Minimum volatility - simply factor exposure

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
Topic Author
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Minimum volatility - simply factor exposure

Post by Robert T »

.
Of all styles with recently launched funds, minimum volatility has received a significant share of assets. For example, since its launch in Oct 2011, the iShares MSCI US Minimum Volatility ETF (USMV) has accumulated $2.9 billion in assets. Across the full suite of ishares minimum volatility ETFs across all markets, accumulated assets are about $8 billion. From my analysis, this is not a separate factor, but simply a combination of market, size, value, and quality exposure. Here are the factor regression results for the full period for which MSCI has backtested data for the US minimum volaility index.
  • MSCI US Minimum volatility index - June 1988 - August 2014

    Alpha = -0.01
    Mkt = +0.84
    Size = -0.15
    Value = +0.19
    Momentum = +0.01
    Quality = +0.19
After accounting for factor exposure, the residual alpha is essentially zero. The min vol series has a signficant load to value and quality of the same order of magnitude (+0.19 load), has lower beta (0.84), and is large cap (-0.15 size load). It also has a much higher allocation to 'defensive' sectors (consumer staples, healthcare, and utilities) than the market (42% vs. 24%). FWIW, in earlier analysis I found that consumer staple on their own had similar mkt beta (+0.85) and signficant value and quality factor loads (higher for quality). http://www.bogleheads.org/forum/viewtop ... 7#p1813977 . So, nothing unique about minimum volatility beyond (mkt, size, value, quality) factor exposure.

Robert
.
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Minimum volatility - simply factor exposure

Post by jaab »

I am wondering about quality/profitability and the F/F term/duration risk factor. In at least one study pushed out by Hsu (RAFI) and his co-authors there seems to be a duration exposure for LovVol. They did not include PMU or QMJ however, if I remember correctly. You also have the duration factor exposure of utilities for exmple, which seem to be a traditionally "higher quality" sector. What happens if you regress one on the other, include both or something? Thoughts?
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

Jaab
I've written several pieces on low volatility

Here's one http://www.etf.com/sections/index-inves ... -leap.html
Here's another http://www.etf.com/sections/index-inves ... ccess.html
and this
http://www.etf.com/sections/index-inves ... tocks.html

The research and logic does show a term exposure which is logical as these firms tend to have more stable earnings (quality/profitability and/or Betting against beta factor) and that means their earnings are more bond like, so shouldn't be a surprise.

I've got another piece coming shortly on this issue

Larry
Last edited by larryswedroe on Sun Sep 28, 2014 10:17 am, edited 1 time in total.
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

The MSCI index is specifically trying to avoid non-beta factor exposure in part by imposing sector and stock weighting constraints.

Different factor models will of course attribute returns to somewhat different factors, as for example: https://www.capitaliq.com/media/100971- ... folios.pdf.
Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Minimum volatility - simply factor exposure

Post by Browser »

So, is simply factor exposure a problem? Seems like maybe a good way to get it.
We don't know where we are, or where we're going -- but we're making good time.
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Minimum volatility - simply factor exposure

Post by jaab »

Not a problem, quite the opposite. For decision making this is good news, because it keeps things simpler. Many mathematical portfolio optimizations seem to end up in the well known factor exposures, not just LowVol. There was a thread about this in 2013: http://www.bogleheads.org/forum/viewtopic.php?p=1881632

The potential problem I see is that in the future such optimizations may end in different exposures, because they are just a by-product. You don't know what you get. Seems (too) uncertain to me, as there are products availabe that more or less directly target factors, like RAFI.
Last edited by jaab on Sun Sep 28, 2014 8:20 pm, edited 2 times in total.
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

"Simply factor exposure" is a problem when it's not that.

For example, the DFA crowd claims that lov vol performance is due to value exposure. Yet, when you plot the exposure across time, you get a chart that looks like this: http://1.bp.blogspot.com/-TMnzlBYGEHE/T ... onroll.png. It's hard for me to say that min. vol. = value tilting looking at that chart (full article where the chart comes from: http://falkenblog.blogspot.com/2011/09/ ... -play.html). In short, when the exposure of a strategy to a particular factor swings all over the place across time, it's not "simply" factor exposure because it averages out to a non-zero number.

And, of course, it all depends on which factors you choose. The factor model Robert T uses says that min. vol. loads on value. Yet, the paper I linked to above that uses a different factor model says there is no value exposure.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

stlutz
It's not JUST value exposure but exposure to other factors, including as I mentioned term and quality/profitability (or Betting against Beta)
And FWIW low vol stocks now look completely different than they used to as cash flowed in. IMO not a good bet now as they don't look anything like value stocks any more
Larry
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

And FWIW low vol stocks now look completely different than they used to as cash flowed in. IMO not a good bet now as they don't look anything like value stocks any more
Look at the chart I linked to. You'll see that low vol. has been in the "non-value" area for long periods historically. It's simply not the case that low vol. always = value up until suddenly flipping in 2010.

Moreover, it's also the case the low. vol. has much more marketlike valuations than it did a couple of years back (as one would expect, these stocks went up less during the big up years of 2012 and 2013. For example, USMV currently trades at 24 times earnings while EUSA (the index that USMV is based off of) trades at 23 time earnings. You'll see about the same thing when you compare SPLV vs. the S&P 500.

