Christina M. Qiu

[Draft] Endogenous Transfer Networks Under Spatial Risk (Job Market Paper)

This paper studies how digital payment technologies affect the amount of insurance households can attain under spatially-correlated shocks when risk-sharing links are endogenously formed. I develop a model with three key features of transfer partnerships: source substitutability, strategic complementarities, and a motive for spatial diversification. Households jointly agree to strengthen their partnerships but can unilaterally weaken them, leading to multiple stable network configurations. In the model, the partnerships of a low-productivity household show fragility: when negative shocks raise the household's expected needs for transfers, their ability to attract profitable sources weakens as more-productive partners scale back engagement. I use the model to quantify welfare gains from transfer networks in Tanzania after the 2008 introduction of mobile money technologies, which lowered the costs of transporting money across space. By 2028, the diffusion of mobile money generates average welfare gains of 0.8%, about two-thirds of which arise from network reorganization in response to the technology. Welfare gains are progressive, driven by an enhanced redistributive role of transfers but diminished insurance benefits. Reduced transfer frictions can induce network changes that undermine informal insurance in equilibrium—offering a new explanation for the persistent vulnerability of low-income households despite technological improvements in financial access.

[Draft] AI and Scale: A Quantitative Task-Based Theory of Automation

with Danial Lashkari, Wensu Li, and Neil Thompson

This paper develops a quantitative task-based theory of automation that allows a direct mapping to empirical measures of AI exposure and AI automation costs. The theory accounts for the increasing returns to scale property of AI, where fixed training costs often outweigh variable inference costs. Firms’ automation decisions across tasks are shaped not only by the relative productivity of machines to labor, but also by patterns of scale advantage, i.e., a task’s output to fixed cost ratio. We derive task, firm, and economy-wide elasticities of substitution between labor demand and capital prices under firms’ endogenous adoption of AI with fixed costs. We directly estimate the training and inference costs of AI using robust empirical relationships between AI performance and its computational inputs. To do so, we employ a worker survey to uncover information on relevant task characteristics, including complexity, i.e., a task’s number of potential outcomes, and required accuracy, i.e. the minimum proportion of “successful” attempts a worker can make to be considered qualified to conduct a task. We calibrate our baseline model to AI adoption rates published by the Census Bureau in 2023. Quantitatively, labor becomes less substitutable with AI when capital prices decline, as remaining workers are highly productive and become increasingly difficult to replace.

[Paper] Regionalized liquidity: A cross-country analysis of mobile money deployment and inflation in developing economies

World Development, 2022

Mobile money has been hailed as a serious innovation in the pursuit of financial inclusion and poverty alleviation; however, its macroeconomic effect is not fully understood. This study presents a regional theory of inflation and argues that limited market integration contributes to mobile money's inflationary effects. Household survey data from Kenya confirms increased use of mobile money after village- and supra-village-level shocks due to risk-sharing between liquidity-flexible and liquidity-constrained regions. A difference-in-differences empirical assessment indicates that mobile money deployment increases national consumer price indices. Findings support that the power to distribute equals the power to generate money supply in developig countries.