Note: There's no typo. I really mean ABs, not ABCs. Stick around, you'll see. And by the way, here's the punch line: Invesco PowerShares comes off looking good, and so, in an indirect way, does (no surprise here) Warren Buffett.
I love reading prospectuses of newly issued ETFs. And yes, I do have a life. It's a professional thing. I'm in the business of model-based equity strategies, and after feeling like a maverick for so many years, I enjoy having company. So I was thrilled when PowerShares (now Invesco PowerShares) launched their model-based set of "Dynamic" ETFs about a decade ago. And I've been continually delighted by the never ending stream of new ideas for how to create what sponsors like to think of as alpha-producing indexes, and how sponsor lawyers amusingly shove the language of active strategy building into the lingo of ETF passive index tracking: The Super Duper Brilliant E=MC2 Investment Trust will invest in securities that passively match the performance of the Super Duper Brilliant E=MC2 Index, an index that's so active it can lead to planetary destruction if constituent companies overly pursue high EaR (existence-at-risk) strategies.
But do these things work? Do they succeed in doing what they promise to do (subject to the usual disclaimers to the effect that if they screw up, qué será será). How can you separate the wheat from the chaff, the pretenders from the contenders?
This is an important question if you are or are considering robo-investing (a field which, by the way, I think is very much worth considering). You're constantly going to be seeing portfolios designed to accomplish specific things. You need to be able to assess them quickly and without having to get a PhD in Finance.
Many fancy ETFs are still too new to properly evaluate, but by now, we have enough with at-least five-year histories to warrant consideration.
I'm Sorry, I Really Am, But I Have To Do This
I know you want to get to the good stuff, identifying the successful nouvelle ETFs. So do I. But to make sense of them, we're going to have to plow through some formulas.
The math is easy, so easy the free calculator app on your smart phone can zip right through it. It's the idea behind it that contributed to 1990 Nobel Prize for Economic Sciences awarded to William Shapre, one of the originators of the main formula known today as the Capital Asset Pricing Model, or CAPM. (For the record, though, there is some politics here; Jack Treynor seems to have pushed these ideas first in 1962-62, with Sharpe independently producing it in 1964, John Lintenr in 1965 and Jan Missin in 1966).
Let's pretend you hired the fictional investment firm Buffett Cramer LLC (how good would it to be a fly on the wall when that investment committee meets) to manage your stock portfolio. It returns 22% in 20X5. Good stuff, you think. But then, in 20X6, your portfolio drops by 7%. Uh oh. Maybe Warren and Jim don't blend skills effectively. Time to fire them and hunt for another firm, or, perhaps, a robo-adviser.
It's understandable to assume that up is good and down is bad. But acting basesd solely on this information is no more sensible than bundling up in an overcoat, scarf and fur-lined gloves just because the weather forecast is for cooler temperatures than we've had in the past week. Has it cooled from the low thirties to the low twenties? Or has it cooled from the low nineties to the mid eighties? Context is crucial.
When evaluating investment performance, the are two kinds of context that must be considered:
1. The market as a whole
In my introductory post, I took a swipe at CAPM based on its inability to guide forward-looking investment decisions in the real world. The model's genius, however, lies in how it can help us evaluate what has happened in the past in the way it teaches us to systematically deconstruct asset returns in a way that spotlights the relative contributions of the market, risk, and the unique contribution of the investment decision-maker.
We start here:
RR = ER + UR
RR = Realized (or actual) Return
ER = Expected Return
UR = Unexpected Return
RR is easy. We can look at our account statements to see that. The challenge is with ER. We can't base our expectation on hopes, dreams, wishes, etc. ER is based on systemic logic, and that's where CAPM comes in.
What Return Can You Legitimately Expect?
The CAPM, the formula we use to calculated ER, expected return, is follows:
ER = RF + (B * (RM - RF))
ER = Expected Return
RF = Risk-free return
RM = Return on the Market as a whole
B =Beta, a numerical index that measures risk relative to the market (If Stock A is 10% riskier than the market, it's beta would be 1.10. If stock B is 25% less risky than the market, its beta would be 0.75. As you probably guessed, the market beta and the beta for assets exactly as risky as the overall market is 1.00.)
