Rob Arnott: What I'd like to do is to talk about "Smart Beta, Monkeys and Upside Down Strategies."
Monkeys like to hang out upside-down, so why shouldn't we turn our smart beta strategies upside down and see how they do.
CAPM, capital asset pricing model, has evolved over the years from the use of one factor into many factors. The notion of a market premium, the market return plus, now, priced factors, where you get a reward for value tilt, a reward for size tilt, perhaps a reward for low volatility, or a liquidity bias, where you get rewarded for accepting illiquidity, or a momentum premium.
In other words, cap-weighting has evolved into a whole array of smart betas, and the way this works is that we used to have a world of passive strategies and active strategies. And smart beta has come onto the scene, creating a whole wealth of strategies that the indexes powerfully resent being called indexes, but the Oxford unabridged doesn't specify cap-weight anywhere in its definition of index, and index is a measure of most anything.
Strategies that are, in effect, quantitatively defined, that are rigorously defined, that will target market exposure, that will typically have relatively low turnover, broad market coverage, can allow us to have low fees and expenses, transparency, and seeks to take from the active management community its raison d'être, which is the effort, the quest, for outperformance.
Now the big difference is with active strategies, you're operating in the same universe as the passive arena. The passive arena replicates the behavior of the market, and so it produces market returns. Strip the indexes out of the picture, and you're left with the playground for the active managers. It's the same list, with the same weights. Collectively they must produce market returns, less their trading costs and less their fees. And because their trading costs are higher and their fees are higher, collectively they cannot, cannot win.
And so then the question is, why should smart beta do any better? The answer to that would appear to be that there are some systematic biases among most active managers. They seek comfort. They seek to pursue growth. They seek to pursue assets that are popular, and a strategy that doesn't do that can capture structural alphas.
So, there are many offerings and many marketing claims in the smart beta arena. How much better are these strategies than market cap? What similarities can we identify? What are the critical differences between these various strategies? And then how do we use smart beta strategies. And this is all very important to the ETF arena, because ETFs are a natural vehicle for bringing these ideas to market. What could be easier than taking an alternative index, building a daily traded strategy out of a publicly available liquid index that just doesn't happen to be cap-weighted? It's marvelous for the ETF arena.
So, let's take high-risk strategies as one example. The claim: Investors are compensated for taking risk. The equity risk premium is called a risk premium, because you're taking risk. So why not take more risk? Why not have a risk-weighted strategy that loads up on high-beta strategies, high-beta stocks, on high-volatility stocks, or on stocks with high downside semi-deviation--the companies that have the biggest downside risk--because that's the really frightening risk.
And what do we find? Boy, it works.
Cap-weighting going back half a century would've given us almost 10% a year, 15% volatility. The volatility-weighted, beta-weighted, downside semi-deviation weighted, they all add about 2%-2.5% per year for a half a century. Well, there are active managers who would happily shoot their grandmother to be able to add 2.5% a year. And the notion of being able to do so with a mechanistic strategy that simply loads up on beta, or loads up on volatility--my goodness, that's amazing. But of course, you're taking on more risk, and so your Sharpe ratio doesn't go up as much as the return, but it goes up. You're getting more return per unit of risk by loading up on risk. So, this thesis that going for the risk must work, obviously makes sense.
Or maybe not.
Let's take each of these strategies and turn it upside down. Let's take volatility-weighting. Take its largest holding, and make it its smallest holding. Take its smallest investment, and make it its largest investment. Simply totally reverse the portfolio. And what do we get? My goodness, we get a higher return.
What about beta weighting? Load up on low beta. We get a jump of another 160 basis points above beta-weighting, and of course, by loading up on lower-volatility companies, the risk comes down, and now the Sharpe ratio just goes to the moon.
So, if boosting risk seems to work consonant with the theory that you're getting a risk premium (so why not ramp up the risk premium?), how can lowering risk work even better?
