Factor Inversion :-(

For those who are interested, I just posted a research piece to my blog site:

https://actiquant.com/2018/11/08/factor-inversion-when-up-is-down-good-is-bad-dumb-is-smart-and-right-is-wrong/

Here’s a link if you want the full pdf version:

https://portfolio123dotblog.files.wordpress.com/2018/11/acti-quant-research-factor-inversion-110818.pdf

Thanks Marc.

Is the power gauge available on the site somewhere?

As participation wanes at tops (i.e. overvalued markets) and mega caps lead the market up, broad-based stock picking in general and value investing in particular become very difficult.

See 1999. Or 2018.

AQR’s long short market neutral fund is down 14% this year. I believe that one uses a leverage of 1.8. That’s a benchmark for how a market neutral quant approach fared.

Marc, thanks! You have been doing great work for years but this is my favorite article from you.

There are some pre-set ranking systems using Chaikin technical indicators (in the All Stars categfory) but Power Gauge, reflecting Chaikin’s own specific ideas as to how to work with the factors, is proprietary and available only on his subscription site at chaikinanalytics.com and/or via two Chaikin ETFs run by Index IQ (CLRG for large caps and CSML for small caps).

From an article on Bloomberg today about AQR:
"Asness and his colleagues argue recent moves haven’t negated their funds’ diversification benefits – they’re uncorrelated over the long haul, not hedges for day-to-day losses.

“We have seen relatively normal factor volatility, factor correlations and transaction costs,” said Ronen Israel, a principal at AQR. “There’s no evidence that the underlying themes are behaving in a way that is symptomatic of factor deleveraging.”"

What is factor deleveraging?

https://www.bloomberg.com/news/articles/2018-11-08/aqr-plays-defense-as-crisis-of-confidence-looms-for-quant-land?srnd=premium

Market neutral sounds great on paper, but it can be dangerous. When you get hit with factor inversion, you double the pain since you get to be wrong on two side of the market — you buy stocks that go down or underperform, and at the same time, you short stocks that go up or oiutperform.

I believe that’s what happend to AQR.

Thank you.

Hi Marc, really enjoyed your post. I wanted to mention a small typo: “hefty does”

Interesting observation regarding momentum outperformance vs. low vol being conducive environment for factor inversion.

Below are some charts I wanted to share on similar topic. These are from earlier this year (dates on files are May) so it’s not been updated for recent time period, but it speaks to the idea of potential factor decay and the rate of decay. The slides shown are on a small cap universe. Top 10pct of each factor shown. I do see the case that maybe we should expect the factors to work less well over time as the average rate of outperformance (measured vs. RSP, not top bucket vs. bottom bucket as you showed) tends to trend downwards for most factors.

Here’s some charts of some various combinations of Value, Quality, Sentiment, and Growth factors over time for the a small cap uni.



Are you really seeing factor decay? If you look at the charts from 2005 to now I don’t see any factor decay at all. All I see is a big difference between 1999-2004 and 2005-2018. It could be that 1999-2004 was a unique period for small caps, and that since then factors have worked intermittently like they always have.

While I wouldn’t out and out rule out the possibility of some genuine decay, my inclination is to agree with Yuval in his suggestion that 1999-2004 was the unusual period (factor bloat I might call it).

Also relevant is the consideration of small caps only. Personally, I don’t necessarily believe in size as an independent factor. i see it as being a member of the Quality family of factors (with small suggesting lesser quality) based on the structural considerations that make small caps behave as they do (greater operating leverage, separate and apart from whatever financial leverage might be present) and diminished operational diversification. So to me, the decay after the early years reflects a market that was less generous in rewarding risk taking.

Decay is real, but it is also definitely related to size.

Marc’s article motivated me to respond to questions regarding the decline and invertibility of factors. For these questions, I used the Fama-French datasets.

l found that value, profitability, and investment anomalies have decayed to essentially zero over the past 55 years, judging from the linear trendlines. While this decay has not occurred as rapidly as would appear based on the 20 years charts, it has resulted in the complete decay of the HML (i.e., 2 sized-ranked portfolios sorted by book-to-market) anomaly, which is often used a proxy for “value”. For example, the alpha on the 3-Factor equal-weighted HML declined by about 13 basis points per year since 1963. But 13 basis points times 55 years is about equal to its expected return intercept. Moreover, the rate of the decreasing appears to be increasing (from 1926 to 1963, there was essentially no decay in HML).

In addition, the trendline on the alphas of the return-based anomalies (i.e., momentum and reversion) portfolios have also declined to essentially zero, and much more rapidly so.

If current trends persists (and are assumed to be linear), it is conceivable that factors could indeed invert. I suspect, however, that the trends are not linear, but rather exhibit a power/exponential declines curves which are over the long term drawn to zero. While factor inversion could come about as a result of demand over-crowding, I doubt there is a reason why relatively low-priced stocks and outperforming companies would persistently underperform.

My intent is not to sow despair, but rather to set my own realistic expectations regarding just how difficult it is to generate alpha. Excess returns come only to those who either have some sort of arbitrage or other competitive advantage. For retail investors, since small cap stocks are more resilient to decay, one advantage might that our own small sizes allow us to gobble up pricing inefficiencies that institutions ignore.

