In this paper, I study the behavioural motivations underlying job referral decisions. In a field experiment in Addis Ababa, Ethiopia, I show that when choosing who to refer for a job, individuals trade off personal benefits and altruistic considerations, with important implications for the efficiency and equity of the referral process. Using complete data on urban social networks and generating a panel of real work and referral opportunities over multiple sessions, I first show that workers rely on reciprocity. This leads to both significant on-the-job productivity losses and persistence in the exclusion of less connected individuals. This dynamic reciprocity is reduced under incentivised referrals, where workers are paid according to the output of their referral, which makes them screen more productive workers. Second, I find that peripheral workers use job referrals strategically to enlarge their network: they are more likely to establish new and reciprocated links, with connections persisting after 18 months. I show that these findings are consistent with a network-based job referral model where individuals trade off social payoffs and altruistic considerations. My findings suggest that conventional job referrals through social networks can reinforce labour market inequalities and prevent less socially connected individuals from getting access to jobs. However, when given referral opportunities, these individuals can manage to escape exclusion even in the long-run. Policy-makers could exploit this and provide subsidised one-off work and referral opportunities for individuals detached from the labour market, with the goal of alleviating long-term unemployment.



Do policy interventions disrupt social networks? We study how a job-search assistance intervention in Addis Ababa, Ethiopia, affects the people who exchange labour market information with programme participants prior to treatment. The intervention is a financial subsidy to use public transport which increases job search intensity among treated individuals. We find that the job-search partners of treated individuals reduce their job search efforts compared to the partners of untreated jobseekers. This is not because they receive more information about vacancies from their treated friends. On the contrary, we document less information sharing and social interaction between job-search partners. We present suggestive evidence that this may be because cooperation in job search becomes harder when one jobseeker has access to more resources than the other.


Bringing young people into productive work is a key labor market challenge in both developing and developed economies, and a multitude of labor market interventions have been implemented to assist vulnerable youths. To assess whether these interventions have succeeded in improving young people's labor market outcomes, this study systematically and quantitatively reviews 113 impact evaluations of youth employment programs worldwide. Of a total of 3105 effect estimates we extract from these studies, one-third are positive significant. The unconditional average effect size across all programs is small, both for employment-related outcomes (Hedges' g = 0.05, SE = 0.02) and earnings-related outcomes (Hedges' g = 0.04, SE = 0.02). We analyze correlates of success in a meta-regression framework. We find that (i) programs are more successful in middle- and low-income countries; (ii) the intervention type is less important than design and delivery; (iii) programs integrating multiple services are more successful; (iv) profiling of beneficiaries, individualized follow-up systems and incentives for services providers matter; and (v) impacts are of larger magnitude in the long-term. Some of these findings provide new and important insights about the design and delivery of interventions, whereas others confirm those of previous reviews. Ultimately, our findings provide practitioners with an improved evidence base about how certain design features contribute to successful youth employment programs in different contexts.


The Market-Reach of Pandemics: Evidence from Female Workers in Ethiopia’s Ready-Made Garment Industry

[Forthcoming, World Development]

(with Morgan Hardy, Gisella Kagy, Christian Johannes Meyer, and Eyoual Tamrat)


In a globalized world, pandemics transmit impacts through markets. We document employment changes, coping strategies, and welfare indicators of garment factory workers in Ethiopia’s largest industrial park during the early stages of the Coronavirus Disease 2019 pandemic. We field a phone survey on female workers employed at the start of the crisis during a two month period in which cases are rapidly rising globally, but not locally. Our data suggest significant changes in employment, high levels of migration away from urban areas to rural areas if women are no longer working, and high levels of food insecurity. These findings compel a research and policy focus on documenting and mitigating the market-reach of pandemics on low-income women at the margins.

The Impact of COVID-19 on the Lives of Women in the Garment Industry: Evidence from Ethiopia

(with Morgan Hardy, Gisella Kagy, Christian Johannes Meyer, and Eyoual Tamrat)

[project description] [materials]

Global Behaviors and Perceptions in the COVID-19 Pandemic

[Revise & resubmit, PNAS]

(with Thiemo Fetzer, Lukas Hensel, Jon Jachimowicz, Johannes Haushofer, Andriy Ivchenko, Stefano Caria, Elena Reutskaja, Christopher Roth, Stefano Fiorin, Margarita Gómez, Gordon Kraft-Todd, Friedrich Gotz, Erez Yoeli)


We conducted a large-scale survey covering 58 countries (N = 108,075) at the onset of the COVID-19 pandemic between March 20th and April 7th 2020—to explore how beliefs about citizens’ and government’s response to the COVID-19 pandemic, and the actions taken by governments, affected mental well-being. Our analyses reveal three findings. First, many respondents indicate that their country’s citizens and government’s response was insufficient. Second, respondents’ perception of an insufficient public and government response was associated with lower mental well-being. Third, we exploit time variation in country-level lockdown announcements, both around the world and through an event-study in the UK, and find that strong government actions—i.e., announcing a nationwide lockdown—were related to an improvement in respondents’ views of their fellow citizens and government, and to better mental well-being. These findings suggest that policy-makers may not only need to consider how their decisions affect the spread of COVID-19, but also how such choices influence the mental well-being of their population.


