Owners of a range of large mutual funds are going to get a nasty surprise when they open their statements in fall of 2008: a range of large funds held big positions in some of the worst financial disaster stocks this year.
Fidelity Magellan loaded up on AIG as recently as June of 2008. This article reports that Fidelity Magellan had $475M in AIG at the end of May of 2008, and increased this position up to $865M in June. This made AIG the sixth largest holding in the fund. During June, the price of AIG was between $26 and $37 per share. Today (September 17, 2008), AIG is trading at less than $5 a share.
Needless to say, the very poor performance of these funds’ top holdings has led to substantial under-performance:
Both of these funds are down on the order of 20% in the last six months, when IVV (which tracks the S&P500) is down less than 10%. The under-performance is not due to the style orientation towards growth, either. The iShares Morningstar Large Cap Growth ETF (JKE) has performed just a hair better than the S&P500 over this period.
The presence of these concentrated positions at large mutual funds is an issue that investors need to be aware of—not because concentration is bad, but rather because concentrated positions have a large influence on portfolio performance. I have no beef with funds taking concentrated positions—there is considerable evidence that funds must take meaningful stakes in order to generate market-beating returns. There is an enormous difference, however, between a concentrated position in a stock like AIG and a concentrated position in a stock like JNJ. This is where risk management comes in. Anyone could have looked at a stock like AIG in June and seen that its price had dropped a lot. This made AIG a value buy only if the market had over-stated the implied loss of earning power or increase in default risk (or both). Sometimes a price decline signals a value—but sometimes it signals that the market is correctly evaluating a substantial increase in default risk.
The issue with these Fidelity funds having large stakes in these two stocks (in particular) has to do with risk management. Large financial firms are assumed to follow standards of practice in risk management. Fidelity states that it does just this. The opening sentence to their Risk Compendium [PDF file] is the following:
At Fidelity, we understand the importance of risk management; we have built a comprehensive risk management structure
Fair enough. The question that strikes me is where risk management was in the process of having two very large funds take on substantial concentrated positions in some very risky stocks.
To start this sort of analysis, we must consider two factors. First, were the risks associated with AIG and WB calculable ahead their massive declines? Second, were these risks (if they were calculable) consistent with the funds’ risk profiles? I am certain that there is a legal standard for approaching the second question, but I am not an attorney and my interest comes from the perspective of the interests of an investor.
I am going to start from the position that these Fidelity funds are supposed to be broadly diversified and thereby strongly constrain the total portfolio exposure to individual stocks. In particular, I suggest that one of the primary reasons that investors invest in funds is to mitigate default risk in any single holding. If the risks associated with AIG and WB were un-knowable ahead of time, we cannot really fault the portfolio managers for their large stakes in these funds. On the other hand, if the extremely high risks associated with stocks like AIG can be reasonably estimated ahead of time, it will behoove investors to look closely at the investment and risk management practices of the mutual funds in which they put their money. Further, this discussion can help to motivate investors and wealth managers to look closely at their own risk management practices.
For investors thinking about where to put their money, it is worthwhile to have at least a passing understanding of portfolio risk management practices. Risk management is all about calculating and managing the risk of loss. Professional portfolio managers typically have access to mathematical models that estimate the amount of money that they can lose in the worst case scenario. At some firms, they calculate the worst 1% losses projected one day out. The name for the amount of money that a portfolio can lose over some specific period is Value-at-Risk [VaR]. VaR and related metrics are standard tools at firms with trading operations (source1) (source2). Large trading operations generate daily VaR reports. Traders and portfolio managers are often given specified VaR limits, which tell them how much risk they are allowed to expose the parent firm to.
Risk management models are intended to estimate the exposure of a portfolio to losses. A portfolio of individual stocks (such as a mutual fund) is exposed to a range of risks, including market risk and company-specific risk. For an examination of concentrated positions in stocks like AIG, we need to look at both market risk and company specific risk. While most investors think of default risk and market-driven risk (volatility) as fundamentally different types of risk, the current evolution of risk modeling supports a convergence [PDF file]:
So we are still faced with challenges in the interactions of credit and market risk, but these are different challenges from what we anticipated ten years ago. The first challenge is in how we define risk, and has to do with breaking the artificial distinctions between credit and market risks.
In a recent article, I discussed the various tools that are available for estimating default risk. In brief, there are five broad categories of statistics that risk managers can use to estimate default risks in the stocks that they buy:
1) Credit ratings
2) Default models like Z-scores (and more recent brethren)
3) Implied volatility in options
4) Credit Default Swaps [CDS]
5) Forward-looking portfolio models
In practice, data from one or more of the first four feed into the portfolio risk management model—item 5. In the article linked above, I explain the relationships between the first four methods—and they are closely related. Implied volatility in options prices on a stock are highly correlated to where CDS trade on that stock. Credit ratings are partly developed using models and inputs that may be quite similar to Z-scores.
