2023 FALL SEMINAR SERIES Seminar: Justin Marlowe University of Chicago

Smart Summaries: Generative Artificial Intelligence and Municipal Bond Investing

Generative artificial intelligence tools can transform how investors acquire and process complex financial information. This seems especially promising for the municipal bond market, a space known for fragmentation, opaque pricing and poor liquidity. To explore this potential, I ask ChatGPT to summarize the narrative sections of audited financial statements and bond offering documents for the largest US cities, counties and school districts. It produces summaries that convey the same sentiment as the original narratives but are 80-90% shorter and written in much simpler language. I show these summaries are more effective than the original narratives at explaining secondary market price/yield dynamics for bonds from the jurisdictions in question. I then extend these findings to compute a measure of information "bloat" and show that longer-than-necessary disclosures associate with measures of municipal bond investor uncertainty. These results suggest that generative artificial intelligence can add considerable value to the municipal market, and that the narrative component of government financial reporting has much room for improvement.