Draft Only
The Dynamics of New Y ork State Social Welfare Reform Finance at the County
Level: A Feedback View of System Behavior
Tsuey-Ping Lee
I-Shou University
Kao-Hsiung, Taiwan, R.O.C.
Abstract
The 1996 welfare reform gave states an opportunity to review their own welfare
programs. One of the important tasks for states recently is to examine the past resource
allocation policies, and to explore the factors preventing these policies from meeting pre-
set goals. In order to help state welfare decision makers approach the 1996 welfare
reform challenge and to understand the welfare system as a whole, this study tries to
explore the unexpected effects that offset the intended impacts of several welfare fiscal
policies from a system dynamics point of view.
This study builds on the tradition of studying the feedback mechanisms that
generate unintended policy outcomes with the aim of improving policy innovations in
public policy systems. A highly aggregated system dynamics model is presented for the
purposes of understanding implied feedback mechanisms underlying the welfare reform
financing. The model in this research elucidates feedback mechanisms where fiscal
policies that were intended to achieve goals related to administrative rules or controls
touch off unforeseen consequences. This research examines the dynamic impacts and
consequences of various fiscal policies by analyzing the interactions among major welfare
actors. This analysis may provide information regarding the factors causing past financial
policies’ failure to control the welfare expenditures. This thesis argues that past policies
are piecemeal and fragmented because they lack insight into the feedback structure of this
system.
How Policies Designed to C ontrol the C ost of Welfare Reform Can Have
Unintended C onsequences: An Introduction
The social welfare system that provides assistance to low-income and needy
families has been running for decades. The purposes of the welfare system include
providing cash assistance to needy families while reducing caseload and costs, promoting
work incentive, increasing earnings, promoting self-sufficiency of post-welfare employees
and so forth. Not a single welfare policy can achieve all these goals. Those welfare
policies, especially regarding how to improve the welfare system or how to allocate
welfare resources more efficiently, could easily draw public attentions and discussions.
Behind each welfare expenditure allocation policy / strategy, there are goals that
the governments intended to meet. For example, the federal mandatory participation
fraction which requires a specific fraction of a state's nonexempt welfare recipients to
participate in the job-related program is a strategy to help more welfare recipients become
employed. Accordingly, a reducing welfare caseload as well as welfare cost is expected.
1
Raising job-training program quality is another policy to help welfare recipients quickly
find jobs and leave the welfare system. As a result, the welfare caseload and spending
could be decreased. The purpose of setting expenditure limits, for example, administrative
cap and budget limits, is to control the administrative expenditures and the growth of total
welfare expenditures.
However, not every welfare policy fully meets its goals. For example, the goal of
increasing the work participation fraction was not completely met as can be seen by
looking at the history of welfare caseload. The mandatory minimum participation rate,
under the Aid to Families with Dependent Children (AFDC) program, increased from 7
percent in Fiscal Y ear (FY ) 90-91 to 20 percent in FY 95. The increase in the
participation fraction seemed not to have had much influence on the caseload. Figure 1, a
plot of AFDC caseload and national unemployment rate, suggests that the national
economics has been influencing the change of welfare caseload.
Figure 1
Average Monthly AFDC Caseload vs. Unemployment Rate
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
Year
[E=_Arerage Monthly Caseload —— Unemployment Rate |
Sources: Committee on Ways and Means, U.S. House of Representatives, 1998 Green Book. P. 413
Bureau of Labor Statistics, Department of Labor, February, 1998.
Compensating feedback can cause unintended policy effect. Unexpected policy
effects may offset the intended influences of a policy. This study hypothesizes that the
pattern of system's behavior and the implied endogenous feedback mechanism create the
unexpected effects of a policy so that the policy's goals can not be fully reached. Using
feedback thinking linked to a formal simulation model, this study intends to focus on the
unexpected effects of the mandatory participation fraction, higher job-related program
quality, administrative expenditures cap and total budget limits. This study addresses the
following questions: Do these welfare policies which intended to control costs and
caseload using fiscal instruments fully reach intended goals and produce expected effects?
If not, what unexpected factors may be the ones that cause these problems? Finally, what
kind of policy improvement could solve problems associated with unintended effects due
to feedback mechanisms that exist in the cost control policies?
This paper opens with a problem statement and the major purposes of this study
followed by a brief introduction of the base of the study. Then, a brief literature review
provides theories and empirical studies within which the key relationships of model WF
3.0 are found. The literature review is organized based upon the three major actors in the
system: the federal government, the state/local govemments as the welfare service
provider, and the welfare recipients. A fter the literature review, a research methodology
of this study is introduced and followed by a description of modeling tasks of this study.
An example of the simulation results and conclusions drawn from the simulation results
are presented at the end of this paper.
The Base of the Study
This study is built upon an Welfare Reform Research Project sponsored by New
Y ork State Department of Social Services (DSS). The system dynamics modeling team of
University at Albany, Rockefeller College, cooperating with Office of Temporary and
Disability Assistance (OTDA) of New Y ork State (NY S), and Department of Social
Services (DSS) of Cortland, Dutchess and Nassau county, have modeled the social
welfare reform issues at county level from January 1997 to September 1998. This project
is looking at the welfare client flow taking a full system view. Different from that research
project, this paper will only focus on dynamic budgeting and welfare expenditure issues.
The DSS Welfare Reform Research Project has been through several group model
building processes and produced several products including meeting reports, flight
simulators and parameter booklets. Those products provided important information for
this study. Generally, the research project builds a strong foundation for the modeling
tasks of this thesis. The model, WF 3.0, of this study is a higher aggregated model based
upon the welfare model, Phase_8e, of the NY S DSS Welfare Reform Research Project
(Center for Policy Research, May 1997; August 1997).
This research focuses on budget dynamics and financial policy analysis of
Temporary Assistance to Needy Families (TANF) program (previously, Aid to Families
with Dependent Children). Other benefit program like Supplemental Security Income
(SSI), Medicaid, Child Support Enforcement, Social Security benefit, Food Stamps, and
Home Relief are out of this research boundary.
Welfare Policy Experiences and Behaviors of System Actors
The formulation of WF 3.0 model is based upon theories and research that study
the behaviors of welfare system actors and the interactions among them. There are three
major system actors in the welfare system. They are federal government, state and local
governments and the welfare recipients. The higher level sets the stage for the level
below. The system actors’ behaviors, including their expectations for the welfare policies,
their responses to changes in welfare system, and the interactions among them, lead to the
formulation of WF 3.0 simulation model.
Currently, the Personal Responsibility and Work Opportunity Reconciliation Act
(PRWORA) of 1996 placed less responsibility on the federal government and granted the
state and local levels primary responsibility. The PRWORA changed the social welfare
funding system of the United States dramatically. No later than July 1, 1997, each state
began operating a program of assistance to needy families funded under the Temporary
3
Assistance for Needy Families (TANF) Block Grant. The provisions broadly put state
governments in the business of operating their own welfare systems in states by tuming
over functions such as determining the eligibility of welfare recipients and designing their
own welfare programs.
1) Federal Government
In the current welfare system, federal government works as a national welfare-
funding provider. In addition to provide funding, federal agencies set the general
guidelines regarding detailed welfare funding formula for state and local welfare services.
Additionally, these general guidelines contain ways of federal government's supervision
and evaluation over the implementation of local welfare services.
There are different intended effects behind federal welfare financial policies.
