Weekly Research Rundown – July 26, 2019

Weekly Research Rundown – July 26, 2019

Who Owns U.S. CLO Securities?

As more loans are purchased by collateralized loan obligations (CLOs), the U.S. leveraged loan market has recently grown by a substantial amount. This post breaks down CLO investors by location and investor type, providing information essential to a better understanding of investors’ exposure to the leveraged loan market. The authors find that U.S. CLOs are generally held by U.S. insurance companies, mutual funds, and depository institutions.

Read More: https://www.federalreserve.gov/econres/notes/feds-notes/who-owns-us-clo-securities-20190719.htm


Did Local Factors Contribute to the Decline in Bank Branches?

While the total number of bank branches increased from the mid-1990s until the financial crisis, that number has been on the decline since. This paper finds a strong correlation between local conditions like population and employment and the aggregate number of branches in a county. The authors find that, post-crisis, the influence of local demographic and economic factors on branch openings and closings has weakened while local market competition has played a greater role.

Read More: https://www.kansascityfed.org/publications/research/er/articles/2019/3q19senguptadice-did-local-factors-contribute-decline


Forecasting High-Risk Composite CAMELS Ratings

This paper finds that statistical learning models can improve bank supervisors’ ability to forecast commercial banks’ CAMELS ratings, and thereby off-site monitoring ability. Such models, more often used outside of economics, may generate better early-warning signs of commercial bank risk.  Moreover, the combination of forecasts from various models improves the accuracy of the forecasts slightly.

Read More: https://www.federalreserve.gov/econres/ifdp/files/ifdp1252.pdf


Machine Learning Implications for Banking Regulation

As the amount and variety of financial data continues to grow, banks are likely to rely increasingly on machine learning models to make decisions about risk and portfolio management, customer experience, and fraud detection. This allows for improvements in bank decision making as machine learning makes prediction cheaper and more accurate. However, these models are more opaque than traditional statistical models, creating new challenges for regulators including an update to the regulatory guidance on model risk management.

Read More: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3423413&download=yes