaluations can help answer the following sorts of questions:
Is an existing program working? Can it be improved? Should it be terminated? How do
different programs compare? Should a program that is being operated in one place be
considered for adoption in other locations?
An analysis that focuses solely on costs, such as the one described in this guide, pro-
vides only some of the information needed to respond fully to such questions — but it is
a necessary first step in evaluating a program. A more formal analysis that compares
program costs with program benefits, which is the most valuable way to determine a
program’s overall worth, goes beyond the scope of this guide. Readers who would like
more information on conducting cost-benefit analyses of E&T programs might want to
consult chapter 9 in Cost-Benefit Analysis: Concepts and Practice, by Anthony E.
Boardman et al. (Upper Saddle River, NJ: Prentice-Hall, 1996).
1.3 How to Use This Guide

The guide was conceived as a template for a very comprehensive cost analysis but,
because resource and time constraints may make such a comprehensive analysis infeasi-
ble in practice, shortcuts are suggested throughout. Nonetheless, readers should be
aware that a more extensive analysis relies on the full template presented here, and that if
compromises must be made, it is very important that both those conducting the analysis
and those using the findings understand how their results will vary from those that
would emerge from a more extensive analysis. They should be aware, for example, of
costs that are left out or are imperfectly estimated. To illustrate how cost analyses of E&T
programs can be conducted and presented, we often draw on findings from cost analy-
ses of real programs that have been conducted by the Manpower Demonstration
Research Corporation (MDRC).
Chapter 1: Who This Guide Is For and How It Can Be Used 5
The sections of this guide are highly interdependent, and concepts developed in the
early chapters are often used later on, so it is recommended that you read the chapters
consecutively. To assist readers who are actually conducting a cost analysis (as opposed
to using findings from cost analyses), each chapter begins with an overview of its con-
tents. In addition, text boxes are used throughout the guide to supplement the narrative.
Moreover, examples are frequently used to illustrate the concepts. Examples are indicat-
ed by the symbol e.g. in the margin. The guide also includes a 3.5" floppy disk with
Excel spreadsheets that are modeled after the step-by-step approaches developed in the
text. You can use these spreadsheets for developing your own cost estimates. Whenever a
?W2@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@6Xg?
?
?
?
?
?
?
?
?
?
?
?
?7@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@)X?f?
?@@(M?e?@@?
?@@Hf?@@?
?@@?f?@@?
?@@?f?@@?
?@@?f?@@?
?@@?e?I'@1?f?
?@@?fN@@?f?
?@@?f?@@?f?
?@@?f?@@?f?
?@@?f?@@?f?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@?f?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@? ?@@?f?@@?f?
reference to the disk is made, the symbol appears in the margin.

?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@?f?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@?
?@@?f?@@? ?@@?f?@@?f?
?@@?f?@@?f?
?@@?f?@@? J@@?f?@@?f?
?@@?f?@@? ?O&@@?f?@@?f?
?@@?f?3@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@5?f?@@?f?
?@@?f?V4@@@@@@@@@@@@@@@@@@@@@@@@@@@@0Y?f?@@?f?
?@@? ?@@?f?
?@@?
?@@? ?@@?f?
?@@?f?
?@@? ?@@?f?
?@@?
?@@? ?@@?f?
?@@?f?
?@@?heW2@@@@@@@@@@@@@@@@@6K?he?@@?f?
?@@?h?W&@@@@@@@@@@@@@@@@@@@@@he?@@?f?
?@@?h?7@@@@@@@@@@@@@@@@@@@@@@he?@@?f?
?@@?h?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@?h?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@?h?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@?h?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@?h?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@?h?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@?h?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@Lh?@@@g?@@@@@@@@@@@@@he?@@?f?
?@@)X?g?@@@g?@@@@@@@@@@@@@he?@@?f?
?3@@)Xg?@@@g?@@@@@@@@@@@@@he?@@?f?
?V'@@)X?f?@@@g?@@@@@@@@@@@@@he?@@?f?
V'@@)Xf?@@@g?@@@@@@@@@@@@@heJ@@?f?
?V'@@)K?e?@@@g?@@@@@@@@@@@@@h?O&@5?f?
V'@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@H?f?
?V4@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@g?
?
?
?
?
?
?
?
?
?
?
Table 1.1 (page 6) will help you determine which sections are most useful to you.
The following is a brief overview of how the remainder of this guide is organized:
■ Chapter 2 focuses on conceptual issues, particularly issues concerning how costs should
be defined and what costs should be included as part of a cost analysis. It will be seen
that deciding which costs to include in a cost analysis depends on the purposes to be
served by the analysis.
■ Chapter 3 provides step-by-step instructions for conducting a cost analysis of an ongo-
ing E&T program, and discusses the data needed to do this.
■ Chapter 4 presents detailed findings from a previous cost analysis, one conducted by
MDRC of an E&T program for welfare recipients. The chapter illustrates how the meth-
ods described in chapter 3 can be applied in practice and how to display cost informa-
tion from ongoing E&T programs. Considerable emphasis is given to the factors that
cause costs to vary among programs.
■ Chapter 5 describes the steps required to conduct cost analyses of not-yet-implemented
programs. Particular emphasis is given to how findings from cost analyses of ongoing
E&T programs can be used in costing out not-yet-implemented programs.
■ Chapter 6 describes some additional ways in which the cost analyses described in the
preceding chapters can be used.
■ Appendix A summarizes the program features and main evaluation findings of pro-
grams discussed in this guide.
■ Appendices B–F provide various values that were used in MDRC cost analyses of ongo-
ing programs. As described in chapter 5, these values can be used to conduct cost analy-
ses of not-yet-implemented programs.
■ The Glossary lists and defines many of the terms used in this guide. It can be found at
the end of this publication.
W2@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@6X?g
■ The pocket on the inside back cover contains a disk with Excel spreadsheets. These files
7@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@)Xg
@@(Mf@@ @@fI'@1g
@@H?f@@
@@g@@ @@f?N@@g
@@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ ?J@@g@@g
O&@@g@@g
@@g3@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@5g@@g
@@gV4@@@@@@@@@@@@@@@@@@@@@@@@@@@@0Yg@@g
@@
@@ @@g
@@g
@@ @@g
@@ @@g
can be used as worksheets for adopting the steps described in chapters 3 and 5 in cost-
@@ @@g
@@
@@he?W2@@@@@@@@@@@@@@@@@6Khf@@g @@g
@@heW&@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he7@@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@L?h@@@?g@@@@@@@@@@@@@?he@@g
@@)Xh@@@?g@@@@@@@@@@@@@?he@@g
3@@)X?g@@@?g@@@@@@@@@@@@@?he@@g
V'@@)Xg@@@?g@@@@@@@@@@@@@?he@@g
?V'@@)X?f@@@?g@@@@@@@@@@@@@?h?J@@g
V'@@)Kf@@@?g@@@@@@@@@@@@@?hO&@5g
?V'@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@Hg
V4@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@?g
ing out your own program.

W2@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@6X?g
Table 1.1
7@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@)Xg
@@(Mf@@
@@H?f@@ @@fI'@1g
@@f?N@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ ?J@@g@@g
O&@@g@@g
@@g3@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@5g@@g
@@gV4@@@@@@@@@@@@@@@@@@@@@@@@@@@@0Yg@@g
@@
@@ @@g
@@g
@@ @@g
@@ @@g
@@ @@g
@@
@@he?W2@@@@@@@@@@@@@@@@@6Khf@@g @@g
@@heW&@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he7@@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@L?h@@@?g@@@@@@@@@@@@@?he@@g
@@)Xh@@@?g@@@@@@@@@@@@@?he@@g
3@@)X?g@@@?g@@@@@@@@@@@@@?he@@g
V'@@)Xg@@@?g@@@@@@@@@@@@@?he@@g
?V'@@)X?f@@@?g@@@@@@@@@@@@@?h?J@@g
V'@@)Kf@@@?g@@@@@@@@@@@@@?hO&@5g
?V'@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@Hg
V4@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@?g
WHICH SECTIONS OF THE GUIDE ARE FOR YOU?
If you are . . . The most useful sections for you are . . .

A policy or cost analyst Chapters 1–6.
at a government agency
(SKIP sections 5.3.5.a and b, which present details about
computing unit costs.)
Files on disk.
A program planner Sections 2.1 and 2.2.

Chapter 4 for an illustration of the implications of
program design for costs.
Sections 5.2 and 6.1 for discussion and illustrations of
sensitivity analyses.
File on_going on disk.
A community-based Chapter 3 and section 6.1 for help in costing out current
organization that is thinking operations and proposed changes, and in conducting
about expanding its current sensitivity analyses to determine how programmatic changes
E&T operations will affect costs.
A community-based Chapter 2 for major concepts that should be considered

organization that has not in preparing cost estimates.
previously been involved Chapter 5, which includes a worksheet for predicting costs.
in an E&T program but is
File new_prog on disk.
planning to be
Chapter 6 for help in conducting sensitivity analyses and
developing a program budget.
A government agency that is Sections 2.1 and 2.2.
preparing requests for bids on Chapters 4–6, which include a uniform framework for
E&T components and wants providing cost estimates.
bidders to use a uniform frame-
work for providing cost estimates
An agency or organization Sections 2.1 and 2.2.

that is bidding to operate all Chapters 4–6 for help in preparing cost estimates for your bid.
or part of an E&T program
An internal auditor or program Chapters 3 and 4 for procedures for obtaining cost
administrator concerned about information that can be used to monitor ongoing programs.
the efficiency and performance of
Section 6.1.
an existing E&T program
An individual responsible for Chapter 2 for conceptual issues in considering costs.
overseeing a program (e.g., a Chapters 4 and 6 for information that can be obtained from
comptroller or a staff member of cost analyses of E&T programs.
a legislative committee)
Sections 3.3 and 5.1 for help in understanding the steps
required for conducting analyses of E&T programs.
(SKIP the detailed explanations of how to complete each step.)
Chapter 2
Basics of Conducting
Cost Analyses
This chapter starts by considering the issue of perspective — that is, who will be affect-
ed, and how, by various program costs? To appropriately measure costs, this question
must first be answered. The costs to include in a cost analysis depend on the perspective
that is most important — that of trainees or that of the agencies and organizations run-
ning the program. If the latter, a further decision must be made about whether to
include costs accruing to all the agencies and organizations in the program, or just a
subset.
The chapter also relies on some standard definitions used by accountants and econ-
omists to explore various decisions that must be made in measuring costs. In conduct-
ing cost analyses, these decisions are often made implicitly, without considering how
they will affect findings from the analysis — but, as will be seen, the findings are some-
times quite sensitive to such decisions. One objective of this chapter is to clarify the cir-
cumstances under which a particular decision is most appropriate.
The final section of the chapter describes the role of discounting, adjusting for infla-
tion, and depreciation in cost analysis. This section will mainly be of interest to readers
who are conducting cost analyses, rather than to readers using findings from them.
7
2.1 The Issue of Perspective

A key issue in conducting cost analyses concerns the groups, agencies, and organizations
to which costs accrue. In brief: Whose costs should be counted? As discussed below,
there are several dimensions to this issue.
2.1.1 Alternative Perspectives

As previously suggested, E&T costs, as well as benefits, can accrue to several groups
and institutions, including the various government agencies and nongovernmental
organizations (NGOs) that operate program activities or provide program services, and
to participants themselves. It is also possible to view costs (and benefits) from the per-
spective of society as a whole, where “society” is defined to include all the groups and
institutions just mentioned and “everyone else” — that is, all other groups, institutions,
and individuals in society apart from program participants. The differences between the
perspectives of various groups and institutions are considerable. Similarly, there are
major differences between the broad social perspective and each of the narrow perspec-
tives.1 Some costs — such as day care costs that are not reimbursed by the government
— accrue only to participants, and are counted from the participant’s perspective as well
as from a societal perspective, but not from a government perspective. Other E&T pro-
gram effects — such as stipends paid to participants — are a cost to the government,
but a benefit to the participants themselves, so from a societal perspective, but not a
government perspective, they cancel out. Still other E&T effects — such as reductions in
welfare payments — are benefits to the government and a cost to participants, so they,
too, cancel out from a societal perspective.
These various perspectives are summarized in Table 2.1. The table also depicts a
slightly more complex example in which a nongovernmental organization provides
Table 2.1
EXAMPLES OF HOW E&T EFFECTS ARE VIEWED

FROM ALTERNATIVE PERSPECTIVES
Effect Government NGO Participant Society

Nonreimbursed day care costs n/a n/a Cost Cost
Stipends paid to participants Cost n/a Benefit Neutral
Reductions in welfare payments Benefit n/a Cost Neutral
Provision of vocational training n/a Cost n/a Cost
Payment to vocational training
provider Cost Benefit n/a Neutral
NOTE: n/a means not applicable.
1. Cost-benefit analysis often refers to a taxpayer perspective, which is almost identical to a government budget
perspective. For more details, see James Riccio, Daniel Friedlander, and Stephen Freedman, GAIN: Benefits, Costs,
and Three–Year Impacts of a Welfare-to-Work Program (New York: MDRC, 1994).
Chapter 2: Basics of Conducting Cost Analyses 9
vocational training for E&T participants and is paid by the government for providing
this service. The payment made by the government is listed separately from the provi-
sion of training because, in practice, the amount paid by the government could cover
either more or less than the costs incurred by a vocational training provider, depending
on the specific contractual arrangement.
2.1.2 Choosing Among Perspectives

A key decision in a cost analysis is the choice among the different perspectives
described above. Consider first the choice between the societal perspective and the more
narrow perspectives of the agencies and organizations that operate E&T programs. In
other words, should the participant’s perspective be ignored? The answer to this ques-
tion turns largely on how the cost analysis will be used.
For many purposes, cost analyses need consider only costs that are incurred by the
agencies and organizations involved in the operation of an E&T program. Information
on such costs, for example, can be used in costing out proposed E&T programs for bud-
geting purposes or for monitoring ongoing programs. For other uses — such as deci-
sions about whether to adopt a new program or expand or scale down an existing pro-
gram — it would be ideal to conduct a formal cost-benefit analysis. Such an analysis
would include the participant’s perspective as well as the perspectives of government
agencies and NGOs, and would involve measuring both benefits and costs. However,
such an analysis requires an investment of considerable time and resources. Indeed,
firms that specialize in cost-benefit analyses of E&T programs can best conduct such
evaluations. While less than ideal, a cost analysis that ignores benefits and the partici-
pant’s perspective may be all that is feasible and is still very helpful.
A government agency with overall responsibility for operating an E&T program,
referred to here as the “program agency,” obviously incurs costs as a result. However,
other government agencies and institutions (for example, community colleges) and
NGOs may also incur costs. For some purposes, such as resource planning and monitor-
ing, each agency or organization may be interested only in those costs that it incurs. For
these purposes, even the program agency will mainly be interested in its own costs. A
nongovernmental organization that is bidding to provide a particular E&T service, for
instance, will obviously be concerned with predicting its own costs, but not those that
occur elsewhere. When agencies and institutions want to focus only on their own costs,
the cost analysis is considerably simplified because many costs can be ignored.
Moreover, it is typically easier for an agency or organization to obtain information on its
own costs than on costs accruing elsewhere.
For other purposes, however, a program agency should consider all costs, regardless
of where they occur. This would be true, for example, of cost analyses that are being
used to decide which of several competing programs to adopt or whether to expand or
cut back on an ongoing program. In such cases, however, it is still often useful to report
costs separately for each agency and organization that incurs costs. In conducting a
comprehensive cost analysis in which costs incurred by the program agency, as well as
by other agencies and organizations, are included, it is important not to double-count.
For example, the cost incurred by a vocational training provider and the program
agency’s payment to the vocational training provider should not be summed.
In the operation of E&T programs, costs may be incurred at each level of govern-
ment — federal, state, and, local — with the precise allocation among levels somewhat
arbitrarily determined by various pieces of legislation and negotiated agreements. Each
level of government will, of course, be especially concerned with the costs that it incurs.
Thus, in a comprehensive cost analysis of an E&T program that attempts to incorporate
all costs, it is useful to report costs separately for each government level.
2.2 Costs
2.2.1 Definition of Costs
As previously discussed, this guide is aimed at persons within the agencies and
organizations funding and operating E&T programs and providing services used by
E&T programs. Consequently, the remainder of the guide focuses on costs accruing to
these agencies and organizations and ignores the perspective of participants and, hence,
that of society as a whole.
Therefore, we define the costs of an E&T program as expenditures on the resources
required to run the program. These expenditures result directly from purchases that are
necessary to provide services under the program, including expenditures on staff
salaries and fringe benefits; purchases of goods, services, office space, and equipment;
vendor payments; and day care and transportation. They also include special payments
that are an integral part of operating the program, such as subsidies paid to employers
who hire program participants and stipends or allowances paid to the program partici-
pants themselves.
Under this definition, costs do not include program-engendered changes in transfer
payments — for example, Temporary Assistance for Needy Families (TANF), Medicaid,
and unemployment compensation — even if these payments increase unintentionally, or
changes in the administrative costs of running transfer programs. Similarly, effects on
the earnings and tax payments of program participants are excluded, even if participant
earnings and tax payments decline. Costs borne by program participants are also not
included under this definition — for example, work-related expenses such as child care
and transportation costs that result because the program increases employment or
hours worked. None of these items represents expenditures that are directly incurred by
agencies and organizations in running E&T programs.
For most purposes, the value of donated equipment or time would also not be count-
ed as costs because such donations do not result in expenditures by the agencies and
organizations operating E&T programs. However, if the program is being considered for
adoption elsewhere, it is important that the agencies and organizations considering it be
provided information on the market value of donated time and equipment because they
may not receive these goods as donations, but have to purchase them instead.
As should be apparent, a cost analysis of an E&T program provides only a very par-
tial look at program effects. For that reason, the interpretation of a cost analysis should
account for what is left out as well as what is included. It would make little sense to
attempt to minimize expenditures on an E&T program without also considering the
implications for benefits from the program — for example, whether the effects of the
E&T program on earnings or on the receipt of welfare payments will diminish with the
reduction in E&T costs.
2.2.2 Gross Costs Versus Net Costs

Costs were previously defined as expenditures on the resources required for operat-
ing an E&T program. These expenditures can be measured in either gross terms or net
terms. Gross costs are relatively straightforward to understand and measure. They are
simply direct outlays required to operate a program. Gross costs are usually sufficient for
use in monitoring an ongoing E&T program or in administering performance contracts.
Information on gross costs is also useful in planning for a new program. It is also usually
appropriate to use gross costs in selecting one program model over another. In addition,
if another site is considering adopting the program, information on the program’s gross
costs is useful because it allows the site to compare the gross costs of its own current
program with a possible alternative.
Net costs are more complex than gross costs and considerably more difficult to com-
pute. Net costs are the change in costs that result from a specific decision — for example,
a decision to adopt a new program, to expand or scale down an existing program, or to
modify some component of an existing program (for example, to increase class size).
Hence, net costs are more appropriately used in making decisions of this sort than are
gross costs. Net costs, like gross costs, are also useful for budget planning purposes.
Finally, net, rather than gross, costs should always be used to conduct formal cost-bene-
fit evaluations of E&T programs. Thus, information on net costs is useful to decision-
makers at both government agencies and private-sector service providers.
2.2.3 Start-Up Costs

The costs required to initiate a new E&T program — expenditures on curriculum
development, writing regulations, developing computer software, and so forth — are
often called start-up or development costs. Start-up costs also include costs that result
from learning how to operate a new program. For example, costs could fall over time
because of various improvements that are made in program efficiency.
There is an important distinction in the treatment of start-up costs when estimating
the net costs of a not-yet-implemented E&T program versus an ongoing program. If an
E&T program has not yet been implemented, then whether start-up costs are actually
incurred depends on whether or not the program is adopted; they are part of the cost of
adopting the program and should be included in estimating the program’s net costs.
However, in the case of an ongoing program, start-up costs have already been incurred
and no decision that is made with respect to the ongoing program, including terminat-
ing it, can affect these costs: they are fixed, or “sunk.” Thus, start-up costs should not be
included as part of the net costs of ongoing programs.
WHAT IS THE DIFFERENCE BETWEEN NET COSTS AND GROSS COSTS?
To understand how net costs differ from gross costs and why they are more appropriately
used for certain purposes, consider a decision on whether to replace an existing E&T program
with a new one. If the proposed new program is adopted, then expenditures on the existing
program will no longer be incurred. To compute the change in costs resulting from adopting
the new program (that is, its net costs), the gross cost of the existing program must be sub-
tracted from the gross cost of the proposed program. This may not be the end of the process,
however.
For example, the agency that would run the proposed program might already have furniture
e.g.
and classroom space that it is not using in running the current program, but would use for the
new program. Unless this furniture and classroom space will be used for some non-E&T pur-
pose, but cannot serve this purpose once the new E&T program is implemented, using the
existing furniture and equipment in operating the E&T program imposes no additional costs.
Thus, in computing the net cost of the new program, these elements are appropriately subtract-
ed from the program’s gross cost.
Similarly, in decisions concerning whether to go ahead and adopt a proposed program, the
cost issue typically concerns how much more the new program will cost than an existing one,
not its absolute (gross) cost. An E&T program that results in little or no additional costs is not
very expensive, even if its gross costs are large. In budgetary planning for the program, the
important issue is not the total amount of staff and equipment required, but the additional staff
that must be hired and the additional equipment that must be purchased. Likewise, if a decision
is being made about whether to expand or contract an ongoing program, the relevant issue is
the additional costs or savings that would result — in other words, the change in net costs.
To help ensure that start-up costs are excluded, cost analyses of ongoing programs
are sometimes delayed until after the program has reached a steady state — that is, until
after the program has reached the point at which its costs over time are expected to be
relatively stable.
2.2.4 Average Costs Versus Marginal Costs

The average cost of an E&T program is simply the cost per participant. Marginal
costs are the additional costs that result from serving an additional participant or the
cost savings that result from serving one less participant.
In conducting cost analyses of E&T programs, marginal costs are often very difficult
to measure. Thus, average costs are typically used instead. Although this decision is
appropriate for some purposes, such as resource planning for a new program or moni-
toring an existing program, it is not appropriate in decisions concerning expanding or
contracting an ongoing program. For that purpose, marginal costs should ideally be
used instead. However, in practice, as will become evident in chapters 3 and 5, it is not
always easy to estimate marginal costs, and average costs are often used as the best avail-
able approximation.
e.g. EXAMPLES OF AVERAGE VERSUS MARGINAL COSTS

Marginal cost can differ considerably from average cost. Consider the following
examples:
• Certain pieces of equipment and certain supervisory activities may be required

regardless of whether an E&T program is large or small. For example, whether an
E&T program is in every county in a state or in only a few counties may have rela-
tively little effect on the number of persons involved in administering it at the state
level. If that is the case, as the number of program participants expands, program
costs will not grow as quickly. Consequently, marginal costs will be less than average
costs.
• If the size of an E&T class is expanded, it may not be necessary to increase the
space used or the number of teachers. In that case, the marginal costs will again be
less than the average costs.
• If it is necessary to make major new purchases or to rent new facilities in order to

expand, marginal costs may exceed average costs. This may be especially impor-
tant in the case of small organizations that begin bidding to provide E&T services.
• In expanding an E&T program, the new participants may differ in important

respects from those currently enrolled. For example, they may have fewer or
greater barriers to employment and, hence, require less expensive or more
expensive services. In the first instance, the marginal costs will be less than the
average costs; in the latter case, they will be greater.
2.2.5 Joint Costs

Some resources are used for several purposes at once. For example, the same com-
puter terminals at a welfare agency might be used to administer welfare payments and to
track clients enrolled in E&T programs. When resources are used for multiple purposes,
their costs are sometimes called joint costs.
In a cost analysis, resources that are used jointly raise the difficult issues of what
proportion of the total cost of the resource to assign to the program being analyzed. Any
treatment of joint costs is likely to be somewhat arbitrary. One possibility, and the one
most commonly used in practice, is to allocate cost on the basis of the proportion of the
resource used for each purpose.
e.g. For example, if one-fourth of the time spent using computer terminals involves the E&T
program being analyzed, then 25 percent of the total cost of the terminals would be
assigned to the program in question. In contrast, 100 percent of the cost of terminals
used exclusively to track E&T participants would be allocated to the program.
A better approach would be to attempt to determine whether the resource would

have been purchased at all in the absence of the program. If the answer is yes, then the
program engenders no additional costs and none should be allocated to the program. If,
however, the resource is purchased only as a result of the program, then its total cost
should be allocated to the program, even if it is also used for other purposes.
2.2.6 Regular Versus Special Costs

Regular costs occur in the normal operation of an E&T program; they are essential
to running the program. Special costs, as the name implies, are not essential to the opera-
tion of a program. A good example is provided by the cost of conducting cost analyses
themselves. Cost analyses conducted in connection with implementing a new E&T pro-
gram or to monitor an existing program would usually be viewed as regular costs. It is
difficult to imagine implementing an E&T program without first costing out the pro-
gram, and analyzing cost data for monitoring purposes is a routine part of operating an
E&T program. Conducting a cost analysis as part of a cost-benefit evaluation of an
ongoing E&T program, in contrast, is a special cost. Although useful, it is not a necessary
component of operating the program. While regular costs should be counted in con-
ducting a cost analysis, special costs usually should not be.
2.3 The Time Dimension

Cost analyses of not-yet-implemented E&T programs typically attempt to predict the
annual costs that will result from adopting the program. Annual cost figures are espe-
cially appropriate for budgetary and planning purposes. However, as discussed in some
detail in chapter 5, annual costs may vary considerably from one year to the next, and it
is important that this variation be taken into account.
As indicated in chapter 3, cost analyses of ongoing E&T programs, in contrast to
those of not-yet-implemented programs, usually examine the costs incurred by specific
groups (or cohorts) of persons who enrolled in the program during a fairly narrow time
span — for example, over a three-month period. Many (or even all) of the members of a
cohort of E&T enrollees may remain active in the program for far less than a year. More-
over, individuals from outside the cohort may also enroll in the program during the
same year. Thus, costs incurred by the cohort do not correspond to annual program
costs.
Even if all the program costs incurred by a cohort accrue over a relatively short
period of time, however, the E&T program may still affect members of the cohort for
many years after the program ends. For example, as a result of the program, the earnings
received by a former participant may increase, and this increase may continue to be
received as long as the participant remains in the work force. While these long-term
effects must be taken into account in cost-benefit evaluations of ongoing E&T programs,
they rarely need to be considered in pure cost analyses of these programs.
2.3.1 Role of Discounting in Conducting Cost Analyses

Viewed from the present, dollars that need not be paid until some future date do not
impose as great a cost as the same amount of dollars that must be paid today because
they can be used for some other purpose until they are paid. For example, they could be
invested and earn interest. If E&T costs are projected and summed over several years, as
THE MECHANICS OF DISCOUNTING
The procedure required for discounting is relatively simple. For each future year for which costs
are estimated, the following formula can be used:
PV(C y ) = C y /(1+d) y
In this formula, y is a number that represents the future year for which costs are estimated (year
1, year 2, year 3, and so forth), C y is the dollar value of the estimated costs for that year, PV(C y )
is the present value of C y , and d is the discount rate. The value of d is controversial among
economists, but, in practice, values between .03 and .1 are typically used. A good approach is to
select two or more values within this range and see whether the results of the cost analysis are
very sensitive to using these alternative values.
To illustrate discounting, assume that it is predicted that a not-yet-implemented E&T

e.g.
program will cost $1,500,000 in its first year of operation, $1,200,000 in its second year, and
only $900,000 in its third year. Without discounting, total costs over the three-year period
would be (inappropriately) predicted to be $3,600,000. Using a discount rate of .05, total costs
would be predicted to be $3,294,461 and are computed as follows:
$1,500,000/(1+.05)1 + $1,200,000/(1+.05)2 + $900,000/(1+.05)3 = $3,294,461
they might be for budgeting purposes, it is necessary to adjust to account for the fact
that current costs impose a greater burden than costs that occur in the future. This ad-
justment is called discounting. The discounting procedure, in effect, allows one to deduct
the interest that could be earned until the costs are incurred and must be paid. The
resulting dollar figure is called the “present value” of the costs. When present values for
different years have all been similarly adjusted, they can be appropriately summed.
2.3.2 Adjusting for Inflation

The purchasing power of a dollar erodes over time because of inflation. Conse-
quently, if the data used in a cost analysis are several years old, it is helpful to those inter-
preting the analysis if costs are stated in current dollars. This is especially important if
changes in the costs of a program over time are being examined or if two or more pro-
grams are being compared and the cost data for the different programs were collected
during different years.
Cost figures can be converted to current dollars by multiplying them by an adjust-
ment factor obtained from the Consumer Price Index (CPI). Alternatively, the values
used by the U.S. Department of Commerce to adjust estimates of the gross domestic
product for inflation (the GDP deflator) can be used instead. It does not matter very
much whether the CPI or the GDP deflator is used; they result in adjustments of similar
magnitude. Up-to-date values for the CPI and the GDP deflator can be obtained from
the Monthly Labor Review and the annually published Economic Report of the President,
respectively.
THE MECHANICS OF THE INFLATION ADJUSTMENT
The inflation adjustment factor is easily computed as follows:

Adjustment Factor = I cy /I py ,
where I cy is the value of either the CPI or the GDP deflator for the current year and I py is the
corresponding value for the year for which the cost data are available.
e.g.
For example, if the cost data were two years old and the CPI equals 200 for that year and 220 for
the current year, the two-year-old cost figures can be adjusted to current prices by multiplying
them by 1.10 (or 220/200). Specifically, a two-year-old cost of $100 would be multiplied by 1.10
in order to obtain a current dollar value that equals $110.
2.3.3 Adjusting for Large Equipment Purchases

