Академический Документы
Профессиональный Документы
Культура Документы
Private
Profits and
Socialized Risk
watchdogs. This is like an airbag that works all the time, on many of them for the same reason. This is why in recent
except when you have a car accident. periods the investment banks had quarterly write-downs
By ignoring the tails, VaR creates an incentive to take that were many times the firmwide modelled VaR.
excessive but remote risks. Consider an investment in a
coin-flip. If you bet $100 on tails at even money, your VaR Ratings: Flawed Opinions from
to a 99% threshold is $100, as you will lose that amount Understaffed Agencies
50% of the time, which obviously is within the threshold. Which brings us to the third question, what were the watch-
In this case, the VaR will equal the maximum loss. dogs doing? Let’s start with the credit rating agencies. They
Compare that to a bet where you offer 127 to 1 odds have a special spot in our markets. They can review non-
on $100 that heads won’t come up seven times in a row. public information and opine on the creditworthiness of the
You will win more than 99.2% of the time, which investment banks. The market and the regulators assume
exceeds the 99% threshold. As a result, your 99% VaR is that the rating agencies take their responsibility to stay on
zero, even though you are exposed to a possible $12,700 top things seriously. When the credit crisis broke last sum-
mer, one of the major agencies held a public conference call
to discuss the health of the investment banks.
The gist of the rating agency perspective was, “Don’t
Everyone knows about the 20% worry.” The investment banks have excellent risk controls,
and they hedge their exposures. The initial reaction to the
incentive fees in the hedge fund and credit crisis basically amounted to “everyone is hedged.” A
few weeks later, when Merrill Lynch announced a big loss,
private equity industry. Nobody that story changed. But initially, the word was that every-
talks about the investment banks’ one was hedged. Securitization had spread the risk around
the world, and most of the risk was probably in Asia,
50% compensation structures, Europe, Dubai or at the bottom of the East River. The
banks were in the “moving” business, not the “storage”
which have no high-water mark and business, so this was no big issue. I wondered whether any-
one saying this had actually looked at the balance sheets.
actually are exceeded in difficult Of course, this raised the following question: how did
times in order to retain talent. everyone hedge and who were the counterparties holding
the bag? I pressed star-1 during the conference call and
asked the rating agency analyst how everyone hedged the
massive apparent credit risks on the balance sheets. The
loss. In other words, an investment bank wouldn’t have analyst responded that the rating agency had observed
to put up any capital to make this bet. The math whizzes enormous trading volumes on the Chicago Mercantile
will say it is more complicated than that, but this is the Exchange (“the Merc”) in recent days.
basic idea. The Merc offers products that enable one to hedge inter-
Now we understand why investment banks held enor- est rate risk, not credit risk. I called the rating analyst back
mous portfolios of “super-senior triple A-rated” whatever. to discuss this in greater depth. At first he told me that you
These securities had very small returns. However, the risk could hedge anything on the Merc. When I asked how to
models said they had trivial VaR, because the possibility of hedge credit risk there, he was less familiar. I came to sus-
credit loss was calculated to be beyond the VaR threshold. pect that the rating agency analyst viewed his role as one to
This meant that holding them required only a trivial restore confidence in the system, which the rating agency
amount of capital, and a small return over a trivial amount call did do for a while, rather than to analyze risk.
of capital can generate an almost infinite revenue-to-equity I later had an opportunity to meet a recently retired senior
ratio. VaR-driven risk management encouraged accepting a executive at one of the large rating agencies. I asked him how
lot of bets that amounted to accepting the risk that heads his agency went about evaluating the creditworthiness of the
wouldn’t come up seven times in a row. investment banks. By then, Merrill had acknowledged large
In the current crisis, it has turned out that the unlucky losses, so I asked him what the rating team found when it
outcome was far more likely than the backtested models went to examine Merrill’s portfolio in detail.
predicted. What is worse, the various supposedly remote He answered by asking me to refocus on what I meant by
risks that required trivial capital are highly correlated; you “team.” He told me that the group covering the investment
don’t just lose on one bad bet in this environment, you lose banks was only three or four people and they have to cover
that do not have a ready market under the current net capi- probability that a large broker-dealer could go bust.
tal rule will reduce the liquidity standards.”
