Temporally Recursive Regression and Social Historical Inquiry: An Example of Cross-Movement Militancy Spillover

Abstract

Our focus here is on time-series regression as a formal analytic tool in social historical inquiry. We have three interrelated purposes. First, we argue that conventional time-series regression is typically ill-suited for social historical inquiry because ahistorical assumptions and conventions regarding time undermine the historical character of social "process-as-analyzed". Second, we present a modified time-series approach -temporally recursive regression -that takes time seriously and provides a more adequate analytic vehicle for social historical inquiry. Finally, we illustrate the promise of temporally recursive regression by using it to analyze how workplace militancy in post-war America was fueled by massive insurgency waves during successive phases of the civil rights movement. A H I S T O R I C A L C H A R A C T E R OF C O N V E N T I O N A L T I M E -S E R I E S R E G R E S S I O N 1Conventional time-series regression contains a conception of time that is ahistorical in character. 2 It is fairly termed ahistorical, we believe, because time enters the analysis solely as a means of analysis, an instrumental marker for purposes of ordering the time unit observations (e.g. years). Treating time, and therefore history, as simply means rather than also object of analysis shapes the practice of conventional time-series regression in such a way that unduly masks significant historical context, meaning and nuance in social process. Simply put, ahistorical time severely limits what conventional * We thank Larry Griffin and Marcel van der Linden for comments on prior versions of the paper. Isaac's co-authors are listed in alphabetical order. 1. This section draws on: Larry

Old Times Are Not Forgotten

“…Given that no academies were closed during the years NCES reported private enrollment data, event history analysis was not employed. Nevertheless, if more longitudinal data were available this approach would provide a more appropriate method for examining temporally dependent dimensions of academy enrollments as it would elucidate the time-varying effects of explanatory determinants in conjunction with the onset of specific events (see Griffin and Korstad 1999;Issac et al 1999;McCammon 1999 for examples and review). Thus, it is important to bear in mind that the findings reported here reflect data collected from cross-sectional "snap shots" of locations at specific of periods of time; additional research is needed to further insight into the local unfolding of events and their contingent influence on the academy enrollments through time.…”

Innovative Methods in the Study of Protest: Editor’s Introduction

“…How does one identify such historical or contextual breaks? In their article, Heidi Reynolds-Stenson and Jennifer Earl address this with a special time series method for addressing what types of policing respond to demonstrations-temporally moving regressions (also known as "temporally recursive regressions" [Isaac et al, 1998]). This technique was originally developed by Isaac and Griffin (1989) to identify historical turning points and qualitative contingencies in temporal processes.…”

The Puzzle of Protest Policing Over Time: Historicizing Repression Research Using Temporal Moving Regressions

“…Isaac and Griffin's collective work address these concerns using a set of techniques that alter how data are pooled to estimate a regression. Going under the names "recursive regression" (Griffin & Isaac, 1992), "temporal recursive regression," (TRR; Isaac et al, 1998), "time-varying parameter (TVP) time series" (Isaac et al, 1994), or simply "moving regressions" (Brown et al, 1975), this ensemble of similar approaches essentially creates minipools of cases by reducing and changing the period the pooling covers iteratively; specific techniques vary in how minipools are constructed. If estimates from smaller pools of time differ from other minipools or from data pooled for the entire study period, that is evidence of temporal heterogeneity and the differences in model results allow researchers to uncover and examine it.…”

“…In this article, we argue for the use of temporal moving regressions (TMR), 1 which are the same as moving regressions (Brown et al, 1975) and diagonal TRR (Isaac et al, 1998), to identify and make sense of temporal heterogeneity in relationships more inductively. TMR takes a larger pooled data set and breaks it into a moving set of minipools on which the same regression is reestimated over and over again.…”