Content area
Abstract
Municipal bonds make up one of the largest fixed income markets in the United States. At over $3.7 trillion of outstanding debt obligations, it is a major source of funding for states, local governments, their agencies, and even some corporations. However, portfolio managers struggle to accurately value municipal bonds due to the time gap between bond issuance and agency rating guidance. Most municipal bonds are issued in late January while the earliest available rating reports from credential agencies like Moody’s are usually released in June. Since credit quality as measured by bond ratings has a decisive influence over bond pricing, errors in estimating bond ratings potentially leads to either overlooking quality bonds or overpayment for bonds.
The current credit valuation of municipal bonds when no agency ratings are available is mostly dependent on the portfolio manager’s subjective estimation. Past research on bond rating prediction has mainly focused on corporate bonds. Moreover, the credit rating agency’s rating framework and analytic scheme for the municipal bonds is quite opaque, making it difficult for investors to formulate a deterministic credit rating prediction model. In this paper, we construct several supervised machine learning multiclassification models as well as traditional statistical models to predict the agency’s municipal credit rating and propose a new systematic credit evaluation pipeline to boost the investment quality and efficiency.
