Research Paper

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/305639587

DECISION MAKING THROUGH SIMULATION IN PUBLIC POLICY

MANAGEMENT FIELD

Conference Paper · March 2016

DOI: 10.21125/inted.2016.0911

CITATIONS

0 READS

705

3 authors:

Some of the authors of this publication are also working on these related projects:

Bateratzen View project

Maria Ruiz

Mondragon Unibertsitatea

11 PUBLICATIONS   0 CITATIONS   

SEE PROFILE

Noemi Zabaleta

Mondragon Unibertsitatea

22 PUBLICATIONS   38 CITATIONS   

SEE PROFILE

Unai Elorza

Mondragon Unibertsitatea

10 PUBLICATIONS   32 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Maria Ruiz on 26 July 2016.

The user has requested enhancement of the downloaded file.

DECISION MAKING THROUGH SIMULATION IN PUBLIC POLICY MANAGEMENT FIELD

Maria Ruiz1, Noemi Zabaleta1, Unai Elorza1 1Mondragon University (SPAIN)

Abstract

Managing complex systems in which heterogeneous agents act according to non linear behaviors turns into a difficult task. Public policy evaluators are dealing with difficult problems to define efficient public policy management. Until now many methods have been used basing on archaic methodologies or intuitions with no empirical evidences. Simulation seems to be the best option to solve this problem and make efficient decisions in management fields. Due to its flexible and evidence based nature enables a better understanding of the whole system and the effects of their decisions. Moreover, multiple scenes could be simulated while avoiding risky real situations, these characteristics are considered powerful and key functionalities, converting simulation into the best alternative.

The whole simulation process is based on a hybrid framework, two techniques are combined, Agent Based Modelling (ABM), and System Dynamics (SD).This paper will present the context of social simulation, the necessity of evidence based empirical tools for decision making, and the simulation model as an alternative to the evaluation of public policies.

Finally, it could be stated that proposed process in this paper consists in three steps: Introduction to social simulation, proposed methodology for the evaluation (theoretical pillars, multi-paradigm framework, simulation techniques) and construction of the model.

Keywords: Simulation, Public policies, Management, Evidence-based, Agent Based Modelling, System Dynamics.

1 INTRODUCTION

A lot of research has been done around the topic of simulation, simulation is increasing as a method to develop theory about strategy and organizations; as a result many research efforts have used it and many definitions are formulated about the topic. One of the main advantages of simulation is that it can be helpful in the interpretation of complex theoretical relationships among constructs. It can also contribute in the specification of basic assumptions of some theories. Moreover, it can add insight regarding the interactions between different organizational and strategic processes. From these perspectives simulation can be an effective method for extending theory in different ways, as for example assessment. [1]

As a result of this power, simulation models can have several objectives, including: Testing of new ideas, predicting the impact of policy or technology, developing a theory, determining the need for a mechanism which specifically targets decision-making, predicting future directions, scenarios What-if, critics suggest experiments, evaluate the impact of different variables, assumptions or simulate any factor within a experiment.[2]

For the psychologists’ community, the objective of characterizing and theoretically understanding social and psychological phenomena (it is involved in public policies) needs a specific understanding and knowledge of such interactive and dynamic processes. Smith defined most commonly theory- building and modelling techniques in this field as not enough effective for this purpose. They believe that social simulation is able to capture and understanding those complex, dynamic, and interactive processes that are so important in the social world.[3]

There are characteristics that define simulation and are in contrast to the technical "traditional” ones, those characteristics are the following [4] :

• They have to deal with unstructured problems composed of multiple actors, multiple perspectives, conflicts of interest, more uncertainties and factors that cannot be quantified.

• Modeling should be the way to open alternatives.

• They and their associated models must be accessible to all stakeholders.

• They must be flexible and iterative.

The role of simulation is still not defined within the management community, but the simulation itself is legitimate, powerful and should be disciplined for scientific research in the field of management. [4]

2 METHODOLOGY

This chapter will explain the theoretical base of the model on the one hand and the development of the simulation model on the other hand.

