matjen wrote:Weird how you left out 2001/2002 and 2008. In 2008 LifeStrategy VSMGX was down 26.5%. QSPIX was down 5% (estimated to be fair). Moreover, you can't just paint the fund with the term "quant" and pretend it is like all other quant funds. It has a very different strategy than most.
1) There were significant drawdowns in 2001/2001 and 2008 for the balanced funds. But as an investor, I know that these funds are un-levered. These funds don't have any short positions. These funds won't go to 0. As the economy recovers, over time, my position will be in the black again. With levered long-short hedge funds, the wipe out to $0 is a very real possibility, and has happened plenty of times in the past.
2) The "strategy" is the issue. After 3 pages of discussions amongst possibly the best investors around, there is no clear idea on what exactly the "strategy" is except that the strategy is go long on assets that will go up and go short on assets that will go down - and implement that with leverage.
It is one thing to long-only tilt towards value and small (the penalty for being wrong there is not catastrophic). It is quite another thing to put 10% or 20% of your assets in a levered long-short quant fund. You are totally at risk of the (data mined) models working in your favor.
If you're wondering about the theoretical justification for security selection and the styles, it's all in the literature and has largely been implemented in real funds for decades. For example, 1/4 of it is value investing. Surely we don't need to go back all the way to Graham/Dodd or more closely Fama/French?
Long/short implementations of styles go way back to Fama/French and before. Recall that the decomposition of returns is by market minus risk free, small minus big, high minus low. Small minus big means long small stocks, short large stocks. High minus low is long value, short growth. You can look at Kenneth French's dataset for how diversified baskets of these long/short strategies would have performed in equities (before trading costs/implementation issues, which is the huge caveat), over the decades.
In any case, there are real risks from the long/short implementation. The upside is actually removing the market beta and having much lower correlation. With long-only tilts you still get much of the same behavior as the market itself. With the long/short you can use a much smaller amount and have something to rebalance with. For example, instead of 40% market equities, 20% tilted equities, 40% bonds, then something like 57% market equities, 5% long/short styles, 38% bonds is possible probably with higher Sharpe. The fund is not going to blow up from a few short positions going sour, and even if it did, it's only that 5% of the portfolio or however much you have to allocate to it to get the desired exposure.
And this fund in particular is long/short in much more than just equities and in multiple styles. Also, the amount of leverage is cause for concern, but it's targeting levels and volatility far below the spectacular failures of the past. I fully expect more quant meltdowns in the future and models to go wrong, but everything is a matter of degree. More leverage is used on less volatile assets than on volatile ones like stocks, so not all leverage is created equal and is equally dangerous.
As a minor curiosity, Vanguard's market neutral fund appears to be over 4000% short and 4000% long in individual stocks. Now that has the potential to blow up when bets are wrong.
(as an additional aside, the 2008/2009 loss of 20% or more was with the old management team... of course)
edit: checking the Vanguard fund's filings, I think it's more likely that Morningstar is confused than that it's really that leveraged.
matjen wrote:This is crazy talk. QSPIX haters can't argue about the theories and why QSPIX may be useful in a factor-based approach, they can't argue about the actual real world performance, so they are left with pulling boilerplate language out of the prospectus and speculating about end of the world scenarios. Laughable. For those interested in how QSPIX may have performed in 2008 and 2002 I present Ronen Israel. AQR examined all of this. And yes, of course, it is back tested. That is all you can do but it is a a heck of a lot better than rank speculation just to spread FUD...fear, uncertainty, and doubt.
Every strategy in the past that failed, at first, backtested well. That's how people come up with new strategies... They back-test 100 strategies, 99 fail, 1 appears to do really well in the past... They start using that, showing doubters how well it back-tested.
But the next crisis never looks like the last 4 crises (I had to look up the plural of crisis), and the strategy implodes.
I agree in principle, but the core strategies here are nothing new, just a combination of old existing styles across as many asset classes as possible. There are always data mining and out-of-sample concerns, and I have them too here. But unless you would take Bogle's stance on say value not working in the long run, there's not much to dispute at a high level. The empirical evidence for the other styles may actually be stronger.
So what's "new" here is just combining all these things together, targeting certain risk levels (i.e. adjusting leverage amount) for each style/asset combination for a total target standard deviation of 10%. In practice, this will not work as well as it did in the backtest if
(1) a style has greater volatility or lower returns than in the past (i.e. lower Sharpe)
(2) the correlation between styles goes up
I think (2) is a relatively remote concern except possibly in a crisis type of scenario, and probably not even then. For example, there is a lot of evidence for value and momentum having negative correlation. An argument against the fund would more have to be (1), in which case you're arguing against a whole lot more than just AQR regarding the way the world (probably? usually?) works. Again, I think these concerns are justified, but representing this as "com[ing] up with new strategies" is not very fair.
Again, it should be pointed out that the raw returns using the most basic definitions of the styles, e.g. price/book for value and restricting only to more liquid stocks (i.e. not getting value or momentum in small stocks), were on the order of 17.4% over the risk-free rate for 23-24 years. Clearly, transaction costs eat into the actual returns, and these four styles were chosen (and not others) because these happened to work well over the period. But I'm not sure you can call data mining all the way to this thing not earning money in practice.
Taylor Larimore wrote:
What are the reasons given? Futures are just financial tools that can be used for a variety of purposes. I don't think it's very useful to just cite an appeal to an authority out of context.
In her acclaimed book, Making the Most of Your Money
, Jane Bryant Quinn devotes six pages
to commodities Futures which she calls "A Loser's Game." I recommend her book to anyone who thinks they can win with Futures (except brokers who sell them).
Taylor, I will have to look for this book then, thanks.
For the time being, I will also just note that the fund we're talking about has long and short positions in futures contracts in much more than commodities.
But for commodities futures in general, that's pretty much the only way any fund has exposure to commodities. The vast majority (perhaps all but certain gold funds, if you count that) roll futures contracts and do not attempt the cost of storing commodities themselves. So a commodity fund, as used by say Blackrock's LifePath target date series, as recommended by Larry Swedroe in some posts here and books, as described in sample allocations by Christine Benz at Morningstar:http://news.morningstar.com/articlenet/ ... ?id=672743
would be long commodities futures. Leveraged discretionary speculation is almost always a bad idea. Any and all uses have their own risks, like any other investment.