Simulation-based research Contents 1 Introduction... 1 2 Application of simulation research... 1 3 Outline of simulation research... 2 4 Strengths and weaknesses of simulation research... 4 5 References... 5 1 Introduction This supplement expands the introduction to simulation-based research studies in Chapter 6. Along with additional details of how to carry out a simulation study it offers an overview of the design s strengths and weaknesses, along with additional references for further reading. In Chapter 5 we introduced simulation as an option for studying how complex processes behave dynamically. In this context, a simulation can be defined as a method for using computer software to model the operation of real-world processes, systems, or events (Davis et al. 2007: 481). In simulation studies of this kind, the researcher develops a model of the phenomenon under investigation and then chooses an appropriate simulation method. The model specifies the concepts of interest and the theoretical logic (Davis et al. 2007: 481) that links them together so that they can be coded into computer software. The model can then be run many times under various conditions to observe the outcomes. In this sense, simulations are sometimes seen as akin to virtual experiments although, as Gilbert and Troitzsch (2005) point out, in a simulation the researcher is experimenting with the model rather than the actual phenomenon. 2 Application of simulation research Most of us have encountered computer simulations being used in training, teaching or entertainment. In research, Gilbert and Troitzsch (2005) claim that simulation can be used to help us get a better understanding of a phenomenon of interest and for the purposes of prediction, for example, when modelling demographic change or the behaviour of a planned Management Research: Applying the Principles 2015 Susan Rose, Nigel Spinks & Ana Isabel Canhoto 1
production system. They also argue that simulation is valuable for social science as a tool for formalising theory. According to Gilbert and Troitzsch the process of specifying and building the computer simulation which involves being precise about what the theory means and making sure it is complete and coherent, is a very valuable discipline in its own right (Gilbert and Troitzsch 2005: 5). Simulation, they suggest, has advantages over traditional mathematical modelling when the interest is in processes and mechanisms rather than associations between variables. In addition, as we noted in Chapter 5, simulation may be particularly suitable when dealing with complex, non-linear phenomena. In research terms, simulations have long been employed in process analysis and evaluation in operations research, investigating diverse topics such as patient appointments in healthcare, supply chain dynamics and production control systems. Whilst their application to general management research has been more restricted (Berends and Romme 1999), Davis et al. (2007) argue for the value of simulation in theory development in strategy and organisations. 3 Outline of simulation research In outline, a simulation study involves the following generic steps (Gilbert and Troitzsch 2005, Davis et al. 2007) as shown in Figure 1: Research question. Identify a research question that is suitable for study by simulation. Model design. Model design involves specification of the target to be modelled in the simulation and the selection of an appropriate simulation method. There are a number of different methods from which to choose depending on the problem being investigated. Model design will usually involve some data collection to inform the parameters for the model and the initial conditions for the simulation. Model building. The next step is building the simulation model. A number of software programs are now available to support specific simulation methods but if no suitable software package is available you will have to write the program yourself. Model verification. Verification involves running the simulation and testing whether or not the model is working as it should. If there any problems with the simulation these should be corrected. Run the simulation. Simulations can be thought of as virtual experiments during which you run a series of experiments under different conditions that can be varied Management Research: Applying the Principles 2015 Susan Rose, Nigel Spinks & Ana Isabel Canhoto 2