Durable goods purchased by an organization, such as computer terminals and
furniture, have a useful life of many years. The organization's expenditure on such
equipment should not be assigned entirely to the year it is made, but should be spread
over all its years of useful life. Straight-line depreciation is the easiest of the methods
that can be used to do this. For a description of alternative methods, see Kumen H.
Jones et al., Introduction to Financial Accounting: A User Perspective (Englewood Cliffs,
NJ: Prentice-Hall, 1996, chapter 8).
THE MECHANICS OF STRAIGHT-LINE DEPRECIATION
Straight-line depreciation can be accomplished in three steps. First, the number of years the
equipment to be depreciated will be used must be determined. Second, if the equipment will
then be sold, its sales price (which is usually called its “salvage value”) must be predicted. Third,
the annual cost of the equipment is determined by deducting its salvage value, if any, from its
purchase price, and then dividing this total by the number of years it is expected that the equip-
ment will be used.
e.g.
For example, if equipment that is purchased for an E&T program for $50,000 will be sold after
five years for $10,000, then $8,000 (or [$50,000–$10,000]/5) would be an appropriate value to
use for the equipment in determining the annual cost of the program.
Chapter 3
Costing Out Ongoing Programs

The objective of this chapter is to provide the information you need to conduct a cost
analysis of an ongoing employment and training program for the purpose of monitor-
ing or evaluating this program. The methods described in this chapter can also be used
in determining the costs of individual program activities (such as vocational training)
and support services (such as child care). The chapter delineates the data needed to con-
duct the analysis and describes the required steps.
Cost analyses of ongoing E&T programs are described before cost analyses of not-
yet-implemented programs are discussed because the data required for the former are
available from the program itself. Cost analyses of not-yet-implemented programs, in
contrast, require predictions; but, as discussed in chapter 5, these predictions can usually
be based, in part, on information about the costs of similar ongoing programs.
Moreover, as indicated in chapter 2, in determining the net cost of a not-yet-implement-
ed program, the costs of the existing program should be subtracted from the predicted
costs of the proposed program. Thus, it is helpful to understand how cost analyses of
ongoing E&T programs are conducted before conducting analyses of not-yet-imple-
mented programs.
Readers who use the information from cost analyses but do not plan to conduct one
of their own will be mainly interested in sections 3.1 and 3.2 and the text box found at
the end of section 3.4. Section 3.1 discusses why costs are typically estimated on a per
case basis, while section 3.2 describes the data needed for conducting a cost analysis.
The text box at the end of this chapter provides a brief summary of the steps that should
be followed in conducting cost analyses of ongoing programs.
17
3.1 Costs per Case Versus Aggregate Costs

As mentioned in chapter 1, employment and training programs may be broadly divided
into two categories: voluntary or mandatory. Voluntary E&T programs provide training
for individuals who apply for them and meet certain criteria of need, such as being
unemployed or having low incomes. Mandatory E&T programs require participation by
recipients of government transfer payments who meet certain other criteria — for
example, with respect to the length of time they have been on the transfer program rolls.
The distinction between voluntary and mandatory E&T programs is rather formal
in theory, but may blur in actual practice. Not every qualified person who applies to a
voluntary program, for example, is necessarily accepted and, owing partially to imper-
fect enforcement, not every person who meets the criteria for a mandatory program
necessarily participates. Nonetheless, the distinction is important because cost analyses
of the two types of programs differ in important respects. Cost analyses of ongoing vol-
untary E&T programs usually compute the costs engendered by a typical individual
who has been accepted into the program (costs per accepted case). Often, this is further
limited to a typical accepted case that actually participates (that is, receives services) in
the voluntary program (costs per participant).1 Cost analyses of ongoing mandatory pro-
grams, in contrast, generally calculate the costs generated by a typical transfer payment
recipient who has been assigned or referred to the program (costs per referred case). For
mandatory programs, unlike voluntary programs, cases that receive program services
and those that do not must be included in computing average costs because both incur
costs. Service recipients obviously engender costs; cases that do not receive services
engender costs as a result of expenditures needed to enforce the program’s participation
requirement.
As just indicated, program costs for both voluntary and mandatory E&T programs
are typically computed on a per case basis, rather than being aggregated. However, there
are situations in which it may be desirable to determine aggregate costs — for example,
for budgeting purposes. The costs of ongoing programs are usually computed on a per
case basis (be it case accepted or referred), rather than being aggregated, for several rea-
sons. First, in comparing costs over time or among different programs, the total number
of persons served by the program will drive aggregate costs. Thus, better comparisons
can usually be obtained by looking at costs per case. Second, as discussed earlier, for
many purposes, the costs of E&T programs should be compared with the benefits from
these programs, such as increases in earnings and reductions in transfer payments.
However, these benefits are almost always measured on a per case basis. Consequently, if
program costs are to be compared with program benefits, they must also be computed
on a per case basis. Third, information on the costs of ongoing E&T programs often
serves as a key input in costing out programs that have not yet been implemented. As
1. For simplicity, in the remainder of this guide, we refer only to costs per accepted case in discussing voluntary pro-
grams. However, it is equally valid to estimate costs per participant for voluntary programs. This is not valid for
mandatory programs because, as discussed in the text, those who receive services and those who do not both incur
costs.
Chapter 3: Costing Out Ongoing Programs 19
shown in chapter 5, use of this information is greatly facilitated if it is reported on a per

case basis. Thus, this chapter focuses on the computation of program costs per case.
In interpreting costs that are reported on a per case basis, it is important to keep in
mind that they are an average of higher costs for some cases and lower, or even zero,
costs for others.
e.g. In the case of a voluntary program, for example, some unemployed persons may find
jobs between the time they are accepted into the program and the time services actually
become available; other cases drop out shortly after they begin to receive services.
Similarly, in the case of a mandatory program, some individuals may leave the transfer
program rolls (perhaps because they have found employment) between the time they
are referred to the program and the time services are ready to begin. Moreover, other
referred individuals may be excused from participation prior to the time when services
become available (perhaps because of illness or because they find part-time employ-
ment). Still others may refuse to comply with the requirements of the mandatory pro-
gram, even if their transfer benefits are reduced as a result. Thus, costs-per-case esti-
mates include small values for cases that incur small costs (for example, only costs
resulting from a mandatory program enforcement process) and zero values for cases
that incur zero costs.
3.2 Data Needed for Cost Analyses of

Ongoing E&T Programs
3.2.1 Records Program Agencies Should Keep
For administrative purposes, many agencies that are operating E&T programs
already maintain all or most of the data needed for cost analyses of these programs,
often in computer-retrievable form. The only items that must be added are those that
provide information not already being collected. As indicated in later sections, because
different agencies maintain data in different ways, procedures used to conduct a cost
analysis of an E&T program operated by one agency will differ somewhat from proce-
dures used to conduct an analysis of a similar program run by another agency. In other
words, to some extent, the form of the data determines the nature of the analysis.
Moreover, even within the same agency, efforts may have to be made to reconcile data
sets that are maintained in different formats.
To the extent time permits, it is important to attempt to ensure that the data collect-
ed for the cost analysis are as accurate as possible. If there are errors in the data, they
may need to be “cleaned” by first making detailed inquiries about individual data entries
that seem inconsistent, implausible, or otherwise suspect, and then correcting these
items. If this is not possible, other sorts of adjustments may be required.
The following data items are critical to cost analyses of ongoing E&T programs:
■ time spent by program participants in each program component;
■ allowances paid to program participants — for example, for day care or clothing;
■ vendor payments made on behalf of program participants — for example, for day care;
■ subsidies paid to employers who hire program participants;
■ stipends paid to those participating in the program;
■ the salary and fringe benefits of each staff member involved in the operation of the E&T
program being analyzed, even if many of these persons devote only part of their time to
the program;
■ special purchases (such as forms, computers, and furniture) made specifically for the
program’s use;
■ office overhead rates or, alternatively, the costs of the various items that go into overhead
(for example, the rental value of office space, telephone bills, and equipment costs), even
if these items are used for multiple purposes and, hence, are not exclusively devoted to
the E&T program being analyzed.
e.g. A $5 daily allowance was paid for attendance at job search workshops run as part of an
E&T program for Aid to Families with Dependent Children (AFDC) applicants con-
ducted during the 1980s.2 By comparing the number of persons who attended the work-
shops with the number for whom allowances were recorded, the MDRC research team
determined that there were allowance records for only about 5 percent of those who
actually received allowances. To adjust for this discrepancy, MDRC inflated the recorded
allowance amounts.
3.2.2 Data Needed from Outside the Program Agency

The agency with direct responsibility for an E&T program may not perform many
of the services participants in the program receive. A welfare agency may operate an
E&T program for public assistance recipients, for instance, but much of the actual train-
ing may be conducted by local community colleges. Depending on contractual arrange-
ments, the welfare agency may or may not make tuition payments to the community
colleges. However, even when such payments are made, they may not accurately reflect
the full cost incurred by the colleges. The colleges may depend on revenue collected
from property taxes, as well as tuition, but the welfare agency may be required to pay no
more than the standard tuition amount. Other payments to outside organizations — for
example, those made to day care providers — may, in contrast, reflect the full costs of
providing a service. In the latter case, the cost analysis can rely entirely on the program
agency’s records, but in the former case, information must be obtained directly from the
service provider.
2. Barbara Goldman, Daniel Friedlander, and David Long, Final Report on the San Diego Job Search and Work
Experience Demonstration (New York: MDRC, 1986).
Cost information provided by organizations other than the program agency are
unlikely to be as detailed as the agency’s own records since the agency has no control
over external records. However, in most instances, the information provided by external
sources is sufficient.
3.3 Steps in Estimating the Gross Costs

of Ongoing Programs
This section lists the seven steps required to determine the gross costs of ongoing pro-
grams, discusses why each step is needed, and describes approaches that can be used to
complete each step. Section 3.4 then describes how net costs can be obtained, once gross
costs are estimated.
A spreadsheet that is based on the seven steps is available on the accompanying

disk.
A word of caution: The seven steps described in this section can be used to estimate
the gross costs of virtually any ongoing program. The procedures for completing each
step, however, depend on the data and time available to the analyst. Specific program
circumstances may require the analysis to vary in detail from some of the suggestions in
this section, and approximations and compromises may be necessary. Those who are
extremely time-constrained might want to consider using some of the values for other
E&T programs that are provided in Appendices B–F (although, as discussed in chapter
5, these values are really more appropriately used in costing out a not-yet-implemented
E&T program). It is important to be aware of the errors that may result from any neces-
sary approximations and compromises, and to ensure that users of the cost analysis are
apprised of potential errors.
Step1. Enumerate Program Components

Each of the activities and support services provided by the E&T program undergoing a
cost analysis needs to be listed. Program activity categories might include orientation,
assessment, job search, basic education, vocational training, work experience, and so
forth. Support service categories could include child care, transportation, and
allowances paid to program participants.
The exact categories will, of course, be determined by the nature of the program
itself. The categories should be sufficiently inclusive so that each cost engendered by the
program falls under some category. To help ensure that program costs do not get over-
looked, it may be helpful to develop a detailed flow diagram of the sequence of E&T
activities. (Examples of such a diagram appear in the following chapter in Figures 4.1
and 4.2.) In general, it is better to use more detailed, rather than less detailed, categories,
since categories can always be combined later. Information about the costs attributable
to a narrowly defined program activity or support service is useful, for example, in
forming judgments as to whether expenditures on the component are too high or too
low. Realistically, however, each category must be defined so that data on unit cost, par-
ticipation rates, and length of participation that correspond to the category are obtain-
able. Hence, the category definitions will be at least partially determined by the data
available for the cost analysis. Moreover, the number of categories should not be so large
that they are unwieldy or confusing to work with.
Step 2. Compute Unit Costs

Unit costs are simply the costs of providing a particular program component to one per-
son over a specified time period. The appropriate time period to use in computing unit
costs will vary with the component being examined. For example, the unit cost of pro-
gram orientation is usually most readily computed as the cost of providing one session
to one person, while the unit cost of job search, work experience, transportation
allowances, or child care could be computed as the cost of providing these activities and
support services for either a week or a month. If the cost of providing a particular pro-
gram component is shared by two organizations — as when, for example, the E&T
sponsoring agency and a community college divide the cost of providing vocational
training — it is usually both easier and more useful to compute the unit cost incurred
by each organization separately.
a. The Basic Computation. Unit costs can be determined by first calculating the
total cost of providing a given program component over a specific time period (for
example, a session, a school semester, a week, a month, or a year) and then dividing by
the number of individuals who participated in the activity or received the service over
the corresponding time period. Thus, data are needed on both total cost and number of
participants during the time period. (As discussed in chapter 2, start-up costs should
usually be excluded from estimates of the costs of ongoing programs. Thus, the time
period used for computing costs should ideally be a “steady state” period for the pro-
gram — that is, a period of relative stability.)
e.g. Assume that an ongoing E&T program spent $100,000 on job search during a one-year
steady-state period and that, during this period, an average of 100 persons participated
in job search each month. Then, the unit cost for job search equals
$100,000 annual costs/(100 participants x 12 months)

= $83.33 per participant per month.
It is important to note that in calculating unit costs in this manner, any case man-
agement costs that are part of the activity will be included in the measure of unit costs.
One example is the cost of enforcing the participation requirement in a mandatory
program. Moreover, all the costs of a program are ascribed to those individuals who
actually participate in the activity. This could be important if some persons who are
assigned to the activity fail to show up, as these “no shows” may nonetheless engender
costs as a result of staff time required to process paperwork and encourage them to
attend. These costs will be included in the numerator of the unit cost measure, but not
in the denominator.
e.g. If, in the illustration appearing above, $95,000 of the $100,000 expended on job search
was incurred by the 100 job search participants, and the other $5,000 was engendered by
20 no-shows, the entire $100,000 would be ascribed to the 100 participants — because it
took $100,000 to produce 100 participants. Moreover, as the remaining calculations are
all based on persons who participate in each program activity, costs engendered by no-
shows would be inappropriately left out of the final cost estimates if costs incurred by
no-shows, as well as those engendered by job search participants, were not included in
the numerator of the unit cost ratio.
b. Determining Staff Time. If each E&T program staff person served program
participants exclusively and engaged in only one activity (for example, instruction, case
management, counseling, or administering voucher payments) and resources such as
rental space and computers were similarly exclusively dedicated, computing unit costs
would be greatly simplified. In practice, however, each staff member often serves both
persons in the program being analyzed and persons outside the program, and engages
in a variety of activities (such as orientation sessions, counseling, and individual job-
search monitoring). Physical resources may similarly serve a multiplicity of functions.
For the purpose of cost analysis, it is critical to determine the proportion of total
staff time and personnel costs devoted to the E&T program being analyzed. The cost of
physical resources must be similarly allocated between program and nonprogram uses.
In addition, allocations must be made among the various program components so that
costs can be appropriately assigned to the various program component categories listed
under step 1.
There are at least three different ways to allocate staff time among different activities:
time studies, time sheets, and interviews. Many government agencies routinely engage in
time studies in which selected staff members are monitored by others to determine how
their time during a workday is allocated. Findings from these studies can be readily
adopted for use in an E&T cost analysis if the time studies collect information in a man-
ner consistent with the needs of the cost analysis. For instance, if there is a desire to dif-
ferentiate between time spent in helping prepare a client for a job interview and time
spent in counseling a client on resolving internal family problems, then the instrument
used in the time study may have to be modified to distinguish between these two activi-
ties. Furthermore, if the information is not otherwise available, it may be necessary to
record the number of persons who receive each type of counseling. If possible, it is also
often useful to differentiate between time spent in performing a task for program partici-
pants and time spent performing the same task for persons outside the program.
Time sheets are similar to time studies, except that staff members themselves are
responsible for keeping track of the time they spend on different tasks. The time sheets
are usually filled out over a period of several weeks. Data on the number of persons who
participate in each program activity during these weeks could also be recorded on the
time sheets if not available elsewhere. Once again, the instrument used must be consis-
tent with the requirements of the cost analysis. It may be desirable, for example, to sepa-
rate staff time spent in case management from time spent in other ways; this can be
accomplished by using a carefully designed time sheet. For illustrative purposes, the
time sheet used by MDRC in its recent cost analysis of two welfare-to-work programs in
Minnesota — the Minnesota Family Investment Program (MFIP) and STRIDE, the
state’s version of the federal-state Job Opportunities and Basic Skills Training (JOBS)
program — is shown in Table 3.1.
In interviews, staff members (or their supervisors) are asked about the amount of
time they typically devote to each activity of interest. This approach is, in general, inferi-
or to the use of time studies or time sheets because it relies on the staff ’s ability to recall
their time allocations retrospectively. This is likely to result in errors, especially when rel-
atively small amounts of time are devoted to a particular activity. Thus, if possible, the
use of interviews should be limited to supplementing information obtained from time
studies or time sheets.
Once staff time is allocated among activities, further steps may be necessary to allo-
cate time devoted to performing a particular activity between persons in the E&T pro-
gram that is being analyzed and persons outside the program. This is obviously unnec-
essary if the task is performed exclusively for those in the program or if staff performing
the task can differentiate between the time they devote to each group. This is not always
feasible, however; counselors may not even know whether a particular client is in the
E&T program. In such circumstances, time can often be allocated on the basis of the
number of persons in each group — assuming that, on average, the time spent perform-
ing the task for a program participant is roughly the same as the time spent performing
the task for a nonparticipant. This assumption may not, of course, be valid.
c. Treatment of Overhead. Once the fraction of time each staff member devoted to
various activities is known, this information can readily be converted to dollar figures by
multiplying the fractions by the staff members’ salaries over an appropriate time period,
such as one year. After doing this, it is important to distinguish between the costs that
are included in the resulting estimates and the costs that are left out.
e.g. Specifically, the estimates clearly include the program costs that accrue to line staff who
are directly involved in various E&T activities. However, costs associated with support
staff, such as secretaries, and supervisors may be left out of the estimates. In addition,
fringe benefits may or may not have been included in the salary figures used to obtain
the estimates. Finally, the costs of various physical resources, such as computer systems,
furniture, physical facilities, and telephones, will definitely be left out of the estimates.
For the purpose of this discussion, costs that are included in the estimates are
referred to as personnel costs and costs that are left out are referred to as overhead costs.
The estimates of personnel costs can be used in assigning overhead costs to various
E&T activities. This is usually done in one of two ways. The first method is to multiply
an organization’s overhead rate — that is, the organization’s total annual expenditures
divided by its total annual expenditures on staff salaries — by the estimate of the staff
Table 3.1
MINNESOTA FAMILY INVESTMENT PROGRAM/STRIDE TIME SHEET Staff Members’ Time Allocation to Program Components
Name County
Title Organization/Provider
Date TOTAL
Activity 11/11 11/12 11/13 11/14 11/15 11/18 11/19 11/20 11/21 11/22 HOURS
1 Up-Front Mandatory STRIDE Orientation
2 MFIP/STRIDE Orientation/Intake/lnitial Assessment
3 Operation and Monitoring of MFIP/STRIDE Career Exploration and Skill Building Workshops
4 Operation and Monitoring of MFIP/STRIDE Job Club
5 Operation and Monitoring of MFIP/STRIDE Job Search Classes/Workshops
6 Operation and Monitoring of MFIP/STRIDE Individual Job Search Activities
Chapter 3: Costing Out Ongoing Programs
7 Operation and Monitoring of CWEP

8 Basic Education Support and Monitoring
9 Post-Secondary Education and Vocational Training Monitoring and Support
10 MFIP/STRIDE Employment-Related Services and Monitoring
11 MFIP/STRIDE Recruitment
12 MFIP/STRIDE Child Care Referrals/Arrangements
13 MFIP/STRIDE Job Development/Job Placement/lnformal Job Search
14 MFIP/STRIDE Ongoing Case Management
a) MFIP/STRIDE Sanctioning/Non-Compliance Notification
b) Coordination with Other Agencies
c) Other MFIP/STRIDE Case Management Activities
15 Other MFIP/STRIDE Functions (not related to individual participants)
16 Operation and Monitoring of Food Stamp Employment and Training
17 Operation and Monitoring of Other Programs
18 Vacation, Sick Leave, Breaks
25
TOTAL HOURS
NOTE: A copy of the instructional material that accompanied this time sheet can be obtained upon request from MDRC.
SOURCE: MDRC.
personnel costs for each activity. In making this calculation, the salary values used to
compute the denominator of the overhead ratio must include only the salaries on which
the estimates of personnel costs were based. The numerator, however, should include the
costs of all the personnel and physical resources that are either directly or indirectly
involved in running the E&T program. The costs of personnel and resources that have
nothing to do with the program should be excluded from the numerator, however. As
the numerator of this rate will include both personnel costs and overhead costs, while
the denominator will include only the former, the rate itself will exceed one.
e.g. An overhead rate of 2.5 would imply that the total costs of the program are two and a
half times larger than the personnel costs. Thus, multiplying the estimate of personnel
costs of an activity by the overhead rate results in inflating or marking up the estimate
so that it includes both the personnel costs and the overhead costs associated with the
activity. If personnel costs were, say, $50,000, and the overhead rate was 2.5, total costs
(personnel costs plus overhead costs) would equal 2.5 x $50,000, or $125,000.
The second approach is to multiply total annual expenditures on each overhead

item (secretaries’ salaries, supervisors’ salaries, computers, rental space, telephone, and
so forth) by the fraction of the organization’s total staff costs devoted to each activity.
e.g. Say that 5 percent of the total staff costs of an organization that was operating an E&T
program was required to pay staff for aiding program participants in their search for a
job and that the organization’s rental expenditures were $100,000 a year. Given this
hypothetical situation, $5,000 (or .05 x $100,000) of rental expenditures would be
assigned to the job search component of the program.
In using either of the approaches just described, it is important to recognize that

durable equipment purchased by an organization, such as computer terminals and fur-
niture, have a useful life of many years. As discussed in chapter 2, a depreciation method
should be used to spread the organization’s expenditures on such equipment over the
equipment’s years of useful life.
Step 3. Select the Analysis Sample

Steps 4 and 5 in the cost analysis of ongoing E&T programs require estimating measures
of participation for individuals accepted into voluntary programs or referred to manda-
tory programs. But it is often unnecessary to obtain the data needed to make such an
estimate for every individual who has ever been accepted into or referred to the program
being analyzed. A subgroup, if sufficiently large, will often suffice. If a subgroup of indi-
viduals is used for the cost analysis, however, it is important that this subgroup be repre-
sentative of all those who are accepted or referred.
e.g. The analysis sample might consist of a cohort of individuals who were accepted or
referred over a given time interval, ideally a period during which the program was in a
steady state. If this cohort is sufficiently large, then a subsample of members of the
cohort could be randomly selected. If only those with social security numbers ending in
3 or 7 were selected, for instance, 20 percent (two numbers out of a possible ten) of the
cohort would be randomly selected.
Step 4. Determine Participation Rates

After the sample to be used in the cost analysis has been selected, the fraction of the
sample that participated in each of the activities and received each of the support ser-
vices listed in step 1 must be determined. In completing this step, it is often useful to
first define the minimum level of activity required before an individual will be counted
as participating in a particular program component — for example, receipt of at least
one day of vocational training or reimbursement for at least one week of child care.
Three sources of data can be used to determine participation rates: (1) a computer-
ized management information system that tracks the E&T activities in which individual
members of the analysis sample participated and the support services they received,
(2) a review of the individual hard-copy case files of members of the analysis sample,
and (3) surveys administered either in person or by telephone to each member of the
analysis sample. A computerized tracking system that contains complete, accurate, and
clean data provides an ideal means of computing participation rates. If such a system is
not available, however, then the other two data sources must be used instead. Unfortu-
nately, surveys are subject to recall error, while case files are laborious to use and may
contain incomplete information. However, using surveys and case files to verify and
supplement one another can increase accuracy. Moreover, if the computerized tracking
system exists but contains incomplete or inaccurate information, its reliability can be
checked and improved by using data from the other two sources.
NONPROGRAM COSTS
Some E&T participants may enroll in various training and education activities — for example,
a computer training course or college — even after they have formally completed the E&T
program, perhaps stimulated by their experiences in the program. Such activities, which are
sometimes called nonprogram activities, are outside the scope of the E&T program. Yet, they
obviously engender costs, even though the program agency will bear little or none of these
costs. For most purposes, these nonprogram costs can be ignored. However, in conducting a
very comprehensive cost analysis, perhaps as part of a cost-benefit analysis, they should be
estimated. To do this, it is quite likely that surveys will have to be used to determine participa-
tion rates in nonprogram activities, as the necessary information is unlikely to be available in
either computerized management information systems that are operated by the program or in
program case files. Because program agencies will, of course, be more interested in program
costs than nonprogram costs, the two types of costs should be recorded separately.
Step 5. Determine the Average Length of Participation

For each member of the analysis sample who participated in a given E&T activity or
received a given support service, the duration of participation or receipt must be deter-
mined. Averages can then be computed for each program component. The time period
that is used to compute the length of participation must be consistent with the unit cost
measure.
e.g. If the unit cost of vocational training is measured as the cost of providing vocational
training for one month to one person, then length of participation must be measured
as the number of months that a typical vocational training participant received such
training.
The data sources that can potentially be used to determine average lengths of par-
ticipation are the same as those that can be used to determine participation rates: com-
puterized tracking systems, case files, and surveys. Again, data from each of these
sources can be used to check the reliability and make adjustments in data from the other
sources. It should be noted that survey data on length of participation in a given pro-
gram activity are almost surely subject to greater recall error than survey data on
whether individuals participated in the activity at all.
Step 6. Compute Gross Cost per Case for Each Program Component
Step 6 is mechanical in nature once steps 1–5 have been completed. The cost of
each program component for a typical person who has been accepted into or referred to
the program is determined by first multiplying the unit cost of the activity or support
service (as determined by step 2) by the average length of participation in the compo-
nent (as determined by step 5). The product is then multiplied by the component’s par-
ticipation rate (as determined by step 4). Thus, costs for the component are averaged
over both participants and nonparticipants in the component, even if the costs incurred
by the latter group are zero.
e.g. Suppose it costs $130 to provide one month of vocational training to one person and
that those persons who received such training remained in it for an average of six
months. Then, the cost of vocational training would be $780 (or 6 x $130) for each per-
son who received it. However, if only 10 percent of the analysis sample actually partici-
pated in vocational training, the cost per case for this program component would be
only $78 ( or .10 x $780).
Step 7. Compute Total Gross Cost per Case

This step is also mechanical. An E&T program’s total gross cost per accepted or referred
case is computed by simply summing all the cost-per-case values obtained in step 6.
It is important to recognize that the cost-per-case values for the various program
components can be appropriately added together only because they are all computed by
averaging across the same group of individuals. Specifically, all the members of the analy-
sis sample are included in the averages, regardless of whether those individuals actually
participated in particular program components or not. Averages that are instead comput-
ed for groups that differ in composition and size cannot be appropriately summed.
e.g.
All 1,000 members of an E&T program analysis sample participate in an orientation ses-
sion at a cost of $10 per person, while one member of the sample receives a four-year
college education at a cost of $50,000. Total program cost per case is obviously not
$50,010 [or $10 + $50,000] but is instead only $60 [or $10 + ($50,000/1,000)]. In other
words, the average costs of the two program components cannot be summed unless they
are computed over the same group of individuals.
3.4 Converting Gross Costs into Net Costs

Chapter 2 suggested that for certain purposes, such as cost-benefit evaluations of ongo-
ing E&T programs and assessing what it cost to replace a previous E&T program with
the current program, the costs of ongoing programs are more usefully measured in net,
rather than gross, terms. Gross costs remain relevant, however, because measuring the
net costs of an ongoing E&T program requires that its gross costs be measured first.
Except in the case of a formal cost-benefit evaluation, net costs can, at least in prin-
ciple, be fairly readily determined if estimates are available of the gross cost per case of
both the ongoing program and the program it replaced (assuming there was one). As
mentioned in chapter 2, both estimates should ideally be obtained after the programs
reach a steady-state level of operation so that they exclude start-up costs. Given such
estimates, computing net costs would involve the following four steps. (If there was no
previous program, then the first three of these steps can be ignored.)
1. If the cost-per-case estimates for the two programs pertain to two different years, then
the estimate for the older program should be adjusted for inflation using the method
described in chapter 2 (section 2.3.2).
2. Costs per case should be transformed to aggregate annual costs for both estimates by
multiplying costs per case by the number of cases accepted or assigned during the year
for which each estimate was obtained (see the box below). This step will account for the
possibility that the number of cases in the old program and the current program may
have differed.
3. The aggregate annual cost of the old program should be subtracted from the aggregate
annual cost of the current program.
4. As discussed in chapter 2, the cost of any equipment that was not used in the old pro-
gram, but is used in the current program and would not otherwise be used for some
non-E&T purpose, should be subtracted from the estimate obtained in step 3, above.
Notice that in the four steps above, the net cost estimates rely on comparing indi-
viduals who are accepted into or assigned to two different programs. These two groups
of individuals may, of course, differ in a number of respects — demographic composi-
tion, education level, work experience, motivation, and so forth — and these difference
may, in turn, create cost differences between the two programs. For most purposes, this
is quite appropriate. After all, if a decision is made to replace one program with another,
the costs resulting from that decision are accurately reflected by the difference in the
costs of the two programs regardless of the causes of these differences.
THE TRANSFORMATION FROM COSTS PER CASE TO AGGREGATE ANNUAL COSTS
Costs per case can be used to calculate aggregate annual costs for ongoing programs by
multiplying costs per case by the number of cases newly accepted (in the case of a voluntary
program) or newly assigned (in the case of a mandatory program) during a year. Cases that
are already in the program at the beginning of the year incurred costs during both that year
and the previous one. Although the costs they incurred during the current year are omitted
from the calculation, this is roughly offset by the fact that some of the costs incurred by the
newly accepted or assigned cases will not be incurred until the following year. For purposes of
computing annual aggregate cost, the estimate of the average length of participation (step 5)
may have to be scaled back for those program components in which individuals often
participate for longer than 12 months, such as college or child care. To do this, no values for
individual cases should be permitted to exceed 12 months in determining the average length
of participation.
In cost-benefit evaluations, in contrast, net costs are usually estimated by comparing

costs engendered under two program regimes by two groups of individuals who are as
similar as possible. Persons assigned to or accepted by the program being evaluated are
usually referred to as the “program group,” while those under an alternative program
regime are usually called the “comparison group.” The reason for comparing two similar
groups is that the effects of the program being evaluated on such outcomes as earnings
and welfare payments would ideally be determined by comparing the earnings and wel-
fare payments of the same group of individuals under both program regimes. Because
individuals cannot be under two program regimes at the same time, this is obviously
impossible. Consequently, an approximation is obtained by comparing the earnings and
welfare payments of the program group with the earnings and welfare payments of a
similar comparison group. To maintain comparability, exactly the same approach is used
to determine net costs in a cost-benefit evaluation.
Because members of a comparison group do not have access to the services provid-
ed by the E&T program, they are likely to obtain those services from a wide variety of
other sources, and comprehensive data on such nonprogram E&T services are essential.
Moreover, these data must be similar to those for members of the program group so that
appropriate comparisons can be made. In addition, the funding sources of E&T services
received by members of the comparison group must be carefully tracked because the
agencies and organizations that provide these services may differ from those operating
the E&T program being analyzed.
Obtaining an appropriate comparison group for a cost-benefit analysis, measuring
the E&T costs incurred by this group, and then comparing these costs with those
incurred by the program group involves numerous technical issues that are well beyond
the intended scope of this guide. Thus, the discussion here is limited to brief comments
on a few general points. (There is an extensive professional literature that focuses on
these and related topics. Agencies that wish to conduct a cost-benefit evaluation of an
E&T program should obtain help from experts with experience in using comparison
groups to conduct such analyses.)
3.4.1 Obtaining Comparison Groups for Cost-Benefit Analyses

