The defensive anomaly

Speaking of the brown stuff reminds me of fund managers. Here's another tale of their utter uselessness.
In my day job, I've been pointing out for years that low-risk stocks do better than they should. This chart shows my point for this year. It shows the performance of a basket of the 20 stocks with the lowest volatility of monthly returns in the previous five years; it's rebalanced every quarter.
Not only has this basket out-performed the market. In the last six months it's easily out-performed every fund in Trustnet's database of all companies unit trusts. In fact, similar baskets of defensives have out-performed almost all funds ever since 1999.
But here's the scandalous bit. The tendency for defensive stocks to do well is one of the oldest-attested quirks in the market. It was pointed out in this classic paper way back in 1972.
And yet, judging by performance, no unit trust manager seems to exploit this anomaly. Like our lame-brained politicians, they think their "judgment" is better than the evidence.

Comments

[Matthew]

What is the definition of a "defensive" stock?

[Piyush Pant]

Not sure you can draw this conclusion based on the performance of a 20 stock portfolio versus the 350 stock FTSE 350 index.
I remember some analysis quoted in Robert Hagstorm's book on Buffett where he found that there is a higher probability of beating the market with a fewer number of stocks - implying that if you randomly put together 20 stock portfolios based on arbitrary criteria, a higher number of them would outperform the market than a similar collection of portfolios based on 100 stocks each for example. Of course the returns would be more volatile but the point is that just because a portofolio of 20 stocks picked according to some criteria outperforms the market doesn't necessarily mean that there is there is any correlation with the criteria. It could simply be increased reward for the higher risk of a smaller number of stocks. It is quite possible that a collection of 25 stocks starting with the letter 'A' also outperforms the market.This may not totally go against what you're suggesting but it seems that you need to show more data to make the argument. I know that 15 stocks diversify away something like 80% of non market risk but that is an average figure. Hagstorm's numbers still show the increased probability of beating the market with fewer stocks in a portfolio. Unfortunately i can't find his research anywhere on the web but he's the manager for the legg mason growth trust ( same fund family as bill miller - who has beaten the s&p 15 years in a row !).

[Piyush Pant]

Read your original post once again - I hadn't picked up the fact that the portfolio is rebalanced according to the low volaility criteria every quarter. This makes my point much less valid.

[chris]

Piyush - in fact, the daily volatility of the portfolio in the table is lower than that of the market - an annualized 13.81 against 15.21.
As you say, we need more data. But I've got data since 1999 for a mean-variance optimal portfolio of FTSE 100 stocks which shows that low-risk outperforms. And the Jensen, Black and Scholes paper I cited shows big out-performance in 1926-66 in the US.
Matthew - the definition doesn't much matter, as it seems that various specifications out-perform. It's true of low-volatility stocks; low- beta stocks and (a slightly different thing), the lowest volatility portfolio on a mean-variance efficient frontier.

[dsquared]

really? Not one single one? So for example, you'd say that Neil Woodford at Perpetual doesn't? That's an awfully broad generalisation across fund managers, seemingly unsupported by evidence.

Unless you mean that there is no ETF-equivalent which takes this strategy as a quasi-index approach. In which case the reason is that quant funds don't sell.

[Piyush Pant]

Chris,
It would be really interesting to compare these results against random collections of 20 stocks. I have a suspicion that a similar argument could be made for stocks assembled according to weird criteria ( all stocks starting with 'A' for example).
I think the case you make is pretty strong but wouldn't it be interesting to see if there are any interesting ways of making up portfolios that leads to market outperformance on a long term basis ? Must be research on this somewhere...

[chris]

D2 - sorry about that. I was referring to unit trusts in the IMA's actively managed sector, rather than its equity income sector, where Woodford's high income fund is. According to Trustnet, his fund tops the table over the last 6 months with a 10.7% return. This is consistent with a defensive bias - as indeed, is his 5-yearly return. But it's well below the 16.6% return (admittedly before dealing costs) made by the notional low-risk portfolio.
There are around 350 funds in the equity income and active managed sectors. Even if you take "no fund manager" literally, this means my error is only 0.3%.
Piyush - yes, many random portfolios do out-perform, such as those based on letters of the alphabet. I've done stuff on this in the past, but it's all behind the IC's firewall. But the defensive anomaly lasts much longer than the out-performance of random portfolios.

[Piyush Pant]

Re: fund managers, abysmal though their general record is, given the choice between a 100% chance of underperforming the market by tracking a broad index ( due to management expenses and cost) and a miniscule but finite chance of outperforming in an actively managed fund, it is clear that a lot of people prefer the latter. There must be a term in behavioural finance to explain why people continue to have faith in highly unlikely but not impossible events ( like a fund manager outperforming) even when statistics clearly show this is unlikely. Unfortunately i don't know what that term is.

Obviously passive strategies like the one you outlined above pretty much guarantee outperformance but I suspect the reason people still flock to fund managers is the attraction of the 'fat pitch' - however improbable.

[chris]

Another thing is that people prefer to trust human beings than statistics - hence, as D2 says, the fact that quant funds don't sell.

[dsquared]

Dealing costs are a big factor here though; unit trusts commit to buy and sell units at the mid price more or less without limit, send you monthly statements and generally act as a savings product, which notional portfolios obviously don't. I'm not at all happy with comparing returns on a portfolio with returns on a financial services product - I can just about bear this when the comparison is an actual tracker like Vanguard or some such, but for a notional strategy, I am not so sure.