Free mug with every unsolicited rating.

Twitter Is Junk. Why Are You Surprised?

Matt Levine is a Bloomberg View columnist. He was an editor of Dealbreaker, an investment banker at Goldman Sachs, a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz and a clerk for the U.S. Court of Appeals for the Third Circuit.
Read More.
a | A

Standard & Poor's gave Twitter an unsolicited BB- credit rating yesterday, and I sort of wonder if that was a stunt cooked up by its social media team. Why give any company an unsolicited rating, really? The defining feature of an unsolicited rating is, no one asked you. S&P's statement explains the reasoning behind the rating, but not the reasoning behind doing it in the first place, so I will just blithely assume was to increase its follower count on Twitter.

My favorite reason to do an unsolicited rating is to punish the issuer for not paying you to do a solicited one. That sometimes happens, or at least, ratings agencies sometimes accuse each other of doing that. But it doesn't seem to be what's going on here. Twitter has only convertible debt outstanding, which is not normally rated, so it didn't exactly snub S&P. It's not like it paid anyone else to rate those bonds.

Also that BB- doesn't seem particularly punitive? It seems sort of fine? Fine to generous? The thing about ratings is that they are typically a lagging indicator: The market usually knows a company's creditworthiness before the ratings agencies do. So it's a bit strange that Twitter's stock was down 5.9 percent yesterday on the back of its surprise junk rating. Did people think that Twitter was an investment-grade company?

I mean, maybe. Its financials are ... I guess more Internet-grade than either investment-grade or high-yield. It's got $1.8 billion of debt (two convertible bonds it issued last month), $3.6 billion of cash, negative net income and around $200 million of EBITDA. On present income it looks pretty junky, though S&P is optimistic about "healthy growth in monthly active users and revenues, the possibility of positive discretionary cash flow in 2016, and ongoing minimal debt leverage."

But of course Twitter issued debt two months ago. And, without the benefit of S&P's wisdom, the market went and figured out Twitter's creditworthiness. And you can back out the market's thoughts from how that debt priced two months ago.

That is a bit confounded by the fact that Twitter's debt came in the form of convertible bonds, but we can deal with that. One way to deal with it is to use a convertible bond pricing model to back out the credit component, though that requires some inputs. (And a convertible bond pricing model. ) The main inputs are the trading prices of Twitter's convertible and its stock, both of which you can more or less figure out, and the expected volatility of Twitter's stock over the next five or seven years, which is harder. That makes pricing a convertible more of an art than a science, and I have somewhat lost the knack over the last three years. But my best guess would be a credit spread of about 300 basis points, meaning that if Twitter issued non-convertible bonds today they'd come at a rate of a bit under 5 percent for five years, or a bit under 5.5 percent for seven.

You don't have to trust my guess, though. You can trust Twitter's guess! A convertible bond is a combination of a bond (pays interest, pays back principle at maturity) and a call option on Twitter's stock. For accounting purposes, Twitter was required to split up the convertible by separately valuing the bond and the call option, using some basic bond math. Twitter issued a 5-year bond with a 0.25 percent interest rate, and a 7-year with a 1 percent interest rate. Those are not the "right" interest rates for Twitter bonds. Apple just did an 8-year bond with a 1 percent interest rate, and Twitter is no Apple. So Twitter went and applied its "right" interest rate to the bonds, to figure out what they should be worth. And that "right" interest rate is right there in the 10-Q from a week ago:

In accordance with accounting guidance on embedded conversion features, the Company valued and bifurcated the conversion option associated with the 2019 Notes and 2021 Notes from the respective host debt instrument, which is referred to as debt discount, and initially recorded the conversion option of $214.6 million for the 2019 Notes and $267.4 million for the 2021 Notes in stockholders’ equity. The resulting debt discounts on the 2019 Notes and 2021 Notes are being amortized to interest expense at an effective interest rate of 5.75% and 6.25%, respectively, over the contractual terms of the notes.

Emphasis added, and probably ignore the rest. Twitter is saying that a regular non-convertible Twitter bond would pay interest of 5.75 percent for five years, or 6.25 percent for seven. That's higher than my guess, some of which may be due to new-issue premiums, and some of which may be due to conservatism in accounting. A third and even stranger way of doing the math gets a credit spread of around 260-270 basis points.

What does that tell you? Well, it lets you construct hypothetical debt instruments for Twitter -- say, a 5-year credit default swap spread of about 300 basis points, or a hypothetical new 7-year high-yield bond price of about 5.25 to 5.5 percent. Then you can compare those instruments to real instruments. For instance, E*Trade just priced a new 8-year high-yield bond at 5.375 percent. E*Trade is rated Ba3 by Moody's and B+ by S&P, or one notch below Twitter. And here are some credit default swap spreads for other companies with BB- ratings:

So if you believe my pricing, Twitter looks like a weak BB-, maybe a B+. If you believe Twitter's own pricing of its debt in its financial statements, it looked more like a single-B-area credit.