2.1 Evidence based nature The main base of the simulation, as it has been mentioned before, is the evidence based nature that it should have in the field of public policies assessment. Usually, management and senior management provide solutions to problems based on outdated techniques learned earlier, without validating old methods and models learned from experience. In the field of medicine they have begun to seek, identify and implement new management methods that are clinically relevant. It's time for managers to begin to do the same. The risk at the organizational level is much higher, since it is not difficult for some people to be considered leading expert and therefore his/her decisions based on simple feelings are not even evaluated.[5]

According to Rousseau is necessary a more systematic approach to studying the complexities of organizational behavior, he believed that the simulation has unique advantages in this regard. This allows acquiring knowledge about the development of different theories and their consequences. [6]

Improvement of management skills is a direct target of the evidence-based management. Managers need real learning, not deception or false conclusions. In this way that senior management acquires systematic knowledge about how organizations govern human behavior and the risk of making bad decisions is reduced. Evidence is derived from valid learning and continuous improvement rather than gridded races based on false assumptions.[6]

Introne states that many fails related to decision making are an effect of human generalizations about facts, lack of evidence when making an inference and opt for easier and heuristic shortcuts. The model they present brings addition of evidence through a computational tool to engage evidence- based reasoning. It is also valuable for individuals in the creation of knowledge tools that can be used to manage problems in the same field. This computational engines for evidence based decision making require training and careful integration into daily routines in order to gain the maximum potential of the technique.[7]

2.2 Multi-paradigm framework In the healthcare field for example, there is no way to experiment in a real system different approaches before implementing them, because of the agents involved in the field of public policies it would suppose high costs and time. Organizational simulation, which is evidence based decision making social simulation in this case, can be the alternative to this gap and bring a wide range of solutions for improving this management context. Moreover, this learning could be applied in other fields. [8]

According to this evidence based needed framework and multi agent complex system in which the model is developed, multi paradigm has been chosen for the model, a paradigm defined by different agents acting and using different techniques (multi-method) and in which different agents act (multi- agent).A multi-method simulation is the one which uses different kind of techniques. Due to the existing limitations in the space, models are expected to be multi functional. This variety produces complexity, and this leads into more conflicts between different agents’ expectations. Multiple actors from different organizations and interest groups compete for the same resources in different scenarios and all of them should be taken into account in the public policy evaluation. This results in both conflict interests and the necessity of using multi method tools for the virtual representation of the system.[9]

Multi-agent models, however, are defined as the ones which are defined by individual heterogeneous agents. Indeed, they are used for analyzing “complex social systems”, especially because of the multiple interacting parts which composed the system and the non-linear behavior of the agents. Due to the nature of these models, they are applied to study a variety of policy domains. Researchers, on the other hand used these models because of reasons related to ethics, cost, timeliness and appropriateness.

As a result of the level of complexity, and dynamic non-linearities, multi-agent and multi-method techniques are considered the best option to examine and simulate this kind of environments.[10]

2.2.1 Simulation techniques

According to Carley the most common types of simulation techniques are Agent Based Modeling (ABM), System Dynamics (SD) and Discrete Event Simulation (DES). One example of ABM could be the model of “Colonist household decision making and land-use change in the Amazon Rainforest: an agent-based simulation”[11]. As an example of SD could be shown the case of Automobile Leasing Strategy in [12]. And finally, examples of DES related to social topics, are mostly related to operational part of the organizations, uses discrete-event simulation modeling to health care clinics and systems of clinics (for example, hospitals, outpatient clinics, emergency departments, and pharmacies). [13]

These three simulation techniques are major paradigm in the field of simulation; there are also dynamic, less known systems, because they are used to simulate physical processes. Technically DES and SD work more with continuous processes, while AB works more discrete times, that is crossing from one event to another.