However, not all of the pre-set goals are fully met. For example, one of the officially
listed goals of PRWORA is to end the dependency of needy parents on government
benefits by promoting job preparation, work and marriage (Section 401 (a)). However,
this specific goal may not be fully achieved if welfare recipients’ earnings are not
promoted. The dependency of welfare recipients will not be promoted only by
employment. Their eamings should be raised in order to stay away from the welfare
system as long as possible. Otherwise, the possibility of these post-TANF employees to
return to the welfare assistance could keep high.
2) State and Local Governments
State and local governments blend federal policies with their existing programs to
adjust them to local customs and norms. Under TANF, state and local governments have
broad discretion to customize their welfare program to meet local requirements. State and
local govemments are allowed to determine the eligibility standards and benefit levels for
welfare assistance recipients in their jurisdictions, to decide the resources allocation among
different welfare programs or services, to set the budget limitations on welfare programs
and so forth. The shift of the funding formula from matching funds to block grant could
change the ways that state and local governments design their welfare programs and
allocate welfare resources.
Unintended effects of state and local governments’ welfare policies happened
under the new law. With a broader discretion on local welfare services, states have the
freedom to find their own ways to release financial pressure which may caused by
insufficient financial resources accompanied by the increase in welfare caseload. Studies
show that the welfare benefit level could easily become a target of reduction (Fisher, 1996;
Gold, 1995). In addition, state and local governments have discretion on the quality of
welfare-to-work programs. Studies suggest a possible trade-off between income gain for
participants versus budgetary savings for govermments. This trade-off, according to the
study, may be greatly influenced by the format of a welfare-to-work program (Friedlander
& Burtless, 1995). Steuerle and Mermin (1997) argued that the TANF block grant
provides low incentive for states to spend additional funds on low-income families with
children. Previously, any additional dollar a state chose to spend on cash assistance for
low-income families with children would be matched by federal funds. Under TANF, any
additional dollar a state chose to spend is state-only dollar. Therefore, as Peterson (1995)
4
argued, rapid state spending growth will be halted because now states would enjoy 100
percent of any program savings they generate rather than sharing savings with the federal
government.
3) Welfare Recipients
The welfare recipient’ s high length of stay has been a critical issue in this system
(Committee on Ways and Means, U.S. House of Representative, 1996:89; Dutchess
County meeting report, June 1997). One of the means to decrease length of stay of the
welfare recipients is to raise their work incentives. However, there is a concer that high
benefit levels would discourage lower-income people to work (Lerman, 1995:17).
Appel’s study in 1972 had the similar conclusions. Appel found that a welfare system with
programs to encourage work incentives could make welfare system so attractive that it
decreases the number of families leaving AFDC and increases the number entering AFDC.
Accordingly, the welfare costs and caseloads increase.
Although a welfare recipient leaves the system, his or her dependency on welfare
benefits does not necessary end permanently. According to Pavetti’s study based on
monthly data in 1993 and 1995, 58 percent of those who exit AFDC come back within 24
months. Those who leave AFDC because of employment remain off the program
somewhat longer than those who leave for other reasons. However, leaving AFDC due
to work only accounts for slightly less than half of all exits within a 5-year period
(Committee on Ways and Means, U.S. House of Representative, 1996: 89). A recent
study analyzing 57 districts outside New Y ork City shows that 79 percent of the cases that
left the welfare rolls in the first quarter of 1997 stayed off welfare during a one-year
follow-up period (Nathan and Maxwell, 1999, iv). In other words, about 21 percent of
the leaving cases ever retumed to welfare. Some local data even shows that new opening
cases only account for around 50 percent of the average monthly cases opened (Parameter
Booklet of Dutchess County, 1997). This data implies that the other 50 percent of cases
opened are actually re-opened cases. A high recidivism rate drove the AFDC caseload to
increase about 23 percent from 1985 to 1996, and in effect raised total welfare
expenditures (Committee on Ways and Means, U.S. House of Representative, 1998).
The Major Welfare Expenditures in this Study
In this study, welfare expenditures are divided into three major areas:
administrative expenditures, census-driven benefit expenditures, and employment service
program expenditures. A variety of functions and types of expenditures make up the
totality of administrative costs. The administrative expenditures defined here are those
costs paid for staff administering the TANF program (previously, AFDC and JOBS).
These administrative expenditures include staff salaries and fringe benefits. Non-personal
administrative costs including costs associated with the operation of physical office space,
heat, light, etc. are not included in the administrative expenditures defined here because
those are fixed costs that are uncontrollable in the short term. The census-driven benefit
expenditures are paid for welfare cash benefit. Employment service program expenditures
are defined here as payment for purchasing employment and related services including day
care, transportation, skill training, and so forth. The purposes of this spending is to move
people to employment and keep them employed as long as they can to avoid long-term
welfare dependence.
All of the expenditures mentioned above are paid for the purpose of providing
basic needs and promoting self-sufficiency of welfare recipients. However, most of the
past financial policy interventions did not fully meet their intended goals due to
unexpected effects. For instance, the welfare-to-work program that successfully improved
recipients’ earnings encountered the growth of government spending. The program
emphasizing both govemment budget savings and recipients’ immediate employment failed
to improve recipients’ earnings. Federal mandatory participation fraction may successfully
place more TANF recipients in job-training programs, but insufficient administrative
support may offset the intended effect of this federal regulation. In review, most financial
policies failed to foresee the effects and consequences of an individual policy from every
dimension. This failure could prevent a policy from reaching its intended goals.
Therefore, this study tries to explore the unexpected effects that offset the intended
impacts of several resource allocation policies from a dynamic point of view.
Research Method and Analytic Framework
“The goal of a system dynamics policy study is understanding: understanding the
interactions in a complex system that are conspiring to create a problem, and
understanding the structure and dynamic implications of policy changes intended to
improve the system’s behavior.” (Richardson, 1991: 162) The purposes of this research
are to build a highly aggregated system dynamics model in order to understand implied
feedback mechanisms underlying welfare reform financing, to understand how the
feedback mechanisms meet the intended purposes of policy implementation, and to show
how they produce unintended effects.
Figure 2 illustrates the research framework of this study and most importantly, the
role of model WF3.0 and its relationships with other characters in this research
framework.
As shown in figure 2, model WF3.0 is based on the welfare model that was funded
by New Y ork State Welfare Reform Group Modeling Project. The welfare model is a
large and comprehensive model with 651 equations involving the detailed welfare services
modeling. Since the purpose of this thesis is to analyze the welfare financing system from
a more aggregated view, a detailed modeling of welfare services as in the model is not
necessary for this particular purpose. To fulfill this purpose, a higher aggregated model
with major feedback mechanisms, WF 3.0, was built. The structures, parameters and
welfare service effects in model WF3.0 were basically formulated upon the welfare model.
Model WF3.0 was calibrated to Dutchess county. The historical administrative data of
Dutchess county was employed in the formulation of the model and as the reference mode
for model evaluation. Gathering the strengths of the welfare model, model WF3.0 has
rich feedback mechanisms with transparent processes of parameters and welfare service
effects calibration.
Primary Methods of
System Dynamics
rd
Group Time Series
Modeling Analysis
=
Figure 2 The Research Framework
Model WF3.0
Welfare Model
|g
en)
Professional Knowledge
Administrative
Data
Generalizabilit
Empirical Studies
(Literature Review)
Finance at Welfare
Reform Sites
In addition to the welfare model, a minor part of the basis of model WF3.0 is
empirical studies in the field of social welfare. These empirical studies provided model
WF3.0 with richer ideas of existing problems in the welfare system.