These adjustments reduced the amount of required capi- The Fall of Bear Stearns
tal to engage in increasingly risky activities. The SEC esti- I don’t know what effect the new rules had on Bear Stearns.
mated at the time the rule was proposed that the broker- The information the broker-dealers provide the SEC to show
dealers taking advantage of the alternative capital compu- their compliance with these regulatory capital requirements
tation would realize an average reduction in capital deduc- is confidential. It would be interesting to know how ade-
tions of approximately 40%. From my reading, the final quately capitalized Bear and other large broker-dealers
rule appears to have come out even weaker, suggesting that would have been under the rules as they existed before 2004.
the capital deductions may have been reduced even further. In response to this possible regulatory failure,
Obviously, since the rule was implemented, the broker- Christopher Cox, the SEC chairman, said recently that this
dealers have modified their balance sheets to take advan- current voluntary program of SEC supervision should be
tage of it. They have added lots of exposure to low-return made permanent and mandatory. Reuters reported that
bonds with credit risk perceived to be beyond the VaR Cox said that the current value of the SEC supervisory pro-
threshold, and they have added more no ready market gram “can never be doubted again.”
securities — including whole loans, junior pieces of struc- Rather than looking at its own rules — which permitted
tured credit instruments, private equity and real estate. increased leverage, lower liquidity, greater concentrations
If this wasn’t enough, the 2004 rule also changed what of credit risk and holdings of no ready market securities —
counts as capital: “In response to comments received, the the SEC is conducting an investigation to see if any short-
Commission has expanded the definition of allowable capi- sellers caused the demise of Bear by spreading rumors.
tal … to include hybrid capital instruments and certain Of course, Bear didn’t fall because of market rumors. It
deferred tax assets,” the SEC explained. fell because it was too levered and had too many illiquid
The rule also permits the inclusion of subordinated debt assets of questionable value and at the same time depended
in allowable capital. The SEC permitted this because subor- on short-term funding. With the benefits of the reduced
dinated debt “has many of the characteristics of capital.” I capital requirements and reliance on flawed VaR analysis,
find this one particularly amazing; apparently, it doesn’t Bear — like the other investment banks — increased its risk
actually have to be capital. For everyone else except the profile over the last few years.
broker-dealers, subordinated debt is leverage. The commis- While VaR might make sense to the quants, it has led to
sion considered but stopped short of allowing the broker- risk-taking beyond common sense. If Bear’s only business
dealers to count all long-term debt as capital. was to have $29 billion of illiquid, hard-to-mark assets,
In reading through the rules and the SEC’s response to supported by its entire $10.5 billion of tangible common
comment letters, it seems that the SEC made concession equity, in my view, that by itself would be an aggressive
after concession to the large broker-dealers. I won’t bore investment strategy. However, as of November 2007, that
you by describing how the rule eased the calculations of sliver of equity was also needed to support an additional
counterparty risk, maximum potential exposures and mar- $366 billion of other assets on Bear’s balance sheet.
gin lending or how the rule permitted broker-dealers to When Bear’s customers looked at the balance sheet and
assign their own credit ratings to unrated counterparties. also noticed the increased cost of buying credit protection on
My impression of this is that the large broker-dealers con- Bear, they had to ask themselves whether they were being
vinced the regulators that the dealers could better measure compensated for the credit risk and counterparty risk in
their own risks, and with fancy math, they attempted to show doing business with Bear. Many decided that they weren’t and
that they could support more risk with less capital. I suspect did the prudent thing to protect their own capital by curtail-
that the SEC took the point of view that these were all large, ing their exposure. Bear suffered a classic “run on the bank.”
well-capitalized institutions, with smart, sophisticated risk
managers who had no incentive to try to fail. Consequently, Perils of a Reverse-Robin Hood System
they gave the industry the benefit of the doubt. When I came up with the title for this discussion, it was
In the cost-benefit analysis of the rule, on the benefit before Bear Stearns failed. I was going to point out that we
side, the SEC estimated the “value” to the industry by tak- were developing a system of very large, highly levered,
ing advantage of lower capital charges to earn additional undercapitalized financial institutions — including the
returns. In the “cost” part of the analysis, the SEC carefully investment banks, some of the large money center banks,
analyzed the number of hours and related expense of the the insurance companies with large derivatives books and
monitoring and documentation requirements, as well as IT the government-sponsored entities (GSEs). I planned to
costs. It did not discuss the cost to society of increasing the speculate that regulators believe all of these are too big to
mortgage assets into Level 3. Last year, Lehman reported its have come to regret them.
