This dissertation studies empirically and theoretically how firms should and do respond to biased consumers. Specifically, I study the magnitude and implications of left-digit bias, the tendency of people to overweight left-most digits, on firms pricing behavior and profits.
In the first chapter, I ask why do so many prices end with 99 cents? Firms arguably price at 99-ending prices because of left-digit bias, the tendency of consumers to perceive a $4.99 as much lower than $5.00. Using retail scanner data on thousands of products and dozens of retailers, I provide reduced-form support for this explanation. I then structurally estimate the magnitude of left-digit bias, and find that consumers respond to a 1-cent increase from a 99-ending price as if it were a 15-25 cent increase. Next, I analyze how firms should respond to left-digit biased demand. I solve and estimate a model that makes three key predictions: (1) prices should bunch at 99-ending prices; (2) there should be ranges of missing prices with low price-endings; (3) these ranges of missing prices should increase with the dollar digit. Qualitatively, these predictions hold. Firms respond to the bias with high shares of 99s and missing low-ending prices. Quantitatively, however, firms price as if the bias were much smaller and demand were more elastic, so they use dominated prices. I estimate that the retailer is forgoing 1-3 percents of potential gross profits due to this misperception.
In the second chapter, I ask to what extent do firms understand their consumers? When consumers are left-digit biased, demand is discontinuous at round numbers making them dominated prices. I present evidence that firms exploit left-digit bias in the long-run. They act as if they know this demand structure, using round numbers only for 1%-2% of posted prices. However, due to a slight reform in the set of admissible prices, supermarket chains in Israel fail to respond optimally - Round numbers post-reform account for 20% of prices, making pricing sub-optimal. After about a year firms re-optimize, and the shares of round numbers drop. Suggestive evidence show that having more spatial competitors is correlated with faster re-learning. Combining the well-understood bias, data, and reform, allows a unique window to look into firms sophistication level. I conclude that firms do respond to biases in the long-run, but do not have a model of these biases in mind and hence fail to respond in the short-run.