Social Network Structure and The Trade-Off Between Social Utility and Economic Performance

The objective of this paper is to identify the key mechanisms allowing the social network structure to affect individuals' social trust, willingness to cooperate, economic performance and social utility, and to trace how these individual-level outcomes aggregate up to the society level. To this end, we study how social networks give rise to the accumulation of social capital, defined as the aggregate of resources accessible to individuals through their social networks (Bourdieu, 1986), and how in turn social capital enables the creation of trust and cooperation. Even under Bourdieu's definition, however, social capital remains an ambiguous, complex concept. In this paper, we handle this complexity by considering four key types of individuals' social capital: (i) network degree, (ii) centrality, (iii) bridging and (iv) bonding social capital. The contribution of this study to the literature is to construct a novel computational multi-agent model, based on Watts and Strogatz (1998) network structure and extending it to incorporate the aforementioned quadripartite decomposition of social capital as well as accommodate several other findings from the associated socio-economic literature. The details of the model setup also draw from our empirical findings for the Polish society (Growiec, Growiec and Kamiński, 2017), based on a unique, detailed survey dataset. Implications of the model, however, reach beyond the specificities of this particular society and should be tested at the cross-country level. While this may be partially hindered by the lack of internationally comparable data on the detailed social capital measures, we make a first step in this direction by presenting some preliminary evidence from European Social Survey (ESS) data as well as confronting model outcomes with implications from theory. We demonstrate that our computational agent-based model, whose properties have been analyzed following a systematic simulation design, is a useful tool for simulating social capital stocks, trust, cooperation, and economic performance at the aggregate and individual level in the economy. Assuming that different countries or communities may feature different topologies of social networks and exhibit different social norms (e.g., how much value is attached to social ties with family members), we investigate if these differences lead to varying levels of social capital, trust and cooperation in the economy and if there is a trade-off between aggregate social utility and economic performance which they ultimately convey. In particular we specify the conditions under which small world type networks of social ties (as observed in most real-life economies) can be socially optimal. We also address the micro--macro linkages, implicit in the model, by answering the question, how the aggregate variables affect individual-level trade-offs such as, e.g., the trade-off between individual social utility and economic performance. Our key findings are as follows: (i) societies that either are globally better connected, exhibit a lower frequency of local cliques, or have a smaller share of family-based cliques, record relatively better economic performance; (ii) social utility presents an inverted U-shaped relationship with network density and a negative relationship with the frequency of family-based local cliques; (iii) if contacts with family are highly valued in the society, then there is a trade-off between aggregate social utility and economic performance, and then small world networks are socially optimal, otherwise they are outperformed by highly diversified, inclusive networks; (iv) in dense networks, social ties are individually less valuable; (v) social trust is a functional substitute to social networks: in trustful societies, social ties are individually less valuable, and vice versa; (vi) in dense networks and trustful societies, there is a trade-off between individuals' social utility and economic performance, and otherwise both outcomes are positively correlated in the cross section.