Level 3 assets actually had $400 million of realized and Lehman wants to concentrate on long investors; in fact,
unrealized gains. Lehman has more than 20% of its tangi- it went to great lengths to tell the market that it sold all of
ble common equity tied up in the debt and equity of a single its recent convert issue to long-only investors. Putting aside
private equity transaction — Archstone-Smith, a real estate the fact that some of the clearing firms have told us that
investment trust (REIT) purchased at a high price at the end this wasn’t entirely true, companies that fight short sellers
of the cycle. Lehman does not provide disclosure about its in this manner have poor records. The same goes for com-
valuation, though most of the comparable company trading panies that publicly ask the SEC to investigate short selling,
prices have fallen 20-30% since the deal was announced. as Lehman has done. There is good academic research to
The high leverage in the privatized Archstone-Smith would support my view on this point.
suggest the need for a multibillion-dollar write-down. As I have studied Lehman for each of the last three quar-
Lehman has additional large exposures to Alt-A mort- ters, I have seen the company take smaller write-downs
gages, CMBS and below-investment-grade corporate debt. than one might expect. Each time, Lehman reported a
Our analysis of market transactions and how debt indices modest profit and slightly exceeded analyst estimates that
performed in the February quarter would suggest Lehman each time had been reduced just before the public
could have taken many billions more in write-downs than it announcement of the results. That Lehman has not report-
did. Lehman has large exposure to commercial real estate. ed a loss smells of performance smoothing.
Lehman has potential legal liability for selling auction-rate Given that Lehman hasn’t reported a loss to date, there is
securities to risk-averse investors as near cash equivalents. little reason to expect that it will any time soon. Even so, I
What’s more, Lehman does not provide enough trans- believe that the outlook for Lehman’s stock is dim. Any
parency for us even to hazard a guess as to how they have deferred losses will likely create an earnings headwind
accounted for these items. It responds to requests for going forward. As a result, in any forthcoming recovery,
improved transparency grudgingly, and I suspect that Lehman might underearn compared to peers that have been
greater transparency on these valuations would not inspire more aggressive in recognizing losses.
market confidence. Further, I do expect the authorities to require the broker-
Instead of addressing questions about its accounting and dealers to de-lever. In my judgment, a back-of-the-envelope
valuations, Lehman wants to shift the debate to where it is calculation of prudent reform would require 50-100% cap-
on stronger ground. It wants the market to focus on its liq- ital for no ready market investments; 8-12% capital for
uidity. However, in my opinion, the proper debate should what the investment banks call “net assets”; 2% capital for
be about Lehman’s asset values, future earning capabilities the other assets on the balance sheet; and an additional
and capital sufficiency. charge that I don’t know how to quantify for derivative
In early April, Lehman raised $4 billion of new capital exposures and contingent commitments.
from investors, thereby spreading the eventual problems Only tangible equity, not subordinated debt, should
over a larger capital pool. Given the crisis, the regulators count as capital. On that basis, assuming that Level 3 assets
seem willing to turn a blind eye toward efforts to raise capi- are a good proxy for no ready market investments —
tal before recognizing large losses; this holds for a number assigning no charge for the derivative exposure or contin-
of other troubled financial institutions. gent commitments and assuming its asset valuations are
The problem with 44 times leverage is that if your assets fairly stated — Lehman, based on its November balance
fall by only a percent, you lose almost half the equity. sheet, would need $55-$89 billion of tangible equity, which
Suddenly, 44 times leverage becomes 80 times leverage and would be a three- to-five-fold increase.
confidence is lost. It is more practical to raise the new So what do I expect to happen? I just finished a book on
equity before showing the loss. Hopefully, the new Allied Capital and the lack of proper and effective regula-
investors understand what they are buying into, even tory oversight. Based on my book and the current regulato-
though there probably isn’t much discussion of this ry environment, the pessimistic side of me says that regula-
dynamic in the offering memos. Some of the sovereign tors will probably decide to send me a subpoena and send
wealth funds that made these types of investment last year Lehman a Coke. ■
✎ DAVID EINHORN is the president and founder of Greenlight Capital, a long-short value-oriented hedge fund. Since its inception in 1996,
Greenlight has generated a greater than 25% annualized net return for its partners. Einhorn is the chairman of Greenlight Capital Re, Ltd.
