SEC’s $5M Bloomberg BVAL fine targets ‘dark magic’ in fixed income pricing

This week, lawyers for providers of evaluated pricing services are likely burning the midnight oil, double- and triple-checking that their disclosures and marketing materials align exactly with the methodologies—in principle and in practice—that they employ to value fixed income securities. This follows last week’s announcement that the Securities and Exchange Commission (SEC) had settled charges against Bloomberg for violating its disclosed methodology about how it prices fixed income securities on its evaluated pricing service, BVAL.

The data giant paid a $5 million fine, made additional disclosures to its valuations methodologies, and agreed to refrain from future violations. For customers—some of whom are subject to new SEC rules requiring them to ensure pricing sources are anchored in reality—the settlement has raised broader questions about the reliability of evaluated pricing services, which use a combination of trade data, market quotes and algorithms to price illiquid assets. 

According to the SEC’s order, from at least 2016 through October 2022, Bloomberg omitted in disclosures to BVAL customers that the valuations for thinly traded, illiquid securities could be based on a single data input, such as a broker quote, which did not adhere to methodologies it had previously disclosed. The affected securities constituted a “very small fraction” of BVAL’s total reported valuations, which cover 2.5 million securities daily. 

Notably, the SEC did not find that the violations led to any pricing errors not reflective of the market, but sources say the punishment—which they view as little more than a slap on the wrist for Bloomberg—fit the crime.

“They did something that they just didn’t list as possible. It’s not good, but not horrible,” says a fixed income portfolio manager at a US hedge fund. “Fixed income pricing on less-liquid items can be dark magic sometimes, no matter how hard we all try to systematize. [I’m] not sure that excuses anyone, though, from strictly following their own published methodology.”

In its statement announcing the settlement, Osman Nawaz, chief of the SEC’s Division of Enforcement’s Complex Financial Instruments Unit, said the matter underscores that the regulatory authority will hold service providers accountable for misrepresentations that impact investors. But it also suggests that further regulation is on the horizon for the fixed income market, which has been “electronifying” for decades but remains plagued by manual paper-and-phone processes and patchy data.

“Bloomberg wants to have as much instrument coverage as possible—and charges a ton for it—so a breach in their methodology is worth going after,” says a chief technology officer at a second US hedge fund. But sources interviewed for this story also describe a larger market structure problem, to which Bloomberg, and others, have fallen prey.

It’s not good, but not horrible. Fixed income pricing on less-liquid items can be dark magic sometimes, no matter how hard we all try to systematize

A fixed income portfolio manager at a US hedge fund

In 2020, the SEC fined Intercontinental Exchange’s Ice Data Pricing and Reference Data (PRD) business $8 million for delivering prices to clients based on single-broker quotes. However, in a notable contrast with the BVAL investigation, the SEC in that case found that Ice had also failed to adopt and implement policies and procedures designed to address the risk that these prices would not reasonably reflect the value of the securities.

The commission’s order regarding Ice also found that the company’s quality controls for prices based on single-broker quotes were not effectively or consistently implemented, and that these failures impaired Ice Data PRD’s ability to assess the reliability of quotes it received from market participants and determine whether a quote provider was an accurate source of information.

A spokesperson for Ice declined to comment.

The act of using single-broker quotes—while generally considered a “no-no” following the 2008 credit crisis—is not the punishable offense in either case. Rather, the disparity between stated and actual methodologies is what drew the SEC’s attention. But single-sourced data carries the potential to harm investors when evaluated pricing services, such as BVAL and PRD, are used to determine net asset value calculations or ensure best execution. But the act of using single-broker quotes is itself the less-than-perfect answer to a thus-far unsolvable problem: what do you do when you don’t have enough data?

How BVAL works

BVAL receives data from a variety of sources, including contracted data suppliers; publicly-available trade reporting sources such as MSRB, Trace and Mifid data; market quotes from global and regional banks; broker-dealers and exchanges; and executable levels from Bloomberg’s electronic trading platforms and services. (It should be noted that the action was for Bloomberg’s BVAL product and does not include BVAL OTC, which is a separate product despite the similar name.)

All prices are generated using a weighted combination of two proprietary algorithms: direct observables, which may consist of trades, executable levels, or indicative quotes for “target bonds,” which must be corroborated by at least one other executable level or an indicative quote obtained from a different dealer; and indirect observables, which is data regarding bonds that are comparable to the target and which is used in the absence of sufficient observables.

A former Bloomberg employee says they were shocked by the SEC fine because using singular data inputs was never the intention of BVAL.

“The approach starts off as market-driven and transitions to model-driven as market data gets scarce. … One of the value props of BVAL was historically the transparency with which the service illustrates how the price was derived,” they say.

