Protest in America: Univariate Arima Models of the Postwar Era

T WO COMPETING views of the outbreaks of protest over time coexist in conflict studies. One tradition maintains that protest activity occurs randomly over time. The alternative position is that protect activity occurs systematically over time in that "protest breeds protest." The different expectations are rooted in competing views of the conflict process within nations. This paper will attempt to address this controversy about the behavior of protest over time by modeling protest in the United States during the postwar period with univariate Box-Jenkins (ARIMA) procedures. Theories of protest that are consistent with both the contagion and randomness arguments are discussed next. The following section explores previous research on the contagion of protest. We then show how univariate BoxJenkins (ARIMA) models formalize many of the verbal expectations about the over time contagion and randomness of protest by specifying the form that these contagion and randomness effects take. The data for the test, which consist of a time series of the number of day/locales of protest events in the United States, is then discussed. The next section applies univariate ARIMA models to a time series of protest in the United States. In other words, an autocorrelation function which specifies the degree of dependence of incidents of protest at any time point on incidents of protest at previous time points is used to examine the pattern of protest in the postwar United States. Yearly, quarterly, and monthly models, using raw and per capitized data for two separate eras in the postwar period, are successively fit to the data. The final model, which disaggregates to months, per capitizes the data, and is fit to the post-1960 period, shows that protest appears to be a "white noise" or random process over time. The final section of the paper discusses the implications of our findings for theories of protest.