RF is easy. Use a U.S. Treasury rate. We can and do argue over how long a term treasury security we should use, and if we go out more than few months, there may be market risk if you have to sell before maturity. But for convenience, we just ignore that.
RM, the expected market return, is a total monstrosity. Nobody has any idea what that can possibly be. For forward-looking investing, we cannot use any historical data because the market has gone up and down all over the place and we don't know which periods will be relevant in the future. And the future is important because we're dealing here with expectations and nobody invests in risky assets assuming as negative return, which occurred on many occasions, and nobody with an ounce of common sense will use an oddball period of exceptional gains. (That doesn't mean crazies don't exist. Shortly before 2008, my boss, a Quant with a PhD in math, wouldn't allow me to deviate from historical observations regarding the Chinese stock markets despite the fact that the only data we had was for a wild speculative buying panic; needless to say, that project failed as did that entire operation. So be careful if you read "white papers," etc. making reference to use of Nobel Prize wining models.)
Because we can't even guess what RM is, we're really up the creek when it comes to RP (the risk premium), the difference between RM and RF. If you just want a theoretical understanding of the market, that's OK. We don't really need a number. But if for whatever reason you absolutely positively want to compute something, just assume a risk premium of 4%-5% (you can split the difference and go with 4.5%). It's sort-of a gentleman's agreement that let's us stop endlessly debating and get on with the calculating. Ultimately, trying to disprove the validity of 4%-5% and come up with something else is more trouble than it's worth.
Because we're not using CAPM to make a going-forward investment using real money but are instead looking to evaluate what actually happened in the past, we have a big advantage. We actually can consider historic market results. Phew.
Beta is also easy to observe and calculate from historic data. (If you're into quant concepts, it's the coefficient attached to "x," the independent variable that represents market return, in a linear regression in which x is used to determine y, the dependent variable or the asset return.) For non-quants, just stick to the basic definition: a barometer of risk relative to the market. The problem is that the easily calculated betas based on history can be erratic or even useless to truly assess risk because as the lawyers love to remind, past performance doesn't necessarily mean squat.
When assessing the risk of individual stocks, beta can lead to some pretty idiotic conclusions. For example, shares of Itty Bitty Biotech Research Enterprises Inc. might fluctuate wildly as drug trials succeed and fail and the company might constantly be only one bad test away from bankruptcy and liquidation. But the testing calendar may be such that Itty Bitty gets good news and peaks at times when the market tanks because of something the Fed and vice versa. Because its volatility is uncorrelated with the market, Itty Bitty will likely have a very low Beta, one that leads quants to see it as a very conservative stock. But I've found that such oddities can be mitigated when we look at Beta on a portfolio-wide basis that reflects a bunch of stocks, rather than just one.
Back to Buffett Cramer LLC
Here, again are the realized returns (RR) posted by our fun-if-was-real advisory firm:
Our initial impression was: They're slipping. Let's see it that's really true.
We need more information. We need to know RF, RM and B.
Assume RF was 2%. Assume B, in both years, was 1.5, meaning Warren and Jim took 50% more risk than could have been the case simply by investing in the SPDR S&P 500 EFT ($SPY). I suppose we also have to assume Warren was out sick the day the investment committee met and made that choice. Let's assume, too, that RM (return on the market) was plus 20% in 20X5 but that it dropped 6%, in 20X6.
We can now determine that ER, expected return, in 20X5 was 29%, based on the following:
· RM = 20%
· Cool, with a 22% realized return, we beat the market! But wait, there's more.
· RP = RM - RF = 20% - 2% = 18%
· ER = RF + (B * RP) = 2% + (1.5 * 18%) = 29%
· Ouch. Given the amount of risk taken under those market conditions, the portfolio should have earned 29%. Warren, get well soon!
It's now 20X6 and Warren is back and making his voice known.
