UC Berkeley

This dissertation consists of two essays in the fields of Industrial Organization and Behavioral Economics. It studies the issue of consumers' biased beliefs that result in misestimation of personal consumption and excessive spending on the products that will never be used in the future.

Chapter 1 introduces the issue of overspending in the videogames market. The excessive purchasing exhibits itself in the behavior when consumers purchase games and never open them later. According to the descriptive statistics, around 29\% of games in people’s game libraries are unopened as of March 18, 2019. This setting is very beneficial for the analysis in question as besides purchasing decisions one can observe consumption patterns as well, which are usually problematic to track. I gather the main dataset from Steam, one of the biggest online gaming platforms. The data contain daily purchasing and consumption statistics as well as daily prices in all supported currencies for the period of 9 months. In this chapter I study economic and behavioral mechanisms that may be causing such conduct. In particular, I perform survival analysis, discuss search costs, study the effects of game heterogeneity, check for the presence of the projection bias, and consider intertemporal substitution of purchases as possible explanations of the puzzle. I show that none of the above mentioned mechanisms are supported by the dataset and arrive to the conclusion that the main reason for the excessive spending is consumers' biased beliefs regarding their future willingness to try new products.

Chapter 2 builds on the results of the first chapter. In this part of the thesis I develop a structural model that takes into account people's biased beliefs at the point of purchasing new products. It is a static model with the elements of dynamic behavior where every period consumers decide how many games they want to purchase based on how many games they stopped playing in the previous period and then decide how much time to devote to playing. I discuss functional and distributional assumptions that I impose in order to identify the parameters of the model. Next I simulate the model and run three counterfactuals that could potentially alleviate the ovespending behavior. In particular, I consider debiasing consumers, removing sales, and decreasing variety (i.e. imposing a lower bound on game quality). According to the simulated results, removing sales only slightly decreases overspending. Decreasing variety also reduces excessive spending, but it is the most effective for low userscore thresholds. One has to determine the optimal threshold that minimizes the effect of the bias. Debiasing consumers is the most effective in the reduction of expenditures, and the total effect depends on the initial magnitude of the bias.