Welcome back to TPI’s Research Roundup, our semi-regular compilation of recent outside research of interest to tech policy nerds. Papers this month include studies on internet adoption, machine learning, Uber Eats, e-commerce, and cryptocurrencies. If you’ve read a paper you think might be interesting to include in the next roundup, feel free to send it to email@example.com
DISCLAIMER: The papers and authors are not affiliated with TPI. We do not necessarily agree with everything, or even anything, in these papers, but find them interesting and informative.
Predicting Cell Phone Adoption Metrics Using Satellite Imagery by Edward J. Oughton and Jatin Mathur
What it is: A method of predicting cell phone and telecom adoption using machine learning and computer image processing.
What they find: A neural network is better at capturing variance in cell phone adoption in Malawi and Ethiopia than commonly used predictors like population density. However, the neural network the authors use requires training different models for each country. Combined with the long amount of computer time required to train one version of the model, more work is needed for this approach to be practical for other countries.
Why it matters: Accurate measures of communications usage and adoption data are important for measuring and addressing the digital divide, and often ideal data for doing so is not available.
When are Google data useful to nowcast GDP? by Laurent Ferrara and Anna Simoni
What it is: A method for using Google Trends to predict current GDP.
What they find: Google search trends data can improve the accuracy of GDP nowcasts, especially at the beginning of quarters when official data is not available.
Why it matters: More frequent data and better predictive accuracy allows for better real-time decision making.
COVID-19 and Digital Resilience: Evidence From Uber Eats by Manav Raj, Arun Sundararajan, and Calum You
What it is: A study of the effects of the pandemic on restaurants using Uber Eats data.
What they find: The number of restaurants newly offering Uber Eats delivery more than doubled in March compared to January or February across five major U.S. cities. The total number of orders placed on the platform increased significantly even though many restaurants were open fewer hours or closed completely. The increase in online orders for restaurants using the platform is due to the increase in demand and reduction in supply. The competitive effect of other restaurants offering the same cuisine on Uber Eats increased relative to prior to the pandemic—a ten percent increase in open restaurants would cause a 6.35% decline in orders post-pandemic, but a 5.14% before. The study does not estimate the net benefits to restaurants because Uber provided data only on the number of deliveries, not revenues or order details.
Why it matters: As delivery services become more important in a socially-distancing world, understanding their economics becomes more important.
Gains from Convenience and the Value of E-Commerce by Yugeng Huang and Bart J. Bronnenberg
What it is: A study of consumer surplus gains from e-commerce based on the ability to shop from home rather than traveling to stores using Dutch individual-level retail data and store openings and closings.
What they find: 27% of the benefits of e-commerce, representing 5.9% of consumer spending, are specifically related to convenience.
Why it matters: Understanding why consumers shop online allows for better e-commerce policy.
The Microeconomics of Cryptocurrencies by Manna Halaburda, Guillaume Haeringer, Joshua S. Gans, and Neil Gandal
What it is: A review of microeconomic models of cryptocurrencies.
What they find: The authors discuss the models that build the microeconomic foundations for understanding cryptocurrencies, including game theory models of proof-of-work, proof-of-stake, and longest chain attacks and bitcoin adoption and competition between Bitcoin and other cryptocurrencies. They use insights from this review to model Bitcoin’s price changes.
Why it matters: The microeconomic foundations of cryptocurrencies are important, especially for non-currency blockchain projects. Formalizing these models also helps researchers to figure out how to improve on proof of work/proof of stake.