FWIW:

I believe that factor decay is attributable to the leveling of the information battlefield (i.e., the transience of informational asymmetry). The data and my beliefs are consistent with the weak and semi-strong forms of the EMH.

But David, is it really “decay” when a factor works very well for a long period and then simply stops working for a long period, as is the case with the price-to-book ratio? What if there were some logical explanation WHY it stopped working? Or two explanations, or three, perhaps? First, there began to be an increasingly large number of funds predicated on “value” that used the price-to-book ratio as the basis for their investing. This, for obvious reasons, eroded the effectiveness of the factor. Second, as leverage became increasingly cheap, heavily leveraged companies outperformed unleveraged ones, so that if you compared two companies with roughly the same sales, the leveraged one (the one with the lower book value) would outperform the unleveraged one (with the higher book value). Third, book value is the portion of a company’s assets that constitutes its equity, which has the highest cost of all of its assets. The less a company has in equity, the greater control it has over its future, and more opportunity costs it can take advantage of. That’s one reason among many that stock buybacks have replaced dividend payments. Dividend payments don’t reduce a company’s equity, but stock buybacks do. Stock buybacks reduce a company’s book value, and increase its price, thereby nullifying the price-to-book factor every time a company buys back stock. And stock buybacks became a big deal just about the same time the price-to-book value factor stopped working.

If you were a Martian visiting the stock market for the first time twenty or thirty years ago, and you were a pretty smart Martian too, and you were to foresee that rather than placing an order with your broker by phone and having him charge you $50 for the order, everyday investors would be able to just type up an order on their laptop at any time of day and have it placed immediately, within seconds, for $5 or less, what kind of effect on momentum, reversion, and other purely price-based strategies do you think this would have? Exactly. It would make those well-documented factors more or less disappear. They would cease to be useful and reliable forms of arbitrage.

Factors don’t just decay because they get old. They’re not made of radioactive material. They don’t have a half-life. They appear and disappear for good reasons. I use a number of factors that were not documented prior to twenty years ago. I don’t know how long they’ll continue to work, but they’ve worked pretty well in the 21st century. I also use some older factors that have always worked. But I know the reason why they work. They make good, financial sense. And I know that when they stop working, there will be a good financial reason for that too.

One such reason is over-exploitation. The greatest investors have always looked for unexploited arbitrage opportunities. Some of those can still be found in the stock market if you try hard enough. Once they become widely exploited, the greatest investors stop using them. And maybe that’s what you and others mean by “factor decay.” But there are plenty of other reasons that a factor might stop working besides over-exploitation.

Excerpts from Yuval:

YES!!!

Let’s remember what a factor is. It’s a convenient description of something from the real world. That’s all it is; a description. Thinking about a factor without thinking about the reasons is like reading a review of a movie that doesn’t exist. It’s the movie that counts. The review is just a description and commentary that may or may not be accurate or well reasoned.

Let’s also remember that that Fama French were just data mining. Nothing in their papers suggest any understanding at al of why something may or may not work and as I said in the paper, their dividend factor research was a complete mess because their hypothesis was completely illogical. Value is always logical and it’s the responsibility of investors to understand when low ratios will “work” and when they won’t. You get none of that from Fama French and the like. It comes from financial theory. Value must and will work if the so-called value stock you choose is priced too low relative to future growth and/or quality. A stock with a P/E of 100++ can be cheap (ask Bill Miller the noted value guy who invested a truckload in AMZN way back in the early days and was right because the stock did, indeed, turn out to be dirt cheap relative to future growth). Conversely, a stock with a P/E of 8 can be way overpriced, something unappreciated by sell side analysts who sent me hate email after I panned some homebuilder stocks on Reuters back in 2006.

The value factor is just a convenient description that may or may not be useful depending on business and market conditions. Value investing, something very different, can never be wrong (unless one is willing to pay $10 to purchase a $1 bill). Sometimes it’s easy. Sometimes it’s hard. Sometimes we make correct choices. Other times, we get it wrong. We’re human and the world is complicated. But logic is logic and is not impacted by any of that.

Ditto for any other factors.

Primus, don’t despair. The Fama French type game may be decaying as the market shifts away from a 35+ year valuation regime (driven by falling interest rates) toward an emerging earnings driven regime (accompanied by some other regime changes including shifts in attitudes toward risk that once made a small-cap effect seem like a bona fide factor even though its really just a subset of quality/risk). Nobody can ever be expected to be hot all the time, especially at a time like the present, when regime transition is in progress and everybody needs time to recognize and adjust), but by aligning with financial logic, you can still accomplish quite a lot. (Correction: It’s easy to outperform now: Momentum, momentum and more momentum. The question is whether one feels confident and agile enough to jackrabbit in and out quickly enough considering the way these stocks get super hot and then super cold so suddenly. I’m not in that category. But for those who are, go for it. You might want to try timing models based on trends in relative performance of style-oriented ETFs relative to one another. I did this on my blog on 10/12 with MTUM vs USMV. But my analysis was backward looking. I haven’t yet developed something that looks to be predictive. But if you check that post, maybe it’ll inspire some research ideas.)

