The most commonly used statistical models of civil war onset fail to correctly predict most occurrences of this rare event in out-of-sample data. Statistical methods for the analysis of binary data, such as logistic regression, even in their rare event and regularized forms, perform poorly at prediction. We compare the performance of Random Forests with three versions of logistic regression (classic logistic regression, Firth rare events logistic regression, andL1-regularized logistic regression), and find that the algorithmic approach provides significantly more accurate predictions of civil war onset in out-of-sample data than any of the logistic regression models. The article discusses these results and the ways in which algorithmic statistical methods like Random Forests can be useful to more accurately predict rare events in conflict data.

Electoral violence is increasingly affecting elections around the world, yet researchers have been limited by a paucity of granular data on this phenomenon. This paper introduces and describes a new dataset of electoral violence-the Dataset of Countries at Risk of Electoral Violence (CREV)-that provides measures of 10 different types of electoral violence across 642 elections held around the globe between 1995 and 2013. The paper provides a detailed account of how and why the dataset was constructed, together with a replication of previous research on electoral violence. We introduce this dataset by demonstrating that the CREV, while measuring the same underlying phenomena as other datasets on electoral violence, provides researchers with the ability to draw more nuanced conclusions about the causes and consequences of violence that occurs in connection with the electoral process. We also present and analyze descriptive data from the CREV dataset.

Hate speech has grown significantly on social media, causing serious consequences for victims of all demographics. Despite much attention being paid to characterize and detect discriminatory speech, most work has focused on explicit or overt hate speech, failing to address a more pervasive form based on coded or indirect language. To fill this gap, this work introduces a theoretically-justified taxonomy of implicit hate speech and a benchmark corpus with fine-grained labels for each message and its implication. We present systematic analyses of our dataset using contemporary baselines to detect and explain implicit hate speech, and we discuss key features that challenge existing models. This dataset will continue to serve as a useful benchmark for understanding this multifaceted issue. To download the data, see https://github.com/ GT-SALT/implicit-hate

Elections are in theory democratic means of resolving disputes and making collective decisions, yet too often force is employed to distort the electoral process. The postCold War increase in the number of electoral authoritarian and hybrid states has brought this problem into relief. In recent years the prevention of electoral violence has played an increasingly large role in the democratic assistance activities undertaken by international agencies, following increased awareness within the international community of the specific security challenges that elections entail. However, there has to date been little systematic evaluation of the success of different electoral violence prevention (EVP) strategies in reforming electoral institutions so as to enable them to maintain the peace during the electoral period. This article assesses the effectiveness of two common types of international EVP activity. Using a new global dataset of EVP strategies between 2003 and 2015, this article finds evidence that capacity-building strategies reduce violence by non-state actors, whereas attitude-transforming strategies are associated with a reduction in violence by state actors and their allies. The findings are relevant both for understanding the dynamics of electoral violence, and also for policymakers and electoral assistance providers in the international community who have responsibility for the design of democratic assistance projects in states at risk of electoral violence.

Electoral violence is conceived of as violence that occurs contemporaneously with elections, and as violence that would not have occurred in the absence of an election. While measuring the temporal aspect of this phenomenon is straightforward, measuring whether occurrences of violence are truly related to elections is more difficult. Using machine learning, we measure electoral violence across three elections using disaggregated reporting in social media. We demonstrate that our methodology is more than 30 percent more accurate in measuring electoral violence than previously utilized models. Additionally, we show that our measures of electoral violence conform to theoretical expectations of this conflict more so than those that exist in event datasets commonly utilized to measure electoral violence including ACLED, ICEWS, and SCAD. Finally, we demonstrate the validity of our data by developing a qualitative coding ontology.