Party nationalization in a multilevel context: Where’s the variance? Thomas Mustillo a, * , Sarah A. Mustillo b a Department of Political Science, Purdue University,100 N. University Street, West Lafayette, IN 47907-2098, United States b Department of Sociology, Purdue University, 700 W. State St., West Lafayette, IN 47907-2059, United States article info Article history: Received 12 May 2011 Received in revised form 14 December 2011 Accepted 20 December 2011 Keywords: Party nationalization Electoral volatility Political party Elections abstract We use estimates of variance in district-level electoral data as a way to identify multiple dimensions of the nationalization of party support, including “static nationalization” and “dynamic nationalization.” The multilevel model we usedalso described as a random coefficient, mixed, growth curve, and hierarchical modeldyields estimates of a party’s mean national trajectory of electoral support (fixed effects), as well as estimates of variability around the mean trajectory parameters (random effects). Using a general model, we present a two-step approach to first identify electoral variability and then account for it. We develop the model, apply it to three political parties, demonstrate its behavior under controlled conditions using data we create, and demonstrate its application for explanatory purposes. Ó 2011 Elsevier Ltd. All rights reserved. 1. Introduction Political scientists use election results to measure core concepts about the nature of parties and party systems, and to draw conclusions about the democratic regime. Electoral returns are versatile data because they are observed over time, across space both sub-nationally and cross-nationally, on parties which themselves are highly variable, and under all sorts of contextual conditions. In this paper we study phenomena pertaining to electoral variability, by which we mean the concepts of electoral volatility and party nation- alization. In a recent study, Morgenstern and Potthoff (2005), extending a line of work initiated by Stokes (1965), identify substantive and methodological flaws in the way scholars have typically used electoral data to compute electoral volatility and various dimensions of the national- ization of political parties. They show that flawed measures result with techniques that compute quantities of one phenomenon without simultaneously taking into consid- eration other phenomena. As an alternative, they advocate a “components-of-variance” model 1 which estimates three quantities togetherdvolatility, static nationalization, and dynamic nationalization 2 dso as not to mistakenly conflate electoral variability of one type with variability of another type. Thus, that technique is not simply an alternative measure to most recent published measures, but it is an alternative paradigm for treating electoral data. While we concur that the “components of variance” approach improves upon alternative measurement strategies in all the ways they suggest, we argue that there is a preferable way to treat time which is more intuitive, nuanced, and precise. * Corresponding author. Tel.: þ1 765 532 6238. E-mail addresses: tmustill@purdue.edu (T. Mustillo), smustill@purdue. edu (S.A. Mustillo). 1 Throughout this paper, we adopt the following terminology: the model used by Morgenstern and Potthoff (2005) and Stokes (1965, 1967) is a “components-of-variance” model. The model we propose is called the “multilevel model.” Note that the multilevel model includes estimates of “variance components,” but that these estimates are distinct from the “components-of-variance” class of model. 2 The concept names Morgenstern and Potthoff (2005) adopt change in subsequent publications which build upon that work, with the exception of “volatility,” which remains stable. We adopt the term “static nation- alization” from Morgenstern et al. (2009), which Morgenstern and Potthoff (2005) call “district heterogeneity.” We adopt the term “dynamic nationalization” from Morgenstern et al. (2009), which Morgenstern and Potthoff (2005) call the “district-time effect” and Morgenstern and Swindle (2005) call the “local effect.” Contents lists available at SciVerse ScienceDirect Electoral Studies journal homepage: www.elsevier.com/locate/electstud 0261-3794/$ – see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.electstud.2011.12.003 Electoral Studies 31 (2012) 422–433