Wednesday, September 02, 2009

Arithmetic Returns For Junk Biased

I noted in my book Finding Alpha that junk bonds have not outperformed investment grade bonds since data on junk bonds really became available, around 1987. This is the real corporate bond puzzle, in direct contrast to the corporate bond puzzle most academics address, which is the anomalously high return premium between BBB and AAA bonds (around 100 basis points annually).

Academics seem to consistently miss the big picture in corporate bonds, which to me is the lame returns on patently riskier junk bonds over the business cycle. For example, Steve Cecchetti’s textbook Money, Banking, and Financial Markets, immediately presents the seemingly straightforward example of how bonds with higher default rates have higher yields: Risk and return rise together. Yet, this is purely an anticipation of the default rates, and so is not risk in the sense of something priced. BBB bonds have, over time, about the same total return as B-rated bonds. One must subtract the expected defaults and the resulting losses from a stated yield regardless of one’s risk tolerance.

The successful and ubiquitous usage of one flavor of 'risk'—the mere statistical volatility and loss estimation—does not imply the second flavor of 'risk' relating to a priced factor affecting future returns as also ubiquitous and essential. The distinction between or default risk by itself and priced risk (a covariance with some systemic metric of aggregate happiness, such as GDP, the S&P500, or unemployment) is a fundamental distinction in modern risk-return theory, yet prominent professors conflate risks when useful for selling the old bromide that risk and return go together like shoes and socks. Financial professionals have a strong, perhaps unconscious, bias toward the big idea: Risk begets average returns.

In addition to stated yield vs. actual returns, or survivorship biases, there is the simple fact of the difference between an annual average, and a cumulative return. Basically, the difference between a geometric mean and an arithmetic mean. In bonds this is huge. As mentioned yesterday, bond returns were down 26% in 2008, but up 40% in 2009. Does that mean they have a 14% total return? No. Think (1-.26)*(1+.40) and you get an 3.6% total return. But remember, those numbers are from Merrill Lynch indices using a collection of highly illiquid bonds closing prices. Actual bond fund returns from the beginning of 2008 have been -1%. See below for a collection of returns from the subperiods, and note how they compare to the entire period.

Below are data through August 2009 for a collection of High Yield bond funds:
2008 Ret2009 Ret2008-9 Ret
BLACK-HI INC SHSHIS-3756-2
NEW AMER HI INCHYB-409417
HIGH YLD PLUS FDHYP-285914
MORGAN ST HI YLDMSY-27456
VAN KAMP HI INC2VLT-4559-12
MFS INTERMEDIATECIF-4056-6
PUTNAM MGD MUNIPMM-23333
MFS HIGH INC MUNCXE-43740
DWS HIGH INCOMEKHI-3039-3
MFS HIGH YIELD MCMU-38621
BLACKROCK-COR HYCOY-39768
BLACKROCK-SR HIGARK-4942-28
WESTERN ASSET INSBI-92211
PACHOLDER H/Y FDPHF-4884-4
FRANKLIN UNIVERSFT-4145-14
HIGH YLD INC FNDHYI-26403
MFS MUNI INC TSTMFM-36688
WESTERN ASSET MUMHF-51610
MORGAN ST MU IN3OIC-28334
DWS STR MUN INCMKSM-204919
MORGAN ST MU INOIA-3137-6
BLACKROCK-APEX MAPX-2736-1
FEDERATED I MUNIFPT-174521
MORGAN ST MU IN2OIB-333510
Average-3250-1

9 comments:

The Recovering Banker said...

Of course, the fees are higher for junk bonds (for the underwriter, the fund manager etc). There is a similar issue for fund of hedge funds, which have fixed income type returns.

I wonder how often the benefits of a riskier and/or illiquid and/or esoteric asset class accrue more to the financial industry than to the investor.

Anonymous said...

Eric,

Which data are you using for the junk bonds?

If you look here:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=943326

The average yield is 4.87 and the realized return is 2.28 (2.22 geometric). The weighted average default rate is ~5% and recoveries ~50%.

yield = treasury rate + expected loss + taxes + risk premium

Ignoring taxes:
4.87 - 5%*50% = 2.37% about the actual realized return.

Am I missing soemthing?

Eric Falkenstein said...

That paper is about one year, 2005. The data above are for 2008 and 2009. My general proposition comes from High Yield funds and indices since 1987.

Anonymous said...

Eric,

In the paper, Altman provides returns for the 1978-2005 period. The numbers I mentioned are averages for the 27 yr period (page 18)

The risk premium (and the average return) is about 2.3% for high yield bonds.

Which dataset are you using for junk bonds?

Eric Falkenstein said...

I'm using Merrill Lynch's High Yield Master II, which is a total return index. On the overlapping period 1987-2005, that has an 8.5% return vs. Altman/Citi's 9.0% return, pretty close. But I also note that looking at actual junk funds, these reported a return a couple percentage points lower, reflecting the high transaction costs in this space (this, using Bloomberg, 19 funds, with a survivorship bias).

The data prior to 1987 are pretty bad. There really weren't any good prices prior to 1987, which is why Mike Milken made so much money (he would offer 20 point wide markets in the mid eighties).

Robert H. Heath said...

Eric -

You say (1-.26)*(1+.40) equals a total return of 8%. But the math says 3.6%. Am I missing something, or is this just the capital gain/loss portion, and there's another 4.4% yield over the same period?

Thanks.

Eric Falkenstein said...

oop. math error. Tx. I was using -46% at one iteration of writing that.

Anonymous said...

Eric,

I don't think I understand the point about the risk premium.
The average historical spread on BBB is 1.25% and 5% for B-. The default rates are 5% and 0.4% respectively.

Expected return differential is:

5% - 5%*50% = 2.5%
1.25% - 0.4%*50% = 1%


The expected risk premium between the two is 1.5% Are you saying that this does not show up in the realized returns?

Eric Falkenstein said...

There is a difference between junk returns and spreads-defaults*LGD because of the mark-to-market, the convexity of prices in relation to yields, and the fact that most bonds are sold when the become truly distressed in price. For example, a bond goes from a spread of 500 to 2000, as people think it will default. The fund sells it at a big loss, much greater than any gain it gets going from 500 to 300.
Such bonds often don't default.

Why don't funds simply hold on to every bond and reap the returns suggested from your math? Well, many have it in their prospectus they don't actively manage less than B- bonds, so they want out, and in the depth of a crisis, with only 22 years of good data, not many people are so confident the sample mean is the population mean.