Can AI-Driven Techniques Improve Opaque Bond Market Pricing?
New research from SMU Cox explores how AI-based pricing techniques could address long-standing transparency and liquidity challenges in corporate bond markets. The study finds that AI-driven reference prices can meaningfully improve price discovery, even in less liquid markets.
Price discovery is a primary function of financial markets. Equity markets easily facilitate this function with transparent, continuously updated trade and quote data. But in the less-liquid fixed income markets, price discovery often breaks down, leading to stale and less informative pricing. In new research, SMU Cox Finance Professors Kumar Venkataraman, Stacey Jacobsen and David Xu analyze whether AI-driven pricing techniques can address long-standing challenges in the opaque corporate bond market.
“The potential for AI models to advance bond price discovery is significant,” Venkataraman explains. “Intraday AI-based reference prices could be a game changer, providing a faster, more reliable way to assess a bond's fair value and in turn encourage more trading.” For the many securities that trade infrequently, including mortgage-backed securities, municipal bonds, and structured credit products, AI-based reference prices offer a promising path towards greater transparent and efficiency.
AI models and data challenges
The study examines a proprietary AI-based pricing engine called “CP+”, developed by MarketAxess, one of the industry’s leading bond trading platforms. “We're analyzing an algorithm from a major market participant that combines publicly available data with proprietary trading information on their platform,” Jacobsen explains. “If AI models are going to improve price discovery, this is exactly the kind of high-quality, timely, and reliable bond data they need.” The CP+ algorithm applies machine learning techniques to an exceptionally rich set of trading inputs.
Venkataraman illustrates the unique challenges of the bond market. “When securities trade infrequently, which is very common in fixed income markets, the price discovery role of the market isn't really working,” he explains. “It also limits the trade data available to train the AI model.” When a bond does trade, it sends out a price signal that other participants can view to assess fair value. During a six-year study period from 2017 to 2023, 60% of corporate bonds did not trade in institutional (non-retail) size on any given day, and 27% did not trade for an entire week. This infrequency of trading shows the nature of the price discovery obstacle, unlike equity markets.
A further challenge is that institutions trade bonds over-the-counter, and no systematic record of these bilateral negotiations between traders exists. “The informal conversations convey important market color for price discovery,” Venkataraman adds, “but that qualitative data is not available to the AI model.” Given sparse trading and the lack of relevant qualitative data, the study asks whether AI-driven techniques can improve price discovery in the fixed income market.
The study finds that CP+ contributes meaningfully to price discovery. It is more informative about future trade prices than the last trade; CP+ responds systematically to market-wide and bond-specific information, and provides broad coverage across bonds and trading days. Even during periods of stress, such as the onset of COVID, the algorithm tracked efficient prices effectively, but less so during episodes of fire sales. Xu notes, “During fire sales, the trade prices deviate from fundamental values, and since the algorithm is designed to predict the next trade price, it takes longer for CP+ to converge back to fundamentals.”
Why It Matters
The benefits of greater price discovery are substantial. Buyers worry about paying too much; sellers worry about accepting too little. A reliable reference price that signals the bond’s fundamental value helps market participants overcome trading hesitation. Jacobsen notes that better reference prices can unlock substantial gains, including improved liquidity and secondary market trading volume. Xu sees a natural evolution in AI models, with future models offering multiple reference prices—one for the next likely trade and another for fundamental value during stressed markets.
AI-driven models continue to evolve, but they already show promise for enhancing price discovery in less liquid, infrequently traded markets. “Our study shows that alternative mechanisms based on AI models can signal bond value in the absence of timely transaction and quotation data,” Venkataraman concludes.
The paper “Illiquidity Meets Intelligence: AI-Driven Price Discovery in Corporate Bonds,” by Kumar Venkataraman, Stacey Jacobsen and David Xu of Southern Methodist University, Cox School of Business, is under review.
Written by Jennifer Warren.
Video
Can AI Address Opaque Bond Marketing Pricing? A Conversation on Research
See the accompanying vlog on Professor of Finance Kumar Venkataraman's research on AI's ability to address opaque bond market pricing.
