Sunday, March 17, 2013

DeMuth's Affluent Investor Highlights Hindsight Bias

I received a copy of Phil DeMuth's Affluent Investor, which at around $11 is a bargain. Alas, advice books like these, for the layman, are kind of futile, as good advice is ignored by those who need it precisely because those who need it would never seek it. So, it's best as a gift. If there's a 10% chance they can learn, it's worth $11.

For example, he nicely informs the reader what investment advisers think is wealthy: $100k to $1 million is 'affluent', $1MM to $10MM is 'high net worth', above that is 'ultra high net worth'. From a practical perspective, one should target getting 25 times one's income when one retires, probably a good definition of 'comfortable.'

He makes the interesting point that gambling is fun and savings is hard. But savings, investing, and gambling, are all different shades of grey, and highlights that something may seem the same at one level (not spending) is very different at another (buying Tbills vs. lottery tickets).  Another good point is that some stocks, like utilities, have generated profits and dividends for decades, and some like Facebook are predicated on something unconventional, monetizing something novel. These should be thought of as totally different, yet, these are both 'stocks', and classed the same (FB)

He seems to be plugged in to the big names in finance, and so co-authored several books with Ben Stein, a well-know author and commentator (and son of famous economist Herb Stein), and seems to know the Peter Bernstein crowd (eg, Black, Scholes) pretty well.

Interestingly his book mentions low-beta investing in  a couple of pages, and even includes this blog as charitable reference, which I greatly appreciate. Alas, I don't agree with his portrayal of the low-beta phenomenon.  He promotes an alternative history, one consistent with Frazzini and Pedersen's betting-against-beta theory. The basic idea is that low beta stocks outperform high beta stocks on a risk adjusted (but not absolute) basis, and this is because of borrowing constraints.

DeMuth mentions Myron Scholes, Peter Bernstein, Perry Mehrling and others, in creating his narrative, that these guys were big early believers in the Frazzini and Pedersen model (before these guys were born), but were thwarted by regulation. That is, all the big players knew low beta stocks outperformed since 1969, and activily advocated levering up low-beta stocks. In practice, however, this aggressive leveraging is hardly, if ever, observed, as a result of which the volatility effect can persist, because people are leverage averse. That is, the big players knew about the low volatility investing anomaly all along (just ask them!).

True, Black did show early on why the Security Market Line (SML) is too flat because of borrowing constraints (see Brennan (1971) and Black (1972)) . Facts are important, and Black missed the big one here, which is the SML is flat--not insufficiently upward sloping--but flat (at best). A flat SML doesn't generate a rational reach for high beta that underlies the Frazzini and Pedersen (and Black, Scholes, Mehrling, etc.) model, because you don't prefer the higher returning 1.5 beta stocks if, actually, those stocks don't generate higher returns.

 That is, 1.5 beta stocks don't outperform 1.0 beta stocks, so saying investors prefer them because they have higher returns means their beliefs are irrational. It's like normal form game where everyone is wrong about the other guy. This is general not considered an equilibrium, and so, back in my day, nonpublishable. Times change. I suppose this is the mainstream academic's (eg, Fama, Campbell) favored explanation.

The low volatility (aka low beta) anomaly is not merely that low beta (vol) stocks have higher than expected returns, it's that their returns are higher than average. Just look at the performance of Acadian, Robeco, or Analytic, which have been running low vol funds since 2006ish.

If borrowing constraints were key, we would see a risk premium where they are not applicable, such as futures, but we don't. We also see the opposite in options, where the negative risk premium seems simply related to volatility. Saying this is consistent because investors believe they are amplifying positive returns, while in fact they lose something like 20% per expiration, implies an insane dichotomy between expectations and reality. For this reason I think Frazzini and Pedersen or Scholes are wrong to say this all fits in their theory, because their theory is based on everyone acting on a different model than they assume: maximizing returns personally, while assuming everyone else is maximizing variance-adjusted returns. Usually, the dichotomy is someone thinks everyone else is acting irrational, here, the agents act irrational and assume everyone else is rational.