In many situations, some of the persons who would otherwise be accepted into or
referred to an E&T program are instead randomly assigned to a “control” group. While
ineligible to take part in the E&T program, members of the comparison group can
receive whatever employment and training services are otherwise available to them. The
major advantage of this “random assignment” strategy is that — when research samples
are large — persons in the comparison group will, on average, be very similar to persons
in the program group in terms of education, employment history, motivation, and so
forth, differing only by chance. Thus, E&T costs incurred by members of the program
group and members of this comparison group can legitimately be compared.
Sometimes, however, random assignment is not appropriate or feasible. For exam-
ple, agencies running E&T programs may be reluctant to follow the random assignment
strategy because they do not want to deny services to some persons who would other-
wise qualify, or they may simply find it inconvenient administratively to treat otherwise
similar persons differently. Further, there may be programs for which the supply of pro-
gram “slots” and potential applicants are approximately equal, and diverting some peo-
ple to a comparison group would lead to a reduction in services. Finally, in some pro-
grams the goal may be to completely “saturate” a community or group of families with
services, so random assignment would undercut the program model.
In such instances, a number of possible alternative strategies are available for
obtaining comparison groups. We mention only two here. (For a more comprehensive
discussion, see Daniel Friedlander, David H. Greenberg, and Philip K. Robins,
“Evaluating Government Training Programs for the Economically Disadvantaged,”
Journal of Economic Literature, December 1997, pp. 1809–1855.) In the first, the costs
incurred by persons accepted into or referred to the E&T program being analyzed are
compared with E&T costs incurred by similar persons prior to the introduction of the
program. The second possibility is to compare E&T costs at locations where the pro-
gram operates with E&T costs at locations where it does not.
One problem with nonrandom assignment approaches such as these two is that the
environmental circumstances — economic conditions, for example — facing the pro-
gram and comparison groups may differ in important respects. These differences could
affect participation rates and length of stay in E&T programs, which would, in turn,
affect E&T costs. A second problem is that the personal characteristics of the two groups
being compared may differ in important respects, and these differences may affect the
amount of E&T services they receive and, hence, the E&T costs they incur. Some of
these differences (for example, demographic differences) can frequently be controlled
for through various statistical procedures, but others (differences in motivation) often
cannot.
A word of warning: the comparison group should never be limited to only those
persons who actually receive E&T services. Instead, the comparison group should
include persons who are as similar as possible to those accepted into or referred to the
program of interest, regardless of whether or not they receive E&T services. Limiting the
comparison group to only those who actually receive services ignores possible differ-
ences in program participation rates that cause E&T costs to differ.
A SUMMARY OF THE STEPS FOR ESTIMATING THE GROSS COSTS

OF AN ONGOING PROGRAM
Step 1. Enumerate program components. List all the components of the E&T program — that is, all
the program activities and support services.
Step 2. Compute unit costs. Compute the average cost of providing one unit (for example, an hour or
a month) of each of the program activities and services listed in step 1 to one person.
Step 3. Select the analysis sample. Select a representative group of cases that have been accepted into
or referred to the E&T program. Program costs will be estimated for these cases.
Step 4. Determine participation rates. Estimate the fraction of the cases selected in step 3 that
participated in each of the program activities and received each of the support services listed
under step 1.
Step 5. Determine the average length of participation. Estimate the average length of time over
which those cases that participated in each program activity remained active and those cases
that received each support service continued to obtain the service.
Step 6. Compute gross cost per case for each program component. Compute the cost per case for
each program component as follows:
unit cost (step 2) x participation rate (step 4)
x average length of participation (step 5)
Step 7. Compute total gross cost per case. Compute the total gross cost per case for the E&T program
by summing all the cost-per-case values obtained in step 6.
Chapter 4
A Real-World Cost Analysis of an

Ongoing Program: Implications of
Program Design for Program Costs
This chapter identifies the key factors that cause cost estimates to vary and presents
detailed findings from an actual cost analysis — specifically, MDRC’s cost analysis of
JOBS programs in three different communities. The tables from the MDRC analysis pro-
vide examples of how information from your own cost analysis can usefully be present-
ed. The material in section 4.2 demonstrates how the steps described in chapter 3 can be
used to determine the cost of an actual ongoing E&T program. Our intent is to clarify
how these steps can be applied in practice. Finally, the factors that caused costs to vary
among the JOBS programs in the three communities are examined in some detail, with a
focus on one community, Riverside, California. These factors are important because cost
estimates from ongoing E&T programs provide a major source of information that can
be used for costing out not-yet-implemented programs. The factors that cause costs to
vary from one program to another, however, must be taken into account if information
from ongoing programs is to be most appropriately used for this purpose.
Readers who use the presented estimates to conduct cost analysis of a not-yet-
implemented program will find the cost figures presented in this chapter useful for con-
ducting sensitivity analysis of the sort described in chapters 5 and 6.
Readers who are interested in a cost analysis of ongoing programs will find Table 4.8
of particular interest as it illustrates how the steps developed in chapter 3 can be readily
used in a spreadsheet. Table 4.8 can be replicated using the file on_going on the disk.
33
4.1 JOBS Cost Analysis

The JOBS program was created by the 1988 Family Support Act and was the successor
to the Work Incentive (WIN) program.1 JOBS mandated participation in employment
and training activities by AFDC recipients with minor children more than three years
old.2 It established participation targets and required welfare agencies to provide a mix
of services with an emphasis on job search, education, and training. As part of the
seven-site JOBS Evaluation — now titled the National Evaluation of Welfare-to-Work
Strategies3 — MDRC recently completed cost analyses of JOBS programs in three sites:
Grand Rapids, Michigan; Atlanta, Georgia; and Riverside, California. The evaluation is
funded by the U.S. Department of Health and Human Services (HHS), with additional
support from the U.S. Department of Education.
For purposes of the evaluation, Grand Rapids, Atlanta, and Riverside each ran two
parallel programs: one that focused on rapid job placement, called the Labor Force
Attachment (LFA) model, and a second program that focused on education and train-
ing, the Human Capital Development (HCD) model. Both of these program models
have been widely used in the United States. The HCD model focuses on skill-building
and education prior to employment. Proponents of the HCD approach believe that
high-paying, stable jobs will keep welfare recipients from again needing to seek public
assistance. They assert that allowing welfare recipients to prepare for “better” jobs
through education and training will help them stay off welfare permanently. The LFA
program model, in contrast, focuses on fast placement of program participants in the
labor market. The position of supporters of the LFA approach is that almost any job is a
positive first step. They argue that advancement will come through acquiring a work
history and training on the job.
In the JOBS evaluation, MDRC has utilized a random assignment research design.
Thus, in Grand Rapids, Atlanta, and Riverside, MDRC randomly assigned AFDC recipi-
ents to one of three research groups, which were established in each site: a group that
received JOBS services under the HCD model, a group that received JOBS services
under the LFA model, and a control group that did not receive any JOBS program ser-
vices but were free to seek out, on their own initiative, training and education programs
available in their communities. Random assignment assured that the individuals in the
three groups at each site had similar characteristics. As a result, the three groups are
comparable. This means that the costs of the two program models can be appropriately
compared for each of these three sites.
1. The specific provisions of JOBS (but not its overall aims) have been largely superseded by the federal Personal
Responsibility and Work Opportunity Reconciliation Act. The terminology “JOBS” and “JOBS programs,” however,
remains in common use.
2. States had the option of extending the mandate to welfare recipients with children older than one year.
3. Referred to as “the evaluation” or “the JOBS evaluation” in the remaining chapters.
Chapter 4: A Real–World Cost Analysis of an Ongoing Program 35
4.2 Findings
The costs of different employment and training programs vary for a number of reasons,
including differences in client characteristics and economic conditions across E&T sites
and differences in the program approaches selected by state and local government agen-
cies. Differences in program approaches help determine the program activities provided,
the level of participation in these activities, and the level of support services provided.
These factors are well illustrated by the three-site JOBS cost analysis. Table 4.1 summa-
rizes the cost findings for the LFA and HCD approaches at the three JOBS sites over a
two-year follow-up period beginning with each sample member’s entrance into the
study (month of random assignment). In general, the HCD program approach was more
expensive than the LFA approach. However, even within one approach, costs varied sub-
stantially among the three sites. For example, the Riverside LFA site incurred costs that
were less than half those at the Grand Rapids LFA site.
Table 4.1
GROSS COSTS OF THE LABOR FORCE ATTACHMENT AND HUMAN CAPITAL

DEVELOPMENT APPROACHES WITHIN TWO YEARS AFTER ORIENTATION
Cost per Referred Case

Labor Force Human Capital
Site and Cost Component Attachment (LFA) Development (HCD)
Atlanta
Operating costs $1,956 $2,901
Support services 883 1,020
Total 2,839 3,921
Grand Rapids
Support services 297 417
Total 3,109 4,739
a
Riverside (full sample)
Operating costs $1,105 —
Support services 122 —
Total 1,227 —
Riverside (subsample without a
high school diploma or GED)
Support services 103 231
Total 1,163 2,990
SOURCES: Gayle Hamilton, Thomas Brock, Mary Farrell, Daniel Friedlander, and Kristen Harknett, National
Evaluation of Welfare-to-Work Strategies: Evaluating Two Welfare-to-Work Program Approaches — Two-Year Findings
on the Labor Force Attachment and Human Capital Development Programs in Three Sites (Washington, D.C.: U.S.
Department of Health and Human Services and U.S. Department of Education, 1997). The Riverside LFA sub-
sample data are unpublished data from MDRC’s research.
NOTES: Estimated costs are in 1993 dollars.

Where data are not applicable, dashes are used.
a. The Riverside HCD approach was available only to sample members without a high school diploma or GED
certificate. The cost data for this subsample are shown in the last section of the table.
Without being aware of the key factors that cause the costs to vary across sites and
approaches, it is tempting to cost out a not-yet-implemented program by using cost
estimates calculated for other programs before making necessary adjustments to these
estimates. Unfortunately, you cannot simply choose from these estimates in costing out
your own E&T program; rather, you need to come up with your own estimates to cost
out your program. Estimates from ongoing programs do provide a useful starting
point, however. Therefore, this chapter provides you with cost estimates from the JOBS
evaluation. Appendices C–F present additional estimates from the JOBS evaluation and
from various other programs evaluated by MDRC. Additionally, chapter 5 presents
tools that will help you in tailoring these estimates to your program. All this is intended
to help you save time and effort in costing out your program. However, note that the
JOBS program and the programs presented in the appendices are mandatory pro-
grams. Consequently, their findings might not be directly applicable to voluntary pro-
grams.
This section focuses mainly on cost findings for Riverside. The sample for
Riverside’s HCD program included only individuals without high school diplomas or
GED certificates. Riverside’s LFA program approach, like Atlanta’s and Grand Rapids’,
included people with high school diplomas or GED certificates, as well as persons with-
out these education credentials. In order to examine the differences in the two
approaches in the Riverside program site, it is important to use samples that are as sim-
ilar as possible. Therefore, the findings for Riverside that are presented in this chapter
for both the LFA and HCD approaches are based on only program participants without
a high school diploma or GED certificate. Findings based on the full LFA sample are
presented in Tables C.1, D.1, E.1, and F.1. Findings for Riverside’s full LFA sample and
the subsample without a high school diploma or GED certificate differ only slightly,
however. Members of the subsample were much less likely than members of the full
sample to attend college or receive vocational training or work experience, but were
slightly more likely to participate in basic education and job search activities.
The remainder of this section covers the following: First, the section explores the
effects of different programmatic approaches on costs. This is followed by a discussion
of unit costs. Next, the section examines the effects of the length of stay in an activity on
costs. Then, the way in which the rates of participation in various program activities
influences costs is covered. Support service costs are considered next. The section con-
cludes with a discussion of how site and area characteristics affect program costs.
4.2.1 Effects of Program Models on Costs

Figures 4.1 and 4.2 depict the intended service sequence for the LFA and HCD pro-
grams, respectively. Because the LFA approach focuses on quick job placement, clients
were generally strongly encouraged to seek employment after an initial assessment. To
help them do this, primary LFA components included job search activities such as job
clubs and job development. The HCD approach, in contrast, focuses on in-depth educa-
tion and training for program participants. Thus, the majority of LFA cases were
assigned to participate in job search. In contrast, only a few HCD cases were assigned to
job search as their first activity, with the majority referred to basic education, vocational
Figure 4.1
INTENDED SEQUENCE OF ACTIVITIES IN A

LABOR FORCE ATTACHMENT WELFARE-TO-WORK PROGRAM
Orientation and
Appraisal
Job Club
including classroom
instruction and phone
room
Reappraisal leading to
Employment
Development Plan
Exit from
AFDC
and/or
Employment
Individual Basic Vocational Work

Job Club Job Search Education Training Experience
(short-term) (short-term) (short-term)
Reappraisal and update

of Employment
Development Plan
SOURCE: Gayle Hamilton et al., National Evaluation of Welfare-to-Work Strategies, 1997.

Figure 4.2
INTENDED SEQUENCE OF ACTIVITIES IN A

HUMAN CAPITAL DEVELOPMENT WELFARE-TO-WORK PROGRAM
Orientation and
Appraisal leading to
Employment
Development Plan
Basic Vocational College Work

Education Training Experience
Exit from
AFDC
and/or
Employment
Reappraisal and update

of Employment
Development Plan
Job Club,
including
classroom Individual Basic Vocational College Work
instruction Job Search Education Training Experience
and phone
room

training, and work experience instead. In Riverside, for example, 68 percent of the LFA
cases were assigned to job search upon their program entry, compared with only 2 per-
cent of the HCD cases.
These differences in assignment rates between the two program models have impor-
tant implications for costs. As derived from Table 4.2, in Riverside’s LFA program, 74
percent of the gross costs4 were allocated to job search activities and only 7 percent to
basic education. The Riverside HCD program, in contrast, allocated 64 percent of its
gross costs to basic education and only 19 percent to job search.
Table 4.2 presents Riverside’s gross costs per referred case for each service. The esti-
mates pertain to costs incurred by members of the LFA and HCD research samples
within two years following their assignment to the program. As discussed in chapter 3,
these cost estimates are averaged over both those persons who participated in the vari-
ous service components and those who did not. The costs of the individual activities
include the case management costs associated with these activities, including enforcing
requirements to participate in them.
4. The definition of gross costs as used in this guide differs from the definition used in the JOBS evaluation. As
defined in step 7 in chapter 3, gross costs include only costs incurred by the program. However, in the JOBS evalua-
tion, nonprogram costs, as well as costs incurred by the program, are included in the gross cost measure. For a dis-
cussion of nonprogram costs, see chapter 3.
Table 4.2
GROSS COSTS OF RIVERSIDE’S LABOR FORCE ATTACHMENT

AND HUMAN CAPITAL DEVELOPMENT APPROACHES WITHIN
TWO YEARS AFTER ORIENTATION
Labor Force Attachment (LFA) Human Capital Development (HCD)

Welfare Non-Welfare Gross Welfare Non-Welfare Gross
Activity Dept. Cost Dept. Cost Cost Dept. Cost Dept. Cost Cost
Orientation and appraisal $103 $0 $103 $96 $0 $96
Formal assessment 6 0 6 11 0 11
Job search 865 0 865 559 0 559
Basic education 26 60 86 872 1,042 1,914
College 0 0 0 0 0 0
Vocational training 0 0 0 37 142 179
Work experience 0 0 0 0 0 0
Subtotal (operating costs) 1,000 60 1,060 1,575 1,184 2,759
Child care 59 0 59 157 0 157
Other support services 44 0 44 74 0 74
Total 1,103 60 1,163 1,806 1,184 2,990
SOURCES: For LFA: unpublished data from MDRC research. For HCD: Gayle Hamilton et al., National Evaluation of
Welfare-to-Work Strategies, 1997.
NOTES: These data are for the subsample without a high school diploma or GED certificate. Estimated costs are in 1993 dollars.
Table 4.2 reports costs borne by the welfare agencies and by other agencies and
organizations separately. In general, the welfare department in Riverside financed the
day-to-day operations of JOBS, including providing case management, conducting ori-
entations and assessments, and operating job clubs, while basic education, vocational
training, and college were financed mainly by other agencies and organizations such as
community colleges, adult schools, and proprietary schools. For budgetary purposes, it
is obviously important to distinguish between costs borne by welfare departments and
costs borne by other agencies and organizations.
Table 4.2 clearly demonstrates how the focus on different service components affects
costs. Riverside spent only $86 on basic education per referred case per month under the
LFA approach, but under the HCD approach, which focuses on skills training, Riverside
spent $1,914 on basic education per referred case. Overall, Riverside’s LFA approach
incurred costs that were well under half of its costs under the HCD approach. As will be
seen, this difference in costs is ultimately attributable to differences under the two
approaches in participation rates in various services and average lengths of time pro-
gram participants received various services.
4.2.2 Unit Costs

Table 4.3 presents the unit costs of each service component of Riverside for both the
LFA and the HCD approach. Unit costs incurred by the welfare department represent
Table 4.3
UNIT COSTS OF RIVERSIDE’S LABOR FORCE ATTACHMENT AND

HUMAN CAPITAL DEVELOPMENT APPROACHES

Welfare Non-Welfare Welfare Non-Welfare
Dept. Dept. Dept. Dept.
Unit Cost Unit Cost Unit Cost Unit Cost
Per Participant a
Per Participant a
Activity Per Month Per ADA Per Month Per ADA
Orientation and appraisal b $ 79 $ — $ 79 $ —
Formal assessment b 535 — 535 —
Job search 682 — 682 —
Basic education 229 2,100 229 1,911
College 110 3,008 110 2,966
Vocational training 110 2,604 110 2,510
Work experience 514 — 514 —
NOTES: These data are for the subsample without a high school diploma or GED certificate. Estimated costs are in 1993 dollars.
a. One ADA refers to one unit of Average Daily Attendance, an attendance measure used by California community colleges and
adult schools, and is defined as a block of 525 hours of attendance.
b. Cost per session for one participant.
the average cost per month of serving one case. Non-welfare department costs are dis-
played as the cost per average daily attendance (ADA), an attendance measure used by
California community colleges and adult education schools. One unit of ADA is defined
as a block of 525 hours of attendance. Thus, under the LFA model, it costs $2,100 to pro-
vide 525 hours of basic education to one person. Table 4.3 indicates that unit costs are
very similar across the two approaches. This is due to the fact that both program
approaches were run by the same welfare department and service providers. As will be
seen next, the variation in cost is instead attributable to whether the program model
calls for referring individuals to high-cost or low-cost activities and how long individu-
als stay in various program activities once they begin participating.
4.2.3 Participation Rates

Table 4.4 shows participation rates in program activities under both approaches in
Riverside. The rates are based on all members of the JOBS research samples. The partici-
pation pattern reflects directly the philosophy behind the two approaches. Under the
LFA approach, 55 percent of all referred cases participated in at least one day of job
search. Under the HCD approach, in contrast, only 34 percent of the cases participated
in at least one day of job search. The basic education service component was also subject
to divergent participation patterns: 2 percent of the LFA program members participated
in basic education compared with 55 percent of the HCD program members. Much of
the difference in program costs can be attributed to such differences in participation
Table 4.4
PARTICIPATION RATES FOR EMPLOYMENT AND TRAINING ACTIVITIES

IN RIVERSIDE WITHIN TWO YEARS AFTER ORIENTATION
Of Those Referred to the Program,

Percentage Who Ever Participated

Activity Attachment (LFA) Development (HCD)
Job search 54.7% 33.6%
Basic education 2.1 54.8
Vocational training 0.0 4.3
Post-secondary education 0.0 0.0
Work experience 0.0 0.0
Ever participated in any activity 48.9 51.1
SOURCES: For LFA: unpublished data from MDRC research. For HCD: Gayle Hamilton et al., National Evaluation
of Welfare-to-Work Strategies, 1997.
NOTES: These data are for the subsample without a high school diploma or GED certificate. For the LFA
full-sample results, see Table D.1.
rates. As indicated by Table 4.3, for example, the unit costs for job search were $682 for
both the LFA and the HCD programs in Riverside. As reported in Table 4.2, however, the
gross cost of job search was $865 per LFA case, but only $559 per HCD case.
4.2.4 Length of Stay in a Program Component

Like differences in participation rates, differences in average length of stay in vari-
ous E&T activities also cause program costs to differ. Average length-of-stay figures for
Riverside’s LFA and HCD approaches are displayed in Table 4.5. These figures are aver-
aged over only those individuals who actually participated in each of the indicated
activities. For example, Table 4.5 indicates that persons who participated in job search
for at least one day averaged about 2.3 months under the LFA approach and 2.4 months
under the HCD approach.
Average lengths of stay in specific program components are driven by the program
approach (LFA versus HCD), as well as by the nature of the activity. For example, once
individuals begin basic education, they generally remain for several months. However,
on average, participants under the HCD approach stayed approximately one and a half
months longer than under the LFA approach.
Table 4.5
AVERAGE LENGTH OF STAY IN EMPLOYMENT AND TRAINING ACTIVITIES

IN RIVERSIDE WITHIN TWO YEARS AFTER ORIENTATION
Average Number of Months of Participation

Activity Attachment (LFA) Development (HCD)
Job search 2.3 2.4
SOURCES: Unpublished data from the National Evaluation of Welfare-to-Work Strategies based on MDRC
calculations from the two-year survey, adjusted using MDRC-collected case file data from the programs.
NOTES: These data are for the subsample without a high school diploma or GED certificate. For the LFA
full-sample results, see Table C.1.
4.2.5 Support Services

The cost estimates presented so far do not include costs for child care, transporta-
tion, or other ancillary services, all of which must be included in a comprehensive cost
analysis. Table 4.6 shows unit costs, the average length of service usage, costs per indi-
vidual who received support services, the percentage of cases that received services, and
gross costs per case. The figures are shown for both approaches and for all three sites.
Table 4.6
SUPPORT SERVICE COSTS OF WELFARE-TO-WORK–RELATED ACTIVITIES

WITHIN TWO YEARS AFTER ORIENTATION
Cost per Percentage

Referred of Referred Gross Cost
Average Average Case Who Cases Who per
Monthly Months of Received Received Referred
Site and Service Payment Payments Service Service Case
Atlanta
Labor Force Attachment
Child care $225 9 $2,254 31% $709
Transportation 38 3 126 53 67
Ancillary services 36 3 113 94 106
Total 882
Human Capital Development
Child care $247 9 $2,230 29% $648
Total 1,020
Grand Rapids
Labor Force Attachment $214 7 $1,415 19% $270
Child care n/a n/a n/a n/a 26
Transportation n/a n/a n/a n/a 1
Ancillary services 297
Total
Child care $218 7 $1,551 25% $383
Transportation n/a n/a n/a n/a 32
Ancillary services n/a n/a n/a n/a 2
Total 416
Riverside (full sample)a
Child care $143 3 $435 17% $ 73
Total 122
Riverside (subsample
without a high school
diploma or GED)
Child care $134 3 $341 17% $ 59
Total 103
Child care $143 5 $648 24% $157
Total 231
SOURCES: Gayle Hamilton et al., National Evaluation of Welfare-to-Work Strategies, 1997. For Riverside LFA (subsample without
a high school diploma or GED certificate): unpublished data from MDRC research.
NOTES: Estimated costs are in 1993 dollars.
Where data are not applicable, n/a is used.
Rounding may cause slight discrepancies in calculating sums.
a. For Riverside HCD, full-sample data are not available.
For purposes of discussion, the focus here is on child care costs under the LFA
approach. Table 4.6 indicates that unit costs for child care ranged from $134 to $247
across the three sites. Gross costs per case, however, had much more variation, ranging
from a low of $59 in Riverside (sample without high school diploma or GED certificate)
to $709 in Atlanta. Much of this disparity can be explained by differences in the length
of service receipt and in participation rates, very much the same factors that cause varia-
tion in the costs of program activities. For example, among those referred to the LFA
program who received any child care support, the average number of months of receipt
was nine in Atlanta but only three in Riverside. While this helps explain Atlanta’s high
and Riverside’s low child care costs, it is only part of the story. The percentage of cases
that received child care services affects gross costs as well. The percentage of cases that
received child care payments ranged from 17 percent in Riverside to 31 percent in
Atlanta.
The lesson here is that decisions on program design have a very direct influence on
support service costs. Localities can choose along a continuum that ranges from the
provision of very generous to very basic support services. Such decisions strongly affect
the gross costs of a program.
4.2.6 Site-Specific Factors

Various factors that are specific to a site running an E&T program can have major
effects on the costs of the program. These site-specific factors include the manner in
which the local agency operating the program implements its program model, welfare
grant levels at the site, the demographic characteristics of the population served by the
program at the site, and the characteristics of the local labor market.
a. Agency Discretion. Besides making decisions on the overall philosophy of a pro-
gram and basic service offerings, program agencies must decide on specific program
practices. These decisions affect program costs. For example, Riverside’s unit cost for job
search exceeded that at the other two sites under both the LFA and HCD approaches, in
part because, to a greater extent than the other two sites, Riverside used job developers
on staff, who canvassed the local job market for employment opportunities for partici-
pants. Similarly, Riverside’s unit cost for basic education was relatively high because it
made payments to basic education providers to supply the welfare department with
detailed attendance information on JOBS students, that they were not otherwise funded
to collect and report.
Included among the agency decisions that may affect program costs are staff man-
agement practices, procedures for monitoring and sanctioning clients’ participation
behavior, and guidelines for providing support services. For example, it is apparent that
providing extensive staff training, supervising case managers closely to ensure that they
counsel clients effectively, and focusing great effort on completing administrative tasks
all increase program costs, although there may be good reasons for all these efforts.
b. Client Characteristics. The population an E&T program serves affects the aver-
age length of stay in the program. For example, the welfare population can be catego-
rized into two groups: short-term recipients, who have received public assistance for less
than two years, and long-term recipients, who have been on the rolls for two years or
more. Individuals in the former group usually have acquired a more substantial work
history and, hence, even in the absence of E&T programs are more likely to have skills
that enable them to find jobs. Long-term welfare recipients, in contrast, often lack the
basic skills required in the labor market. Therefore, individuals who have a short welfare
history will, on average, be ready to leave an E&T program earlier than long-term recipi-
ents, who have higher hurdles to surmount. Obviously, a person who participates a
shorter time in a program needs fewer resources and incurs fewer costs than a person
who remains longer.
Table 4.7 presents the percentage distribution of cases covered by the JOBS mandate
at the three JOBS evaluation sites. While 78 percent of cases covered by the JOBS man-
date in Atlanta were long-term welfare recipients, only 54 percent in Riverside were. This
is probably one reason why Riverside had lower gross costs under both the LFA and
HCD approaches than Atlanta.
Table 4.7
WELFARE STATUS OF AFDC RECIPIENTS AT THE BEGINNING

OF THE PROGRAM
Welfare Status Atlanta Grand Rapids Riverside

On welfare less than two years 21.6% 36.5% 45.9%
On welfare two years or more (cumulatively) 78.4 63.5 54.1
Total 100 100 100
c. Welfare Grant Levels. The JOBS evaluation found that state welfare grant levels
affect E&T program impacts and costs. Specifically, in high-grant-level states that make
relatively generous welfare payments, the incentive for welfare recipients to seek work is
diminished. Thus, under otherwise similar circumstances, it can be assumed that welfare
recipients in low-grant states get off the welfare rolls and, hence, leave E&T programs
operated by welfare agencies sooner than recipients in high-grant states. This, in turn,
should lower program costs.
d. Labor Market Conditions. An E&T program that operates in a strong labor mar-
ket is likely to find it easier to place program participants into jobs than a similar pro-
gram in a weak labor market. In addition, a strong labor market might also contribute to
the faster advancement of new hires and to increases in job retention, thereby keeping
individuals out of E&T programs in the first place. By affecting the time in an E&T pro-
gram, such labor market factors may influence the program costs.
4.3 Estimating the Costs of an Ongoing

