Will Lowe’s research while affiliated with Princeton University and other places

During the last decade, much of political behaviour research has come to be concerned with the impact of the Internet, and more recently social networking sites such as Facebook, on political and civic participation. Although existing research generally finds a modestly positive relationship between social media use and offline and online participation, the majority of contributions rely on cross-sectional data, so the causal impact of social media use remains unclear. The present study examines how Facebook use influences reported political participation using an experiment. We recruited young Greek participants without a Facebook account and randomly assigned a subset to create and maintain a Facebook account for a year. In this paper we examine the effect of having a Facebook account on diverse modes of online and offline participation after six months. We find that maintaining a Facebook account had clearly negative consequences on reports of offline and online forms of political and civic participation.

The extensive use of social media for protest purposes was a distinctive feature of the recent protest events in Spain, Greece, and the United States. Like the Occupy Wall Street protesters in the United States, the indignant activists of Spain and Greece protested against unjust, unequal, and corrupt political and economic institutions marked by the arrogance of those in power. Social media can potentially change or contribute to the political communication, mobilization, and organization of social movements. To what extent did these three movements use social media in such ways? To answer this question a comparative content analysis of tweets sent during the heydays of each of the campaigns is conducted. The results indicate that, although Twitter was used significantly for political discussion and to communicate protest information, calls for participation were not predominant. Only a very small minority of tweets referred to protest organization and coordination issues. Furthermore, comparing the actual content of the Twitter information exchanges reveals similarities as well as differences among the three movements, which can be explained by the different national contexts.

The core question of this paper is of whether online engagement with a movement actually leads to offline participation. Social media content cannot only reveal users attitudes towards policy problems, politicians, elections, riots, protests and unrest, but also highlight people’s preferences, willingness to participate and mobilise others. We monitored the ‘indignant citizens’ of Spain and Greece, and the US‐based Occupy Wall Street movements since their origin and harvested more than one million tweets with a text‐analysis tool. Building on previous comparative analyses of tweets, which revealed both similarities and differences in the use of social media to organize protest activities and mobilize political action in the streets, the paper explores whether the exchange of content through microblogging platforms such as Twitter is translated into offline participation. To answer that question we present preliminary results from an online survey addressed to Twitter users. By comparing the offline political participation of Twitterers who did and did not tweet about the movements, we obtain unique empirical information about social network sites’ potential to mobilize citizens offline. The main conclusion is that, in general, online and offline participation are not related. However, involvement in microblogging indeed has a positive impact on participation in offline protest activities, but for a specific cause or issue only.

This article introduces the global Pro-Government Militias Database (PGMD). Despite the devastating record of some pro-government groups, there has been little research on why these forces form, under what conditions they are most likely to act, and how they affect the risk of internal conflict, repression, and state fragility. From events in the former Yugoslavia, Iraq, Sudan, or Syria and the countries of the Arab Spring we know that pro-government militias operate in a variety of contexts. They are often linked with extreme violence and disregard for the laws of war. Yet research, notably quantitative research, lags behind events. In this article we give an overview of the PGMD, a new global dataset that identifies pro-government militias from 1981 to 2007. The information on pro-government militias (PGMs) is presented in a relational data structure, which allows researchers to browse and download different versions of the dataset and access over 3,500 sources that informed the coding. The database shows the wide proliferation and diffusion of these groups. We identify 332 PGMs and specify how they are linked to government, for example via the governing political party, individual leaders, or the military. The dataset captures the type of affiliation of the groups to the government by distinguishing between informal and semi-official militias. It identifies, among others, membership characteristics and the types of groups they target. These data are likely to be relevant to research on state strength and state failure, the dynamics of conflict, including security sector reform, demobilization and reintegration, as well as work on human rights and the interactions between different state and non-state actors. To illustrate uses of the data, we include the PGM data in a standard model of armed conflict and find that such groups increase the risk of civil war.

Automated and statistical methods for estimating latent political traits and classes from textual data hold great promise, since virtually every political act involves the production of text. Statistical models of natural language features, however, are heavily laden with unrealistic assumptions about the process that generates this data, including the stochastic process of text generation, the functional link between political variables and observed text, and the nature of the variables (and dimensions) on which observed text should be conditioned. While acknowledging statistical models of latent traits to be “wrong”, political scientists nonetheless treat the treat their results as sufficiently valid to be useful. In this paper, we address the issue of substantive validity in the face of potential model failure, in the context of unsupervised scaling methods of latent traits. We critically examine one popular parametric measurement model of latent traits for text and then compare its results to systematic human judgments of the texts as a benchmark for validity.

Coding non-manifesto documents as if they were genuine policy platforms produced at election time clearly raises serious issues with error when these codings are used in the standard manner to estimate left-right policy positions. In addition to the long term solution of improving the document base of the Manifesto Project identified by Gemenis (2012), we argue that immediate gains in manifesto-based estimates of policy positions can be realised by using the confrontational logit scales from Lowe et al. (2011), which addresses the problems of scale content and scale construction that are exacerbated by but not unique to the problems found in proxy documents.

Event data are an important source of information about interacting political actors, offering a finer time scale and specificity than perhaps any other source. From a statistical perspective they are also awkward. With few exceptions the current literature adopts measurement by fiat by using predetermined event code to number mappings, implicitly imputes missing data with zeros during aggregation, and ignores the effects of measurement error due to variation in coverage density and selection effects. This approach hinders the resolution of questions such as whether and how events or actors should be aggregated (Shellman, 2004), apparent paradoxes due to event exchangeability ('two riots equals a war'), and practical questions such as how to tell whether conflict and cooperation really form a single scale, and how to combine forecasts from different information sources. More importantly these choices affect substantive inference about conflict processes and suitable policy. Recent work by Schrodt and co-authors has investigated various inductive approaches to event data analysis ranging from discrete latent time series analysis using Hidden Markov Models to topic and IRT models (e.g. Schrodt, 2006, 2007, 2011). There are unappreciated substantive and methodological advantages for the wider class of measurement models from which these examples are taken, including a systematic treatment of temporal aggregation, effective quantification of measurement error, and unified treatment of missing data. This paper offers: 1) a general measurement theory for event data that diagnoses existing event data problems and paradoxes. 2) a broad class of measurement models suitable for event data analysis, some of which have not been applied in this context. In particular, the paper establishes a theoretical link between event data analysis and and text scaling models and shows how manually assigned conflict scores can be recovered automatically using an unfolding-type measurement model.

Several methods have now become popular in political science for scaling latent traits— usually left-right policy positions—from political texts. Following a great deal of de-velopment, application, and replication, we now have a fairly good understanding of the estimates produced by scaling models such as "Wordscores", "Wordfish", and other variants (i.e. Monroe and Maeda's two-dimensional estimates). Less well understood, however, are the appropriate methods for estimating uncertainty around these esti-mates, which are based on untested assumptions about the stochastic processes that generate text. In this paper we address this gap in our understanding on three fronts. First, we lay out the model assumptions of scaling models and how to generate un-certainty estimates that would be appropriate if all assumptions are correct. Second, we examine a set of real texts to see where and to what extent these assumptions fail. Finally, we introduce a sequence of bootstrap methods to deal with assumption failure and demonstrate their application using a series of simulated and real political texts.