In sum, I don't think it's a good explanation now, and historically, Scholes or his colleagues didn't ever discuss the low volatility anomaly for 35 years or so, which makes me think it wasn't something they thought was important, relative to their other ideas over that period. For them to now insinuate they documented the low volatility anomaly back in the 1970's is typical post hoc rationalization you find in regular people.

7 comments:

Anonymous said...

Eric, you really have to stop saying this stuff about the distinction between a smaller sloped SML, a flat one, and a negative one. It's interesting to think about but unlike what you say it does NOT distinguish between theories well.

a) Nobody says there is only one explanation driving this. A great flaw that pops up again and again in finance (and presumably in many things I don't pay attention to) is that we will always find one explanation (e.g., for another finance example, a strategy is either risk or behavioral not some of both and that it can be time varying across them). In this case the multiple arguments for low beta stocks might each have degrees of truth, making your bright line test a pure straw man. Of course that means your favorite explanation can have merit too, but not based on this argument. This argument is based on the sum of the effects, we will need finer tests to really distinguish (if possible and remember it can vary through time) between explanations. FP start down this road by looking at when leverage is more restricted. That is at least directionally what we need to distinguish more between explanations.

2) You must know that the t-stats on the line being flat or negative are far weaker than it being not sloped enough according to the CAPM (by definition). You are thus relying on much weaker statistics than those behind the general success of low vol investing in making your assertion that the line is flat/negative, and the assertion itself is still not dispositive even if true as described above.

3) The evidence outside of individual stocks is in general for a flatter not flat SML (like you claim it should be), and the evidence for stocks itself is not as strong for "flat or inverted" as you imply (of course there are periods). It seems we shouldn't favor one arena over the other and the overall evidence for a flat or negative SML across all the tests (stocks, yield curves, credit, countries, etc.) is pretty consistent with a too flat but perhaps not all flat or negative SML.

I am a fan of your theories too, again, I don't think they are mutually exclusive with ours, but you seem to and I think you're wrong.

So, I'd humbly ask you to modify your comments to "I think my theory is one part of why this works" and not denigrate others!

-- Cliff (though signed in anonymously as I usually fail when I try the others!)

Eric Falkenstein said...

Well, I am co-authoring a paper that surveys evidence and I think we fairly present the other explanations (my co-authors don't prefer my pet theory like I do). Clearly, I have strong opinions, but most people who come up with theories do, as even F&P state their theory explains everything from Buffet to Options. It's OK to see one's theory everywhere as long as one is willing to listen rationally to alternatives (as per seeing it everywhere, my kid was watching the 'Diary of a Wimpy Kid' movie, and the protagonist goes through all sorts of drama as his status, listed as an ordinal rank in his peer group, is shown to fluctuate with various adventures; it's all relative).

I agree that leverage constraints affect asset prices, but in this case (performance of low-vol stocks) I don't see leverage as the driver because one should see stronger SMLs in markets where leverage is easy, or a capm beta that generalizes to other asset classes or instruments. For example, it's straightforward to create a strategy that has a high beta with the US market using dynamic trading rules on futures, yet the returns of these strategies are pretty orthogonal to the beta (ie, the beta is not a price risk factor in general).

I'm sure the SML can comport with any prior, but internationally, the SML is flat at the monthly horizon as a point estimate. The Total Return or geometric average, highlights a negative SML in most countries. I doubt, over the next 40 years of my life, we'll have data any more definitive even if it merely continues as is.

I'm not sure which assets contain a positive SML outside of the short end of the risk-free yield curve, and AAA to BBB bonds, but I think this is more due to the cash-value (eg, hypothecation) of the one side as opposed to risk.

As for other theories about this I'm partial to, I think Meir Statman's work on how people invest is pretty good. That is, people don't optimize so much as set aside X% for risky investments, and then buy assets that can change their life like only a high vol investment could.

I should say, I contacted him about my take a while back and he basically responded with an indifferent shrug. Convincing other people in this business, those with their own ideas, to like your ideas on the same subject, is improbable. Nothing personal. I'm not denigrating so much as criticizing (eg, I think Fama is just an awesome guy). Clearly, I'm much more opinionated than an Antti Ilmanen, but that just highlights the value of true diversity.

