A Hidden Assumption in Participation Models

Explaining protest activities, scholars often use data from surveys based on nationally representative samples. (e.g.: Dalton, Sickle and Waldon 2009, Schussman and Soule 2005, Vrablikova 2013). Statistical model estimations aim to detect the factors that influence political participation. Schussman and Soule seek answers to the question: "Why do some individuals participate in protest?” (1084). In these analyses, researchers measure participation with a set of questions that refer to the past, e.g.: "Have you participated in a protest march or demonstration in the last 5 years?". As some authors have already noticed (Bayerley and Hipp 2006, Saunders 2014), it is very difficult to interpret models which contain, at the same time, a dependent variable referring to an activity in the past (e.g. participation in public demonstrations) and other, independent variables referring to present individual properties. The objective of this article is to draw the limelight to the hidden assumption that made it possible to apply and interpret statistical models. The confusing time-lines of the dependent and independent variables could be resolved just in the case, we assume that predictor factors or individual-level properties (eg.: left-right ideological orientation, satisfaction with the government) of the respondents are stable and unchanging in time. In this paper we shed light on the consequences of this assumption on the regression models. After simulating distributions of the year-by-year changes in explanatory variables we experienced that the uncertainty in time lines has a remarkable effect on the Beta coefficients in the model. We make use of data of the LISS (Longitudinal Internet Studies for the Social sciences) panel administered by CentERdata (Tilburg University, The Netherlands).