We've got a bit of a mystery here. Well, let's look at this globally. Maybe it's an anomaly in the U.S. No, it's not. When we look at cap-weighting globally and here the data goes back not as far for global, inclusive of emerging markets, we only go back about 22 years, but we've got a 7% return down at the bottom, and you ramp up your volatility and what do you find? You find a higher return for all but one of them. You find higher risk for all of them. You flip it, and you load up on the low-volatility, low-beta, low-downside semi-deviation portfolio assets, and you wind up with an even higher return with lower risk. And again an information ratio that goes to the moon. This is obviously a global peculiarity.
What about fundamentals? The claim: Why don't we load up on companies with great fundamentals? Because you don't want companies with lousy fundamentals, and there are those who claim--we don't--that fundamental indexes should work because you're loading up on companies that are big fundamentally, and the prices should be catching up to the fundamentals. That's not our claim for why it works. But there are those who make that claim.
So let's weight companies by book value, or by five-year average earnings, or collectively by fundamental weight, or a totally different fundamental-based index that's got nothing to do with fundamental index. Earnings growth, let's weight based on five-year earnings growth--load up on the companies that look fabulous for five-year earnings growth. They all work. They all beat the market. Boy, beating the market is sure looking pretty easy, isn't it? 2%-3% per year compounded for half a century. All except for earnings growth, you're winding up with similar risks to the market. Inclusive of earnings growth, you're winding up with higher Sharpe ratios. That's marvelous. Boy, this is getting easier and easier.
Let's hang these strategies upside down. Instead of loading up on the companies with the biggest book value weighted in proportion to the book value, let's flip it. Let's put the most money in the company with the tiniest book value, and the company with the biggest book value in the market, let's give it the smallest weight in the portfolio. Let's do the same with five-year earnings weight, with fundamental index weight, with earnings growth weight; they all beat the market. They all do so with much higher risk, but they all beat the market with a Sharpe ratio that, in three out of four cases, is even better than their starting portfolio. Boy, for the guy who came up with fundamental index, this is jarring. [With the] fundamental index, you're adding 2% a year for half a century. Hallelujah! Flip it upside down, and you're adding 4% a year for half a century.
Hey, anyone who want to come license upside down fundamental index from us? Happy to provide it. ExxonMobil is your smallest holding. Joe's Bar and Grill in Butte, Mont., is your largest holding. Joe will be very happy, very happy if you adopt this index, and if you create an ETF based on it. But what we find is that when you turn the strategies upside down, they all work beautifully.
Global results, what do we find? They all work. They all beat the cap-weight: 7% for cap-weight, anywhere from not quite 9% to over 11% for the various fundamentally related, indexing approaches. Turn them upside down, and all except for earnings growth, they work better than they did in their original form. Loading up on companies with horrible earnings growth does not seem to do a particularly nice job for you. That's the only exception.
What about the commercially available smart beta strategies? You've got diversity-weighting, which is really a blend between cap-weight and equal-weight. You dig through the numbers, you dig through the mathematics, the patent, the journal articles, and so forth for diversity weighting--all have endless mathematics--but you peel back the mathematics, all it is kind of half equal-weight, half cap-weight. The mathematics allows them to charge almost an active fee for it. If you did half equal-weight, have cap-weight, you couldn't charge much. But anyway, it's a neat idea, and it works beautifully.
Fundamental index, maximum diversification, minimum variance, risk cluster, equal weight, risk efficient portfolios--these of all been developed into product. They're all readily available. Most of them are available in ETF form. They all work, and they all work very nicely. Incidentally speaking, personally, I look on this not as competition, but as kind of a ratification of non-cap indexing. The diversity weighting, equal weighting, minimum variance. They existed long before fundamental index, but equal weighting, which came out in the '80s, minimum variance, which came out in the '70s, they didn't gain traction until the last half dozen years. So I feel kind of like a proud godfather, helping to open the door for these ideas to gain traction. It's fun watching these ideas gain traction, because they work.