But ultimately the facts are ambiguous at some level, and one is applying prejudice quite a bit, all the time. My bias is to have strong opinions, weakly held.

Anonymous said...

Falken, you need to learn to let other people be wrong.

Is it not clear to you that as your ideas become accepted, history is revised to exclude your contribution - while ensuring that the prestige of the high priests is maintained? F em.

Let them be wrong and stop worrying about it - you will become a much better trader.

Anonymous said...

Eric,

I'm not sure why the geometric return is relevant aside for an investor who doesn't diversify into anything else but this one factor, and (an overlapping explanation) doesn't rebalance if they do diversify. There is a reason the literature uses arithmetic returns to study these things, and with low beta in particular using geometric biases the results.

Even using USA stock data and value weight portfolios (BAB results) the SML is barely negative long-term, and the t-stat against half the predicted SML slope is not lower than -2 (and t-stat vs. zero miniscule). You have no statistical power showing its negative but your basing an argument on it, and that argument doesn't even apply if this is the sum of several affects.

Once you accept that these theories might be applying at once all the arguments are kind of silly as it's an argument about a sum. You need a totally different type test to distinguish between arguments, not anything about the sum of all the effects (the stuff about dynamic betas didn't make much sense to me on first read but may be in the right direction).

But, regarding the supposed negative SML, you have no real evidence, and on the geometric return, you're using the wrong data.

I suggest you stop defending your one particular theory. It makes your arguments weak as they are designed for a purpose not truth.

I am far from sure our leverage aversion story is the ONLY argument and don't feel the need to "defeat" other arguments which may also be contributors. If I had to bet I'd say ours explains more, and you'd say otherwise, but that's probably just our respective biases.

And don't listen to the third anonymous dude, he's on drugs.

-- Cliff

Eric Falkenstein said...

The geometric return is not as obvious as most academics believe. I think there are long-run mean-reversion (or cointegration) forces that generate the omnipresence of that negative SML in that space across markets. Consider, many (25%?) Equity Risk Premium studies have a geometric return estimate as the focal point.

There's a lot of other data that's consistent with my theory and no other, namely, info on the total return to equity investors (which shows a zero risk premium), anthropological data on envy vs. absolute wealth, modeling logic on robust preferences, how no characteristic (eg, courage) has a linear expected return in any other dimension (x vs y), and lastly and most importantly, the absence of a risk factor that generalizes.

But why I strictly prefer my theory is that the others all require either ad hoc constraints (eg, the 60-40 equity/bond rule of thumb binds, leverage aversion) or imply that arbitrage is obvious and ignored (Why do so many buy high vol stocks, which have really low Sharpe ratios, both directly or in their indices?). For years, when I was initially discovered that low volatility stocks did as well as high volatility stocks, and tried to publish this result my explanations were dismissed because professors and reviewers thought any result that implied arbitrage was unlikely. This was pre freakonomics/behavioralist popularity, but I came to see it a valid objection, and still do.

Anonymous said...

I like your story and those kind of arguments (not just data but other logical implications and causes) are just the kind of thing I think our field needs more of (I'm not signing off on every one!).

I do think the closer something gets to an entire portfolio (e.g., equities) the more geometric becomes legitimate and the more something is a "tilt" the more arithmetic is legitimate. Since I would tilt to more things than just low beta (e.g., value, momentum, etc.) I greatly prefer arithmetic.

And, again, my argument has never been anti your story, again, I like it! I just think it's nearly impossible to distinguish between some very good essentially additive stories.

For instance, when the efficient market guys start fighting with the behaviorists, I enjoy pointing out that both might have some share of the truth, and it might vary through time. For example, the value premium can be part risk party anomaly, but all anomaly in late 1999. That makes it very intractable, and no side wins, but very good chance it has some truth...

-- Cliff

Eric Falkenstein said...

Good point about value, which I agree isn't obviously behavioral or risk based; that's a case where moderation works in the sense it still isn't obvious. Yet, think about the buyers of value or small cap: most think it's behavioral, because that's why higher return segments dominate; if they truly thought it was a risk-based story, the return premium would be boring (simple trade-off). For those who don't like value, I would guess they find it a risk result. Knowing how to pitch these guys, and show they are being rational, is valuable.