I too was cautious about the strategy back in 2011 when it had a surge of popularity. The assets bases of the involved funds have stabilized quite a bit (i.e. assets are no longer "pouring in"), and are dwarfed by small/value strategies in all of their various forms, which you don't express concern about being overgrazed.

At current prices, I am no longer concerned about pursuing a low-vol strategy. In threads a couple of years back, I was.

In part, I think we are after different things. I like low. vol. because it basically offers the market return with lower risk. You would be less interested since your main focus is on beating the market. No problem with that--people should purse the strategies that best fit their goals/situation.
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Minimum volatility - simply factor exposure

Post by jaab »

I like low. vol. because it basically offers the market return with lower risk.
Lower volatility yes - if there are lower drawdowns is a different question as I understand? see http://ibd.morningstar.com/article/arti ... &CN=brf295 for example:
One is that the lunch might not in fact be free. It may have a price that is hidden. Such is the argument of James Xiong, Tom Idzorek, and Roget Ibbotson (two Morningstar researchers and the founder of Morningstar's Ibbotson business group) in "Volatility vs. Tail Risk: Which One is Compensated in Equity Funds?" The trio examined 30 years' worth of mutual fund performance and determined that yes, lower volatility does not lead to decreased performance. For U.S. equity funds, the outcome is symmetrical; funds with average standard deviations show the best aggregate gains, while those with the lowest and highest risk land at the bottom (with identical results, as it turns out). With funds owning foreign equities, the effect is more strongly in favor of low-risk funds, as they have the best performance and the high-risk funds have the worst.

However, standard deviation is but one way of measuring risk. Another is to look at so-called tail risk, which examines how a fund performs under extreme conditions. The researchers used a statistic called excess conditional value-at-risk, or ECVaR, that evaluates how a fund behaves during sharp downturns. Effectively, ECVaR measures if a fund suffers particularly bad results when stocks are plummeting. The trio found that if ECVaR is used to calculate risk rather than standard deviation, mutual fund results line up according to standard investment theory. That is, the lowest-risk funds as measured by ECVaR have the weakest future returns, and the highest-risk funds have the best. This holds true for both U.S. and foreign equities.

So, from the perspective of tail risk, the markets are efficient. You get what you pay for.
Link to the mentioned tail risk study: https://corporate.morningstar.com/us/do ... ilRisk.pdf
Valuethinker
Posts: 48944
Joined: Fri May 11, 2007 11:07 am

Re: Minimum volatility - simply factor exposure

Post by Valuethinker »

Robert T wrote:.
Of all styles with recently launched funds, minimum volatility has received a significant share of assets. For example, since its launch in Oct 2011, the iShares MSCI US Minimum Volatility ETF (USMV) has accumulated $2.9 billion in assets. Across the full suite of ishares minimum volatility ETFs across all markets, accumulated assets are about $8 billion. From my analysis, this is not a separate factor, but simply a combination of market, size, value, and quality exposure. Here are the factor regression results for the full period for which MSCI has backtested data for the US minimum volaility index.
  • MSCI US Minimum volatility index - June 1988 - August 2014

    Alpha = -0.01
    Mkt = +0.84
    Size = -0.15
    Value = +0.19
    Momentum = +0.01
    Quality = +0.19
After accounting for factor exposure, the residual alpha is essentially zero. The min vol series has a signficant load to value and quality of the same order of magnitude (+0.19 load), has lower beta (0.84), and is large cap (-0.15 size load). It also has a much higher allocation to 'defensive' sectors (consumer staples, healthcare, and utilities) than the market (42% vs. 24%). FWIW, in earlier analysis I found that consumer staple on their own had similar mkt beta (+0.85) and signficant value and quality factor loads (higher for quality). http://www.bogleheads.org/forum/viewtop ... 7#p1813977 . So, nothing unique about minimum volatility beyond (mkt, size, value, quality) factor exposure.

Robert
.

How are you defining "quality" though?

It would seem quality is low volatility?
Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Minimum volatility - simply factor exposure

Post by Browser »

Naively, I wouldn't expect to be buying a "value" investment with low vol in the first place. If it happened to load on value, that would be a stop at the dessert tray. So, knocking low vol for something it's not advertised to deliver seems pointlesss. If I wanted to own value, I'd own an investment targeted at value, probably SCV. Seems like this might be a good combination with low vol, since low vol is defensive and SCV would be expected to to better under favorable market conditions.
We don't know where we are, or where we're going -- but we're making good time.
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Minimum volatility - simply factor exposure

Post by jaab »

But low vol is advertised by Falkenstein and others as unique and free lunch (others make mistakes) trading strategy. It then showing time varying factor exposures that in the longer run end in a mix of lower market beta, value, duration, quality, ... with no significiant alpha makes it look not so unique and more unreliable to predict. And it does not even seem to really reduce crash risk, just volatility. I don't buy the buzz about low vol so far. :? :|
Last edited by jaab on Mon Sep 29, 2014 8:55 am, edited 1 time in total.
Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Minimum volatility - simply factor exposure

Post by Browser »

Maybe low vol is like your cash-back credit card with rotating categories. :D You get cash back, but you just don't know which categories are paying out each period.
We don't know where we are, or where we're going -- but we're making good time.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

Stlutz
Few more thoughts
First, here's problem IMO. Stocks with low vol and HIGH SHORT interest perform miserably. It's only the low vol with low short interest that outperform. These tend to be quality stocks. Note that high vol with low short interest perform fine. During the “dot-com” bubble an equally weighted high volatility/low short interest portfolio had an annualized compound return 3.5 percent greater than that of the CRSP value weighted index. During the financial crisis, that same gap was 13.3 percent.This demonstrates its the other factors that low vol got you that mattered, at least IMO.

