Robots Can Price Your Bonds for You


(Bloomberg) -- Bond pricing.

Let’s say you’re the chief financial officer of a company with some bonds outstanding, and you want to issue some new bonds, and you are trying to figure out what interest rate you’ll have to pay on the new bonds. How can you do that? The traditional way to do it is to open up your office door, where you will find a pack of investment bankers waiting to talk to you. Each of them will have a pitchbook, and in each pitchbook there will be a page that says “Indicative new-issue pricing for XYZ Corp.” (if they’re careful they will have replaced “XYZ Corp.” with the name of your company), and you can just look at that pricing and have a decent idea of what new bonds will cost you. If you turn the page, you will probably see some supporting data—charts of the yields on your existing bonds, new-issue concessions for other recent issuers in your sector with similar credit ratings, etc.—which serves both to justify the indicative pricing and to show you how much work the banks have done for you and how grateful you should be to them.

The alternative way to do it—which is somewhat common at big sophisticated frequent bond issuers—is to just go find that supporting data, look at it, and draw your own conclusions. If you have a lot of bonds outstanding and they trade a lot, then the trading levels of those bonds will give you a pretty good indication of what interest rate you’d pay on a new bond. If you have long and recent experience of issuing new bonds, and of observing your competitors who issue bonds, you’ll probably have some feel for how much extra you’d pay over those trading levels to do a new bond. It is not exactly rocket science, and if you’ve been doing this long enough, and paying enough attention, you may have as good a feel for it as your bankers do. And so you will rely less on bankers. I mean, you’ll still hire bankers to do your bond deal, but your relationship with them will be more commoditized. You won’t feel any gratitude to them for making pretty charts of where your bonds trade. You know where your bonds trade.

But most companies aren’t big sophisticated frequent bond issuers, and most CFOs have better things to do than pay attention to bond trading levels and new-issue concessions, and it is perfectly sensible for them to rely on the expertise of their investment bankers. The investment bankers spend all day thinking about bonds; of course they know more about bonds than the issuers do. They are getting paid for that expertise, and that expertise consists not merely of raw knowledge—the trading levels of comparable bonds, the database of recent deals—but also of a certain Fingerspitzengefühl, an intuitive sense of the market that comes from long experience collecting and interpreting that knowledge.

Some of that stuff just seems real, real easy to teach a robot though:

Overbond, a financial-technology startup in Toronto, wants to change all that. ...

Its main offering is a set of machine-learning algorithms powered by neural networks, a type of artificial intelligence, that predict the timing and pricing of new bond issues. The service is already fully in place for the Canadian corporate-bond market, and partly so for the American one. The algorithms crunch through credit ratings and real-time data on secondary trading for a firm and its peers, among other things. Recent predictions for the yield on new bond issues have been, on average, off by less than 0.02 percentage points.

A subscription buys tailored estimates of demand for new bonds, including the interest rate the market is willing to bear. This helps corporate treasurers gauge market conditions and decide when to issue bonds and in what maturity. Of the 200 or so Canadian corporations that issue debt frequently, 81 are signed up.

I am sure that the neural networks are very smart and deep, but you could write a dumb simple algorithm (take the spreads on the issuer’s liquid bonds closest in maturity to the proposed new bond, interpolate, add that spread to the Treasury rate with the right maturity, average the new-issue concessions on the last few deals with similar credit ratings, add that to your rate, stop) that gets you much of the way there. The question here is not “how can I write a symphony that will achieve immortality” or “how can we achieve the greatest good for the greatest number of people” or “how can I win the heart of my beloved?” It is “here are the prices of a bunch of bonds, what should the price of a similar bond be?” A computer really should be able to figure that out.

Again, this is not all that debt issuers rely on banks for. Even once you know what your bond should cost, you still gotta hire a bank to actually underwrite it. And when you do, you’re not just paying the bank for its expertise and market knowledge. Mostly you’re paying it for its contact list—it has to call up investors and sell them the bonds for you—as well as repaying it for doing previous undercompensated work (or making previous undercompensated loans) to you. You’re also paying for all sorts of other bank-y expertise and skills and processes—writing the bond offering documents, collecting money from buyers and wiring it to you, etc.—that are hard to get anywhere else.

Still, debt capital markets bankers might find this all just a little bit disturbing. The contact lists could be automated too. Whatever investment banks are actually selling, a big part of what they think they’re selling—what they hold themselves out as selling, what they sell themselves as selling—is financial expertise. When companies want to interact with financial markets, they go to investment banks, because the investment bankers know the financial markets. Some of that knowledge of financial markets is subtle and intuitive and multifaceted and hard to reduce to an algorithm, or even a neural network. But a lot of it … probably … isn’t? If computers can know the market better than the bankers do, the case for the bankers is undercut.

Deutsche Bank.