Model Focus and Boundary
This study focuses only on the budget dynamics and the financial policies analysis
of the local welfare system. The influence of non-financial factors, such as self esteem,
time cost, the proportion of specific race in the low income population, the education level
of the welfare recipients, the low income family size and so forth, are excluded from the
model. The national economy that is indicated by the unemployment rate is included in
this research. However, since the national economy has been an important factor
influencing welfare caseloads, the effect of the unemployment rate on welfare caseloads is
deactivated when fiscal policies are tested so that the pure effects of welfare fiscal policies
could be easily shown and understood.
Model WF3.0 was calibrated into a county such like Dutchess County. Dutchess
County is a median size county in New Y ork State. The county Department of Social
Services designs local welfare programs and provides local welfare services. The
reference mode of model WF3.0 for model evaluation is AFDC/TANF caseloads of
Dutchess County from 1984 to 1998. The pattern of the changing caseload is of more
concer here than the numbers. The model is not developed for projecting point by point
welfare caseload and expenditures. Neither is it developed for explaining what specific
factors influence the growth of the caseload and expenditures and which factors do not.
Modeling and Model Evaluation
The conceptual foundation of model WF3.0 is based upon welfare model that was
drawn from several group-modeling conferences held in Cortland and Dutchess County.
Knowledge elicited from the discussion at these group-modeling conferences and data
calibration conferences provides the foundation of causal linkages in the model. The
information elicited from these conferences also provides the direction and polarity of a
relationship and the magnitude of a relationship. In addition to the knowledge and
information obtained from those group modeling sessions, some empirical studies from
literature review are sources for model formulation. The division of model sectors is
based upon key system actors. The major two sectors are TANF recipients and state/local
governments (including the budgeting process dynamics). Since the national financial
welfare policies made by the federal government have impacts on local budget processes,
the function of the federal government is blended into the TANF recipients sector and the
state/local government sector. Figure 3 shows an overview of the model sectors of model
WF 3.0.
“Confidence in a system dynamics model accumulates gradually as the model
passes more tests and as new points of correspondence between the model and empirical
reality are identified” (Forrester & Senge, 1980: 209). Four tests are conducted for
testing consistency and suitability of model structure and model behavior (Richardson and
Pugh, 1981: 312). “Structure and parameter verification” is employed to examine the
model structure consistency with the reality. “Extreme condition test” is for the purpose
of model structure suitability. “Behavior reproduction test” is to examine if the model
8
Figure 3 Model Overview
TANF Clients
Enroll Families
Recidivism
fd
t
TANF > Post-TA NF}
<< Caseload | Employees Tits Masten
Departing T Job Finding
ta N Eamings
Ad Cash
Racnates Payments Program Slots Length-Of-Stay
Needed Needed Needed te Work Capability
l
At Administrative Longer Program Service
Total Budget Effectiveness of Pro.
Requested Coverage Program Slots Coverage LOS Effect Quality
, f
\, Average Unit
7 Cost of Program
Maximum. Total Budget Administrative:
[> Total Budge? Granted Support Ratio
Program Resource
Shortage Pressure
Cash Payment
Expenditures Admin. Staff Program Slots
Cash Benefit State/Local Government
Level
Federal TANF Administrative
Block Grant Cap
Mandatory National Funding Provider:
Participation Fraction Federal Government
behavior is consistent with the reality. “Behavior sensitivity test’ is for suitability of model
behavior (Forrester & Senge, 1980: 212-223; Richardson & Pugh, 1981: 310-318).
By structure and parameter verification, the assurance on the conceptual and
numerical correctness of model's structure and parameters is build. Through extreme
conditions test, the suitability of model structure is ensured. The behavior reproduction
test shows that the model- generated behavior fairly matches observed behavior of the real
system (as shown in figure 4). A point-by-point measure of goodness-of-fit between
model-generated and observed data is 0.75 (adjust R square). In behavior sensitivity test,
the confidence in the model behavior is enhanced because the model behavior is not
sensitive to changes in parameter values. A fter a fair confidence in the model WF 3.0 is
build up, this model is ready for policy tests.
Figure 4 Simulated Caseload, Actual Caseload and Unemployment Rate of
Dutchess County
4,000 Families
0.2 Dimensionless
2,000 Families
0.1 Dimensionless
0 Families a
0 Dimensionless
1984 1991 1998
Time (Y ear)
Simulated AFDC Caseload —_—— _ Families
Dutchess AFDC Historical Caseload Families
Simulated AFDC Caselaod in Equilibrium ~~ Families
Dutchess Historical Unemployment Rate - Dimensionless
Model Analysis and Policy Testing
The simulation results generated by Model WF3.0 are analyzed by examining
longitudinal graphs over time of some key variables. Each graph is a result of the
interaction among all model variables under a specific scenario and policy condition. The
simulation time period is 16 years from 1994 to 2010.
Possible impact of different financial policies are examined by comparing their
output values of key variables over time with the base run and among one another. The
comparison among different runs is based upon three aspects. First, the pattern of the
output values of key variables over time is scrutinized by examining the data of the output
10
values. These output measurements are TANF caseload, total expenditures, earnings and
recidivism fraction. Through these data, the part of the model that is accountable mainly
for the trends of output behavior could become clearer. Second, percentage changes from
the beginning to the end of simulation time are compared to observe the different impacts
in magnitude. Third, 16-year cumulative values of key variables are compared to examine
the total long- run effects. The cumulative values of financial variables for policy analysis
are shown in net present values.
In this analysis, the base run is based on a world without the influence of national
economics, administrative expenditure caps, higher job-training program quality, and the
budget limit. Seven fiscal policies are examined in this study. They are:
1. Setting the administrative cap:
An administrative cap is designed to prevent excessive administrative
expenditures. In the policy run of setting the administrative cap, the
administrative budget ceiling is fixed at 15 percent of the total budget granted.
2. Setting the budget limit:
Under PROWRA, for the first couple of years most of the states will get a little
more money than what they would have received under AFDC (Edelman,
1997: 49-50). Therefore, in this study, the total budget limit is fixed at 5
percent above the equilibrium (original) budget level so that the welfare system
does not encounter budget pressure at the beginning of the simulation time.
However, there is a very real possibility that states will run out of federal
money woon (Edelman, 1997:50). The policy test results sho that the budget
ceiling is hit very soon. This budget includes the program budget,
administrative budget and TANF cash benefit payment budget.
3. Raising program quality:
Raising program quality is indicated as higher unit cost of program slots. A
major goal of the welfare system is to move recipients from public assistance to
stable employment. Therefore, job-related program becomes a focus to be
expanded and intensified for this purpose. In this study, a mechanism is set to
raise the unit cost of program slots gradually after year 1997 until the unit cost
is twice the original cost at year 2002.
Combination of the administrative cap with the budget limit.
Combination of the administrative cap with raising program quality.
Combination of setting the budget limit and raising program quality.
Combination of setting the budget limit, administrative cap and raising program
quality.
The following is an example of policy testing and analysis. The dynamic structure
of this policy will be discussed first and followed by the simulation results.
