Research
My research draws on a broad range of methods — the analysis of large-scale datasets using structural, reduced-form causal, and descriptive econometrics, machine learning, and historical and qualitative analysis, to advance theory and understanding. Much of this work draws on data assembled from many organizations and platforms. What sets a large share of it apart is that the work requires running large-scale field experiments with live organizations and platforms — and, often, building the very systems I study. In collaboration with industry partners, scientific organizations, and grant-making bodies — and, where needed, by assembling my own engineering and data-science teams — I design and deploy real digital infrastructure, prototype platforms, and scale their adoption in the field, then run explicit experiments that test, measure, and optimize these systems while they operate. The aim is two-sided — to advance fundamental knowledge and to improve the systems themselves — pairing hands-on engineering and the scientific management of live operations with economic, business, and quantitative analysis. Across London Business School, Harvard, the NBER, and Northeastern, I have hired more than fifty research assistants in computer science, data science, engineering, and economics — many now at Amazon, Google, Microsoft, Netflix, and Figma, and several in faculty positions of their own.
Data and platforms behind this work span Palm Computing, Handango, the Apple App Store, the Apple jailbreak community, Unknown Worlds Entertainment (a games studio), GE, a university experiential-education platform and organization (Northeastern’s engineering co-op), TopCoder, the NASA Tournament Lab, Harvard Medical School and Harvard Catalyst, and a university strategic grants office — alongside platforms built and run in-house, including the IoT Open Innovation Lab, NUgig, the Cyber-Physical Marketing Challenge, and Digital Scientific Twins.
The same questions, approached with different tools. This table shows which method each empirical paper actually uses — a paper often uses more than one.
On the econometrics column. These papers are tagged with a single “econometrics” label, so causal and descriptive work are not yet separated here. That split has to be made paper by paper and has not been done.
Practitioner articles and handbook chapters. They draw on the empirical work above rather than carrying a primary method of their own, so they are listed rather than tabulated.