Essays in financial economics

Abstract

The three chapters in this thesis study agent behavior, whether it be that of retail investors (chapter 1), the reporting behavior of home-buyers for the purpose of taxation (chapter 2), or that of firms choosing what data to collect on their consumers (chapter 3). All three chapters employ data-driven approaches and novel methods to explore heterogeneity within their respective fields of study and draw conclusions that have implications for policymakers, market participants, and other stakeholders. In the first chapter, I study the heterogeneity of retail investors based on the stocks they chose to discuss. I use Reddit WallStreetBets data totalling more than 1.5 million posts and 35 million comments from 2018 to the end of 2021. I find retail investors to be diverse, with distinct subgroups showing interest in stocks with attention-grabbing features, or fundamental-based investing. I also find those attributes to be connected to echo chamber effects, likely causing confirmation bias, finding that subgroups do not mix. Finally, I find subgroups to be of differential importance in a predictability exercise. Effects commonly attributed to retail investors, such as increased volatility, might be coming from a sub-sample of them but also a sub-sample of stocks. In the second chapter, we develop a new method to estimate under-reporting and employ it on large, granular administrative data from the Mumbai real-estate market. The approach compares bunching of reported values around government-assessed guidance values with the distribution of a third-party measure of market values. We estimate that 11 percent of Mumbai real-estate value is under-reported between 2013 and 2022. We detect a decline in bunching post-demonetization, but no change in the overall under-reporting rate. Properties with mortgages from banks with high nonperforming assets and from public-sector banks exhibit greater under-reporting. 5 In the third chapter, we scrape a comprehensive set of US firms’ privacy policies and study these policies alongside firms’ web data extraction behaviour. We find considerable and systematic variation in privacy policies along multiple dimensions including ease of access, length, readability, and clarity, both within and between industries. Surprisingly, firms’ data extraction, measured by the extent of cookies they place on consumers browsing their websites, is strongly and positively related to the length and complexity of their policies. Firms with intermediate levels of technical sophistication have longer, more complex policies than firms with very high or low technical sophistication, controlling for a range of other firm attributes. A simple signalling model of firms engaging in data extraction in an economy with both myopic and sophisticated consumers, where privacy is a “shrouded attribute” of firm-consumer interactions, helps to rationalize these findings.