SS) 1
An Example of Policy Testing and Analysis - Setting Administrative C ap
1) Dynamic Structure of Administrative Cap
There are four important feedback loops in figure 5. Loop A is named program
coverage loop. In this loop, higher TANF caseloads need more budgets for building
program slots. Under unlimited budget, the amount of program slots will be provided as
requested, except budget delays. Higher program coverage is expected to help more
11
TANF recipients find jobs so that the TANF caseload could decrease accordingly. Loop
B is named administrative coverage loop which demonstrates the similar story as what
loop A does. In loop B, a percentage administrative cap is set to control over the
maximum administrative budget. Loop C is administrative- support acceleration loop.
Loop C demonstrates that, by multiplying a higher federal mandatory participation
fraction, the increased TANF caseload will raise the need for program slots.
Consequently, the amount of program slots will be increased. The key point in this loop is
the administrative-support ratio. If the number of administrative staff grows slower than
the number of program slots does, the administrative-support ratio becomes smaller. A
smaller administrative-support ratio indicates fewer administrative supports to program
slots. A smaller administrative-support ratio drives the job-finding rate even lower, and
thus, raises the TANF caseload. This acceleration loop can offset the intended effect of
the mandatory participation fraction.
Loop D is an administrative-support compensating loop. A higher TANF caseload
needs more administrative staff to support the system. Increasing administrative staff
drives the administrative-support ratio higher and, accordingly, the job-finding rate is
increased and the TANF caseload is decreased. However, since the number of
administrative staff is changed based on the size of TANF caseload, the growth of the
number of administrative staff usually cannot catch the dramatic growth of program slots.
Therefore, the administrative-support acceleration loop (loop C) is usually more
dominating than the administrative-support compensating loop (Loop D).
The administrative cap places a ceiling on the maximum administrative budget.
The maximum administrative budget is formulated as the total granted budget multiplied
by the administrative cap percentage. This maximum administrative budget is a
mechanism to keep administrative expenditures under control. The federal government
mandates that total administrative expenditures must not go over 15 percent of total
expenditures. Therefore, the administrative cap percentage is formulated as 0.15 in this
policy test.
2) Policy Run : Setting Administrative Cap
Figure 6 shows the administrative cap run compared with the base run. The
administrative cap, as expected, drives the system to spend less money than in the base
run. In the graph of cumulative total expenditures in present value, the administrative cap
run spends five millions less than in the base run along the simulation time. The savings is
caused by the reduced administrative expenditures. However, setting the administrative
cap creates more TANF caseload with a higher number of program slots than in the base
run. The job-finding rate in the administrative cap run is much lower than in the base run.
In summary, although the administrative cap saves administrative expenditures, the job-
finding rate becomes lower, and accordingly, the TA NF caseload increases.
12
Figure 5 The Dynamic Structure Including Administrative Cap
ITANF Caseload => Employed Post
| Tarrndaa_| TANE
f Length-of-Stay
b Findin
_—_ J Faden Needed numbers of slots
(Due to Mandatory
Admin — r Participation Fraction)
Coverage i wees ) Ch) onatam Ps Needed numbers
Adie to =. Coverage of Admin Staffs
Program Ratio A 4)
Program
~ Slots Total Requested
D 4) Budget
— Total Budget
Admin Stafts Granted Budget Delay
Maximum. —_—
Admin Budget
~— Percentage
Admin Cap
How does the administrative cap function? When the need for more administrative
staff grows to reach maximum administrative budget, the administrative-support ratio
declines. This is because administrative budget is no longer available for hiring more staff,
on the one hand, the number of program slots keeps increasing due to the increased
mandatory participation fraction, on the other hand. The lower the administrative-support
ratio, the less the job-finding rate. Therefore, TANF caseload becomes higher than in the
Tun without the administrative cap (base run). This story shows that cutting administrative
expenditure could cause the program slots to be less productive. Additionally, this
simulation result implies that any policy design without examining the whole picture of the
system could cause unexpected problems.
13
Figure 6 Simulation Results of Base Run vs. Administrative Cap Run
Program Slots -
Base vs. Administrative Cap
1000 gga ean on +106%
700 Pa
400 [/———T]
1994 1998 2002 2006 2010
Time (Y ear)
Program Slots: Base 9#§=——————————— Slots
Program Slots : Base with Admin Cap------------- Slots
[ob Finding -
Base vs. Administrative Cap
1,000
ee Ee
800 a = a 9.3%
600
1994 1998 2002 2006 2010
Time (Y ear)
Job Finding: Base |§©§ ——————————— Families
Job Finding : Base with Admin Cap------------ Families
14
TANF Caseloads -
Base vs. Administrative Cap
2,000
ee 13%
1,625
1,250
1994 1998 2002 2006 2010
Time (Y ear)
TANF Caseload : Base —————_ Families
TANF Caseload : Base with Admin Cap ------- Families
Total Expenditures -
Base vs. Administrative Cap
20M
— 48.2%
EN 0
1M ——p= +4.8%
10M
1994 1998 2002 2006 2010
Time (Y ear)
Total Expenditures: Base 9 ———————_ Dollars
Total Expenditures : Base with Admin Cap ----- Dollars
15
Total Expenditures in Present Value (C umulative) -
Base vs. Administrative Cap
300 M a
b.7 Million
Difference
150M
0
1994 1998 2002 2006 2010
Time (Y ear)
Total Expenditures in Present Value : Base — Dollars
Total Expenditures in Present Value : Base with Admin Cap-- Dollars
Summary and Conclusion
The purpose of this thesis is to build a highly aggregated system dynamics model in
order to understand implied feedback mechanisms underlying the welfare reform financing.
In addition, how the feedback mechanisms meet the intended purposes of fiscal policy
implementation and how these policies produce unintended effects are also the major
concerns of this research. This study does not attempt to project TANF caseload point by
point. The concer is the pattern of the system’s behavior and the implied feedback
mechanisms.
1) The Mechanisms of Intended and Unintended Effects of Fiscal Policies
There are always one or more expected effects behind each policy. The expected
effects are usually drawn from policy makers’ understanding of the welfare system.
However, lacking a comprehensive view of the entire welfare system, such an
understanding might be partial and incomplete. Unexpected effects could be caused by
endogenous problems. In this section, the intended effects of various fiscal policies are
described and followed by an analysis of feedback mechanisms that cause the side effects.
A. The Effects of Mandatory Participation Fraction
i) The Optimistic View of Mandatory Participation Fraction
Figure 7 depicts the pathway of how the mandatory participation fraction is
expected to reduce TANF caseload. The mandatory participation fraction increases from
25% to 50%. Accordingly, more TANF recipients are required to participate in job-
related programs, and then, more TANF recipients are expected to find jobs. As a result,
the TANF caseload is reduced.
16
Figure 7 Expected Effect of Mandatory Participation Fraction
TANF
Caseload) Se
TANF Recipients
Finding Jobs
TANF Recipients in
Job-Related Program
\ Mandatory
Participation Fraction
ii) The Shortage of Administrative Support Reduces the Intended Effects
of Mandatory Participation Fraction
The effects of the administrative support shortage counteract the expected effects
of increasing the mandatory participation fraction. As the self-reinforcing loop shown in
figure 8, the mandatory participation fraction increases the total of TANF recipients in
job-related programs. Meanwhile, the amount of administrative staff is not enough to
support the dramatic increase of the amount of TANF recipients in job-related programs.
Thus, the job-related programs are not utilized in a productive and effective way. Asa
consequence, fewer TANF recipients in the programs can find jobs, and therefore, the
TANF caseload increases. Due to the increased mandatory participation fraction, the
more the TANF caseload is, the more the TANF recipients are required to participate in
job-related programs. Under this circumstance, the administrative support shortage
becomes even worse. Hence, if the amount of administrative staff does not change to
correspond with the amount of the TANF recipients in job-related programs, the
mandatory participation fraction could produce unexpected effects that lessen its intended
effects.
