arXiv subjects

Daniel Björkegren

Publications and source records attributed to Daniel Björkegren.

9 recordsLinked to original sources

arXiv2026

Market Beliefs about Open vs. Closed AI

Market expectations about AI's economic impact may influence interest rates. Previous work has shown that US bond yields decline around the release of a sample of mostly proprietary AI models (Andrews and Farboodi 2025). I extend this analysis to include also open weight AI models that can be freely used and modified. I find long-term bond yields shift in opposite directions following the introduction of open versus closed models. Patterns are similar for treasuries, corporate bonds, and TIPS. The different movements suggest that that markets may anticipate open and closed AI advances to have different economic implications, and that the cumulative impact of AI releases on bond yields may be more muted.

econ.GN↗

arXiv2025

Could AI Leapfrog the Web? Evidence from Teachers in Sierra Leone

Only 37% of sub-Saharan Africans use the internet, and those who do seldom rely on traditional web search. A major reason is that bandwidth is scarce and costly. We study whether an AI-powered WhatsApp chatbot can bridge this gap by analyzing 40,350 queries submitted by 529 Sierra Leonean teachers over 17 months. Each month, more teachers relied on AI than web search for teaching assistance. We compare the AI responses to the top results from google.com.sl, which mostly returns web pages formatted for foreign users: just 2% of pages originate in-country. Also, each web page consumes 3,107 times more bandwidth than an AI response on average. As a result, querying AI through WhatsApp is 98% less expensive than loading a web page, even including AI compute costs. In blinded evaluations, an independent sample of teachers rate AI responses as more relevant, helpful, and correct answers to queries than web search results. These findings suggest that AI can provide cost-effective access to information in low-connectivity environments.

cs.CY↗

arXiv2024

Are LLMs Useful in the Poorest Schools? TheTeacher.AI in Sierra Leone

Education systems in developing countries have few resources to serve large, poor populations. How might generative AI integrate into classrooms? This paper introduces an AI chatbot designed to assist teachers in Sierra Leone with professional development to improve their instruction. We describe initial findings from early implementation across 122 schools and 193 teachers, and analyze its use with qualitative observations and by analyzing queries. Teachers use the system for lesson planning, classroom management, and subject matter. Usage is sustained over the school year, and a subset of teachers use the system more regularly. We draw conclusions from these findings about how generative AI systems can be integrated into school systems in low income countries.

cs.CY↗

arXiv2022

(Machine) Learning What Policies Value

When a policy prioritizes one person over another, is it because they benefit more, or because they are preferred? This paper develops a method to uncover the values consistent with observed allocation decisions. We use machine learning methods to estimate how much each individual benefits from an intervention, and then reconcile its allocation with (i) the welfare weights assigned to different people; (ii) heterogeneous treatment effects of the intervention; and (iii) weights on different outcomes. We demonstrate this approach by analyzing Mexico's PROGRESA anti-poverty program. The analysis reveals that while the program prioritized certain subgroups -- such as indigenous households -- the fact that those groups benefited more implies that they were in fact assigned a lower welfare weight. The PROGRESA case illustrates how the method makes it possible to audit existing policies, and to design future policies that better align with values.

econ.GN↗

arXiv2022

Instant Loans Can Lift Subjective Well-Being: A Randomized Evaluation of Digital Credit in Nigeria

Digital loans have exploded in popularity across low and middle income countries, providing short term, high interest credit via mobile phones. This paper reports the results of a randomized evaluation of a digital loan product in Nigeria. Being randomly approved for digital credit (irrespective of credit score) substantially increases subjective well-being after an average of three months. For those who are approved, being randomly offered larger loans has an insignificant effect. Neither treatment significantly impacts other measures of welfare. We rule out large short-term impacts either positive or negative: on income and expenditures, resilience, and women's economic empowerment.

econ.GN↗

arXiv2020

Balancing Competing Objectives with Noisy Data: Score-Based Classifiers for Welfare-Aware Machine Learning

While real-world decisions involve many competing objectives, algorithmic decisions are often evaluated with a single objective function. In this paper, we study algorithmic policies which explicitly trade off between a private objective (such as profit) and a public objective (such as social welfare). We analyze a natural class of policies which trace an empirical Pareto frontier based on learned scores, and focus on how such decisions can be made in noisy or data-limited regimes. Our theoretical results characterize the optimal strategies in this class, bound the Pareto errors due to inaccuracies in the scores, and show an equivalence between optimal strategies and a rich class of fairness-constrained profit-maximizing policies. We then present empirical results in two different contexts -- online content recommendation and sustainable abalone fisheries -- to underscore the applicability of our approach to a wide range of practical decisions. Taken together, these results shed light on inherent trade-offs in using machine learning for decisions that impact social welfare.

cs.LG↗

arXiv2020

Manipulation-Proof Machine Learning

An increasing number of decisions are guided by machine learning algorithms. In many settings, from consumer credit to criminal justice, those decisions are made by applying an estimator to data on an individual's observed behavior. But when consequential decisions are encoded in rules, individuals may strategically alter their behavior to achieve desired outcomes. This paper develops a new class of estimator that is stable under manipulation, even when the decision rule is fully transparent. We explicitly model the costs of manipulating different behaviors, and identify decision rules that are stable in equilibrium. Through a large field experiment in Kenya, we show that decision rules estimated with our strategy-robust method outperform those based on standard supervised learning approaches.

econ.TH↗

arXiv2020

The Effect of Network Adoption Subsidies: Evidence from Digital Traces in Rwanda

Governments spend billions of dollars subsidizing the adoption of different goods. However, it is difficult to gauge whether those goods are resold, or are valued by their ultimate recipients. This project studies a program to subsidize the adoption of mobile phones in one of the poorest countries in the world. Rwanda subsidized the equivalent of 8% of the stock of mobile phones for select rural areas. We analyze the program using 5.3 billion transaction records from the dominant mobile phone network. Transaction records reveal where and how much subsidized handsets were ultimately used, and indicators of resale. Some subsidized handsets drifted from the rural areas where they were allocated to urban centers, but the subsidized handsets were used as much as handsets purchased at retail prices, suggesting they were valued. Recipients are similar to those who paid for phones, but are highly connected to each other. We then simulate welfare effects using a network demand system that accounts for how each person's adoption affects the rest of the network. Spillovers are substantial: 73-76% of the operator revenue generated by the subsidy comes from nonrecipients. We compare the enacted subsidy program to counterfactual targeting based on different network heuristics.

econ.GN↗

arXiv2019

Behavior Revealed in Mobile Phone Usage Predicts Loan Repayment

Many households in developing countries lack formal financial histories, making it difficult for firms to extend credit, and for potential borrowers to receive it. However, many of these households have mobile phones, which generate rich data about behavior. This article shows that behavioral signatures in mobile phone data predict default, using call records matched to repayment outcomes for credit extended by a South American telecom. On a sample of individuals with (thin) financial histories, our method actually outperforms models using credit bureau information, both within time and when tested on a different time period. But our method also attains similar performance on those without financial histories, who cannot be scored using traditional methods. Individuals in the highest quintile of risk by our measure are 2.8 times more likely to default than those in the lowest quintile. The method forms the basis for new forms of credit that reach the unbanked.

cs.CY↗