Comment columns are awash today with talk of Morgan Stanley’s trading prudence, as evidenced by its lower Q2 Value at Risk number.
For instance, here’s Graham Bowley at the New York Times:
. . . After years of chasing Goldman, [Morgan Stanley CEO John] Mack is charting a safer course, albeit a less profitable one. Morgan Stanley is emerging from its near-death experience last year as a more conservative operation, one that, unlike Goldman, is far less willing to take big risks in the markets.
That new conservatism was highlighted in the bank’s second-quarter results released on Wednesday. In it, the company’s trading VaR, a way of measuring the expected frequency of portfolio losses, is described as averaging $113m at the 95 per cent confidence level in the quarter. That’s up slightly from the $100m value in Q2 2008, but down from $115m in Q1 2009. Hence the notion that Morgan Stanley is de-risking.
Now VaR is clearly a flawed tool when it comes to risk management, but it is one of the quickest and easiest ways we have of gauging risk appetitite at the banks, so we won’t dispute the basic premise that Morgan Stanley appears to be trying to limit its trading risk. What we will dispute, however, is that the lower VaR value and de-risking has something to do with Morgan Stanley’s rather lacklustre set of results yesterday. From the the NYT’s Dealbook:
The results were in sharp contrast to rivals Goldman Sachs and JPMorgan Chase, which both reported strong second-quarter profits last week. Those two banks in particular have rebounded more quickly, mostly by taking on more risk in trading for themselves and their customers. But Morgan Stanley, which was burned more severely by the crisis, has moved to reduce its risk taking and try to build a stable, less volatile firm.
However, we suspect Morgan Stanley isn’t losing money because it’s not taking risks, but because, unlike banks like Goldman, it appears to have a tendency to lose money rather than gain it when fat tail events — outliers outside the normal distribution assumed in VaR – do occur.
This, for instance, is from a October FT story, when Morgan Stanley disclosed its rather massive Q3 trading losses:
Morgan Stanley disclosed that its quantitative strategies traders lost $390m in one day in August because of what the bank yesterday said was widespread selling by investors. . .
The disclosure came as Goldman Sachs said it lost more than $100m on six trading days over the quarter but earned more than $100m on 23 days. The information came from its quarterly filings with the Securities and Exchange Commission. Goldman said it lost money trading on 18 days in the quarter, nearly double the number of days in the second quarter.
Morgan Stanley lost money trading on 13 days, including four on which it lost more than $125m. Morgan Stanley made more than $125m on eight days.
Now, we don’t have much detail yet on Morgan’s trading profits & losses distribution for last year, but with Goldman’s 2008 distribution looking something like this, we think it’s safe to say that Morgan Stanley may be behind its competitors when it comes to the VaR money-making curve.
In short, it’s not the VaR in your jar, but the dispersion in your version.
Related links:
On Goldman’s fat tail risk – FT Alphaville