One problem with Deutsche Bank’s constant strategy changes, as it tries to figure out if it will fight to remain an ambitious global full-service bank or (the current plan) retrench to be “more Deutsche, less bank,” is that you get a lot of this:

The U.S. division, for example, will probably end up firing some of the dealmakers it lured with multimillion-dollar compensation packages in recent years, according to executives, who, like others inside Deutsche Bank, spoke on the condition of anonymity to avoid jeopardizing their careers.

Already last week, the bank told staff it would close its Houston office. Not even four years ago, it made a big investment in that business, poaching a team of oil and gas bankers from Citigroup Inc. One senior executive emphasized that the bank still views its presence in New York, San Francisco and Chicago as indispensable for its corporate finance units.

Oops! On the one hand, you must feel a bit silly if you were recently poached by Deutsche Bank into a business that it is now shutting down. On the other hand, presumably those “multimillion-dollar compensation packages” were meant, in part, to pay for just this uncertainty. If you were poached by Deutsche Bank during a previous strategy, presumably you knew that things were unstable and charged it a premium for that instability. You sold Deutsche the option to get rid of you, and now that option is paying out, but I hope you at least charged for it.

“People close to the firm said job cuts may be expanded to affect about 20 percent of the U.S. workforce,” which seems like a lot in a basically benign environment for banks. “Employees who remain should feel encouraged by the overhaul, which in many areas will focus on underperformers, said Mark Fedorcik, co-head of the investment bank’s operations in the Americas, in an interview on Thursday.” I am not sure! Surviving layoffs in a core business is encouraging: If you are confident that the bank needs to keep someone around, and they keep you around while firing other people, then that’s a good sign that you’re indispensable. Surviving layoffs in a disappearing business just prolongs the agony: If you’re worried that the bank won’t keep anyone around, you might as well get it over with; the people who go first often get the best severance packages and the easiest time getting jobs elsewhere. And this is Deutsche Bank’s “fourth global turnaround plan in three years,” so I don’t know why you’d take anything as a particularly long-term signal.

Elsewhere in personnel moves, here is a story about how Bank of America Corp. employees in London are not excited to move to Dublin, “after observing the trajectories of BofA staff who moved to Chester.” Still it could be worse: 

Not all BofA are disparaging about Dublin. One said he’d keep an open mind. Another suggested that a move to the Irish capital would at least be better than a move to Charlotte, North Carolina, where BofA’s head office is based.

The crypto. 

I read occasionally about how blockchain technology will allow companies to “tokenize” their shares, but I never quite understand why they would want to, beyond the general notion that everything is better on the blockchain. Here’s one story like that: “The blockchain wing of Morgan Creek Capital Management, an investment advisor, announced Wednesday that it is working with Anexio Technology Services to convert all of its physical shares into an ERC-20 token.”

"By tokenizing the business, we took the cap table, the paper shares - every company in the world has paper share certificates - and we tokenized it. We swapped, one for one, the paper shares for tokens, which allow for a global investor base and trades on a security exchange."

Oh come on. Yes, finally, unlike trading in Apple Inc. stock, where you have to laboriously transport around paper stock certificates, which limits trading to people within a few blocks of each other in Lower Manhattan, these blockchain people have finally found a way to allow for electronic global trading on a securities exchange. We talk sometimes around here about how cryptocurrencies keep rediscovering the basic ideas in finance, but I guess you can make that look more impressive by just pretending that no one else has discovered them. 

Elsewhere: “A lot of people were looking for ways to spend their cryptocurrencies,” is a sentence in this article, and you will need a stronger stomach than I have if you want to find out what they came up with. And: “This Cryptocurrency Billionaire Is Trying to Split California Into Three States.” I guess that’s an appropriately silly way to spend cryptocurrencies, though I would like to think that if I was a Bitcoin zillionaire I would try to split April into three months or put Rhode Island on the moon or increase the atomic mass of cobalt or something. 

And: “Nvidia Corp. revealed for the first time how much revenue it generated from chip sales to the cryptocurrency market—and said that figure will be much smaller in the second quarter.” And: “What’s the Deal With Facebook and the Blockchain?” I don’t know, but good lord, I hope it is anything but this:

At least one cryptoblogger speculated that Facebook might launch its own tokens and distribute them to shareholders and users. Users whose posts hit some kind of engagement metric would earn tokens. If users held FaceBucks, or whatever they're called, they’d have a personal financial stake in the platform’s success. 

Bond marking.

Here is a(n alleged) scam that I would not have thought of. Actually let’s start with a more basic obvious scam. Let’s say you are a hedge fund and you have some weird bonds worth $50. You get paid a management fee that is calculated as a percentage of your assets under management; the more the bonds are worth, the higher that fee is. You also get paid a performance fee that is calculated as a percentage of your performance; the more the bonds are worth, the better your performance is. Also of course the better your performance, the more likely your investors will be to stick with you, and the higher your fees will be. All in all you’d rather have the bonds be worth $60 than $50. 

