We conduct a randomized experiment involving 7,063 late-paying clients of a large Colombian bank to compare the effects of text messages that leverage different behavioral motives on loan delinquency. Our results show that receiving a message decreases the likelihood of borrowers being late by 4 percent. The effects are more pronounced and persistent when messages leverage social norms. Using machine learning tools, we find that the effects are higher among borrowers with higher credit scores and unsecured loans. A second experiment shows that this type of message is ineffective in preventing on-time borrowers from falling into loan delinquency

This article proposes an innovative method to assess borrowers’ creditworthiness in consumer credit markets by conducting machine-learning-based analyses on real-time video information that records borrowers’ behavior during the loan application process. We find that the extent of borrowers’ micro-facial expressions of happiness is negatively associated with loan delinquency likelihood, while the degree of fear expressions is positively associated with delinquency risk. These results are consistent with two economic channels relating to the adequacy and uncertainty of borrowers’ future income, drawn from the extant psychology and economics literature. Our study provides important practical implications for fintech lenders and policymakers

I ask whether machine learning (ML) algorithms improve the efficiency in lending without compromising on equity in a credit environment where soft information dominates. I obtain loan application-level data from an Indian bank. To overcome the problem of the selective labels, I exploit the incentive-driven within officer difference in leniency within a calendar month. I find that the ML algorithm can lend 60% more at loan officers’ delinquency rate or achieve a 33% lower delinquency rate at loan officers’ approval rate. The efficiency is maintained even when the algorithm is explicitly prevented from discriminating against disadvantaged social classes

Using transaction data from a U.S. account aggregator, I study how household balance-sheet conditions drive within-person variation in marginal propensities to consume, repay debt, and invest across three rounds of pandemic stimulus. Using a machine learning imputation estimator, I measure the sensitivity of responses to time-varying financial circumstances. Spending responses fall as liquid assets increase, while debt repayments crowd out consumption only for those with binding borrowing limits. Transfers also encourage retail investment in stocks and cryptocurrencies at both intensive and extensive margins. The findings show how liquidity constraints and debt overhang influence the allocation of transfers across spending, deleveraging, and financial assets

Using new data from US and global markets, we revisit market risk premium predictability by equity anomalies. We apply a repertoire of machine-learning methods to forty-two countries to reach a simple conclusion: anomalies, as such, cannot predict aggregate market returns. Any ostensible evidence from the USA lacks external validity in two ways: it cannot be extended internationally and does not hold for alternative anomaly sets—regardless of the selection and design of factor strategies. The predictability—if any—originates from a handful of specific anomalies and depends heavily on seemingly minor methodological choices. Overall, our results challenge the view that anomalies as a group contain helpful information for forecasting market risk premia

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