Second, we know that valuation matters, and it matters a lot. Here's table I put together for piece just finished

The table below is based on Morningstar data as of September 22, 2014.
SPLV USMV IWB IWD
Price-to-Earnings 18.1 19.3 17.8 16.1
Price-to-Book 2.4 2.9 2.4 1.7
Price-to-Sales 1.7 1.6 1.7 1.4
Price to Cash Flow 10.0 11.0 7.8 5.7

Third low vol stocks outperform only the highest vol stocks. Simply better off IMO screening those out. If you screen out the low priced stocks too not much there in low vol.

Fourth, Don't know where you get the idea that my goal is to beat the market. If that was the case I would have high beta exposure too. It's getting market like returns with lower risk- same as you state for your goal. And if want lower vol IMO far better way is to use low beta and high quality bonds with high tilt. Far more evidence and economic theory to support that strategy IMO while low vol is an anomaly that is behavioral. No risk story there.

Fifth, as mentioned low vol seems to have exposure to term risk. At least that should be taken into when deciding on term of your bond holdings. But low vol has benefited from bond rally.

And finally, while the data is impressive, I'm not convinced on quality going forward as the risk story is pretty darn thin. IMO it's behavioral and one that can be arbed away by cash flows, which we are certainly seeing. So if low vol was getting you quality exposure too, that seems to be an issue.
Bottom line to me is that if you want less vol, and less tail risk just raise tilt and lower beta, at least there you have risk story and logical way correlations tend to work


Larry
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

Larry,

Yes, the short-interest anomaly is a powerful one. Perhaps you can point me to the low short interest ETF? I'm not aware that one exists. For someone picking their own stocks, it's a good indicator to use. Regardless, I don't see how the short-interest anomaly is a problem for low vol. So, short-interest is a more powerful indicator for beating the market than is low vol. I agree 100%. As noted, my interest here is not in beating the market. And there aren't that many stocks with low volatility and high short interest regardless. Certainly you could juice-up your low vol. returns (or the returns of any other factor strategy) by incorporating short interest as well. When the ETF that does that comes to market, I know I'll take a look!

I agree that valuation matters. I've said so in every thread we've had on this topic. A couple of years ago low. vol. portfolios were priced at 20% premium to the market, which definitely caused me to pause. Now it's more like 5%, which I'm comfortable with.

Again, low vol. stocks have provided very close to the market return with significantly less volatilty. The MSCI index which Robert T refers to at the beginning of the thread has provided identical returns to the overall MSCI USA index since inception (5/1988). Of course, it's a backtested index, so caution is always warranted. However, low vol. backtests generally show returns equal to or a little less than the market over time with [obviously] lower volatility.

I get that idea that you are looking to beat the market by the rest of your paragraph. If you pick the "best" stocks, you can either get higher portfolio performance or you can then add more bonds and get the same performance with less risk. Either way, the key to the strategies you promote is to pick equities that will beat the market. As a value tilter, I have no objection to that!

Finally, as you know I'm not particularly persuaded by risk-based explanations of any of the anomalies we talk about, as the notion that risky=all but guaranteed to beat the market makes no sense to me. We've hashed through that before and I seem to be a minority on that count, so those who are more interested can just search old threads.

Jaab--

Actually, Falkenstein doesn't really argue that low vol is a free lunch. Instead, he argues that it seems anomalous because of most in academic finance misunderstand risk (according to him). His argument is that risk is and is perceived to be relative. Happiness comes from being better off than your friends/neighbors etc. not from reaching some absolute level of accomplishment. In investing, this means that "risky" = failure to beat to the market. To avoid this risk, one must pick investments with higher potential payoffs--i.e. high volatility stocks. In this framework, risk and behavioral explanations mesh into one. His approach may or may not be correct (I'm not sure, personally), but it's interesting and his book, "The Missing Risk Premium" is worth a read for those who find such material interesting. :happy
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

stlutz
The problem is that the stocks that have high vol and low short interest have great returns shows it's likely isn't high volatility that is the problem but other factors are explaining low volatility.
I'd also note the one study found that all the excess returns happened in first month after formation and thus likely high turnover would eliminate the benefits
Xi Li, Rodney N. Sullivan, and Luis Garcia-Feijoo, authors of the study “The Limits to Arbitrage and the Low-Volatility Anomaly,” which appears in the January-February 2014 issue of the Financial Analysts Journal (January-February 2014), found that the excess return associated with forming the long low volatility/short high volatility portfolios are basically present only in the first month after formation, and that they are largely subsumed by high transaction costs associated with stocks with low liquidity (such as low priced/high volatility stocks). They also found that the anomalous returns within value-weighted portfolios are largely eliminated when omitting low priced (less than $5) stocks, and are not at all present within equal-weighted portfolios. In fact, the average price of stocks in the highest volatility quintile was just over $7, indicating that many, if not most would be considered “penny stocks”. And finally, they found that the low-risk effect has been noticeably weaker since 1990 — new regulations were passed in that year aimed at reducing fraud associated with trading penny stocks. (I’d add that many of the high beta stocks simply disappeared after the dot.com crash — the number of stocks on public U.S. exchanges has shrunk dramatically since then). The authors concluded: “Our finding cast some doubt on the practical profitability of a low risk trading strategy.”
I just don't find any logic to the low volatility strategy, and the evidence to me seems pretty strong it's other factors. And it's really all about avoiding high volatility, the well known lottery effect

