Logo: to the web site of Uppsala University

Over the past decade, the field of conflict forecasting has undergone a remarkable metamorphosis, transforming from a series of isolated efforts with low predictive power into large, globe-spanning projects with impressive performance. However, despite this evolution, many challenges still remain. First, while we are good at predicting absolute risks, we are poor at predicting conflict dynamics (onsets, escalations, de-escalations and terminations). Second, we are over-reliant on spatio-temporal features and mechanistic models due to the nature of the event-data we use, thus excluding actor agency. Third, we do not handle either data or model uncertainty. Fourth, we are lagging behind the state-of-the-art in machine-learning. This dissertation attempts to resolve some of these salient difficulties, by contributing to six core elements of current-generation forecasting systems. First, time, by looking at the substantive effects and uncertainties of the temporal distance between data and forecast horizons. Second, space, by looking at the inherent uncertainties of high-resolution geospatial data and proposing a statistical method to address this. Third, feature space, by tackling the extreme feature sparsity in event-data and proposing a novel, deep active learning approach to mine features from existing large conflict-related text corpora. Fourth, substantive knowledge, by combining findings from the previous papers to take a fresh look at the microdynamics of conflict escalation. Fifth, the forecasting process itself, by building models that directly forecast from text, eliminating the intermediate step of manual data curation. Finally, the frontier of event-data, by looking at whether the news-media heavy way we collect violent fatal events can be extended to the collection of non-violent events. Methodologically, the dissertation introduces state-of-the art methods to the field, including the use of large language models, Gaussian processes, active learning and deep time series modelling. The six papers in the dissertation exhibit significant performance improvement, especially in forecasting dynamics.

This study advances the field of conflict forecasting by using text-based actor embeddings with transformer models to predict dynamic changes in violent conflict patterns at the actor level. More specifically, we combine newswire texts with structured conflict event data and leverage recent advances in Natural Language Processing (NLP) techniques to forecast escalations and de-escalations among conflicting actors, such as governments, militias, separatist movements, and terrorists. This new approach accurately and promptly captures the inherently volatile patterns of violent conflicts, which existing methods have not been able to achieve. To create this framework, we began by curating and annotating a vast international newswire corpus, leveraging hand-labeled event data from the Uppsala Conflict Data Program. By using this hybrid dataset,  our models can incorporate the textual context of news sources along with the precision and detail of structured event data. This combination enables us to make both dynamic and granular predictions about conflict developments. We validate our approach through rigorous back-testing against historical events, demonstrating superior out-of-sample predictive power. We find that our approach is quite effective in identifying and predicting phases of conflict escalation and de-escalation, surpassing the capabilities of traditional models. By focusing on actor interactions, our explicit goal is to provide actionable insights to policymakers, humanitarian organizations, and peacekeeping operations in order to enable targeted and effective intervention strategies.

The proliferation of large-scale, geographically disaggregated data on armed conflicts, protests, and similar events has opened new avenues of research, but has also introduced significant data quality challenges. A notable yet often overlooked issue involves observations with “known geographic imprecision” (KGI), where event locations are unknown and instead arbitrarily assigned by dataset authors. Although this issue is widely recognized and accounts for up to a quarter of observations in datasets like UCDP GED, it is rarely addressed by users. This paper presents a stochastic method derived from the multiple-imputation literature, employing spatio-temporal Gaussian processes and leveraging latent actor-conflict features in the data to enhance location accuracy. Extensive Monte-Carlo simulations demonstrate that this approach substantially enhances the accuracy of these observations and improves predictive performance beyond the state-of-the-art when applied out-of-sample. Additionally, an adapted version of the UCDP GED dataset that employs this new procedure is provided, showcasing the practical application and benefits of the methodology.

In this article, we present the most up-to-date, fine-grained, global dataset on external support in armed conflicts: the UCDP External Support Dataset (ESD). The dataset encompasses data on states and non-state actors as both supporters and recipients and provides detailed information on the type of support provided to warring parties in armed conflicts between 1975 and 2017. We use it to highlight three broader trends in the provision of external support: (1) a dramatic increase in the number of external supporters, (2) a larger share of pro-government interventions, and (3) the rise of direct military intervention as the predominant mode of external support. In conclusion, we identify several avenues worthy of future inquiry that could significantly improve our understanding of external support in armed conflicts.

Temporally and spatial disaggregated datasets are commonly used to study political violence. Researchers are increasingly studying the data generation process itself to understand the selection processes by which conflict events are included in conflict datasets. This work has focused on conflict fatalities. In this research note, we explore how non-fatal conflict events are reported upon and enter into datasets of armed conflict. To do so, we compare reported non-fatal conflict events with the population of events in two direct observation datasets, collected using a boots-on-the-ground strategy: mass abductions in Nepal (1996-2006) and troop movements in Darfur. We show that at the appropriate level of aggregation media reporting on abductions in Nepal largely mirrors the "true" population of abductions, but at more disaggregated levels of temporal or spatial analysis, the match is poor. We also show that there is no overlap between a media-driven conflict dataset and directly-observed data on troop movements in Sudan. These empirics indicate that non-fatal data can suffer from serious underreporting and that this is particularly the case for events lacking elements of coercion. These findings are indicative of selection problems in regards to the reporting on non-fatal conflict events.

Common political economy models point to rationalist motives for engaging in conflict but say little about how income shocks translate into collective violence in some cases but not in others. Grievance models, in contrast, focus on structural origins of shared frustration but offer less insight into when the deprived decide to challenge the status quo. Addressing these lacunae, we develop a theoretical model of civil conflict that predicts income loss to trigger violent mobilization primarily when the shock can be linked to pre-existing collective grievances. The conditional argument is supported by results of a comprehensive global statistical analysis of conflict involvement among ethnic groups. Consistent with theory, we find that this relationship is most powerful among recently downgraded groups, especially in the context of agricultural dependence and low local level of development, whereas political downgrading in the absence of adverse economic changes exerts less influence on ethnic conflict risk.

Although substantive agreement exists on the role of climate variability and food scarcity in increasing violence, a limited number of studies have investigated how food resources affect violent conflict. This article explores the complex linkages between climate variability, agricultural production and conflict onset, by focusing on the spatial distribution of crop production in a cross-country setting. We hypothesize that spatial differences in crop production within countries are a relevant factor in shaping the impact of climate variability on conflict in agriculturally -dependent countries. To test this hypothesis, we rely on high-resolution global gridded data on the local yield of four main crops for the period 1982–2015 and aggregate the grid-cell information on crop production to compute an empirical indicator of the spatial concentration of agricultural production within countries. Our results show that the negative impacts of climate variability lead to an increase in the spatial concentration of agricultural production within countries. In turn, the combined effect of climate extremes and crop production concentration increases the predicted probability of conflict onset by up to 14% in agriculturally dependent countries.

This article presents an update to the ViEWS political Violence Early-Warning System. This update introduces (1) a new infrastructure for training, evaluating, and weighting models that allows us to more optimally combine constituent models into ensembles, and (2) a number of new forecasting models that contribute to improve overall performance, in particular with respect to effectively classifying high- and low-risk cases. Our improved evaluation procedures allow us to develop models that specialize in either the immediate or the more distant future. We also present a formal, 'retrospective' evaluation of how well ViEWS has done since we started publishing our forecasts from July 2018 up to December 2019. Our metrics show that ViEWS is performing well when compared to previous out-of-sample forecasts for the 2015-17 period. Finally, we present our new forecasts for the January 2020-December 2022 period. We continue to predict a near-constant situation of conflict in Nigeria, Somalia, and DRC, but see some signs of decreased risk in Cameroon and Mozambique.