James Sams

Ph.D., Quantitative Marketing
Graduate School of Business
Stanford University

CV (pdf)
Email: jsams@stanford.edu
Office Address: Stanford University
Graduate School of Business
655 Knight Way, E150
Stanford, CA 94305-7298

I completed my Ph.D. in quantitative marketing at Stanford University's Graduate School of Business June 2019.

My research integrates modern machine learning and economic theory with standard econometric models to explore questions related to consumer learning, innovation, socially responsible marketing, and policy and technology.

Before joining the Ph.D. program at Stanford, I worked at the Kilts Center for Marketing at the Booth School of Business, cleaning, building, and documenting the Kilts-Nielsen data warehouse for the now widely-used consumer and retail scanner panel data. Before that, I worked as a software consultant, developing databases and integration and aggregation tools for small businesses. My undergraduate degree is from the University of Chicago with concentrations in Political Science and Economics.

I joined Facebook July 2019.


Job Market Paper

Learning or Herding? Understanding Social Interactions and the Distribution of Success on a Social Music Sharing Platform

Digital sharing platforms like YouTube and SoundCloud crowdsource the process by which users can discover high quality new products among an increasingly vast flow of new products, acting as on-going digital test markets. Social features on these platforms can accelerate the discovery process by encouraging sharing of information and facilitating learning, thereby reducing the number of people sampling poor quality products. This may more quickly concentrate platform traffic on higher quality alternatives. Social features may also include a feedback loop if people care about consuming the same products as their peers. Given previous research showing that social feedback loops can distort or even invert the relationship between product quality and product popularity, if such feedback loops exist, the discovery and filtering capabilities of crowdsourcing may be compromised, emphasizing the need to understand the nature of social interactions on such platforms. Utilizing data from SoundCloud, a music sharing and streaming site, I develop an approach to separately identify and measure these two separate endogenous social effects with and without feedback loops. Results suggest that the platform's social features do have informative effects but that the feedback loop plays a dominant role for the most successful songs.


Consumption Experiences and the Production of New Ideas: Evidence from Artists' Behavior on SoundClound (with Harikesh Nair, Navdeep Sahni, and Florian Stahl)

The production of new ideas is a fundamental component of modern society. Economists have recognized that new ideas are a key element of economic growth; businesspeople seek them out for competitive advantages; researchers try to push forward the barrier of knowledge; and artists attempt to develop deeper understandings of the human condition. But where these ideas come from remains a mystery. This paper seeks to understand the influence of artists' consumption experience on their production of new ideas. Using a metric of song similarity learned from revealed preference data, combined with data on artists' consumption and production, we explore whether there is a plausible causal connection between novel consumption experiences and artists' creation of new music.

Does Increased Grocery Access Affect the Nutritional Composition of Grocery Purchases: Heterogeneous Treatment Effects and Food Deserts

A substantial literature has posited that one cause of the high rates of obesity among low-income people is the lack of access to reasonably priced nutritional food, with such neighborhoods frequently termed "food deserts". A related but alternative hypothesis is that a high density of readily available junk food through e.g. bodegas and convenience stores, creates exceptionally low barriers to casual, calorie-dense snacking, termed "food swamps". I use an event study framework utilizing the movement of people to new neighborhoods and the opening and closing of grocery stores to identify these effects as a series of small-scale natural experiments. I then utilize recent developments in the measurement of heterogeneous treatment effects to better identify which groups may be most responsive to the treatment.


  • BPR.jl: A set of functions implementing Bayesian Personalized Ranking and hyperparameter tuning, with several, high performance, modular data structures that vary the trade-off between memory and computational intensivity.
  • CenteredSparseMatrix.jl: A sparse matrix implementation that re-centers data without compromising the sparsity structure of your data, especially useful if losing sparsity compromises your ability to compute due to memory usage. Useful with e.g. PCA where we usually want to re-center our data.


This browser does not support PDFs. Please download the PDF to view it: Download PDF