You want less vol, use low beta and high tilt, and less tail risk, and logic besides that low vol doesn't

Larry
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

Larry--

In case I didn't make it clear before, I have absolutely no objection to pursuing a strategy of picking high volatilty stocks with low short interest. I choose not to do so because a) I don't invest in individual stocks; and b) such a strategy is not available in ETF form. If it was, I would at leaset consider investing. Growth stocks with high momentum also outperform. Does this mean that that value effect is likely irrelevant? I say no.

Also, I don't really care about long/short high-turnover hedge-fund strategies. Again, I can't invest in them. If I could, I wouldn't want to--as the paper you cite indicates, there are too many costs involved. The turnover ratios of the low vol. ETFs that actually exist average around 30% per year--just a bit higher than most smallcap value funds. These ETFs have been tracking their indexes very closely since they were formed, which indicates that they do not face issues with trading costs.

With all factor strategies, if past patterns continue into the future, they will be good choices. If they don't, well that's the risk of making bets on various corners of the market. No matter how strongly we want the disclaimer to be false, there is simply no guarantee that past performance indicates what future performance will be.

Nonetheless, given the fairly univeral acceptance of the value premium in the financial community as opposed to the skepticism of low vol. in large segments of it, I would bet that the later would be more likely to persist than the former. The dollars chasing value in all of its forms right now dwarf those going after low vol.

Best,

stlutz
User avatar
Topic Author
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Minimum volatility - simply factor exposure

Post by Robert T »

.
Here's the analysis from the OP broken down into 1990s, 2000s, 2010s, starting again with the full period with a bit more detail added.

The factor loads are relatively stable in 1990s and 2000s – showing a value and quality tilt, and once accounting for exposure to market, size, value, and quality, the residual alpha is essentially zero i.e. nothing unique beyond exposure to these factors. The factor loads are less stable from 2010 with less value and more momentum (I note that the iShares fund tracking this index was launched in 2011). Why are the factor loads less stable since 2010? Possible explanations: (i) noise over a shorter time period, (ii) fund inflows (relative to 1990s-2000s) have reduced ‘value’ and added ‘momentum’ to the stock holdings, (iii) the 20 year period 1990-2009 was not a good representation of the long-run characteristics of the series. My sense, its some of (i) and (ii). When fund inflows settle, perhaps the long-term characteristics will be closer to 1990-2009 period than since 2010.

I have no real problem with a low volatility strategy, in that its just a combination of factor exposures (not unique, positive alpha beyond factor exposure). If these are the factor combinations an investor desires, all well and good. My interest in "low volatility" stemmed from reading Swensen’s view on diversification – i.e. a strategy that could lower volatility without the use of bonds (equity-like returns but with equity-bond type volatility). Re:

  • “By identifying high-return asset classes that show little correlation with domestic marketable securities, investors achieve diversification without the opportunity costs of investing in fixed income”. – from Pioneering Portfolio Management.
What I found is that there is nothing unique to a minimum volatility strategy beyond factor exposure (at least based on historical evidence). It hasn't added unique “Swensen type’ diversification beyond factor exposure. In all US back-tests I did, the historical mean-variance optimum was a small cap value tilted portfolio with bonds, over minimum volatility - but not huge differences.

Here's another example - the MSCI All Cap World Minimum Volatility Index had a similar 2008 downside to my 75:25 stock:bond portfolio, and similar returns, 10.7% vs. my 11.0% (2003-2013) - so owning bonds has not cost me much relative to an all equity minimum volatility portfolio, and the all equity portfolio did not yield higher returns (re: 'Swensen diversification'). In any event, an interesting single fund solution (if just going to own one fund). I'm staying the course with my global small cap value tilted portfolio.

Obviously no guarantees going forward on either. We each need to decide.

Robert

------------------------

FF mkt, size, and value, and Asness et al’s QmJ ‘quality’ factor is used for the analysis
t-stats in parentheses
  • Full Period (06/1988 – 08/2014)
    Alpha = -0.01 (-0.22)
    Mkt = +0.84 (48.23)
    Size = -0.15 (-6.84)
    Value = +0.19 (9.68)
    Momentum = +0.01 (0.50)
    Quality = +0.19 (6.05)
    R-squared = 0.92

    1990s (01/1990 – 12/1999)
    Alpha = -0.14 (-1.45)
    Mkt = +0.86 (31.74)
    Size = -0.16 (-4.52)
    Value = +0.18 (4.50)
    Momentum = +0.00 (0.10)
    Quality = +0.15 (2.55)
    R-squared = 0.93

    2000s (01/2000 – 12/2009)
    Alpha = -0.03 (-0.42)
    Mkt = +0.83 (32.76)
    Size = -0.14 (-5.59)
    Value = +0.26 (12.26)
    Momentum = +0.02 (1.22)
    Quality = +0.11 (2.84)
    R-squared = 0.96

    2010s (01/2010 – 08/2014)
    Alpha = 0.21 (1.17)
    Mkt = +0.76 (13.50)
    Size = -0.25 (-2.37)
    Value = -0.03 (-0.33)
    Momentum = +0.13 (1.94)
    Quality = +0.27 (2.47]
    R-squared = 0.83
Valuethinker
Posts: 48944
Joined: Fri May 11, 2007 11:07 am

Re: Minimum volatility - simply factor exposure

Post by Valuethinker »

How are we defining 'quality'?

