This dissertation investigates the determinants of policy network structure, and the role of belief systems and biased learning in shaping networks. Understanding the interplay between beliefs and networks is important in order to move towards a theory of institutional design to support policy and social learning in the face of complex planning issues. Three theoretical frameworks are adopted to explain network structure. The Advocacy Coalition Framework (ACF) asserts that actors are subject to systematic cognitive biases that polarize networks along ideological lines. This causes entrenched political conflict and tends to inhibit successful learning. The Institutional Analysis and Development framework asserts beliefs and values do not drive network structure, and that actors seek to position themselves close to political resources. Social Capital Theory asserts that policy brokers are able to span divides within networks by making investments in collaborative ties.

Hypotheses are tested using survey data on the beliefs and networking behavior of policy elites involved in transportation and land use planning in five regions of California (N = 752). Results lend support for the ACF view that networks are polarized through cognitive biases, but also demonstrate that shared beliefs are not a sufficient explanation for network cohesion. Power-seeking may be a strong driver of networking within ideologically-similar groups (as predicted by Institutional Analysis and Development), and policy brokers can help to create social capital both within and across groups through transitivity effects. Finally, this dissertation underscores the need to develop better methods of survey-based network measurement to support future research on policy networks.