as required (Davis et al. 2007). Harrison et al. (2007) identify five elements to such an experiment: the initial conditions, the time structure, outcome measurement, the number of iterations and any variation in model parameters or initial conditions. Variation allows different assumptions to be tested in order to answer the research questions and also to test the sensitivity of the model to changes in parameters. Model validation. Validation involves confirming that the simulation is a good representation of the chosen target. This can be done by comparing results of the simulation with empirical data. Validation can be a challenging process due to the nature of simulation and potential limitations on available empirical data. Nevertheless, as Pidd (2004) highlights, it is important that the model is sufficiently credible that people are confident to act on the insights it produces. Credibility, he suggests, is established over time by the model-building process, the actions of the researchers and the insights offered by the simulation. Findings and conclusions. As with other research designs, your findings and conclusions should be formulated in response to the research questions and the results should be disseminated. Gilbert and Troitzsch (2005) note that providing enough detail for the study to be replicated while avoiding burying the reader in detail can be a particular challenge when reporting simulation research. Figure 1 Steps in a simulation-based study Management Research: Applying the Principles 2015 Susan Rose, Nigel Spinks & Ana Isabel Canhoto 3
Given both the range of simulation techniques and their specialised nature, we do not cover simulation approaches in detail. See the references at the end of this chapter for additional sources on simulation techniques. Research in practice 1 presents an example of a simulation study. Research in practice 1 Example simulation study A simulation study of managing uncertainty in supply chains Modern supply chains are complex networks of suppliers and partners on which firms rely to deliver goods to their customers in increasingly uncertain and cost-sensitive markets. Datta and Christopher s (2011) study sets out to investigate different information-sharing and coordination mechanisms for improving performance under such conditions of uncertainty. Their chosen method is a simulation study of a paper-tissue manufacturing supply chain using a technique known as agent-based modelling (ABM). In ABM, a system such as a supply chain is modelled as a collection of autonomous decision-making entities, called agents. Agents can represent distribution centres, planners or factories within the system. Each agent makes its decisions autonomously, based on a set of rules. In addition, over time, agents can learn and evolve. In Datta and Christopher s study, a simulation is run for five different configurations of the supply chain, a baseline model which represents the actual set up and four additional models with different configurations of coordination and information-sharing mechanisms. The baseline model is validated against historical data and by testing with experienced managers. The results from the simulations of each configuration are then compared. Findings showed that decentralised decision making and centrally coordinated material flow along with daily local stock and global inventory information based production planning, and increased shared-information based ordering decisions help in improving the performance of the make-to-stock supply chain in all aspects (Datta and Christopher 2011: 795). 4 Strengths and weaknesses of simulation research The key strength of simulation is its ability to support investigation of phenomena that are hard to research by more conventional means. Davis et al. (2007) highlight its potential, for example, to show the outcomes of interacting processes over time or in situations where empirical data are limited. In such situations simulation can have advantages over analytic statistical modelling of the kind typically used, for example, in correlational studies. Management Research: Applying the Principles 2015 Susan Rose, Nigel Spinks & Ana Isabel Canhoto 4
Simulation-based research does, however, face some significant challenges. As Harrison et al. (2007) point out simulation is vulnerable to misspecification of the model itself. Simulation can be technically challenging and mistakes can be made in developing the computer program. They also note that generalisation must be treated with caution beyond the parameters specified in the model. A further problem is that simulation methods are not as widely understood or accepted as some other, better-known research approaches. 5 References Berends, P. and Romme, G. (1999). Simulation as a research tool in management studies, European Management Journal, 17(6), 576 583. Datta, P. P. and Christopher, M. G. (2011). Information sharing and coordination mechanisms for managing uncertainty in supply chains: A simulation study, International Journal of Production Research, 49(3), 765 803. Davis, J. P., Eisenhardt, K. M. and Bingham, C. B. (2007). Developing theory through simulation methods, Academy of Management Review, 32(2), 480 499. Gilbert, N. and Troitzsch, K. G. (2005). Simulation for the social scientist. 2nd ed. Maidenhead: Open University Press. Harrison, J. R., Zhiang, L. I. N., Carroll, G. R. and Carley, K. M. (2007). Simulation modeling in organizational and management research, Academy of Management Review, 32(4), 1229 1245. Pidd, M. (2004). Complementarity in systems modelling. In: Pidd, M. (ed.) Systems modelling. Theory and practice. Chichester: Wiley. Management Research: Applying the Principles 2015 Susan Rose, Nigel Spinks & Ana Isabel Canhoto 5