On the Role of Agent‐based Modeling in the Theory of Development Economics

• DOI: http://doi.org/10.1111/rode.12264
• Date: 01 August 2017
• Author: Florian Chávez‐Juárez
• Published date: 01 August 2017

On the Role of Agent-based Modeling in the

Theory of Development Economics

Florian Ch

avez-Ju

arez*

Abstract

In this article I discuss the potential role of agent-based modeling techniques in development economics.

Development economics has recently seen a strong rise of experimental evidence from the field and the

laboratory. At the same time, there is a debate on how theory should adapt to this new approach and its

findings. I argue in this paper that the agent-based modeling approach is a promising complement to the

traditional modeling techniques, as it can easily incorporate the non-standard findings of the

experimental literature. Moreover, I emphasize the opportunity of a mutually beneficial interplay

between experiment-based empirical research and agent-based models.

1. Introduction

This article aims at contributing to the debate on the role of theory in development

economics by discussing the potential benefits from an increased use of agent-based

modeling techniques. The starting point of the article is the call of several

researchers for a new or revised theory in development economics (Mookherjee,

2005; Banerjee, 2005; Bradhan, 2005; Rodrik, 2008). This call is essentially driven

by recent developments in the empirical literature on development economics,

which have increasingly used the experimental approach and produced findings that

are challenging the classical theory. Incorporating these findings in the theory is

difficult when using the standard modeling approach. In this article I argue that

agent-based models could be a promising complement that could facilitate this

incorporation. While other authors such as Foley and Farmer (2009) have argued

that agent-based models should be used in economics in general, this article focuses

on the field of development economics and in particular on the usefulness of agent-

based modeling for the above-mentioned call for a new or revised theory.

In a first step I briefly review the core ideas of the agent-based approach using an

every-day example to illustrate the different elements. I then discuss in more detail

the origins of the call for a new or revised theory. After these two introductory

discussions, I start to discuss how the agent-based approach could be beneficial for

the field of development economics and in particular for the incorporation of these

new empirical findings. In this respect, the article discusses methodological issues

rather than the actual content of models in development economics. The goal is not

to contrast agent-based models to the more established mathematics-based

approaches that produce analytical solutions, but to highlight benefits the field

might expect from an increased used of these computer-based techniques.

*Ch

avez-Ju

arez (Corresponding author): Centro de Investigaci

on y Docencia Econ

omicas (CIDE),

Carretera M

exico Toluca No. 3655, Col. Lomas de Santa F

e, Delegaci

on Alvaro Obreg

on, 01210, Mexico

City, Mexico. E-mail: florian@chavezjuarez.com

Review of Development Economics, 21(3), 713–730, 2017

DOI:10.1111/rode.12264

©2016 John Wiley & Sons Ltd

I group these potential benefits into two main categories. The first category of

benefits is directly linked to the above-mentioned call for a new or revised theory in

development economics. This call is based on various insights such as non-optimal

behavior, heterogeneity in both behavior and characteristics and the important role of

interactions among agents and the information they have access to.

The second category of benefits has a more general character and includes

properties of agent-based models that can facilitate the development, use and

analysis of theoretical models. Here I will focus essentially on the increased

flexibility in the modeling approach and the possibility to easily track the evolution

of an economic system.

Of course, agent-based models also come with some drawbacks, which might

explain the rather scarce use of these models in economics nowadays.

1

The risk of

creating too complex models resulting in black boxes or the lack of analytical

solutions are two examples of such potential drawbacks. I will not argue that these

drawbacks should be disregarded and that agent-based modeling is always a good

alternative to the traditional tools. However, my argument is that agent-based

models represent an interesting complement that might—in some cases—provide

more promising results.

I then sketch the idea of an iterative research approach, in which the elaboration

of theoretical agent-based models and the experimental approach are mutually

beneficial. Based on the benefits discussed and the proposed iterative research

approach, I see essentially three main fields of application for the agent-based

modeling approach within the field of development economics. First, agent-based

models can be used ex-ante to evaluate various policy alternatives. Such a

theoretical ex-ante analysis of different alternatives can help us to improve the

efficiency of programs. The second application could be the use of agent-based

models to extrapolate empirical findings to another context. This is particularly easy

with the agent-based approach because the environment of the model can be easily

adapted. Finally, in situations where it is difficult to empirically validate theoretical

proposition, agent-based models can be an excellent vehicle to perform this task

using an artificial world and comparing the outcome with actual data.

2. The Increasing Importance of Empirical and Experimental Studies in

Development Economics and their Challenging Insights

Development economics moved from a strongly theory based field to a much more

empirical field over the last years. The increased availability of high quality data

from middle- and low-income countries is a major reason for this development.

Moreover, the use of experiments has sharply increased over the last years (Guala,

2005), especially through randomized field trials (RFT). Randomized field trials are

often associated with the evaluation of a program. For instance, the PROGRESA

program in Mexico was implemented as a randomized field trial with a control and

a treatment group (Schultz, 2004; Skoufias et al., 2001; Parker et al., 2006).

However, the experimental approach is not limited to the evaluation of a specific

policy measure. Many experiments—both in the field and the laboratory—seek to

answer a particular research question with no direct link to the evaluation of a

program. For instance, Betrand et al. (2010) use an experimental approach to study

the effect of marketing strategies on the take-up rate of credits and the credit

behavior of individuals more generally. Meier and Sprenger (2010) and Bauer et al.

(2012) use field experiments to study the effect of the present bias on credit card

714 Florian Ch

avez-Ju

arez

©2016 John Wiley & Sons Ltd