Essays on urban job networks in developing countries

• Main Author: Witte, Marc
• Format: Dissertation
• Language: English
• Published: University of Oxford 2018

Across the globe, social networks influence labour market outcomes. This thesis investigates job networks-networks in which individuals exchange jobs and information about vacancies - in cities of developing countries. Using the example of Addis Ababa, Ethiopia, I describe such networks, analyse how individuals use them for their economic benefit, and show how the networks respond to external shocks. Chapter one presents experimental evidence on job referrals in informal urban labour markets, investigating the relationship between social network centrality and persistence in labour market outcomes. I provide day work opportunities to young casual workers and vary the conditions by which they can refer their neighbourhood contacts to the same job. I find that central individuals are more likely to receive job referrals, and that there is a strong norm of reciprocal referring. While both factors lead to productivity losses and continuous neglect of peripheral workers, I show that socially excluded workers establish and consolidate new links when they can refer someone to the job. This suggests that individuals can integrate into job referral networks when given the opportunity. Chapter two studies information sharing about job opportunities in the social networks of young job-seekers, who collaborate in job search by sharing expenses and information about vacancies across the city. A job search assistance intervention, targeting only one individual in a pair of job-seekers, can break such collaborative informal arrangements. It is important to consider such social network externalities when evaluating the impact of labour market activation programmes. Chapter three contributes to the literature on the measurement of urban social networks, with a particular focus on job information exchange among individuals. I show that using truncated or local network data - such as from census-blocks - can introduce a bias when the network of interest is spread over larger spaces like whole cities. I compare two different network elicitation methods - pre-populated census-based rosters and open-ended, memory-based listings. I show that within geographically limited and tractable entities such as neighbourhoods, the former method yields more connections than the latter, which systematically underreports women and older contacts and relies more strongly on bilateral links.