Alan Al Yussef
Postdoctoral researcher
Research
Research area:
Somewhere between political economy, microeconomics, political science, economic sociology
Publications:
» Journal articles:
Abstract. Polarization is increasingly seen as a phenomenon with far reaching and precarious consequences that extend beyond politics. Despite the extensive attention polarization has received, two crucial aspects of this phenomenon remain relatively understudied: electoral polarization and its socioeconomic drivers. We argue that socioeconomic anxieties—the fear of losing one’s social position due to economic insecurity—play an important role in driving electoral polarization. To measure the level of electoral polarization, we develop an intuitive metric based on the dyadic ideological distance between voters. Using elections data from federal elections in Belgium, we study the electoral outcomes of the 300 Flemish municipalities, where we find a significant link between risk of poverty, income, and unemployment, on the one hand, and the level of electoral polarization on the other. This provides empirical support to our overarching hypothesis that socioeconomic anxiety plays a centrifugal electoral impact, amplifying the appeal of polar parties, thereby increasing the level of electoral polarization.
Abstract. Political selection is crucial for the functioning of democracy. However, the practice—in education and sports contexts—of artificially dividing school-age children into different age groups leads to a considerable bias in this selection. The probability of becoming a (successful) politician depends on individuals’ relative age. Being born shortly after the cut-off date significantly increases the probability that an individual will be politically successful later in life. Using a regression discontinuity design, we find strong evidence of such relative age effect (RAE) among a large sample of Belgian federal parliamentarians over the period 1950–2019 (N = 4032), but not among municipal councillors (N = 7387), nor among municipal candidates (N = 36,740) in the 2018 election. The estimated overrepresentation of federal members of the parliament (MPs) born immediately after the cut-off date is up to 90% compared to politicians born just before the cut-off date. The overrepresentation is observed over the whole period and thus seems to be deeply rooted in the political system. We find the RAE to have a gendered dimension: The effect is driven by early-born male politicians’ overrepresentation. No significant RAE was found among female politicians.
» Book chapters:
Working papers:
Turnout and the Distortion of Representation: The Case of Electoral Polarisation (Accepted for publication in Public Choice, 1 June 2026)
Alan Al Yussef
Abstract. This paper examines the relationship between voter turnout and electoral polarization. While much of the existing literature focuses on how polarization affects turnout, we analyze the next step in the causal chain: how turnout shapes electoral outcomes. We propose a simple theoretical model in which turnout depends on voters’ aversion to the expected governing coalition, which increases with the distance between the voter’s ideal point and the anticipated policy platform. This generates heterogeneous participation across the ideological spectrum, leading to a compositional reweighting of the electorate. When the governing coalition is expected to be centrist, centrist voters are underrepresented while voters at the ideological extremes are overrepresented, resulting in higher levels of observed electoral polarization. We test the theoretical predictions empirically, using municipality-level data from Dutch general elections. The results show that turnout has systematically heterogeneous effects across parties depending on their ideological position, with more extreme parties benefiting disproportionately from lower turnout. Consistent with the theory, lower turnout is associated with higher levels of electoral polarization. These findings highlight a compositional mechanism through which turnout influences electoral outcomes, resulting in a distorted picture in which electoral polarization exceeds what would be implied by the underlying distribution of ideological preferences in the population.
A Political Economy of Culture Wars: Polarisation, Sticky Preferences and Mobilisation
Alan Al Yussef
Abstract. This paper presents a theoretical model explaining—from the supply side—why political parties may increase campaign resources devoted to certain issues, even without shifts in public demand. The core mechanism, examined within a two-dimensional issue space, is that heightened internal polarisation on one dimension decreases the marginal productivity of campaigning on that dimension, pushing parties to redirect resources toward the other dimension. Thus, the model illustrates how intra-group polarisation, by altering relative campaign costs across issue dimensions, contributes directly to broader inter-group polarisation. Additionally, external shocks affecting voter positions across both issue dimensions can encourage parties to reallocate resources away from the more volatile dimension to the one characterised by more sticky preferences, enhancing campaign cost-effectiveness. This dynamic helps clarify why, following significant economic disruptions, political parties often–paradoxically–increase their emphasis on cultural and identity-related issues rather than economic policy. Similarly, rising income inequality can intensify internal polarisation on economic issues, making campaigning on cultural and identity-related issues relatively more efficient.
Work in progress
Ambiguously human: LLMs Over-ideologize Political Statements
Alan Al Yussef
Abstract. Large language models (LLMs) are increasingly proposed as substitutes for expert judgment in the production of democratic information goods. This paper examines whether LLMs can replicate one specific form of political expertise: the ideological classification of policy statements. Using 74 statements from the Belgian Voting Aid Application (VAA), we compare classifications produced by four leading LLMs with benchmark labels assigned by expert political scientists. We find that LLMs classify clearly ideological statements with impressive accuracy, achieving scores exceeding 97% accuracy. However, all models struggle with statements that experts classify as neutral. Rather than exhibiting a consistent left- or right-wing bias, the models display a systematic tendency to assign ideological meaning to statements that human coders regard as non-ideological. We refer to this pattern as over-ideologization. The findings suggest that LLMs can substantially reduce the costs of political coding and may serve as useful assistants in the production of democratic information goods. However, they remain imperfect substitutes for expert judgment because they systematically under-recognize political ambiguity.