Crystal ball discovered? New model forecasts manager success

Risk Management 15 Apr 2007

Active managers – particularly hedge fund managers – are notoriously inconsistent. This fact is often held up as proof of efficient markets. The assumption: all managers eventually “revert to the mean”, “hot hands don’t last”, and “what goes up must come down.”

As a result, endless resources have been poured into the quest to find some way to predict manager performance other than to simply rely on historical returns. Naturally, such a finding could be lucrative for advisers – particularly those who benefit from the performance of other managers like, say, funds of hedge funds.

The latest attempt to do this has some intriguing possibilities. In a paper published in this month’s Journal of Finance, Marcin Kacperczyk of the University of British Columbia and Amit Seru of the University of Michigan propose a new metric called the “Reliance of Public Information” (RPI) to measure the extent to which a manager’s performance is correlated – not with the markets – but with “public information” (captured by consensus sell-side analyst recommendations).

In a sense, the researchers are assuming the ultimate passive portfolio is not the market, as CAPM suggests, or based on fundamentals as Rob Arnott would suggest, but rather on the predictions given by sell-side analysts. This alone is an interesting take on alpha. Kacperczyk and Seru are saying that alpha is actually the amount of return generating by swimming against analyst picks (or at least orthogonally to them), not against the market itself. Using analyst picks to represent “public information” raises fundamental questions about market efficiency. After all, markets themselves represent all “public information”. So if analyst picks are a proxy for “public information”, then analyst picks should be “neutral” across the board. The sell-side would essentially be saying, “Dear Investor, As usual, the price of this stock reflects everything we know about it. It’s a fair price.”

(Regular readers of AllAboutAlpha might remember a paper published back in the fall by a couple of profs from Yale. In that paper, Martijn Cremers and Antti Petajisto proposed another measure of active-management – one that was based on the correlation of portfolio weights to market weights, not solely on market correlation. Kacperczyk and Seru take a similar tack. But instead of comparing portfolio weights to market weights, they compare portfolio weights to the weights a manager would have if she simply followed sell-side advice to a tee.)

But leaving this question aside, these academics conclude that manager returns are negatively correlated with their reliance on public information (their “RPI”). This makes intuitive sense to value investors (and raises fundamental questions about the value of sell-side research). They regress stock weightings against previous (consensus) analyst recommendations. If the portfolio weights of all stocks are exactly what previous analyst picks would suggest, then the manager is deemed to rely entirely on “public information”. If those weights can’t be explained at all by the analyst recommendations, the manager is deemed not to rely at all on “public information”.

They describe the construction of the RPI as a two-step process:

“In the first step, we find how much of the average percentage changes in a fund’s quarterly holdings can be attributed to changes in analysts’ recommendations…In the second step, we construct the measure of reliance on public information for fund m at time t…In simple terms, RPI equals the unadjusted r-squared of regression (…of percentage change in stock split-adjusted holdings…(against) a change in the recommendation of the consensus forecast…)”.

Their central conclusion:

“…we find that mutual funds with lower RPI, that is, those relying less on information in the public domain, tend to exhibit higher returns adjusted for commonly used risk/style factors such as market, size, value, and momentum.”

Managers with a hot hand often attract assets. Let’s face it, in aggregate, investors are “performance chasers”. But can we predict which managers will raise the most assets regardless of recent performance? Apparently, the answer is “yes”.

“…controlling for past fund performance and other fund-specific characteristics, funds with low RPI and rewarded with higher money flows, suggesting that outside investors learn about managerial skills from RPI and allocate their wealth accordingly.”

(We assume the author’s aren’t actually giving investors the credit for explicitly identifying a manager’s RPI, but rather that they implicitly consider “RPI-like” factors when allocating capital).

The study also finds that smaller funds tend to rely more on public information than do larger funds. It posits that this is due to either a) less skilled managers at smaller funds, or b) the inability for larger funds to easily and quickly adjust weightings in response to new analyst picks. Conspicuously absent from this list is an apparent no-brainer: smaller funds can’t do their own research and are barraged by the sell-side all day long.

In a rare example of justice in asset management pricing, funds that rely more heavily on publicly-available information also tend to have lower fees. Also interesting is the fact that older funds have a lower RPI, meaning that newer funds tend to rely more heavily on public information.

While the study finds that a fund’s reliance on public information is the result of fund-specific factors such as age and size, it also identifies manager-specific factors. In addition, RPI seems to fall, on average, when there is manager turnover. In other words, incoming new managers tend to stray from “public information”. This makes intuitive sense when you consider that a new manager is likely chomping at the bit to put their stamp on a fund regardless of the views of the market at that time. This is a phenomenon also identified in Professor Ross Miller’s paper on manager-changes at Fidelity’s Magellan.

But what if a manager’s portfolio weights aren’t correlated with analyst recommendations simply because the fund is passive, not active? Or worst yet, what if the portfolio weights are uncorrelated with recommendations because the manager is a chimpanzee throwing darts at the stock pages? The authors confront this possibility, but say that, while uncorrelated with analyst picks, these funds should not produce outsized returns:

“RPI might be low for reasons not necessarily related to the use of private information. For example, if relying on information in the public domain were observable and penalized, managers with no private information would have incentives to follow investment strategies based on noise. Such investors, though unskilled, would also exhibit low RPI. Another possibility is that RPI could be low if investors followed passive strategies without considering any information. However, although managers following any of these alternatives would have low RPI, they would be unable to deliver abnormal performance, as measured by traditional measures.”

In conclusion, the authors suggest that RPI might be a very useful way to determine manager skill level. In an approach not that different from setting a benchmark hurdle rate for a manager’s bonus, the authors suggest that managers should not benefit from simply baking analyst picks into their portfolios as soon as they are published.

Notwithstanding the leap of faith taken by imputing sell-side recommendations with all “public information”, this paper is a must read. But don’t take our word for it – it’s the lead article in this month’s Journal of Finance.

– Alpha Male

Be Sociable, Share!

Leave A Reply

← The Value Premium: value pricing, or just a flat tire? Schneeweis on Alpha, the SEC and Hedge Fund Replication →