But they don't work for the reasons that they're claimed to work. And the way you can prove that is by turning them upside down. Relative to cap-weight, they all work. They add anywhere from 1%-3% per year. They take the Sharpe ratio up anywhere from a little too a lot. Minimum variance, because of its effort to hammer down the volatility, gives you the best Sharpe ratio. Fundamental index, because of its relatively low tracking error, gives you the best information ratio, and the others all give you a blend of the two. It's really neat stuff. And then you flip them upside down, and most of them work better upside down than they did in the first place!
So, if the argument is minimum variance works, because the market has this interesting inversion of the capital market line, and so you're better off sliding up the capital market line, down the risk spectrum (and isn't that wonderful?), well then, why does it work better turned upside down than it does in its original form? And you can go down the list. Every single one of these works roughly the same or better when you flip it upside down. Fundamental index is no exception. It works 2% per year better if you turn it upside down. However, a cautionary note, it is such a bizarre portfolio that I defy anyone to try running it.
Burt Malkiel famously in his 1972 book, A Random Walk Down Wall Street, said a blindfolded monkey throwing darts at a newspaper's financial pages could select a portfolio that would do just as well as one carefully selected by experts. Burt Malkiel was wrong. A blindfolded monkey could do much better than the experts.
We took 100 monkeys and gave them darts. Now these were silicon monkeys buried deep in the bowels of a computer, because they don't eat as much. And we let them throw 30 darts per year at the financial pages, and pick 30 stocks at random for each of the last 49 years, and that was the portfolio, equal-weighted 30 stocks. 49 years of equal-weighted 30 stocks. We did it a 100 times, and what did we find? The average monkey beat the market by 160 basis points per year. Risk was a little higher than the market because it's a concentrated portfolio, and some of the 30 stocks were pretty volatile. So, the Sharpe ratio wasn't drastically better, but it was an improvement. And 2 out of the 100 monkeys lagged the market. Now we know that experts on average don't beat the market, so we know that, at most, two monkeys lagged the experts, so Burt Malkiel was wrong.
Global results, the 100 monkeys over the last 22 years beat the cap-weighted portfolio by an average of 1% a year. The Sharpe ratio for the global cap weight was improved from 0.26 to 0.31. Information ratio is not very impressive, but averaged across 100 monkeys, it's statistically significant, so pretty interesting findings.
All of these price-indifferent strategies outperform. Why do they do that? Every last one of them breaks the link between price and the weight in the portfolio. Every last one of them has some form of rebalancing. When a company goes up, it's likely to be trimmed. When a company goes down, it's more likely to be bought. And as a consequence, what you find is that across all of these strategies, you have a valuation-indifferent index. Why does that matter? Cap-weighting is the only index in this entire study that we did that links the weight to the price, explicitly links the weight to the price, so that the more expensive a stock gets, the bigger the weight it has in your portfolio. Why do we want to do that?
Even Jack Bogle in recent weeks came out and made direct comments, saying, why do we want to do this in bonds? Governments are issuing massive amounts of new bonds, and cap-weighting in bonds means that we're enablers of bad behavior. We're buying all of this, forgive the language, crap, and we're being forced to by the cap-weighting. He doesn't say that about the stock side. He does about the bond side--and he didn't use the bad language.
But in any event, with stocks, you had Cisco at the peak of the tech bubble, priced at $600 billion market cap at a time when it had 24,000 employees. That means every single employee was worth $25 million of market capitalization. You hire a new executive assistant--bingo--your stock goes up 25 million bucks. I'm sorry. That doesn't seem to make a whole lot of sense.
So, the notion of having the price linked to the weight in the portfolio seems flawed. Yes, I know the theory--I understand CAPM, and I understand why in theory it's correct--but Bill Sharpe would be the first to acknowledge that theory is an approximation of the way the world works, and that theory is not the real world.