Merriënboer et al. define SD as an holistic technique which assumes a higher level view of the whole project, focusing on human factors and managerial policies, also he remarks an inherent flexibility which enables modelers to incorporate a wide range of influences specific to particular scenarios [14]

System dynamics have a wide range of uses, Brailsford et al. describe an application of system dynamics to a very large, complex system: the whole delivery of emergency or 'unscheduled' care in the city of Nottingham, England [15]. Cooke et al. presents a dynamic model of the systematic causes for patient treatment delays in emergency departments[16]. Esensoy et al.explained a qualitative model for the Ontario Ministry of Health and Long-term care aimed to design patient flow policies, and to generate hypotheses on their in-tended and unintended consequences[17]. Ford and Sobek present a model to test the hypothesis that Toyota uses real options to switch among alternatives to operationalize set-based development and 2) propose and test a hypothesis of how real options at Toyota add value[18]. Dynamic Modeling of Product Development Processes is another area which work with system dynamics [19]

The Discrete Event Simulation (DES) focuses on the concept-based modeling entities and describing flows and share sources. The entities are passive objects that represent people, parts, documents, messages, etc. These entities are moved by the flow where they stay in queue, are delayed, they are processed, divided, combined, etc. Its scope is typically Service, production, logistics. In general, this technique can be defined as an algorithm of global entity, typically with stochastic elements [20]. According to Brailsford et al., DES simulation has been developed upon Monte Carlo methods, and is mostly settled in design and operation of manufacturing plants. DES models could be understood as queuing systems through time, and they are represented by entities, queues, activities and resources.[15]

ABM uncovers multiple interrelated agents and institutions which act responding to established rules. Those agents are skilled to learn and adapt to modifications in the environment and the method is focus on unique characteristics of individuals[21]. Bonabeau reinforces that theory, " ABM is modeled as a collection of autonomous decision-making entities called agents. Each agent individually assesses its situation and makes decisions on the basis of a set of rules. Agents may execute various behaviors appropriate for the system they represent[22]. Epstein defends the ability of agent-based modeling to change the social sciences in a variety of ways, especially facilitating generative explanation [23]. Heckbert et al define the ABM as a representation of autonomous entities, with specific characteristics as heterogeneity and dynamism, those agents' activities have macro scale effects to obtain quantitatively study complex systems[24]. Robinson emphasizes the power of ABM for generative modelers[25]. Thorne et al. state the possibility AB brings to observe phenomena that are not easily testable in the laboratory.

The Agent Based Modelling (ABM) has been used in different disciplines, including artificial intelligence, complexity science, game theory, etc. There are no universally accepted definitions in this area, and it is still under discussion which should be the characteristics of an agent to be considered as such, reactivity, spatial ability, learning ability, social skills, intellect, etc. [20]. Alderton et al. present a model of Human African Trypanosomiasis (HAT), or sleeping sickness using ABM[26]. Allen and Kanamori present a model to measure the potential for earthquake early warning in southern California[27]. Artel et al. build an agent-based model for the investigation of neo vascularization within porous scaffolds [28]. Conner et al. present a model to evaluate coyote management strategies using a spatially explicit, individual-based, socially structured population model[29]. Agent based simulations are also used in agriculture, as an example Colonist household decision making and land-use change in the Amazon Rainforest[11]. Reducing Moose-Vehicle Collisions through Salt Pool Removal and Displacement is the Agent-Based Modeling approach proposed by[30]. Janssen et al. presented an adaptive agent model for analyzing co-evolution of management and policies in a complex rangeland system[31]. Lin et al. presented an agent-based model to simulate tsetse fly distribution and control techniques in Nguruman, Kenya[32]. Palmer et al. build the diffusion of residential photovoltaic systems in Italy[33]. Robinson et al. modeled farmer household decision-making and its effects on Land use/cover change in the Altamira Region [25].

2.3 CASE STUDY: SIMULATION MODEL

Here the general steps followed to build the model will be explained. Apart from the representation of the process of each program, the model gives the opportunity to create different scenarios in order to evaluate the impact of each parameter in the number of requests and in the number of concessions.

The first step consists on collecting the needed data in order to build a model as close to the reality as it is possible. Information and details about deadlines, criterion, and phases of policy evaluation are crucial for the construction of the model.

A combination of Agent Based Modelling (ABM) and System Dynamics (SD) is used in the model where a detailed description of the process is shown taking into account applicant’s preferences in previous years.[34]

States’ diagrams are helpful to follow visually agent’s development using graphics and colours. In fact, thanks to the SD, agent’s exits and entries could be visualized. Moreover, parameters allow the control of fluxes, creating a dynamic and interactive model, with changes, and modifications simultaneously to the simulation.

