Stansky’s Monster: A Critical Examination of Fidelity Magellan’s Frankenfund
By: Ross Miller, State University of New York (SUNY), Albany
Published: February 2007
Yesterday’s posting discussed Fidelity’s desire to add performance fees to its lineup in order to address recent underperformance. Today, we cover one example of said “underperformance”: Magellan’s hard luck since the turn of the century. Fidelity’s Magellan is commonly cited for its extremely high r-squared to equity markets. In other words, it’s been an ETF in disguise.
So who better to tell the story than Professor Ross Miller of SUNY Albany – a man whose paper on index hugging is in the Portable Alpha Hall of Fame. Miller released this ringing indictment of Magellan last month.
It starts with an entertaining and easy-to-read history of the fund and its colourful managers – from Peter Lynch’s “ten baggers” (some sourced from his wife’s grocery preferences) to Jeffrey Vinik’s prematurely defensive positioning in 1996 to Robert Stansky’s closet indexing of the late 90’s and early part of this decade.
But then Miller throws off the gloves:
“In the cinematic adaptations of Mary Shelly’s classic horror novel, Dr. Victor Frankenstein pieces his monster together from body parts obtained from several corpses and reanimates it using the high technology of the early 19th century, electricity. Things go badly for Frankenstein’s creation and mayhem ensues.
“Robert Stansky constructed Magellan’s portfolio from holdings of 200 or more stocks. Over time, Magellan’s portfolio became more than a collection of favored stocks, it had become a synergistic entity that took on a life of its own. Stansky had created what might best be termed a frankenfund that would ultimately destroy billions of dollars of its investor’s wealth. Fortunately for Mr. Stansky, torch-wielding fund holders never ran him out of Boston. Magellan did, however, experience massive withdrawals and was ousted from many defined contribution plans, a major key to Fidelity’s pre-eminence in the mutual fund industry.”
“Torch wielding fund holders” may not have run Stansky out of Boston, but Miller employs his trademark alpha/beta thinking to show why perhaps they should have. In doing so, Miller illustrates how a forensic investigation of track records can possibly reveal hidden trading strategies.
But rather than simply analyzing monthly returns, Miller digs into the daily results of Magellan to determine the correlation between the fund and its S&P 500 benchmark. The results are astounding. Miller says the r-squared of daily returns went above 0.95 almost immediately after Stansky assumed management responsibilities and dropped off almost immediately after he left the role in 2005.
How can this be when it would appear from the fund’s annual reports that it was making bets that were materially different from the market? Miller hypothesizes that Fidelity was employing the latest portfolio optimization software to essentially re-create the index without actually investing in it – or even in the same proportions as the index. (As an aside, tracking the index without literally owning it has implications for previous research that gives weight to both track record and actual holdings when determining closet-indexing.) This portfolio optimization software, muses Miller, may have even sowed the seeds for Magellan’s eventual underperformance.
“Normally, portfolio-optimization software provides benefits because it is able to eliminate inefficiencies that even the most diligent portfolio manager would never detect. Robert Stansky, possibly employing optimization technology proprietary to Fidelity, managed to create a fund that was nominally managed in an active manner and yet never strayed far from the S&P 500. Could it be that what Robert Stansky did to keep Magellan on track was linked to the fund’s underperformance?”
Using a methodology Miller first proposed in 2005 (and contained in his Hall of Fame paper), he shows that a 0.99 r-squared translates into an “active share” of about 9% – meaning the fund could be re-constructed by putting 91% in an ETF and investing 9% actively (for example, in a hedge fund). He explains this approach the following way:
“The derivation of the active share is based on the observation that any mutual fund that maintains a consistent investment style can be viewed as bundling together a passive leveraged investment in the benchmark with a purely active investment that is statistically uncorrelated with that benchmark. When the benchmark is a tradable market index, such as the S&P 500, the active part of the mutual fund is equivalent to a market-neutral hedge fund.” [our emphasis]
In fact, as we have advocated on this website, he brands the “active” part “Magellan’s Inner Hedge Fund” (we have used the term “embedded hedge fund”). Here’s how Magellan’s inner hedge fund performed over the 2002-2004 period:
To be fair, with an 8.85% “active share”, Magellan was one of the most extreme examples of a closet-indexer on the Fidelity shelf. Overall, Fidelity funds ranged from 7.85% to 37.53% active share during the period of the study. But Magellan’s active share was dramatically lower than those of its large-cap mega-fund peers. The remainder of the six largest “large-cap blend” funds in the US had active shares between 17.72% and 28.62%. In other words, ceteris paribus, Magellan’s fee-for-active share was 2 to 3 times higher than its closest competitors.
With apologies to big green men with flat heads and limited verbal skills, this paper is a great case study in the use of regression analysis to study what we at AllAboutAlpha have sometimes called the mutual fund “genome”.