Introduction

Abstract

Prediction and forecasting have now fully reached peace and conflict research. We define forecasting as predictions about unrealized outcomes given model estimates from realized data, and predictions more generally as the assignment of probability distributions to realized or unrealized outcomes. Increasingly, scholars present within-and outof-sample prediction results in their publications and sometimes even forecasts for unrealized, future outcomes. The articles in this special issue demonstrate the ability of current approaches to forecast events of interest and contributes to the formulation of best practices for forecasting within peace research. We highlight the role of forecasting for theory evaluation and as a bridge between academics and policymakers, summarize the contributions in the special issue, and provide some thoughts on how research on forecasting in peace research should proceed. We suggest some best practices, noting the importance of theory development, interpretability of models, replicability of results, and data collection.Keywords forecasting, out of sample evaluation, peace research, prediction, theory testingNo matter how I turn it over in my mind, the number one task of peace research always turns out to be that of prediction [...] (J David Singer, 1973)

Conflict forecasting and its limits

“…Scholars and policy-makers have also relied on algorithms analyzing large amounts of data using various techniques, ranging from the more traditional regression techniques -typically logistic, due to the binary nature of the outcome -to more intricate random forest or neural network models (Rummel [41]; Brandt and Freeman [6]; Muchlinski et al [34]; Hegre et al [28]). In the fields of international conflict and civil war, the majority of this work has until recently relied on structural variables such as regime type, GDP, ethnicity, or terrain (Beck, King, and Zeng [2]; O'Brien [35]).…”

Common statistical patterns in urban terrorism

“…One of these challenges is the protracted political violence that plagues many urban regions in the world [2]. While creating data-driven regression models can yield insights into ongoing violence [3,4], statistical patterns can also yield insight into common trends [5,6]. In this paper, we focus our analysis on urban attacks, where the majority of attacks take place (e.g.…”

Spatio-temporal violent event prediction using Gaussian process regression

“…The social system is sufficiently complex and unpredictable that forecasting large-scale wars or local-scale violence with good accuracy is difficult, though recent advances hold promise [7]. The GLM model used in [21] provides a comparison to the Gaussian process models developed here.…”