Program: An Illustration
In chapter 3, a seven-step procedure for estimating the gross costs of an ongoing pro-
gram was developed. This section uses the information provided in some of the tables
presented earlier in this chapter to demonstrate how those seven steps can be applied in
practice. More specifically, an illustration is provided here of how the seven steps can be
used to replicate gross cost estimates for the Riverside LFA program. For simplicity, non-
program costs are ignored in doing this. As indicated in Tables 4.2 and 4.6, respectively,
the gross costs of program activities were estimated to be $1,060 and the gross costs of
program support services were estimated to be $103 for Riverside’s LFA program. Thus,
including support services costs, total gross costs were $1,163. In Table 4.8, the seven
steps are used to derive this estimate.
By using the disk accompanying the guide, you can obtain estimates for your own
program using a format similar to that used in the table.
Table 4.8
ESTIMATING THE COSTS OF AN ONGOING PROGRAM

Two-Year Costs of Riverside’s Labor Force Attachment Approach
Step 3
Select the Analysis
Sample for Steps 4 and 5 a
Step 1 Step 2 Step 4 Step 5 Step 6 Step 7

List Unit Cost Participation Average Gross Cost per Total Gross
Program Estimates of Rate b Length of Referred Case Cost per
Components Program Stay c per Program Referred
Components Component Case
PROGRAM ACTIVITIES
Orientation and appraisal
Welfare department $ 79 1.30 d 1 session $ 103
Non-welfare department $ 0 1.30 1 session $ 0
Formal assessment
Welfare department $ 535 0.01 1 session $ 6
Job search
Welfare department $ 682 0.55 2.3 months $ 865
Non-welfare department $ 0 0.55 2.3 months $ 0
Basic education
Non-welfare department $ 528 e 0.02 5.4 months $ 60
College
Welfare department $ 110 0.00 0 months $ 0
Non-welfare department $ 500 f 0.00 0 months $ 0
Work experience
Non-welfare department $ 0 0.00 0 months $ 0
SUPPORT SERVICES
Child care
Transportation
Ancillary services
Total gross cost per referred case $ 1,163
SOURCE: Gayle Hamilton et al., National Evaluation of Welfare-to-Work Strategies, 1997. For support services: unpublished
data from MDRC research.
NOTES: These data are for the subsample without a high school diploma or GED certificate.
Rounding may cause slight discrepancies when calculating sums.
a. The data on participation rates (step 4) and average length of participation (step 5) were obtained from a sample of 193
cases that were assigned to Riverside’s LFA program.
b. The percentage of the analysis sample participating in each activity and receiving each service within two years of attend-
ing orientation.
c. Average months participating in each program activity and receiving each support service per participant within two
years of attending orientation. Estimates are based on only those who participated in the activity or received the service.
d. On average, 1.3 sessions were attended.
e. $4 per hour for 132.0 hours of basic education.
f. $5.73 per hour for 87.3 hours of college.
Chapter 5
Predicting Costs of Programs

That Have Not Yet Been
Implemented
This chapter outlines the procedures necessary to predict the annual costs of an employ-
ment and training program that has not yet been implemented. As discussed in chap-
ter 1, it is important to cost out programs prior to implementation for purposes of bud-
geting and resource planning and, if necessary, to modify various program provisions to
make sure that program costs do not exceed the available budget.
The chapter is divided into five major sections. Section 5.1 outlines the nine steps
required to predict the gross costs of a not-yet-implemented program during the first
year of program operations. Section 5.2 illustrates how these steps can be applied in
practice by demonstrating their use in costing out a simple hypothetical E&T program.
As will be seen, a number of the steps require values that are difficult to obtain in prac-
tice. Section 5.3 describes how these values might be obtained, emphasizing particularly
the possible use of information from cost analyses of previously implemented programs.
Section 5.4 discusses the reasons why such estimates are unlikely to hold for subsequent
years and provides some sense of how costs are likely to change over time. The chapter’s
final section depicts techniques that can be used to accomplish the task of obtaining
estimates of the net costs of a not-yet-implemented E&T program. Sections 5.1, 5.2, and
5.4 should be of interest to users of findings from cost analyses, while sections 5.3 and
5.5 are intended mainly for those responsible for costing out not-yet-implemented E&T
programs.
The accompanying disk includes a template for use in costing out not-yet-imple-
mented programs. The spreadsheet is modeled after the steps developed in this chapter.
In contrast to chapter 3, which described procedures for estimating the costs of
ongoing E&T programs, regardless of whether they are voluntary or mandatory, this
chapter focuses specifically on techniques for conducting cost analyses of mandatory
E&T programs operated by welfare agencies, so-called welfare-to-work programs. The
approach needed to conduct cost analyses of voluntary E&T programs would differ in
significant respects from conducting cost analyses of mandatory programs like the one
considered here.
49
5.1 Steps in Estimating the Gross Costs

of Not-Yet-Implemented Programs
This section provides a brief overview of the steps in conducting a cost analysis of an
ongoing program that are required for predicting the gross costs of a not-yet-imple-
mented E&T program during its first year of operation. The following two sections pro-
vide much greater detail about these steps, indicating how they can be used in practice.
The required calculations can be readily made by using the disk provided with the
guide. Although the steps presented here differ considerably from those required to esti-
mate the gross cost of an ongoing program, there are similarities. One important source
of the differences is that in conducting cost analyses of ongoing programs, the number
of people who participate in each program activity is known, but in cost analyses of not-
yet-implemented programs, flows of individuals from one program activity to another
must be predicted.
Step 1. Develop A Flow Diagram

Develop a detailed flow diagram of the proposed program — that is, a diagram that dis-
plays the expected flow of cases into each of the activities and support services to be
provided by the proposed program. (See Figure 4.1 or Figure 5.1 for an illustration.)
Step 2. Predict Total Program Referrals

Predict the total number of cases that will be referred to the program during the first 12
months after it is implemented. Program costs will be estimated for these cases and for
this time frame.
Step 3. Predict the Assignment Rates for Each Program Component

Of those cases referred to the program (see step 2), predict the fraction that will be
assigned during the program’s first year to each of the program activities that appear in
the flow diagram developed in step 1. In addition, predict the fraction of those referred
to the program who will receive each program support service during the first year of
program operations. These fractions must be predicted separately for each program
activity and support service. If some cases will be assigned to certain program activities
or support services more than once during the 12-month period, they should be count-
ed in the numerator of the assignment rate once for each time they are assigned.
However, cases that will not be assigned to a particular program activity until after the
first year of program operations should not be counted in the numerator. Note that it is
likely that all cases referred to the program will be assigned at least once to activities
related to the program enrollment process, such as orientation and assessment.
Chapter 5: Predicting Costs of Programs That Have Not Yet Been Implemented 51
e.g. If all the cases referred to the program will be assigned to job search at least once during
the program’s first year, and one-fifth will be assigned to job search a second time dur-
ing the year, then the job search assignment rate would equal 1.2.
Step 4. Predict the Participation Rate for Each Program Component

Of those cases that will be assigned to each of the program activities, predict the
fraction that will actually participate in the activity. For each program support service,
assign a value of 1.
Step 5. Predict the Average Length of Participation

Predict the average length of time that cases participating in each program activity will
remain in the activity (for example, average months of basic education and average
weeks of participation in job clubs). For cases that receive each support service, predict
the average length of time they will receive the service. For activities that do not really
have a time dimension (such as attending orientation or completing a formal assess-
ment), simply use a value of 1. Note that some cases referred to the program during its
first year will almost surely continue to participate in certain program activities (for
example, college) or continue to receive certain support services (for example, child
care) beyond the program’s first year. Nonetheless, the length-of-participation averages
should be based on length of stay during only the first year of program operations.
Step 6. Estimate Unit Costs

Using time units that are consistent with those used in step 5, estimate the unit cost of
each program component (for example, cost per orientation session, cost of serving one
person in basic education for one month, cost of serving one person for one week in a
job club, and cost of providing one month of child care). If the costs of providing a par-
ticular service component are shared by two organizations — say, a welfare agency and a
community college — it is important for budgeting purposes that the unit cost incurred
by each agency be computed separately.
e.g.
For instance, a welfare agency’s unit cost for vocational training would usually include
any costs incurred in requiring those assigned to vocational training to attend class; the
community college’s unit cost estimate would not incorporate those enforcement costs.
Step 7. Compute Costs per Referred Case for Each Program Activity
and Support Service
For each program activity and support service, the costs per referred case can be pre-
dicted by computing the product of the values obtained in the previous steps.
Specifically, make the following computation for each program activity and support
service:
step 3 x step 4 x step 5 x step 6
Note that by making these computations, costs are averaged across all the cases that
will be referred to the program, regardless of whether or not they will participate in the
program activity or receive the support service. In effect, in computing the average cost
for a given program component, those cases that will not participate in the component
are assigned a zero cost value for that component. Only by doing this is it appropriate to
sum the average cost values obtained in this step.
Step 8. Compute Total Costs per Referred Case

To obtain a prediction of total program costs per referred case, sum all the average cost
values obtained in step 7.
Step 9. Compute Aggregate Program Costs

To obtain a prediction of aggregate program costs, multiply the value obtained in step 2
by the value obtained in step 8.
The sequence actually followed in completing steps 2–6 can vary from the order in
which they are presented here. For example, in designing an E&T program, it may be
helpful to first determine the unit costs of various program activities (step 6) or at least
some of the determinants of unit cost, such as the number of seats in a particular voca-
tional training classroom. This information can then be used to make decisions about
assignment rates for each program component and the length of time participants will
be allowed to remain in various program activities. Steps 7–9, however, each depend on
information from the steps preceding them and, therefore, must be completed in the
order indicted.
5.2 Estimating the Costs of a Not-Yet-

Implemented Program: An Illustration
This section guides you through using the steps described above to predict the gross
cost of a mandatory not-yet-implemented welfare-to-work program for TANF recipients
during its first year of operation. The program used here is a very simple, hypothetical
one. It assumes, for example, that no one participates in any program activity more than
once during the program’s first year of operation and that there are no nonprogram
costs. Once you understand the steps, you will be able to apply them easily to more com-
plex program designs.
Although the values for unit cost and participation that are used in costing out this
simple hypothetical program are rooted in the JOBS evaluation described in chapter 4,
these values are not for any real program. Section 5.3 describes how the values needed to
cost out an actual program might be obtained.
For illustrative purposes, the hypothetical program is costed out twice. Example 1
(Tables 5.1–5.4) works through each of the steps required to obtain cost estimates of the
program as initially designed. In example 2 (Tables 5.5–5.8), a key element of the pro-
gram design is changed to show how that would affect program costs. Examples 1 and 2
can be easily replicated by using the spreadsheet provided on the accompanying disk.
Example 1. Costs of the Program as Initially Designed
Figure 5.1 presents a flow diagram for a hypothetical welfare-to-work program (step
1). In this simple program, all TANF recipients who are referred to the program are
required to go through an initial orientation and assessment process. Based on the
results, they are then assigned to one of the following four program activities: job club,
basic education, vocational training, or unpaid work experience. In addition to program
activities, it is assumed that the program provides child care, but no other type of sup-
port services.
Tables 5.1–5.4 provide a detailed breakdown of the calculations required to work
through steps 1–9, using hypothetical numbers for purposes of the example.
Example 2. Testing for Sensitivity
In costing out a not-yet-implemented E&T program, the computations required by
steps 7–9 can easily be performed using a spreadsheet. Such a spreadsheet is provided in
the disk that accompanies this guide. The major advantage of a spreadsheet is that it
allows you to readily change the values obtained from steps 2–6 and to see how doing so
changes the estimate of total program costs. This could be valuable for two reasons.
First, a welfare agency has considerable discretion over the number of cases that are
referred to an E&T program (step 2) and the various program activities to which these
cases are assigned and the support services they receive (step 3). It is useful to determine
how different decisions concerning these issues affect program costs. For example, if
your annual budget is less than the total cost obtained in step 9, you might want to see
how this cost estimate is affected by policy decisions that differ from the ones on which
the estimate was based. Second, as discussed in section 5.3, considerable uncertainty will
inevitably surround the values for participation rates (step 4), average length of time in
program activities (step 5), and unit costs (step 6) that are used in costing out a not-yet-
implemented program. Thus, rather than selecting a single value, it may be better to
determine a reasonable range, and then see how sensitive the cost estimates are to using
alternative values within this range.
To demonstrate the usefulness of testing the sensitivity of cost estimates to changes
in the values used to obtain the estimates, this section examines a simple example in
which everything is held constant except the program assignment rate. Tables 5.5–5.8
provide a detailed breakdown of the calculations required for the nine cost-estimating
steps, as done for example 1. This second example, however, shows how sensitive the cost
estimates for the hypothetical E&T program are to assigning 80 percent, rather than just
40 percent, of the cases referred to the program to job clubs and, as a result, assigning
fewer cases to other program activities. The sensitivity of total program costs to changes
in the assignment rates is demonstrated by comparing costs under these two alternative
Figure 5.1
FLOW DIAGRAM FOR A HYPOTHETICAL WELFARE-TO-WORK PROGRAM
Referred to the Program

by the Public Assistance Department
Orientation
and
Assessment
Basic Vocational Unpaid

Job Club Education Training Work
Experience
Employment
Table 5.1
ESTIMATING THE COSTS OF A NOT-YET-IMPLEMENTED PROGRAM

HYPOTHETICAL EXAMPLE 1
Steps 1-5
Step 1 Step 2 Step 3 Step 4 Step 5

List Number 3A 3B 4A 4B Average
Program of Cases Length of
Components Referred Number Assignment Number Participation Stay
to the of Cases Rate of Cases Rate
Program Assigned Participating
to Each in Each
Program Program
Component Component
PROGRAM ACTIVITIES
Orientation 10,000 10,000 1.00 10,000 1.00 1 session

Assessment 10,000 10,000 1.00 10,000 1.00 1 session
Job club 10,000 4,000 0.40 3,000 0.75 1 month
Basic education 10,000 2,000 0.20 1,500 0.75 10 months
Vocational training 10,000 1,000 0.10 800 0.80 10 months
Unpaid work experience 10,000 3,000 0.30 1,800 0.60 3 months
SUPPORT SERVICE
Child care 10,000 4,000 0.40 4,000 1.00 8 months
NOTES: Explanation of Step 3: Assignments to various program components are expressed as rates. Thus, the number of
cases assigned to a component is divided by the total number of cases referred to the program. The information is needed in
this form for later steps. For purposes of the illustration, assume that all 10,000 cases will be referred to the E&T program
orientation and assessment sessions. Thus, the assignment rate for each of these program activities is 1.0. It is also assumed
that after the orientation and assessment sessions, it will be decided that 4,000 of the 10,000 referrals will be channeled into
job clubs, 2,000 into basic education, 1,000 into vocational training, and 3,000 into work experience. Thus, it is assumed for
simplicity’s sake that no cases will be unassigned during the year. Finally, it is assumed that 4,000 cases will receive child
care.
Explanation of Step 4: The level of participation in various program activities is expressed as a rate. This is necessary for later
steps. Thus, the participation rate is determined by dividing the number of cases predicted to participate in each program
activity by the number of cases assigned to the activity. This example assumes that all 10,000 cases referred to the program
and assigned to orientation and assessment will actually participate in these activities. Thus, the participation rate for these
activities is 1.0. However, although 4,000 cases will be assigned to job clubs, it is assumed that only 3,000 will actually partic-
ipate in this activity. Similarly, not all the cases that will be assigned to basic education, vocational training, and work experi-
ence are assumed to participate. Those cases that will be assigned to a particular program activity but will not participate
might find employment prior to receiving the service, be out of compliance with program requirements, or have other rea-
sons for not participating.
Explanation of Step 5: This example assumes that it takes one month, on average, to complete a job club assignment; that
cases participating in basic education or vocational education during the program’s first year will remain in these activities
for 10 months, on average, during the year; and that those participating in unpaid work experience will remain in this activi-
ty 3 months, on average, during the year. It is further assumed that cases receiving child care during the program’s first year
will receive it for 8 months, on average.
Table 5.2

Step 6
Unit Cost
Welfare Department Cost Non-Welfare Department Cost
Personnel Overhead Subtotal Personnel Overhead Subtotal Total
PROGRAM ACTIVITIES
Orientation $ 60 $ 20 $ 80 $ 0 $ 0 $ 0 $ 80
Assessment $240 $ 80 $320 $ 0 $ 0 $ 0 $320
Job club $450 $150 $600 $ 0 $ 0 $ 0 $600
Basic education $112 $ 38 $150 $263 $ 87 $350 $500
Vocational training $ 97 $ 33 $130 $450 $150 $600 $730
Unpaid work experience $187 $ 63 $250 $ 0 $0 $ 0 $250
SUPPORT SERVICE
Child care $120 $ 40 $160 $ 0 $ 0 $ 0 $160
NOTE: Explanation of Step 6: All unit costs for the hypothetical program, except for orientation and assessment, are
expressed as costs per month per participant in the indicated activity. Unit costs for orientation and assessment are on a per-
session basis. The example assumes that 75 percent of subtotal cost is accounted for by staff personnel costs; 25 percent is
accounted for by overhead costs. It is also assumed that the job clubs are run entirely by the welfare agency and that child
care is paid for entirely by the welfare agency. Therefore, other agencies or institutions incur no costs. Non-welfare agency
costs are also assumed to be zero for unpaid work experience because these agencies and institutions receive labor free of
charge.
Table 5.3

Steps 7 and 8
Cost per Referred Case for Each Program Activity and Support Service
Step 7 Step 8
List Assignment Participation Average Unit Average Total
Program Rate Rate Length Cost Component Gross
Components of Stay Cost Cost per
(Step 3) x (Step 4) x (Step 5) x (Step 6) = Referred
Case
PROGRAM ACTIVITIES
Orientation
Welfare department 1.00 1.00 1 session $ 80 $ 80
Non-welfare department 1.00 1.00 1 session $ 0 $ 0
Assessment
Job club
Welfare department 0.40 0.75 1 month $ 600 $ 180
Non-welfare department 0.40 0.75 1 month $ 0 $ 0
Basic education
Welfare department 0.20 0.75 10 months $ 150 $ 225
Non-welfare department 0.20 0.75 10 months $ 350 $ 525
Vocational training
Unpaid work experience
SUPPORT SERVICE
Child care
TOTAL
Subtotal
Welfare department $1,556
Non-welfare department $1,005
Total gross cost per referred case $2,561

Table 5.4

Step 9
Aggregate Costs
Step 9
Total Number Aggregate
Cost of Cases Costs
Referred
to the
Program
(Step 8) x (Step 2) =
AGGREGATE WELFARE DEPARTMENT COSTS
$1,556 x 10,000 = $15.56 million
AGGREGATE NON-WELFARE DEPARTMENT COSTS
$1,005 x 10,000 = $10.05 million
AGGREGATE TOTAL PROGRAM COSTS
$2,561 x 10,000 = $25.61 million

Table 5.5

Steps 1-5
Step 1 Step 2 Step 3 Step 4 Step 5

List Number 3A 3B 4A 4B Average
Program of Cases Length of
Components Referred Number Assignment Number Participation Stay
to the of Cases Rate of Cases Rate
Program Assigned Participating
to Each in Each
Program Program
Component Component
PROGRAM ACTIVITIES
Orientation 10,000 10,000 1.00 10,000 1.00 1 session

Assessment 10,000 10,000 1.00 10,000 1.00 1 session
Job club 10,000 8,000 0.80 6,000 0.75 1 month
Basic education 10,000 500 0.05 375 0.75 10 months
Vocational training 10,000 500 0.05 400 0.80 10 months
Unpaid work experience 10,000 1,000 0.10 600 0.60 3 months
SUPPORT SERVICE
Child care 10,000 4,000 0.40 4,000 1.00 8 months

Table 5.6

Step 6
Unit Cost
Welfare Department Cost Non-Welfare Department Cost
Personnel Overhead Subtotal Personnel Overhead Subtotal Total
PROGRAM ACTIVITIES
Orientation $ 60 $ 20 $ 80 $ 0 $ 0 $ 0 $ 80
Assessment $240 $ 80 $320 $ 0 $ 0 $ 0 $320
Job club $450 $150 $600 $ 0 $ 0 $ 0 $600
Basic education $112 $ 38 $150 $263 $ 87 $350 $500
Vocational training $ 97 $ 33 $130 $450 $150 $600 $730
Unpaid work experience $187 $ 63 $250 $ 0 $ 0 $ 0 $250
SUPPORT SERVICE
Child care $120 $ 40 $160 $ 0 $ 0 $ 0 $160

Table 5.7

Steps 7 and 8
Cost per Referred Case for Each Program Activity and Support Service
Step 7 Step 8
List Assignment Participation Average Unit Average Total
Program Rate Rate Length Cost Component Gross
Components of Stay Cost Cost per
(Step 3) x (Step 4) x (Step 5) x (Step 6) = Referred
Case
PROGRAM ACTIVITIES
Orientation
Assessment
Welfare department 1.00 1.00 1 session $320 $ 320
Job club
Welfare department 0.80 0.75 1 month $600 $ 360
Non-welfare department 0.80 0.75 1 month $ 0 $ 0
Basic education
Welfare department 0.05 0.75 10 months $150 $ 56
Non-welfare department 0.05 0.75 10 months $350 $ 131
Vocational training
Non-welfare department 0.05 0.80 10 months $600 $ 240
Unpaid work experience
SUPPORT SERVICE
Child care
TOTAL
Subtotal
Welfare department $1,425
Non-welfare department $ 371
Total gross cost per referred case $1,796

Table 5.8

Step 9
Aggregate Costs
Step 9
Total Number Aggregate
Cost of Cases Costs
Referred
to the
Program
(Step 8) x (Step 2) =
AGGREGATE WELFARE DEPARTMENT COSTS
$1,425 x 10,000 = $14.25 million
AGGREGATE NON-WELFARE DEPARTMENT COSTS
$ 371 x 10,000 = $ 3.71 million
AGGREGATE TOTAL PROGRAM COSTS
$1,796 x 10,000 = $17.96 million

assignment rates. The steps that are affected by this change in assignment rates are
shown in Table 5.9. Examples 1 and 2 assume that 10,000 cases are assigned to the pro-
gram.
The change in the assignment rates for different program activities causes the pre-
diction of aggregate gross costs to fall from $25.6 million to $17.9 million. The reason
for this reduction is that the number of cases assigned to job clubs, which are relatively
inexpensive due to their short duration, was doubled, while far fewer cases were assigned
to more expensive program activities such as basic education, vocational training, and
unpaid work experience. The effect of this change in assignment rates is indicated in
Table 5.9.
Table 5.9
EFFECT OF A CHANGE IN ASSIGNMENT RATES ON THE PREDICTED COSTS

PER CASE REFERRED TO THE PROGRAM
Assignment Rate Costs per Referred Case

(Step 3) (Step 7)
Program Component Example 1 Example 2 Example 1 Example 2
Orientation 1.0 1.00 $ 80 $ 80

Assessment 1.0 1.00 320 320
Job club 0.4 0.80 180 360
Basic education 0.2 0.05 750 187
Vocational training 0.1 0.05 584 292
Unpaid work experience 0.3 0.10 135 45
Child care 0.4 0.40 512 512
Total Gross Cost — — $2,561 $1,796
5.3 Obtaining the Values Needed to Cost Out

an Employment and Training Program
This section discusses how the values required by steps 2–6 can be obtained in practice.
The values needed to cost out a not-yet-implemented program are predictions, so they
can never be entirely accurate. However, greater accuracy is possible with more time and
resources. The appendices at the end of this guide contain values from MDRC studies of
various welfare-to-work programs. The quickest way to cost out a not-yet-implemented
E&T program is simply to choose from among these values. (This section provides some
guidance for doing that.) However, because the values provided in Appendices B–F are
for programs that will differ from the one you are costing out, none is likely to fit your
needs perfectly. Consequently, these values should be adjusted in various ways that are
described in this section. Furthermore, it is usually better to obtain your own values,
ones that pertain directly to the program you are costing out. The extent to which you
are able to adjust values obtained from other E&T programs or obtain values for your
own program, however, will depend on the time and resources that you have available.
5.3.1 Predicting the Total Number of Cases Referred (Step 2)

The number of cases referred to an E&T program during a year is mainly a policy
decision. This decision, however, obviously has major budgetary implications, as
decreasing the number of cases referred is a direct way of decreasing program costs.
In practice, the number of cases referred to a mandatory E&T program will be
strongly influenced by the program’s eligibility criteria. Perhaps there is a minimum age
requirement for all the children in the case, or a specified number of months on the wel-
fare rolls. Thus, to predict the number of cases that will be assigned to a program during
the year after its adoption, it is useful to determine the proportion of cases currently on
the rolls that meet the program’s eligibility criteria. If the program has eligibility criteria
that are tied to the length of time on the welfare rolls, it is also useful to determine the
proportion of cases that will become eligible during the next 12 months. In making
the latter determination, it is necessary to account for the rate at which cases leave the
welfare rolls. (See the discussion in section 5.3.2, below.)
Determining the proportion of cases that meet program eligibility criteria is most
readily accomplished if there is an automated data file on the current caseload that
contains the information required for making a determination for each case. If such a
data file does not exist, a determination can be made from the hard-copy case files of
a randomly selected sample of current cases.
5.3.2 Predicting Assignment Rates (Step 3)

Appendix B provides information on assignment rates at three JOBS program sites.
However, these assignment rates are likely to differ greatly from those for your program.
Like the number of cases referred to a mandatory E&T program, assignment rates for
various program activities and support services are strongly influenced by policy deci-
sions concerning the program design and rules. For example, cases may be assigned to a
prescribed sequence of activities, which ends either after a job is obtained or a time limit
is reached. Similarly, whether a case receives certain support services will be determined
by program rules and will depend on the particular activity in which the person is par-
ticipating and certain case characteristics; the receipt of child care, for example, may
depend on the number and ages of children in the case. Depending in part on available
financing, there may be periods when a person who is referred to the program will not
be assigned to any activity. Depending on program rules, this may also occur when a
person is faced with certain circumstances — for example, an illness or a transportation
problem.
Because assignment rates are computed by dividing the number of cases assigned to
an activity by the total number of cases referred to the program, they are affected not
only by policy decisions reflected in the program design and rules, but also by the num-
ber of referred cases available for assignment. The number of cases available for assign-
ment must first be predicted, therefore, to determine assignment rates. Decisions con-
cerning the program rules and design should help guide the effort to predict assignment
rates.
e.g.
The program rules may stipulate that only cases with particular characteristics qualify to
participate in certain program activities or receive certain support services. If so, infor-
mation from welfare agency automated data files or a sample of hard-copy case files
should be used to determine the fraction of referred cases that will meet these criteria.
Similarly, the program design may specify that math and literacy tests or other such
devices will be used to determine the program activities to which individuals will be
assigned. Individuals with low scores may be assigned to basic education and those with
high scores to job search. In such instances, if sufficient time is available, information
useful in predicting assignment rates can be obtained by administering the tests to a
small representative sample of program-eligible individuals at the time the effort to cost
out the program is made.
The difficulty in predicting the number of cases available for assignment to a partic-
ular program activity depends largely on where in the program sequence assignment to
the activity occurs. Presumably, virtually all the cases referred to the program will be
available for assignment to enrollment activities such as orientation and assessment.
Some cases, however, will fail to complete the enrollment process (they may, for exam-
ple, find a job first or be out of compliance with the program rules) and will be unavail-
able for assignment to the second round of activities. Other cases will complete the
enrollment process, but leave the welfare rolls before the assignment to the second round
of activities is made. Still other cases will receive deferrals from assignment because of
part-time employment, health and transportation problems, lack of child care arrange-
ments, and so forth. If the assignment occurs fairly soon after the enrollment process is
completed, however, this “leakage” from the program will probably be fairly small, and
predictions of it are likely to be relatively accurate. It is in predicting the number of cases
available for assignment to the third and subsequent rounds of program activities that
substantial errors are likely to occur. For by then, program leakage will likely be substan-
tial, but its size is difficult to forecast accurately.
e.g.
The importance of leakage is illustrated in Figures 5.2 and 5.3 for 100 typical TANF
cases that were assigned to the JOBS labor force attachment models in Riverside and
Grand Rapids and attended the program orientation. (See chapter 4 and Appendix A for
a description of this model.) As can be seen, while specific flow patterns differed
between the two sites, leakage was high in both sites. Thus, the number of cases available
for assignment diminished rather rapidly as cases moved through the sequence of pro-
gram activities. Consequently, while a large fraction of the initial 100 cases in both sites
were assigned to job search — the first activity after orientation — only 17 cases in
Grand Rapids and only 14 cases in Riverside were ever assigned to any other activities.
Moreover, many of those cases that were assigned to job search did not actually partici-
pate. This last point is discussed in greater detail in section 5.3.3.
Figure 5.2
THE FLOW THROUGH RIVERSIDE’S LABOR FORCE ATTACHMENT (LFA) PROGRAM APPROACH
FOR 100 TYPICAL LFA SAMPLE MEMBERS WITHIN TWO YEARS AFTER ORIENTATION
100 28
Attended an Never assigned to an
orientation activity after orientation
72
Assigned to an activity
after orientation
12 6 4 4
Deferred for a Employed No longer Other (these
4 68 non-employ- (deferred or mandatory do not add to
Assigned to a non- Assigned to a job ment reason no longer within one the total
job search activity search activity within one mandatory) month after because of
month after within one orientation rounding)
orientation month after (for a reason
orientation other than
28 40 employment)
Did not participate in the initially Participated in job search
assigned job search activity as an initial activity
6 18 16
Completed job search Dropped out of job Dropped out of job search because of employ-
without a job search without a job ment or completed job search with a job
5 1 9 1 1 1 6
Assigned to Deferred for a Assigned to Left AFDC Sanctioned or Employed
another non-employ- another in the month referred to part-time
activity ment reason activity with- of dropout sanction within 30 Other
in 30 days or in the within 30 days after
after drop- following days after dropping out
ping out month dropping out
66
SOURCE: Unpublished data from the National Evaluation of Welfare-to-Work Strategies.

