By contrast, sociological institutionalism and constructivist approaches in international relations defined institutions much more broadly to include informal norms and conventions as well as informal rules. They argued that such institutions could ‘constitute’ actors, shaping their identities and hence their preferences in ways that rational-choice approaches could not capture (see next section).
Historical institutionalists took up a position between these two camps, focusing on the effects of institutions over time, in particular on the ways in which a given set of institutions, once established, can influence or constrain the behaviour of the actors who established them. In its initial formulations (Hall 1986; Thelen and Steinmo 1992), historical institutionalism was seen as having dual effects, influencing both the constraints on individual actors and their preferences, thereby making the theory a ‘big tent’, encompassing the core insights of the rationalist and constructivist camps.
What makes historical institutionalism distinctive, however, is its emphasis on the effects of institutions on politics over time. In perhaps the most sophisticated presentation of this thinking, Paul Pierson (2000) has argued that political institutions are characterized by what economists call ‘increasing returns’, insofar as they create incentives for actors to stick with and not abandon existing institutions, adapting them only incrementally in response to changing circumstances. Thus, politics should be characterized by certain interrelated phenomena, including: inertia, or ‘lock-ins’, whereby existing institutions may remain in equilibrium for extended periods despite considerable political change; a critical role for timing and sequencing, in which relatively small and contingent events at critical junctures early in a sequence shape events that occur later; and path-dependence, in which early decisions provide incentives for actors to perpetuate institutional and policy choices inherited from the past, even when the resulting outcomes are manifestly inefficient.