Financially engineered.

Photographer: Simon Dawson/Bloomberg.

SEC Investigates Bank of America for Doing Nothing

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

Here's a Wall Street Journal article about a Securities and Exchange Commission investigation into "whether Bank of America Corp. broke rules designed to safeguard client accounts" by doing "an array of complex trades and loans" with clients. Generically, I delight in this sort of thing, but specifically, I don't know what those complex trades were or how they worked. But here we are on the Internet, so let's not let that slow us down. Instead, let's just reason the trades out from first principles.

We do have one important clue, which is that "one variety of the strategy, launched around 2009, was called 'leveraged conversion.'" Let's just start with the name: "leveraged conversion." Sheldon Natenberg, who wrote a good book on options, defines a conversion as a trade "where the underlying contract is offset by the sale of a synthetic position": buy stock, sell a call, buy a put with the same strike price. Schematically:

"Leverage" also has a well-known meaning. It means borrowing money. The Journal says that, "A small team from Bank of America’s equities desk recruited a handful of clients, including wealthy individuals, to put up token amounts of their own money -- a few million dollars in some cases -- and in exchange receive loans of nearly 100 times those amounts." So a leveraged conversion would be: take a loan, use it to buy stock, then sell a call and buy a put. Schematically:

So Bank of America Merrill Lynch lends the client money. The client uses the money to buy some stock. The client also buys a put option, protecting the downside of the stock, and sells a call option, giving up the upside.

Now that we've built a leveraged conversion in our imaginations, let's collapse it. First of all, let's deal with this put and call. The point of a conversion is, as Natenberg says, to offset the underlying thing (the stock) with "the sale of a synthetic position." The put and call have the same strike price and expiration, and combine to create a "synthetic" forward sale of the stock. In a conversion on Apple, you might buy Apple stock for $129 a share, sell your bank a call with a strike price of $130 and buy a put from your bank with a strike price of $130. The put might cost you about $5.25 a share, and the call might bring in about $4.30 a share.  You've spent $129 on the stock and about $1 net on the options, for a total price of $130. Then you wait a few months or whatever, and the put and call expire. If the stock price is above $130, the bank will exercise the call you sold it: The bank pays you $130, and you give the bank the stock (which is worth more than $130). If the stock price is below $130, you will exercise the put you bought from the bank: The bank pays you $130, and you give the bank the stock (which is worth less than $130).  Whatever happens, you end up with (1) no stock and (2) $130, which happens to be the amount you spent up front. You just have to wait for the options to expire. 

The put and the call together are commonly referred to as a "put-call combo," and look just like a forward sale. Graphically:

(Remember to flip the call graph upside-down because you're subtracting it!) So now our leveraged conversion is simplified a bit. Substituting:

We can keep going. The client has bought some stock from the bank now and has agreed to sell it back to the bank later. What is that? Well, there is a commonly used name for what happens when you buy a thing and immediately agree to sell it back in a little while. The name is "repo," short for "repurchase agreement," because that's what it is: buying a thing with an agreement that the seller will repurchase it. One well-known fact about a repo is that it is a loan: The initial buyer is lending money to the initial seller, secured by the stock that the seller delivers to the buyer. This is such a well-known fact about repo that repos are normally accounted for as loans, not sales and repurchases.  

This trade is not exactly a repo, but it has the same structure (buy stock and simultaneously agree to sell it forward). So now we have:

At this point anyone can play along. How do you collapse a loan that the bank makes to the client, with a loan that the client makes to the bank? That's an easy one:

This, by the way, is a good way to check that your financial-engineering transaction worked. If you collapse all the bits of your trade and are left with nothing, you have achieved pure financial engineering. If you are left with something rather than nothing, then you have entered into an economic transaction. Maybe you wanted to do that, but it detracts from the purity of your financial engineering.

The point of the engineering, traditionally, is to achieve no or minimal economic effects, while achieving some regulatory (or tax, or accounting, or whatever) effect. That seems to have happened here. Here's a very schematic summary:

  1. Bank of America Merrill Lynch lends client $130.
  2. Client buys $129 worth of stock from BAML.
  3. Client sells stock forward to BAML for $130, via a put-call combo for which client spends $1. 
  4. At expiry of the forward (put-call combo), client delivers stock back to BAML in exchange for the agreed forward price of $130.
  5. Client delivers the $130 back to BAML to close out the loan.

Nothing enters or leaves the system.  Energy is conserved. The client gets paid a bit of money -- that is, the financing rate on the synthetic forward is higher than the financing rate on the loan -- for his trouble.

