Vous êtes sur la page 1sur 4

Dropping Acid in Disneyland: Thoughts on the New Normal TM

JM
April 30, 2010

The boom and bust cycles of the new normal (as it wishes to be called) can be characterized. This characterization
provides clues about the nature of its endgame and progress toward that end.

 Economic growth fueled by easy monetary policy leads to excessive investment in specific assets
(bubbles), resulting in more extreme boom and bust cycles.
 Government policies to regulate the cycles lead to further crises that impact higher levels of capital
structure.
 In particular, very steep yield curves result in a generalized carry trade.
 Equity valuations always get smashed in a crash. However, in the new normal the top of the corporate
capital structure is not as immunized as it has been.
 The next to get smashed? Sovereign credit.
 The Endgame? The United States lives within their means, or U.S. t-bills get hammered.

Remembrance of Meltdowns Past


Equities Bonds
MSCI All- MSCI MSCI
World Developed Emerging iBoxx HY iBoxx IG Sovereign U.S. T-
Date Index Markets Markets Premium Premium Bonds Bills
LTCM Stumblebum Aug-98 -14.20% -13.50% -29.30% -2.20% -1.50% 2.50% 0.90%
Tech Bubble Screw-Job Nov-00 -6.20% -6.10% -8.80% -1.30% -0.70% 2.00% 0.70%
September 11, 2001 Sep-01 -9.10% -8.80% -15.50% -1.40% -0.70% 0.80% 1.00%
Idiot Quant Crisis Aug-07 -0.20% 0% -2.10% -0.80% -0.90% 1.60% 0.70%
Lehman Clusterfluke Oct-08 -19.80% -18.90% -27.40% -7.40% -6.30% -1.90% 0.70%
Average Loss -9.90% -9.46% -16.62% -2.62% -2.02% 1.00% 0.80%
Note: Red cells indicate losses greater than average over the listed crashes.

The typical stuff you often read is that “averaging these losses out, optimal return is obtained with the following
asset class allocation… blah, blah, blah...” The real point is quite different. Observe that the boom-bust cycles
exhibit a tendency toward greater losses at the top the capital structure. This is not just in corporate names. If
you permit the analogy, sovereigns were marginally affected in the latest bust (see above). You can see it written
all over SovX quotes. It will get worse until policymakers stop rewarding failure.

My conjecture is simple: more and more capital structure senior assets are synchronized to a measure of
disruption in the subordinate capital structure: VIX. This instability is the new normal until another new normal
comes along.

Focus on 2008

A closer look at those capital structure losses in the last bust gives a feel for the cloth. That Spoos vaporize is
nothing new. CDX.IG is an index of big liquid CDS names. There was a tremendous amount of stress on cash
markets, shown by the IG bond index blow-out, and that the term structure of SP500 vol was inverted for a good
chunk of Q4 2009.

The Year the Music Died, selected quotes


5Y Swap CDX.IG iBoxx IG
Milestone Insolvencies Date VIX ATM 5Y Vol SP500 Rate spread Bond spread
1/2/2008 12.0% 18.7% 1,417 5.06% 0.33% 0.14%
Bear Sterns 3/17/2008 32.2% 27.0% 1,277 3.17% 1.85% 4.64%
Merrill Lynch 8/29/2008 20.7% 25.4% 1,283 4.03% 1.43% 3.43%
Freddie, Fannie 9/8/2008 22.6% 25.2% 1,268 3.91% 1.38% 3.41%
Lehman 9/15/2008 31.7% 25.3% 1,193 3.69% 1.94% 5.04%
AIG Failure 9/16/2008 30.3% 26.3% 1,214 3.45% 2.00% 5.07%
12/31/2008 40.0% 35.3% 903 2.10% 2.14% 8.13%

Data Porn: Piecewise Regression and Correlation Functions

Imagine a financial experiment across a spectrum. At one extreme are very solid senior credits, at the other
extreme are residual securities, subordinate in terms of capital structure. Shocks to one extreme manifest in
higher credit spreads to the “risk-free” reference rate; the other extreme results in selling. I use daily bond
spreads and VIX quotes for this. VIX is good for this purpose because it is traded and thus demonstrates extreme
effects better. And yes… I too am stunned, benumbed, and burned by the hateful chasm between ratings and
reality. The agencies themselves engage in at best pro-cyclical cheerleading. At worst they are... pathetic. Close
eyes, think of England.

The model: (Baa spreads)t =  0 +  1(AAA spreads)t +  2(Closing price of VIX)t = t

t = daily quotes
AAA, Baa spreads = ratings based on Moody’s seasoned bonds, daily quotes.

2
From 1990 to 1999, the fit was near perfection (R ), and Baa spreads were dominated by AAA spreads (t-value).

Regression: Jan 1990- Dec 1999


Parameter
Variable N Estimate t Value Pr > |t| R2
Intercept -0.38698 -14.41 <.0001
0.9804
VIX 2620 0.01144 24.48 <.0001
AAA Spread 2625 1.12232 354.96 <.0001

From 2000 to 2005, the model says essentially the same thing, though the overall fit is marginally diminished.

Regression: Jan 2000 - Dec 2005


Parameter
Variable N Estimate t Value Pr > |t| R2
Intercept 1.20161 33.24 <.0001
0.9546
VIX 1447 0.02261 26.24 <.0001
AAA Spread 1435 0.88106 134.28 <.0001

From 2006 to April 2009, the situation is different. The overall fit now admits specification issues (it’s still real
good fit), but more to the point, VIX now has more explanatory power over Baa spreads than AAA spreads (bold t-
value).
Regression: Jan 2006 - Apr 2010
Parameter
Variable N Estimate t Value Pr > |t| R2
Intercept 2.28122 11.96 <.0001
0.8183
VIX 1149 0.05061 66.36 <.0001

AAA Spread 1137 0.62153 17.81 <.0001

In these regressions I lagged VIX as well. There was no appreciable difference in fit are parameter estimates. This
implies that either the relationship between VIX and Baa spreads is simultaneous or the effect in time is somewhat
complex.

Physics or Technical Analysis of Cap Structure

The correlation function is a way to determine how the relationship of VIX and Baa spread evolves in time. The
correlation function is a mapping that measures the correlation of changes in one time series to changes in the
other time series.

The analysis below uses daily quotes; back-testing on intraday quotes shows evidence of a power-law relationship.
Simulations haven’t performed as well on single reference entities as it has on indexes for a high-frequency cap
structure trading strategy.

The chart below takes a given Baa spread in time and measures the strength of correlation to the closing price of
VIX from 24 days before to 24 days after. The closer a line is 0 on the horizontal axis, the less relationship there is
between Baa spreads and VIX. The more positive or negative the line becomes shows how positively or negatively
correlated Baa spreads and VIX are to each other when looking at Baa spreads on, say, Jan 1 and at VIX 24 days in
the past and 24 days in the future.

Changes in the Baa spread is negatively correlated to the closing price of VIX on the same day. However, Baa
spread rises made VIX more likely to rise then next two days afterward. Note also that the correlation between
Baa spreads and VIX has more extreme positive and negative correlations in the 2006-2010 period. These
extremes seem to characterize the new normal.

There are other details you can determine on your own.

Vous aimerez peut-être aussi