17
Figure 8 The Shortage of Administrative Support Reduces the Effect of Mandatory
Participation Fraction
TANF
Caseload Sees
TANF Recipients
Administrative Finding Jobs
Staff
+
Administrative Support TANF Recipients in
To Job-Related Programs Job-Related Program
Mandatory
Participation Fraction
B) The Effects of Raising Program Quality
i) The Expected Effects of Raising Program Quality
In this study, the program quality is measured by the unit cost of program slots. A
higher unit cost of program slots indicates a higher program quality, and vice versa.
Raising the program quality is expected to reduce TANF recipients’ length-of-stay.
TANF recipients who participate in job-related programs can receive better job-related
services than they would have had in a lower-quality program. Those TANF recipients
who receive better job-related services are expected to find jobs and leave TANF system
quickly. In other words, these TANF recipients’ length-of-stay will be shorter, and
therefore, the TANF caseload is reduced. The intended effects of raising the program
quality are displayed as a causal diagram in figure 9.
18
Figure 9 The Expected Effects of Raising Program Quality
TANF
Caseload
TANF Recipients
Finding Jobs
TANF Recipients’
Length-Of-Stay
Program Quality
+
Unit Cost of Program
Slots
ii) The Focus of Job-Related Programs Influences the Intended Effects of
Higher Program Quality
The expected effects of raising the program quality could be offset if job-related
programs emphasize a long-term vocational training and education. The more the job-
related programs emphasize long-term vocational training and education, the more the
expected effects of higher program quality are offset. As the pathway shown on the left of
the figure 10, the more the job-related programs emphasize long-term vocational training
and education, the longer the TANF recipients are required to stay for job-training.
Therefore, TANF caseload increases.
19
Figure 10 Longer Length-Of-Stay Effect
TANF
Caseload rs ani
TANF Recipients
Finding Jobs
TANF Recipients’ TANF Recipients'
Length-Of-Stay For Length-Of-Stay
Job-Training {
+
Program Quality
+
Unit Cost of Program
Slots
C) The Function of the Administrative Cap
i) The Expected Effects of The Administrative Cap
The purpose of setting the administrative cap is to limit the administrative
expenditures. Currently, the local administrative expenditures, including local shares, are
beyond 20 percent of the total TANF/AFDC expenditures. When local governments’
administrative expenditures surpass the administrative cap, the local shares are increased
to cover the exceeding expenditures. When increasing the local shares becomes a strategy
of local governments to cover exceeding administrative expenditures, the original purpose
of administrative cap - preventing the administrative expenditures from overspending -
becomes even more difficult to reach. Thus, this study examined the policy consequences
of setting the administrative cap under the assumption that local governments do not
increase local shares. If so, as shown in figure 11, the lower the administrative cap, the
lower the available administrative budget. The lower the available administrative budget,
the fewer the number of the administrative staff. As a result, the administrative
expenditures could be under control.
20
Figure 11 The Expected Function of The Administrative Cap
Administrative
Expenditures
+
Administrative
Staff
f
Admin Budget
Available
Administrative
Cap
ii) The Side-effects of Setting the Administrative Cap
Without increased local shares, the administrative cap restricts the available
administrative budget. Thus, the amount of administrative staff is limited. Facing the
dramatic increase of TANF recipients in job-related programs, the limited number of
administrative staff worsens the shortage of administrative support. Due to the
administrative support deficiency, the job-related programs could not be as productive and
effective as with sufficient administrative support. Consequently, less TANF recipients in
these programs can find jobs. The TANF caseload increases accordingly. In this case,
more TANF recipients are required to join job-related programs which, therefore, cause
higher program expenditures. As shown in figure 12, the administrative cap restrains the
administrative expenditures, but its side-effects produce higher program expenditures.
Apparently, the expected governments’ savings from capping administrative expenditures
are offset by the higher program expenditures.
21
Figure 12 The Side-effects of Setting the Administrative Cap
Program
Expenditures .*
TANF Recipients in
Administrative Job-Related Program
Expenditures
+
Administrative _
Staff Administrative S ‘t TANF
+ 4 inistrative Suppo: Cascload. ==
Ee To Job-Related TANF Recipients
Programs Finding Jobs
Admin Budget _——
Available
Administrative
Cap
D) The Effects of Total Budget Limits
i) Economize By Setting Total Budget Limits
The purpose of setting total budget limits is to control the growth of total TANF
expenditures. The lower the total budget limits, the lower the budgets available for both
administrative and program expenditures. Consequently, the amount of both the
administrative staff and the TANF recipients in job-related programs are limited.
Therefore, the actual administrative and program expenditures are under control and so
are total expenditures. Figure 13 illustrates the diagram of the expected effects of setting
total budget limits.
22
Figure 13 Economize By Setting Total Budget Limits
Total
Expenditures
a a
+
Administrative Program
Expenditures Expenditures
+
+
Administrative TANF Recipients in
Staff Job-Related Program
+
+
Program Budget
Admin Budget
‘Available Available
NO
Total Budget Limit
ii) The Unintended Effects of Setting Total Budget Limits
Total budget limits could definitely restrict the total expenditures. However, the
possible side effects of setting total budget limits should not be ignored. Part of the
possible side effects is the same as the one caused by administrative cap. The lower the
available administrative budget, the less the administrative support to the increased TANF
recipients in job-related programs. Hence, fewer TANF recipients in job-related programs
can find jobs, and as a consequence, the TANF caseload increases. However, the extent
of the side effects caused by total budget limits is less than the one caused by
administrative cap. Since a total budget limit restrains both administrative and program
expenditures at the same time, the shortage of administrative support is not as serious as in
the policy of administrative cap.
The program budget pressure that is resulted from the capped program budget
causes the other possible side effects. When governments’ available program budget is
not sufficient to fulfill the needs for more program slots, such a budget pressure could
motivate governments to reduce the unit cost of program slots. However, the lower
program quality (indicated as lower unit cost of program slots) implies that the
govemments can not afford a long-term job-training program. Since the TANF recipients
no longer receive a long-term job-training, they are not as competitive in the labor market
as they otherwise would have been after a long-term training. In this case, the possibility
for them to get jobs with higher wages could become lower. With low-wage jobs, these
post-TANF employees may not be self-sufficient and independent from TANF assistance.
23
Hence, the possibility of recidivism becomes higher, and accordingly, the TANF caseload
increases. Apparently, both side-effects together cause higher TANF caseload. The
trade-off between governments’ savings and higher TANF caseload is worthy to be
considered in the welfare policy making processes.
Figure14 The By-Effects of Setting Total Budget Limits
Total Expen
ditures
Administrative
Expenditures
TANE | sg__gs) Post TANF
7 Employee
+ Post TANF
Employees’ Eamings
+
Administrative Support TANE Recipients in 1 ;
Adminisiaive; To Job-Related Job-Related Program (+4 FANE Reape
Staff ——e Programs : a Length-Of-Stay For
~_——— Job-Training
* +
* +
Admin Budget Program Budget
Program Budget
Available ‘Available a soe Program Quality
: P
x - +
+
Total Budget Limit Unit ce of Program
3) Summary of Simulation Results
Model WF3.0 is a structurally oriented model. All the model simulation outputs
formed in graphs over time are structure-generated. The pattem of system’s behavior is of
more concern in this study than point-by-point projection. By means of model simulation,
several welfare fiscal policies were tested and analyzed. The base world for the base run is
a world with policy variables and exogenous influences taken out except the increasing
mandatory participation fraction. These fiscal policies were analyzed by comparing the
system's behavior between policy runs and base run. Such a comparison can show the
pure effect of a fiscal policy while all the other factors are held the same. The simulation
results and analyses of the fiscal policies examined in this study are summarized below.