Figure 5.3
THE FLOW THROUGH GRAND RAPIDS’ LABOR FORCE ATTACHMENT (LFA) PROGRAM APPROACH
FOR 100 TYPICAL LFA SAMPLE MEMBERS WITHIN TWO YEARS AFTER ORIENTATION
100 1
Attended an Never assigned to an activity
orientation after orientation
99
Assigned to an activity
after orientation
13 86
Assigned to a non-job Assigned to a job
search activity search activity
40 46
Did not participate in the initially Participated in job search
assigned job search activity as an initial activity
16 15 15
Dropped out of job search Completed job search Dropped out of job search
because of employment or com- without a job without a job
pleted job search with a job
Chapter 5: Predicting Costs of Programs That Have Not Yet Been Implemented
14 1 3 2
Assigned to another Deferred for a non- Assigned to another Sanctioned or referred to 5
activity employment reason activity within 30 days sanction within 30 days Other
after dropping out after dropping out
3 2 0
No longer mandatory Left AFDC in the Employed part-time
within 30 days after month of dropout or within 30 days after
dropping out in the following dropping out
month
67
SOURCE: Unpublished data from the National Evaluation of Welfare-to-Work Strategies.

In predicting the number of cases that will be available for assignment to a par-
ticular activity, two prior predictions must be made. The first is the length of time after
being referred to the program that it will take a typical case that is still active in the pro-
gram to reach the point at which the assignment might occur. The second prediction is
the number of cases that will still be in the program and not be deferred by that point.
e.g.
USING TWO PREDICTIONS TO FORECAST ASSIGNMENT RATES:
AN ILLUSTRATION
An examination of the assignment rate for basic education in a hypothetical not-yet-

implemented E&T program can illustrate how predictions about the length of time to
reach assignment and the number of cases that will still be in the program by that time
might be used in forecasting assignment rates. Assume that under this program, cases
will be required to complete the enrollment process and two months of job search before
becoming eligible for basic education. Further assume that, as a result, four months will
pass after a typical case is referred to the program before it will be ready for assignment
to basic education. Assume that by then, however, only 80 percent of the original group
of referred cases will still be in the program (15 percent will voluntarily leave the welfare
rolls and 5 percent will fail to comply with program rules and, consequently, be sanc-
tioned and no longer active in the program). In addition, another 15 percent of the cases
will be deferred. That leaves only 65 percent of the originally referred cases available for
assignment. Assume that of those available cases, the program managers decide to assign
20 percent to basic education and the remainder to post-secondary education, vocation-
al training, work experience, or a job club. Thus, the assignment rate for basic education
will be .65 x .20 = .13.
a. Length of Time to Reach Assignment. The length of time it will take a typical
person to reach the assignment point depends on the time she or he spent in previous
program activities and the time she or he previously spent without an assignment.
Approaches to predicting time that will be spent in previous program components are
described later in this section. Predictions of time that will be spent without an assign-
ment by a typical program enrollee will have to be based on judgment. However, consid-
erable relevant information is likely to be available. For example, the extent to which
cases are not assigned will be strongly influenced by the resources available for program
activities. It will also be influenced by the extent to which program rules allow tempo-
rary deferrals to be granted. Information concerning these factors should be available
for the agency considering a new program. Moreover, MDRC evaluation reports, some
of which are listed in the references at the end of this guide, provide considerable infor-
mation on the frequency with which such deferrals have been granted in previous
mandatory E&T programs.
b. Predicting the Number of Cases. To determine the number of cases that will still
be active in a program at various assignment points after they are initially referred, a
welfare agency can use its own case file records. The agency would select a sample of
cases from some recent point in time — say, one or two years back. The characteristics
of the selected cases should be as similar as possible to the cases that would be referred
to the proposed program. For example, only those cases that meet program eligibility
criteria should be selected. The agency can then determine the fraction of cases that
were still on the rolls after one month, after two months, after three months, and so
forth. These figures should be adjusted downward to account for the likely effects of the
not-yet-implemented program itself. For example, as a result of the program, some
enrollees who would otherwise remain on the rolls will obtain employment and leave
the rolls. Other enrollees will refuse to comply with program rules and, hence, will never
make it to the assignment point (although they may be sanctioned). MDRC evaluation
reports provide considerable information useful in determining the likely size of each of
these two groups. Appendix Table F.1 indicates the effect of the JOBS program on AFDC
receipt.
5.3.3 Predicting Participation Rates (Step 4)

As illustrated in Figures 5.2 and 5.3, not all those assigned to a program activity will
actually participate. Some might leave the welfare rolls before the activity begins, while
others will not show up and, consequently, be out of compliance with the program. As
discussed earlier,“leakage” of this sort also affects assignment rates.
It is difficult to predict accurately the rate of participation among those assigned
to any specific program activity; one must rely largely on educated guesses. However,
these guesses can be made more accurate by paying special attention to two factors.
First, participation rates will decline as the interval between assignment to a program
activity and the start of the activity increases. Some idea of how the length of this inter-
val is likely to affect participation rates can be obtained by examining the rate at which
cohorts of cases leave the welfare rolls. (See the discussion in section 5.3.2.)
Second, some program activities (for example, vocational training) are likely to
be relatively more attractive to those assigned to them than others (for example, unpaid
work experience). Thus, frequency of noncompliance may be greater among those
assigned to the latter than to the former activities. However, noncompliance may, per-
haps, be reduced if sanctions are vigorously enforced. Some idea of the frequency of
noncompliance among cases assigned to particular program components can be
obtained by examining previous experiences at both the sites considering the proposed
program and elsewhere. (Again, see the discussion in section 5.3.2.)
5.3.4 Predicting the Average Length of Participation (Step 5)

The average time spent participating in a particular program activity or receiving a
particular program support service will depend upon the characteristics of those partic-
ipating, the local economy, and the rules of the program.
e.g.
Project Independence, a welfare-to-work program in Florida that was evaluated by
MDRC several years ago, required two weeks of independent job search for cases
referred to the program that were deemed job-ready. However, this two-week period
should be viewed as a maximum, as some of those who began the job search left the
program earlier because they found a job or for other reasons. This was especially likely
because cases were assigned job search only if they were considered job-ready. It was
also likely because unemployment rates in Florida were quite low at the time the evalua-
tion was conducted.
In predicting the average time spent in a particular program activity or receiving

a service, it is useful to examine information on previous experience with a similar
program component. Doing this is especially likely to be helpful if the information
pertains to programs currently or previously run at the sites considering the not-yet-
implemented program, because it will help control for the effects of participants’ charac-
teristics and the condition of the local economy on time spent in various program
components. Even then, however, adjustments will have to be made for differences
between the old and new programs in program rules, economic conditions, and the
characteristics of those likely to participate in the program component. Participants’
characteristics are especially likely to differ if the not-yet-implemented program will
offer a particular component at a different point in the program sequence than was
previously the case. The necessary adjustments will have to be made on the basis of
informed judgments about these differences, as there is unlikely to be a firmer basis
on which to make them.
If a similar program activity or support service has not been previously offered at
the program sites considering the not-yet-implemented program, or if information does
not exist on average time spent in the program component, then the prediction of the
average length of participation can be based on the experiences of other sites that have
offered the component. Average time estimates for three JOBS program sites are present-
ed in Appendix C, which also includes estimates from MDRC evaluations for ongoing
programs in a number of additional sites. In this instance, however, the adjustments may
have to be quite substantial. Economic conditions, the characteristics of participants,
and program rules are all likely to differ considerably between the sites considering a
not-yet-implemented program and sites included in the MDRC studies. Obtaining
information on all these differences will be helpful in forming the judgments necessary
to make the required adjustments. Such information is available in MDRC’s evaluation
reports.
Estimates of average length of participation derived from existing programs
may also need to be scaled back to account for the fact that the costs of the not-yet-
implemented program are being predicted for a 12-month period, while cost analyses
of ongoing programs often measure length of participation over a longer time period.
This is especially likely in the case of lengthy program activities, such as vocational
education and college, and certain support services, such as child care.
5.3.5 Estimating Unit Costs (Step 6)

An estimated unit cost of a particular activity in a not-yet-implemented E&T
program can be obtained in three different ways:
1. It can be based on the unit costs of a similar activity in an ongoing program at
the sites considering implementing the proposed program.
2. Less ideally, it can be based on the unit cost of a similar activity in ongoing
programs at other sites.
3. It can be estimated by a program unit cost approach. Under this approach, the
cost of providing a group of individuals with the activity is first estimated (for
example, the cost of a month-long basic education class). Then, the estimated
cost is divided by the size of the group (for example, the number of persons in
the class) to obtain a measure of unit costs (in the case of our example, costs
per person per month).
Unit costs that are obtained under the first and third of these approaches rely on
data from the sites considering the proposed program. Thus, once unit costs are predict-
ed using these approaches, it will usually also be possible to allocate unit costs among
the components — that is, the percentage accounted for by salaries paid to classroom
teachers, by support staff salaries, by rent for classrooms, and so forth. Consequently, as
discussed in greater detail in chapter 6, unit cost estimates obtained under the first and
third approaches can be used to develop an operational budget for the program. Unit
cost estimates that are based on data from other sites (the second approach) typically
cannot be used to do this because insufficient information is available to allocate unit
costs among the components.
Because the first two of the approaches listed above rely on unit cost estimates that
already exist, they often require somewhat less effort than the third approach. However,
appropriate existing estimates are not always available for all the activities proposed as
part of a not-yet-implemented program because the program may differ substantially in
some respects from previous E&T programs. Under such circumstances, the third
approach can be used instead. As discussed next, however, even if appropriate existing
unit cost estimates are available, they still must usually be adjusted in various ways
before they can be adopted. Although these adjustments are typically greater when the
estimates are from sites other than those considering the proposed program, the adjust-
ments are sufficiently similar that the first two approaches can be discussed together.
The third approach is quite different from the first two and, therefore, is discussed sepa-
rately.
a. Using Existing Estimates of Unit Costs. Chapter 3 provides details on how unit
costs can be computed for ongoing programs. Chapter 4 presents unit cost estimates for
the Riverside JOBS program site, and Appendix E presents unit cost estimates for a
number of additional ongoing E&T programs that have been evaluated by MDRC. In
examining these estimates, it is apparent that unit costs for the same activity vary con-
siderably among E&T programs and even among different sites running the same pro-
gram. The tables presented in Appendix E indicate specific reasons why unit costs in
some of the programs and sites are especially high or low.
SOME SOURCES OF VARIATION IN UNIT COSTS
• differences in salary and fringe benefit levels;

• differences in staff/client ratios;
• if case management costs are included in unit costs, as they often are, differences in case
management/client ratios;
• whether or not incentive bonuses are paid to staff or to contractors operating the activity;
• differences in how classes are run — for example, a fixed cycle, in which enrollees must begin
at the start of the cycle, or open enrollment, in which someone else can fill the open slot that
results when a participant drops out.
In using unit cost estimates from an ongoing program to predict unit costs for
a not-yet-implemented program, two important adjustments should be made. The
first adjustment is needed to take account of differences in salary and fringe benefit
levels between those staffing the ongoing program and those who will staff the not-yet-
implemented program. This adjustment is likely to be minor if the new program will
operate at the same sites as the former program. Still, adjustments should be made if
staff salary and fringe benefit levels have changed appreciably over time or if the salaries
and benefits of personnel for the not-yet-implemented program will be higher or lower
than those in the ongoing program.
The adjustment may be considerably more important if unit cost information for an
ongoing program is obtained from different sites than the ones in which the not-yet-
implemented program will operate because staff salary and fringe benefit levels are
likely to vary considerably between the two sets of sites. In this instance, it is helpful to
obtain data on staff salary and fringe benefit levels for both sets of sites whenever feasi-
ble. Then, unit cost estimates that were obtained from the ongoing program can be
adjusted by using the ratio of staff salaries and fringe benefits in the new sites to staff
salaries and fringe benefits in the ongoing program sites. In making this adjustment,
however, a distinction must be made between personnel costs (including those engen-
dered by support staff) and the costs of physical resources. Differences in salaries and
fringe benefits affect personnel costs, but not the costs of physical resources.
e.g.
Consider a situation in which salaries and fringe benefits are 25 percent higher in sites
considering a not-yet-implemented program than in the sites from which unit cost esti-
mates were obtained. If personnel costs account for 80 percent of the total cost at the
sites considering the new program and physical resources account for the remaining 20
percent, the unit cost estimates from the sites with the ongoing program can be adjusted
by multiplying them by 1.2 [(1.25 x .8) + .2 = 1.2].
The second adjustment is required to account for differences in staff/client ratios

between the ongoing program and the not-yet-implemented program.
e.g.
Consider a not-yet-implemented program in which the numbers of persons attending
orientation sessions are expected to be twice as large as in the ongoing program, and
there are no other differences between the two programs — the size of the staff leading
the orientation sessions remains fixed, the same room is used for the sessions, and no
additional equipment is needed. Under such circumstances, the unit cost estimate for
orientation that was obtained from the ongoing program should be reduced by half
before it is used to predict the costs of the not-yet-implemented program.
The adjustment is more complex if some costs remain fixed when class size
increases, while other costs rise.
e.g. Assume that the size of a vocational training class is doubled, but that the number of
instructors remains fixed. However, all other costs (for example, classroom rental and
equipment costs) are doubled (that is, they increase in proportion to the increase in class
size). Further assume that instructor costs account for 40 percent of the total cost of
vocational training in the ongoing program and, hence, other costs account for the
remaining 60 percent. In this case, a unit cost estimate for vocational training from the
ongoing program can be adjusted by multiplying it by .8 [(.5 x .4) + .6 = .8].
One can readily imagine even more complicated situations — some costs might
remain fixed, some might increase by 50 percent, and still others might double. In such
cases, it can be very difficult to determine the size of the adjustment factor. Under these
circumstances, it may be better to forecast unit costs for the not-yet-implemented pro-
gram directly, rather than attempting to adjust an estimate from an ongoing program.
This possibility is addressed next.
b. Program Unit Approach. Unit costs measure expenditures incurred in providing
an E&T service to one individual over a specified time period. However, individuals
often participate in E&T program activities as part of a group — for example, in a class-
room setting. When this occurs, unit costs are usually most readily predicted by first
estimating the cost of serving the entire group and then dividing group costs by the
number of persons in the group.
e.g. Predicting the unit costs for a job club, for example, might be done as follows:
1. Determine the duration of the job club and the number of hours it would meet each
day — say, four hours a day over a three-week period.
2. Determine the average number of persons who would be served by the job club
during the three weeks it would operate. One might anticipate, for example, that the
job club would start with 10 people, but that two members would find jobs by the
beginning of the second week and that two more would find jobs by the beginning
of the third and final week. Unless new club members replace those who leave, an
average of 8 people would participate in the job club over the three-week period,
although it has the capacity to serve 10.
3. Determine the resources required for running one job club for three weeks — the
number of chairs, the number of instructors/coaches, additional case manager time,
additional support staff time, the number of phones, and so forth.
4. Determine the cost of each of the required resources — for example, the salary and
fringe benefits paid per hour to each of the required personnel, the costs for space
and equipment (such as telephones), the cost of necessary supplies, and so forth.
5. Determine the total cost of running one job club by summing all the resource costs
obtained from step 4. (As previously discussed, the cost of resources that are available
for use by the job club and would otherwise be unused, such as an empty classroom,
should not be included in this sum.)1
1. As explained in chapter 2, using a resource that would otherwise be unused for an E&T program is essentially
cost-free to the organization with the resource. However, if a community-based organization were operating job
clubs under subcontract to a welfare agency, it would usually incorporate the costs of otherwise unused resources
into its charges to the agency.
6. Determine unit costs by dividing the sum obtained from step 5 by the average
number of persons served, as determined by step 2. It is important that the average
size of the group, rather than its maximum size, be used. The average size allows for
the fact that the unit cost of a class that has excess capacity is larger than the unit cost
of a class that is always filled to capacity. It is also necessary to use the average group
size in order to maintain consistency with the prediction of the average length of
participation.2
An important byproduct of using these six steps to predict unit costs is that they
provide information that is useful in planning for a proposed E&T program. Once class
size and the average duration of a course are determined, for example, these values can
be combined with predictions of the number of persons who will participate in the
activity during a year (as determined by steps 2–4 in estimating the gross costs of not-
yet-implemented programs) to estimate the number of classes needed during the year
and the number that will have to be run at any point in time. If outside organizations are
to be solicited to deliver the service, the information on class capacity, number of class-
es, cost per class, and so forth can be used to specify a Request for Proposals.
e.g.
If about 1,000 persons were expected to be enrolled in job clubs during a year and each
job club began with 10 members, about 100 job clubs would have to be run over the
year. Thus, two new job clubs would have to be initiated each week. Moreover, if a job
club lasted three weeks, six clubs would be running at any point in time. By running
three classes in the morning and three in the afternoon, classroom requirements could
be held to three.
5.4 Changes in Program Costs Over Time

As indicated earlier, the procedures described here for predicting the cost of a not-yet-
implemented mandatory E&T program are directed at costing out the program during
its first year after full implementation. Once past the first year, however, program costs
could change dramatically: the program design might be changed, perhaps because
experience with operating the program suggests improvements that can be made; unit
costs could increase or decrease over time — for example, as a result of changes in prices
or in the client mix in the E&T program; or the economy might change, altering both
2. For example, in our illustration, six cases remain in the job club for all three weeks, two cases participate for two
weeks, and two cases participate for only one week. Thus, the average length of participation is 2.4 weeks (or
[(6 x 3) + (2 x 2) + (2 x 1)]/10) and the average number of cases over the three weeks is eight. Therefore, if the total
cost of operating the job club for a week is, say, $1,000, then unit cost (cost per participant per week) is $125
(or $1,000/8) and the cost incurred by one typical job club participant while the job club is in session is $300
(or 2.4 x $125). Now consider an alternative possibility: all job club participants remain in the job club for the full
three weeks, but the total costs of operating the club for one week remain at $1,000. Under these circumstances, unit
costs would fall to $100 (or $1,000/10), but the cost incurred by one typical participant would still be $300
(or 3 x $100).
the number and characteristics of the persons entering the program. These changes and
their effects on costs can rarely be predicted with any reliability.
Even if the program design and the economy remain stable, however, program costs
will still change over time, probably diminishing. Moreover, these changes may be quite
sizable. One reason program costs may fall over time is that the number of cases referred
to a mandatory E&T program will probably decline. A second reason is that the propor-
tion of cases in the program that are assigned to relatively expensive activities is likely to
decline and the proportion assigned to less expensive activities is likely to increase.
e.g.
To see this, consider a new E&T program. Although our illustration is grounded in the
provisions of TANF, several simplifying assumptions about the new program are made
initially to keep the example as clear as possible. First, let us assume that all cases that
have been on the TANF rolls for at least two years are required to participate in the
new program, but that no cases that have been on the rolls less than two years are
allowed to participate. Let us further assume that the welfare agency that will operate
the new program began running the clock on a five-year lifetime time limit for all
TANF recipients one year prior to beginning the new program. As a result, TANF
recipients will begin to become subject to the time limit four years after the E&T pro-
gram is put into place and, as they do, they will begin to leave the program. Finally, let
us assume the welfare agency that is implementing the new program is not currently
operating an E&T program. (Some of these assumptions will be modified later.)
Given the assumptions just made, all TANF cases that have been on the rolls for over
two years — most of the caseload — will be referred to the E&T program during its
first year. Now, consider the program’s second year. On the one hand, while some of
the originally enrolled cases will remain in the program, others will leave (for example,
some will find jobs). On the other hand, still other cases will begin their third year on
the TANF rolls and, hence, be newly required to enter the E&T program. However, if
the program is successful, it will increase the number of cases leaving the TANF rolls.
Consequently, more cases will leave the E&T program than enter, and the total number
of cases receiving E&T services will fall. This reduction will continue for several years,
causing program costs to fall over time.
Now consider the end of the program’s fourth year. At that time, most of the pool of
cases that were referred to the program in its first year and who remain on the TANF
rolls will become subject to the five-year time limit. Hence, they will lose their eligibili-
ty for TANF and leave the program. At the end of the program’s sixth year, only cases
that are in their third, fourth, or fifth year on TANF will be in the program; longer-
term welfare cases will be gone. This will cause a reduction, probably a substantial one,
in the number of cases in the program, again causing program costs to fall.
Assignment rates to different program activities will also change over time. Cases
that remain in the program that were assigned to comparatively expensive activities such
as basic education and vocational training in the program’s first year will eventually
complete these activities. Upon doing so, they are likely to be reassigned to relatively less
expensive activities such as work experience. These cases will be replaced in the more
expensive activities by cases that newly enter the E&T program, but only partially.
e.g.
To see this, imagine that during the program’s first year, all those cases in the E&T pro-
gram that would appear to benefit from basic education and vocational training are
assigned to these activities, but at the end of the first year, these cases are all reassigned
to work experience. During the program’s second year, additional cases will be assigned
to basic education and vocational training. However, these cases will be drawn exclusive-
ly from among those cases that are in their third year on the TANF rolls and, hence,
newly referred to the E&T program. Older cases will have been previously assigned to
basic education and vocational training. Thus, during its first year, the program will have
a much larger pool of cases from which to draw for assignment to basic education and
vocational training than during its second year. Hence, the proportion of cases in basic
education and vocational training will fall between the program’s first and second year,
while the proportion in work experience will grow. This change, in turn, will decrease
program costs.
All the factors discussed above will cause the costs of a newly adopted E&T pro-
gram to diminish over time. However, if the new program is implemented gradually,
costs might increase until the implementation is complete. Thus, gradual implementa-
tion could cause program costs to decline more slowly over time, or perhaps not at all.
5.4.1 An Assessment of the Issue

To acquire some sense of the importance of the factors that cause E&T costs to
diminish over time, we used a well-established mathematical tool, Markov Chain
Analysis, to develop a model.3 This mathematical model allows us to examine the extent
to which the number of cases in the E&T program and the number of cases assigned to
various program activities will change over time under different sets of circumstances.
In utilizing the Markov model for this purpose, let us assume that a welfare agency
that does not currently have an E&T program is introducing one for cases on TANF that
have accumulated more than two years of benefits. It is assumed that, at the same time,
the welfare agency also adopts the five-year lifetime time limit required by the Personal
Responsibility and Work Opportunity Reconciliation Act of 1996. Consistent with the
provisions of the Act, let us also assume that 20 percent of the cases receiving TANF will
be exempt from the time limits. In addition, assume that the agency fully implements the
E&T program immediately upon introducing it, rather than gradually implementing it.
Finally, assume that prior to the introduction of the program and the time limit the
agency’s TANF caseload has been stable. That is, each case that left the TANF rolls was
replaced by a new case and, hence, the size of the caseload was not changing over time.
As discussed earlier, after the program and time limits are introduced, the caseload will
fall for several years and, consequently, the number of cases in the E&T program will
also fall.
3. Briefly, Markov Chain Analysis permits one to determine movements of individuals off and on the TANF rolls
and the distribution of recipients by number of years on the rolls under steady-state conditions. A Markov chain is a
dependent stochastic process. The key assumption is that the probability of an individual moving onto the TANF
rolls is not influenced by how long she has been off the rolls, and the probability of an individual moving off the
TANF rolls is not influenced by how long she has been on the rolls. Future manifestations of a Markov chain are
treated as completely determined by the present state of the system — that is, as independent of the past.
Using the Markov model requires that an assumption be made about the rate at
which cases were permanently leaving the TANF rolls prior to the introduction of the
new E&T program and the five-year time limit. The rate that is most appropriate to use
for this purpose is not apparent. Prior to passage of the 1996 welfare law, the rate at
which AFDC cases permanently left the rolls appears to have varied considerably among
welfare agencies and over time. For example, MDRC evaluations imply that, depending
on the specific site and time period studied, between 0.5 and 1.5 percent of AFDC cases
were permanently leaving the rolls each month prior to the Personal Responsibility and
Work Opportunity Reconciliation Act.4
As a starting point for this analysis, a rate is used that falls near the top of this range,
1.2 percent. This particular rate was selected because when it is used, the Markov model
implies that prior to introducing an E&T program and time limits, 74 percent of a wel-
fare agency’s caseload would have been on the rolls for two or more years and 46 percent
for five or more years. This appears reasonably consistent with the empirical evidence.
For example, estimates that were made by Sheila Zedlewski and Isabel Sawhill just prior
to passage of the 1996 law indicate that 70 percent of the single-parent AFDC caseload
had been on the rolls for over two years and 40 percent for over five years.5
Based on the 1.2 percent assumption, the findings indicate that the five-year time
limit will have a dramatic effect on the number of cases in the E&T program after it
takes effect. Specifically, for every 100 cases enrolled in the program at the time it is initi-
ated, the Markov model predicts that only 56 cases will be enrolled once the caseload
has fully adjusted to the time limit. The number of cases declines even further if the
monthly rate at which cases permanently leave the TANF rolls increases as a result of
the E&T program. For example, if the rate increases from 1.2 to 2.0 percent and the five-
year time limit remains in place, the Markov model indicates that for every 100 cases
initially in the E&T program, only 50 cases will be enrolled once the caseload adjusts.6
Apparently, therefore, the combined effect of the time limit and the increase in the rate
at which cases exit from the TANF rolls is to greatly reduce program costs over time.
a. How Sensitive Are the Findings to the 1.2 Percent Assumption? It is now
assumed that the rate at which cases are leaving the TANF rolls each month is only 0.5
percent, a figure that is at the bottom of the range. It is also assumed that the rate of leav-
4. This 0.5 to 1.5 range is best viewed as a rough approximation. MDRC evaluations do not attempt to determine the
rate at which cases permanently leave the welfare roles each month. However, they do determine the percentage of
cases in their evaluation samples that are still on the rolls each month after the samples were initially drawn. It is this
information that we used to obtain the estimates upon which the 0.5 to 1.5 percent range is based. To obtain these
estimates, it was necessary to distinguish first between cases that left the rolls permanently and cases that left the
rolls only temporarily. This, in turn, required us to make certain strong simplifying assumptions. The estimates are
probably sensitive to these assumptions.
5. Zedlewski, S., and I. Sawhill. 1995.“Assessing the Personal Responsibility Act,” Welfare Reform Briefs, No. 5,
Washington, D.C.: The Urban Institute Press.
6. Because we first examined the effect of the five-year time limit on the number of cases in the E&T program, the
relatively small further effect of the increase in the rate at which cases permanently leave the TANF rolls is not sur-
prising. The increase affects only cases that have been on the rolls for less than five years and cases that are exempt
from the five-year time limit. All other cases left the rolls as a result of the time limit.
ing will increase to 1.0 percent per month, rather than to 2.0 percent, as a result of the
program. Given these new assumptions, the Markov model implies that for every 100
cases in the program when it is initially introduced, only 33 (rather than 50) cases will
be in the program once the caseload fully adjusts to the time limit and to the increase in
the rate at which cases exit the TANF rolls. This suggests that specific estimates based on
the Markov model are quite dependent upon specific assumptions about rates of leaving
TANF and, consequently, should be used with great caution. Perhaps more important,
however, regardless of the specific assumptions about exit rates, the model consistently
implies that the number of E&T participants and, hence, program costs would greatly
diminish over time.
In this example, most cases that are referred to a mandatory E&T program are
assigned to fairly expensive activities such as basic education and vocational training at
the time the program is initiated. However, the Markov model implies that a few years
after the program has been initiated, over 70 percent of the cases participating in the
program will have been in the program for over a year. These cases are likely to be reas-
signed to fairly inexpensive program activities such as workfare. As a result, program
costs may decline substantially.
b. Effect of Policy Decisions. The findings discussed above pertain to a specific
E&T program. It seems useful to examine how different policy decisions regarding the
program design affect the results. To investigate this issue, we change only one program
design characteristic at a time, leaving all the others unchanged. We also once again
assume that at the time the E&T program is implemented, 1.2 percent of the cases are
leaving the TANF rolls each month and that this rate increases to 2.0 percent as a result
of the program. The findings can, therefore, be compared with the earlier reported result
that for every 100 cases initially in the E&T program, only 50 cases would be in the pro-
gram after the caseload adjusts to the time limit and the increase in the rate at which
cases exit the TANF rolls. Four different program design changes are examined in the
examples that follow. Again, the Markov model is applied to estimate the results of the
changes.
e.g.
The 20 percent exemption to the five-year time limit is weakly enforced and, conse-
quently, 30 percent of the cases in the TANF caseload are actually exempt. Finding:
For every 100 cases in the E&T program when it is initially introduced, only 63 cases
will be in the program once the caseload fully adjusts to the time limit and the
increase in the rate at which cases exit the TANF rolls.
e.g. Instead of exempting 20 percent of the caseload from the five-year time limit, a deci-
sion is made to exempt no cases. Finding: For every 100 cases in the E&T program
when it is initially introduced, only 38 cases will be in the program once the caseload
fully adjusts to the time limit and the increase in the rate at which cases exit the TANF
rolls.
e.g. A decision is made to refer cases to the E&T program after they have accumulated one
year, rather than two years, of TANF benefits. Finding: For every 100 cases in the pro-
gram when it is initially introduced, only 62 cases will be in the program once the
caseload fully adjusts to the time limit and the increase in the rate at which cases exit
the TANF rolls.
e.g. A decision is made to reduce the lifetime limit on the TANF rolls to three years, rather
than five years. Finding: For every 100 cases in the program when it is initially intro-
duced, only 27 cases will be in the program once the caseload fully adjusts to the time
limit and the increase in the rate at which cases exit the TANF rolls.
These findings suggest that the amount by which the number of cases in a man-
datory E&T program and, hence, program costs would diminish over time would vary
greatly with the specific program design and the manner in which the program is
implemented. Regardless of the specific program design, however, it seems evident
that program costs would greatly diminish over time. Anyone costing out a not-yet-
implemented E&T program or using the resulting cost estimates should be aware of
this possibility.
5.5 Predicting the Net Costs of a