Due to the modification of parameters, a comparison of different scenarios is possible (when more agents in the last phase, better macro results), what is helpful in order to know accurately program’s way of working and their clue evaluation parameters.

The process followed is:

 One SD built for each policy.  Addition of parameters and variables to the SD’s.  Addition of controllers to each parameter (Sliders), like this different scenarios with different

parameters are possible.  Construction of ABM diagrams (States, transitions and codes).  “Parameter variation” new scenarios.

For the SD model “Flow” and “Stock” elements are needed, one SD will be built for each policy and with the corresponding phases for each policy, Fig.1.

Figure 1: SD diagrams for each public program

Program A

Program B

The next step consists on adding parameters and variables in order to define the entries and data behaviour. Variables’ function was to add agents according to each company’s behaviour inside the same statechart. Each SD will have three variables and four parameters, each of them connected by a link. The function of the variables is the following, counting the number of requests, the number of evaluated requests and the number of accepted requests (changeable), whereas parameters have fixed values (average values) for each scenario. The following Table 1 shows the name and type of parameters and variables.

Table 1: Variables used

VARIABLES TYPE

Nwant (Potential agents number) Int

NSol (Requests ) Int

NCon (Accepted) Int

PARAMETERS

Solicitudesprom (Num. of requests) Double

Tiempoderec (Time of reception) Time

Teimpodeval (Time of evaluation) Time

Promotorgadas (Num. of accepted requests) Rate

The last step consists on adding controllers to each parameter, that is, slider elements with the purpose of allowing the modification of parameters and consequently, it would be possible to analyze models’ behaviour due to the change of those values. Each SD has its own assigned controller and the minimum and maximum values for its parameter.[20]

In order to define the ABM diagram an analysis was done where the different states passed during the process were the following:

1) Potential agents (Agents which could ask for that subsidy) 2) Request or not 3) Evaluation or not 4) Subsidy or not 5) Objective gained or not

The states will be added to the frame, they will be connected using transitions, and the codes inside the states will have the function of subtracting and adding agents according to the exits and entries, Fig.2 and Fig.3. Inside the state is the code to add the colour to the agents depending on the state in which they are in that moment, that is very helpful for the user.[35]

Figure 2: ABM diagram (process for each program request) Figure 3: Appearance of the general final ABM diagram

The next task is to analyze the different scenarios according to the parameter modification and the interaction between the two programs, this type of simulation is called Parameter Variation. First of all, a Data Set is built in the last stock in order to collect all the parameters that have impact in the results, time units are defined in order to know which time frame will be analyzed during the simulation, and finally, a Time Plot is added using a code in order to visualize the results.

As we have commented before, the SD system is useful to control and change internal parameters which could be determinant for the final result of the received subsidies, the following Table 2 shows the variable parameters existing in the model, and the measured outputs.

Table 2: Variable parameters and outputs

VARIABLE PARAMETERS OUTPUT

Number of requests (1st program) Number of accepted subsidies

Number of requests (2nd program) Number of accepted subsidies

Reception time (1st program) Number of accepted subsidies

Reception time ( 2nd program) Number of accepted subsidies

Evaluation time (1st program) Number of accepted subsidies

Evaluation time (2nd program) Number of accepted subsidies

Once the scenarios are built, the results which will be shown in the model are the grant subsidies for each of the programs, taking into account the indicators defined.

Thanks to an analysis about which parameters are used to control and performance, the results of the model are more accurate. Therefore, the selected parameters were: number of requests in programmes, evaluation period on each and reception period of time. The analyzed periods are composed of four years, being this cycle, the time corresponding to the possible govern change.

2.3.1 Results of the model

The final conclusion states that the stabilization regarding number of requests and number of accepted subsidies begins in the second year after the implementation, and the increase is continuous. Moreover, the principal conclusions could be itemized in the following:

 If the number of requests is increased in the second program, the number of subsidies in the first is less. If two programmes with similar characteristics are published, when you increase the number of requests in one of them the second will have less grant subsidies.

