My research focuses on building more realistic economic models of markets and crowds through measuring information from textual sources, capturing the role of relationships in social and financial networks, and incorporating new innovations and policies. Although I am trained as a Ph.D. economist, my work draws on methods from natural language processing, network theory, and machine learning combined with econometrics and economic theory.
“Editorial Decisions and Information Contents of Reviews” [pdf]
I measure information on distinct facets of quality from a corpus of reviews and characterize how decision-makers integrate this information present in text with that available through other channels. Specifically, I demonstrate that referee comments at a scholarly journal contain information on submissions’ future citation impact above and beyond information available in referee scores. I measure this signal on future citation impact and show that it does not enter into editorial decision-making directly but rather through an interaction that amplifies the information content of referee scores: the more citations a low- or mediocre-scoring paper is likely to get the less likely it is to be published. Secondly, I describe referee comments that are highly predictive of greater citations. Papers that referees say have access to unique datasets, or are written on topics of relevance to ongoing debates or government applications receive greater citations on average. Third, I show the appearance of favoritism amongst editors who accept a higher share of papers that cite themselves is partly a reflection of an ability to draw and select for papers that receive more citations. Finally, I characterize budget constraints on publication space and referee capital and provide some guidance on what types of information editorial systems could capture to promote transparency in future analyses while protecting privacy of authors or referees.
“Strategic Interaction and Network Formation in Human Insurgencies” [Request Draft]
I study the behavior of individuals in a network setting wherein link formation as well as investment into a collective good are endogenous and costly. I show that in this framework, equilibrium networks are characterized by multi-tiered stars with the number of tiers determined by the population size and cost of maintaining links. When individual payoffs are strictly increasing in interactions between both group effort and individual participation, stable equilbria for individual investment exist; however, when these payoffs are weakly negative in individual participation, multiple equilibria may exist but they will not be stable. I explain how this model is related to human insurgency and may provide a theoretical underpinning for empirical findings of political scientists about the sizes of war and human insurgency over the last century. The findings suggest that managing media, winning “hearts and minds” or other methods for shifting individual payoffs on participation may be key to disrupting insurgency. Furthermore, shifting the cost structure of network links can affect the number and magnitude of insurgent behavior. Extensions to research collaboration networks are also discussed.
“Mortgage Markets with Lender Competition and Asymmetric Information” [pdf]
This analysis introduces a theoretical framework for assessing the empirical discussion of asymmetric information amongst mortgage lenders and adds in the idea of lender competition. The results are generally consistent with existing empirical findings that diversified lenders will act as uninformed investors, making poorer loans, selling most of their mortgages and making less profit off each mortgage while concentrated lenders do the opposite. When lenders face potential competition, however, they are able to use the bidding process as a means of generating additional information. Competition drives down profits, making it more difficult for diversified lenders to operate but also improves the information they have access to, both of which address to some extent the role of uninformed lending others have characterized in the run up to the recent mortgage market crash. If lenders can resell loans while obscuring ex post learning (either directly or via securitization), then some lenders may find an incentive to deliberately pose naive offers resulting in reduced loan quality across the market.
“Industry Structure and Its Determinants” [Request Draft]
Using textual analysis of business descriptions from public filings of over 24,000 companies in the CRSP/COMPUSTAT merged database, this work develops a network representation of relationship strength between firms. The study findings together with past work on textual determinants of firm competitors make a strong case for the networked nature and complex structure of relationships between companies rather than distinct markets or even isolated multi-sided market relationships typically studied in economic literature. This complexity of relationships between firms and markets raises questions about the nature of concentration and the role of a firm’s environment in determining positive/negative spillovers it experiences, incentives for innovation and other firm behavior, as well the lifecycle and dynamics of firms and markets.
Here, I extend past work on industry structure by decomposing the economic factors that contribute to the strength of textual overlap. By estimating an econometric model of co-movement amongst various firm factors, I identify and decompose textual factors associated with different types of firm relationships including co-movement of growth in R&D, sales, advertising, and labor force. I show that these factors are independent of each other and significant for understanding market structure and predicting innovation outcomes. Finally, I develop a theoretical framework for studying why such multi-sided relationships between firms emerge and evolve.
“How do Navy Officers Respond to Changes in Pay & Commitments: A Dynamic Retention Model for Evaluating Changes in the Critical Skills Retention Bonus” [Request Draft]
This paper offers a dynamic retention model of the Navy Surface Warfare Officer population. I estimate officers’ annualized decisions to stay in the military, leave, or make additional commitments in exchange for a Critical Skills Retention Bonus using a thirty-year time horizon, random-effects expected utility model incorporating a taste distribution for military service that is fitted to data on actual stay and leave rates. Results show significant expected variation between retention of officers with different levels of prior enlisted service, which points to the potential for meeting labor demand with prior-enlisted officers at mid-career levels, and show significant potential cost-savings may be possible by increasing bonuses associated with additional commitments.