The STELLA model makes use of a traditional
macroeconomics concept - the circular flow of income and spending - recast in
stock and flow terms as the nominal "demand side" of an economic system. The
"supply side" features a classic system dynamics representation of aggregate
flows of real goods and services--production and sales--buffered by an
inventory stock. Sales trends and inventory conditions provide information
feedback that affects production goals and employment factors of production.
With real sales driven by nominal spending, and with nominal income driven by
real production, the loop is closed between the demand and supply sides. The
full model consists of about 200 equations, including fiscal policy and
monetary policy sectors.
The model economy takes shape weekly as MacroLab
uses STELLA's "story-telling" feature to display and explain the structure of
sectors being added to the model, and suggest new behavior that can be
expected. After reading the story, students participate in simulation
activities that compare behaviors of evolving structures while exploring
traditional macroeconomics topics. They answer questions designed to assess
understanding of model structure revealed in the story, they model behavior
observed during simulation experiments, and they explore the connection between
structure and behavior. Students post their work online and receive instructor
email feedback soon thereafter.
The GDP experimental model is a small, simplified portion of the full model.
The purpose of the GDP experiment is to compare the learning that takes place
with different methods of delivering essentially the same information about
gross domestic product to two groups of students. The two delivery methods are
(1) simple narrative only, and (2) the same narrative, accompanied by model
structure revealed in stages. During the past year, 126 subjects in two states
participated in trial runs of this experiment. At Harvard Public Schools in
Massachusetts, 68 junior and senior economics students participated, thanks to
Larry Weathers, who is actively engaged in implementing systems thinking and
system dynamics modeling at his school. In Virginia, the 58 participants were
political science and macroeconomics students at Virginia Western and Dabney
Lancaster Community colleges.
Subjects were randomly assigned to a Control Group and Experimental Group and
pre-tested. Later, each group used the mechanics of STELLA's story-telling
feature--clicking on a button that activated a "story" about GDP, and pressing
the spacebar to progress through the story. Each page in the story read by the
Control Group contained text-only information about the meaning of GDP, its
measurement, and how it fits in an overall economic system as a concept of
"production." Experimental Group subjects read a story with the same textual
information, but their story was accompanied by an unfolding stock-and-flow
diagram that revealed the structure of the simple economy.
The test questions were simple, but were intended to probe three different
types of learning. One was a straightforward factual question, where merely
recalling information contained in the story would produce a correct answer.
Another sought awareness of an analogous stock-and-flow relationship, which
required a higher-order level of thinking than mere recall. No such analogy was
stated in the narrative information read by the Control Group. Nor was there an explicit
analogy accompanying the "unfolding model structure" observed by the
Experimental Group participants, but they did see the stock-and-flow
relationships in Figure 1.
The top relationship (A) refers to the flow of dollars from business bank
accounts to household bank accounts when income payments are made to the
factors of production, and the return flow of dollars when those households
purchase goods and services from businesses. The bottom relationship (B) refers
to the flow of goods and services in the production process into an inventory
stock that is reduced by subsequent sales. The textual narrative (identical for
both groups) mentions both types of stock and flow relationships, but not in
the same context and not in a way that would imply an analogous relationship. However,
the visual analogy implicit in Figure 1- part of STELLA's unfolding model
feature - was available to the Experimental Group.
Another interesting question probed the participants' sense of dynamics - their
grasp of GDP's behavior over time. The narrative available to both groups
emphasized that GDP is the production of final goods and services. Also, both
groups received a series of identical statements that traced the impact of
production on income, the impact of income on sales, and the effect of sales on
production, with the conclusion that "...production, income, and sales are part
of a mutually-reinforcing process." However, the Experimental Group had access
to an unfolding reinforcing feedback loop that accompanied that narrative
(figure 2).
The results of this trial experiment are consistent with the hypothesis that
the Experimental Group - with access to STELLA's unfolding model feature -
would learn more. Overall, participants in the Experimental Group gained an
average of 30 points between pre- and post-tests, compared to an average gain
of 18 points by those in the Control Group. For both groups, the pre-test
scores were about 55, but the post-test scores were about 85 and 73 for the
Experimental Group and Control Group, respectively. Thus, the Experimental
Group appears to have learned more. Use of a one-tail t-test for statistical
significance suggests that the probability of this result occurring by chance
is less than 7 percent. The learning gains are displayed graphically in Figure
3.
Over forty years ago, educational psychologist Jerome Bruner concluded that
"...the most basic thing that can be said about human memory...is that unless
detail is placed into a structured pattern, it is rapidly forgotten." Bruner's structured
pattern
is what Jay Forrester means by a system dynamics "framework" where facts can be
placed so that learning becomes more relevant and meaningful. In the GDP
experiment, both groups received behavioral descriptions, but only the
Experimental Group was able to see diagrams of structure that could infer such
behavior.