24
A) Misallocation of A dministrative-Support Can Drive Up TANF Caseload and Costs
As mentioned above, high program expenditures with increased TANF caseload is
not as optimistic as expected. The main reason for the pessimistic results of the base
tun is that the amount of administrative staff does not increase corespondent with the
increase of TANF recipients in job-related programs. The mandatory participation
fraction requires more people to be placed in the job-related programs. However, the
shortage of administrative support causes the results that the job-related programs can
not be effectively and productively implemented. Therefore, in the base run, even
though the total of program slots become more than doubled, the TANF caseload does
not go down as expected.
B) Administrative Cap Can Exacerbate the Welfare Administration
Given the pessimistic system’s behavior of the base run, the administrative cap
makes the system behave even worse. Under the administrative cap, the TANF
caseload becomes higher because the job-related programs are not as effective as
expected due to the shortage of administrative support. Setting an administrative cap
is able to bring down the growth in administrative expenditures. However, the
program expenditures are raised to fulfill the needs for the higher TANF caseload.
C) Total Budget Limits Can Motivate Governments to Lower the Program Quality
At first glance, setting the total budget limits seems worse than setting the
administrative cap because both the administrative and program expenditures are
constrained. One of the problems caused by the total budget limits is the same as the
one caused by setting the administrative cap. However, because the total budget limit
cuts across administrative expenditures as well as program expenditures, the shortage
of administrative support, indicated as the ratio of program slots to administrative
staff, is not as serious as in the policy of setting administrative cap only. Another
problem caused by the total budget limits is the program financial pressure. The
financial pressure could motivate governments to lower the program quality that
weakens the post-TANF employees’ capability to become independent from TANF
assistance. Therefore, the recidivism rate increases and so does the TANF caseload.
D) Different Job-Related Program Designs Can Generate Different Outcomes
Raising the program quality does not necessarily promote the number of job-
finding people. A policy design that emphasizes long-term job-training and education
could keep TANF recipients staying longer in the job-related programs. The more the
job-related programs emphasize long-term job-training, the longer the TANF
recipients are required to stay in the programs. As a result, the job-finding rate is
comparably lower than in the programs emphasizing quick job-finding. However, such
a long-term job-training program could help to increase the post-TANF employees’
earning levels, and therefore, to decrease the recidivism rate.
3) Implications of Model WF3.0 - Corresponding A dministrative-Support With J ob-
Related Program
25
The analysis of model structure and simulation results suggests that, to be
effective, the amount of administrative staff should vary to correspond with the amount of
TANF recipients in job-related programs. This study examined this policy suggestion with
a revised model structure. In this revised model, the amount of administrative staff is
increased when the present administrative support is lower than its initial level. This
mechanism is formulated for the purpose of alleviating the administrative support
shortage.
The simulation results show that the system behaves better under the new model
structure. The TANF caseload decreases 3.8% compared with equilibrium level. In
addition, this policy suggestion costs $0.6 million more in total expenditures at the end of
simulation time. In summary, changing the number of administrative staff correspondent
with the amount of TANF recipients in job-related programs could make the system
perform better with a little more costs.
4) Research Results vs. Research Goals
As mentioned above, the purposes of this thesis are to build a highly aggregated
system dynamics model in order to understand implied feedback mechanisms underlying
the welfare reform financing, and to understand how the feedback mechanisms meet the
intended purposes of policy implementation, or how they produce unintended effects.
Furthermore, this study intends to provide policy suggestions that could alleviate those
unexpected effects. The study of this thesis does achieve these goals. A highly
aggregated welfare finance model, WF3.0, is built upon a complex welfare model which
involving about 90 local welfare managers’ professional knowledge. Model WF3.0 isa
feedback-rich model and it has passed several model evaluation tests before it was
employed for policy tests. These model evaluation tests were used to examine if the
model structure and behavior are consistent with the real world and suitable to the goals.
Passing these model evaluation tests built up the modeler’s confidence to this model.
The feedback mechanisms in model WF3.0 has helped to understand what the
intended effects of various fiscal policies are and how these fiscal policies are expected to
function. Furthermore, by means of several feedback mechanisms, the analysis of model
structure and the results of policy tests explained why and how the intended effects of
fiscal policies are offset. In summary, clarifying the pathways and feedback mechanisms
which cause unexpected effects of the fiscal policies is the major contribution of this
thesis. Concluding that the welfare system could be better off if the amount of
administrative staff changes to correspondent with the amount of TANF recipients in job-
related programs is also a main contribution of this thesis.
Model WF3.0 does not lean toward any extreme point of view in order to obtain a
specific system’s behavior. It covers both conservative point of view (the performance of
the welfare system is not sensitive to the amount of investment) and liberal point of view
(the performance of the welfare system is sensitive to the amount of investment). The
sensitivity analysis showed that the pattern of system’s behavior does not change among
different points of view. In addition, this thesis not only provided policy suggestions
based on the implication of the model, but also revised the model in order to examine the
policy suggestions. This thesis experimented with different scenarios (higher productivity
26
scenario) to examine if the conclusions drawn from the model simulation results vary
under different scenarios. The experiment showed that the conclusions are solid.
5) Research Limitations and Suggestions for Future Research
There is always a trade-off between elaboration for causal structure and the
desired level of analytic disaggregation (Andersen, 1983: 234). Model WF3.0 lacks the
detailed cost breakout often expected by welfare practitioners. Welfare practitioners who
are used to such details sometimes find it difficult to accept the model’s results. Of
course, the detailed feedback structure could be replicated for each service category,
thereby making detailed cost projections available. For example, the job-related programs
could be disaggregated into job-training program, job-searching program and so forth.
TANF recipients could be differentiated into high need and low need as in welfare model.
The advantage of such a disaggregation is to provide detailed cost breakdown so that the
model's results could be easily accepted and understood by practitioners, and the policy
suggestions provided by the study could be more specific. However, the cost of providing
these information is a extreme complicated model with numerous feedback mechanisms
which may lose the focus of the model.
TANF recipients’ work incentives is a very important and controversial issue in the
field of social welfare. Even though this issue is excluded from the research boundary of
this study, the model could be expanded to include this issue for the future research.
Modelers will find it very difficult to formulate work incentives not only because it is very
hard to operationalize work incentive but also because it is a complex and controversial
issue that may include opposite arguments at the same time. However, it is worthy to try
to formulate welfare recipients’ work incentives because, by means of model simulation,
policy makers could have a more comprehensive view on work incentives when a related
policy is considered.
Another ambitious extension to the model is involving the issue of “race to the
bottom”. “Race to the bottom” means a race for localities to cut welfare benefits faster
than their neighbors in order to reduce the attraction to poor immigrants. This issue might
be more serious in the situation that states have more discretion on deciding their welfare
benefit packages. Since under the TANF system, state and local governments have more
discretion on designing and implementing their local welfare programs, the issue of “race
to the bottom” becomes very important. There are various ways to formulate this issue.
