E-piphany

I mentioned a week and a half ago that I’d had a “really cool” idea that I had mentioned to a member of the Fed’s Open Market Desk, and I promised to write about it soon. “It’s an idea that would simultaneously be really helpful for investors and help the Fed reduce a balance sheet that they claim to be happy with but we all really know they wish they could reduce.” First, some background.

It is currently not possible to directly access any inflation index other than headline inflation (in any country that has inflation-linked bonds, aka ILBs). Yet, many of the concerns that people have do not involve general inflation, of the sort that describes increases in the cost of living and erodes real investment returns (hint – people should care, more than they do, about inflation), but about more precise exposures. For examples, many parents care greatly about the inflation in the price of college tuition, which is why we developed a college tuition inflation proxy hedge which S&P launched last year as the “S&P Target Tuition Inflation Index.” But so far, that’s really the only subcomponent you can easily access (or will, once someone launches an investment product tied to the index), and it is only an approximate hedge.

This lack has been apparent since literally the beginning. CPI inflation derivatives started trading in 2003 (I traded the first USCPI swap in the interbank broker market), and in February 2004 I gave a speech at a Barclays inflation conference promising that inflation components would be tradeable in five years.

I just didn’t say five years from when.

We’ve made little progress since then, although not for lack of trying. Wall Street can’t handle the “basis risks,” management of which are a bad use of capital for banks. Another possible approach involves mimicking the way that TIGRs (and CATS and LIONs), the precursors to the Treasury’s STRIPS program, allowed investors to access Treasury bonds in a zero coupon form even though the Treasury didn’t issue zero coupon bonds. With the TIGR program, Merrill Lynch would put normal Treasury bonds into a trust and then issue receipts that entitled the buyer to particular cash flows of that bond. The sum of all of the receipts equaled the bond, and the trust simply allowed Merrill to disperse the ownership of particular cash flows. In 1986, the Treasury wised up and realized that they could issue separate CUSIPs for each cash flow and make them naturally strippable, and TIGRs were no longer necessary.

A similar approach was used with CDOs (Collateralized Debt Obligations). A collection of corporate bonds was put into a trust[1], and certificates issued that entitled the buyer to the first X% of the cash flows, the next Y%, and so on until 100% of the cash flows were accounted for. Since the security at the top of the ‘waterfall’ got paid off first, it had a very good rating since even if some of the securities in the trust went bust, that wouldn’t affect the top tranche. The next tranche was lower-rated and higher-yielding, and so on. It turns out that with some (as it happens, somewhat heroic) assumptions about the lack of correlation of credit defaults, such a CDO would produce a very large AAA-rated piece, a somewhat smaller AA-rated piece, and only a small amount of sludge at the bottom.

So, in 2004 I thought “why don’t we do this for TIPS? Only the coupons would be tied to particular subcomponents. If I have 100% CPI, that’s really 42% housing, 3% Apparel, 9% Medical, and so on, adding up to 100%. We will call them ‘Barclays Real Accreting-Inflation Notes,’ or ‘BRAINs’, so that I can hear salespeople tell their clients that they need to get some BRAINs.” A chart of what that would look like appears below. Before reading onward, see if you can figure out why we never had BRAINs.

When I was discussing CDOs above, you may notice that the largest piece was the AAA piece, which was a really popular piece, and the sludge was a really small piece at the bottom. So the bank would find someone who would buy the sludge, and once they found someone who wanted that risk they could quickly put the rest of the structure together and sell the pieces that were in high-demand. But with BRAINs, the most valuable pieces were things like Education, and Medical Care…pretty small pieces, and the sludge was “Food and Beverages” or “Transportation” or, heaven forbid, “Other goods and services.” When you create this structure, you first need to find someone who wants to buy a bunch of big boring pieces so you can sell the small exciting pieces. That’s a lot harder. And if you don’t do that, the bank ends up holding Recreation inflation, and they don’t really enjoy eating BRAINs. Even the zombie banks.

Now we get to the really cool part.

So the Fed holds about $115.6 billion TIPS, along with trillions of other Treasury securities. And they really can’t sell these securities to reduce their balance sheet, because it would completely crater the market. Although the Fed makes brave noises about how they know they can sell these securities and it really wouldn’t hurt the market, they just have decided they don’t want to…we all know that’s baloney. The whole reason that no one really objected to QE2 and QE3 was that the Fed said it was only temporary, after all…

So here’s the idea. The Fed can’t sell $115bln of TIPS because it would crush the market. But they could easily sell $115bln of BRAINs (I guess Barclays wouldn’t be involved, which is sad, because the Fed as issuer makes this FRAINs, which makes no sense), and if they ended up holding “Other Goods and Services” would they really care? The basis risk that a bank hates is nothing to the Fed, and the Fed need hold no capital against the tracking error. But if they were able to distribute, say, 60% of these securities they would have shrunk the balance sheet by about $70 billion…and not only would this probably not affect the TIPS market – Apparel inflation isn’t really a good substitute for headline CPI – it would likely have the large positive effect of jump-starting a really important market: the market for inflation subcomponents.

And all I ask is a single basis point for the idea!

[1] This was eventually done through derivatives with no explicit trust needed…and I mean that in the totally ironic way that you could read it.