Systems Simulation. Expectation formation in behavioral simulation models

Abstract

The nature and rationality of expectations are hotly debated in economics and management science. Expectations are usually portrayed in behavioral simulation models and system dynamics as adaptive learning proceeses. This paper presents a behavioral model of trend expectation formation. The model assumes expectations about the growth rate of a quantity are formed adaptively from the recent growth rate of the input variable itself. The model is then tested directly against actual forecasts in two quite different domains: short-term expectations of inflation and long-term energy demand forecasts. In both cases the model replicates the evolution of the expectations quite well over extended time periods. The results support the use of adaptive expectations and trend extrapolation. The results also suggest the presence of additional judgmental heuristics which can have dynamic and policy significance. In particular, there seem to be substantial conservatism in both inflation expectations and energy demand forecasts: forecasters systematically underestimate the growth rate of the input. Such conservatism are consistent with the empirical literature on judgment and decision making. The results show it is possible t o test the expectation formation processes assumed in behavioral simulation modele; implications for use of adaptive expectations in behavioral models when empirical data are unavailable are ale0 die-KEY WORDS: individual, organizational and societal systems; system dynamics; forecaets; inflationary expectations; energy demand; computer simulation; behavioral simulation models c*J EXPECTATION FORMATION IN BEHAVIORAL SIMULATION MODELShe nature and rationality of expecta-T tions are hotly debated in economics and management science. Expectation formation is particularly important in behavioral simulation models, a class of dynamic models which share the following characteristics (Simon, 1982;Cyert & March, 1963;Nelson & Winter, 1982;Forrester, 1961;Morecroft, 1983Morecroft, , 1985Sterman, 1985;Sterman & Richardson, 1985): (i) A descriptive rather than normative representation of human behavior. Decision making behavior is portrayed in terms of the heuristics and routines used by the I am indebted to the members of the MIT system dynamics group for their comments and criticisms, and to several anonymous referees for useful suggestions. Tim Schiller of the Federal Reserve Bank of Philadelphia supplied me with the Livingston data.190 Behavioral Science, Volume 32, 1987 actors in the system rather than as the behavior which maximizes utility.(ii) The limitations of human cognitive capabilities are explicitly accounted for in modeling behavior.(iii) The availability and quality of information is explicitly treated including possible bias, misinterpretation, distortion, and delay.(iv) The physical and institutional structure of the system is explicit, including organizational design such as task and goal segmentation, the stock and flow networks that characterize the physical processes under study, and lags betwee...

Simulating Dynamic Capabilities and Value Creation in Family Firms

“…This anchoring process is well rooted and documented in the management literature (Sterman, 1987;Lant, 1992;Schneider, 1992;Sastry, 1997). As discussed earlier, each construct can be accumulated over time; it can both increase, as well as decrease depending on the dynamics of the two corresponding flow variables.…”

Boom, Bust, and Failures to Learn in Experimental Markets

“…If desired production were higher than current capacity, additional capacity would be ordered to reduce the backlog. The decision rule for competitor capacity acquisition has been used extensively in simulation models and is well supported both empirically and experimentally (Senge 1980, Sterman 1987a, 1987b). …”

Results of a beer game experiment: Should a manager always behave according to the book?

“…The time parameter used in the expected loss term is ignored in our study because expectation formation is out of the scope of this paper. For expectation formation, see Sterman (1987b Alternatively, a weight coefficient can be used in the supply line adjustment term so that a single adjustment time can be used rather than using two separate adjustment times.…”