Developing Theory Through Simulation Methods

Abstract

We describe when and how to use simulation methods in theory development. We develop a roadmap that describes theory development using simulation and position simulation in the "sweet spot" between theory-creating methods, such as multiple case inductive studies and formal modeling, and theory-testing methods. Simulation strengths include internal validity and facility with longitudinal, nonlinear, and process phenomena. Simulation's primary value occurs in creative experimentation to produce novel theory. We conclude with evaluation guidelines.Simulation is an increasingly significant methodological approach to theory development in the literature focused on strategy and organizations (e.g., Adner, 2002;Lant & Mezias, 1990;Repenning, 2002;Rivkin & Siggelkow, 2003;Zott, 2003), Indeed, several influential research efforts (e.g., Cohen, March, & Olsen, 1972;March, 1991) have used simulation as their primary method. Yet while simulation has become an important methodology, its value for theory development remains clouded and even controversial.On the one hand, some argue that simulation methods contribute effectively to theory development. For example, simulation can provide superior insight into complex theoretical relationships among constructs, especially when challenging empirical data limitations exist (Zott, 2003). It can also provide an analytically precise means of specifying the assumptions and theoretical logic that lie at the heart of verbal theories (Carroll & Harrison, 1998;Kreps, 1990). In addition, simulation can clearly reveal the outcomes of the interactions among multiple underlying organizational and strategic processes, especially as they unfold over time (Repenning, 2002). From these perspectives, simulation can be a powerful method for sharply specifying and extending extant theory in useful ways.On the other hand, some researchers maintain that simulation methods often yield very little in terms of actual theory development. They suggest that simulations are simply "toy models" of actual phenomena, in that they either replicate the obvious or strip away so much realism that they are simply too inaccurate to yield valid theoretical insights (Chattoe, 1998;Fine & Elsbach, 2000). For example, simulation research is usually based on at least some clearly unrealistic assumptions, such as zero search costs (Rivkin, 2000) and all strategic rules are effective (Davis, Eisenhardt, & Bingham, 2007). In addition, simulation constructs are often "measured" by empirically distant approaches, such as "0" and "1" bit strings as representations of organizations (Bruderer & Singh, 1996) and strategies (Rivkin, 2001). The results of research using simulation methods can also be dynamically indeterminate and overly complex (Fichman, 1999). From these perspectives, the value of simulation methods for theoretical development is tenuous.The controversy surrounding the value of simulation methods for theory development partially arises, in our view, from a lack of clarity about the method and its related link to th...

Agent-based modeling in managerial science: an illustrative survey and study

“…This reduces the fear that the model is rigged to produce the desired result'' (Rivkin 2001, p. 280). Davis et al (2007) provide an extensive overview of structured simulation approaches used in the area of managerial science-for simulation studies in general, not only for ABM-and, in particular, mention system dynamics, NK fitness landscapes, genetic algorithms, cellular automata and stochastic processes. From these approaches, cellular automata and NK fitness landscapes are outlined more into detail in the following-cellular automata since, according to Dooley (2002;see also Fioretti 2013), they could be regarded as a rather simple form of ABM and thus are helpful to demonstrate the idea of ABM, and NK fitness landscapes since they were widely used in agent-based models in managerial science (see Sect.…”

“…Beyond these rather ''formal'' classifications, Davis et al (2007), in studying the potential contributions and pitfalls of simulation methods for theory development, argue that the choice of the simulation approach is crucial ''because of its framing of research questions, key assumptions, and theoretical logic'' (p. 485). The authors distinguish between structured and non-structured simulation approaches.…”

“…While cellular automata mainly allow the mapping of evolutionary processes, NK fitness landscapes might be regarded as an approach for representing adaptive search and optimization processes (Davis et al 2007). NK fitness landscapes were originally developed in evolutionary biology (Kauffman 1993;Kauffman and Levin 1987) to study how effectively and how fast biological systems adapt to reach an optimal point.…”

“…if only direct neighbors affect each other). Agent-based models implemented via cellular automata mainly seek to study the emergence of macrolevel patterns resulting from local interactions like competition, diffusion or segregation in a set of agents (Davis et al 2007). For example, Lomi and Larsen (1996) apply cellular automata for investigating the characteristics of global dynamics in a branch (e.g.…”

“…The latter means that simulation is used to develop structural insights and gain understanding of fundamental processes within a certain area of interest (e.g. Davis et al 2007;Dooley 2002;Harrison et al 2007;Gilbert and Troitzsch 2005). Axelrod regards simulation ''a third way of doing science '' (1997a, p. 3) or a ''third research methodology '' (1997b, p. 17) which has in common with deduction the explicit set-up of assumptions (though it does not prove theorems by the ''classical'' mathematical techniques) and which generates data from a set of rules (rather than from measurement of real world data as it is typical for induction).…”