While there are different ways to calculate the extreme risks associated with a stock, I am going to use only one to provide examples here—a forward looking portfolio model. Quantext Portfolio Planner [QPP] is a portfolio management tool that projects tails risk for individual stocks. In past research, I have shown that QPP’s projected risks of default for individual stocks map well with implied volatility and to credit ratings derived from credit default swaps (source). I am going to look at the projected probability of default for some financial stocks (including AIG and WB) using QPP—and I am quite sure that a range of other models would have given somewhat consistent results, based on my earlier work.
Calculating Risk of Major Loss
Let us now move from theory to practice. I have used QPP to estimate the projected 1% tail loss at a time horizon of one year for a series of stocks at two different points in time (using data available up to these dates): 12/31/2007 and 6/1/2008. I chose the 6/1/2008 date simply because we know that Fidelity Magellan loaded up on AIG in June. I used the 12/31/2007 date to give a sense of the direction of risk levels. The projected 1% loss is what an asset is expected to meet or be below in the worst 1% of cases. These results are shown below, using all default settings (and QPP users can easily verify these numbers):
As of 6/1/2008, for example, QPP projected that the 1-year / 1% worst loss would meet or exceed a loss of -84% for Merrill (MER). This projected risk had grown dramatically since the end of 2007. Lehman (LEH) has a projected 1-year / 1% worst loss of -100%--also a substantial increase in risk from the end of 2007.
To provide a sense of context, we have previously mapped these projected tail losses to Market Implied Ratings [MIR] from Moody’s and shown that a projected tail risk of minus 50% or worse corresponds (on average) to stocks with sub-investment grade ratings from MIR (source). This is important because the MIRs are derived from the implied default risk in Credit Default Swaps [CDS]. This does not mean that these companies are bad companies—simply that they face high levels of uncertainty in future performance and subsequently elevated chances of extreme loss—up to, and including, default. These high risks are a function of both company specific risk factors and market risk. The results using data through 6/1/2008 show that AIG and WB looked quite risky—but far less so than Merrill (MER) or Lehman (LEH). What is most notable in looking at the projected risks is how much they increased from 12/31/2007 to 6/1/2008—and this is shown in the last column in the table above. These substantial increases in projected risks suggest that the market’s perception of the risks in these stocks increased dramatically in the first five months of 2008—and these increases in risk were not at all symmetric across this sample of stocks.
As a point of reference, I have calculated the projected tail losses for IYF, the iShares Dow Jones U.S. Financial Sector ETF. As of 12/31/2007, QPP projected a 1% 1-year loss of -29%. For the Year-To-Date (through September 16), IYF is down about 25%. A great deal of the substantially higher risks associated with the individual stocks in the list above can be diversified away—as shown by the fact that the projected loss potential for IYF is so much lower than the risks of the individual stocks.
In previous work, I have discussed the fact that QPP’s projected volatility was consistent with implied volatility. Further, I have shown that QPP’s projected tail risks map well into market-implied default risk. These connections are well known in operational risk management, so I would expect that many institutional-grade risk management platforms would have provided somewhat similar estimates of the substantial increase in distress risk for stocks like Merrill, Lehman, AIG, and Wachovia. How, then, do we explain the high exposure of some very large mutual funds to these stock tickers? Perhaps these funds believed that they held such well diversified portfolios that their large positions in these stocks were not, in the aggregate, significant. Perhaps these funds feel that large positions in stocks with high default risks are within their investment mandates. I have a hard time believing that the funds which took such large positions in individual stocks that the market was signaling were distressed were unaware of the risks—although we have seen some substantial failures in risk management in the past years.
The lesson for investors is not to avoid individual stocks—even risky ones. Risky assets, combined effectively, can result in portfolios with the highest possible risk-adjusted returns. The key is that investors need to understand the risk profiles of the stocks and funds in which they invest. If I invested in a portfolio intended to buy up low-priced distressed stocks, I would not be surprised if there were significant exposure to a company like AIG. From what I observe, however, I think that many investors do not understand the risk management practices of the portfolio managers in whose hands they put their money—and that is a recipe for problems. For mutual fund investors, the suggested solution is to dig deeper into the risk management protocols for the funds that they hold. Personally, I have been surprised that such large and (theoretically) diversified funds had taken on so much exposure to stocks with substantial default risks.