If you believed that Twitter was an investment-grade company ... why? The convertible bond market knew that Twitter was junk months ago. And Twitter told everyone a week ago. If stock investors are surprised that Twitter is not an investment-grade company, it's only because they weren't listening.

  1. That's over the last 12 months. Page 27 of the 10-Q has $169.4 million of adjusted EBITDA over the last nine months, plus another $45 million from the last quarter of 2013.

  2. Fortunately the Bloomberg terminal has one (OVCV). I am not aware of a plausible public domain model. You can fake out a model using a bond calculator, a Black-Scholes option calculator, and some hope, but that's no way to live, and also your model will be wrong.

  3. Based on Twitter convertible prices on Wednesday, the day before S&P's rating, for purposes of predictiveness. Here's the 5-year, based on trading prices of the bonds and stock at the close on Wednesday, Nov. 12:

    [imgviz image_id:iZk1qBPnZgOU type:image]

    And the 7-year, based on trading prices of the bonds and stock just before noon that day:

    [imgviz image_id:i5XibQjntnIA type:image]

    Source: Bloomberg OVCV function. Bond prices from Trace (Bloomberg TDH function). That 40 flat volatility is perhaps a bit aggressive; Twitter's historical volatility is in the 50s but Bloomberg implied volatility is around 44.

  4. I mean. Five-year swaps are around 1.75 percent, and seven-years are around 2.1 percent, so figure 4.75 percent for 5-year bonds and 5.1 percent for 7-years, plus a bit of new issue premium.

  5. That's a credit spread of a bit over 378 basis for a 5-year, or 387 for a 7-year. On September 12, the pricing date, I see 5-year U.S. Libor swaps at 1.97 percent, and 7-years at 2.38 percent.

  6. Extra credit: Twitter did the math in a different way for tax and accounting. Along with the convertible, Twitter also did what's called a "call option overlay," or "call spread." It "entered into privately negotiated convertible note hedge transactions" with its banks, in which it essentially bought back the call options embedded in the convertible bonds (with a strike price of $77.64 per share, versus Twitter's stock price of $52.64 at the time), and then it sold warrants to those banks with a strike price of $105.28 per share. The effect is, roughly speaking, to synthetically raise the conversion price of the convertible from $77.64 to $105.28 per share.

    (Disclosure: I used to do these trades for a living! At Goldman Sachs, which did some of Twitter's trades! But there's no inside information or secret sauce here, just public filings and standard bond math.)

    For tax reasons, the note hedge (where Twitter buys back the call options) and the warrants (where it sells higher strike call options) are treated as separate transactions, and paid for separately. What that means is that you can ignore the warrants and pretend that Twitter issued synthetic straight debt by combining the convertible bonds and the note hedges. It issued $1.8 billion of convertible bonds, and paid $387.5 million to make them non-convertible. So for tax purposes, it valued that conversion option at $387.5 million.

    For accounting purposes, notice, it valued it at $482 million. (That's $214.6 million plus $267.4 million, from the 10-Q passage quoted in the text). So, off by about a hundred million. I assume that's just conservatism: For accounting, it's conservative to make your interest expense high (to reduce income); for tax, it's conservative to make it low (to increase the taxes you pay). (Not that Twitter pays a lot of taxes.)

    Twitter does not disclose how the cost of the bond hedge is split among the two series of bonds, but you can make a reasonable guess, because you want the 7-year rate to be roughly 50 basis points higher than the implied rate of the 5-year. (As it was in Twitter's accounting. Basically the Treasury and swaps curves are around 40 basis points steep there, and you add a bit for credit.) So that's:

    • $900 million of 5-year bonds with a 0.25 percent coupon rate, sold at an initial issue price of $728 million, and
    • $900 million of 7-year bonds with a 1 percent coupon rate, sold at an initial issue price of $685 million.

    The bond math there works out to about 4.6 percent for the 5-year and 5.1 percent for the 7-year. Using the same swap spreads as in footnote 4, you get a credit spread of around 260-270 basis points.

  7. This is kind of guesswork. Note 7-year swaps are now about 2.1 percent, add a bit of new issue premium and some curve to the 300 basis points and you get 5.25 to 5.5.

  8. Source for CDS data is Bloomberg (CMAN). Names chosen from the Markit CDX.NA.HY.23 index at whim, with a goal of being sort of but not exclusively tech-focused.

  9. If you use the call spread numbers (footnote 6), which are also Twitter's, you get something more solidly BB-.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

To contact the author on this story:
Matthew S Levine at

To contact the editor on this story:
Toby Harshaw at