Connections and Political Survival in Personalist Autocracies: The Case of Russia’s Governors

What role do personal connections play in political careers within personalist autocracies? The existing scholarship on patronage and elite mobility has primarily focused on party regimes—especially China. While the impact of political ties varies across levels of government and between party and state hierarchies, scholars have strong evidence to argue that advancement to the upper echelons of China’s party nomenklatura is nearly impossible without close connections to senior party cadres. The extensive knowledge of elite mobility in China and other party regimes (e.g., the USSR) contrasts sharply with the limited evidence on the most common nondemocratic systems today: personalist autocracies. Unlike party regimes, they typically lack a single institutionalized hierarchy—such as a party apparatus—through which careers and connections are formed. Promotion decisions are concentrated in the leader’s hands rather than collective bodies like the Politburo. Additionally, because personalist autocracies often hold multiparty elections, the ability to deliver votes may matter more than long-term personal loyalty. I test the relative impact of connections vis-à-vis other factors—such as performance and electoral mobilization—on political careers by examining one case of personalist autocracy: Russia and its subnational leaders (governors). Their tenures vary substantially in length and subsequent career trajectories. I ask whether shared career or educational ties with federal elites help explain this variation. Does a governor’s position within the elite network, rather than only direct ties to patrons, shape career prospects? Does the quality of ties matter (e.g., early-career, long-lasting, or repeated co-working)? Or, given the regime’s personalist nature, are connections to the leader and his close circle ultimately decisive? Beyond presenting empirical findings, I also address the methodological challenges of collecting data on career connections. The analysis relies on two different approaches, both leveraging large language models (LLMs). In the first, I fine-tune an LLM (OpenAI 4-o model) to detect whether pairs of biographical texts contain evidence of shared work experience among all politicians in the dataset (n ≈ 40,000 pairs). A small set of biographies with manually validated connections serves as the ground-truth training data. In the second approach, I use the same model to convert raw biographical texts into structured CV-like records and then identify shared career episodes using programming script. The paper compares the outcomes and limitations of these approaches. Preliminary results highlight at least three conditions under which naïve LLM-based identification of career ties becomes less reliable. First, unstable institutional environments—frequent renaming, inconsistent abbreviations, and organizational restructuring—complicate accurate matching of organizational names. Second, large variations in organizational size mean that detecting meaningful shared experience often requires identifying suborganizational units, which increases prompt complexity and faces data limitations. Third, identifying ties formed within bureaucratic hierarchies requires reconstructing chains of command—a dimension rarely specified in biographical sources and one that often depends on expert knowledge.