Between replication and docking: "Adaptive agents, political institutions, and civic traditions" revisited

This article has two primary objectives: (i) to replicate an agent-based model of social interaction by Bhavnani (2003), in which the author explicitly specifies mechanisms underpinning Robert Putnam's (1993) work on Civic Traditions in Modern Italy, bridging the gap between the study's historical starting point-political regimes that characterized 14th Century Italy-and contemporary levels of social capitalóreflected in a 'civic' North and an 'un-civic' South; and (ii) to extend the original analysis, using a landscape of Italy that accounts for population density. The replication exercise is performed by different authors using an entirely distinct ABM toolkit (PS-I) with its own rule set governing agent-interaction and cultural change. The extension, which more closely approximates a docking exercise, utilizes equal area cartograms otherwise known as density-equalizing maps (Gastner and Newman 2004) to resize the territory according to 1993 population estimates. Our results indicate that: (i) using the criterion of distributional equivalence, we experience mixed success in replicating the original model given our inability to restrict the selection of partners to 'eligible' neighbors and limit the number of agent interactions in a timestep; (ii) increasing the number of agents and introducing more realistic population distributions in our extension of the replication model increases distributional equivalence; (iii) using the weaker criteria of relational alignment, both the replication model and its extension capture the basic relationship between institutional effectiveness and civic change, the effect of open boundaries, historical shocks, and path dependence; and (iv) that replication and docking may be usefully combined in model-to-model analysis, with an eye towards verification, reimplementation, and alignment.