But the bonds are worth $50. But they don’t trade very much. The way you know they’re worth $50 is basically because you ask some dealers and the dealers say “I’d bid $48 for them and offer them at $52,” and you take the midpoint and call it $50. But you could ask different dealers. You could for instance ask dealers who do a lot of business with you and who owe you a favor. They could just say “I’d bid $58 for them and offer them at $62.” If you then say “okay fine sold to you for $58,” then the dealer would be pretty annoyed, but if the dealer was confident that you were just asking for the bid to fake your books, and not to trade, then maybe he’d do you a favor. I mean it is not advisable, for him, or for you, but it clearly happens from time to time.

Anyway that is all a fairly standard scam. But let’s say, instead of being worth $50, your bonds are worth, like, $5. For instance they might be weird bits of mortgage-backed securities that don’t trade anywhere near 100 cents on the dollar. Then you might go out to dealers and get quotes and come back with quotes of, say, $4.75 bid at $5.25 offered. That 50-cent spread is 10 percent of fair value, which is pretty wide, but remember these are creepy weird bonds. 

Here’s the scam I wouldn’t have thought of. Instead of just taking the midpoint of the bid and ask and saying that the bonds are worth $5, you could take just the bid, and add a “normal” bid-ask spread to it to get an “imputed” ask, and then figure out an “imputed” midpoint between the two, and use that as your fair value. For instance, if the average mortgage-backed security has a bid-ask spread of, say, $3, then you could assume that your bonds are $4.75 bid at $7.75 offered, for a midpoint of $6.25. 

Is this legitimate? Hahaha no, not really, but it is surprising. Like, you just wouldn’t expect that to be a scam, or to work. Here’s how the Securities and Exchange Commission describes it, in its lawsuit against an investment adviser called Premium Point Investments LP:

Unbeknownst to at least most investors, Premium Point itself routinely derived, or “imputed,” the mid-point prices of securities, even where Premium Point could easily obtain a midpoint price for the security from a broker. To do so, Premium Point took a bid price for a particular security and added half the spread between the bid and ask prices on a broad sector of securities— not the spread on that particular security—to “impute” a mid-point price for that security.

When you just read it like that it sounds conservative. You would expect weird illiquid securities to have wider bid-ask spreads than “a broad sector of securities.” If the broad sector has a $3 spread, and your bond is weird and rarely trades, you might expect it to have a $6 spread, say $47 at $53. Just taking the bid ($47) and adding half the “normal” spread ($1.50) to it gets you a fair value ($48.50) that is lower than the actual fair value.

The trick is that if you do this with very low-dollar-price bonds you get absurd results. Also you can combine it with the first trick (inflating the bid) to get even more absurd results. The SEC gives an example of one bond where a pricing service provided a $2.63 midpoint price, Premium Point allegedly got a friendly brokerage to provide a bid of $4, and then it allegedly applied an average bid-ask spread of $3.68 for the bond’s sector—a bid-ask spread wider than the price of the bond—to get a $5.84 “imputed mid-point price.”

I don’t know, it’s just an impressive bit of (alleged) nonsense. The rest of the complaint is not terrible either:

For example, in July 2015, Majidi told Shor: “[W]e need to show performance…. [W]e’re like the bottom third of hedge funds.”

Similarly, in October 2015, Majidi told Shor: “I mean, if we show down 3% this month [in one of our portfolios], game over…. [The investor] would do a complete redemption.”


In one conversation with Majidi in approximately January 2016, Shor said: “I am overmarked by a bunch of millions… I am at 3 [dollar spread] on 6 dollar bonds…”

Things happen.

Leveraged Loan Safety Net May Prove to Be Frayed for Investors. Bond investors do not care if Argentina is solvent in 100 years. Record Buybacks Help Steady Wobbly Market. In the Long Run, Fear of Short-Termism Is Mostly Bunk. JP Morgan revamps its stock trading business in a sign of the changes sweeping through Wall Street. JPMorgan applies to re-enter China securities market. Goldman Sachs, Apple Team Up on New Credit Card. Wells Fargo Says It Will Take More Time Before It Can Grow. Barclays CEO Fined About $870,000 on Whistle-Blower Incident. Eight of 20 Best-Paid CEOs of 2017 Are No Longer on the Job. U.S. Raises Pressure on Iran With Sanctions on Currency Exchange. “Private equity is of two minds about the work that it does. It’s not quite sure if it wants to believe its own self-justification or if ambient societal criticism has some merit.” “The next generation of burger lovers is looking for a more food-forward, cosmopolitan approach.” Hitler-loving stalker sent guy 65K texts after one date: cops. Man Allegedly Used Change Of Address Form To Move UPS Headquarters To His Apartment

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Thanks! 

To contact the author of this story: Matt Levine at

©2018 Bloomberg L.P.

BQ Install

Bloomberg Quint

Add BloombergQuint App to Home screen.