Specifically, on the Bloomberg Professional Terminal, a user can initiate a command—BVAL<GO>—and see how many trades have been utilized, how old they are, and whether they were one-sided trades. From there, Bloomberg performs a regression analysis to identify the yield, which allows the user to “back into” the price.

Each priced instrument also receives a BVAL score, which is designed to gauge the amount and consistency of market data used to arrive at a generated price. The BVAL score is measured on a scale of 1 (the lowest) to 10 (the highest) and changes according to the market data available for a particular instrument. It, however, is not a price confidence score.

“If a single quote was indeed used, then the question begs, was this disclosed to the client—which historically it has been—and were the BVAL evaluators [experts who monitor BVAL’s algorithmically generated scores] challenged on it?” the former employee says.

Missing the mark(et)

A source at an agency broker says that evaluated pricing sources in the US have come under more scrutiny since the introduction of SEC Rule 2a-5 in September 2022, which requires fund managers to demonstrate greater control over the pricing sources they use for daily valuations and to ensure those pricing sources are anchored to real price observations.

“The issue for the likes of Bloomberg and Ice is that their models depend on indicative contributions from banks, brokers, and others. These are typically not ‘observable’ price points, and in many cases, they have legacy agreements with their contributors that do not permit them to disclose that they are using the prices—observable or not,” the broker says.

Relying on counterparties to consistently and reliably provide information, such as quote data, can be a risky part of business, though there often isn’t another option.

Catherine Shalen, director of research at Cboe Global Markets from 2002 to 2018, recalls a plan from several years ago to create a swap rate volatility index. The plan was eventually scrapped after Cboe was only able to secure two swap dealers to contribute their data, instead of the five or six minimum needed to make the index credible. To this day, Shalen still isn’t sure why dealers declined to contribute to the project, the point of which was to provide better transparency into the OTC derivatives market, which she considered a noble pursuit. But she faced the same problem as Bloomberg—sometimes, there just isn’t enough data.

Users of evaluated pricing services typically perform analyses against both the output and the methodology, and many firms will use only one independent price evaluation source, or at least one per market, given that these services come with a hefty price tag. As a result, users tend to have a heavy dependency on their chosen pricing products and often expend significant effort evaluating feeds for accuracy, methodology and controls.

This dependency—and potential for investor harm—can trickle downstream. Relationships between providers and users would have to be signed off by risk functions, and likely front offices. Once reviewed, the firms likely make certain disclosures to investors and clients about how they use the price feeds. Data from evaluated pricing providers is also consumed by benchmark operators for bond and portfolio evaluation, and BVAL data primarily underlies the Bloomberg Fixed Income Indices (formerly the Bloomberg Barclays Fixed Income Indices).

Though the SEC found no pricing irregularities in the case of BVAL’s violation, all evaluated pricing providers—including Ice’s PRD and offerings from S&P Global/IHS Markit, and Refinitiv—must at times contend with a fundamental inadequacy in the data available, namely that if important calculations are performed using incomplete data, and if users aren’t aware of how incomplete it is, it would be all too easy to manipulate a price, the agency broker says.

“Say a trader has an aged position on a bond with no observable data and hence does not have a BVAL score. If that trader sends an indicative quote that is picked up by BVAL and there are no [other data points], that bond will be priced at the level of that quote,” the broker says. “Now, people may not believe the price, but what about the automated processes that take in the BVAL feeds via API? Does that bond get priced in an end-of-day NAV? All bonds without a BVAL price probably go to some manual process to be priced, but anything with a BVAL may not be flagged.”

Bloomberg declined to comment for this story.

Not long after the SEC announced its settled charges, two SEC commissioners, Hester Peirce and Mark Uyeda, issued their own joint statement, dissenting to bringing the charges in the first place. The pair criticized the SEC order, stating that it was “notably vague” about the frequency of BVAL’s use of single-broker quotes for thinly traded, illiquid securities.

According to the order, Bloomberg violated Section 17(a)(2) of the Securities Act of 1933, which prohibits the use of materially false statements to obtain money or property in the offer or sale of securities. The dissenting commissioners wrote that they believed Bloomberg’s statements to customers of its pricing services in no way resemble how the Supreme Court defined “misrepresentations,” or materially false statements, when the court explained that “in the offer or sale” encompasses the entire selling process.

In other words, Bloomberg’s omissions were not material enough to have influenced a customer’s decision to buy the product.

“It could be that, in light of the importance of their services, the Commission should consider whether and how providers of pricing services should be subject to oversight. … A one-off enforcement action that rests on a strained reading of Securities Act Section 17(a)(2) is not the right way to make regulatory policy,” they wrote.

While it’s yet unclear as to whether SEC chairman Gary Gensler will impose further regulation on pricing service providers during his tenure (the SEC did not respond to a request for comment) it seems that those in his crosshairs (and apparently in his chambers) are anticipating—and should be preparing for—greater scrutiny of the sector.

With additional reporting by Anthony Malakian