Place-based industrial policies are widely used to attract investment and stimulate employment in targeted locations. Yet subsidized firms can expand without increasing local employment. I examine where the labor supporting this expansion comes from, using Colombia’s Free Economic Zones (FEZs) and a novel georeferenced panel linking municipalities, firms, and workers to the staggered rollout of zones between 2005 and 2018. I find that FEZs increase municipal informality by roughly five percentage points while leaving total employment unchanged. Firms entering the zones are skill-intensive and draw skilled workers from incumbent employers, which raise wages in response. Small, less-productive firms contract or exit, while less-educated workers become 6 percentage points more likely to be informal. These patterns reveal a reallocation of workers across firms rather than an expansion of local employment. The findings show that a large informal sector does not imply an elastic supply of the skills expanding firms demand: when those skills are locally scarce, industrial policy can expand targeted firms while increasing informality and generating no net employment gains.

SPANDE 2026*, European Development Economics Group (EDEG) Conference 2026, EAYE 2026, CSAE Oxford 2026, LACEA 2025, UNU-WIDER Seminar, JDD Job Market 2025, Helsinki Economics Group Development Seminar, LSE Workshop of Early Career Women in Economic Geography and Spatial Economics, IFS-UCL-LSE/STICERD Development Seminar, HEC PhD Seminar 2025, Junior Workshop ENS de Lyon 2024, LAGV 2024, Konstanz Brown Bag Seminar, Urban Economic Association Summer School 2024, QMUL Workshop 2024, AMSE PhD Seminar.

Domestic violence against women and children is persistent worldwide. This paper examines Bogotá’s Care Blocks, a place-based care infrastructure policy that provides free integrated services to women who perform unpaid domestic and care work, as well as the dependents in their care, mainly children. Using the staggered rollout of Care Blocks across neighborhoods, along with variation in households’ proximity to the centers, we estimate the causal effect of access to care infrastructure on domestic violence incidents using a difference-in-differences framework and administrative records. We find that exposure to Care Blocks reduces domestic violence by 22%, driven by declines in intimate partner violence (IPV) and violence against children (VAC). To shed light on the potential mechanisms, we exploit survey data to show that exposure to Care Blocks services likely reduces time spent with potential perpetrators at home by shifting daily routines. We find that women are more likely to enroll in educational and training programs, while children spend more time in leisure activities such as sports and recreation. Our findings suggest that public investments in care infrastructure generate substantial welfare and social returns, and that place-based care policies can offer an effective, scalable approach to preventing domestic violence in developing countries.

LISER Gender and Economics Workshop*, AMSE DEVPOL Informal Seminar.

Developing artificial intelligence (AI) technologies requires access to data, yet little is known about the economic forces that shape firms’ access to commercially valuable data. This paper studies whether access to foreign markets contributes to AI innovation by expanding firms’ access to data generated through interactions with foreign customers. We combine firm-level financial information from CMIE Prowess, patent records from the Indian Patent Office, and newly developed crosswalks linking firms’ products and services to detailed trade classifications. Exploiting plausibly exogenous variation in foreign demand through a shift-share design, we find that stronger export opportunities increase both the probability and intensity of AI patenting and are accompanied by higher expenditure on data-related activities. To examine whether commercially generated data constitute the underlying mechanism, we exploit the implementation of the European Union’s General Data Protection Regulation (GDPR), which increased the costs of collecting and processing customer data while leaving firms’ access to foreign markets largely unchanged. Firms with greater pre-existing exposure to the European Union experienced a decline in AI patenting intensity following the GDPR, despite showing little evidence of changes in exports or broader firm performance. Our findings indicate that commercially generated data are an important mechanism through which access to foreign markets fosters AI innovation.

Presentations (* by co-author): ETSG Athens (2024), U Bologna (2025), AMSE (*2026).