Page 1
Group Model Building to Support Welfare Reform
Part II: Dutchess C ounty
Robert Allers, Commissioner Robert Johnson
Department of Social Services Office of Temporary and Disability Assistance
Dutchess County New York State
David F. Andersen, Tsuey Ping Lee, George P. Richardson,
John Rohrbaugh, Aldo A. Zagonel-Santos
Nelson A. Rockefeller College of Public Affairs and Policy
University at Albany, SUNY
This paper continues the effort to document how group model building takes place at the
University at Albany, as a follow-up to last year’s report presented in Istanbul (Rogers et al.,
1997). The product of this work is a model elicited from groups of county managers responsible
for implementing federal and state mandates regarding welfare reform. Preliminary results
suggest that the group model building (GMB) method is a valuable tool to overcome the
complexity involved in the welfare system, to focus group discussions, and to help build and
sustain interorganizational relationships (IORs) among the public and private agents involved in
delivering social services.
Approaches to systems thinking (Senge, 1990; Richardson et al., 1994; Morecroft &
Sterman, 1994) and strategic planning (Bryson, 1995; Eden, 1989) increasingly are coming to
rely on the practice of building a system dynamics (SD) model directly with a group as a method
to accelerate a management team’s work (V ennix et al., 1992; Vennix, 1996). Drawing upon
nearly two decades of experience with decision conferences (McCartt & Rohrbaugh, 1995;
Reagan-Cirincione et al., 1991; Rohrbaugh, 1992), researchers at the University at Albany
recently have set out to document a particular approach to building SD models directly with
groups (Richardson & Andersen, 1994; Andersen & Richardson, 1997). These efforts have been
part of a larger movement both to document how various consultants build models with groups
(Richmond, 1987), as well as to document the outcomes of these efforts (Huz et al., 1997).
Welfare Reform in New Y ork State
In 1996 President Clinton signed into law the Personal Responsibility and Work Opportunity
Reconciliation A ct (PRWORA) that mandated broad changes in how welfare would be
administered in the United States. In brief, the act ended federal entitlement, and it limited
benefits to a cumulative five-year period per family --after which the family looses eligibility to
the federally-funded Temporary Assistance to Needy Families (TANF) program. The law also
switched the funding mechanism from matching- to block-grant, thus providing states and
localities with more managerial flexibility. In 1997 New Y ork State (NY S) Governor Pataki
proposed a package of reforms at the state level that would provide for the implementation of
federal reforms within the specific NY S context. The NY S reform maintained the “entitlement”
Page 2
by creating the Safety-net (SN) program, thus, respecting Article 17 of the NY S Constitution,
which mandates that the state provide for its poor.
As Governor's Pataki’s reform package was scheduled for debate in the NY S Legislature,
researchers within the NY S Department of Social Services (DSS) approached the University at
Albany with a proposal to undertake a series of group model building sessions designed to focus
on how local providers might respond to both federal and state reforms. (In NY S, it is the
counties that actually provide social welfare services, and those services must conform to state
and federal guidelines and funding policies.)
These group modeling sessions were scheduled to take place in one rural county, one
medium-sized county, and one large county. The purpose of these efforts was three-fold: (1) to
assist the three participating counties to think through welfare reform strategies using a SD
modeling framework, (2) to provide state policy makers with opportunities to observe the group
model building process and, hence, to lean how local communities were responding to state and
federal initiatives, and (3) to create a “management flight simulator” for welfare reform viewed
from the county level. The three NY S counties directly involved in this project were Cortland,
Dutchess, and Nassau.
Drawing upon the collective knowledge of “management teams,”! the “modeling team” ”
at the University at Albany produced a series of SD models. In Cortland, the focus of our
discussions was on TANF; in Dutchess, it was on the safety-net. In preparation for Nassau, we
developed a combined TANF and safety-net model. In building these models, we followed the
GMB method devised in Albany (Richardson & Andersen, 1994; Andersen & Richardson, 1997;
Rogers et al., 1997).
A Glimpse At The Model’s Structure
The full model contains 30 views and 642 equations, as illustrated in Diagram 1. The first seven
views describe the patterns of client-flow through the welfare system, as conceptualized by the
management teams in Cortland County and Dutchess County.
' The “management teams” are the practitioners engaged in the group modeling effort. In this project, the
participants were primarily, but not exclusively, high-level local managers of social welfare services. Other agencies
involved were Health, Mental Health, Labor, and Education. State personnel often observed the sessions and
sometimes participated more actively (in particular, when the combined TANF and safety-net model was being
built). On occasion, a local legislator was present. Some non-governmental actors also participated (in particular,
representatives of private charities). At each site, the results of the modeling work were presented to a broader group
of community leaders.