So, when we do disentangle this, we find that using Fama-French, including the Carhart momentum exposure, we find that there are common denominators. There's a value exposure to every one of these. Why is that? It's because if you break the link with price, price means that the higher-priced stocks get higher weight. Take price out of the equation, and the higher-priced stocks no longer are biased to have the higher weight. There is more of the lower-priced stocks, the lower valuation multiple stocks, than there are the high valuation multiple stocks--far more. And so if you strip that linkage away, you're going to wind up with a value tilt. You're likely to wind up with a size tilt, because there are a lot more small companies than big companies. So, you find a common denominator. They all have a size tilt, they all have a value tilt. And, very interesting, they all have an alpha net of the Fama-French Carhart model. I will come back to that in a minute, because that's an interesting peculiarity.
So, when we look at this globally, we find much the same thing, a size tilt and a value tilt across the board on all of these. When we look at the fundamental measures and their inverse, we find a value tilt on all of them, the direct and the inverted. We find a size tilt on all of them--but especially the inverted. Note that fundamental weight has only a 0.05 on the size tilt. Five-year average earnings weight has a 0.00 on size tilt. It's a myth out there that fundamental index has a deep size tilt … a big small-cap bias. It doesn't. It has a tiny small-cap bias. It has a big value tilt.
But if you flip it upside down, and all of a sudden because you're putting Joe's Bar and Grill at the top of the list and ExxonMobil at the bottom, now you've got a huge size tilt, and you still have a huge value tilt, and so you wind up with a much more powerful Fama-French tilt in your portfolio. Global findings, same thing. You wind up with a value tilt, size tilt, and amplifying typically both when you invert the strategies.
When you look at the smart beta strategies, the commercially available smart beta strategies, some of them have very little size tilt--risk cluster, equal weight, diversity weight, fundamental weight--hardly any size tilt. Minimum variance, maximum diversification, risk efficient, much more of a size tilt. The value tilt, minimum variance and fundamental index have the big value tilt, and some of the others like diversity weight have very little value tilt.
So there are different characteristics--think of them as different flavors: vanilla, chocolate strawberry--but they all have big Fama-French tilts in one flavor or another. You flip them upside down, and they all suddenly have huge Fama-French tilts on size and value. That's why they work better. When you turn them upside down, they have a big size tilts, big value tilts, and that's why they work better.
Notice across all of these, the annualized alpha, net of the Fama-French Carhart factors, is almost always positive, sometimes even statistically significantly positive. Why is that?
Well, clearly if it's positive across all of these strategies, Fama-French Carhart is missing something. It's not a complete model of the real world. Surprise, surprise. We're doing some work on trying to figure out what's missing, and our research is kind of centering on long-horizon mean-reversion, that long-horizon mean-reversion seems to be a driver that's missing in Fama-French Carhart. But the evidence is that it's weaker than these other factors, and that's why these T statistics are mostly not quite significant.
When we look at the global findings for these various publicly available smart beta strategies, it's the same thing--big value tilt. Some have a big size tilt, some don't. You flip them upside down, they always have the big size tilt, they always have the big value tilt, and they almost always have some residual alpha, not explained by Fama-French.
So what about Malkiel's monkeys? Here you find a value tilt. It's modest. Here you find a big size tilt, because darts are going to mostly land on small companies. And here you find a negative Fama-French Carhart alpha. Why is that? Because you have a concentrated 30-stock portfolio, and the noise winds up driving the alpha negative. And when we globalize it, we find much the same thing. The alpha swings slightly and insignificantly positive, but you find the same risk characteristics. They all win for the same reason. Every last one of these smart beta strategies wins because they're breaking the link between price and the weight in the portfolio.
I'd like to spend the next 10 minutes taking you through this formula, so you can understand it very well--no, actually maybe I won't. Jonathan Berk noted that value and size factors generate returns because they sort stocks based on prices. He noted that back in the '90s, and that was an important insight. I kind of hope he finally wins a Nobel Prize on that finding, because it was an important finding. He attributed it to hidden risk factors, which I think was a mistake, but he did find that really you're capturing price, and cap-weighting is the only strategy in this study that has negative skill. It has negative skill, because it's loading up on companies that are high-priced.