Not-Yet-Implemented Program
As shown in chapter 2, estimates of the net costs of a proposed E&T program are more
useful for certain purposes, such as determining whether the proposed program should
replace an existing program, than are estimates of the program’s gross costs. For such
purposes, the net costs of the proposed program can be determined by subtracting an
estimate of the aggregate gross costs of the current program from the estimated aggre-
gate gross costs of the proposed program. In other words, the difference between these
two estimates provides a measure of the cost of replacing the old program with the pro-
posed one.
An estimate of gross cost per case referred to the current program can be obtained
by using the seven-step procedure described in chapter 3. Because the estimate of the
cost of the proposed program is predicted for a 12-month period, it is important that
the estimate of gross cost per case for the current program also pertain to a year. Thus,
cost information for the research sample used to conduct the cost analysis of the ongo-
ing program should be collected over a 12-month period. Otherwise, the cost estimates
for the two programs will not be comparable.
The estimate for the current program can be converted into an aggregate measure
by multiplying it by the number of cases referred to the current program over 12
months. In doing this, it is important to multiply by the number of cases referred to the
ongoing program, not the number that will be referred to the not-yet-implemented pro-
gram. Otherwise, the net cost estimate will not appropriately reflect the effect of differ-
ences between the two programs in the sizes of their caseloads.
A SUMMARY OF THE STEPS FOR ESTIMATING THE GROSS COSTS

OF A NOT-YET-IMPLEMENTED PROGRAM
Step 1. Develop a flow diagram. Develop a detailed flow diagram of the proposed program.
Step 2. Predict total program referrals. Predict the total number of cases that will be referred to the
program during the first 12 months after it is implemented.
Step 3. Predict the assignment rates for each program component. Of those cases referred to the
program (see step 2), predict the fraction that will be assigned during the program’s first year to
each of the program activities that appear in the flow diagram developed in step 1.
Step 4. Predict the participation rate for each program component. Of those cases assigned to each
of the program activities, predict the fraction that will actually participate in the activity.
Step 5. Predict the average length of participation. Predict the average length of time that cases that
participate in each program activity will remain in the activity.
Step 6. Estimate unit costs. Using time units that are consistent with those used in step 5, estimate the
unit cost of each program component.
Step 7. Compute costs per referred case for each program activity and support service. Make the
following computation for each program activity and support service: step 3 x step 4 x step 5 x
step 6.
Step 8. Compute total costs per referred case. To obtain a prediction of total program costs per
referred case, sum all the average cost values obtained in step 7.
Step 9. Compute aggregate program costs. To obtain a prediction of aggregate program costs,
multiply the value obtained in step 2 by the value obtained in step 8.
Chapter 6
Using Cost Information in

Designing, Bidding, and
Budgeting
The earlier chapters of this guide provide detailed suggestions on how to conduct a cost
analysis of an ongoing E&T program and how to predict the costs of a proposed E&T
program. Once the information required for either type of cost analysis has been gath-
ered and assembled, there are several additional ways in which it can be used. This chap-
ter first explains why conducting sensitivity tests in performing cost analyses is useful
and important, and how such tests can be conducted. The chapter then discusses how
community-based organizations that are bidding to provide employment and training
program services can use the guide in developing cost estimates, and how program
agencies can use the guide to structure bids so they can be readily compared. The last
section of the chapter describes how the information provided in the guide can be used
to develop a budget for a proposed employment and training program — specifically,
how total program costs can be broken down into the program’s operational
components.
81
6.1 Sensitivity Test

Chapter 3 described a seven-step method for determining the gross costs of currently
operating employment and training programs. In Chapter 4, Table 4.8 used the Riverside
LFA program to illustrate how these steps can be applied to an actual ongoing program.
For the reader’s convenience, the required computations are repeated in Table 6.1.
W2@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@6X?g
7@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@)Xg
@@(Mf@@
@@H?f@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@g@@
@@fI'@1g
@@f?N@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
@@g@@g
The calculations that appear in Table 6.1 were made by the computer spreadsheet
@@g@@
@@g@@ ?J@@g@@g
O&@@g@@g
@@g3@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@5g@@g
@@gV4@@@@@@@@@@@@@@@@@@@@@@@@@@@@0Yg@@g
found on the disk accompanying this guide. By using this spreadsheet, it becomes very
@@ @@g
@@
@@ @@g
@@g
@@ @@g
@@ @@g
@@
@@he?W2@@@@@@@@@@@@@@@@@6Khf@@g @@g
@@heW&@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he7@@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@L?h@@@?g@@@@@@@@@@@@@?he@@g
@@)Xh@@@?g@@@@@@@@@@@@@?he@@g
3@@)X?g@@@?g@@@@@@@@@@@@@?he@@g
V'@@)Xg@@@?g@@@@@@@@@@@@@?he@@g
?V'@@)X?f@@@?g@@@@@@@@@@@@@?h?J@@g
V'@@)Kf@@@?g@@@@@@@@@@@@@?hO&@5g
?V'@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@Hg
V4@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@?g
easy to conduct “sensitivity tests” — that is, to determine how the costs of a current pro-
gram will change if the program is modified (for example, if the number of cases
assigned to various program activities is changed or if a new activity is introduced). The
remainder of this section illustrates a number of such sensitivity tests. (Example 2 in
section 5.2 illustrates the use of sensitivity tests in conducting cost analyses of not-yet-
implemented employment and training programs.)
Table 6.2 lists a number of (hypothetical) modifications to the Riverside LFA pro-
gram that a welfare agency might consider. Predictions of the changes in costs that
would result from these program modifications are reported in Table 6.2. As should be
evident from the table, determining the cost effects of modifying an ongoing program
sometimes requires predictions of unit costs, participation rates, and the average length
of stay in program activities. These values can be obtained by using the techniques
described in chapter 5.
Table 6.2 indicates that some of the hypothetical program modifications would
increase costs and others would decrease costs. Moreover, relative to the total cost of
$1,163 per referred case, some of the modifications would result in very modest changes
in costs, while others would cause substantial changes. While each program modifica-
tion is considered separately here, several could obviously be combined. Some of the
savings from reducing the length of time in basic education, for example, could be used
to pay for increasing unit costs by enriching this activity while cases are participating,
perhaps by reducing student/teacher ratios. Furthermore, the cost implications of more
fundamental changes in the program, such as those that might be required by a chang-
ing case mix or adopting a different program model, can be examined by conducting
additional, somewhat more elaborate, sensitivity tests. Two examples of these more basic
modifications are considered next.
e.g.
Costs would change if the Riverside LFA model were to serve a more difficult case-
load — for instance, if relatively job-ready TANF recipients were served first by the
program, and the program later began to enroll recipients with less job experience and
education, or if the scale of program operations increased and, as a consequence, less
job-ready, as well as more job-ready, cases were served.
Table 6.3, which was computed using the disk that accompanies this guide, provides a
W2@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@6X?g
7@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@)Xg
@@(Mf@@ @@fI'@1g
@@H?f@@ @@f?N@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
@@g@@g
@@g@@ @@g@@g
@@g@@
@@g@@ @@g@@g
?J@@g@@g
@@g@@ O&@@g@@g
prediction of the changes in costs that could result if less job-ready cases were to be
@@g3@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@5g@@g
@@gV4@@@@@@@@@@@@@@@@@@@@@@@@@@@@0Yg@@g
@@ @@g
@@
@@ @@g
@@g
@@ @@g
@@
@@ @@g
@@g
@@he?W2@@@@@@@@@@@@@@@@@6Khf@@g
@@heW&@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he7@@@@@@@@@@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@he@@@?g@@@@@@@@@@@@@?he@@g
@@L?h@@@?g@@@@@@@@@@@@@?he@@g
@@)Xh@@@?g@@@@@@@@@@@@@?he@@g
3@@)X?g@@@?g@@@@@@@@@@@@@?he@@g
V'@@)Xg@@@?g@@@@@@@@@@@@@?he@@g
?V'@@)X?f@@@?g@@@@@@@@@@@@@?h?J@@g
served but the program model were to remain basically intact. The table is meant only to
V'@@)Kf@@@?g@@@@@@@@@@@@@?hO&@5g
?V'@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@Hg
V4@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@?g
Chapter 6: Using Cost Information in Designing, Bidding, and Budgeting 83
Table 6.1

Two-Year Costs of Riverside’s Labor Force Attachment Approach
Step 3
Select the Analysis

PROGRAM ACTIVITIES
Welfare department $ 79 1.30 d 1 session $ 103
Formal assessment
Job search
Basic education
College
Non-welfare department $ 500 f 0.00 0 months $ 0
Work experience
SUPPORT SERVICES
Child care
Transportation
Ancillary services
SOURCES: Gayle Hamilton et al., National Evaluation of Welfare-to-Work Strategies, 1997. For support services: unpublished
data from MDRC research.
NOTES: These data are for the subsample without a high school diploma or GED certificate.
Rounding may cause slight discrepancies in calculating sums.
a. The data on participation rates (step 4) and average length of participation (step 5) were obtained from a sample of
193 cases that were assigned to Riverside’s LFA program.
b. The percentage of the analysis sample participating in each activity and receiving each service within two years of
attending orientation.
d. On average, 1.3 sessions were attended.
e. $4 per hour for 132.0 hours of basic education.
f. $5.73 per hour for 87.3 hours of college.
Table 6.2
ILLUSTRATIONS OF CHANGES IN COSTS RESULTING FROM ALTERNATIVE

MODIFICATIONS TO THE RIVERSIDE LFA PROGRAM
Change in Costs per Referred Case

Resulting from the Modification
Welfare Non-Welfare
Program Modification Department Department Total
Increase in the job search participation

rate from .55 to .7 $233 $ 0 $233
Decrease average length of stay in basic
education by two months -10 -24 -34
Enrich basic education so that non-welfare
unit costs increase by $200 0 19 19
Eliminate transportation allowances -32 0 -32
Provide vocational training to 5% of the
cases referred to the programa 43 165 208
NOTE: a. The following values were used to make these calculations: unit cost for the welfare department = $110; unit cost
for non-welfare departments = $422; predicted average length of stay = 7.8 months. These values were adopted from the
Riverside HCD program, which, unlike the Riverside LFA program, offered vocational training.
illustrate the sort of simulation exercise that could be conducted to ascertain the cost
implications of moving from a more job-ready to a less job-ready caseload. Thus, it is
based on hypothetical assumptions.
A comparison of Table 6.3 with Table 6.1 indicates, not surprisingly, that costs per case
would increase substantially (from $1,163 to $2,010) were the program to serve a more
difficult caseload. A more detailed comparison of the two tables suggests that changes in
costs would result because numerous different adjustments would be necessary as case-
load composition changes. Indeed, one of the major advantages of using a spreadsheet
to simulate the effects of a change in caseload composition on costs is that a multitude of
adjustments can readily be taken into account at once. At the same time, however, the
simulation is also helpful in illuminating the major factors that would cause costs to
change.
In the illustration, much of the cost increase results because it is assumed that many
more members of the less job-ready caseload would be assigned to basic education than
members of the more job-ready caseload, and these cases would remain in basic educa-
tion longer. In addition, it is further assumed that unit costs for basic education would
rise because additional case management would be required. Substantially higher costs
also result because it is assumed that to more effectively serve the less job-ready case-
load, the program agency would hire an outside contractor (a retention service) to
follow up on those cases that obtain jobs. The contractor would help resolve problems
that arise and receive an incentive payment for each person who remains employed for
at least 90 days.
Table 6.3

Serving a Harder-to-Serve Caseload:
A HYPOTHETICAL EXAMPLE
Step 3
Select the Analysis
Sample for Steps 4 and 5

Program Estimates of Rate a Length of Referred Case Cost per
Components Program Stay b per Program Referred
PROGRAM ACTIVITIES
Welfare department $ 100 1.27 c 1 session $ 127
Non-welfare department $ 0 1.27 c 1 session $ 0
Formal assessment
Retention services
Job search
Basic education
Non-welfare department $ 516d 0.14 6 months $ 433
College
Welfare department $ 110 0 0 months $ 0
Non-welfare department $ 500 e 0 0 months $ 0
Work experience
SUPPORT SERVICES
Child care
Transportation
Ancillary services
Welfare department $ 72 0.25 1 month $ 18
Non-welfare department $ 0 0.25 1 month $ 0
NOTES: a. The percentage of the analysis sample participating in each activity and receiving each service within two years
of attending orientation.
b. Average months participating in each program activity and receiving each support service per participant within two
c. On average, 1.27 sessions were attended.
d. $4 per hour for 129.0 hours of basic education.
e. $5.73 per hour for 87.3 hours of college.
e.g. To illustrate a simulation test of a fundamental change in an E&T program, we consider

a TANF agency that is considering shifting its E&T program from a human capital
development model to a labor work force attachment model, but leaving the composi-
tion of the program caseload unchanged. Table 6.4 provides cost estimates of the HCD
approach used in Riverside. (See chapter 4 for a description of the HCD program
model.) For purposes of this example, these estimates are used as measures of costs
under the current program model. Table 6.1, as previously mentioned, provides cost
estimates of Riverside’s labor force attachment model. For illustrative purposes, it is
assumed that these estimates represent simulation predictions of the costs of an LFA
E&T approach.
A comparison of the two tables indicates that costs per case could be cut by more than
half by shifting from the HCD approach to an LFA approach (from $2,990 to $1,163).
A comparison of Tables 6.1 and 6.4 further suggests that although numerous factors
cause costs to vary between the two program models, the key reason the LFA approach
is less expensive than the HCD model is that it makes much less use of basic education.
Under the LFA approach, only 2 percent of the cases assigned to the program participate
in this activity. Under the HCD model, in contrast, over half the cases assigned to the
program participate in basic education. Moreover, on average, they remain in basic edu-
cation for nearly seven months, more than one and a half months longer than under the
LFA approach.
The examples presented above demonstrate that while a sensitivity test is usually
easy to conduct, it can be a powerful tool. As mentioned in chapter 5, simulation tests
should be conducted whenever uncertainty exists concerning predictions (such as
predictions of unit costs and participation rates) that are required in costing out a not-
yet-implemented E&T program. Such tests help indicate whether poor predictions are
likely to result in large errors in the cost estimates. Perhaps more important, sensitivity
tests can be used to examine the cost implications of numerous alternative policy
choices, and this can be done in the context of either an ongoing program or a proposed
program. Even when an ongoing program is being substantially modified, it is often
unnecessary to treat it as an entirely new program and, hence, conduct a cost analysis
that requires use of the rather elaborate procedures described in chapter 5. A sensitivity
test of the current program can frequently be used instead.
6.2 Bids for Program Components by

Community-Based Organizations
E&T program agencies have frequently “farmed out” the actual operation of certain
program components — for example, vocational training and day care — to outside
organizations. Since the passage of the 1996 federal welfare law, welfare agencies
have increasingly encouraged community-based organizations (CBOs) to bid on the
components of welfare-to-work programs.
Table 6.4

Riverside’s Human Capital Development Approach
Step 3
Select the Analysis

PROGRAM ACTIVITIES
Formal assessment
Job search
Basic education
Non-welfare department $ 274 d 0.55 7.0 months $ 1,042
Vocational training
SUPPORT SERVICES
Child care
Transportation
Ancillary services
NOTES: Rounding may cause slight discrepancies in calculating sums.

a. The data on participation rates (step 4) and average length of participation (step 5) were obtained from a sample of
435 cases that were assigned to Riverside’s HCD program.
b. The percentage of the analysis sample participating in each activity and receiving each service within two years of
attending orientation.
d. $3.64 per hour for 75 hours of basic education.
e. $4.78 per hour for 88 hours of vocational training.
With small modifications, the first six of the nine steps that were developed in chap-
ter 5 for costing out a not-yet-implemented welfare-to-work program provide a useful
framework that can be used by both program agencies in writing requests for bids and
CBOs in responding to these requests. CBOs should find the framework helpful in
developing a budget for their bid. Moreover, by requiring CBOs to use the framework,
program agencies can help ensure that the bids they receive will be more uniform and,
consequently, more readily compared.
The remainder of this section describes the six steps and suggests how each one can
be used in the bidding process.
Step 1. Develop a Flow Diagram

The program agency should include a detailed flow diagram of the overall program in
its request for bids so that potential bidders will know how the component on which
they may bid fits. (See Figures 4.1 and 4.2 for illustrations of flow diagrams.)
Step 2. Predict Total Program Referrals

The request for bids should include an estimate of the total number of cases that will
be referred to the program during the 12 months after the contract with the successful
bidder begins. Bids should be based on these cases and on this time frame.
Step 3. Predict the Assignment Rate for the Program Component

The request for bids should indicate the fraction of those cases referred to the program
(see step 2) that will be assigned the program component for which bids are being
solicited. If some cases will be assigned to a program component more than once during
the 12-month period, they should be counted in the numerator of the assignment rate
once for each time they are assigned.
Step 4. Predict the Participation Rate for the Program Component

The request for bids should predict the fraction of those cases that are assigned to the
program component that are likely to actually participate in the activity. The techniques
suggested in chapter 5 can be used to obtain this prediction. If bidders believe that the
actual participation rate will be smaller or larger than that suggested in the request for
bids, they should indicate so in their bid and explain why. A CBO that is bidding on a
job club, for example, may plan to make special efforts to encourage those assigned to
job clubs to participate.
Step 5. Predict the Average Length of Participation

The request for bids should contain a prediction of the average length of time that par-
ticipating cases will continue to participate in the program component. The techniques
suggested in chapter 5 can be used for deriving this prediction. CBOs should be invited
to suggest an alternative prediction in their bids and to state the reasons why this alter-
native should be used in estimating costs. For example, bidders on basic education or
vocational training may plan on making special efforts to discourage participants from
dropping out. Note that for certain program components, such as child care, some cases
will continue to participate for more than 12 months. Nonetheless, the length of partici-
pation averages should be based on a 12-month period so that the budgets submitted by
bidders pertain to a year.
Step 6. Estimate Unit Costs

The request for bids should require bidders to provide an estimate of the unit costs of
providing the service for which they are bidding. This estimate, which should use a time
unit that is consistent with that used in step 5, can be based on the group cost approach
described in chapter 5 and should be justified in detail. This will be a key part of the bid.
Once the six steps are completed, bidders can then determine the number of cases
that will participate in the program component by making the following computation:
step 2 x step 3 x step 4
Bidders can obtain an estimate of costs per participant by making the following
computation:
step 5 x step 6
Total costs for the program component can then be computed by multiplying the
predicted number of participants by the estimate of costs per participant.
6.3 Developing a Program Budget

Prior to implementing a new employment and training program, it is necessary to devel-
op a program budget that indicates the projected expenditures for the program period
(usually one year). If unit costs for the new program are estimated with data from the
sites adopting the new program, it is usually relatively simple to develop a program bud-
get, as the required information on the cost items that must be included in a program
budget — for example, salaries paid to classroom teachers, salaries paid to support staff,
rent for classrooms, equipment costs, and so forth — is typically now available.
e.g.
How might a program budget be developed as part of the same process as estimating
unit costs? Recall the example used in chapter 5 to describe how the program unit
approach could be used to forecast the unit cost of a job club. In that example, it was
assumed that the job club would meet for four hours a day over a three-week period,
that 10 people would be enrolled in each job club, and that 1,000 people would partici-
pate during a year. Consequently, about 100 job clubs would be run over the year, with
two initiated each week, and, because each club lasts three weeks, six would need to
operate at any point in time. If three classes operated in the morning and three in the
afternoon, three classrooms would be required.
Before this information can be used to develop a program budget, a number of

rather detailed questions first need to be answered.
■ Can an instructor conduct one class in the morning and another in the afternoon so
that only three instructors would be required? Or would preparation time be required? If
so, how much? Would instructors also be responsible for following up with those placed
(or not placed) by the job club? If so, the number of required instructors would obvious-
ly increase.
■ If instructors are not responsible for follow-up, who would be and how much of their
time would be required? For example, case managers might be used.
■ How much case management time would be required for outreach and tracking (and,
possibly, follow-up)?
■ How much supervisory time would be required? How much support staff time would be
required? Is there a standard ratio, perhaps specified by civil service regulations, that can
be used for determining these requirements?
■ Would all 10 students in a classroom need telephones at once for calling prospective
employers, or would they alternate?
■ What other equipment or supplies would be needed?
Once these and other relevant questions are answered, a similar process can be
followed for each of the other planned program activities and services. Then, the
program budget itself can be readily developed. An illustrative prototype of such a
budget appears in Table 6.5.
Table 6.5
PROGRAM BUDGET PROTOTYPE
Instructors
Number x salary and fringes per instructor $
Supervisors
Number x salary and fringes per supervisor $
Support staff
Number x salary and fringes per support staff person $
Case managers
Number x salary and fringes per case manager $
Classroom rental
Number of rooms x rent per room $
Phone banks
Number of phones x cost per phone $
Other equipment and supplies

$
$
$
Child care allowances

Number of cases receiving allowance x allowance per case $
Transportation allowances
Number of cases receiving allowance x allowance per case $
Appendix A
Selected Employment and Training Programs:

Summary Descriptions and Main Evaluation
Findings
• California’s Greater Avenues for Independence (GAIN) Program
• Florida’s Project Independence (PI)
• JOBSTART Demonstration
• Minnesota Family Investment Program (MFIP)
• National Evaluation of Welfare-to-Work Strategies
• National Job Training Partnership Act (JTPA) Study
93
1. California’s Greater Avenues for Independence (GAIN) Program1
Summary. This evaluation, conducted during the early 1990s, tested the effects of a
mandatory welfare-to-work program that provided basic education, job search, skills
training, and work experience to AFDC recipients. GAIN’s effects were estimated for a
sample of 33,000 persons which included single parents (AFDC-FGs) and unemployed
heads of two-parent households (AFDC-Us). The follow-up period covers five years.
Treatments Tested. Program group members were subject to the GAIN participation
mandate and had access to case management and services. The control group was pre-
cluded from receiving services from the program, but could seek other services in the
community on their own.
Target Population. AFDC recipients. Four of the six counties had resources to include
all mandatory registrants; two focused on long-term recipients.
Number and Location of Sites. Six counties in California: Alameda, Butte, Los Angeles,
Riverside, San Diego, and Tulare.
Major Findings. Effects varied by county.

1. Averaged across the six counties (with each county given equal weight), the GAIN
program increased the percentage of AFDC-FGs who worked for pay during the five-
year follow-up by 4.3 percentage points and raised average earnings by $2,853.
Employment impacts generally decreased over time, whereas earnings gains were
largest during years 4 and 5.
2. For AFDC-FGs, five-year AFDC savings averaged $1,496 across the six counties.
Moreover, the percentage reduction in AFDC payments was somewhat larger during
the last two years of follow-up than during years 1, 2, or 3.
3. Five-year earnings gains and AFDC savings for AFDC-FGs were achieved in all six
counties, although for some effects and some counties the program–control group
differences were small and not statistically significant.
4. Riverside’s GAIN program produced the largest increase in total earnings ($5,038) for
AFDC-FGs and the largest reduction in AFDC expenditures ($2,705).
5. GAIN increased the percentage of AFDC-Us who found employment by 6.3 percent-
age points over five years. Earnings gains totaled $1,906 over five years and reached a
maximum in year 5.
1. The information on the GAIN evaluation, JOBSTART Demonstration, and National JTPA Study is excerpted
from David Greenberg and Mark Shroder, The Digest of Social Experiments, Second Edition (Washington, D.C.:
The Urban Institute Press, 1997). Used by permission of The Urban Institute Press. For this guide, the GAIN
findings have been updated to reflect a full five years of follow-up, as presented in The GAIN Evaluation,
Working Paper 96.1, Five-Year Impacts on Employment, Earnings, and AFDC Receipt.
Appendix A: Selected Employment and Training Programs 95
6. GAIN reduced AFDC payments to AFDC-Us by an average of $1,432 over five years.
However, AFDC savings declined substantially during years 4 and 5.
Time Trends in Findings. The earnings impacts grew progressively stronger over time
and were largest during years 4 and 5, whereas the impacts on AFDC receipt tended to
level off over time.
Information Sources
Stephen Freedman, Daniel Friedlander, Winston Lin, and Amanda Schweder. The GAIN
Evaluation. Working Paper 96.1. Five-Year Impacts on Employment, Earnings, and AFDC
Receipt (New York: MDRC, 1996).
James Riccio, Daniel Friedlander, and Stephen Freedman. GAIN: Benefits, Costs, and
Three-Year Impacts of a Welfare-to-Work Program (New York: MDRC, 1994).
2. Florida’s Project Independence (PI)
Summary. PI is Florida’s JOBS program. The program addressed the question of

whether welfare costs can be reduced and the employment and income of welfare recipi-
ents increased. Subjects were followed for a two-year period.
Treatments Tested. Education and training, job search.
Target Population. Single-parent AFDC recipients and applicants.
Number and Location of Sites. Nine counties in Florida — Bay (Panama City),
Broward (Fort Lauderdale), Dade (Miami), Duval (Jacksonville), Hillsborough (Tampa),
Lee (Fort Myers), Orange (Orlando), Pinellas (St. Petersburg), and Volusia (Daytona
Beach).
Major Findings
1. Sixty-four percent of program group members participated in employment-related
activities through PI and other sources.
2. Forty percent of control group members participated in employment-related
activities.
3. PI produced a substantial increase in the use of employment-related activities among
program group members. Most of this increase was accounted for by their much
greater use of relatively low-cost independent job search activities.
4. For the full sample, PI produced a modest decrease in program group members’
AFDC and Food Stamp receipts — a decrease that persisted over the two-year
follow-up period.
5. For the full sample, the modest earnings gains achieved by program group members
in the first year of follow-up declined greatly in the second year. Overall, PI produced
only a small increase in the average earnings of program group members over the two
years.
6. PI produced AFDC savings for both sample members with preschool-age children and
those with only older children.
7. From a benefit-cost standpoint, PI was most successful for the early group of AFDC
applicants and recipients with no preschool-age children. For this group, the program
produced budgetary savings and made program group members better off financially.
Information Source
James Kemple, Veronica Fellerath, and Daniel Friedlander. Florida’s Project Independence:
Benefits, Costs, and Two-Year Impacts of Florida’s JOBS Program (New York: MDRC,
1995).
3. JOBSTART Demonstration
Summary. This demonstration, conducted between 1985 and 1992, tested the effects of
education, training, and support services on a large sample of economically disadvan-
taged school dropouts. Subjects were followed for four years.
Treatments Tested. Education and vocational training, support services, and job place-
ment assistance. Support services included assistance with transportation, child care,
counseling, and incentive payments.
Target Population. Economically disadvantaged school dropouts ages 17–21 who read
below the eighth-grade level and were eligible for Job Training Partnership Act (JPTA)
Title II-A programs or the Job Corps. Some sites screened out youth with problems that
the program was not equipped to handle. These problems included emotional problems,
drug and alcohol abuse, health problems, unstable living conditions, poor motivation,
and those who were likely to prove dangerous or disruptive.
Number and Location of Sites. Thirteen sites — Buffalo and New York, New York;
Atlanta, Georgia; Hartford, Connecticut; San Jose, Monterey Park, and Los Angeles,
California; Chicago, Illinois; Pittsburgh, Pennsylvania; Dallas and Corpus Christi, Texas;
Denver, Colorado; and Phoenix, Arizona.
Major Findings
1. Significantly more treatment group youths completed high school or its equivalent
as compared to the control group (42 percent versus 28.6 percent). This impact was
fairly large for all subgroups studied.
2. Employment was significantly greater for the control group in the first year of follow-
up. In the second year, significantly more treatment group youths were employed.
There were no significant differences in the third and fourth years regarding this vari-
able.
3. Similarly, control group members earned more in the first two years after follow-up,
whereas the treatment group earned more in the third and fourth years. This finding
was statistically significant only for year one.
4. Few significant findings are found in subgroup analyses regarding employment.
Trends are similar to those for the full sample.
5. For women who were not custodial mothers when the program started, there was a
consistent pattern of reductions in AFDC receipt and payments, and many of these
impacts were statistically significant. However, for women who were custodial moth-
ers, JOBSTART significantly increased child bearing and had no impacts on AFDC
receipt.
6. There is some indication that JOBSTART led to a reduction in criminal activity
(arrests and drug use), although impacts are generally not significant.
7. The San Jose site had higher earnings impacts than any other site. This finding was
significant at the .05 level. The reasons for this are unclear. At the San Jose site, train-
ing and placement efforts were closely linked to the labor market, education and
training efforts were coordinated, and the program had a clear organizational mis-
sion. Any or all of these factors may have affected program impacts.
Time Trends in Findings. Treatment group payoffs did not usually occur until after
year 2.
Information Sources
George Cave, Fred Doolittle, Hans Bos, and Cyril Toussaint. JOBSTART: Final Report on
a Program for School Dropouts (New York: MDRC, 1993).
George Cave and Fred Doolittle. Assessing JOBSTART: Interim Impacts of a Program for
School Dropouts (New York: MDRC, 1991).
4. Minnesota Family Investment Program (MFIP)
Summary. MFIP was implemented on a field trial basis in three urban counties and four
rural counties to test financial incentives and welfare-to-work mandates for applicants
and recipients.
Treatment Tested. Financial incentives, mandatory employment-related activities, direct

child care payments, and simplification of public assistance rules.
Target Population. AFDC applicants and recipients.
Number and Location of Sites. Three urban counties — Hennepin, Anoka, and
Dakota; four rural counties: Mille Lacs, Morrison, Sherburne, and Todd.
Major Findings
1. For single-parent, long-term recipients in urban areas, MFIP substantially increased
employment and earnings during the first 18 months.
2. MFIP substantially reduced poverty for long-term, single-parent recipients in urban
areas by increasing their earnings and limiting the reduction in their welfare benefits
(compared with AFDC) when they worked.
3. It is MFIP’s combination of financial incentives and mandatory employment-focused
activities that achieved the goals of increased employment and reduced poverty.
4. For applicants in urban areas, MFIP produced a modest increase in employment and
no increase in earnings during the first 18 months, and it increased the payout of
welfare benefits.
5. MFIP increased income and reduced poverty among single-parent applicants. In con-
trast to long-term recipients, applicants’ increased income came entirely from the
increase in welfare payments to applicants who worked.
6. MFIP produced no sustained increase in employment or earnings among long-term
recipients in rural areas, but it did increase welfare receipt and reduce poverty.
Information Source
Cynthia Miller, Virginia Knox, Patricia Auspos, Jo Anna Hunter-Manns, and Alan
Orenstein. Making Welfare Work and Work Pay: Implementation and 18-Month Impacts of
the Minnesota Family Investment Program (New York: MDRC, 1997).
5. National Evaluation of Welfare-to-Work Strategies2
Summary. MDRC is in the midst of the most comprehensive national evaluation of

welfare-to-work programs ever undertaken — the National Evaluation of Welfare-to-
Work Strategies. The study encompasses 11 programs in seven sites and 55,000 sample
members.
Treatment Tested. In three of the program sites (Atlanta, Georgia; Grand Rapids,
Michigan; and Riverside, California), the evaluation measures the effectiveness of two
alternative employment preparation strategies: a human capital development approach
(HCD), in which individuals who could potentially benefit from education and training
2. Formerly known as the JOBS Evaluation.