(Nasdaq: GLRE), and serves on the boards of the Michael J. Fox Foundation for Parkinson’s Research and Hillel: The Foundation for Jewish
Campus Life. Einhorn is the author of Fooling Some of the People All of the Time:A Long Short Story (Wiley, May 2008).
This article is a reprint of a speech given April 8, 2008, at Grant’s Spring Investment Conference
agree with David Einhorn that the financial dealer sys- later is less adequate. Basel I treated every asset like a loan,
positions and normal markets. You evaluate VaR by back- day VaR has to operate for about three years before you
test; you compute it every night on the portfolio, and check trust it. A 99.97% one-year VaR, which some people use
the next day how much that fixed portfolio would have for economic capital, requires 26,000 years for the same
gained or lost (this is not the same as the actual gain or loss level of confidence. That makes deep tail VaR a matter of
if there is trading during the day). You check that the loss faith and assumptions, not something you can observe with
exceeds VaR on 1% of the days (plus or minus expected reasonable statistical certainty over a moderate time inter-
statistical variation) and test that the break days are both val. Moreover, assumptions like fixed positions and normal
independent in time and independent of market factors markets make VaR far less relevant over longer periods at
(most importantly, the level of the VaR). smaller probabilities.
VaR is only as good as its backtest. When someone Most important, some days have undefined losses due
shows me a VaR number, I don’t ask how it is computed, I to events like market closures, extreme liquidity events,
legal uncertainties and so forth. And however careful you
are, there is some chance your VaR estimate is significantly
misstated due to data or calculation errors (including those
VaR is only as good as its backtest. induced by rogue traders and embezzlers and third-party
crooks). These might happen two or three times a decade,
When someone shows me a VaR about 0.1% of the time. VaR excludes these days, which is
reasonable in a 99% measure. But in a 99.9% measure,
number, I don’t ask how it is com- you would be excluding as many days as you had VaR
puted, I ask to see the backtest. If I breaks, making the number pretty useless; 99.97% is
entirely pointless.
think I could make money betting
New and Improved!
either side at 100 to 1 on whether A once-popular alternative to VaR, which has enjoyed a
or not a break will occur tomor- resurgence among academics but not practitioners, is
expected shortfall (ES) — defined as how far you expect to
row, I disregard the VaR. If anyone go beyond VaR, given that you exceed it. ES has some
desirable mathematical properties and tells you more about
argues with me, I challenge them to the tail than VaR. However, expected loss, conditional or
take my bet over the next year. unconditional, is not observable because there is always the
possibility of a small-probability, large-impact event that is
missing from historical data. Standard deviation, which
depends on the square of P&L movements, is even more
ask to see the backtest. If I think I could make money bet- sensitive to this problem.
ting either side at 100 to 1 on whether or not a break will Another alternative is to consider maximum loss within
occur tomorrow, I disregard the VaR. If anyone argues with the VaR interval. But this is hard to define with securities
me, I challenge them to take my bet over the next year. trading in different markets and time zones. You can esti-
Also, it’s essential that VaR is actually calculated before the mate it with a parametric model, but you have to trust your
start of trading. That means there will be errors: mis- model; you can’t prove the results with transactions.
booked trades, missing feeds, stale prices and other prob- Useful numbers are compromises between what we
lems. The noise from these errors has to be included in want to know and what we can measure well. VaR can be
VaR, because they are real risks. measured better than alternatives, but what does it tell us?
If you report a $50 million VaR, then revise it to $200 It’s not the worst-case loss: in fact, we expect to lose more
million after a $150 million loss, you did nothing useful than VaR two or three times a year. Sometimes people
and I don’t trust either the original or the revised number. I assume that a multiple of VaR, say three times VaR, is a
want to see a VaR backtest based on the actual values pub- good almost-worst-case number. This is a terrible assump-
lished before the start of the period they reference. tion on both theoretical and empirical grounds.
Hypothetical backtests based on scrubbed data are suspect, VaR is a tool of risk management, not risk measure-
and tests with after-the-fact corrections to the VaR for data ment. For one thing, putting in a VaR system always leads
or processing errors are worthless. to tremendous improvements in information systems and
Why not go deeper into the tails? Because you won’t communication within a firm and always turns up some
have enough data for convincing validation. A 99% one- important surprises. You could throw away the final num-
is Mediocristan, outside is Extremistan. Tools that work in capital equal to worst-case loss. Capital does not protect
one place are useless in the other. against all risks.