All of the talking heads in the investing media seem to be focusing on ‘quality’ now.

Good call, Marc.

My high quality/low beta port is doing fine so far this year, as is my high quality/dividend growth model.

This is a great read. Thanks.

FWIW, Patrick O’Shaughnessy just released a 90 minute interview with Cliff Asness on his Invest Like the Best podcast, and they hit a lot of these topics. A good listen. AQRs research seems to believe this type of scenario, where there’s not a particularly strong factor that can prop up all the ones that are out of favor, generally happens every 10 years or so. It’s unpleasant to go through, but nothing really atypical about what is going on historically.

I’m here because I read Marc’s Factor Inversion article and found it to be one of the best things I’ve read all year. That, and I need to marshal any and all resources that help me explain to my investment committee why our equity portfolio is underperforming.

Thank you.

Your need, explaining to investment committees, is a perfect example of what motivated the project. (I’ve been dealing with a similar situation lately.) We’re saying the right things but I want to put evidence behind it to combat the impression that we’re just whining or making excuses.

I’ve also decided to take it a step further since this need is not going away any time soon, if at all. I’m working on a comprehensive Factor Almanac (free PDF resource) that will give information and comparisons for each factor for each month from January 1999 through the present. I already have the data. Now comes the work of organizing and presenting it.

Ultimately, my goal is to make this topic more rationally discuss-able, to get away from the prevailing atmosphere that treats equity investing as series of a short-term races between a portfolio manager and a benchmark and that treats instances of underperformance as incompetence or dark forces.

Hi Marc,

you have demonstrated that indeed very well. However, with backtesting we all fail to make good models for the future. We look at a mere 19 years of data and analyse each factor for what worked in the past and then combine the best performing factors to what was the best strategy in the recent past. None of this helps to build robust models for the future.

What I ask myself is what strategy to follow to be fit for the future. For instance:

  • what sectors / industries are likely to outperform in the future? Along these lines Steve has designed his cloud computing R2G which has faired quite well.
  • what will be the next crisis and which companies will be most affected by it? Many seem to agree that corporate debt is becoming the next big issue, so staying with debt-free companies might be a safe strategy.

Lastly, there are times where no factor will save us from a big drawdown, hence market timing is always an important component. Looking at my 8-component market timer, we are very close to drop below 50 (i.e. stay in cash).

Florian


Hi Florian,

I wouldn’t put it quite that way. I’ve always stressed the importance of knowing WHY our test results are what they are and not necessarily extrapolating anything into the future. In fact, in some recent Seeking Alpha articles, I wrote about screening ideas and was candid in showing that backtest results were weak – bit I explained why I think the future will be different from the past. Now, with the market transitioning from a falling-interest-rate-valuation driven regime to a rising-rate-earnings-driven regime, it is especially important t be willing to refuse to extrapolate. Do you remember, for example, a couple of years ago when I suggested in the forum that p123 should remove TLT (the 20+ year Treasury) from among the list of allowable timing-hedge vehicles? Do you remember the vehement pushback I got from member who cherished TLT because of backtests that demonstrated its efficacy as a great equity alternative (some of who were especially vehement not realizing that I was joking)? Now we know. Rates were, indeed, poised to break away from the long downward regime and as it turned out, TLT has been and still is a horrible hedge vehicle for use with real money (although long-sample backtests continue to suggest otherwise). Here’s a little mathematical formulation I just concocted:

My belief is that an understanding of how factors work in theory (that’s the P/E = 1/(R-G) framework I’ve been plugging) and sharpening our ability to relate this to what goes on in the real world (that is the goal of the sort of study I did and plan to do more of in the future) helps us understand why our test results are what they are and improves our ability to make assumptions regarding how they’ll play out in the future.

The idea of a sector or industry having such-and-such prospects is equivalent to a statement along the lines of: This group of companies has important business characteristics in common and as a result, there is likely to be a set of core expectations that the companies will be similar in terms of important factors. Whether or not that group will do well going forward depends on the extent to which the market will or won’t reward strength in those factors going forward. (We know, for example, that the market has lately been rewarding and paying up for expectations of future growth – in cloud computing and elsewhere.) Whether one says "I want cloud computing"or “I want strong showing is such and such factors,” we’re getting to the same place. Thinking in terms of the story (cloud computing) is actually a pretty good way of tuning into how the factors play out in a real world setting and can be a valuable way to help one fine tune an understanding of the past and expectations about the future.

That’s a completely different thing. Factors (at least the sort of factors we discuss in terms of stocks) necessarily presume a focus on equities. A highly successful factor, one that, for example, separate best from worst be 10 percentage points, won’t make you happy if the market as a whole dropped 50% and you wound up with minus 45% as opposed to minus 55%. You still had an undesirable drawdown. So when it comes to timing, you’re in a different area that requires judgement and/or a different model geared toward this particular task; different kinds of factors that grow from a different framework (not P/E = 1/(R-G).