Matías Golman

This study examines the demand for Big Five personality traits in the UK labour market and their interaction with artificial intelligence (AI) adoption. Using text retrieval methods on 11.7 million online job vacancies (2017-2022), I find that 72 per cent of vacancies require at least one personality trait. Extraversion is most common (44 per cent), followed by Conscientiousness (34 per cent) and Openness (30 per cent). After accounting for occupational changes, personality demands remain mostly stable over time, with only Conscientiousness showing a decline (-2.1 percentage points) and Extraversion a modest increase (1.7 percentage points) by 2022. I use two approaches to study how AI affects personality demands. At the vacancy level, jobs requiring AI skills are more likely to demand Openness (7.2 percentage points) and less likely to demand other traits, even within the same firm and occupation. At the firm level, after companies adopt AI for the first time, they increase their demand for Openness (3.6 percentage points) across all subsequent hiring. These findings suggest that as firms adopt AI technologies, they increasingly value workers with traits associated with creativity and adaptability, both for AI-demanding positions and across their broader workforce.

This study examines the relationship between personality trait mentions in job advertisements and wage offers. Using text analysis of 11.7 million UK online job postings from 2017 to 2022, I identify Big Five personality traits in vacancy descriptions. Jobs mentioning personality traits offer 4.6 per cent lower wages on average, creating a paradox: why do employers advertise for traits if these are penalised? I test four explanations: correlation with low-pay job characteristics, compensating differentials, linguistic intensity effects, and demographic targeting. Occupation fixed effects produce the largest reduction in trait penalties, but significant penalties remain. Penalties are larger in high-skill occupations but smaller in female-dominated roles. Trait penalties decrease when combined with traditional benefits and workplace culture signals, but increase with remote work. Linguistic intensity analysis shows that low-intensity mentions carry minimal penalties, but high-intensity mentions show substantial penalties. Machine learning analysis of gender targeting contradicts simple demographic targeting predictions. The findings suggest employers use personality-trait language for multiple purposes rather than a single mechanism: signalling requirements at low intensity, facilitating compensating differentials with appropriate amenity bundles, and serving organisational signalling functions at high intensity.

This paper presents the first comprehensive meta-analysis of formalisation policies in Latin America, analysing 79 impact evaluations with 527 estimates from interventions implemented between 1990 and 2020. We calculate standardised mean differences for 444 estimates and conduct rigorous publication bias analysis using FAT-PET-PEESE methods. Unlike previous meta-analyses, we examine interventions targeting both worker and firm formalisation, include Spanish-language studies and unpublished government reports, and provide the first systematic assessment of effect magnitudes for the region. Our findings show that formalisation policies predominantly produce positive effects, with adverse impacts being rare (5.1% of estimates). However, systematic publication bias substantially reduces effect sizes when corrected. For formality outcomes, active labour market policies show consistent positive effects (0.238 standard deviations, p < 0.05), as do information campaigns (0.226 standard deviations, p < 0.10) and labour inspections (0.320 standard deviations, p < 0.10). For earnings outcomes, social protection programmes with activation components demonstrate the largest effects (1.258 standard deviations, p < 0.01), though negligible formality impact (0.033 standard deviations, not significant). This divergence suggests these programmes primarily affect income rather than formal employment status. Programme benefits strengthen over time, with long-term effects exceeding short-term impacts by approximately 0.10 standard deviations for formality and 0.32 standard deviations for earnings. Multi-component interventions outperform single-component policies by 0.07 standard deviations. These findings underscore the need for realistic expectations while supporting evidence-based, multifaceted approaches to formalisation.