I’m not talking about the VaR you find in the “risk man- The classic example is more cash in the vault doesn’t pro-
agement” section of “management’s discussion and analy- tect against the danger of the vault being robbed. Physical
sis” of the financial statements. That considers only the robbery is not a significant risk for modern financial insti-
market risk of the firm’s trading and portfolio positions tutions, but there are other scenarios in which the more
and usually has an unimpressive backtest. Instead, I mean capital you have, the more capital can be taken away. What
the VaR concept applied to all firm risks: define the risk if the Justice Department goes after top managers, as hap-
precisely enough to specify an observable measure, calcu- pened to Drexel? What if the US government defaults on
late and backtest the measure, then use scenario and stress treasuries, as almost happened in 1995? What if Bear
testing to ensure foreseeable risks are within an order of Stearns had collapsed and frozen all financial institution
magnitude of the observable risk. transfers for a week?
Other risks can be measured in dollars, but plausibly can
scale larger than any reasonable level of capital. If a rogue
trader can cost Societe Generale $7 billion, why can’t the
I’ve argued that capital should be next one cost five or 10 times as much? The 9/11 terrorist
attack caused the stock market to close for a week during
risk-sensitive and that VaR is the which stocks fell 5%, and what financial institution can
hold enough capital to guard against markets being closed
appropriate first step to measure for a month while prices move 25% or 50%?
The simplest view of capital is to consider what happens
risk. Given the VaR, how much cap- if a firm gets into trouble and decides to liquidate. If it can
ital should a bank hold? No multi- sell its assets to repay its liabilities, it can have an orderly
liquidation, with all claimants paid in full. Therefore, capi-
ple of VaR is a safe worst-case loss tal is the excess in value of assets over liabilities, and the
capital requirement should be set such that there is very
estimate. But it’s silly to set capital low probability of a firm going from well capitalized to
equal to worst-case loss. Capital insolvent (negative capital) before a regulator can force liq-
uidation.
does not protect against all risks. Of course, this does not mean that the preferred reaction
to losses is liquidation. If you can liquidate and repay
everyone — that is, if you have positive equity value —
then you have other options, such as raising new capital or
Regulators were wise, not foolish, to allow firms to use selling yourself. If your assets are worth less than your lia-
internal models to estimate risk. The quality of the risk pre- bilities, you are no longer in control of your bank, and
dictions can be judged without looking at the internal some third parties are going to take losses.
workings (the regulators check under the hood as well). No This seems to be what David Einhorn is thinking of with
set of rules could possibly cover the complexities of a mod- his prudent reform proposal. Consistent with the liquida-
ern diversified financial institution, and if they could, they tion view, he wants to count only tangible equity as capital.
would be out of date before they were implemented. That assumes intangible assets are worthless in a liquida-
The regulators offer simpler fixed alternatives for risk- tion: 100% capital for illiquid assets assumes they cannot
challenged institutions, but the top financial institutions all be sold in a liquidation; 50% assumes a steep discount.
invested in Basel II “advanced” approaches. Basel II does An 8% to 12% write-down — about $30 billion to $45
not only require risk estimation for capital adequacy: mod- billion for Lehman — seems to be in the ballpark for the equi-
els and results are reviewed by supervisors and exposed to ty a prime broker would demand to finance the portfolio of
market participants. Lehman’s marketable assets. The collateralized lending assets
are considerably safer: 2% (or about $6 billion) for Lehman
How Much is Enough? seems like a conservative but not absurd potential loss figure.
I’ve argued that capital should be risk-sensitive and that If Lehman meets this standard one month, a regulator can
VaR is the appropriate first step to measure risk. Given the have an appropriate level of confidence that the next month’s
VaR, how much capital should a bank hold? No multiple report will show enough equity cushion for an orderly liqui-
of VaR is a safe worst-case loss estimate. But it’s silly to set dation with full payout to creditors.
✎ AARON BROWN is a risk manager at AQR Capital Management and the author of The Poker Face of Wall Street (John Wiley & Sons,
2006). He is also serves on the GRR editorial board and is the former executive director and head of credit risk architecture at Morgan Stanley.
He can be reached at Aaron.Brown@aqr.com.This article expresses the personal opinions of the author, which are not necessarily shared by
his employer or any other entity.