I am an Economist in Development Impact at the World Bank. I completed my PhD in Agricultural and Resource Economics at UC Berkeley in 2019. You can view my CV here.
Contact: jloeser@worldbank.org
Working papers
A few good masks: Evidence from mask manufacturing in Rwanda during the COVID-19 pandemic (w/ Kieran Byrne, Florence Kondylis, and Denis Mukama)
Conditionally accepted, Journal of Development Economics
(2025 Jan) show abstract pdf working paper (WPS; 2022)
Can trade frictions limit access to improved health technologies? Rwanda encouraged and licensed domestic production of high-quality masks by a few selected textile manufacturers at the start of the COVID-19 pandemic. We exploit spatial variation in exposure to mask manufacturing through pre-licensing medical and textile trade networks within an event-study design using receipt-level transaction data. Local markets less exposed to mask manufacturing had higher mask prices, purchased fewer masks, and experienced faster growth in COVID-19 infections proxied by demand for anti-fever medicine. The dynamics of our results suggest that mask quality, rather than quantity, explains reduced infections caused by manufactured masks.
Decentralizing agricultural demonstration to accelerate social learning (w/ Florence Kondylis, Mushfiq Mobarak, Maria Jones, and Dan Stein)
(2024 Sep) show abstract pdf working paper (WPS)
Can decentralizing demonstration accelerate learning about new technologies? In a field experiment, farmers learn about a new flood-saline-resilient seed from a single, centralized demonstration plot in some Bangladeshi villages, while the same demonstration kit is distributed to multiple farmers in other villages. The decentralized delivery allows more farmers to experience the technology for themselves or through their near-neighbors, and increases adoption in the short run, especially in flood-affected villages where the benefits of adoption are largest. These effects disappear after a year. A Bayesian learning model can rationalize these adoption/disadoption patterns. Structural estimates imply that self-experimentation is seven times more valuable than learning from others, yet positive learning externalities generated in the decentralized demonstration arm exceed the cost of the demonstration kit.
Cash transfer size and persistence (w/ Florence Kondylis)
(2024 Jul) show abstract pdf working paper (WPS; 2021)
Do larger cash transfers increase consumption more cost effectively? And should poverty traps motivate increasing the size or the scope of cash transfers? We leverage 38 experimental estiamtes of dynamic consumption impacts of cash transfers and complementary interventions from 14 developing countries. First, increasing transfer size decreases cost effectiveness and does not affect persistence of impacts. Second, while adding complementary interventions increases impacts and persistence, it reduces cost effectiveness and increases heterogeneity. The absence of poverty traps is neither necessary nor sufficient for these results.
Intertemporal choice bracketing and the measurement of time preferences (w/ Yonas Alem and Aprajit Mahajan)
(2024 Jun) show abstract pdf working paper (EfD)
The implications of commonly used money earlier or later (MEL) games for intertemporal behavior depend critically upon subjects' choice bracketing. If subjects bracket narrowly, responses reflect preferences independent of subjects' financial environment. Alternatively, if subjects bracket broadly, responses reflect subjects' marginal returns to investment. We test both hypotheses in a lab-in-the-field experiment, which involves repeated MEL games, a large unconditional cash transfer, and an illiquid savings product. Subjects do not narrowly bracket - randomized cash transfers induce greater patience in MEL choices. Subjects do not broadly bracket either - they fail to arbitrage across equivalent MEL and savings opportunities. We provide a conceptual framework and present evidence that narrowly bracketing subjects drive the predictive power of MEL outcomes for financial choices, justifying the common practice of interpreting MEL choices as a proxy for time preferences rather than financial environment.
Local average treatment effects for multiple treatments with multiple instruments
(2024 May) show abstract pdf
When individuals self-select into multiple treatments, additional assumptions are needed to identify treatment effects with instrumental variables. Suppose each treatment is targeted by a single instrument, and one of three equivalent assumptions holds: no defiers, identical compliers, or additive random utility. With continuous instruments, average marginal treatment effects for individuals indifferent between control and K treatments ("MTE-K") are identified. With price instruments in a 3x3 experimental design, absent income effects, local average treatment effects for two treatments for common compliers ("LATE-2") are informatively partially identified: as experimental price variation shrinks, LATE-2 bounds converge to MTE-2.
Consumer surplus with incomplete markets: Applications to savings and microfinance
(2023 Jun) show abstract pdf working paper (WPS)
The household welfare gains from financial inclusion are empirically elusive. I establish that household welfare gains from a financial technology are equal to the area under dynamically compensated demand in a household model with incomplete financial markets, and general technology, preferences, and choice sets. I then estimate compensated demand for financial technologies leveraging three randomized control trials that introduce experimental variation in interest rates. Welfare gains per dollar lent or saved are small as compensated demand elasticities are large, but still correspond to large aggregate welfare gains from financial inclusion.
Publications
Does gender tagging public works increase women's participation? Experimental evidence from Haiti, Kenya, and Rwanda (w/ Tanay Balantrapu, Paul Christian, Lelys Dinarte-Diaz, Felipe Dunsch, Jonas Heirman, Dahyeon Jeong, Erin Kelley, Florence Kondylis, and Gregory Lane)
AEA Papers and Proceeedings 115 pp. 318-323
(2025 May) show abstract pdf
Public works programs often fail to induce participation by women in dual-headed households, with potential implications for closing gender gaps in autonomy. We randomize "Gender Tagging", labeling programs as "for women", in cash-for-work programs that target poor households in three countries (Haiti, Kenya, and Rwanda). Gender Tagging increases women's participation by 11-to-27 percentage points (29-190%) in contexts where women have low labor market attachment. We apply recent econometric methods to test effect heterogeneity across households, to show that Gender Tagging generates catch up: women least likely to participate under the status quo program experience the largest increase in participation from Gender Tagging.
Factor market failures and the adoption of irrigation in Rwanda (w/ Maria Jones, Florence Kondylis, and Jeremy Magruder)
American Economic Review 112(7) pp. 2316-2352
(2022 Jul) show abstract pdf online appendix
Factor market failures can limit adoption of profitable technologies. We leverage a plot-level spatial regression discontinuity design in the context of irrigation use by farmers provided free access to water. Using irrigation boosts profits by 43-62%. Yet, farmers only irrigate 30% of plots because of labor costs. We demonstrate inefficient irrigation use, by showing farmers irrigating one plot reduce their irrigation use on other plots. This inefficiency is largest for smaller households and wealthier households, suggesting labor market frictions constrain use of irrigation.
Other publications
Sectoral heterogeneity in the COVID-19 recovery: Evidence from Rwanda (w/ Kieran Byrne, Saahil Karpe, Florence Kondylis, and Megan Lang) featured in Arezki, R., Djankov, S., and Panizza, U., "Shaping Africa's Post-Covid Recovery" (CEPR/VoxEU, February 2021)
(2021 Feb) show abstract pdf
Following the initial COVID-19 shock, developing countries have begun to transition to a COVID-19 economic recovery characterized by eased lockdowns and fiscal stimulus. We leverage high frequency administrative tax records from Rwanda on firm sales and employment to characterize the impacts of the COVID-19 shock and recovery. We show that the aggregate shock peaked in April 2020, with aggregate turnover and employment recovering to pre-COVID-19 levels by September. The aggregate recovery masks meaningful heterogeneity: while the initial shock impacted sectors in which in-person work was most necessary, the sectors in which face-to-face interactions with consumers are most necessary continue to experience a protracted recovery.