Think of it this way: With cap-weighting, you're going to be loading up on growth stocks. You're going to be chasing momentum, and you'll, in a sense, have a popularity-weighted index. Why on earth should we expect a lofty equity risk premium for a portfolio that is growth tilted, that is momentum chasing, and that's popularity weighted? Does it feel risky if you have a portfolio that has got a lot of growth, a lot of momentum, and a lot of popularity in it? No, it does not. Therefore, you should expect a diminished risk premium for that. If you hold a portfolio that looks like the broad macroeconomy or that looks like a minimum variance hodgepodge of illiquid, small, low-volatility companies, it's going to feel riskier when you look at the individual names. Of course, you should expect a bigger risk premium.
So, I think really all that's going on here is that cap-weighting is the only strategy that self-selects on the companies that by dint of being higher priced carry a lower risk premium. It's a very simple risk premium argument.
So, if these strategies are all basically the same, how do you choose? And here I'm going to turn to some analysis that may just seem a little self-serving, but bear with me and forgive me for this. Investors will typically want to maximize capacity and liquidity, and if they're looking for a core portfolio, they're going to want to maximize economic representation that spans the broad macroeconomy. They're going to want to minimize turnover and trading costs.
And here what we find is that there are stark differences. Cap-weighting has huge average market cap. Contrary to some of the mythos out there, fundamental index does not have a huge small-cap tilt. It has a tiny small-cap tilt, because the little companies at lofty multiples, which carry a significant weight in cap-weighting, are de-emphasized back down to their fundamental economic footprint. They get less weight. Reciprocally some of the jumbo companies that are out-of-favor and are deemphasized in cap-weight, get a full weight in fundamental index. So, fundamental index has a slight small-cap tilt that exactly matches cap-weighting's slight small-company tilt, subtle nuance but very interesting and fun to play with. And the whole spectrum of other smart beta strategies wind up having much lower market cap on average.
Adjusted daily volume, the liquidity, the daily liquidity of the portfolio--cap-weighting, of course, leads the herd there, as it should, as it would be expected to. Fundamental index not far behind, about 90% of the daily liquidity of cap-weighted. The other strategies fall off pretty sharply from there: Diversity weighting by dint of being half cap-weight, half equal weight, winds up being a pretty close third on that roster.
What about trading costs? Trading costs are a function of things like your average annual turnover, and your bid/ask spread, and here we find fundamental index and diversity weighting are both very low turnover, not drastically higher than cap weight, and other smart beta strategies are materially higher than cap weight in their annual turnover.
We can also look at things from a risk spectrum. You've got information ratio, and you've got Sharpe ratio. The question is, do you want high information ratio or do you want to willingly take on large tracking error, to tolerate large tracking error, in order to get a marvelous high Sharpe ratio? And there what we find is a clustering off to the right, where fundamental index has wonderful information ratio, and an improved Sharpe ratio relative to cap weight. And the low-vol and the min-var strategies have fabulous Sharpe ratio, but a sharply diminished information ratio, because their tracking error relative to the market is very large. Why is that?
You can't have a beta of 0.7 without having very large tracking error. It's not possible. So you are going to have a low information ratio. I'll avoid reading this detailed notes section to you, and open it up at this stage for questions.
Thank you very much for your time.
Unidentified Speaker: Flipping everything on its head is an interesting concept, but the problem is, there's not enough of Joe's Bar & Grill floating out there to whet the appetites of the investors that would view it as a valuable way to do it--because all of a sudden, that becomes a big-cap stock. How do you get around the conundrum of supply?
Arnott: Well this is an intellectual exercise. It's a fun experiment. Please do not mistake this as a recommended strategy, not in any way, shape, or form. You take those upside-down strategies, you look at their holdings, they are bizarre, and their capacity would be tiny. And so you're absolutely right. There's not enough Joe's Bars & Grills to go around. I did have somebody come up to me when we were circulating a draft copy, saying, can we license that from you? And I said, no, but you can have it. I don't recommend it. No. This is free IP. Anyone who wants to do an upside-down strategy, go for it.