Is it not the same thing as 'low volatility'?
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Minimum volatility - simply factor exposure

Post by jaab »

stlutz wrote:[...]His argument is that risk is and is perceived to be relative. Happiness comes from being better off than your friends/neighbors etc. not from reaching some absolute level of accomplishment. In investing, this means that "risky" = failure to beat to the market. To avoid this risk, one must pick investments with higher potential payoffs--i.e. high volatility stocks. In this framework, risk and behavioral explanations mesh into one. His approach may or may not be correct (I'm not sure, personally), but it's interesting and his book, "The Missing Risk Premium" is worth a read for those who find such material interesting. :happy
This is a good argument, similar to the one mentioned in Rekenthaler's recent posting at M*: http://news.morningstar.com/articlenet/ ... ?id=664536 I have to look more into that.

In general I find it pretty unreasonable that absolute financial return (% or terminal wealth) is claimed to be the only legitimate investment goal priced into the market. As far as I understand Falkenstein sees the market as pretty efficient, just not on a pure absolute return <> risk to that goal basis. Interestingly Fama's EMH does neither state markets being perfect (just good to best available guesses) nor that they are solely priced for absoute returns. Because EMH is not an asset pricing model. It just states that the available information is priced in, depending on people's preferences. For some reasons the behavioural finance school of thought does not seem to accept things like relative returns to others (peers, market) as a legitimate investment goal either - they are called irrational errors, mistakes, biases. A strange thing for economists actually, who like to accept people's utility functions as given. And even in the 101 homo economicus definition "rational" means optimizing for one's preferences, which include emotions, likes/dislikes, social goals and so on.

PS: as far as I understand "beating the market" is used for "better (expected) riskadjusted returns" with "same (expected) return for lower risk" as one possible application. Not just higher (expected) returns which a proportional increase in risk. This would be pointless - although salespersons will still use it this way.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

stlutz
First, I'm not arguing for investing in stocks with high vol and low short interest, likely to be highly concentrated portfolio. Though I certainly would argue to avoid stocks with high vol and high short interest. In fact I'd avoid all high short interest stocks. And I'd note that there are firms like DFA and Bridgeway that put on do not buy lists when stocks go "on special" in the securities lending market (lending rates are high due to high demand to short the stocks).

Second the past patterns should persist if nothing changes and there were good reasons for the pattern to exist. But very clearly low vol (and for example high div) strategy has morphed due to cash flows. Robert's data shows it was value oriented and no longer is. And I showed that as well by looking at the value metrics. It's the same as a high dividend strategy that has morphed away from a value strategy due to cash flows.

Also I just don't see any logical reason for low vol to work. If you do I would be interested in hearing the explanation of why, since there aren't limits to arb at work here (as there would be in small stocks--these tend to be larger stocks). Finally I would add its quality exposure that seems to be driver, along with value. But I at least have the same problem with quality as with low vol, there's no real good risk story (though you can come up with one). With low vol it's mostly about avoiding bad high vol, and same with quality I think, just avoid the low quality lottery stocks

But to each his own.

As to value premium, yes cash flows have shrunk the premium, this was noted in the blog posts I wrote on Claude Erb's papers on "overgrazing"---but he found that while this was true in large value, not so true in small value. And the premiums have shrunk, not disappeared---While premiums like for size and low vol can disappear, they shouldn't with value (though they clearly are time varying) because you are always buying something like the bottom 30% of stocks as ranked by price to some metric. Valuations matter, and there is nothing in low vol that is related to price. And without that metric there is nothing that prevents the premium from going negative.

Good luck
Larry
Beliavsky
Posts: 1233
Joined: Sun Jun 29, 2014 10:21 am

Re: Minimum volatility - simply factor exposure

Post by Beliavsky »

larryswedroe wrote:stlutz
Also I just don't see any logical reason for low vol to work.
(1) Lots of people judge performance by raw returns and won't give managers credit for achieving returns with low volatility.

(2) Institutions that will not or cannot lever may choose to get their effective leverage from high vol stocks.

(3) Minimum vol can mean
(a) choosing individual low-volatility stocks
(b) constructing a minimum variance portfolio
Doing (b) with thousands of stocks and getting sensible results is a nontrivial exercise.
Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Minimum volatility - simply factor exposure

Post by Browser »

If low vol is such a bad idea, why is Vanguard in that space?
We don't know where we are, or where we're going -- but we're making good time.
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Minimum volatility - simply factor exposure

Post by jaab »

Why has Vanguard 1/3 or so of AuM in active funds? :mrgreen: Don't think this thread says it's a bad idea. If it is a good one is another question, let alone if it is a superior or unique one.
Last edited by jaab on Tue Sep 30, 2014 9:58 am, edited 1 time in total.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

Beliavsky
Your points argue for avoiding high vol, not investing in low vol
If you see that a strategy's returns are explained by other factors then you should IMO go after the exposure to those factors, at least the ones you want. Reason is low vol may not be exposed to those factors in future--which IMO is clearly the case now.
Larry
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

Browser
Like many firms Vanguard responds to investor demands. And of course they MIGHT also believe low vol will continue. Only time of course will tell.
Larry
User avatar
steve r
Posts: 1299
Joined: Mon Feb 13, 2012 7:34 pm
Location: Connecticut

Re: Minimum volatility - simply factor exposure

Post by steve r »

larryswedroe wrote: Fifth, as mentioned low vol seems to have exposure to term risk. At least that should be taken into when deciding on term of your bond holdings. But low vol has benefited from bond rally.