? Richardson and Andersen (1995) refer to the “modeling team” in a GMB session as ideally composed of
individuals performing five roles: (1) facilitator, (2) modeler/reflector, (3) process coach, (4) recorder, and (5)
gatekeeper. Except for the gatekeeper, a role filled by “a person within, or related to, the client group” (p. 115), all of
the other roles are performed by Rockefeller College faculty and students. In addition to these five roles, this
particular intervention had a model builder (6), and it was supported with model calibration (7) and interface
development (8). On occasions, the modeling team was joined by a substance expert (9), Irene Lurie, an economist
who is also a specialist in social welfare policy.
Diagram 1: Sample view of the model
phs
Output He
Ty we
mL ell eM
TANF Sector
Safetynet Sector
High-need Families
Low-need Families
PostT ANF Families
Individuals
Other Families
Populations on Welfare System
Co-flow for Time Used-up
Fr. losing TANF eligibility
Behavioral Effects of WR
{tor Bs
Unit oost of BSs
L——
Totelfect of
attact onnew entie
Perception delay
Page 3
“*~—_| TANF & Diversion payments
Prevention Services
Child Support Enforcement
TANF Assessment & Monitoring
Effects of Wotk Patticipation Targets (TANF)
Time to set
egional cost Pevoeived celatve
attractiveness
N
Tot effect ofl
attract on fr departing
andi leaving
i
to samomnd tami
TANF Services to High-need Families
| TANF Services to Lowneed Families
. Sel-suficiency Promotion ‘ona ses otras
Safety-net Monitoring
Effects of Wotk Participation Targets (SN)
[Employment Services on the SN
|Job Maintenance Services
i} Emergency Services
|Actual Expenditures
Financial Shares
[Effects From Unemployment Rate
New Entiants Scenario
Present Evpenditures
4 New" 2
‘#2 Buld_[ Bl Baxc Sevess —] Ties New Ronan 12] [ols A SSI 22
Onit cost of
BSS rin
——
BSebaremin
cost of
/<Samotioned "PD fmeiat
+]
Diagram 2 is a summary view of the “stock-and-flow” structure. The model contains a TANF
sector and a safety-net sector. Within each sector, the clients are distinguished according to:
TANF high and low-need families; and post-TANF families, individuals and other families on
the safety-net.?
The main stock-and-flow dynamic of both sectors is:
¢ the inflow from the mainstream economy (and in TANF, the enrollment into the
program);
¢ the “pumping” of clients on assistance to employment outcomes; and
¢ the subsequent return of clients on to the rolls via recidivism (and in TANF, also via
going back at risk).
The consequence is that, at any one point in time, only a fraction of the clients served actually
make it back into the mainstream economy.
Two other stock-&-flow processes were important to the Cortland and Dutchess
management teams: (1) the possibility of diverting many of the potential TANF enrollees, and (2)
sanctioning, as an important monitoring tool.
3 “Post-TANE” families are those families that exhausted their TANF eligibility (after five cumulative years on the
TANF rolls) and, therefore, fell into the NY S safety-net program. The “individuals” constitute the clientele of the
existing home-relief program, now incorporated into the safety-net. “Other” families are those families that do not
qualify for TANF, such as legal aliens.
Diagram 2: TANF and safety-net Gient “stod<and-flow”
Page 4
Page 5
Diagram 3: TANF dient “stock-and-flow” with resource impacts and unemployment soenario
Child Support
Enforcement.
Self Sufficiency
Promotion
® Employment
‘Unemployment rate Services to High
, Need Families
Assessment and
Monitoring
Employment
Services to Low
Need Families
Page 6
Diagram 4: Safety-net Gient “stod<and-flow” with resource impacts and unemployment scenario
Job Maintenanc
Services
Child Support
Enforcement
Basic
Services
Employment
Services
Page 7
The remaining views of the model contain the formulations of the resources and scenarios that
control the rates-of-flow in the stock-and-flow structure, and compute the costs of the welfare
system.
Diagram 3 illustrates how the resources and scenarios affect the rates-of-flow in the
TANF sector. In the “front-end” of the system, prevention resources help to reduce the inflow of
families falling from the mainstream economy, into poverty. In the “middle,” TANF assessment
and monitoring promotes diversion programs, helps clients get jobs and other services needed to
move into employment, and sanctions clients who are not complying with drug-free regulations,
and/or mandates to participate in work preparation and work-fare. Employment services, in turn,
help to “pump” clients into paid jobs.
