Research

Company Research

Research

Note: A significant portion of my earlier research efforts have shifted into projects at Machine Learning X Doing and Development Economics X under different titles and modules. While individual papers are not fully delineated here, I invite interested collaborators and other parties to also explore the breadth of work under the “Company Research” sections above based on their interests.

Does Raising Police Salaries Lower Petty Corruption? A Policy Experiment on West African Highways (with Jeremy Foltz) (2020). [International Growth Centre Working Paper].  Policy Solutions to Combat Corruption, London School of Economics Connect Magazine, Seeker: A Discovery Digital Network, Seeker Daily, The Economist, Cherokee Gothic, World Bank Africa Can End Poverty Blog, African Development Bank Evaluation Matters

In one of the most ambitious public sector reform experiments in Africa, the Ghana government doubled its police officer salaries in 2010 in part to mitigate petty corruption on its roads, while leaving salaries for other officials unchanged. Neighboring countries in the West African region left their police salaries unchanged. Using unique data on bribes paid from over 2,100 truck trips in West Africa and representing over 45,000 bribe opportunities, we evaluate impacts of higher police salaries on petty corruption using a difference-in-difference method that exploits the exogenous policy experiment. By following bribes paid by the same trucks in different countries as well as to different civil servants in Ghana we identify whether salaries affect the effort to seek bribes, their value and the total amount paid by truckers. Rather than decrease petty corruption, the salary policy significantly increased the police efforts to collect bribes, the value of bribes and the amounts given to truck drivers to policemen in total. Robustness checks show the higher bribe efforts and amounts are stable across alternative specifications.

Computational Ethics (with Edmond Awad, Sydney Levine, Michael Anderson, Susan Leigh Anderson, Vincent Conitzer, M. J. Crockett, Jim A.C. Everett, Theodoros Evgeniou, Alison Gopnik, Julian C. Jamison, Tae Wan Kim, S. Matthew Liao, Michelle N. Meyer, John Mikhail, Jana Schaich Borg, Juliana Schroeder, Walter Sinnott-Armstrong, Marija Slavkovik, and Josh B. Tenenbaum). Trends in Cognitive Sciences, forthcoming.

Econometric Causal Inference for Computer Vision: Image Natural Experiments Inspired by the Economic and Social Sciences (2021). Beyond Fairness: Towards a Just, Equitable, and Accountable Computer Vision. 2021 Conference on Computer Vision and Pattern Recognition (CVPR ’21) Workshop, June 19-25, 2021.

Predicting Petroleum Fields in Ethnic Regions with Social and Economic Data: Evidence from Africa (Poster) (2021). COMPASS ’21: ACM SIGCAS Conference on Computing and Sustainable Societies (COMPASS), Virtual Event, Australia, June 2021.

Deep learning and the development economics of natural resources.

Scaling Up Peer Education with Farmers in India (2017). (with Bhaumik Shah and Tapan S. Parikh). Proceedings of the Ninth IEEE/ACM International Conference on Information Technologies and Communication for Development (ICTD ’17), 15: 1-15. November 16-19. 

Indian farmers are better at predicting other farmers’ listening habits on social media than data-driven message targeting.

Encountering Poverty: Thinking and Acting in an Unequal World (2016). (with Ananya Roy, Genevieve Negrón-Gonzales, and Clare Talwalker). University of California Press.

In Modeling Poverty, development economists use economic models to change the world.

Mobilizing Impact Evaluations: Mobile Survey Micro-Experiments for Sustainable Development (2015). Evaluation Matters, (3rd Quarter), 24-29. in Emerging Solutions to Development Challenges, Volume 1. African Development Bank Group. UC Berkeley News Media Blum Center for Developing Economies, CALIFORNIA Magazine, BORGEN Magazine

Mobile phone surveys can help policy makers in Africa understand their citizens.

Upload:  Race, Financial Misconduct and the Customer-Value Economics of the Firm 

This paper investigates racial bias in bank consumer complaints across the US financial industry and evaluates the first database policy for financial product complaints. The incidence of financial complaints often exhibits a sharp change when white people become the majority group. I find significant selection in how companies responded to the database policy experiment. However, I find a correlation between exposing financial companies with responding to complaints with relief. Comparing credit card complaints with credit reporting complaints provides more convincing causal evidence, although race seems not to be a mechanism. The findings empirically support the customer-value theory of the firm. 

Sports Championships and Farmer Productivity Spillovers in Ghana

One small fishing community in Ghana has produced a world boxing champion for the country every decade since the 1970s. World boxing competitions are widely covered on radio, reaching 90% of the country. In these cases, do sports affect rural agricultural productivity by enhancing sentiments of national patriotism? Using comprehensive microdata on rural agricultural productivity and radio access, I find evidence consistent with the argument that sporting events can enhance workplace productivity. 

Political Economy of Conflict

The Economics of Terrorism and Social Media (forthcoming). In Handbook for the Economics of Terrorism, forthcoming, Eds. Atin Basuchoudhary and Guenther Schulze. Cambridge University Press.

Innocent until Stereotyped Guilty? Terrorism and US Immigration Court Decisions  (2020). (with Justice Tei Mensah). reject-and-resubmit, Journal of Human Capital [OSF Working Paper] Stanford Institute for Theoretical Economics (SITE), 2nd Geospatial Analysis for International Development, CEGA, UC Berkeley 

Domestic terrorist attacks (where the attacker is not a foreigner) caused judges in the US immigration court system to deny asylum applications, suggesting that stereotypes can actually be powerful enough to affect legal decisions even when they are not relevant.

Econometrics and Causal Inference 

Deep Learning and Differences-in-Differences

Deep learning for program evaluations.

Economics for Human-Centered Artificial Intelligence (2019) detailed proposal requested, Nature Machine Intelligence. (in progress)

I argue that to better understand how to overcome the constraints in using AI to augment human capacities, artificial intelligence researchers should look to economics, which has focused on institutional behavior in a rich literature. I focus on the positive implication on three constraints algorithms face: fairness, transparency and ethics.

Economics of Research Transparency

Blockchain and the Scientific Method (2021) (with James A. Evans, Krishna Ratakonda, Kush R. Varshney, and Lav R. Varshney) in Position Papers for the ASCR Workshop on Cybersecurity and Privacy for Scientific Computing Ecosystems, Eds. Stacy Prowell, David Manz, Candace Culhane, Sheikh Ghafoor, Martine Kalke, Kate Keahey, Celeste Matarazzo, Chris Oehmen, Sean Peisert, and Ali Pinar. United States. doi:10.2172/1843573.

Using Blockchain for Scientific Transparency (2017).

US Department of Energy Advanced Scientific Computing Research Workshop on Cybersecurity and Privacy for Scientific Computing Ecosystems, American Academy for the Advancement of Science Annual Meeting Symposium

Trusted timestamping can help make social science research more transparent.

Time-Inconsistency and Research Replications: Overcoming Quasi-Hyperbolic Discounting with Prereplication Plans (2018).  An Urgency for Evidence and Transparency in Economic Analysis and Policy Conference 2017, Association for Integrity and Responsible Leadership in Economics and Associated Professions (AIRLEAP)  Berkeley Initiative for Transparency in the Social Sciences Blog, Development Lunch Seminar, Department of Economics, December 12, 2017, UC Berkeley

Like original authors, replicators may place more weight on transparency in the present than in the future. A replication is a graph of code and results that is time-inconsistent when schadenfreude impulses overturn a published result. A proposed commitment device is a dynamic generalization of a static preanalysis plan.