And finally, while the data is impressive, I'm not convinced on quality going forward as the risk story is pretty darn thin. IMO it's behavioral and one that can be arbed away by cash flows, which we are certainly seeing.
Larry
First off ... this is a great thread.

I have noticed ... but not tested ... day to day bond yield rallies lead to under performance that day with minimum volatility (I do not watch low vol). This is not because min vol is tilted utilities ... it is not. I wonder if any empirical studies confirm this???

I agree the phenomenon is behavioral, but I have seen no evidence of it being arbed away. Knowledge of the phenomenon dates back to at least the 1970s. You claim we are certainly seeing this ... I wonder if you can tell me more.

Steve
"Owning the stock market over the long term is a winner's game. Attempting to beat the market is a loser's game. ..Don't look for the needle in the haystack. Just buy the haystack." Jack Bogle
User avatar
steve r
Posts: 1299
Joined: Mon Feb 13, 2012 7:34 pm
Location: Connecticut

Re: Minimum volatility - simply factor exposure

Post by steve r »

larryswedroe wrote:
I just don't find any logic to the low volatility strategy, and the evidence to me seems pretty strong it's other factors. And it's really all about avoiding high volatility, the well known lottery effect

You want less vol, use low beta and high tilt,

Larry
I would think if you want want a specific factor tilt -- tilt that factor. The idea of tilting low vol to capture value seems problematic. This was discussed above.

But, it seems the reverse is true. If you want low vol, tilt it. Tilting low beta or value or something else to get low vol seems problematic. I guess this depends on your belief in the theory behind the low vol anomaly.

How do we know value (and other factor) tilting is not simply capturing a low vol anomaly? Why must it be the other way around? I guess this (too) depends on your belief in the theory behind the low vol anomaly. I guess this is also the case because the value factor is far more established, accepted, and what not.

As far as tail risk, these indexes did relatively well in 2008 -- losing about a third less. This is after the indexes were established, but before ETFs existed.
"Owning the stock market over the long term is a winner's game. Attempting to beat the market is a loser's game. ..Don't look for the needle in the haystack. Just buy the haystack." Jack Bogle
User avatar
steve r
Posts: 1299
Joined: Mon Feb 13, 2012 7:34 pm
Location: Connecticut

Re: Minimum volatility - simply factor exposure

Post by steve r »

Robert T wrote:.
Here's the analysis from the OP broken down into 1990s, 2000s, 2010s, starting again with the full period with a bit more detail added.


FF mkt, size, and value, and Asness et al’s QmJ ‘quality’ factor is used for the analysis
t-stats in parentheses
  • Full Period (06/1988 – 08/2014)
    Alpha = -0.01 (-0.22)
    Mkt = +0.84 (48.23)
    Size = -0.15 (-6.84)
    Value = +0.19 (9.68)
    Momentum = +0.01 (0.50)
    Quality = +0.19 (6.05)
    R-squared = 0.92

    ....
I wonder if it would be possible to get some risk measure (annual std. dev. or other) during the periods for comparison sake.

I have seen data for the overall period, just not broken down. That is viewed favorably with the overall market. I am not sure with the specific factors (or factor funds).
"Owning the stock market over the long term is a winner's game. Attempting to beat the market is a loser's game. ..Don't look for the needle in the haystack. Just buy the haystack." Jack Bogle
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

Steve
First value is not low volatility, in fact value has higher not lower SD over the long term
Second, IMO if want low vol best way is to go low beta and higher tilt and use only safe bonds. Historically that has been the highest Sharpe Ratio. That's what's discussed in Reducing the Risk of Black Swans
Third, yes with low vol you have heavier allocations to some sectors, like utilities. Low beta sectors tend to be low vol.
Fourth, IMO we are seeing evidence of it changing by looking at the value metrics as I did, clearly it's changed. That's what Robert T showed.

Larry
User avatar
steve r
Posts: 1299
Joined: Mon Feb 13, 2012 7:34 pm
Location: Connecticut

