Characterizing Punctuated Equilibrium in Comparative Analysis

Abstract

In policy process theory, the literature on punctuated equilibrium deals with fundamental questions about how governments allocate attention to policy problems. This paper revisits its methodological foundations and focuses on three problems that, while not critical at the early stages of theory development, appear to get in the way of comparative punctuated equilibrium. These problems highlight the need for greater sensitivity to the substantive differences of alternative measures of government attention allocation. The analysis is particularly relevant to scholars who attempt to extend the punctuated equilibrium literature, or policy studies generally, to non-democratic and/or non-Western contexts (e.g. Chan and Zhao forthcoming; Lam and Chan 2014). First, kurtosis estimates were first applied to characterize the non-incremental dynamics of individual systems of attention allocation. Given the relatively restricted scope of analysis, the choice between alternative measures of change did not change the outcome. However, switching between them can lead to different estimates and therefore different conclusions when that these measures are used for direct comparisons of multiple systems of attention allocation. Second, while the level of instability of a political or policy system is estimated from longitudinal data on attention allocation, there is no agreement on the length of the window of observation. Some analyses rely on longitudinal data that cover at least a century of budgetary allocations. In some cases, they are directly compared to others that draw on less than 20 years of data. In the case of long data streams, estimates of policy instability may become biased if the observations are drawn across major regime transitions. On the other hand, bias may arise in short data streams because the observations can either fall between instances of great disruption or fall on short-lived but particularly volatile intervals. These concerns about sampling are relevant to punctuated equilibrium studies because punctuated changes are by definition rare events. Thirdly, researchers begin to move from system-level characteristics of attention allocation to the modeling of individual instances of punctuated change. For example, more recent applications involve modeling individual punctuated changes as a function of past punctuations. Since alternative measures of change capture different dynamical features of attention allocation, instances that are categorized as punctuated shifts with one measure may turn out to be incremental using a different measure. Since the discrepancies have substantive implications for the choice of modeling strategies, I evaluate the extent that the classification of allocation changes as “punctuation” is contingent on how one quantifies change.