Unidentified Speaker: (Question Inaudible)
Arnott: The question was on the construction of the low-volatility strategy. What we did was to mimic the construction of the MSCI low-vol strategy using the rules that are published. For some reason, I just don't know why, MSCI doesn't want to share with us the historical composition of their indexes. They just don't want to. But in any event, we took their basic rules, applied them, took off the turnover constraint, and just ran with it and said, if you did a low-vol strategy with some of the sector neutrality constraints and tried to make it a nice tractable minimum variance portfolio, optimized for minimum variance, how would it look? How would it perform? And that's what we came up with. So I just view it as a generic approximation of what a very lightly constrained minimum variance portfolio would look like. Our starting point was the 1,000 largest--think of it as the Russell 1000, the 1,000 largest market-cap companies in the U.S.--or for the global portfolio the 3,000 largest.
Unidentified Speaker: (Question Inaudible).
Arnott: The question was, if we paid MSCI, would they give us the back data? If there is anyone from MSCI in the audience, please let us know if there's a willingness to sell us the data at your rack rates; we are very happy to pay.
Unidentified Speaker: Thanks for your presentation. It was very enjoyable. I have a question regarding the impact of correlation on some of these portfolios. Some of the strategies you walk through completely ignore correlation. Others are more correlation based, like a minimum variance approach. I am wondering, if you have thoughts … It's kind of surprising, given the history of market theory, that one might even consider just completely ignoring correlation. But I'd be curious to hear your thoughts on how impactful it can be?
Arnott: You make a very good point. You could loosely partition the strategies we tested into rules-based on optimization-based, and if you make that loose partitioning it really doesn't matter. The optimization-based may have tighter risk metrics slightly, but rules based with broad diversification is going to have fairly similar risk attributes.
Harry Markowitz is a dear friend of mine and probably wouldn't be offended by my saying that the optimizer itself is a tool, which in a broadly diversified portfolio isn't likely to make a huge difference. And so, when you're looking at an optimization-based or heuristics-based approach, they come up with different answers, but broadly looking at the sweep of time, they wind up with startlingly similar results.
The big difference among these strategies is how do they sever the link between price and the weight in the portfolio, and how does that severing the link lead to a large or a small size or value tilt in the portfolio, and how does it lead to a high- or a low-volatility portfolio.
And so if you wind up with relatively low tracking or relatively low volatility, and stark separation of price from the weight in the portfolio, it doesn't matter if you are using an optimizer or a heuristics-based approach. You're going to wind up with a marvelous Sharpe ratio and a marvelous information ratio. The severing of the link between price and the weight in the portfolio is the driver of the alpha. It is not the nuances of the construction of the portfolio.
When somebody comes to you and says, I have got a smart beta strategy, here it is, and look how carefully we craft our optimizer and the alpha comes from this linkage with downside semi-deviation with the risk control thus and such way. Listen politely with a smile on your face, and just know in your heart, it's not any of that stuff. It's just breaking the link with price. Yes?
Unidentified Speaker: What's the case reducing the RAFI benchmarks instead of say a mid-cap value fund that's cap weighted? What's the case for the RAFI benchmarks instead of some of the other smart beta benchmarks?
Arnott: The question is using RAFI as a benchmark versus mid-cap value versus other smart beta. Well, mid-cap value is not a representative broad economy portfolio. It leaves out all the big companies, and it may behave somewhat similarly to fundamental index, but it's not the broad macroeconomy. So it would be very different.
I don't necessarily advocate using fundamental index as a benchmark and the reason for that is that the market isn't fundamentally weighted. Its cap-weighted. And so, using cap-weighted indexes as a benchmark, to me, makes total sense. Using fundamental index as a benchmark makes sense in the more narrow application of, if you're running a fundamental index enhanced strategy, measuring whether the enhancements are working, yes. But measuring, for example, your active equity managers against fundamental index, well firstly they are usually going to flunk, and so you will usually want to fire them fairly soon, but more to the point, it's not the market, and so I don't think it's necessarily that useful as a benchmark.
Other smart beta indexes as a benchmark, most of them don't look even vaguely like the macroeconomy or the macro market.