 Increasing the evaluation period doesn’t increase accepted subsidies, the behaviour shown is opposite to the result. However, it is determinant between the two programmes, that is, if evaluation time is increased in one program the other one will have more subsidies.

After these results have been analyzed the validation of the model is based on running it using past data in order to prove that the results correspond to the reality.

3 CONCLUSIONS

It is evident the current problems for making good decisions, this phase is especially important in the case of public policy assessment, as many agents are involved. Although different techniques have been used, almost anyone has the requirements needed to offer accuracy.

As an alternative to this paradigm simulation seems to be the best option, it allows the analysis of different scenarios in order to assess the possible risks depending on the decision. In this way, future predictions, supported by evidence instead of intuitions, could be done.

Due to the complex nature of these targeted huge systems in which public policies are applied, societies, evidence based methodologies are needed in order to make good decisions which could be empirically validated. The heterogeneity of the agents derives in multiple decisions and actions impacting in different individual processes which finally impact in the whole macro system.

Apart from that, the multi paradigm framework selected for the model is enough flexible and clear to cover all the points a policy maker should taken into account. The combination of ABM and SD results in a wide range of functionalities, the first one makes possible the general representation of the process of acceptance each agent is obliged to pass. However, the second one is used to create different scenarios depending on the variable that should be changed in order to understand their links and the effects they have on the result of conceding or not the subsidy.

In conclusion, simulation appears as the best alternative to assess public policies in an efficient way within gaining good decision making skills. This is presented as a promising scenario for the organizational management in which different dynamic objects could behave, act, decide and consequently, provoke different effects in a more macro level.

REFERENCES

[1] 1. Davis, J.P.e.a., Develop Theory Through Simulation Methods. Academy of Management Review, 2007. 32(2): p. 480–499.

[2] 2. Carley, K.M., Computational modeling for reasoning about the social behavior of humans. . Computational and Mathematical Organization Theory, 2009. 15(1): p. 47-59.

[3] 3. Smith, E.R.C., F.R., Agent-Based Modeling: A New Approach for Theory Building in Social Psychology. Personality and Social Psychology Review, 2007. 11(1): p. 87-104.

[4] 4. Harrison, J.R., Carroll, G.R. & Carley, K.M., 2007, SIMULATION MODELING IN ORGANIZATIONAL AND MANAGEMENT RESEARCH University of Texas at Dallas. . Academy of Management Review, 2007. 32(4): p. 1229–1245.

[5] 5. Pfeffer, J.S., R.I., Evidence-based management. Harvard business review, 2006. 84(1): p. 62.

[6] 6. Rousseau, D.M., Is there such a thing as “evidence-based management”? Academy of Management Review, 2006. 31(2): p. 256-269.

[7] 7. Introne, J.I., L., Improving decision-making performance through argumentation: An argument-based decision support system to compute with evidence. . Decision Support Systems, 2014. 64: p. 79-89.

[8] 8. Basole, R.C., Bodner, D. a. & Rouse, W.B., Healthcare management through organizational simulation. . Decision Support Systems, 2013. 55(2): p. 552-563.

[9] 9. Ligtenberg, A.e.a., A design and application of a multi-agent system for simulation of multi-actor spatial planning. . Journal of Environmental Management, 2004. 72(1-2): p. 43-55.

[10] 10. Louie, M.a.C., K.M., Balancing the criticisms: Validating multi-agent models of social systems. . Simulation Modelling Practice and Theory, 2008. 16(2): p. 242-256.

[11] 11. Deadman, P.e.a., Colonist household decisionmaking and land-use change in the Amazon Rainforest: an agent-based simulation. Environment and Planning B: Planning and Design, 2004. 31(5): p. 693-709.

[12] 12. D.Sterman, J. System Dynamics: Systems Thinking and Modeling for a Complex World. in ESD Internal Symposium. 2002.

[13] 13. Jun, J.B., Jacobson, S.H. & Swisher, J.R., Application of discrete-event simulation in health care clinics: A survey. . Journal of the Operational Research Society, 1999. 50(2): p. 109-123.

[14] 14. Merriënboer, J.J.G.v., Perspectives on Problem Solving and Instruction Computers & Education, 2012. 64: p. 153-160.

