Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Look up keyword
Like this
6Activity
0 of .
Results for:
No results containing your search query
P. 1
No, Small Probabilities are Not "Attractive to Sell"

No, Small Probabilities are Not "Attractive to Sell"

Ratings: (0)|Views: 17,471 |Likes:
Published by caitlynharvey
pdf
pdf

More info:

Published by: caitlynharvey on Nov 28, 2012
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

12/04/2012

pdf

text

original

 
No, Small Probabilities Are Not “Attractive to Sell”: A Comment
Nassim N. TalebVERY PRELIMINARY DRAFT FOR DISCUSSION (NOTSUBMISSION),CANNOT BE CITED
Owing to the convexity of the payoff of out-of-the money options, an extremely small probability of alarge deviation unseen in past data justifies rationally buying them, or at least justifies excessive cautionin
not being expose
to them, particularly those options that are extremely nonlinear in response tomarket movement. On needs, for example, a minimum of 2000 years of stock market data to assert thatsome tail options are “expensive”. The paper starts by discussing errors in Ilmanen (2012), whichprovides an exhaustive list of all arguments in favor of selling insurance on small probability events.The paper goes beyond a discussion of Ilmanen (2012) and presents an approach to analyze the payoff and risks of options based on the nonlinearities in the tails and heuristics to measure fragility of exposures.
Background
To the question,
 Do Financial Markets Reward Buying or Selling Insurance and Lottery Tickets?
Ilmanen (2012) offers a negative answer tobuying, seemingly backed by a great deal of "empirical" examination, and citing a large number of studies. And he concludes that investorsshould not be insured, or should perhaps sell such insurance.Perhaps too many papers and arguments for comfort. Just as in a complicated detective novel, where it is often the character with the largestnumber of alibi who turns out to the the murderer, the enumeration of back-up arguments fails to mask that a central element is missing in thepaper: nonlinear effects and asymmetries. Just adding these nonlinear responses of tail payoffs does more than reverse the result. And thepaper also misses asymmetries in reasoning required for decision-making and statistical decidability. But Ilmanen (2012) is priceless because itincludes a review-study of all supporting arguments against the purchase of small-probability events and it turns out that there is not a singleconvincing study to that effect.
 Problems Proper to Ilmanen(2012)
More specifically, the elephant in the room in Ilmanen(2012) is in the two large misses:
Problem 1-A)
The paper excludes the stock market crash of 1987 from the analysis.
Problem 1-B)
The paper ignores the disastrous effect of bank loans (small probability selling) from the 2008 debacle (and those of the1991 and 1982 credit problems) while citing explicitly Taleb(2004), a paper that includes them as a domain of tail selling. The losses of 2008, as estimated by the IMF at ~$5 trillion (before the government transfers and bailouts) would offset every single gain from tail sellingin the history of economics.Excluding the crash of 1987 and bank loans would be like claiming that the 20th century was extremely peaceful by excluding the first andsecond world war. These two fallacies on their own would be devastating for the entire idea. The very addition of these events invalidates theentire conclusion of the paper, but let us examine some more errors related to nonlinearities.
Problem 2-A)
The paper makes the severe error of ignoring the effect of Jensen’s inequality from the nonlinearity of the differencebetween VIX and delivered volatility (the VIX,
by design
, delivers a payoff that is closer to the variance swap. Comparing the VIX todelivered volatility is erroneous: say the VIX was “bought” at 10% (that is the component options are purchased at a combination of volatilities that corresponds to a VIX at that level), because of Jensen’s inequality, it could benefit from an episode of 4% volatilityfollowed by an episode of 14%, for an average of 9.5% while Ilmanen seems to treat this .5% gap as a loss).
Problem 2-B)
Using VIX is inappropriate to gauge small probabilities. The VIX is not quite representative of the “tails”; its value isdominated by at-the-money options, and the fact that at-the-money options can be expensive has no bearing on the argument since we areconcerned with the tails (this author, when betting on fat-tailedness, tends to sell at the money options as one can safely say, in agreementwith Ilmanen (2012) that, owing to their linearity, that they are patently expensive; and such a statement is robust).The paper has other related confusions from missing the nonlinear nature of the payoffs, such as:
Problem 2-C)
Citing the phenomenon called "long shot bias" while referring to papers on bounded payoffs and payoffs of a binary naturein unrelated domains (what Taleb (2007) calls the "ludic fallacy"). These packages in papers cited such as Golec and Tamarkin (1998) andSnowberg and Wolfers (2010) are not sensitive to fat-tails (there are no true exposure to extreme payoffs). For a discussion of the problemsee a short note by the author (Taleb, 2012).and
Problem 2-D)
The study conflates long volatility trading (a more or less convex strategy) with investment in high or low volatility stocksand uses the successful historical performance from investing in low volatility stocks to back up the idea of selling tail volatility.Finally, we go beyond Ilmanen(2012) to present a general approach in deriving conclusions from a time series. Even if the Ilmanen paper weredevoid of the previous problems, a more general way to look at risk from times series would invalidate it as follows.
 
