This project has received funding via the EUI Research Council call 2024.
This project aims to study wartime regime loyalty in Nazi Germany 1939 – 1945: a closed autocracy that severely repressed the public expression of political dissent. Our approach leverages information from more than 150,000 newspaper issues published during WWII. We use image analysis and LLMs to detect expressions of regime approval in private obituaries. This will allow us to provide the first granular, time-varying measurement of popular support in a closed autocracy during wartime. In the second step, we use this data to study the determinants of regime loyalty, focusing on exposure to (i) air raids, (ii) military casualties, and (iii) economic deprivation.
Maintaining popular support is crucial for dictatorships to sustain power during wartime. The loss of such support can threaten regime survival by fostering domestic resistance (Adena et al., 2021), reducing combat motivation (Barber and Miller, 2019), and creating fractures within the elite coalition (Wintrobe, 1998). Understanding the factors that influence regime loyalty in wartime is crucial for at least two reasons. First, it deepens our understanding of how dictatorships function under the extreme pressures of war. Second, it provides insights into the dynamics within contemporary autocracies engaged in warfare, such as, for example, Russia today.
However, studying regime loyalty in wartime presents significant challenges. In dictatorships, reliable data on political attitudes is scarce, particularly regarding sensitive topics like regime loyalty. The threat of repression fosters preference falsification (Kuran, 1995), and regimes have strong incentives to manipulate available data. The high risks and high stakes in international wars often exacerbate these dynamics and increase the challenges of measuring the “true” political perceptions and attitudes of citizens. While innovative survey methods like list experiments have highlighted the prevalence of self-censorship in autocracies, they are unsuited for measuring changes in regime support at a high temporal and spatial resolution. Ultimately, regime loyalty can only be accurately measured based on a behavioural measure that remains independent of regime interference and repression.
This project addresses these challenges by leveraging unique data from a historical case: Nazi Germany during World War II, where the massive extent of regime repression poses formidable challenges for a reliable measurement of citizens’ wartime loyalty.
Our key innovation is measuring regime loyalty through observed behaviour by analysing wartime obituaries in local newspapers. Hume (2000) argues that obituaries reflect “what society values and wants to remember.” Thus, obituaries' content, language, and symbolism can serve as proxies for communities’ values and ideals. Importantly, historical research has established that wartime obituaries for fallen soldiers during WW2 constituted one of the few channels of public expression that have long remained relatively free of heavy regime censorship. Obituaries, thus, provide a unique opportunity to observe citizens’ attitudes toward the war and the incumbent regime (Kershaw, 2001). Some obituaries openly display support for the war and loyalty to the regime by invoking Adolf Hitler and celebrating heroism for the Fatherland. In contrast, others avoid such nationalist rhetoric, highlight the tragedy of death, and depart from the regime’s glorification of the war–contradicting the regime’s war propaganda.
In this project, we will use advanced machine learning and AI methods to systematically measure variation in regime loyalty in these obituaries on a large scale. We classify and extract precise wordings and references within these obituaries from a corpus of about 1.8 million historical and digitized newspaper pages from North-Rhine Westphalia, Germany's most densely populated region. Specifically, we will retrieve (1) information on soldiers’ names, hometowns, and death dates, (2) the names of soldiers’ relatives that published the obituaries, as well as (3) various references to regime narratives of the war (e.g., the use of the phrase “for the Führer, the People and the Fatherland”). The digital newspaper archive “Zeitpunkt NRW” provides access to high-resolution scans of historical newspapers, including metadata on newspaper distribution areas. This will allow us to provide the first measure of regime sympathy in Germany during WW2 with unprecedented spatial and temporal detail.
In the second step of the project, we use this novel measure to examine how citizens’ personal exposure to the costs of the war shaped their attitudes toward the regime. We focus on three types of costs. First, we draw on data from the US Department of Defense, covering 178,000 Allied sorties, to measure local exposure to Allied bombing attacks. Second, we draw on counts of obituaries to measure variation in localities’ exposure to previous battlefield losses. Finally, we exploit temporal changes in food wartime food rationing to capture exposure to the economic costs of the war. In our main analysis, we investigate how citizens’ exposure to bombings, losses, and food scarcity influence expressions of regime loyalty in obituaries. We expect fewer expressions of loyalty in areas with relatively high exposure to the costs of the war.