Using Dynamic Simulation for Resource Management Policy Design at the
Minnesota Department of Transportation
Kristina Wile David Smilonich
Associate District Engineer
GKA Incorporated Minnesota Department of Transportation
125 CambridgePark Drive Highway 10 W
Cambridge, MA 02140 Detroit Lakes, MN 56502
ABSTRACT:
This paper describes the application of systems thinking and system dynamics
modeling for the development of improved resource management policies in the
Minnesota Department of Transportation. The structure explored was that of one of
twelve operating units responsible for maintenance of ~1600 miles of highway. Family
testing has shown its applicability to the other operating units.
The paper identifies some of the policy insights and learning that occurred during
model building and testing, as well as notes about process techniques. Insights
included the effectiveness of hiring policies, side-effects of outsourcing strategies,
effectiveness of training and cross-bargaining of human resources. Process techniques
included group model testing, strategy and scenario building to frame model purpose
and testing, the use of learning laboratories using a simulation model to engage the
leadership of the organization, and a management flight simulator to disseminate the
insights gained to the ranks in the organization.
BACKGROUND:
This organization has a deeply functional culture, and as many government agencies,
faces some fiscal uncertainty. The project, sponsored by the district's top executive and
other department managers, was intended to unfreeze and broaden the perspective of
the management for more effective policy-making, and to leverage any learning
throughout this district and across the wider organization. The clients were very
involved in every aspect of building their model.
OVERVIEW OF POLICY INSIGHTS:
Effectiveness of hiring policies
In the organization there is pressure to hire people whenever discretionary money is
available. The organization is split into Construction workers and Maintenance
workers. Managers of the two functions lobby vigorously to hire into their particular
function, with a goal of maintaining headcount at the local level.
SU
In the modeling process, the team posited a more systemic ways to determine how
many people are needed. Specifically, the size of the backlogs of work which are
driven by the performance measures were identified as key drivers.
Inputs Transformation Outputs Outcome
- maintenance people - construction projects - pavement quality - customer
- program people - maintenance work - roadside service level satisfaction
MNDOT
= District 4 = Ke
Zz M1
Local view Systemic view
im ! |
Side-effects of outsourcing strategies
Outsourcing of maintenance work can be done on a short term basis to handle peaks in
the workload. The backlog is decreased, and discretionary dollars are allocated to it.
Because the backlog is reduced, the suggested staffing level is decreased slightly.
Additionally, available discretionary funding is drawn down making longer term
financial commitment to hiring permanent staff impossible. Because the financial
hurdle to outsource this work is always lower than that to hire a permanent person,
outsourcing can become a slippery slope.
Allocation
S Hiring g
\
Budget
Backlog
‘o> TOSS
Sere bargaining
The team realized that although outsourcing is significantly more cost effective than
doing this work internally, and seems the obvious answer, choosing which work to
outsource has long term strategic applications.
So
iminishing re! a ini an resource:
Cross-bargaining, or flexible workforce, is becoming a popular notion in the
organization as a way to increase the efficiency of the organization. Through the testing
of the model, the team noted that 20% cross-bargaining capability captured all the
efficiency of a completely flexible workforce. The benefit was due largely to the out of
phase seasonal peaks in the two kinds of work.
This result has implications for a plan to train the entire workforce to align the skill base
with market salary levels. This policy would not have the intended benefits unless the
organizational strategy involved increasing the amount of construction work or
changing the balance of work.
namic understanding of the effects of different reconstruction project types
The process of building the model helped to clarify the relationships between different
types of construction projects and their effect on pavement quality. Although the
organization has a management information system that recommends projects types for
particular stretches of highway, the logic embedded in this system was not well
understood. The different project types, each having different costs, delays in
construction, and wear characteristics are effective in different situations.
The team also noted that if construction funds were increased substantially, it would
cause the pavement quality to increase significantly, and overall spending, although
higher in the short term, would be lower in the long term. This would result in large
fluctuations in workforce levels, however, and require a shift in policies to prevent a
long term oscillation.
Effects of fragmented budgets
Minnesota has been decreasing the amount of fragmentation in funding to the agency.
This has increased the need to understand how best to allocate those resources.
The team's testing showed that even more flexibility in allocating both operating dollars
and construction dollars could improve organizational efficiency even further. One
underlying assumption for this conclusion is that management at the local levels has a
dynamic understanding of their system and environment.
PROCESS TECHNIQUES:
The client team had much of the responsibility for testing, validating, and calibrating
the model. There were six people on the core modeling team who were from different
parts of the organization and at different levels of responsibility. The roles on this client
team included the gate keeper, the internal modeler, and content experts. Consulting
associates filled the facilitator and modeler/reflector roles. Stakeholders and other
experts were included at many points in the process.
si
woup my testing for validati
The team invited the management staff of the organization to take part in testing the
results and assumptions in the model. Team members had frequent dialogue checking
particular assumptions with people in different areas of the organization.
Strategy and scenario planning to frame mi ind testin
The team used classical scenario planning tools to develop coherent strategies for the
organization: --They also identified key uncertainties and-developed scenarios from
them. They were able to use the model to test the strategy scenario combinations to get
a better understanding of their multiple possible futures.
earning laboratories to engage leadership of orga i
In order to determine whether the structural assumptions in the model were valid for
other districts in the organization, the top executives from all the districts participated
in a group model explanation and demonstration where they experimented, questioned
assumptions, and validated the behavior. This meeting helped to build the confidence
of these leaders, which will facilitate the roll-out to the wider organization.
Management flight simulator development to disseminate insights
The team learned a tremendous amount about their systems and environment. In order
to leverage that learning to the rest of the district and the wider organization, a
management flight simulator was developed to allow teams to experientially learn some
of what the core modeling team learned during development.
REFERENCES
de Geus, A. P., Modeing to predict or to learn?", European Journal of Operational
Research 59(1992), pp. 1+
Forrester, J. and Senge, P., "Tests for Building Confidence n System Dynamic Models,"
System Dynamics TMS Studies in Management Science, Vol. 14, pp. 209-228, (1980)
Homer, Jack B., "Partial-model testing as a validation tool for system dynamics,"
Proceedings of 1983 International System Dynamics Conference, Boston, MA, pp. 919-
931
Isaacs, William, Senge, Peter, ‘Overcoming limits to learning in computer-based
learning environments," European Journal of Operational Research, 59(1), pp. 183-196
Lane, David C., "Modelling as Learning," European Journal of Operational Research,
59(1), pp. 64-84
Richardson, George P., Andersen, David F., "Teamwork in group model building,”
System Dynamics Review, Vol. 11, no. 2 (Summer 1995). pp. 113-137
Senge, P., Ross, R., Kleiner, A., Roberts, C., Smith, B., The Fifth Discipline Fieldbook:
Strategies and Tools for Building A Leaming Organization Doubleday, 1994
Sterman, John D., "A behavioral model of the economic long wave," Journal of
Economic Behavior and Organization, 1985, 6, pp. 17-53
Sy
