Tradeweb has launched a new AAA Municipal Yield Curve based on combining Municipal Securities Rulemaking Board data and pre- and post-trade data obtained from institutional market participants trading on the Tradeweb platform.
Tradeweb’s “more timely, granular view” of institutional activity in the municipal bond market allows it to identify trends developing in real-time, often before they show up in other yield curves currently offered by financial market data vendors, according to the firm.
The new dataset also includes the ability to deep-dive into buy/sell ratios and bid wanted flows to get a read on liquidity and demand in the muni market, the firm said.
Tradeweb decided to create a curve that “leverages cutting-edge technology” by using a machine learning algorithm “to help us price the market, as well as leverage the vast amount of trade data that we have,” said Daniel Conde, data and analytics product manager at Tradeweb.
The model examines outstanding municipal securities, both tax-exempts and taxables, and it looks for similarities in bond attributes, such as maturities, call dates, use of proceeds and coupon structures, Conde said.
“The models then tries to figure out where are bonds similar and are they trading similarly,” he said, adding it can tie those various attributes together quickly to effectively price the market.
Tradeweb had been working on its AAA curve for the last few years, starting to develop in house in 2018 after hearing from its clients there was a need for the market for more precise pricing, Conde said.
“We were looking to leverage new advanced machine learning and artificial intelligence techniques and unique data to replicate what humans do,” said Lisa Schirf, managing director and global head of data and analytics at Tradeweb.
Understanding that humans cannot price a million securities within a few minutes of the market close in an accurate way, Tradeweb saw a window of opportunity to use technology to assist in the process.
“We’re able to do it in a fully automated fashion without human biases … based on the actual data, but replicating the process that a human would use,” she said.
Early feedback has been positive.
“When there’s a big move in the market, you don’t have to wait two to three days to actually get to where you should be, and so our curve adjusts in a much more seamless way, rather than having these big jumps,” Schirf said.
Additionally, clients like that Tradeweb offers “full explainability” to users on how it came up with the price, she said.
Tradeweb is also exploring expanding its curve beyond just the AAA GO curve given the positive feedback it has received, Conde said.