are provided with these services before they seek work, under the theory that they will
then be able to get better jobs, and a labor force attachment approach (LFA), in which
individuals are encouraged to gain quick entry into the labor market so that they can
build up their work habits and skills and advance themselves on the job. In a fourth site
(Columbus, Ohio), the evaluation is measuring the effectiveness of a traditional case
management approach, in which different case workers handle income maintenance and
welfare-to-work case management, as opposed to an integrated case management
approach, in which one worker handles both functions. The evaluation also tests the
effectiveness of welfare-to-work programs in Detroit, Michigan; Oklahoma City,
Oklahoma; and Portland, Oregon.
Target Population. AFDC recipients and approved applicants in six sites; AFDC appli-
cants who were not yet approved for welfare in Oklahoma City.
Number and Location of Sites. Seven sites: Atlanta, Georgia; Grand Rapids, Michigan;
Riverside, California; Columbus, Ohio; Detroit, Michigan; Oklahoma City, Oklahoma;
and Portland, Oregon.
Major Findings
Over the last seven years, this project has provided support for 20 major books, reports,
and papers. Presented below is a summary of findings from two of the most recent
reports. The first report presents implementation, participation patterns, costs, and
two-year employment and earnings impacts for the LFA and HCD programs in Atlanta,
Grand Rapids, and Riverside. The second report presents implementation, participation
patterns, costs, and two-year employment and earnings impacts for Portland’s program.
1. Both the LFA and HCD programs increased individuals’ two-year cumulative employ-
ment and earnings. On average, one out of every five welfare recipients who normally
would not have worked in an unsubsidized job during the two-year follow-up period
did so as a result of the LFA programs. Two-year earnings were increased by more
than $1,000 per average LFA sample member. The HCD programs in two of the three
sites led to small two-year increases in earnings and employment.
2. The cumulative employment and earnings impacts over the two-year period were
smaller for the HCD programs than for the LFA programs.
3. Both the LFA and HCD programs reduced welfare expenditures within the two-year
follow-up period. Relative to the total welfare payments that the control groups
received over the two years, the LFA and HCD programs reduced welfare expendi-
tures between 6 and 18 percent.
4. The Portland program substantially increased employment and produced unusually
large increases in earnings. The program raised employment levels by 11 percentage
points over two years (relative to the control group). In addition, two-year earnings
were increased by over $1,800 per sample member, a 35 percent increase over the
control group’s earnings.
5. Unlike many programs that produce immediate impacts on employment and earn-
ings, the Portland program increased job quality. At the end of two years, the pro-
gram increased the proportion of people working at full-time jobs by 13 percentage
points and, among those employed, increased average hourly pay by $0.86.
6. The program reduced welfare expenditures by 17 percent over the two-year follow-up
period.
7. Portland’s impacts were widespread: both recipients with relatively few barriers to
employment and those typically considered very hard to place achieved employment
and earnings gains and AFDC reductions. Few other programs have attained such
consistent impacts.
Information Sources
Gayle Hamilton, Thomas Brock, Mary Farrell, Daniel Friedlander, and Kristen Harknett.
National Evaluation of Welfare-to-Work Strategies: Evaluating Two Welfare-to-Work
Program Approaches – Two-Year Findings on the Labor Force Attachment and Human
Capital Development Programs in Three Sites (Washington, D.C.: U.S. Department of
Health and Human Services and U.S. Department of Education, 1997).
Susan Scrivener, Gayle Hamilton, Mary Farrell, Stephen Freedman, Daniel Friedlander,
Marisa Mitchell, Jodi Nudelman, and Christine Schwartz. National Evaluation of Welfare-
to-Work Strategies: Implementation, Participation Patterns, Costs, and Two-Year Impacts
of the Portland (Oregon) Welfare-to-Work Program (Washington, D.C.: U.S. Department
of Health and Human Services and U.S. Department of Education, 1998).
6. National Job Training Partnership Act (JTPA) Study
Summary. This evaluation, conducted from 1987 to 1994, tested the effects of the Job
Training Partnership Act (JTPA) Title II program’s employment and training services on
a large sample of economically disadvantaged adults and youths. Subjects were followed
for 30 months.
Treatments Tested. Access to Title II services under the JTPA. Participants were divided
into three groups by local staff according to which services were deemed appropriate.
They were then randomly assigned to a treatment or control group for each service
strategy. Specific services varied widely across sites, but could include the services shown
in Table A.1.
Table A.1
JTPA SERVICES
Service Strategy
Classroom On-the-Job Training Other

Training (OJT)/Job Search Activities
Specific Program Service Group Assistance Group Group
Classroom training in occupational skills Yes No Yes

OJT No Yes Yes
Job search assistance Yes Yes Yes
Basic education Yes Yes Yes
Work experience Yes Yes Yes
Miscellaneous Yes Yes Yes
Target Population. Eligible JTPA Title-II adults and out-of-school youth. The study
focused on four subgroups: adult women; adult men; female out-of-school youth; and
male out-of-school youth.
Number and Location of Sites. Sixteen sites throughout the United States: Fort Wayne,
Indiana; Coosa Valley, Georgia; Corpus Christi, Texas; Jackson, Mississippi; Providence,
Rhode Island; Springfield, Missouri; Jersey City, New Jersey; Marion, Ohio; Oakland,
California; Omaha, Nebraska; Larimer County, Colorado; Heartland, Florida; Butte,
Montana; Decatur, Illinois; Cedar Rapids, Iowa; and Northwest, Minnesota.
Major Findings. Because the control group was able to receive employment and train-
ing services from non-JTPA providers, impacts reflect the incremental effect of JTPA
services beyond what sample members could have accomplished without access to JTPA.
Impacts were estimated separately by subgroups: adult men; adult women; female youth;
male youth, nonarrestees; and male youth, arrestees.
Adults
1. The treatment group received significantly (p = .01) more employment and training
services than did the control group; on average, men received 169 more hours of ser-
vice and women received 136 more hours.
2. For adult women, average earnings over the 30-month period following random
assignment were $1,176 (9.6 percent) greater for the treatment group than the control
group. This is significant at the .01 level. For men, earnings were $978 (5.3 percent)
greater for the treatment group. This is significant at the .10 level.
3. Earnings gains came more from an increase in hours worked (an employment effect)
than from an increase in average hourly earnings (a wage effect). This was especially
true for women.
4. JTPA resulted in a substantial and statistically significant impact on the attainment of
a high school credential (diploma or its equivalent) for adult female school dropouts.
The findings for adult males were also positive, although not statistically significant.
5. The greatest earnings impact was estimated for women in the OJT/JSA and other
activities subgroups.
6. For adult women, there was no significant program impact on Aid to Families with
Dependent Children (AFDC) or Food Stamp receipt. For men, there was a small, but
significant, increase in AFDC receipt for the treatment group.
Youth
1. JTPA resulted in a significant increase in the amount of employment and training
services for all categories of youth. (Female youth in the treatment group received,
on average, 182 more hours than their control group counterparts; male youth
nonarrestees received 175 more hours; and male youth who had a prior arrest
[arrestees] received 127 more hours.)
2. There were no significant treatment–control group differences for the quarterly earn-
ings of female youth and male youth nonarrestees. For male youth arrestees, there was
a great discrepancy between survey data and data using unemployment wage records.
The former suggests significantly less earnings for the treatment group. The wage
record data suggest no significant treatment–control group difference.
3. JTPA had a significant positive effect on the attainment of a high school credential for
female youth (7.7 percentage points more treatment group females, compared to their
control group counterparts, had a high school diploma or GED 30 months after ran-
dom assignment), but not for male youth.
4. No significant treatment–control group differences were found in welfare receipt for
male or female youth.
Time Trend in Findings. There was a gradual increase in the earnings of all adult
participants and youth nonarrestees — treatment and control group — over time.
Information Sources
Larry Orr, Howard Bloom, Stephen Bell, Fred Doolittle, Winston Lin, and George Cave.
Does Training for the Disadvantaged Work? Evidence from the National JTPA Study
(Washington, D.C.: The Urban Institute Press, 1996).
James Kemple, Fred Doolittle, and John Wallace. The National JTPA Study: Site
Characteristics and Participation Patterns (New York: MDRC, 1993).
Appendix B
Assignment Rates for Employment

and Training Activities
103
Figure B.1
NATIONAL EVALUATION OF WELFARE-TO-WORK STRATEGIES

Assignment Patterns Within a Two-Year Follow-Up Period, by Site:
Labor Force Attachment Approach
Activities to which individuals were initially assigned:

ATLANTA GRAND RAPIDS RIVERSIDE
Basic education 1%
Basic education 4%
College 1% College 1%
Vocational training 3% College 6%
Vocational training 3% Vocational training 3%
Work experience 1%
Employment 1% Employment 6%
Employment 1%
No assignment 1% No assignment 21%
No assignment 5%
Job search 89% Job search 86% Job search
68%
NOTES: The designated activities are those to which individuals were initially assigned or in which they were allowed to continue if they had begun the activity prior to entering the
program. Numbers may not add up to 100% because of rounding.
104
Figure B.2

Assignment Patterns Within a Two-Year Follow-Up Period, by Site:
Human Capital Development Approach
Activities to which individuals were initially assigned:
ATLANTA GRAND RAPIDS RIVERSIDEa
Employment 4%
Employment 2%
No assignment 12% No assignment 15%b No assignment Job search 2%
Job search 4% Job search 8% 24%
Basic education 57%
Work Work
experience experience
4% 9%
Appendix B: Assignment Rates for Employment and Training Activities
Vocational
Vocational training 21%
training 34%
Basic
education 40% Basic education 36% Vocational training 3%
College 2% College 9% Employment 14%
NOTES: The designated activities are those to which individuals were initially assigned or in which they were allowed to continue if they had begun the activity prior to entering the
program. Numbers may not add up to 100% because of rounding.
a. Includes only individuals without a high school diploma or GED.
105
b. Includes many individuals who were initially assigned to a formal assessment and received no further assignments.
Appendix C
Average Length of Stay in Employment

and Training Activities
107
Table C.1

Average Length of Stay in Employment and Training Activities
Within Two Years After Orientation
Atlanta Grand Rapids Riverside
Activity LFA HCD LFA HCD LFA HCDa
Average number of months

of participation in:
Job search 3.1 3.1 2.2 2.4 2.3 2.4
Basic education 7.8 10.7 5.4 9.9 5.6 6.9
Post-secondary education 17.5 9.8 10.5 9.3 12.9 0.0
Vocational training 7.1 7.6 6.6 5.8 0.0 7.8
Work experience 6.6 8.6 4.8 3.5 5.7 0.0
SOURCES: Unpublished data based on MDRC calculations from a survey of clients administered approximately two years
after orientation, adjusted using MDRC-collected case file data from the programs.
NOTES: As discussed in chapter 4, the Labor Force Attachment (LFA) strategy emphasized placing people into jobs quickly,
whereas the Human Capital Development (HCD) strategy emphasized education and training as a precursor to employ-
ment.
a. The HCD sample for Riverside included only individuals without a high school diploma or GED.
Appendix C: Average Length of Stay in Employment and Training Activities 109
Table C.2
CALIFORNIA’S GREATER AVENUES FOR INDEPENDENCE (GAIN) PROGRAM

Two to Three Years After Orientation
County
Activity Alameda Butte Los Angeles Riverside San Diego Tulare

Average number of months
of participation in:
Job search 1.6 n/a 1.7 1.9 2.2 1.9
Adult Basic Education /
General Educational
Development
(GED) preparation 7.9 n/a 9.9 4.6 5.0 9.0
English as a Second Language 12.5 n/a 8.1 4.6 5.0 10.3
Vocational training or
post-secondary education 10.8 n/a 9.5 8.3 8.6 9.2
Unpaid work experience 3.3 n/a __ __ 1.5 2.9
SOURCE: James Riccio, Daniel Friedlander, and Stephen Freedman, GAIN: Benefits, Costs, and Three-Year Impacts of a
Welfare-to-Work Program (New York: MDRC, 1994).
NOTES: Results shown are for single parents in the experimental (program) group who started specified activities within
two to three years after orientation. Estimates include participation in GAIN and non-GAIN activities.
Where data are not available, n/a is used.
Table C.3
FLORIDA’S PROJECT INDEPENDENCE

Within Two Years After Random Assignment
Activity Months
Average number of months of participation in:a
Any independent job search or job clubb 2.5

Independent job search 2.2
Job club 1.6
Any education or trainingb 9.0

Adult Basic Education / General Educational Development (GED) preparation 5.9
English as a Second Language 5.7
Vocational training or post-secondary education 8.8
On-the-job-training 6.1
Average months of participationb 6.7
SOURCE: James Kemple, Veronica Fellerath, and Daniel Friedlander, Florida’s Project Independence: Benefits, Costs, and Two-
Year Impacts of Florida’s JOBS Program (New York: MDRC, 1995).
NOTES: The results shown include participation in Project Independence and non-Project Independence activities.
a. Estimates were based only on individuals who participated in the specified activities.
b. Individuals could participate in more than one activity during the follow-up period; therefore, the sum of months in
specific activities exceeds the number of months in the category.
Appendix D
Participation Rates for Employment and

Training Activities
• Minnesota Family Investment Program (MFIP)
NOTE:
The definition of participation rate presented in chapter 5 is somewhat different from the defini-
tion used in the following tables, which are extracted from MDRC research results. However, the
definition used in the tables is very similar to the one used in chapter 3. In chapter 5,“participa-
tion rate” is defined as the percentage of cases assigned to a program activity who participate in
the activity. In the following tables, as well as in chapter 3,“participation rate” is defined as the
percentage of cases referred to the program (not to a program activity) who participate in an
activity. In case this appendix is used to cost out not-yet-implemented programs, as described in
chapter 5, note that the values presented in the following tables approximate the assignment rate
(step 3B) times the participation rate (step 4B) for cases that were referred to the programs.
111
Table D.1

Participation Rates for Employment and Training Activities
Within Two Years After Orientation
Atlanta Grand Rapids Riverside
Activity LFA HCD LFA HCD LFA HCDa
Of those who were referred

to the program, percentage
who ever participated in:
Job search 62.0% 13.0% 46.9% 16.7% 45.0% 33.6%
Basic education 14.0 33.1 11.7 29.7 1.1 54.8
Post-secondary education 1.3 1.2 16.5 21.4 2.4 0.0
Vocational training 9.3 24.6 9.4 16.1 0.0 4.3
Work experience 14.4 7.0 8.8 12.0 1.7 0.0

NOTES: As discussed in chapter 4, the Labor Force Attachment (LFA) strategy emphasized placing people into jobs quickly,
whereas the Human Capital Development (HCD) strategy emphasized education and training as a precursor to employ-
ment.
Appendix D: Participation Rates for Employment and Training Activities 113
Table D.2

Two to Three Years After Orientation
County
Activity Alameda Los Angeles Riverside San Diego Tulare

Ever participated in:
Job search 32.2% 14.0% 38.0% 34.0% 24.2%
Adult Basic Education /
General Educational
Development
(GED) preparation 42.3 27.2 20.7 18.1 36.9
English as a Second Language 3.2 13.3 6.7 5.2 6.7
post-secondary education 28.4 13.5 26.8 34.8 28.6
Unpaid work experience 2.4 0.0 0.0 2.0 0.7
On-the-job-training 0.0 0.0 0.8 3.7 0.3
NOTES: For Butte County, the program’s sixth site, no participation data were available.
Results shown are for single parents in the experimental (program) group only. Estimates include participation in GAIN and
non-GAIN activities.
Table D.3
MINNESOTA FAMILY INVESTMENT PROGRAM (MFIP)

Participation Rates for Employment and Training Activities, in Urban Counties,
Within One Year After Random Assignment
MFIP MFIP
Full Financial
Activity Program Incentives Onlya

Career workshop 29.6% 21.2%
b
Job search 41.4 11.5
Job search class 19.9 8.7
Job search club 12.8 2.5
Individual job search 29.0 3.8
c
d
English as a Second Language 2.8 3.8
Other 2.7 1.5
Ever participated in any employment-related activityb 58.7 44.9
SOURCE: MDRC calculations from a survey of clients administered approximately one year after random assignment, as
reported in Cynthia Miller, Virginia Knox, Patricia Auspos, Jo Anna Hunter-Manns, and Alan Orenstein, Making Welfare
Work and Work Pay: Implementation and 18-Month Impacts of the Minnesota Family Investment Program (New York: MDRC,
1997).
NOTES: a. For research purposes, some welfare recipients were subject only to MFIP’s financial incentives and not to its
mandatory participation in employment and training activities by long-term recipients (although they could volunteer for
such services). This group is labeled “MFIP Financial Incentives Only.” The sample includes single-parent long-term welfare
recipients only.
b. Individuals could participate in more than one activity during the follow-up period; therefore, the sum of percentages in
specific activities exceeds the category percentage.
c.Post-secondary education is defined as courses for college credit at a two-year or four-year college.
d. Vocational training is defined as training for a specific job, trade, or occupation that does not lead to college credit. It does
not include on-the-job training or unpaid work experience.
Appendix D: Participation Rates for Employment and Training Activities 115
Table D.4

Within Two Years After Random Assignment
Activity Percentage
Any independent job search or job cluba 42.7%

Independent job search 41.0
Job club 13.2
Any education or traininga,b 41.5

Adult Basic Education / General Educational Development (GED) preparation 12.1
English as a Second Language 4.6
Vocational training or post-secondary education 29.2
On-the-job-training 4.0
Ever participated in any employment-related activitya 63.9
SOURCE: James Kemple, Veronica Fellerath, and Daniel Friedlander, Florida’s Project Independence: Benefits, Costs, and Two-
Year Impacts of Florida’s JOBS Program (New York: MDRC, 1995).
NOTES: The results shown include participation in Project Independence and non-Project Independence activities.
a. Individuals could participate in more than one activity during the follow-up period; therefore, the sum of percentages in
specific activities exceeds the category percentage.
b. Includes self-initiated basic education, vocational training, or post-secondary education.
Appendix E
Unit Costs for Employment and

Training Activities
• National Job Training Partnership Act (JTPA) Study
• JOBSTART Demonstration
117
Table E.1

Unit Costs for Employment and Training Activities
(in 1993 Dollars)
Labor Force Attachment Approach (LFA)

Welfare Dept. Unit Costa Non-Welfare Dept. Unit Cost
Activity Average per Average per Average per
and Site Participant-Month Participant-Month ADAb
Atlanta $ 65c — —
Grand Rapids 16c — —
Riverside 79c — —
Formal assessment
Atlanta — — —
Grand Rapids 355 — —
Job searchd
Atlanta $374 $ 55 —
Grand Rapids 233 413g —
Riverside 682 — —
Basic educatione,f
Atlanta $104 — $1,502
Grand Rapids 119 — 3,008
Riverside 229 — 2,100
Post-secondary educatione
Atlanta $ 66 — $4,263
Vocational traininge
Atlanta $138 — $3,449
Work experience
Atlanta $150 — —
(continued)
Appendix E: Unit Costs for Employment and Training Activities 119
Table E.1 (continued)
Human Capital Development Approach (HCD)

Welfare Dept. Unit Costa Non-Welfare Dept. Unit Cost
Activity Average per Average per Average per
and Site Participant-Month Participant-Month ADAb
Atlanta $ 65c — —
Grand Rapids 16c — —
Formal assessment
Atlanta — — —
Job searchd
Atlanta $416 $ 55 —
Grand Rapids 233 377g —
Basic educatione,f
Atlanta $ 89 — $1,460
Post secondary educatione

Atlanta $ 85 — $3,806
Vocational traininge
Atlanta $126 — $3,418
Work experience
Atlanta $182 — —

NOTES: Welfare department costs include the costs of monitoring and sanctioning non-compliant participants.
The following calculation will adjust for inflation and express 1993 dollars in 1997 dollars. Adjustment factor: 1.0984 (based
on the GDP deflator, which is based on the U.S. government fiscal year). 1997 dollars = 1993 dollars x adjustment factor. For
example, orientation in Atlanta (LFA): $65 x 1.0984 = $71.40. Expressed in 1997 dollars, one session of orientation cost $71.
(continued)

a. In general, Riverside’s welfare department unit costs were the highest of the three, in part because Riverside had relatively high
overhead costs and job developers on site, who canvassed the local job market for employment opportunities for participants.
b. One ADA refers to one unit of Average Daily Attendance, an attendance measure used by California community colleges and
adult schools, and is defined as a block of 525 hours of attendance. For comparison reasons, educational activities of all three
sites are expressed per ADA.
c. Cost per session for one participant.
d. Job search costs per participant incurred by the welfare departments tended to be higher than the education and training unit
costs because the welfare department typically paid for case management costs for clients enrolled in job search activities as well
as for the direct expenses of providing job search services. The exception was Grand Rapids, where job search services were pro-
vided and paid for by the local community education center.
e. The education and vocational training unit costs incurred by the welfare department reflect case management costs only. Non-
welfare agencies paid for the education and training services.
f. Riverside made incentive payments to the basic education providers.
g. The estimated unit cost to Grand Rapids non-welfare agencies was $921 per participant.
Table E.2

(in 1993 Dollars)
Welfare Dept. Unit Cost Non-Welfare Dept. Unit Cost

Activity Average per Average per
and County Participant-Month ADAa
Orientationb
Alameda $ 515c __
Butte 345c __
Los Angeles 520c __
Riverside 407c __
San Diego 295c __
Tulare 342c __
Job search activitiesd

Alameda $1,120 __
Butte 607 __
Los Angeles 556 __
Riverside 568 __
San Diego 537 __
Tulare 667 __
Adult Basic Education / General

Educational Development (GED) /
English as a Second Languagee
Alameda $ 180 $2,109
Butte 112 2,106
Los Angeles 288 2,131
Riverside 141 2,160
San Diego 243 2,153
Tulare 94 2,101
post-secondary educationf
Alameda $ 104 $3,151
Butte 192 3,011
Los Angeles 162 3,257
Riverside 101 3,522
San Diego 82 3,252
Tulare 77 3,121
(continued)

example, orientation in Alameda: $515 x 1.0984 = $565.68. Expressed in 1997 dollars, one session of orientation cost $566.
a. One ADA refers to one unit of Average Daily Attendance, an attendance measure used by California community colleges
and adult schools, and is defined as a block of 525 hours of attendance. For comparison reasons, educational activities of all
three sites are expressed per ADA.
b. Los Angeles held a motivational training seminar, which was incorporated into its day-and-a-half-long orientation
session. Alameda had an on-site day care center at the GAIN office for registrants attending a GAIN orientation and meeting
with staff.
c. Cost per session for one participant.
d. Job search activities had the highest welfare department unit costs, reflecting the fact that they included the costs of both
case management and providing the job search activities, such as job club sessions.
e. San Diego had extra on-site counselors in its learning labs for Adult Basic Education /GED and English as a Second
Language participants.
f. Butte and Tulare had extra on-site counselors at some community colleges. Riverside made incentive payments to basic
education providers.
Table E.3

(in 1993 Dollars)
Welfare Department Non-Welfare Department

Unit Cost Unit Cost
Average per Average per Average per
Activity Participant-Month Participant-Month FTEa
Orientation $ 69b n/a n/a
Formal assessment 112 n/a n/a
Independent job search 72 $ 72 n/a
Job club 226 226 n/a
Adult Basic Education / General
Educational Development
(GED) preparation 76 n/a $2,688
English as a Second Language 76 n/a 2,688
post-secondary education 21 n/a 4,768c
On-the-job training 231 231 n/a
Unpaid work experience 30 30 n/a
SOURCE: James Kemple, Veronica Fellerath, and Daniel Friedlander, Florida’s Project Independence: Benefits, Costs, and
Two-Year Impacts of Florida’s JOBS Program (New York: MDRC, 1995).
The following calculation will adjust for inflation and will express 1993 dollars in 1997 dollars. Adjustment factor: 1.0984
(based on the GDP deflator, which is based on the U.S. government fiscal year). 1997 dollars = 1993 dollars x adjustment
factor. For example, orientation: $69 x 1.0984 = $75.79. Expressed in 1997 dollars, one session of orientation cost $76.
Where data are not available, n/a is used.
a. One FTE (full-time equivalent student) represents the total number of scheduled hours for a student attending full-time
for one academic year. For adult schools, vocational training centers, and non-credit courses at community colleges, one
FTE equals 900 scheduled hours of class, which is equivalent to 180 days of five one-hour classes per day..
b. Cost per participant per session.
c. The average cost per FTE is a weighted average of the costs per FTE of vocational training centers and community
colleges.
Table E.4
NATIONAL JOB TRAINING PARTNERSHIP ACT (JTPA) STUDY

(in 1988 Dollars)
Service Provider Unit Cost:

Average per Participant-Day
Activity Adults Youth
All sitesa
Orientation/intake/assessment $159b $166b
Classroom training in occupational skills 16 16
On-the-job-training 19 12
Job search 4 6
Adult Basic Education 17 11
Work experience 25 9
SOURCES: Unpublished data from the National JTPA Study, Abt Associates. Also, Larry L. Orr, Howard S. Bloom, Stephen
H. Bell, Fred Doolittle, Winston Lin, and George Cave, Does Job Training for the Disadvantaged Work? Evidence from the
National JTPA Study (Washington, D.C.: The Urban Institute Press, 1996).
NOTES: The figures shown are the cost per day enrolled in the activity, which may not equal the cost per day of active
participation in the activity.
example, orientation for adults: $159 x 1.3138 = $208.89. Expressed in 1997 dollars, one session of orientation cost $209.
a. Weighted averages of all 16 sites.
b. Cost per participant per session.
Table E.5
JOBSTART DEMONSTRATION
(in 1986 Dollars)

Average per
Activity and Site Month of Participation
Orientation
Allentown, NY $ 33
Atlanta Job Corps 195
Basic Skills Academy, NYC 439
Capitol Region Education Council (CREC), Hartford 291
Center for Employment Training (CET), San Jose 56
Chicago Commons n/a
Connelley, Pittsburgh 14
East LA Skills Center n/a
Emily Griffith Opportunity School (EGOS), Denver 62
Phoenix Job Corps 178
SER Corpus Christi, TX 64
El Centro, Dallas n/a
LA Job Corps 364
Basic educationa
Allentown, NY $ 113
CREC, Hartford 525
CET, San Jose 16
Chicago Commons 1,724b
Connelley, Pittsburgh 661b
East LA Skills Center 15c
EGOS, Denver 63
SER Corpus Christi, TX 642b
El Centro, Dallas 418
LA Job Corps 88
Life skills education
Allentown, NY $ 110
CREC, Hartford n/a
CET, San Jose n/a
Chicago Commons 35b
Connelley, Pittsburgh n/a
East LA Skills Center n/a
EGOS, Denver n/a
SER Corpus Christi, TX n/a
LA Job Corps 60
(continued)

