We introduce a novel application of machine learning to compare pooling and servicing agreements (PSAs) that govern commercial mortgage-backed securities. In contrast to the view that the PSA is largely boilerplate text, we document substantial variation across PSAs, both within- and across-underwriters and over time. A part of this variation is driven by differences in loan collateral across deals. Additionally, we find that differences in PSAs are correlated with ex post loan and bond performance. Collectively, our analysis suggests the importance of examining the entire governing document, rather than specific components, when analyzing complex financial securities

We extend the Fama–MacBeth regression framework for cross-sectional return prediction to incorporate big data and machine learning. Our extension involves a three-step procedure for generating return forecasts based on Fama–MacBeth regressions with regularization and predictor selection as well as forecast combination and encompassing. As a by-product, it provides estimates of characteristic payoffs. We also develop three performance measures for assessing cross-sectional return forecasts, including a generalization of the popular time-series out-of-sample R2 statistic to the cross section. Applying our extension to over 200 firm characteristics, our cross-sectional return forecasts significantly improve out-of-sample predictive accuracy and provide substantial economic value to investors. Overall, our results suggest that a relatively large number of characteristics matter for determining cross-sectional expected returns. Our new method is straightforward to implement and interpret, and it performs well in our application

We show theoretically that the usual estimated investment strategies will not achieve the optimal Sharpe ratio when the dimensionality is high relative to sample size, and the $ 1/N $ rule is optimal in a 1-factor model with diversifiable risks as dimensionality increases, which explains why it is difficult to beat the $ 1/N $ rule in practice. We also explore conditions under which it can be beaten, and find that we can outperform it by combining it with the estimated rules when $ N $ is small, and by combining it with anomalies or machine learning portfolios, conditional on the profitability of the latter, when $ N $ is large

Big data on job postings reveal multiple facets of the impact of COVID-19 on corporate hiring. Firms disproportionately cut new hiring for high-skill positions, with financially constrained firms reducing skilled hiring the most. Applying machine learning methods to job-ad texts, we find that firms have skewed their hiring toward operationally-core functions. New positions display greater flexibility regarding schedules and tasks. While job posting levels show signs of recovery starting in late-2020, changes to job descriptions and skill profiles persist through early-2022. Financial constraints amplify these changes, with constrained firms’ new hires witnessing greater adjustments to job roles and employment arrangements

Much of the extant literature predicts market returns with “simple” models that use only a few parameters. Contrary to conventional wisdom, we theoretically prove that simple models severely understate return predictability compared to “complex” models in which the number of parameters exceeds the number of observations. We empirically document the virtue of complexity in U.S. equity market return prediction. Our findings establish the rationale for modeling expected returns through machine learning

This study examines the investment value of information provided by crypto-influencers, that is, social media influencers covering crypto assets on Twitter. We examine the returns associated with approximately 36,000 tweets issued by 180 of the most prominent crypto social media influencers covering over 1,600 crypto assets for the two years spanning through December 2022. Our primary results indicate that crypto-influencers’ tweets are initially associated with positive returns. However, these tweets are followed by significant negative longer-horizon returns, suggesting they generate minimal long-term investment value. These effects are most pronounced for tweets issued by crypto-influencers proclaiming to be crypto experts, for smaller cap crypto asset securities and for self-described experts with many Twitter followers. In an additional analysis, we use machine-learning methods to classify tweets and find that this pattern of results strengthens when the tweets have a more positive sentiment or relate to buy recommendations

We use variational inference (VI), a technique from the machine learning literature, to estimate a mortality-based Bayesian model of nursing home quality accounting for selection. We demonstrate how one can use VI to quickly and flexibly estimate a high-dimensional economic model with large datasets. Using our facility quality estimates, we examine the correlates of quality and find that public report cards have near-zero correlation. We then show that in contrast to prior literature, higher quality nursing homes fared better during the pandemic: a one standard deviation increase in quality corresponds to 2.5% fewer Covid-19 cases

We investigate whether improvements in banks' organizational capital and control systems facilitate increased loan origination to minority borrowers. We focus on bank supervisors' enforcement decisions and orders (EDOs) against banks and hypothesize that EDO‐imposed improvements in loan policies, internal governance, and employee training mitigate deficiencies in credit assessments and lending decisions that previously disadvantaged minority borrowers. We find that mortgage origination to minority borrowers increases following the resolution of EDOs, and more so for banks with stricter supervisors or more severe EDOs. Using a semisupervised machine learning method to analyze the text of EDOs, we find that such increases are higher for EDOs specifying revisions of loan policies, implementation of formal internal governance procedures, or more employee training. Overall, we find that EDO‐driven improvements in organizational capital generate unintended, positive social externalities that enhance access to credit for minority borrowers

This empirical study investigates whether a strong bank culture may help strengthen, weaken, or have no effect on the relationship between regulatory capital and liquidity creation. Using a machine learning approach and banks’ 10-K reports, we first measure the corporate culture of selected bank holding companies (BHCs) in the United State (U.S.) over the period between 1995 and 2019. We find that bank culture does affect the link between regulatory capital and liquidity creation. In particular, while we find that regulatory capital has a negative impact on bank liquidity creation, a strong culture in a bank weakens this negative association. We also find that an increase in asset-side liquidity creation is the main channel through which bank culture exerts its moderating role. Finally, our results are largely driven by smaller banks, banks with a more traditional funding structure and more profitable banks. The results of this study suggest that regulators should consider bank culture as being a crucial element in the monitoring approach when designing bank regulation and supervision

This paper compares three methods for assessing the contagion of risk among ten Globally Significant International Banks, known as GSIBs, listed on the New York Stock Exchange with daily and weekly data sets from 2007 to 2020, based on Machine Learning and Network Analysis. In particular we identify the banks which are the largest net sources or transmitters of risk, and net receptors of risk. We also examine the response of regulatory actions, in the form of fines and BIS Bin Classification for capital adequacy. Under alternative risk measures, of Range Volatility (RV) of share prices, Credit Default Swap (CDS) premia, and Conditional Value at Risk (ΔCoVar), there is a stronger and significant connection between Contagion and the BIS Bin classifications relative to the connections between Contagion and banking fines, either in the amount or frequency of the fines. These results show that BIS bin classifications respond positively to underlying signals of increased contagion in the form of Range Volatility (RV) and ΔCoVar measures but not to CDS risk premia