[15] 15. Brailsford, S.C., et al., Large Complex Modelling System. Society, 2012. 55(1): p. 34- 42.

[16] 16. Cooke, D., Rohleder, T. & Rogers, P., A dynamic model of the systemic causes for patient treatment delays in emergency departments. . Journal of Modelling in Management, 2010. 5(3): p. 287-301.

[17] 17. Esensoy, A.V.C., M.W., Health system modelling for policy development and evaluation: Using qualitative methods to capture the whole-system perspective. . Operations Research for Health Care, 2015. 4: p. 15-26.

[18] 18. Ford, D.N.S., D., Modeling Real Options to Switch Among Alternatives. . Proceedings of the 21st System Dynamics Conference, 2003: p. 1-15.

[19] 19. Ford, D.N.S., J.D., Dynamic Modelling of Product Development Processes. . System Dynamics Review, 1998. 14(1): p. 31-68.

[20] 20. Borshchev, A., The Big Book of Simulation Modeling, ed. e.A.N. America. 2013, North America.

[21] 21. An, L., Modeling human decisions in coupled human and natural systems: Review of agent-based models. . Ecological Modelling, 2012. 229: p. 25-36.

[22] 22. Bonabeau, E., Agent-based modeling: methods and techniques for simulating human systems. . Proceedings of the National Academy of Sciences, 2002. 99: p. 7280–7287.

[23] 23. Epstein, J.M., Chapter 34 Remarks on the Foundations of Agent-Based Generative Social Science. . Handbook of Computational Economics, 2006. 2(05): p. 1585–1604.

[24] 24. Heckbert, S., Baynes, T. & Reeson, A., Agent-based modeling in ecological economics. . Annals of the New York Academy of Sciences, 2010. 1185: p. 39-53.

[25] 25. Robinson, D.T.R., T., Modelling Farmer Household Decision-making and Its Effects on Land use/cover Change in the Altamira Region, Párá, Brazil. . History. , 2003.

[26] 26. Alderton, S., Noble, J. & Atkinson, P., Simulating Sleeping Sickness: a two host agent- based model. Advances in Artificial Life, 2013: p. 27-34.

[27] 27. Aruoba, S.B. and J. Fernández-Villaverde, A comparison of programming languages in macroeconomics. Journal of Economic Dynamics and Control, 2015. 58(265-273).

[28] 28. Artel, A., et al., An Agent-Based Model for the Investigation of Neovascularization Within Porous Scaffolds. Tissue engineering. Part A, 2011. 17(17-18): p. 2133-2141.

[29] 29. Conner, M.M., Ebinger, M.R. & Knowlton, F.F., Evaluating coyote management strategies using a spatially explicit, individual-based, socially structured population model. . Ecological Modelling, 2008. 219(1-2): p. 234-247.

[30] 30. Grosman, P.D.e.a., Reducing Moose-Vehicle Collisions through Salt Pool Removal and Displacement: an Agent-Based Modeling Approach. . Ecology and Society, 2009. 14(2).

[31] 31. Janssen, M.a.e.a., An adaptive agent model for analysing co-evolution of management and policies in a complex rangeland system. . Ecological Modelling, 2000. 131(2-3): p. 249–268.

[32] 32. Lin, S., DeVisser, M.H. & Messina, J.P., An agent-based model to simulate tsetse fly distribution and control techniques: A case study in Nguruman, Kenya. . Ecological Modelling, 2015. 314: p. 80-89.

[33] 33. Palmer, J., Sorda, G. & Madlener, R., Modeling the diffusion of residential photovoltaic systems in Italy: An agent-based simulation. . Technological Forecasting and Social Change, 2015. 99(106-131).

[34] 34. Terry Flynn, Y.T., Discrete choice, agent based and system dynamics simulation of health profession career paths, in Winter Simulation Conference, S.Y.D. A. Tolk, I. O. Ryzhov, L. Yilmaz, S. Buckley, and J. A. Miller, eds., Editor 2014.

[35] 35. Grigoryev, I., Anylogic in three days, ed. e. Anylogic. 2015.

[36]

View publication statsView publication stats