In the “back-end,” self-sufficiency promotion helps families maintain the jobs they obtain,
move-on to better jobs and, eventually, transition back into the mainstream economy. Finally,
child-support enforcement has wide-spread effects in the welfare system: it prevents families
from becoming at risk; it helps them to depart the rolls; it leads to more sanctioning; and it helps
some single-parent employed families rise above poverty level.
As illustrated in Diagram 4, child-support enforcement has similar effects for families in
the safety-net sector. And, except for prevention, all other resources have counterparts in the
safety-net. These are: monitoring, employment services, and job maintenance services.
In addition to resources, there are also scenarios controlling the rates-of-flow. For
example, unemployment has wide-ranging effects on the rates. Second, there are environmental
responses, and migratory dynamics. Third, “step” and “ramp” changes can be made to the inflow
from the mainstream economy. When simulating the model:
1. funds can be reallocated across resources;
2. scenarios can be changed;
3. different budgeting mechanisms can be chosen (from rigid to flexible); and/or
4. modeling assumptions regarding client behavior can be changed.
The model computes the costs of managing the system, which includes funding the above
mentioned resources, as well as paying for “census-driven” costs --such as diversion services and
payments, TANF payments, basic services, and emergency services. In all, there are 13 spending
categories. The rate of loss of TANF eligibility is controlled by a “co-flow” structure that mirrors
the TANF sector of the model, and that keeps track of the time used-up by the clients on the
TANF rolls.
In sum, the richness of the model permits investigation of system performance through
testing a wide-range of policies. Within its boundary, the model also helps to identify
interdependencies within the welfare system.
Page 8
The “Base-run”*
Graphs 1A and 1B contain the base-run behavior of eight variables. The simulation suggests that
there will be a shift in client populations, in particular after the year of 2002, when TANF
families begin losing federal eligibility. The size of the TANF and safety-net rolls converge.
However, overall, the total populations on welfare rolls does not change significantly. The model
shows a spike in the rate of loss of TANF eligibility at the year of 2002, with a subsequent
stabilization at a lower level. The increased costs observed between 1998 and 2005 are a function
of rising mandated work participation targets (from 25 to 50 percent on TANF, and from 75 to 90
percent on the SN). State and local share of welfare expenditures rise.
Graphs 1A & 1B: The base-run of the model
Six Key Variables Populations on Welfare System & Local Expenditures
2,800. Famies 7560 Cass
200 Cases
240 FamilieyY ear men: ieee a a 6411 M Dollars
31M_ Dollars -=— aaa
a Dineses [Loe F =
ae ee
71495 Cass
Families Se} sat datas
Case uae
0
0
0 Familiey ear
M Dollars
43. Dimensionless
7430 Cases [| 3
1997 2000 2002 2005 2007 6142M Dollars [717+ - a
Time (¥ ear) 1997 2000 2002 2005 2007
Time (Year)
Populations on welfare system - BASE
Local DSS expen- BASE |=» ————————._ Dollars.
Graph 1B portrays a discouraging view of welfare reform, in which the total populations in the
welfare system (sum of all of the population stocks) actually rises by about two percent, while,
welfare expenditures funded at the local level of government rise by approximately four and a
half percent (between 1997 and 2007). The reason why the model projects this rise is because the
safety-net populations are characterized by longer lengths-of-stay, and higher recidivism than
TANF populations. Thus, the shift in clients from TANF to safety-net, due to loss of eligibility,
results in a larger and more costly welfare system.
Improving System Performance
In order to improve system performance, the local Department of Social Services could resolve
to make investments in its current services. In general, DSS has command over resources
expended in the “middle” of the welfare system, such as assessment and monitoring, TANF
diversion, TANF payments, basic services, and employment services. But, DSS has limited
influence over the “front-” or “back-end” of the system. Other actors in the community --public
and private-- are better positioned to improve system performance in the areas of prevention and
self-sufficiency.
“The numbers in the graphs reflect what would be happening to a medium-sized NY S county that did not
implement any policy strategy, under a stable economy over time, and future client behavior --lengths-of-
stay, recidivism, etc.-- conforming with historical patterns of behavior.