Unidentified Speaker: (Question Inaudible).
Arnott: As a strategy as opposed to a benchmark, why use fundamental index instead of a mid-cap value ETF? Very simple. If you want a core portfolio that represents the broad macro market or the broad macroeconomy, cap-weighting looks like the broad macro market. Fundamental index looks like the broad macroeconomy, and historically, over time adds about 2% a year. That's valuable, and it adds 2% a year for reasons that I think are totally defensible, based on a different risk premium, based on Fama-French tilts, and it does so in a fashion that's very low turnover, very low implementation cost. I think using that as a strategy, as a core strategy, makes a ton of sense.
Using a mid-cap value strategy is going to work better than mid-cap, cap-weighted core, but the problem with mid-cap value is you're hiving off all the large cap, all the small-cap, segmenting out big chunks in the market, and then you're chopping off and throwing out all the growth names. So, you're dealing with this narrow universe, and then you're cap-weighting it. You're cap-weighting that small sub-universe. So, within that small sub-universe, you're still going to overweight the overvalued and underweight the undervalued.
Now, what that means is, if you took a small-cap value index and reweighted it on fundamental metrics, I guarantee you that over long periods of time, you'd wind up with higher returns, and so if you want to do that, why not find a mid-cap value portfolio that's reweighted fundamentally. PowerShares actually has one, which is kind of cool.
But the notion of going with a niche category, a mid-cap value, in lieu of a broad market, broad economy portfolio, because you think you're going to get the same return while remaining anchored on cap-weighting, to me it just doesn't make sense. Why not go with something that is broad market, covers the whole economy, adds the same amount of value, and does it with less risk? That's so cool. And if you want special loading on mid-value, why not do it fundamentally reweighted?
Unidentified Speaker: Hi, Rob. Great presentation. It seems like one of your core messages is a lot of these alternative strategies can add value in terms of either Sharpe or information ratio, or both, and then it comes down to the execution. I'm wondering, if your team has done much research on quantifying how much value is detracted with the enhanced turnover, in terms of trading costs, potential tax effects, as well as maybe even pushing around the prices of some of the smaller securities?
Arnott: We've done a lot of work on that, and you can reasonably assume that the tax consequences are going to be tied to the magnitude of the turnover. The higher the turnover, the more the tax consequence is likely to be. It's going to be tied to the nature of the embedded trading in the portfolio. Is it done in ETF form or mutual fund form can be a minor contributing factor. The ETFs do have the advantage of coming into the portfolio with in-kind and out with in-kind for the larger chunks, and so it'll also be based in part on whether the portfolio tends to have a momentum-based trading character to it. So, companies that go up, if they're sold, will trigger taxable gains. So, selling high, buying low, will trigger some tax consequences.
In theory, fundamental index should have slightly worse tax consequences than cap-weighting on that basis, slightly worse than cap-weighting on the basis of turnover, but vastly, vastly better than active management on both dimensions, and if there is somebody from PowerShares in the audience, they could tell me if I'm right or wrong on this. To the best of my knowledge, the PowerShares' ETF, which was launched just about eight years ago, has never had a cap-gains distribution, which tells you that it is really very, very tax efficient--very nearly as tax efficient as a cap-weighting.
Schwab in their mutual fund platform goes back to '07, and so going back six and a half years, to the best of my knowledge, they've never had a cap-gains distribution on their fundamental index mutual funds.
So, I think, the short answer to your question is, these are really, really tax efficient. The longer and more nuanced answer is, yes in theory, they'd be slightly less tax efficient than cap-weighting, but more tax efficient than active strategies, and more tax efficient than the higher turnover smart beta strategies.
Unidentified Speaker: (Question Inaudible).
Arnott: The question was, regression to the mean is a little reminiscent of GMO, Jeremy Grantham's work, and the question was, can we elaborate a little bit.
I think, the most powerful driver at work in the capital markets, and the most widely understudied factor at work in the capital markets, is mean reversion. You find it in earnings. You find it in prices. You find it in valuation multiples.
