CAIDA Talks

1st Talk by Rajesh Vijayaraghavan

Managerial Decisions and Machine Predictions

Abstract: A large body of research in finance and accounting examines managerial decision-making. The managers in this research include executives, such as CEO of firms, or other parties that interact with firms, such as equity analysts, and auditors. These research focus on causal inference, but as argued in Kleinberg at al. 2015, there are many applications, including with policy relevance, where causal inference is not the key. This talk will discuss how machine learning can be used to improve and understand managerial decision-making.

I focus on manager’s predictions of future loan losses of banks.  These accounting decisions, referred to as estimating loan loss allowances, are governed by accounting rules set by standard setters, such as the Financial Accounting Standard Board (FASB). In June 2016 the FASB issued a new rule, effective in December 2019, that will replace current accounting rules (current GAAP) with a model that allows banks to use broader information to estimate loan loss allowances. To empirically examine current GAAP and the new model, I exploit differences in the information sets allowed under the old and the new rules. Using a methodology that combines micro data and machine learning techniques, I provide evidence that it is possible to construct a loan loss recognition model that outperforms the current GAAP without expanding the information set beyond that permitted under the current rule. I find that expanding this model’s information set does not significantly improve its performance. My model’s predicted allowances would have been materially larger at the outset of the recent financial crisis than actual reported bank estimates. The differences are due to that my model consistently assigns larger weights to certain input variables relative to current GAAP. My results provide a novel method to examine aspects of the new accounting rule before it comes into effect. The interpretation is that data-driven machine learning techniques can improve the role of allowance as a buffer for future loan losses.

Biography: Rajesh Vijayaraghavan is an Assistant Professor of Accounting at the UBC Sauder Business School. He received his PhD in Accounting at Harvard Business School, Boston. His research focuses on accounting and risk management in banks and insurance companies, corporate finance, and applying machine learning to social science research. Prior to his PhD, he worked number of years in Wall Street. Outside of academia, Rajesh enjoys cooking, visiting coffee shops, and playing percussion music.

2^nd Talk by Gene Moo Lee

Opportunity Structures: A Machine Learning Approach for Analyzing Industry Dynamics

Abstract

Researchers often model an industry as a network where each node corresponds to an organization and an edge represents an inter-organizational relationship (e.g., competition, acquisition, alliance). Structural holes are an important construct in identifying network opportunity structures. While there have been significant theoretical and empirical works around this concept, there has been limited fine-grained empirical research on the operationalization of the structural hole concept based on organizational self-identified strategic posturing. In this project, we propose an innovative method to quantify self-identified strategic posturing structural holes using a machine learning approach called doc2vec, which transforms textual documents into numeric vector representations. Specifically, we apply the doc2vec model to the collection of 10-K annual reports from U.S. public firms in the 1995-2016 period. To show the effectiveness of our measure, we conducted empirical analyses on firm birth (i.e., IPO) and firm mortality (i.e., delisting) using Compustat data. First, our firm birth analysis, using the generalized linear model, shows that new organizations have an increasing birth rate in structural holes between a pair of existing firms. Second, using the Cox proportional hazard model, we show that organizations entering into a structural hole have a significant decrease in mortality rates. This is the first large-scale empirical study to use self-identified strategic posturing structural holes in the analysis of industry dynamics, and as such provides an advance to both the industry dynamics and network literatures.

Biography

Gene Moo Lee is an Assistant Professor of Information Systems at UBC Sauder School of Business. He received PhD in Computer Science from UT Austin. His research on business analytics has appeared in journals such as MIS Quarterly, Journal of MIS, Journal of Cybersecurity, and Journal of Business Ethics as well as CS conferences such as IMC, EC, and INFOCOM. He held industry positions at Samsung, AT&T, Intel, and Goldman Sachs. He holds 10 patents in mobile technology.