Goldman, liquidity and VAR

July 16, 2009

Goldman Sachs’ second-quarter earnings release showed a continued increase in the amount of market risk held on the firm’s trading book. Its risk appetite has continued to expand at a time when extreme turbulence has forced others to scale back.

True, Goldman’s publicly reported figures may overstate its actual positions. But the Wall Street bank also appears to be taking advantage of its access to liquidity from the Federal Reserve to increase risk.

Total value-at-risk (VAR) averaged $344 million, on a gross basis before diversification effects, up from $303 million in the second quarter of 2008 and $226 million in the second quarter of 2007.  picture-1

Click chart to enlarge in new window.  Additional data available here.

(VAR is a crude measure of the worst loss the firm would expect to report on 19 days out of 20, given prevailing volatility in the market.)

Like other banks, Goldman reports VAR on a net basis after taking account of a “diversification effect”. The diversification effect reflects the fact that risks in different parts of Goldman’s book are not perfectly correlated.

The bank would not expect to lose the maximum amount on all its positions at the same time. So net VAR for the book as a whole is less than the sum of the VARs for the individual components (which Goldman reports as currencies, interest rates, equities, and commodities). Goldman’s net VAR in the second quarter averaged $245 million, up from $184 million in the second quarter of 2008 and $133 million in the second quarter of 2007.

How Much Risk?

Banks understandably prefer to focus on the smaller net figure, but when looking at the amount of market risk on a bank’s book, gross VAR is arguably more useful. The whole point of a crisis is that when it hits, contagion ensures that previously uncorrelated asset classes move in the same direction, and the diversification effect disappears. So it is dangerous to rely too much on the diversification effect to reduce overall risk.

Interestingly, Goldman has been forced to reduce its reported diversification effect from $119 million to $99 million over the last twelve months as correlations have increased, reducing the benefit it receives from diversifying its portfolio, even though the overall trading book appears to have grown.

But whether we use net or gross VAR, Goldman’s risk taking has risen by 50-80 percent over the last two years. Even during the last 12 months, as the financial system has suffered its worst crisis since the 1930s, net VAR is up by 33 percent while the gross figure has risen 14 percent.

Adding Risk in Rates

The additional risk has been added very selectively. All the additional VAR is reported in the firm’s interest rate sector, where the average daily VAR has risen $61 million (42 percent) from $144 million to $205 million. For other components (currencies, equities and commodities) average daily VAR is mostly flat or lower over the last year.

The question is why Goldman has continued to grow its risk-taking even as others have found it prudent to reduce market risk, and why all the extra risk has been added in the fixed income area, when the firm has held the line on risk-taking in other asset classes?

Answers must remain speculative because the firm reports only the minimum detail on average daily VAR by categories required by the Securities and Exchange Commission in its earnings statements and quarterly 10-Q filings.

In the past, though, most changes in Goldman’s VAR have appeared at least partly endogenous. In other words, when the underlying volatility in an asset class has increased, Goldman has preferred to “accommodate” it by allowing traders to continue running positions of roughly the same size, even if total risk is higher, rather than forcing them to cut back positions to keep the VAR level the same.

There is some evidence that the increase in rates VAR was at least partly endogenous in this case. Volatility was especially high in the rate sector during the first two quarters of 2009, as the market struggled to balance hopes of recovery and additional stimulus against fears for continuing recession or an early tightening of monetary policy.

Volatility was far higher than in other asset classes. Goldman may have followed past practice and decided to “accommodate” it rather than push back — especially if the firm concluded its activity in the rates segment was profitable, and that forcing position cuts would weaken its ability to provide liquidity to its customers.

Risk – but Over What Horizon?

That still leaves the question why Goldman felt comfortable allowing VAR to rise so much. One explanation is that the firm is comfortable it can rely on the Greenspan/Bernanke put, now enhanced by its access to the Fed’s discount window as a member of the Federal Reserve System, to limit downside risks in the event of a crisis.

In effect, the put allows the firm to ignore the worst “tail risks”. The firm can ramp up risk-taking confident in the knowledge it will be shielded from the very worst outcomes by the government.

The other explanation has to do the limitations of VAR, and particularly what time horizon to use when measuring volatility.

Like its peers, Goldman publishes VAR on a one-day basis, reflecting price changes from one evening’s close to the next.

But there is nothing special about a one-day horizon (and in some ways it is a very unrealistic measure of risk since a large financial institution would find it impossible to exit all its positions over such a short period).

In practice, banks calculate a whole series of VARs for different time horizons, though only the one-day figures are
currently disclosed. VARs for longer horizons (one week, 10 days, one month etc) may paint a very different picture of risk in a trading book. When markets exhibit strong trending behaviour, with many small daily changes cumulating in a consistent direction, VAR will be higher when evaluated over longer horizons. But when markets are choppy and directionless, with large one-day changes frequently reversed the following session, VAR will be higher over a shorter horizon.

The first quarter saw precisely these choppy trading conditions in the interest rate sector. The volatility in asset price changes was much smaller when measured over a five-day
period than over the course of single trading session.

Benefit of Unlimited Liquidity

In the circumstances, Goldman’s risk-managers may have decided that the one-day VAR overstated the true risk of loss in the firm’s book and decided to focus on VAR levels evaluated over longer periods, which showed less build-up of risk.

That approach would be sensible for assets the firm intended to hold for more than a single day. The only reason to
evaluate VAR over very short periods (daily, hourly) is if liquidity becomes an issue and the firm was not sure it could absorb large one-day profit and loss (P&L) swings and meet margin calls.

But once the firm was sure of unlimited liquidity from the central bank, it could afford to “look through” the one-day P&L swings, holding loss-making positions and waiting for the market to reverse them in coming days.

In one sense, Goldman’s access to unlimited Fed liquidity increased its capacity to absorb short term volatility and conferred a competitive advantage over other institutions, such as hedge funds, that do not have the same access to central bank funding. Armed with a Fed credit line, Goldman understood the risks in its book were smaller than the crude one-day VAR indicated, and could increase apparent risk-taking on this measure without any increase in the real danger to the firm.

The moral is that the basic one-day VAR figures being disclosed by Goldman Sachs and other financial firms in their SEC filings provide a very limited — and potentially misleading — indication of the true amount of risk they are running.

In a world of limited liquidity, VAR measures may substantially understate the true risk. But once central banks step in as market-makers of last resort and guarantee firms against failure arising from liquidity rather than solvency, the one-day VAR measure probably overstates the degree of risk and banks are comfortable ramping it up.


Comments are closed.