Pricing and Systematic Trading of Municipal Bonds
Written by Petter Kolm & Sudar Purushothaman
In this article, the authors propose a systematic approach for pricing and trading municipal bonds, leveraging the feature-rich information available at the individual bond level. Based on the proposed pricing framework, they estimate several models using ridge regression and Kalman filtering. In their empirical work, they show that the models compare favorably in pricing accuracy to those available in the literature.
Additionally, the models are able to quickly adapt to changing market conditions. Incorporating the pricing models into relative value trading strategies, the authors demonstrate that the resulting portfolios generate significant excess returns and positive alpha relative to the Vanguard Long-Term Tax-Exempt Fund (VWLTX), one of the largest mutual funds in the municipal space.
Over the past two decades, numerous inroads have been made in systematic pricing and algorithmic trading in fixed income markets. In contrast to equities, most fixed income markets, such as corporate bonds, mortgages, and municipal bonds, are less liquid and associated with poorer data quality.
This is especially the case for issues that are traded over-the-counter (OTC),since without a centralized exchange the price discovery process becomes lesstransparent.
The municipal bond market presents a unique challenge for systematicpricing and algorithmic trading. Much of the research in this field hasfocused on predicting or explaining the value of the tax-exemption (Livingston, 1982; Mankiw and Poterba, 1996; Kim, Lee, Lile, and Ramsey, 2000; Kalotay, 2018), understanding the impact of liquidity on market pricing (Harris and Piwowar, 2006; Biais and Green, 2019), or yield curve analysis (Kalotay and Dorigan, 2008; Dash, Kajiji, and Vonella, 2018).
Due to a lack of active trading and uniform market structure, the municipal bondmarket is significantly less liquid than corporate and treasury bond markets. Being primarily a buy and hold market, a large number of municipal bonds do not trade, or trade infrequently; often, the bulk of trading activity occurs in the weeks immediately after it is issued (Wu, Bagley, and Vieira, 2018).
Bessembinder, Spatt, and Venkataraman (2020) find that as of 2016, households constitute 51% of municipal bond ownership, compared to just 6% for corporate bonds.
The fragmentation of the municipal bond market is a daunting challenge for electronic trading platforms (Cestau, Hollifield, Li, and Schürhoff, 2019), which explains the slow adoption of algorithmic and systematic trading in this market. Therefore, it is notsurprising that to date most research on systematic and factor-driven approaches in fixed income has focused on corporate bonds (Bektić, Wenzler, Wegener, Schiereck, and Spielmann, 2019; Israelov, 2019; Henke, Kaufmann, Messow, and Fang-Klingler, 2020; Guo, Lin, Wu, and Zhou, 2020; Heckel, Amghar, Haik, Laplenie, and Carvalho, 2020), sovereign and treasury bonds (Brooks, Richardson, and Xu, 2020), or both (Cochrane and Piazzesi, 2005; Ludvigson and Ng, 2009; Brooks, Palhares, and Richardson, 2018; Bektić, Hachenberg, and Schiereck, 2020).
The relatively low liquidity, limited volatility and default risk have contributed to the moderate interest the municipal bond market has received amongst institutional investors and academic research communities.
Needless to say, the fragmentation provides opportunities to leverage modern statistical and machine learning approaches, as there are many features (state of issuance, sector, rating, coupon, call date, to name a few) across which the market can be characterized, analyzed and modeled.