Page 9
The two policy nuns contrasted below compare a DSS-only to a community-wide
investment strategy. In both cases, the investments are of the order of an additional ten percent in
the intensity of services (dollars/client). But, while the DSS strategy involves enriching only the
middle of the system, the community-wide strategy expands the effort to the front- and back-
ends. Of course, the community-wide simulation presumes that all social services agents can
agree upon a “coordinated” strategy.
Graphs 2A and 2B compare the two strategies, to each other, and to the base-run. The
simulations suggest that isolated DSS invests can improve system performance. However, only a
community-wide effort is capable of “deflating” the welfare system.
Graphs 2A & 2B: Comparative graphs
Graph for Populations on welfare rolls Graph for Populations on welfare system
23530 =e 7.360 me
= 4
et wae
2,482 oi - = 7515 a
S w4.° P “|
2435 7a70
~S PA
N Se ser file. rd
2387 _ 7425 Cee
5 ;———_
2,340 Pee | 7,380
1997 7000 7002 2005 2007 1997 2000 2000 2005 7007
Time (Year) Time (¥ eat)
Populations on welfare rolls - BASE Populations on welfare system-BASE == == += 55255 e5- Cases
Populations on welfare rolis- DSS. — — — — Populations on welfare system-DSS_ — ———————- Cases
Populations on welfare rolls - COMMUNITY Populations on welfare system - COMMUNITY ————————._ Cases
As we would expect, DSS can achieve significant improvements in the middle of the system,
where it controls most of the resources. The comparative graphs for populations on welfare rolls
portray that most of the reduction in the rolls is achieved by DSS alone. However, a DSS-only
strategy does little to improve the welfare system as a whole, as suggested in the comparative
graphs for populations on welfare system (sum of all of the population stocks). Here, the
community-wide strategy is much more effective.
Discussion
This project in GMB to support welfare reform has produced a number of interesting insights.
Perhaps one of the most important implications of this work is that it suggests that a coordinated
community-wide welfare reform effort is more likely to produce favorable results. Developing
IORs --horizontally, in the community, and vertically, with NY S agencies-- to arrive at a broader
welfare reform strategy can enhance the quality of life of welfare recipients (helping many to
transition out of the welfare system and into the mainstream economy, as well as preventing
many others from coming in by falling at risk). As suggested in Graph C - Comparative graphs
for local DSS expenditures, such a strategy may be effective enough, as to offset the shifting
financial burden --from federal to state and local-- resulting from PRWORA and its federal time
limits.
Page 10
Graph 3: Comparative graph on local expenditures
Graph for Local DSS expen
6.411M
6.339 M
6.268 M |
a® - al
6.196 M Rar ree Ea
| 4 onl a
6.124M as i
1997 2000 2002 2005 2007
Time (Y ear)
Local DSS expen - BASE
Local DSS expen- DSS_ — — —
Local DSS expen - COMUNITY
Ina preliminary evaluation, the leadership of Dutchess County DSS stated that the group
modeling sessions provided an opportunity for representatives from several agencies within the
community to come together in one setting to plan for welfare reform. As discussions regarding
the complexity of social services programs took place, the value of computer support and
modeling was amply recognized. The model and the simulations permitted detail exploration of
the interactions between different scenarios and alternative welfare reform policies. Finally, the
experience to work with the University at Albany modeling team --in the process of model
conceptualization, formulation and calibration, for the purpose of examining the impact of
changes in the administration of welfare programs-- was both challenging and rewarding.
Photo 1: Dutchess C ounty (Day 4) - The “modeling” team presenting
the SN model to the “management” team that conceptualized it
Page 11
This work is continuing. The model is being refined based upon on-going discussions
with the management teams, and a management flight simulator interface is under development
for the combined TANF & safety-net model. The direction and scope of the work with Nassau
County --a large county neighboring New Y ork City-- is in the process of being defined.
Dutchess County is evaluating the prospects of conducting a community-wide resource allocation
conference. Procedural scripts to “roll-out” the model “cold” to other NY S counties, and
extraneous audiences are being developed. Three doctoral dissertations are underway based upon
this work: (1) exploring the group model building terms of building interorganizational
relationships; (2) investigating the financing of welfare reform; and (3) developing interfaces,
gaming, and learning environments. Finally, these interventions beg a formal evaluation to assess
the long-term impact of GMB in term of state and local implementation of welfare reform.
Addendum On Presentations And Other Products
There have been a number of presentations made on this project: to NY S Department of Social
Services; Cortland County and Dutchess County community leaders; System Dynamics ’97;
NYS Social Services County Commissioners; NY PWA - New Y ork Public Welfare Association;
Nassau County Department of Social Services; LINKS ’98 - Conference on State Governments,
State Universities, and the Public Interest; among others.