Average per
Activity and Site Month of Participation
Skills trainingd
Allentown, NY $ 529b
Basic Skills Academy, NYC 1,965b
CREC, Hartford 2,135b
CET, San Jose 187
Chicago Commons 3,187b
East LA Skills Center 315
EGOS, Denver 59
LA Job Corps 217
Job placement
Allentown, NY $ 48
Basic Skills Academy, NYC n/a
CREC, Hartford n/a
CET, San Jose 56
Chicago Commons 285b
East LA Skills Center 7
EGOS, Denver 3
LA Job Corps 534b
SOURCES: Unpublished data from the JOBSTART Demonstration. Also, George Cave and Fred Doolittle, Assessing
JOBSTART: Interim Impacts of a Program for School Dropouts (New York: MDRC, 1991).
NOTES: Sites had different enrollment levels. For example, if the number of participants “on board” a program in a typical
month is high relative to the number of program instructors, the total monthly instructional costs (and the corresponding
overhead expenditures) will be spread over many people, lowering the average unit cost. Differences in wage scales further
explain some of the variation in unit costs. The average hourly wage paid to instructors at SER Corpus Christi was about half
the hourly rate received by teachers at the East LA Skills Center. EGOS, in Denver, a large public vocational school with more
than 15,000 students, was able to spread personnel costs for instructors and overhead expenditures over many students. In
contrast, the Basic Skills Academy, in New York City, had high unit costs because it enrolled only about half the number of
students the school had the capacity to serve at any time.
example, orientation in Allentown: $33 x 1.3988 = $46.16. Expressed in 1997 dollars, one session of orientation cost $46.
Where data are not applicable, n/a is used.
(continued)
a. Education costs were especially low at CET, San Jose, in part because most of the hours that program participants spent
in that site were in training classes, which included work on basic education skills. (Basic education that occurred in the
context of occupational skills training was counted as skills training.)
b. Per person served.
c. Per hour.
d. Chicago Commons added a separate education class specifically for JOBSTART participants. Chicago Commons
assigned two instructors to all training classes, an unusual practice among the JOBSTART sites, and this raised its unit cost
per participant. Additionally, training classes operated on a fixed-cycle basis (dropouts within a cycle were usually not
replaced with other students).
Appendix F
Program Effects on AFDC Receipt
129
Table F.1

Program Effects on AFDC Receipt Within a Two-Year Follow-Up Period
Atlanta Grand Rapids Riverside Atlanta Grand Rapids Riversidea

Ever Received Any Program Control Program Control Program Control Program Control Program Control Program Control
AFDC Payments (%) Group Group Group Group Group Group Group Group Group Group Group Group
Year 1 or 2 97.7% 98.1% 95.7% 97.4% 93.3% 93.4% 97.3% 98.1% 97.1% 97.4% 93.8% 93.9%
Year 1 97.4 97.8 95.1 97.1 92.7 92.9 97.2 97.8 96.9 97.1 93.1 93.4
Year 2 82.0 86.1 74.5 79.9 62.8 68.7 83.3 86.1 77.1 79.9 69.2 71.8
Quarter 1 98.3 98.3 97.7 98.1 96.4 96.5 98.2 98.3 97.7 98.1 96.4 96.6
Quarter 2 96.7 97.7 92.6 95.7 91.5 91.7 96.7 97.7 95.7 95.7 92.0 92.5
Quarter 3 91.7 93.6 83.5 89.8 80.5 83.8 92.9 93.6 88.1 89.8 81.4 86.2
Quarter 4 86.9 88.2 76.8 84.2 71.2 75.4 88.0 88.2 81.1 84.2 73.7 78.1
Quarter 5 81.9 85.1 71.7 79.3 63.6 69.6 83.4 85.1 76.3 79.3 68.1 72.4
Quarter 6 79.0 82.9 67.7 76.8 58.8 65.2 80.2 82.9 71.4 76.8 65.3 68.2
Quarter 7 75.3 79.6 63.5 71.6 55.3 61.4 77.0 79.6 67.2 71.6 61.1 64.2
Quarter 8 71.5 77.7 61.9 68.9 52.4 58.7 73.4 77.7 62.4 68.9 58.0 61.8
Quarter 9 68.4 74.8 58.1 65.1 50.0 55.9 70.7 74.8 58.8 65.1 54.9 58.9
SOURCE: Gayle Hamilton, Thomas Brock, Mary Farrell, Daniel Friedlander, and Kristen Harknett, National Evaluation of Welfare-to-Work Strategies: Evaluating Two Welfare-to-Work
Program Approaches — Two-Year Findings on the Labor Force Attachment and Human Capital Development Programs in Three Sites (Washington, D.C.: U.S. Department of Health and
Human Services and U.S. Department of Education, 1997).
NOTES: For all measures, the quarter of random assignment refers to the calendar quarter in which random assignment occurred. Because quarter 1, the quarter of random assignment,
may contain some earnings and AFDC payments from the period prior to random assignment, it is excluded from follow-up measures. Thus,“year 1” is quarters 2 through 5,“year 2” is
quarters 6 through 9, and so forth.
130
Glossary
Aggregate program costs. Costs summed over all cases accepted into a voluntary E&T program
or referred to a mandatory E&T program.
Assignment rate. The percentage of cases accepted by or referred to a program who are assigned
to a particular program activity or receive a particular program support service.
Average costs. Costs per accepted or referred case.
Benefits. Effects of a program or policy that makes individuals better off (for example, an
increase in earnings as a result of participating in an E&T program).
Comparison group. Cases that do not have access to the services provided by a program, but are
otherwise eligible for the program. In many evaluations, these cases are compared to cases that
are accepted or referred to the program.
Control group. Cases that, for purposes of evaluation, are randomly assigned to a group that is
not allowed to receive program services, but would otherwise be eligible. These cases are com-
pared to cases that are randomly assigned to a group that is allowed to receive program services.
Cost-benefit analysis. A comparison of a program’s costs with its benefits.
Costs. Expenditures required to operate a program, including expenditures on staff; purchases of
goods, services, office space, and equipment; vendor payments; and purchases of day care and
transportation assistance.
Discounting. A procedure that uses an interest rate to compute the present value of a future
stream of benefits and costs. The interest rate is used to adjust dollars that will be paid out
or received in the future because they are valued less than dollars paid out or received in the
present.
Employment subsidies. Payments made to employers who hire program participants.
Gross costs. The total costs necessary to run a program.
Inflation. A general increase in the price level over time.
Joint costs. Costs that result from using resources for multiple purposes.
Mandatory E&T program. An employment and training program in which people are required
to participate in exchange for receiving benefits (e.g., welfare or unemployment compensation).
Marginal costs. Additional costs that result from serving one more case in a program or the cost
savings that would result from serving one less case.
Net benefits. The benefits of a program less the costs of the program.
131
Net costs. Gross costs incurred by the program less cost savings attributable to the program.
Nonprogram employment and training costs. Employment and training costs that occur out-
side the program being analyzed.
No-show. A case that is assigned to a program activity, but does not show up.
Overhead costs. Costs not counted as part of staff personnel costs. Overhead costs may include
costs associated with support staff, such as secretaries, and supervisors; fringe benefits; and the
costs of various physical resources, such as computer systems, furniture, physical facilities, and
telephones.
Overhead rate. Total annual expenditures divided by staff personnel costs.
Participation rate. For an ongoing program, the participation rate is the percentage of cases
accepted by or referred to a program who participate in a particular program activity or receive a
particular support service. For a not-yet-implemented program, the participation rate is the per-
centage of cases who are assigned to a particular program activity and actually participate in the
activity.
Performance-based contract. Contract which states that all or, more often, part of the fees paid
to a contractor depend on whether (and sometimes the extent to which) the contractor exceeds
or falls below certain goals or standards.
Personnel costs. Costs that accrue to the line staff who are directly involved in various E&T
activities.
Present value. The value today of a future stream of benefits or costs.
Price index. A measure of the changes in the price level over time that result from inflation.
Program agency. A government agency with overall responsibility for operating an E&T pro-
gram.
Program group. The portion of a research sample comprising the cases accepted by or referred
to an E&T program (cf. comparison group, control group).
Random assignment. An evaluation technique whereby people eligible for a program are
assigned randomly to either of two groups — one that is subject to the program and allowed to
receive program services (the treatment or program group) and one that is neither subject to the
program nor allowed to receive program services (the control group). The purpose of random
assignment is to ensure that any differences between the two groups, other than being subject to
the program and receiving program services, is due to chance alone.
Regular costs. Costs that are essential to operating a program.
Sample. A representative subgroup of the cases that are eligible to participate in a program.
Special costs. Costs that are not essential to operating a program.
Start-up costs. The costs required to implement a new program (such as devising curriculum,
writing regulations, developing computer software, and so forth) and incurred in learning how to
operate a new program.
Steady state. A period of relative stability in the life of a program.
Straight-line depreciation. A procedure that allows the cost of durable goods to be spread
equally over their years of useful life.
Glossary 133
Support services. Payments made to or for program participants to help offset the costs of day
care, transportation, and the like that they would otherwise bear as a result of participating in the
program, or the direct provision of these services.
Time study. The use of a survey instrument to allocate staff time among program activities.
Treatment group. See program group.
Unit costs. Costs of providing a particular program component to one case over a specified time
period.
Voluntary E&T program. An employment and training program to which eligible individuals
apply. Some of those who apply may not be accepted into the program.
References and Further Reading
Boardman, Anthony E., David H. Greenberg, Aidan R. Vining, and David L. Weimer, Cost-Benefit
Analysis: Concepts and Practice (Upper Saddle River, N.J.: Prentice-Hall, 1996).
Cave, George, and Fred Doolittle, Assessing JOBSTART: Interim Impacts of a Program for School
Dropouts (New York: MDRC, 1991).
Cave, George, Fred Doolittle, Hans Bos, and Cyril Toussaint, JOBSTART: Final Report on a
Program for School Dropouts (New York: MDRC, 1993).
Congressional Budget Office, Work and Welfare: The Family Support Act of 1988 (Washington,
D.C., 1989).
Freedman, Stephen, Daniel Friedlander, Winston Lin, and Amanda Schweder, The GAIN
Evaluation. Working Paper 96.1, Five-Year Impacts on Employment, Earnings, and AFDC Receipt
(New York: MDRC, 1996).
Friedlander, Daniel, David H. Greenberg, and Philip K. Robins,“Evaluating Government Training
Programs for the Economically Disadvantaged,” Journal of Economic Literature (December, 1997:
pp. 1809-1855).
Goldman, Barbara, Daniel Friedlander, David Long, with Marjorie Erickson, and Judith Gueron,
California: Final Report on the San Diego Job Search and Work Experience Demonstration (New
York: MDRC, 1986).
Greenberg, David, and Mark Shroder, The Digest of Social Experiments, Second Edition
(Washington, D.C.: The Urban Institute Press, 1997).
Hamilton, Gayle, Thomas Brock, Mary Farrell, Daniel Friedlander, and Kristen Harknett,
National Evaluation of Welfare-to-Work Strategies: Evaluating Two Welfare-to-Work Program
Approaches — Two-Year Findings on the Labor Force Attachment and Human Capital Development
Programs in Three Site (Washington, D.C.: U.S. Department of Health and Human Services and
U.S. Department of Education, 1997).
Jones, Kumen H., Jean B. Price, Michael L. Werner, and Martha S. Doran, Introduction to
Financial Accounting: A User Perspective (Englewood Cliffs, N.J.: Prentice-Hall, 1996).
Kemple, James, Fred Doolittle, and John Wallace, The National JTPA Study: Site Characteristics
and Participation Patterns (New York: MDRC, 1993).
Kemple, James, Veronica Fellerath, and Daniel Friedlander, Florida’s Project Independence:
Benefits, Costs, and Two-Year Impacts of Florida’s JOBS Program (New York: MDRC, 1995).
135
Meyers, Roy T., ed., The Handbook of Government Budgeting (San Francisco: Jossey-Bass.
Forthcoming, 1998).
Mikesell, John L., Fiscal Administration: Analysis and Applications for the Public Sector, Fourth
Edition (Belmont, CA: Wadsworth, 1994).
Miller, Cynthia, Virginia Knox, Patricia Auspos, Jo Anna Hunter-Manns, and Alan Orenstein,
Making Welfare Work and Work Pay: Implementation and 18-Month Impacts of the Minnesota
Family Investment Program (New York: MDRC, 1997).
Orr, Larry L., Howard S. Bloom, Stephen H. Bell, Fred Doolittle, Winston Lin, and George Cave,
Does Training for the Disadvantaged Work? Evidence from the National JTPA Study (Washington,
D.C.: The Urban Institute Press, 1996).
Powdar, Juliet Carol, The Operating Budget: A Guide for Smaller Governments (Chicago:
Government Finance Officers Association, 1996).
Riccio, James, Daniel Friedlander, and Stephen Freedman, GAIN: Benefits, Costs, and Three-Year
Impacts of a Welfare-to-Work Program (New York: MDRC, 1994).
Scrivener, Susan, Gayle Hamilton, Mary Farrell, Stephen Freedman, Daniel Friedlander, Marisa
Mitchell, Jodi Nudelman, and Christine Schwartz, National Evaluation of Welfare-to-Work
Strategies: Implementation, Participation Patterns, Costs, and Two-Year Impacts of the Portland
(Oregon) Welfare-to-Work Program (Washington, D.C.: U.S. Department of Health and Human
Services and U.S. Department of Education, 1998).
Zedlewski, Sheila, and Isabel V. Sawhill,“Assessing the Personal Responsibility Act,” Welfare
Reform Briefs, No. 5 (Washington, D.C.: The Urban Institute Press, 1995).
Recent Publications on MDRC Projects
Employment and Community Initiatives
Connections to Work Project

A study of local efforts to increase competition in the choice of providers of employment
services for welfare recipients and other low-income populations. The project also provides
assistance to cutting-edge local initiatives aimed at helping such people access and secure jobs.
Tulsa’s IndEx Program: A Business-Led Initiative for Welfare Reform and Economic Development.
1997. Maria Buck.
Washington Works: Sustaining a Vision of Welfare Reform Based on Personal Change, Work
Preparation, and Employer Involvement. 1998. Susan Gooden.
Jobs-Plus Initiative
A multi-site effort to greatly increase employment among public housing residents.
A Research Framework for Evaluating Jobs-Plus, a Saturation and Place-Based Employment
Initiative for Public Housing Residents. 1998. James A. Riccio.
Section 3 Public Housing Study

An examination of the effectiveness of Section 3 of the 1968 Housing and Urban Development
Act in affording employment opportunities for public housing residents.
Lessons from the Field on the Implementation of Section 3 (U.S. Department of Housing and
Urban Development). 1996. Maxine Bailey, Suzanne Lynn.
Canada’s Earnings Supplement Project

A test of an innovative financial incentive intended to expedite the reemployment of displaced
workers and encourage full-year work by seasonal or part-year workers, thereby also reducing
receipt of Unemployment Insurance.
Implementing the Earnings Supplement Project: A Test of a Re-employment Incentive (Social
Research and Demonstration Corporation). 1997. Howard Bloom, Barbara Fink, Susanna
Lui-Gurr, Wendy Bancroft, Doug Tattrie.
NOTE: For works not published by MDRC, the publisher’s name is shown in parentheses. A complete publications
list is available from MDRC and on its Web site (www.mdrc.org).
137
Reforming Welfare and Making Work Pay
ReWORKing Welfare: Technical Assistance for States and Localities

A multifaceted effort to assist states and localities in designing and implementing their welfare
reform programs. The project includes a series of “how-to” guides, conferences, briefings, and
customized, in-depth technical assistance.
After AFDC: Welfare-to-Work Choices and Challenges for States. 1997. Dan Bloom.
Changing to a Work First Strategy: Lessons from Los Angeles County’s GAIN Program for Welfare
Recipients. 1997. Evan Weissman.
Work First: How to Implement an Employment-Focused Approach to Welfare Reform. 1997.
Amy Brown.
Business Partnerships: How to Involve Employers in Welfare Reform. 1998. Amy Brown, Maria
Buck, Erik Skinner.
Time Limits
Connecticut’s Jobs First Program
An evaluation of Connecticut’s time-limited welfare program, which includes financial work
incentives and requirements to participate in employment-related services aimed at rapid job
placement. This study provides some of the earliest information on the effects of time limits in
major urban areas.
Jobs First: Early Implementation of Connecticut’s Welfare Reform Initiative. 1998. Dan Bloom,
Mary Andes, Claudia Nicholson.
Florida’s Family Transition Program

An evaluation of Florida’s initial time-limited welfare program, which includes services,
requirements, and financial work incentives intended to reduce long-term welfare receipt and
help welfare recipients find and keep jobs.
The Family Transition Program: Implementation and Early Impacts of Florida’s Initial
Time-Limited Welfare Program. 1997. Dan Bloom, James J. Kemple, Robin Rogers-Dillon.
The Family Transition Program: Implementation and Interim Impacts of Florida’s Initial Time-
Limited Welfare Program. 1998. Dan Bloom, Mary Farrell, James J. Kemple, Nandita Verma.
Cross-State Study of Time-Limited Welfare

An examination of the implementation of some of the first state-initiated time-limited welfare
programs.
Implementing Time-Limited Welfare: Early Experiences in Three States. 1995. Dan Bloom, David
Butler.
The View from the Field: As Time Limits Approach, Welfare Recipients and Staff Talk About Their
Attitudes and Expectations. 1997. Amy Brown, Dan Bloom, David Butler.
Financial Incentives
Minnesota Family Investment Program
An evaluation of Minnesota’s welfare reform initiative, which aims to encourage work, alleviate
poverty, and reduce welfare dependence.
MFIP: An Early Report on Minnesota’s Approach to Welfare Reform. 1995. Virginia Knox, Amy
Brown, Winston Lin.
Recent Publications on MDRC Projects 139
Making Welfare Work and Work Pay: Implementation and 18-Month Impacts of the Minnesota
Family Investment Program. 1997. Cynthia Miller, Virginia Knox, Patricia Auspos, Jo Anna
Hunter-Manns, Alan Orenstein.
New Hope Project

A test of a community-based, work-focused antipoverty program and welfare alternative operat-
ing in Milwaukee.
The New Hope Offer: Participants in the New Hope Demonstration Discuss Work, Family, and
Self-Sufficiency. 1996. Dudley Benoit.
Creating New Hope: Implementation of a Program to Reduce Poverty and Reform Welfare. 1997.
Thomas Brock, Fred Doolittle, Veronica Fellerath, Michael Wiseman.
Who Got New Hope? 1997. Michael Wiseman.
An Early Look at Community Service Jobs in the New Hope Demonstration. 1998. Susan Poglinco,
Julian Brash, Robert Granger.
Canada’s Self-Sufficiency Project

A test of the effectiveness of a temporary earnings supplement on the employment and welfare
receipt of public assistance recipients. Reports on the Self-Sufficiency Project are available from:
Social Research and Demonstration Corporation (SRDC), 275 Slater St., Suite 900, Ottawa,
Ontario K1P 5H9, Canada. Tel.: 613-237-4311; Fax: 613-237-5045. In the United States, the
reports are also available from MDRC.
Creating an Alternative to Welfare: First-Year Findings on the Implementation, Welfare Impacts, and
Costs of the Self-Sufficiency Project (Social Research and Demonstration Corporation [SRDC]).
1995. Tod Mijanovich, David Long.
The Struggle for Self-Sufficiency: Participants in the Self-Sufficiency Project Talk About Work,
Welfare, and Their Futures (SRDC). 1995. Wendy Bancroft, Sheila Currie Vernon.
Do Financial Incentives Encourage Welfare Recipients to Work? Initial 18-Month Findings from the
Self-Sufficiency Project (SRDC). 1996. David Card, Philip K. Robins.
When Work Pays Better Than Welfare: A Summary of the Self-Sufficiency Project’s Implementation,
Focus Group, and Initial 18-Month Impact Reports (SRDC). 1996.
How Important Are “Entry Effects” in Financial Incentive Programs for Welfare Recipients?
Experimental Evidence from the Self-Sufficiency Project (SRDC). 1997. David Card, Philip K.
Robins, Winston Lin.
Do Work Incentives Have Unintended Consequences? Measuring “Entry Effects” in the Self-
Sufficiency Project (SRDC). 1998. Gordon Berlin, Wendy Bancroft, David Card, Winston Lin,
Philip K. Robins.
Mandatory Welfare Employment Programs

National Evaluation of Welfare-to-Work Strategies
A large-scale study (formerly known as the JOBS Evaluation) of different strategies for moving
people from welfare to employment.
Adult Education for People on AFDC: A Synthesis of Research (U.S. Department of Education
[ED]/U.S. Department of Health and Human Services [HHS]). 1995. Edward Pauly.
Early Findings on Program Impacts in Three Sites (HHS/ED). 1995. Stephen Freedman, Daniel
Friedlander.
Five Years After: The Long-Term Effects of Welfare-to-Work Programs (Russell Sage Foundation).
1995. Daniel Friedlander, Gary Burtless.
Monthly Participation Rates in Three Sites and Factors Affecting Participation Levels in Welfare-to-
Work Programs (HHS/ED). 1995. Gayle Hamilton.
Evaluating Two Welfare-to-Work Program Approaches: Two-Year Findings on the Labor Force
Attachment and Human Capital Development Programs in Three Sites (HHS/ED). 1997. Gayle
Hamilton, Thomas Brock, Mary Farrell, Daniel Friedlander, Kristen Harknett.
Work First: How to Implement an Employment-Focused Approach to Welfare Reform. 1997. Amy
Brown.
Implementation, Participation Patterns, Costs and Two-Year Impacts of the Portland (Oregon)
Welfare-to-Work Program. (HHS/ED). 1998. Susan Scrivener, Gayle Hamilton, Mary Farrell,
Stephen Freedman, Daniel Friedlander, Marisa Mitchell, Jodi Nudelman, Christine Schwartz.
Los Angeles’s Jobs-First GAIN Program
An evaluation of Los Angeles’s refocused GAIN (welfare-to-work) program, which emphasizes
rapid employment. This is the first in-depth study of a full-scale “work first” program in one of
the nation’s largest urban areas.
The Los Angeles Jobs-First GAIN Evaluation: Preliminary Findings on Participation Patterns and
First-Year Impacts. 1998. Stephen Freedman, Marisa Mitchell, David Navarro. Working Paper.
Teen Parents on Welfare

Teenage Parent Programs: A Synthesis of the Long-Term Effects of the New Chance Demonstration,
Ohio’s Learning, Earning, and Parenting (LEAP) Program, and the Teenage Parent Demonstration
(TPD). 1998. Robert Granger, Rachel Cytron.
Ohio’s LEAP Program

An evaluation of Ohio’s Learning, Earning, and Parenting (LEAP) Program, which uses financial
incentives to encourage teenage parents on welfare to stay in or return to school.
LEAP: Three-Year Impacts of Ohio’s Welfare Initiative to Improve School Attendance Among
Teenage Parents. 1996. David Long, Judith M. Gueron, Robert G. Wood, Rebecca Fisher, Veronica
Fellerath.
LEAP: Final Report on Ohio’s Welfare Initiative to Improve School Attendance Among Teenage
Parents. 1997. Johannes Bos, Veronica Fellerath.
New Chance Demonstration

A test of a comprehensive program of services that seeks to improve the economic status and
general well-being of a group of highly disadvantaged young women and their children.
New Chance: Final Report on a Comprehensive Program for Young Mothers in Poverty and Their
Children. 1997. Janet Quint, Johannes Bos, Denise Polit.
Parenting Behavior in a Sample of Young Mothers in Poverty: Results of the New Chance
Observational Study. 1998. Martha Zaslow, Carolyn Eldred, editors.
Recent Publications on MDRC Projects 141
Focusing on Fathers
Parents’ Fair Share Demonstration
A demonstration for unemployed noncustodial parents (usually fathers) of children on welfare.
PFS aims to improve the men’s employment and earnings, reduce child poverty by increasing
child support payments, and assist the fathers in playing a broader constructive role in their
children’s lives.
Low-Income Parents and the Parents’ Fair Share Demonstration. 1996. Earl Johnson, Fred
Doolittle.
Working with Low-Income Cases: Lessons for the Child Support Enforcement System from Parents’
Fair Share. 1998. Fred Doolittle, Suzanne Lynn.
Other
Can They All Work? A Study of the Employment Potential of Welfare Recipients in a Welfare-to-
Work Program. 1995. James A. Riccio, Stephen Freedman.
Florida’s Project Independence: Benefits, Costs, and Two-Year Impacts of Florida’s JOBS Program.
1995. James J. Kemple, Daniel Friedlander, Veronica Fellerath.
From Welfare to Work Among Lone Parents in Britain: Lessons for America. 1996. James A. Riccio.
Education Reform
School-to-Work Project
A study of innovative programs that help students make the transition from school to work or
careers.
Home-Grown Lessons: Innovative Program Linking School and Work (Jossey-Bass Publishers).
1995. Edward Pauly, Hilary Kopp, Joshua Haimson.
Home-Grown Progress: The Evolution of Innovative School-to-Work Programs. 1997. Rachel A.
Pedraza, Edward Pauly, Hilary Kopp.
Career Academies
The largest and most comprehensive evaluation of a school-to-work initiative, this 10-site study
examines a promising approach to high school restructuring and the school-to-work transition.
Career Academies: Early Implementation Lessons from a 10-Site Evaluation. 1996. James J. Kemple,
JoAnn Leah Rock.
Career Academies: Communities of Support for Students and Teachers — Emerging Findings from a
10-Site Evaluation. 1997. James J. Kemple.
About MDRC
The Manpower Demonstration Research Corporation (MDRC) is a nonprofit social

policy research organization founded in 1974 and located in New York City and San
Francisco. Its mission is to design and rigorously field-test promising education and
employment-related programs aimed at improving the well-being of disadvantaged
adults and youth, and to provide policymakers and practitioners with reliable evidence
on the effectiveness of social programs. Through this work, and its technical assistance
to program administrators, MDRC seeks to enhance the quality of public policies and
programs. MDRC actively disseminates the results of its research through its publica-
tions and through interchanges with a broad audience of policymakers and practition-
ers; state, local, and federal officials; program planners and operators; the funding
community; educators; scholars; community and national organizations; the media;
and the general public.
Over the past two decades — working in partnership with more than forty states,
the federal government, scores of communities, and numerous private philanthropies
— MDRC has developed and studied more than three dozen promising social policy
initiatives.
16 East 34 Street
New York, New York 10016
(212) 532-3200
www.mdrc.org
88 Kearny Street, Suite 1800
San Francisco, California 94108
(415) 781-3800

Cost Analysis Step by Step

Uploaded by

Copyright:

Available Formats

Cost Analysis Step by Step

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cost Analysis Step by Step

Uploaded by

Copyright:

Available Formats

CONNECTIONS TO WORK

Cost Analysis Step by Step

Manpower Demonstration Research Corporation

ROBERT REISCHAUER, Chairman RUDOLPH G. PENNER

PAUL H. O’NEILL, Treasurer MARION O. SANDLER

Cost Analysis Step by Step

Manpower Demonstration Research Corporation

Dissemination of MDRC publications is also supported

The findings and conclusions presented in this guide

For information about MDRC, see our Web site: www.mdrc.org.

Chapter 1 Who This Guide Is For and How It Can Be Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1

Chapter 2 Basics of Conducting Cost Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Chapter 3 Costing Out Ongoing Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Chapter 4 A Real-World Cost Analysis of an Ongoing Program:

Chapter 6 Using Cost Information in Designing, Bidding, and Budgeting . . . . . . . . . . . . . . . . . 81

Appendix A: Selected Employment and Training Programs: Summary Descriptions

Tables and Figures

4.7 Welfare Status of AFDC Recipients at the Beginning of the Program.................................... 45

D.1 National Evaluation of Welfare-to-Work Strategies: Participation Rates for

Who This Guide Is For

■ presenting cost information in the most useful manner possible;

■ dealing with issues and problems encountered in measuring costs.

1.2 Potential Uses of Cost Analyses

1.2.1 Cost Analyses of Programs Prior to Their Implementation

1.2.2 Cost Analyses of Ongoing Programs

Cost analyses of ongoing programs can serve a number of purposes, although, as

1.3 How to Use This Guide

reference to the disk is made, the symbol appears in the margin.

ing out your own program.

WHICH SECTIONS OF THE GUIDE ARE FOR YOU?

If you are . . . The most useful sections for you are . . .

A program planner Sections 2.1 and 2.2.

A community-based Chapter 2 for major concepts that should be considered

An agency or organization Sections 2.1 and 2.2.

2.1 The Issue of Perspective

2.1.1 Alternative Perspectives

EXAMPLES OF HOW E&T EFFECTS ARE VIEWED

Effect Government NGO Participant Society

2.1.2 Choosing Among Perspectives

2.2.2 Gross Costs Versus Net Costs

2.2.3 Start-Up Costs

WHAT IS THE DIFFERENCE BETWEEN NET COSTS AND GROSS COSTS?

2.2.4 Average Costs Versus Marginal Costs

e.g. EXAMPLES OF AVERAGE VERSUS MARGINAL COSTS

• Certain pieces of equipment and certain supervisory activities may be required

• If it is necessary to make major new purchases or to rent new facilities in order to

• In expanding an E&T program, the new participants may differ in important

2.2.5 Joint Costs

A better approach would be to attempt to determine whether the resource would

2.2.6 Regular Versus Special Costs

2.3 The Time Dimension

2.3.1 Role of Discounting in Conducting Cost Analyses

THE MECHANICS OF DISCOUNTING

To illustrate discounting, assume that it is predicted that a not-yet-implemented E&T

2.3.2 Adjusting for Inflation

THE MECHANICS OF THE INFLATION ADJUSTMENT

The inflation adjustment factor is easily computed as follows:

2.3.3 Adjusting for Large Equipment Purchases

THE MECHANICS OF STRAIGHT-LINE DEPRECIATION