Why is it understudied? It acts over long horizons, and people love to study things with lots and lots and lots of data, and very large numbers of degrees of freedom, and high statistical significance. But just as a case in point: 10-year returns on U.S. stocks, rolling 10-year returns, the correlation of 10-year return in one decade with the next or the prior decade is minus 43% going back over the last two centuries. Minus 43%! That means you have a great decade like the '90s, odds are about 75-25, you're going to have a lousy decade to follow. There will be exceptions, but not many. And with 20 samples, it's not very interesting to study. Broaden it to cover all of the developed markets in the world, and the best you can do is maybe 200 independent samples. So, people don't study this stuff. And when you look at long horizon mean reversion in individual stocks, you find the same problem.
Our work suggests that long-horizon mean reversion is probably one of the most understudied parts of the capital markets, and is the root source of value-added for global TAA [tactical asset allocation], for fundamental index, for contrarian investing. It gets back to Baron Rothschild's famous comment … he made a fortune in the Napoleonic wars. He famously said buy when there is blood in the streets. Hardly anyone hears his full sentence, which was buy when there is blood in the streets, even when the blood is your own.
If you're a contrarian investor, you don't know where the bottom is. You're going to buy when it's cheap, buy more when it's cheaper, buy more when it's cheaper still, and by then, you've got some blood in the streets mixed in with that first blood. It's painful. Being a contrarian is painful. Being a fundamental indexer is painful. Buying stocks that are wildly out of favor, feared and loathed, and buying them because they're feared and loathed, is painful. Wrapping it in a fundamental index wrapper suddenly makes it a little more palatable, a little bit more understandable, a little more defensible; it gives people a little more confidence to stay the course through that discomfort. Makes them a little bit more comfortable with the fact that a global fundamental index portfolio has, for two or three years in a row, been trimming Google each year just a little bit and buying Greek and Cypriot banks each year, just a little bit. You single out that trade, it looks horrible. But you embed it in a broad, overall strategy, and that trade doesn't go in your favor a couple of years in a row, it's OK. It's embedded in a larger strategy, and it doesn't hurt you much in the context of the larger strategy. So, you're buying when there is blood in the streets.
Time for one or two more questions.
Unidentified Speaker: (Question Inaudible).
Arnott: The question is [about] the case for RAFI versus other smart beta strategies. Why use a RAFI ETF against some of the other smart beta ETFs.
The other smart beta strategies are highly likely to add value. And before fees and trading costs, are actually--and this pains me to say--are highly likely to perform approximately as well as fundamental index, if they've done a good job of stripping price totally out of the weighting mechanism--if they've done that.
Some of them reintroduce price with sector reweighting towards cap-weight, or with turnover limitations that reintroduce price drift into the portfolio. And to the extent that they do that, they wind up reintroducing price into the weight, and lose some of their alpha. But the ones that don't do that, they should work just as well as fundamental index before costs.
If their turnover is higher, and they trade in less-liquid, smaller companies, you're going to get a bit more of a haircut. If their fees are higher, you're going to get a bit more of a haircut. So, I think we have a small advantage there.
Our big advantage is, these are core strategies. They span the entire macroeconomy, they're the only smart beta strategies that can really be seen as a natural alternative to cap-weighting, a natural core vehicle, a natural complement to cap-weighting for those who want to split their core--half cap-weight, half smart beta--this would be the logical smart beta choice.
The other smart beta strategies are more niche strategy in character, not core in character. At least that's my view, but they're wonderful strategies, and people who want to embrace smart beta, try them, try them. They're neat as concepts. I love the fact that fundamental index has opened the door to a whole array of neat ideas. The ETF arena, and the indexing arena, 10 years ago was as dull as watching the grass grow. It was a dull arena. It was gaining traction, gaining assets, gaining momentum tremendously because it beat active managers, and it beat active managers while delivering better returns at much lower fees.
The smart beta suite is suddenly making the whole arena of indexing and ETFs so much more interesting. It's so fun to watch.
Thank you all very, very much for your time.