Re: Minimum volatility - simply factor exposure

Post by steve r »

larryswedroe wrote:Steve
First value is not low volatility, in fact value has higher not lower SD over the long term
Agreed. And yet people constantly claim low vol is capturing a value factor. Hard to see the connection given value has higher SD and low vol are often pegged by M* as blend or even growth. Let me rephrase this, it is hard to see the "theory" why this will be the case ... (I can see the statistics and data).
larryswedroe wrote: Second, IMO if want low vol best way is to go low beta and higher tilt and use only safe bonds. Historically that has been the highest Sharpe Ratio. That's what's discussed in Reducing the Risk of Black Swans
Agreed with the bonds portion. But IMO if you want low volatility of equities you can get that portion more directly than low beta. You are presumably correct on the history.
larryswedroe wrote:Third, yes with low vol you have heavier allocations to some sectors, like utilities. Low beta sectors tend to be low vol.
Agreed. My question is more on the sector LIMIT weight models of min variance. Is that interest rate sensitive? I suspect it is but have found no evidence to support what I suspect. Today (sharp down day) is a case in point.
larryswedroe wrote:.
Fourth, IMO we are seeing evidence of it changing by looking at the value metrics as I did, clearly it's changed. That's what Robert T showed.
That is less clear to me. To see this you would need P/E and other metrics -- before and now. I see that now it is high now (but declining per STULTZ). For money to be moving in, it would need to start low PE an move to a high PE, would it not?
"Owning the stock market over the long term is a winner's game. Attempting to beat the market is a loser's game. ..Don't look for the needle in the haystack. Just buy the haystack." Jack Bogle
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

steve
But you are seeing higher P/Es now, relatively speaking, and that's the point. They used to look like value stocks, explaining the market like return with lower beta (got the value premium), but now they don't look like value stocks (so losing exposure to the premium meaning they no longer should be expected to get market like return with less volatility, just less volatility, perhaps)
Now they do exhibit exposure to term risk, and rates have fallen, so that should help them. And value underperforming this year, so that should help them. But longer term different story likely IMO
Larry
User avatar
Topic Author
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Minimum volatility - simply factor exposure

Post by Robert T »

According to the linked RAFI article - assets managed in low volatility strategies increased from US$9 billion in 2010 to almost US$60 bn by March 2013. http://www.researchaffiliates.com/Our%2 ... sting.aspx .

This article indicates that historically, low volatility strategies had a value tilt (except in emerging markets) and recently, the value tilt in developed market low volaitlity strategies has disappeared. This is consistent with the analysis I presented earlier in this thread.

From the article:
  • In developed markets, the low volatility group has historically tended to have higher earnings yields and lower book-to-price ratios than the market cap-weighted portfolio. However, over the past 10 years, the cheapness or “valueness” of developed market low volatility stocks seems to have diminished. As of May 1, 2013, the earnings yield and B/P ratio data indicate that low volatility strategies have become more expensive than the market cap-weighted core indices. In emerging markets, the valuation level for low volatility strategies has never been lower, relative to the core equity index—but emerging market low volatility portfolios are growing less attractive, especially considering that the benchmark index has fallen quite a bit this year
    Image
RAFI claims it adds an explicit 'value tilt' to its minimum volaitlity stratgies (as per the article), which maintained a value tilt in the respective series in 2013.

Robert
.
User avatar
Topic Author
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Minimum volatility - simply factor exposure

Post by Robert T »

Valuethinker wrote:How are we defining 'quality'?
In same way as Asness et al, as the above analysis uses their Quality minus Junk (QmJ) factor
http://www.econ.yale.edu/~shiller/behfi ... dersen.pdf
Valuethinker wrote:Is it not the same thing as 'low volatility'?
There is a component in the Asness et al definition of quality that relates to 'safety', other measures relate to companies that are "profitable, growing, and well managed". Companies with low beta and low idyosyncratic volatility explicitly make up 1/3 of the saftey measure, or 1/12 the overall quality measure - as far as I can tell.

Here are some return and standard deviation (volatility) comparisons:

1998-2013 - annualized return/standard deviation

11.4%/18.3 = Assness et al. Large cap quality (one of the components of QmJ)
12.1%/19.7 = MSCI USA Quality
10.3%/18.7 = S&P500
10.7%/14.9 = MSCI USA Minimum Volatility

The two quality series had similar (or higher) volatility (standard deviation) to the S&P500, and much higher standard deviation than MSCI minimum volatility. In this sense quality is not the same and "low volatility". Both quality series had lower downside in 2008 relative to S&P 500 (-30% vs. -37%), but less so in 2000-2002.

2000-2002/2008 losses
-37.2%/-30.0% = Assness et al. Large cap quality (one of the components of QmJ)
-33.9%/-30.2% = MSCI USA Quality
-37.6%/-37.0% = S&P500
-19.7%/-25.7% = MSCI USA Minimum Volatility

Robert
.
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

IMO if want low vol best way is to go low beta and higher tilt and use only safe bonds.
steve r:To translate Larry's recommendation a bit (I'm sure he won't mind :P ), what he means is that you should load up on small cap value stocks and since those are highly likely to exceed the return of the overall market, you can therefore take a higher position in safe bonds, thus getting the same return as you would get with a normal Total Market fund while lowering your overall portfolio risk.

He refers to SV as being "low beta" because they have lower correlation to the overall market. Everyone agrees that these stocks are more volatile than the overall market, however.

I thought it would be good to clarify for anyone who doesn't read all of the Larry threads who might come across this thread in the future.
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

RE: value and low volatility

Something we have to keep in mind is that if low vol has loaded on value historically, the reverse is also true--value has loaded on low. vol. And, interestingly, if you look at the F/F factor data, the value premium has been larger in down market years than in up market years, on average.

If low volatility now loads on growth and continues to do so long into the future, that also has implications for value tilters--they should now expect their portfolios to decline more, not less in future bear markets. And, since so much of the value premium has come during down markets historically, the future long-term value premium would likely be less if this bear-market-advantage did not continue to be the case.

