Path Dependence

The concept of path dependence has been popularized by economic historians and is now widely employed in economics, history, political science, sociology, and geography to describe systems and processes that are governed by their own history. More formally, a path-dependent system or process is characterized by nonergodicity, a term taken from the theory of stochastic systems. While an ergodic system or process will converge toward an invariant, asymptotic probability distribution regardless of its initial conditions, the asymptotic probability distribution of a nonergodic or path-dependent stochastic system evolves as a consequence of its own history. Path dependence is a label for a particular class of dynamic systems or processes, not a theory explaining how those systems behave. It is thus necessary to “develop theories and frame hypotheses regarding the economic circumstances in which path-dependent dynamics are likely to appear, and where the consequences of past events can be expected to exert greatest and most enduring influence” (David, 2007, p. 92).

Work on path dependence differs in the emphasis that scholars give to four elements included in most accounts: “causal possibility, contingency, closure, and constraint” (Bennett & Elman, 2006, p. 252). Causal possibility means that more than one path could have been taken at a specific historical juncture. If we could rerun history and end up with the same outcome every time, then that process is not path dependent. Some authors argue that early events are more important than later events in the sequence, while others distinguish between path independence and path dependence, where the former refers to those outcomes that are independent of the order of events. Contingency entails that the causal narrative be influenced by random or, in some cases, exogenous events. Closure refers to the fact that once a path is chosen, alternative paths become increasingly unlikely. And finally, once a path is chosen, a constraint is required that ties agents to that path.

In economics, three main versions of path dependence exist: path dependence as a technological lock-in, as increasing returns, and as institutional hysteresis. Paul David's work on technological lock-in highlights the impact of contingent accidents, or microlevel chance events, on long-run trends and structures and stresses the importance of early events or decisions for closing alternative paths and strengthening existing ones as network externalities increase the efficiency gap between the chosen and the abandoned paths. The lock-in of suboptimal designs and systems such as the QWERTY keyboard or the VHS video system is often invoked as proof that markets do not necessarily lead to optimal outcomes. Brian Arthur's work highlights the role of various forms of increasing returns to explain how a technology or spatial system becomes locked into a path. Arthur formalizes those processes through the so-called ball-and-urn models to demonstrate how macrostructure emerges from contingent, random micro events and behavior. Mark Setter-field's work on institutional hysteresis recognizes that formal and informal institutions, such as conventions, habits, and traditions, change only slowly over time; structuring human behavior, interaction, and communication, these institutions keep people, places, and countries on specific evolutionary trajectories. The work suggests that the adopted technologies, spatial patterns, or [Page 2134]institutions are not necessarily the most efficient alternatives. The work also stresses that the past does not determine the future but that at each moment in time, the possible future evolutionary trajectories are contingent on the past and current states of the system in question and some paths are more likely than others.

...