· RM = -6%
· Not good. We lost 8%, worse than the market. Patience please.
· RP = RM - RF = -6% - 2% = -8%
· ER = RF + (B * RP) = 2% + (1.5 * -8%) = -10%
· Oh. How about that? Given the amount of risk we took, we should have lost 10%, but in fact, we only lost 7%. You go guys! Welcome back Warren.
Let's look at it this way:
RR 20X5 = +22%
RR 20X6 = -7%
UR 20X5 = RR - ER = 22% - 29% = -7%
UR 20X6 = RR - ER = -7% - -10% = +3%
Notice how our opinion dramatically changes. It seemed, at first, like Buffett Cramer LLC went from great to horrible. But adjusting for market conditions and the amount of risk taken, we see that in fact, the firm delivered much more effectively for us in 20X6.
Reviewing our ABs
Now we're ready for the ABs of investment performance analysis. A stands for Alpha, which is the highfaluting name used to designate UR, or unexpected return. Our goal is investments that deliver positive (or at least non-negative) alpha, if not in every possible time period, then on the whole over a prolonged period.
B stands for - you probably guessed it - beta. It's not inherently good or inherently bad. It helps provide a standard by which we can evaluate RR, realized returns in a proper context, in the context of the market environment (RP and RM) and the amount of risk we took.
So Let's See Some Cool Nouvelle Funds
I screened the Portfolio123 ETF universe to find US equity funds that do not use shorting or leverage and which select stocks using quantitative (Quant) models; i.e. funds that passively track a hip index that ties to deliver alpha. Although for the sake of simplicity, the examples presented above used one-year numbers, in order to hunt for ETFs, I worked with five-year data
To give you an idea how important CAPM analysis is, I ranked all those Quant funds by B, Beta, and by RR over the past year. If all of these funds were doing what skeptics might suggest they're doing (delivering higher returns simply by taking higher risk and vice versa), the correlation between B and RR should be close to 1.00. Actually, though, the correlation is -0.40. In human terms, we square correlation (multiply it by itself) to come up with a metric known as R2 ("r-squared") that tells us the extent to which one item can be seen as having caused the other. The R2 here is 16%, which, together with the minus sign in front of the correlation, means that beta accounts for 16% of the movement in returns achieved by this group of ETFs, and that such movements ran against expectations. High beta is supposed to be associated with higher returns, but in fact, we saw the opposite. That means the funds are telling the truth when they claim to be trying to generate returns that were separate and distinct from those which could be explained away by basic capital market theory.
I have to tell you here that when it comes to defining "the market" (for RM), I did something that is definitely debatable. I used the S&P 500, or more specifically, $SPY. For large-cap ETFs, that's a pretty solid choice. But does it make sense for ETFs that specialize in small cap. Should I have used the Russell 2000? And what about ETFs specializing in, say, value? S&P and Russell have value indexes I could have used. If I were designing the strategy, I would have used a benchmark that better matched the portfolio orientation. That would have given me better information on whether and how my ideas were impacting performance separate and apart form the performance of small cap in general, value in general, etc. But that's not our goal here. I'm viewing this through the eyes of an investor who is deciding whether to make a thoughtful (active) decision as to which ETF or ETFs to choose, or take what for many is the default no-thinking no-deciding course of action; going with the big-gorilla brand name $SPY ETF.
So now, the question becomes: Which funds succeeded on the whole over the past five years? The best-performing ETF in this group delivered a five-year annualized Alpha (Unexpected Return) of plus 35.1%, while the worst performing fund delivered minus 3.2%. Both extremes involved sector-specific ETFs; Energy at the low end and Biotech at the high end. These need to be assessed a bit differently (we should consider sector-relevant benchmarks and ways to decide which sectors we want to consider since yesterday's winners and losers can and often do flip flop on a dime). So I narrowed further to non-sector-specific ETFs.
Betting on Invesco PowerShares
I could go beyond having controlled for sector specialty. I could also control for size (large-cap, mid cap, etc.) and style (value or growth). But eyeballing my results suggests I need drill down that way.