Formal Hiring Practices, Firm Growth, and Inclusive Labour Markets

(with Lukas Hensel and Tsegay Gebrekidan Tekleselassie)

Network Position and Public Good Contributions

(with Simon Franklin)

Economic Impacts of Mental Health Interventions in Low and Middle Income Countries: A Systematic Review and Meta-Analysis

(with Thandi Davies, Johannes Haushofer, Crick Lund, Kate Orkin)

This paper systematically and quantitatively reviews 39 mental health trials conducted in lower- and middle income countries to assess whether the interventions causally changed the economic and employment position of participants. This is the first meta-analysis studying the impact of mental health interventions on economic outcomes.

Urban Job Search Networks Along a Spatial Dimension

This paper introduces a unique set of urban job network data for young unemployed residents of Addis Ababa, Ethiopia. I use these data to test the assumption that locally truncated networks are a reasonable approximation for the complete network structure. Comparing restricted neighbourhood contacts with the city-wide job contacts of a sample of job-seekers, I find that the unrestricted city-wide links display a higher degree of homophily in demographics than local links, but a lower degree of homophily on work-related and aspirational characteristics. This is in line with the intuition of a simple theoretical model on geographical link formation that I present.

Overcoming Gender-Biases in Social Network Elicitation

I discuss two alternative ways of network elicitation: eliciting networks with open-ended listings from memory yields fewer than a third of connections than using pre-populated network rosters, omitting most female job contacts. This underreporting of women could be linked to later discrimination of women at the job referral stage (cf. Beaman et al., 2018). In a survey experiment, I develop and test two different interventions to nudge participants into reporting the correct gender balance of their network in open-ended network surveys.


Can mental health treatments help prevent or reduce intimate partner violence in low- and middle-income countries? A systematic review

[BMC Women's Health, 2019, 19:34]

(with W. A. Tol, S. M. Murray, C. Lund, P. Bolton, L. K. Murray, T. Davies, J. Haushofer, K. Orkin, L. Salama, V. Patel, G. Thornicroft and J. K. Bass)



Epidemiological research suggests an interrelationship between mental health problems and the (re)occurrence of intimate partner violence (IPV). However, little is known about the impact of mental health treatments on IPV victimization or perpetration, especially in low- and middle-income countries (LMIC).


We conducted a systematic review to identify prospective, controlled studies of mental health treatments in LMIC. We defined ‘mental health treatment’ as an intervention for individuals experiencing mental ill health (including substance misuse) including a substantial psychosocial or pharmacological component. Studies had to measure a mental health and IPV outcome. We searched across multi-disciplinary databases using a structured search strategy. Screening of title/abstracts and full-text eligibility assessments were conducted by two researchers independently, data were extracted using a piloted spreadsheet, and a narrative synthesis was generated.


We identified seven studies reported in 11 papers conducted in five middle-income countries. With the exception of blinding, studies overall showed acceptable levels of risk of bias. Four of the seven studies focused on dedicated mental health treatments in various populations, including: common mental disorders in earthquake survivors; depression in primary care; alcohol misuse in men; and alcohol misuse in female adult sex workers. The dedicated mental health treatments targeting depression or alcohol misuse consistently reduced levels of these outcomes. The two studies targeting depression also reduced short-term IPV, but no IPV benefits were identified in the two alcohol-focused studies. The other three studies evaluated integrated interventions, in which a focus on substance misuse was part of efforts to reduce HIV/AIDS and violence against particularly vulnerable women. In contrast to the dedicated mental health interventions, the integrated interventions did not consistently reduce mental ill health or alcohol misuse compared to control conditions.


Too few studies have been conducted to judge whether mental health treatments may provide a beneficial strategy to prevent or reduce IPV in LMIC. Key future research questions include: whether promising initial evidence on the effects of depression interventions on reducing IPV hold more broadly, the required intensity of mental health components in integrated interventions, and the identification of mechanisms of IPV that are amenable to mental health intervention.