However, we need to properly nuance all of this discussion. As Falkenstein shows in the posting I linked to earlier, the extent to which low vol loads on value has varied a lot historically. For example, for most of the 1980s, low vol. tilted toward growth not value. So, today's situation is not something new or unusual. Anyone who adopts a low vol. tilt should expect the same in the future--sometimes it will be growth and sometimes it will be value.

Personally, I don't think we've entered as world where low vol will permanently load on growth. I of course could be wrong.

Finally, it is also the case that low. vol. generally performs better at times that it has a value tilt to it. I suppose that opens the question as to whether one should therefore market-time low. vol. I'll leave it to somebody else to propose that approach, however.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

stlutz
He refers to SV as being "low beta" because they have lower correlation to the overall market. Everyone agrees that these stocks are more volatile than the overall market, however.
Well second part is accurate as SV about 70% more volatile.
But the first part is wrong. It's low beta because of the low equity allocation, low exposure to beta. Say a 30% equity allocation, and the equity portion has beta of 1. So beta loading of portfolio is just 0.3.
Now compare that to say a low vol portfolio with beta of say 0.7 if 100% equity to get same return.

Larry
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

stlutz
Something we have to keep in mind is that if low vol has loaded on value historically, the reverse is also true--value has loaded on low. vol.
This is simply incorrect. Value has not loaded on low vol, in fact value over long term has higher SD than the market, so higher volatility. And beta and vol while not the same are related. And beta of value portfolios are about 1.

The real issue is now low vol has lower expected returns relative to market due to lower value loading and same beta perhaps, so lower expected returns.

As to low vol, the cash flows drove low vol to become growth stocks, that's why they load on growth, not because companies were growthy, but the investors drove prices up so that prices made them growthy.

Larry
User avatar
alpenglow
Posts: 1798
Joined: Tue May 31, 2011 12:02 pm

Re: Minimum volatility - simply factor exposure

Post by alpenglow »

I'm enjoying this thread, even though I'm more or less a three fund investor. Anyway, I thought you might find this follow-up article from RAFI (9/2014) of interest.

http://www.researchaffiliates.com/Our%2 ... ffect.aspx
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: Minimum volatility - simply factor exposure

Post by larryswedroe »

epimedium
Thanks for sharing
The problem that I have with the piece, which is well written, is that it doesn't explain the low vol anomaly. Instead it explains the HIGH VOL anomaly, that high vol stocks (lottery tickets) have very poor returns---both for lottery ticket love (positive skew) and limits to arbitrage.
Larry
whitefish
Posts: 10
Joined: Sat Aug 07, 2010 4:27 pm

Re: Minimum volatility - simply factor exposure

Post by whitefish »

Robert T wrote:.
.... The min vol series has a signficant load to value and quality of the same order of magnitude (+0.19 load), has lower beta (0.84), and is large cap (-0.15 size load).
.
Hi Robert. Can you define the word "load" and how you arrived at the numerical values in your post?

Thanks.
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Minimum volatility - simply factor exposure

Post by jaab »

whitefish wrote:Can you define the word "load" and how you arrived at the numerical values in your post?
There are two easy to read articles by Samuel Lee about it. The first one explains the idea of factors as drivers of returns in general (with inflation, growth, liquidity as examples). The second explains the commonly used micro/fundamental models for stocks and what the math/numbers mean:
http://news.morningstar.com/articlenet/ ... ?id=636550
http://news.morningstar.com/articlenet/ ... ?id=636847 (3 pages)
User avatar
steve r
Posts: 1299
Joined: Mon Feb 13, 2012 7:34 pm
Location: Connecticut

Minimum volatility - RISING INTEREST RATES

Post by steve r »

One (other) concern in this space ... how it may respond in a rising rate.

My thought is that min vol (with ad hoc sector caps) outperform low vol (with heavy utilities)

Not a lot to go by, but the two years wear total bond market lost money with data (1994, 1999) low vol did under perform both years.

Being in such a fund in 1999 would be tough (as is the case with a lot of tilting strategies) ... min vol was flat, low vol down 7 ... market up 21 (of course the reverse happened in 2000)
"Owning the stock market over the long term is a winner's game. Attempting to beat the market is a loser's game. ..Don't look for the needle in the haystack. Just buy the haystack." Jack Bogle
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Minimum volatility - simply factor exposure

Post by stlutz »

One (other) concern in this space ... how it may respond in a rising rate.
Backtesting-wise, the strategy "worked" as expected during the 60s and 70s--i.e. there was no massive underperformance of the overall market.

Another way USMV has changed vs. a couple of years ago is that it no longer has a significantly higher dividend yield than that market.

I'd expect rising rates to provide a headwind, but I think a big jump in rates would provide a headwind to the market overall, so it's probably a wash.
whitefish
Posts: 10
Joined: Sat Aug 07, 2010 4:27 pm

Re: Minimum volatility - simply factor exposure

Post by whitefish »

jaab wrote:
whitefish wrote:Can you define the word "load" and how you arrived at the numerical values in your post?
There are two easy to read articles by Samuel Lee about it. The first one explains the idea of factors as drivers of returns in general (with inflation, growth, liquidity as examples). The second explains the commonly used micro/fundamental models for stocks and what the math/numbers mean:
http://news.morningstar.com/articlenet/ ... ?id=636550
http://news.morningstar.com/articlenet/ ... ?id=636847 (3 pages)
Thanks, jaab.
Post Reply