Policy and Complex Systems
important that he revises his estimate of the wage distribution as offers are made . If his initial estimate is high , an adaptive policy reduces the period of frictional unemployment and , similarly , if his initial estimate is low an adaptive policy makes him avoid “ poor ” offers . Unfortunately , this type of policies are expensive in terms of time and computational effort : for this reason , in addition to saving computational time , automated contracting through computational agents ( a computational agent is an agent whose decisions about its actions can be explained in terms of computation ) can increase search efficiency in certain problem applications .
The purpose of this paper is therefore to advance theoretical understanding of the matching process and its dynamics , focusing on an analysis at the regional level . We propose a different interpretative strategy based on the acknowledgement that both firms and workers live in a complex system that evolves through time , which might be characterized by endogenous , persistent , different agents , which are heterogeneous in their endowments , wealth , behavior , and rationality ( Fagiolo , Dosi , & Gabriele , 2004 ).
Starting from this context , agentbased computational economics ( ACE ) models ( Tesfatsion & Judd , 2006 ) identify a generic model in order to explain a complex phenomenon at a very abstract level . At the same time , this kind of model can be based on empirical data to describe a particular case study . Finally , ACE models include spatial characteristics : a person located in an area with low labor demand will have a lower probability of getting an employment . This fact is seldom covered in standard economic approaches ( Dibble , 2006 ).
This paper follows therefore an agent-based models ( ABMs ) approach trying to link micro-level and macrolevel evaluations and focusing on the interactions between economic agents . The so constructed model will actually try to describe the dynamic matching between labor supply and demand , as well as how it is affected by specific changes in policy and social parameters .
The paper is organized as follows . In Section 2 , we start by describing the convenience of evaluating Labor Market Policies with ACE . In Section 3 , we present the model and we discuss its characteristics , design concepts , and functional specifications according to Dahlem ABM documentation ( Wolf et al ., 2010 ). We next present , in Section 4 , two test experiments and first results . Finally , Section 5 draws some conclusions and perspectives .
2 - Labor Market Policy Evaluation with ACE
Recently , ABMs have been proposed in order to support labor policymakers in their decisions : a similar argument holds for many policy questions , which are typically related to at least some aspects of economy , technology , and actor behavior ( Chappin , Chmieliauskas , & de Vries , 2012 ). With agents in a computer , it becomes possible to recreate the actual world on an artificial basis , to see the effects of the action and interaction .
Computer simulation can effectively be considered as a “ third way ” between verbal argumentation and mathematical equations that can combine an extreme flexibility of a computer code , where it is possible to create agents who act , make choices , and react to the choices of other agents and to the modification of their environment , and its computational power . This approach allows researchers to use the descriptive capabilities of verbal
17