We are leveraging provider, product, and price data on marijuana, beer, and wine markets in the United States to explore the relationship between regulation, market structure, consumer preferences, and price in these markets. We use natural language processing, unsupervised and supervised machine learning, and computational photography and machine vision to explore a number of phenomena, including how consumers and producers understand and evaluate products in murky state-legal marijuana markets and how legitimacy and risk affect price in contested markets. We will compare these findings in the marijuana market to data that we have collected from the well-established wine and beer markets to examine how regulatory uncertainty, changing cultural norms, and the presence of field expertise affect consumer preferences and price.

We have scraped data from online forums used by bitcoin, forex, and commodity traders. We use natural language processing and machine learning to examine how exogenous shocks, that is, unexpected political, cultural, and social changes outside of the market, and endogenous shifts, that is, changes occurring within the market, affect traders' strategies. We further investigate whether trader characteristics, such as age, gender, and nationality; commodity characteristics, such as legality, legitimacy, and cultural meaning of the instrument being traded; or trader philosophies play a larger role in explaining dynamics within these exchange markets.