Study finds links between portfolios' profiles and future performance

(Published in the South China Morning Post on 19 Sep 2010)

Is it possible to find a simple measure that identifies skilful active managers and thus funds that will produce good returns?

This is what the information ratio – a measure of risk-adjusted return – was meant to do, but several studies have shown there is no relationship between it and future fund performance.

However, research by Martijn Cremers and Antti Petajisto of the Yale School of Management, recently endorsed by Morningstar, finds a strong link between active share – the extent to which a fund’s holdings differ from its benchmark’s – and performance.

Unlike the information ratio, which measures performance per unit of volatility, Cremers and Petajisto’s research looks at what is inside portfolios, then finds robust links between portfolios’ profiles and future performance.

But what is active management?

Commonly it is understood to refer to the level of activity in a portfolio; the more transactions, the more active the manager. Even Nobel laureate William F. Sharpe defined it such. But were we to use this definition, then some of the world’s most successful investors would be considered more passive than active. Warren Buffett, Philip Fisher, Peter Lynch and many like them have produced great results by buying and holding, but are considered inactive by the common standard.

Successful investing arguably is about being different from the herd, or rather about finding the right opportunities to be different. Unless you are a high-frequency quant trader like Jim Simons at Renaissance, producing superior returns does not necessitate lots of portfolio activity.

Buffett, Fisher and Lynch all understood that one couldn’t predict share prices over short periods – at least not by using fundamental analysis – but that one could – sometimes – predict business performance over long ones. And, if you could do that, then you could predict share prices over long time frames.

Furthermore, they all appreciated the principle of putting your money where your mouth is, believing that, as Buffett put it, “wide diversification is for people who don’t know what they’re doing”.

A second misconception relates to risk, the generally accepted measure of which is volatility. But does volatility really measure risk?

We think important risk relates to permanent loss of capital such as would be incurred by the bankruptcy of a company or the gradual deterioration of a company’s business rather than normal month-to-month undulations of share prices that should be viewed as temporary losses – and gains – of capital. But there is no neat measure of the potential for permanent loss of capital so we are left with one that is easy to calculate but which is also deeply misleading.

Good performance involves putting yourself in a position to produce good performance.

And while it would be nice to construct a fund that returned 2 per cent per month, month in month out, the reality is that the great long-term performers like Berkshire Hathaway come with a cost, namely high levels of short-term volatility, at least relative to the relevant index.

But this is precisely my point: short-term volatility is not a risk to be avoided but simply the cost of good long-term performance. And once one appreciates the power of compounding – the difference say, between 8 per cent per annum and 10 per cent per annum over 20 years – one can appreciate this is in fact a small cost.

Which brings me back to active share. Cremers and Petajisto looked at 2,647 funds from 1990 to 2003. They categorised funds according to both their tracking error and their active share. It should be noted that while there is clearly a link between tracking error and active share, both measure different things.

Tracking error measures a fund’s net exposure to systemic factors such as sector or size while active share is more a measure of non-systemic – or stock-specific – factors.

Thus funds that have high tracking error but low active share are “factor bet” funds and funds that have high active share but low tracking error are “diversified stock pickers”. Funds with both high active share and high tracking error are concentrated stock pickers”.

The two researchers divided the 2,647 funds into quintiles by both active share and tracking error, then looked at the performance relative to benchmark of each of the 25 groups.

What they found was that funds with low active share underperformed by around 1 per cent per annum regardless of whether they had low tracking error (benchmark huggers) or high tracking error (factor bets).

Diversified stock pickers produced good gross returns but fees wiped them out. But funds with high active share and tracking error that was not excessive produced excellent returns net of fees.

These results have huge implications. If investors know their fund’s holdings and those in its benchmark, they can tell whether their fund is managed properly or by a closet indexer claiming to be active and thus charging for a service he is not providing.

Vanguard’s John Bogle built his multibillion-dollar passive empire by pointing out that the average active fund performed poorly. However, Cremers and Petajisto’s research makes it clear that one should not tar all actives with the same brush. Bogle’s point may be correct, but when was anyone interested in the average runner in a marathon?