On the other hand, there are regulatory effects. Here is how the Journal describes the relevant rule:

Known in industry circles as Rule 15c3-3, a reference to the section where it is tucked within the Securities Exchange Act of 1934, the policy requires banks and other financial firms that handle customer trades to calculate at least once a week their net liabilities to clients -- in other words, how much more banks owe to clients, in the form of deposits and other funds, than they are owed from clients, in the form of assets such as loans.

The greater the overall -- or net -- amount that the banks owe their clients, the more money the banks need to set aside in reserve funds, known as “lockup” accounts, to pay their customers in an emergency. Funds in those lockup accounts must be segregated from other accounts, including the banks’ own.

Having billions of dollars idling in lockup is expensive for banks, partly because the money generally can’t be put to other profitable but potentially risky uses such as trading.

Really simplistically, if you want to reduce the net amount of money you owe clients, you should lend clients some money. But that is an economic transaction; maybe you don't want to lend them money. (Maybe they don't want to borrow money.) If you can lend them money, and then borrow it right back from them, then you have no economic transaction, so it's no problem. And if somehow you can lend them money and call it a loan, and then borrow it back from them and not call it a loan, then you get an A+ on this assignment. The amount they owe you goes into your Rule 15c3-3 calculation, the amount you owe them does not, and you've freed up lockup money without actually taking any risk by lending to your clients.

I assume that that is roughly what happened here. I am no expert on Rule 15c3-3, but you can read it, and the appendix that governs calculation of the "lockup" account: You count up what you owe clients, subtract what they owe you and put the money in a lockup. In this transaction, BAML's loan to the client looks like an amount the client owes, so it reduces the amount BAML needs to put into the lockup. But un-collapse the leveraged conversion again and you'll see that, while BAML's loan to the client is a "loan," the client's loan to BAML is lovingly handcrafted from stock and a put and a call, and isn't a loan at all. BAML's loan to the client goes into the 15c3-3 calculation at its full amount; the client's trade with BAML does not.

How bad is this? These trades don't seem particularly risky, for the bank or for the client: The loans are secured by, basically, each other. There's no realistic risk that BAML could lose money: If its customer defaults, BAML will seize the stock and the options, which will always be worth at least the amount of the loan.  So putting its money into this trade is arguably about as safe as putting it anywhere else -- even into a lockup -- and not worth worrying about too much. On the other hand, I mean, look at those cash flows again. BAML isn't putting money into this trade; it's putting, like, Apple stock into the trade. The trade itself is risk-free, but if it reduces the requirement for BAML to keep cash in a box, then that arguably makes BAML a bit riskier.

The other thing is that the construction and deconstruction of the leveraged conversion that I did above is ... pretty straightforward? I mean, as these things go. One rule of thumb of financial engineering is that it can't be too obvious that the transaction that you're doing is a purely regulatory transaction. It has to have some arguable economic effect, or look like it has an economic effect, or at least just have, like, more steps. If you make it too obvious that your transaction is about getting around rules, that's the sort of thing that the SEC doesn't like.

  1. Those are rough numbers from a glance at June options on Bloomberg at about 12:30 today.

  2. If the stock is worth exactly $130, the earth crashes into the sun.

    No I mean it's fine. In the theoretical case both options expire worthless, and you frictionlessly sell the stock for $130. In the practical case it's also fine.

  3. This is called "put-call parity," sort of.

  4. I mean technically the name is "reverse repo"; a repo is selling the thing with an agreement to buy it back. But they're two sides of the same trade, and I'm going to use the simpler term here.

  5. And one of the big scandals of the financial crisis is that Lehman Brothers accounted for some of its repos as sales rather than as loans.

  6. I mean, the client puts in a little money, but we can reasonably ignore that.

  7. Not literally. There's some very minor transfer of risk; presumably if for instance the bank goes bankrupt during the trade, the customer's only recourse is to take the stock. But the overall thrust of the transaction is to pay the client a little bit of money for not doing very much.

    Also, I am assuming that the two loans have similar maturities: that is, that the "leverage" bit and the "conversion" bit of the "leveraged conversion" mature at the same time. So the put and call expire, the client delivers the stock, the bank delivers the money back to the client, and the client delivers the money back to the bank to pay off the loan. That seems like a reasonable assumption.

  8. I don't really understand if it goes in at all. Conceivably the put that BAML writes to the client is a liability and thus creates some lockup requirements, though I'm not sure about that.

  9. It is a deep mystery that a stock plus a put is always worth at least the strike price of the put. It's a mystery because if you wrote the put, seizing the put from yourself doesn't quite feel like making yourself whole. But it is. If the put is in-the-money, you had a loss on it elsewhere, and now you're rid of that loss.

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:
Matt Levine at

To contact the editor on this story:
Zara Kessler at