The most comprehensive way is to replicate the existing model sectors to be another
locality, and then, to make two localities respond to each other’s benefit levels. Although
conceptually plausible, such an expansion will make the model become a very large one.
The safety-net system could also be included in the model for the future research.
Under the TANF system, those timing- out recipients will be placed in the safety-net.
Governments’ savings from the TANF system may cause more expenditures in the safety-
net. Including the safety-net system in the model could help policy makers to balance the
policy focus between the TANF and safety-net systems. By means of a system dynamics
model, welfare policy makers could have a more comprehensive view of the entire welfare
financing system. Hopefully, a well designed welfare financing policy could benefit both
welfare recipients and governments.
27
References
(1995). Welfare Exodus. USA Today: 3A.
(October, 1997). Code of Federal Regulations.
(1997). Final Report of the Budget 2000 Project. New Y ork , NY, Citizens Budget
Commission.
Acs, Gregory. (1995). Do Welfare Benefits Promote Out-of-W edlock Childbearing?
Welfare Reform: An Analysis of the Issues, I. V. Sawhill. Washington, D.C., Urban
Institute: 51-54.
Allers, Robert, Robert Johnson, David F. Andersen, George P. Richardson, John W. Rohrbaugh
and Aldo A. Zagonel-Santos. (July, 1998) “Group Model Building to Support Welfare Reform
Part II: Dutchess County.” 1998 Intemational System Dynamics Conference Proceedings v.16.
Andersen, David F, Rudy Runko, George P. Richardson (1979). “The Dynamics of
Municipal Debt Accumulation: Unintended Consequences of Deficit Financing” Summer
Computer Simulation Conference, Toronto, Canada.
Andersen, David F. (1983). “A System Dynamics Simulation of Educational Finance
Policies.” Simulation 40(6). pp.227-235
Andersen, David F. (1980). “Using Feedback Simulation to Test Educational Finance
Policies.” Policy Sciences 12. pp. 315-331.
Andersen, David F. (1980). How Differences in Analytic Paradigms Can Lead to
Differences in Policy Conclusions. Elements of the System Dynamics Method. J. Randers.
Cambridge MA, Productivity Press: 61-74.
Andersen, David F. (1990). “Analyzing Who Gains And Who Loses: The Case of School
Finance Reform in New Y ork State.” System Dynamics Review 6(1). pp. 21-43.
Andersen, David F., John Rohrbaugh, Tsuey-Ping Lee, Aldo Zagonel-Santos (1997). Data
Calibration Exercise For Social Services in Dutchess County Meeting Report. Albany,
NY, Center For Policy Research, Rockefeller College of Public A ffairs and Policy,
University at Albany, SUNY .
Andersen, David F. & George P. Richardson (1997). “Scripts for Group Model Building”
System Dynamics Review 13(2): 107-129.
Appel, Gary Louis (1972). Effects of A Financial Incentive on AFDC Employment:
Michigan's Experience Between July 1969 and July 1970. Minneapolis, Minnesota,
Institute for Interdisciplinary Studies.
28
Blank, Rebecca M. (October, 1995). “The Impact of State Economic Differentials on
Household welfare and Labor Force Behavior.” Journal of Public Economics 18: 25-58.
Brodkin, Evelyn and Michael Lipskya. (1983). “Quality Control in AFDC as an
Administrative Strategy.” Social Service Review 57(1): 1-34.
Brown, Charles. C. & Wallace E. Oates (April 1987). “Assistance to the Poor in a Federal
System.” Journal of Public Economics 32(3): 307-330.
Center for Policy Research. Albany, NY: Nelson A. Rockefeller College of Public A ffairs
and Policy, University at Albany, State University of New Y ork
— May 1997. Welfare reform project: Report of the group model building conferences
in Cortland County
— June 1997. System Thinking: A case study on welfare reform in Chugwa County
— August 1997. Welfare reform project: Report of the group model building
conferences in Dutchess County.
— September 1997. Welfare reform project: User’s manual to the TANF flight
simulator, version 1.0.
— October 1997. Welfare reform project: Report of the model calibration meeting in
Dutchess County.
— November 1997. Welfare reform project: User’s manual to the TANF flight
simulator, version 2.0.
— April 1998. Welfare reform project: Parameter booklet for the combined TANF &
Safety-net model of Dutchess County, New Y ork.
— September 1998. Welfare reform project: Parameter booklet for the combined
TANF & Safety-net model of Nassau County, New Y ork.
Committee on Ways and Means, U. S. House of Representatives. (1996). 1996 Green
Book: Overview of Entitlement Programs. Washington, D.C., Government Printing
Office.
Committee on Ways and Means, U.S. House of Representatives (1998). 1998 Green
Book. Washington, DC, Government Printing Office.
Dobelstein, Andrew W. (1986). Politics, Economics, and Public Welfare. Englewood
Cliffs, NJ, Prentice-Hall, Inc.
Edelman, Peter (March, 1997). “The Worst Thing Bill Clinton Has Done.” The Atlantic
Monthly: 44-58.
Fisher, C. Ronald (1996). State and Local Public Finance, Richard D. Irwin, A Times
Mirror Higher Education Group, Inc. company.
29
Forrester, Jay W. (1961) Industrial Dynamics. Portland, Oregon, Productivity Press.
Forrester, J. W. (1975). Counterintuitive Behavior of Social Systems. Collected Papers of
ay W. Forrester. Cambridge MA, Productivity Press: 211-244.
Forrester, Jay W. and Peter M. Senge (1980) “Tests For Building Confidence in System
Dynamics Models” System Dynamics: TIMS Studies in the Management Sciences. Robert
E. Machol (ed). 14: 209-228. New Y ork, NY, North-Holland Publishing Company.
Friedlander, Daniel. & Gary Burtless. (1995). Five Y ears After: The Long-Term Effects of
Welfare-To-W ork Programs. New Y ork, NY, Russell Sage Foundation.
Gold, Steven D. (1995). Spending Policies and Revenue Trends Compared. The Fiscal
Crisis of the States: Lessons for the Future. S. D. Gold. Washington, D.C., Georgetown
University Press.
Gramlich, Edward M & Deborah S. Laren (Fall, 1984). “Migration and Income
Redistribution responsibilities.” Journal of Human Resources 19: 489-511.
Greenberg, Mark & Steve Savner(1996). Waivers and the New Welfare law: Initial
Approaches in State Plans. Washington, D.C., Center for Law and Social Policy.
Greenberg, Mark & Steve Savner (1996). A Detailed Summary of Key Provisions of the
Temporary Assistance for Needy Families Block Grant of H.R. 3734. Washington, D.C.,
Center for Law and Social Policy.
Greenberg, Mark (1997). Welfare-to- Work Grants and Other TA NF-Related Provisions in
the Balanced Budget Act of 1997. Washington, D.C., The Center For Law and Social
Policy.
Greenberg, Mark (February 1997). HHS Policy Guidance on Maintenance of Effort,
Assistance, and penalties : Summary and Discussion. Washington, D.C., The Center For
Law and Social Policy.
Greenhouse, Linda (1995). Justices Take On Welfare Benefits Case. New Y ork Times:
10.
Hagen, J. L., & Irene Lurie (1994). Implementing JOBS : Progress and Promise. Albany,
NY, Nelson A. Rockefeller Institute of Government.
Jones, Del (1996). Private Firms Eye $28 Billion Welfare Prize. USA Today: 1A-3A.
Katz, Jeffrey L. (February, 1996). “GOP Prepares To Act On Govemors' Plan.”
