Political polarisation is a defining feature of modern societies (Boxell 2021). Not only does it shape elections and policy debates, but also social interactions and behaviours beyond the political sphere (Anderson and Robinson 2024, Braghieri et al. 2025). Its potential to influence where people choose to live has recently attracted scholarly attention (Kaplan et al. 2022). However, whether polarisation-driven spatial sorting affects the regional distribution of human capital, a factor with significant implications for economic development (Mankiw et al. 1992), remains unexplored.
Empirical research on this issue faces two main challenges. First, granular data linking individual human capital to revealed allocation preferences is scarce, making even correlational analysis difficult. Second, plausible exogenous variation in regional polarisation is rare, limiting credible identification strategies.
In recent research (Martinez-Miera and Sunyer 2025), we address these challenges in two ways. First, we exploit Spain’s centralised placement system for medical residents, which provides standardised measures of individual human capital and revealed preferences over hospital-specialty pairs. Second, we leverage the sharp, and plausibly unexpected, increase in polarisation triggered by the October 2017 secessionist crisis in Catalonia, a large region in Spain.
Our main measure of human capital is based on the results of a standardised exam that all medicine students aiming to specialise in Spain must take after their university studies. The grade obtained in this exam is the main determinant of our measure of human capital, which we call their score. Each candidate, starting with the one with the highest score, chooses the hospital and specialty in which they want to complete their residency, until all available positions are allocated.
As shown in Figure 1, prior to 2018, the average score of all doctors who chose Catalonia was above that of the rest of Spain and followed parallel trends. However, from 2018 (the first year after the political polarisation shock), the average score in Catalonia declined sharply, falling below the national average. This decline was not short-lived; it persisted for at least seven years, up to 2024, the last year in our dataset. We estimate that the polarisation shock caused a decrease of 0.13 standard deviations in the average score of residents in Catalonia, equivalent to a drop of 4.16 percentiles or 7.7% relative to pre-shock levels.
Figure 1 Average medical residents’ score, Catalonia vs rest of Spain
Importantly, this average effect masks substantial heterogeneity across the score distribution. Figure 2 provides suggestive descriptive evidence of how the composition of individuals choosing Catalonia for their medical residency shifted by score decile before and after 2018. While the bottom decile saw an increase of 20%, the top decile experienced a decline of nearly 30%, falling from an average of 21% to 15%.
Figure 2 Percentage change in the share of individuals choosing Catalonia for their residency, by score decile, before and after 2018
Since the patterns shown in Figure 2 could reflect, among other reasons, changes in the composition of medical residency positions offered in Catalonia relative to the rest of Spain, in Figure 3, we show the results when we control for both hospital×specialty and specialty×year fixed effects. Using quantile regressions, we confirm that the decline in score, while relevant across the whole scoredistribution, was more intense among high-scoring individuals – namely, those with the greatest outside options. Specifically, while the estimated drop at the 5th percentile is -0.096, it rises in absolute terms to -0.180 among the top 5% of performers.
Figure 3 Impact of political polarisation on human capital, quantile regression
Having documented the negative impact on the attraction of human capital, we consider possible underlying reasons for such a decline. We present five potential explanations for this loss of talent.
First, individuals might be concerned about an increase in the mandatory institutional use of Catalonia’s regional language, Catalan. If this were the case, we would expect individuals from Valencia or the Balearic Islands – regions where the regional languages are closer to Catalan than to Spanish – to be less affected by the shock. However, we find no evidence of any differential effect for this subset of individuals compared with the rest.
Second, the observed loss of talent could reflect a decline in the quality of Catalonia’s university education in the years prior (or even after) the shock. However, we find no evidence of changes in score before and after the shock between individuals who graduated from Catalonia’s universities and those who graduated elsewhere in Spain.
Third, the loss of talent might reflect concerns about potential independence rather than political polarisation. However, this explanation seems unlikely, as the effect is not concentrated in the immediate years around 2018. In fact, the data suggest that the loss of talent grows over time, continuing through 2024, when we argue that the potential independence was less probable. Also, we find that the results are more intense in municipalities with a higher proportion of nationalist votes in previous elections. We argue that this is not in line with a potential independence motive, as these municipalities would be equally affected by independence as municipalities with lower proportion of nationalist votes.
Finally, consistent with the idea of ideological sorting, one might expect that after the shock, individuals supporting independence would move to Catalonia while others would leave. However, using individuals’ home addresses as a proxy for pro-independence likelihood, we find no evidence of a differential effect for likely pro-independence individuals compared with the rest.
By exclusion, we conclude that the most plausible mechanism behind our results is that individuals, on average, avoid polarised environments.
We conclude our analysis by examining whether hospital performance in Catalonia declined after the political polarisation shock, relative to that of hospitals in the rest of Spain. As a productivity outcome, we follow the literature and use in-hospital mortality rates (Doyle et al. 2019). We further restrict attention to patients diagnosed with acute myocardial infarction, heart failure, and pneumonia, as the non-deferrable nature of these conditions makes them standard benchmarks for assessing hospital quality (Aghamolla et al. 2024).
Our estimates suggest that mortality rates for these diagnoses increased by 6.9% in Catalonia compared to other hospitals, consistent with a decline in productivity. Unfortunately, we do not have measures of talent for other professional profiles in the health sector that are relevant to health outcomes and likely affected by the polarisation shock, nor can we link doctors to patients. Hence, these results should be interpreted as suggestive, and consistent with the decline in human capital in part of the labour force that we document, rather than causal.
As the debate on the economic consequences of rising political polarisation continues, our findings highlight its potential to shape the spatial allocation of human capital. Given that talent is closely tied to economic performance, our research contributes to a better understanding of the mechanisms linking polarisation with societal welfare.
Source: cepr.org
Artificial intelligence differs from other technological advancements in finance, such as the initial adoption of…
Industrial raw materials such as nickel, cobalt, and rare earths are critical inputs in countless…
As European governments scale up investment, bond market stability is more critical than ever. This…
Economists have long warned of the negative consequences of excessive US public debt (e.g. Friedman…
Financial distress affects roughly one in five adults in OECD countries (OECD 2024). It constrains…
Until 2018, the US-China trade data gap was in line with the discrepancies found in…