In addition, these are some of the other products available from this work:
1. A TANF model (See CPR - Cortland report, May 1997);
2. A series of systems thinking exercises based upon the “Chugwa” County case-
study in welfare reform (CPR - Case-study, June 1997);
A safety-net model (See CPR - Dutchess report, August 1997);
A TANF management flight simulator (See CPR - User’s manuals, Versions 1.0
& 2.0, September & November 1997);
5. A community-wide resource allocation conference for Cortland County (CPR -
Resource Allocation, September 1997);
6. The development of model calibration “scripts” and reports (CPR - Model
calibration report, October 1997; CPR - Parameter booklet, A pril 1998; Lee et al.,
1998); and
7. A combined TANF & SN management flight simulator (in progress).
These presentations and reports have played a useful role in communicating to people
who have not been directly involved in the group model building process. We expect to
find that the project's communications strategy with regard to the larger community will
play an important role in establishing an environment conductive to collaboration and the
promotion of IORs.
Page 12
References:
Andersen, D.F. and G.P. Richardson. 1997. “Scripts for group model building.” In System
Dynamics Review 13(2):107-129.
Bryson, J.M. 1995. Strategic Planning for Public and Nonprofit Organizations: A Guide to
Strengthening and Sustaining Organizational Achievement. Revised Edition. San
Francisco: Jossey-Bass.
Center for Policy Research - CPR. 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.”
(NY S Governor's Office of Employee Relations, The Leadership Classroom.)
.. August 1997. “Welfare reform project: Report of the group model building conferences
in Dutchess County.”
... September 1997. “Welfare Reform Project: Report of the Community-wide Resource
Allocation Conference in Cortland 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.”
Eden, C. 1989. “Strategic options development and analysis.” In J. Rosenhead, ed. Rational
Analysis in a Problematic World. Chichester: John Wiley & Sons.
Huz, S., D.F. Andersen, G.P. Richardson, and R. Boothroyd. 1997. “A framework for evaluating
systems thinking interventions: An experimental approach to mental health system
change.” In System Dynamics Review 13(2):149-169.
Lee, T.P., A. Zagonel, D.F. Andersen, J. Rohrbaugh and G.P. Richardson. 1998. “A judgment
approach to estimating parameters in group model building: A case-study of welfare
reform at Dutchess County.” 16th International System Dynamics Conference: July 20-
23. Québec City, Canada.
McCartt, A.T. and J. Rohrbaugh. 1995. “Managerial openness to change and the introduction of
GDSS: Explaining initial success and failure in decision conferencing.” In Organization
Science 6(5):569-584.
Page 13
Morecroft, J.D.W. and J.D. Sterman, eds. 1994. Modeling for Learning Organizations. Portland,
Oregon: Productivity Press.
Reagan-Cirincione, P., S. Schuman and G.P. Richardson. 1991. “Decision modeling: Tools for
strategic thinking.” In Interfaces 21(6).
Richardson, G.P. and D.F. Andersen. 1995. “Teamwork in group model building.” In System
Dynamics Review 11(2):113-137.
Richardson, G.P., E.F. Wolstenholme, and J.D.W. Morecroft, eds. 1994. System Thinkers,
Systems Thinking, special issue of the System Dynamics Review 10(2-3).
Richmond, B. 1987. The Strategic Forum. Hanover, NH: High Performance Systems.
Rogers, J. et al. 1997. “Group model building to support welfare reform in Cortland County.”
15th International System Dynamics Conference: August 19-22. Istanbul, Turkey.
Rohrbaugh, J. 1992. “Cognitive challenges and collective accomplishments: The University at
Albany.” In R. Bostrom, R. Watson, and S.T. Kinney, eds. Computer Augmented
Teamwork: A Guided Tour. New Y ork: Van Nostrand Reinhold:299-321.
Senge, P.M. 1990. The Fifth Discipline: The Art and Practice of the Learning Organization.
New Y ork: Doubleday/C urrency.
Vennix, J.A.M. 1996. Group Model Building: Facilitating Team Learning Using System
Dynamics. Chichester, England: John Wiley & Sons.
Vennix, J.A.M., D.F. Andersen, G.P. Richardson, J. Rohrbaugh. 1992. “Model building for
group decision support: Issues and alternatives in knowledge elicitation.” In European
Journal of Operational Research 59(1):28-41.
