Research
Peer-Reviewed
Nathan Barker, C. Austin Davis, Paula López-Peña, Harrison Mitchell, Ahmed Mushfiq Mobarak, Karim Naguib, Maira Reimão, Ashish Shenoy, and Corey Vernot. Migration and Resilience during a Global Crisis. Accepted: European Economic Review.
This study explores the relationship between migration and household resilience during a global crisis that eliminated the option to migrate. We link prior data from four populations in Bangladesh and Nepal to new phone surveys conducted during the early months of the COVID-19 pandemic. While earnings fell universally, pandemic-induced declines were 14–25% greater among previously migration-dependent households and urban migrant workers, with household remittance losses far exceeding official statistics. Heightened economic exposure during the pandemic erased prior gains achieved by transnational migrants and caused fourfold greater prevalence of food insecurity among domestic subsistence migrants. Economic distress spilled over onto non-migrants in high-migration villages and labor markets. We show that migration contributed to economic contagion independent of its role in disease transmission. Losing the option to migrate differentially increased the vulnerability of migration-dependent households during a crisis.
Working Papers
Edward Jee, Anne Karing, and Karim Naguib. Optimal Incentives in the Presence of Social Norms: Experimental Evidence from Kenya.
Reputational incentives interact with economic incentives and may mitigate or amplify their effects (Bénabou and Tirole, 2012). We test this prediction in the context of a new deworming program that offers free treatment to 200,000 adults in Kenya. We experimentally vary the cost and ability to credibly signal program participation by assigning communities to different travel distances to treatment points and providing a bracelet or ink on the thumb as a sign of take-up. Our reduced-form results show that (1) bracelets significantly increase deworming take-up, outperforming a material incentive; (2) adults are highly sensitive to distance cost and the signaling treatments have a greater impact on participation at far distances. We build a structural model that mirrors the theoretical framework outlined by Bénabou and Tirole (2012) and explicitly model latent variables such as the private benefit of incentives, the visibility of the signals, and the reputational returns to signaling. The model allows us to estimate counterfactuals in response to manipulations infeasible to conduct in the experimental setting and investigate how changes in the community level of deworming take-up change the returns to signaling. We show that reputational returns increase at lower levels of take-up, and that these increases mitigate the negative impact of cost on take-up. Accounting for these interactions, it is optimal to set treatment locations further apart, allowing for an expansion of the program.
Harrison Mitchell, A. Mushfiq Mobarak, Karim Naguib, Maira Reimão, and Ashish Shenoy. External Validity and Implementation at Scale: Evidence from a Migration Loan Program in Bangladesh.
Many economic policies show promising pilot results that fail to replicate at scale. We demonstrate how delegation of authority to implementing agents can threaten scalability in a randomized evaluation of a migration loan program in Bangladesh. Pilot evaluations found the loan offer to increase temporary migration by 25–40p.p., but this effect fell to 12p.p. at scale. To explain the attenuation, we introduce a theory of delegation risk that leads implementing agents to systematically mistarget intended program beneficiaries. Mistargeting occurs because benefits are concentrated among those enabled to migrate with a loan — i.e. program compliers — but capacity constraints lead effort to be directed toward those already planning to migrate — i.e. always-takers. We present evidence consistent with this theory that the characteristics that predict pre-loan migration are strongly correlated with the likelihood of remembering the loan offer in endline surveying, and we show delegation risk can quantitatively account for the diminished treatment effect. Policy impacts are further tempered by expansion to adjacent geographic regions despite participants being observably similar. We rule out two additional explanations: First, our geographically clustered randomization design reveals treatment intensity crowds in rather than crowds out migrants. Second, changes in population characteristics over time appear to have little influence. Delegation risk identified in this study has the potential to undermine a number of common development policies, and is exacerbated by management practices frequently used by development organizations.
Preprints
Karim Naguib, Roger Berché, Lu Li, Antonia Bevan, Sajan Khosla, Jessica Davies, and Paul Metcalfe. PIONEER: Bayesian Joint Modelling of Mechanistic Tumour Growth and Time-to-Event Endpoints for Dynamic Prediction of Ongoing Oncology Trials. arXiv:2607.17908.
High-stakes decisions in oncology clinical trials must often be made while survival data remains immature: progression-free survival (PFS) and overall survival (OS) are heavily censored, few events have accumulated, and the primary endpoint may be months or years from reading out. What is available at interim data cut-offs is information-rich longitudinal tumour measurements and baseline covariates. We present PIONEER, a Bayesian joint modelling framework that couples a mechanistic two-component state-space submodel of longitudinal tumour size dynamics to a multistate proportional-hazard submodel for competing clinical events, fitted simultaneously under a single posterior. The mechanistic submodel infers latent per-patient tumour trajectories — decomposed into treatment-responsive and refractory compartments with Gompertz-attenuated growth — from sparse, noisy sum-of-longest-diameter (SLD) observations. These latent trajectories feed the multistate hazard as time-varying covariates, while the event data simultaneously refines the tumour dynamics through the joint likelihood. All clinical endpoints (PFS, OS, objective response rate) are derived from the joint posterior in a single forward simulation pass, propagating full parameter uncertainty without any two-stage plug-in. Applied to a case study in extensive-stage small-cell lung cancer (two trials, N = 497), leave-future-out cross-validation demonstrates that at month 4 of enrolment (9 patients) the model produces calibrated PFS forecasts covering the mature month-19 Kaplan-Meier curve, and at month 11 (39 patients) the OS forecast converges — representing at least 8 months of advance forecasting with properly quantified uncertainty. We hope this work paves the way for broader adoption of Bayesian mechanistic state-space frameworks in clinical development, enabling earlier and more informed decision-making from immature trial data.
See also the Writing page for essays and methodological notes.