Congressional Weekly Report 54(7): 394-395.
30
Legislative Commission on Expenditure Review. (1985). Local Social Services
Administrative Costs, Program Audit. Albany, New Y ork.
Lerman, Robert (1995). Increasing the Employment and Eamings of Welfare Recipients.
Welfare Reform: An Analysis of the Issue. I. V. Sawhill. Washington, D.C., Urban
Institute: 17-20.
Levine, L. (1994). Jobs for Welfare Recipients. Washington, D.C.
Long, David A. (1988). “The Budgetary implications of Welfare Reform: Lessons from
Four State Initiatives.” Journal of Policy Analysis and Management 7(2): 289-299.
Lundberg, Shelly and Robert Plotnick (1990). “Effects of State Welfare, Abortion, and
Family Planning Policies on Premarital Childbearing Among White Adolescents.” Family
Planning Perspectives 22(6): 246-251.
Lurie, Irene (1996). “A Lesson From the JOBS Program: Reforming Welfare Must be
Both Dazzling and Dull.” Journal of Policy Analysis and Management 15(4): 572-586.
Lurie, Irene (1997). Temporary Assistance For Needy Families: A Green Light For The
States.
Lurie, Irene (1998). “Welfare Reform in New Y ork State”. Poverty Research News
11(1): 22-23.
McCaffrey, Shannon (1997). State Audit Finds Food Stamp Distribution Costs V ary
Widely. The Sunday Gazette: B6.
Mead, Lawrence M. (1995). An Administrative A pproach to Welfare Reform. Welfare
Reform: An Analysis of the Issues. I. V. Sawhill. Washington, D.C., Urban Institute: 21-
24.
Moffitt, Robert A. (March 1992). “Incentive Effects of the U.S. Welfare System: A
Review.” Journal of Economic Literature 30: 1-61.
Moffitt, Robert (1996). “.” Journal of Policy Analysis and Management 30: 1-61.
Musgrave, Richard A. (1959). The Theory of Public Finance. New Y ork, Mcgraw- Hill.
Nathan, Richard, Terrence Maxwell, etc (1999). After Welfare: A Study of Work and
Benefit Use After Case Closing. Interim Report Submitted to the U.S. Department of
Health and Human Services. The Nelson A. Rockefeller Institute of Government. July
1999,
Newcombe, Tod (1998). “Welfare's New Burden: Feds Tie Down States With Data
31
Reporting Requirements.” Goverment Technology 11(4): 13-15.
O Toole, Daniel E., James Marshall, and Timothy Grewe (1996). “Current Local
Government Budgeting Practices.” Government Finance Review 12(6): 25-29.
Office of the State Comptroller (1997). New Y ork State Department of Social Services’
Food Stamp Program's Administrative Costs. Albany, NY , Division of Management
Audit, NYS.
Peterson, Paul E. and Mark C. Rom (1990). Welfare Magnets: A New Case Fora
National Standard. Washington, D.C., The Brookings Institution.
Peterson, Paul E. (1995). State Response to Welfare Reform : A Race to the Bottom?
Welfare Reform : An Analysis of the Issue. I. V. Sawhill. Washing, D.C., Urban Institute.
Peterson, George E. (1995). A Block Grant Approach to Welfare Reform. Welfare
Reform : An Analysis of the Issues. S. Isabel V. Washington, D.C., Urban Institute.
Peterson, Paul E. (1995). The Price of Federalism. Washington, D.C., Brookings
Institution.
Proctor, Allen J. (1994). “Budgeting for Structural Balance: Illustrations from New Y ork
City.” Government Finance Review 10(4): 7-11.
Riccio, James, Daniel Friedlander and Stephen Freeman (1994). GAIN: Benefits, COsts,
adn Three-Y ear Impacts of a Welfare-to-W ork Program, Manpower Demonstration
Reserach Corporation.
Richardson, George P. & Alexander L. Pugh III. (1981). Introduction to System
Dynamics Modeling with DY NAMO. Portland, Oregon, Productivity Press.
Richardson, George P, David F. Andersen, Irene Lurie and Sauwakon Ratanawijitrasin
(1990). “Simulating the dynamics of Social Program Management: The Case of JOBS
Implementation” Modelling For Management II: Simulation in Support of Systems
Thinking. Richardson (ed.) pp. 155-174, Brookfield, Vermont, Dartmouth Publishing
Company Limited.
Richardson, George P. (1991). Feedback Thought in Social Science and Systems Theory.
Philadelphia, PA, University of Pennsylvania Press.
Richardson, George P. (1991). “System Dynamics: Simulation for Policy Analysis from a
Feedback Perspective”. Qualitative Simulation Modeling and Analysis. P.A. Fishwick and
P.A.Luker (eds). New Y ork: Springer Verlag, pp. 144-169.
Richardson, George P. & David F. Andersen (1995). “Teamwork in Group Model
32
Building’ System Dynamics Review 11(2): 113-137.
Richardson, George P. (1996). Modeling for Management: Simulation in Support of
Systems Thinking. Brookfield, Vermont, Dartmouth Publishing Company Limited.
Saeed, Khalid (1987). “A Re-Evaluation of the Effort to Alleviate Poverty and Hunger”.
Socio-Economic Planning Sciences, 21, pp. 291-304.
Savner, Steve & Mark Greenberg (1997). The New Framework: Alternative State
Funding Choices Under TANF. Washington, D.C., The Center For Law adn Social Policy.
Senge, Peter M (1990). The Fifth Discipline: The Art & Practice of The Learning
Organization. New Y ork, New Y ork, Currency Doubleday.
Smith, Vernon K. (1974). Welfare Work Incentives: The Eamings Exemption and Its
Impact Upon AFDC Employment, Earnings, and Program Costs. Lansing, MI, Michigan
Department of Social Services.
Southwick, Lawrence, Jr., (October, 1991). “Public Welfare Programs and recipient
Migration.” Southern Economic journal 40: 22-32.
Sterman, John D. (1988). “A Skeptic’s Guide to Computer Models” Managing a Nation:
The Microcomputer Software Catalog. Gerald O. Bamey, W. Brian Kreutzer and Martha
J. Garrett (eds). Boulder, Westview Press. pp. 209-229. Modeling For Management:
Simulation in Support of Systems Thinking. (1996) George P. Richardson (ed).
Brookfield, Vermont, Dartmouth Publishing Company. pp. 3-23.
Sterman, John D. (1994). “Learning in and about Complex Systems” System Dynamics
Review, 10: pp. 291-330. ; Modeling For Management: Simulation in Support of Systems
Thinking. (1996) George P. Richardson (ed). Brookfield, Vermont, Dartmouth Publishing
Company. pp. 89-128.
Steuerle, C. Eugene & Gordon Mermin (1997). Devolution as Seen from the Budget.
Washington, D.C., The Urban Institute.
U. S. Bureau of the Census (1993). County Government Finances: 1990-91. Washington,
D.C., U.S. Government Printing Office.
U.S. Congress, H., Committee on Ways and Means (1996). Summary of Welfare Reforms
made by public law 104-193.
U.S. General Accounting Office (1995). Welfare to Work : Most AFDC Training
33
Programs Not Emphasizing Job Placements. Washington, D.C.
Uccello, Cori E. & L. Jerome Gallagher (1996) General Assistance Program : The State-
based Part of the Safety Net. Washtington, D.C., The Urban Institute.
34
CC BY-NC-SA 4.0