Errors Proper to General Inference From Time Series
A common error is made in the interpretation of time series by confusing an estimated value for a “true” parameter, without establishing thelink between estimator from the realization of a statistical process and the estimated value. Take M the true mean of a distribution in thestatistical sense, with
M
N
*
the observed arithmetic mean of the realizations
 N 
.Note that scientific claims can only be made off M, not M*. Limiting the discussion to M* is mere journalistic reporting. Mistaking M* for Mis what the author calls the “Pinker mistake” of calling journalistic claims “empirical evidence”, as science and scientific explanations are abouttheories and interpretation of claims that hold outside a sample, not discussions that limit to the sample.Assume we satisfy standard statistical convergence where our estimator approaches the true mean (convergence in probability, that is,
limit
N
ö¶
P[ |
M
N
*
-M|>
e
]=0). The problem is that for a finite M, when all realizations of the process when the distribution is skewed-left, say bounded by 0on one side, and is monomodal, E[
M
N
*
] > M. Simply, returns in the far left tail are less likely to show up in the distribution, while these havelarge contributions to the first moment.{NOTE TO EDITOR: CAN THE JOURNAL READERS HANDLE THEOREMS? OTHERWISE WEUSE WORDS FOR ABOVE BUT A THEOREM CAN PROVE WHY PRE-ASYMPTOTETICALLYE[
M
N
*
] > M FOR NEGATIVELY SKEWED PROCESSES. }A simple way to see the point: the study assumes that one can derive strong conclusions from a single historical path not taking into accountsensitivity to counterfactuals and completeness of sampling. It assumes that what one sees from a time series is the entire story.
ere ata ten to e m ss ngOutcomesProbability
Figure 1: The Small Sample Effect and Naive Empiricism
: When one looks at historical returns that are skewed to the left, most missing observationsare in the left tails, causing an overestimation of the mean. The more skewed the payoff, and the thicker the left tail, the worst the gap between observedand true mean.
Now of concern for us is assessing the stub, or tail bias, that is, the difference between M and M*, or the potential contribution of tail eventsnot seen in the window used for the analysis. When the payoff in the tails is powerful from convex responses, the stub becomes extremely large.So the rest of this note will go beyond the Ilmanen (2012) to explain the convexities of the payoffs in the tails and generalize to classicalmistakes of testing strategies with explosive tail exposures on a finite simple historical sample. It will be based on the idea of metaprobability(or metamodel): by looking at effects of errors in models and representations. All one needs is an argument for a
very
small probability of alarge payoff in the tail (devastating for the option seller) to reverse long shot arguments and make it uneconomic to sell a tail option. All ittakes is a small model error to reverse the argument.
The Nonlineatities of Option Packages
There is a compounding effect of rarety of tail events and highly convex payoff when they happen, a convexity that is generally missed in theliterature. To illustrate the point, we construct a "return on theta" (or return on time-decay) metric for a delta-neutral package of an option, seenat
t
0
o given a deviation of magnitude
N
K
.
(1)
P
H
N, K
L
ª
1
q
S
0
,t
0
,
d
J
O
J
S
0
e
N
K
d
, K, T
-
t
0
,
K
N
-
O
H
S
0
, K, T
-
t
0
-d
,
K
L
-D
S
0
,t
0
H
1
-
S
0
L
e
N
K
d
N
Where
0
H
S
0
, K, T
-
t
0
-d
,
K
L
is the European option price valued at time
t
0
off an initial asset value
S
0
, with a strike price K, a finalexpiration at time T, and priced using an “implied” standard deviation
K
. The payoff of 
P
is the same whether O is a put or a call, owing to t,owing to the delta-neutrality using
D
S
0
,t
0
(thanks to put-call parity).
q
S
0
,t
0
is the discrete change in value of the option over a time increment
d
(changes of value for an option in the absence of changes in any other variable). With the increment
d
=
1252
, this would be a single businessday. We assumed interest rate are 0, with no loss of generality (it would be equivalent of expressing the problem under a risk-neutral measure). 
2
 
Ilmanen.nb
 
 What Eq (1) did is re-express the Fokker-Plank-Kolmogorov differential equation (Black Scholes),
¶ 
¶ 
t
= -
12
2
 
s  
2
¶ 
2
¶ 
2
in discrete terms, awayfrom the limit of 
d
 
Ø
0. In the standard Black-Scholes World, the expectation of 
(N,K 
) should be zero, as N follows a Gaussian distributionwith mean -
12
2
. But we are not about the Black Scholes world and we need to examine payoffs to potential distributions. The use of 
K
neutral-izes the effect of "expensive" for the option as we will be using a multiple of 
K
as N standard deviations; if the option is priced at 15.87%volatility, then one standard deviation would correspond to a move of about 1%, Exp[
1252
. 1587].Clearly, for all K,
P
[0,K]=-1 ,
P
[
2
p
,K]= 0 close to expiration (the break-even of the option without time premium, or when
T
-
t
0
=
d
,takes place one mean deviation away), and
P
[ 1,K]= 0.
Convexity and Explosive Payoffs
Of concern to us is the explosive nonlinearity in the tails. Let us examine the payoff of 
P
across many values of K=
S
0
Exp[
L
 
K
 
d
], inother words how many “sigmas” away from the money the strike is positioned. A package about 20
s
out of the money , that is,
L
=20, thecrash of 1987 would have returned 229,000 days of decay, compensating for > 900 years of wasting premium waiting for the result. Anequivalent reasoning could be made for subprime loans. From this we can assert that we need a minimum of 900 years of data to start pronounc-ing these options 20 standard deviations out-of-the money “expensive”, in order to match the frequency that would deliver a payoff, and, morethan 2000 years of data to make conservative claims. Clearly as we can see with
L
=0, the payoff is so linear that there is no hidden tail effect.
L=
20
L=
10
L=
0
N
P
H
 N 
L
5 10 15 20200040006000
 
8000
Figure 2:
Returns for package
(N,K= 
S
0
Exp[
L
 
K
] ) at values of
L
= 0,10,20 and N, the conditional “sigma” deviations.
Ilmanen.nb
3

Activity (6)

You've already reviewed this. Edit your review.
1 hundred reads
bodnarti liked this
eric997 liked this
caitlynharvey liked this

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->