The top two entries among the sector generalists are PowerShares S&P 500 High Quality ETF ($SPHQ), with a 5- year annualized alpha of 2.27% and PowerShares Dynamic Large-Cap Growth ($PWB), with annualized alpha of 1.53%.
There are only three sector-generalist Growth Quant funds with five-year histories. Behind $PWB, we have two First Trust Alphadex funds; First Trust Large Cap Growth ($FTC) and First Trust Multi Cap Growth ($FAD) with alphas of -0.23% and -2.68%.
Let's switch to value. Once again, we have three entries, one from PowerShares and two from First Trust. Leading the way is PowerShares Dynamic Large Cap Value ($PWV) followed by First Trust Large Cap Value ($FTA) and First Trust Multi Cap Value ($FAB). The respective five-year annualized alphas are 0.69%, -3.27% and -5.42%.
Are you noticing a large-cap tilt here? You should. But we can't say for sure it's large cap per se.
Betting against Beta
The three good alpha-producing quant ETFs from PowerShares have something possibly more important in common than large-cap orientation; low beta. The numbers are 0.92 for $SPHQ, 1.09 for $PWB and 0.91 for $PWV. Higher-beta ETFs have fared worse on balance.
Do you remember that negative correlation I referred to above suggesting that higher betas correspond to lesser results? It's important.
In a pair of 2013 research papers, it was demonstrated that much of the excess equity return generated by Warren Buffett's Berkshire Hathaway ($BRK-B) came from low-cost leverage (funded by Berkshire's insurance operations) and a market anomaly that has become known as Betting Against Beta. With regard to the latter, it seems that institutional money managers need return (who doesn't) but are not permitted to use margin. So they chase high-beta stocks. Well what happens whenever any group chases anything? You got that right: Demand rises relative to supply, which pushes prices upward. And in the market, paying higher prices usually means earning lesser returns.
Hence the world is turned upside down. High-beta stocks, which should deliver better returns, are in greater demand causing prices to rise and returns to fall. On the other hand, low-beta stocks, which should lag, are not widely sought and hence had lesser valuations and produced better returns.
How can we tell whether the anomaly remains relevant and, hence, whether it pays to go into low-beta funds even if we want superior returns? How can we tell if the anomaly vanishes meaning return seekers must again chase beta? Use the sort of analysis described above. Compare RR to ER and UR, and notice which kinds of funds are producing better UR (alpha). Or forget the whole thing and just own $BRK-B.
Anyway, it appears that at least as of this time, and important characteristic of Invesco PowerShares relative to First Trust is that the former is modeling around the betting-against-beta phenomenon. I can't say one way or the other if it's intentional or if it just turned out that way. With Warren Buffett, we know it's deliberate: While he doesn't openly use language referencing beta (actually, the remarks I heard from him along these lines tend to poke fun at it), the things he has said over and over about predictability, inevitable growth, and so forth are completely consistent with the approach. Whatever the case, it seems to be accomplishing something. And if one is afraid that the market may falter after a long run of strength, betting against beta may be a decent way to go.
By the way, this is relevant to why the large-cap ETFs have been doing better. The smaller cap funds have had higher betas. This is really interesting. There's a well-known market phenomenon known as the small-cap effect, wherein smaller stocks have tended to perform better. Does this contradict "Betting Against Beta?" We don't have to go that far. PowerShares and First Trust seemed to have built models that captured high-beta small-cap stocks. But there are small cap stocks that also feature lower beta. I'll hunt for ETFs that combine these features. If not, I may create something for you on Portfolio123 Ready-to-Go.
Hopefully, you now understand how alpha and beta can help you make sense of models, portfolios and funds you're likely to encounter as you and/or your advisor put together and maintain and investment program. Next up, I'll cover two chic variations on these phrases; smart beta and smart alpha. The former is flying around a lot and some big companies have established themselves based on it. Actually, though, as we'll see next time, smart beta is a largely meaningless phrase and what it offers that's relevant really falls under the heading of the newer less-well-known term, smart alpha.