Is factor investing dead?

Research
1

I have tested the various rating systems against a number of universes, and it seems that for the most part, it is the sentiment indicator that does it best. To some extent, this is confirmed in both Zack Rank and AAII 5% Rev Up, both of which are sentiment systems.

The other Core investing systems do not seem to provide much excess return. The study: “Alice’s Adventures in Factorland: Three Blunders That Plague Factor Investing”, also shows that the factors have not worked well since 2003. The study also shows that the factors work even worse in periods when the stock market rises sharply (such as now), so it may be part of the explanation. ( Alice’s Adventures in Factorland: Three Blunders That Plague Factor Investing by Robert D. Arnott, Campbell R. Harvey, Vitali Kalesnik, Juhani T. Linnainmaa :: SSRN )

Does anyone have any views? Is it mostly just Sentiment that keeps on working?

Factor.jpg
Factor.jpg703×714 166 KB

2

Factor investing is alive and well. (The problem is that you have been brainwashed as to what factors you should use and how you should evaluate them.)

I have a 4-factor growth ranking system that I developed two years ago. I first applied it to the constituents of the SKYY ETF. The 10-stock port (SKYY-Light) has been running since the start of 2020 and is up 375% compared to ~80% for the SKYY ETF. Trades for this portfolio have been fully documented since Jan 1, 2020 at Seeking Alpha.

In addition, I have recently released designer models, all of which are based on the same 4-factor ranking system. Most of these DMs are “piggybacking” off of ETFs, something that people here are denouncing. See graphs below.

SKYY-Light.gif
SKYY-Light.gif1012×518 22.7 KB

DMs.gif
DMs.gif2326×1120 111 KB

3

At least in backtests sentiment is best hands down for systems more liquid than micro-cap strategies.

One can question how much of that is due to look-ahead bias. But there is evidence that it works without look-ahead bias.

I do think sentiment works best with certain fundamental factors.

Zacks does have good out-sample evidence but they are light on details on how Zacks rank has been working recently or whether combining Zacks rank with, say value factors, affects the returns (or how).

Len Zacks had a book about many anomalies that I have read. A good book and one Marc recommended once. Without minimizing the importance of all of the other anomalies Len Zacks says this:''Earnings estimate revisions are the most powerful force impacting stock prices."

Times change. Zacks rank may not work as well as it once did, but probably fundamental factors do not work as well either. I think Zacks still may have it about right.

I leave it to others to discuss micro-caps and if you see an opinion on that above, please let me know so that I can edit it out.

Jim

4

Of the core ranking systems, the sentiment one has indeed worked best. But keep in mind that it’s a very high-turnover system. Almost all the other core systems show much better performance for the 90-100 decile than for the 0-10 decile, so they’re working as they should, even if they’re not giving you huge excess returns.

I would advise you to play around with the other core systems to see if you can improve their performance by eliminating certain factors, adding others, and changing various parameters. For example, momentum factors tend to work better with a one-month lag (e.g. rather than measuring 12-month momentum as the price change from one year ago to today, measure it from 13 months ago to 1 month ago). Quality factors work better if you also focus on earnings quality and the stability of fundamentals. Low volatility works better if you take share turnover into account. Growth works better if you look at consistent growth rather than just recent growth, and if you modify certain factors to punish companies that grow too fast. Core Combination works better with the addition of size factors. And so on.

My own personal experience tells me that factor investing isn’t dead. I have a CAGR of 46% since I started using factor-based ranking systems in late 2015, and I’ve beat the market every year except for 2019. And sentiment factors account for only about 7% of the weight of my ranking systems.

5

Thanks for all the input.

The problem with making too many adjustments to the Core systems (or primary factors) is to make sure I do not over-optimize them for the last 10 years.

I have also tried to look through some of the strategies and screens that are available but there are few that hold up well when tested against (besides Yuvals “Zoo”):

  • different universes
  • different time period
  • out of sample
  • different capsize
  • and different number of positions

I also do not want to throw myself into buying designer models, since the strategy is secret and there is too little learning effect in it.

Do I understand that most of you mostly use ranking systems and to a lesser extent specific buy rules, outside the criteria that govern the price of stock, volume or capzise?

By the way, below is the strategy that AAII uses in its “top 30 revision up”:

There are more than four analysts providing earnings estimates for the current fiscal year (Y0)
The latest earnings per share estimate for the current fiscal year (Y0) is greater than it was one
month ago
The latest earnings per share estimate for the next fiscal year (Y1) is greater than it was one month ago
There has been at least one upward revision in the earnings estimate for the current fiscal year (Y0) over the last month
There have been no downward revisions in the earnings estimate for the current fiscal year (Y0)
over the last month
There has been at least one upward revision in the earnings estimate for the next fiscal year (Y1) over the last month
There have been no downward revisions in the earnings estimate for the next fiscal year (Y1) over the last month
The top 30 companies are those that have had the 30 largest percentage increases in the current-year consensus EPS estimate over the last month.

6

I think this supports what Yuval has said above.

I like Olikea’s Designer Models. He is wicked-smart and highly trained: degrees in physics but probably a professional quant somewhere now. He has left some models with some good ideas. Models that have no resent survivorship bias with considerable out-of-sample data.

His best models are micro-cap strategies. Check-out his models and see if you agree.

As near as I can tell, if you want outsized returns at P123 you go micro-cap or heavy sentiment. Some of this is based on anecdote and personal bias no doubt.

But once P123 addresses any PIT earnings estimates issues , P123 can rightly claim it is the place to be for micro-caps (whether one is using sims and/or machine learning) and for better-than-Zacks sentiment models.

Jim

7

Thanks, again, I’m going to examine Olikea’s strategies. I am reasonably new here, so it takes some time to understand the coding and the platform, but I have learned a lot in a short time.

For me who invests from Norway, it is an advantage if I stick to liquid and preferably mid-large cap shares.

I have looked with great interest at Yuval’s strategies, blog, and webinars.

8

I rely on ranking systems with very few buy/sell-rules, and I’m quite happy with the performance. I also recommend going through the latest webinar series by Yuval & taking a look at Yuvals blog, it’s a goldmine of useful information, as is the forum.

Also: Hello fellow Norwegian! If you invest from Norway (as I do) I don’t think you have to avoid microcaps or other not-very-liquid stuff, but you absolutely have to avoid very high turnover. In addition to Us smallcap I’ve found that Canadian or ex-US stocks listed in the US, of any market cap, works reasonably well. The most challenging corner of the marks seems to be US large cap, I’ve yet to find a ranking system that can beat simple indexing for this segment.

9

Great thread! Apologies if this is long winded, but please hear me out.

I’d agree with much of what has been said, i.e. factor investing works, no question. The catch - you have to put the work in to see just how and when these factors work. I have been at quant investing for about 5 years, the first 3 of which were filled with mistakes, some luck, more mistakes, overconfidence and I lost money. I’ve learned a lot since then, particularly from the fantastic P123 community.

As it turns out, my latest project has been to decompose the returns of these Core systems by time, specifically by business cycle stage.

One of the most controversial topics in investing is “market timing”. Conventional wisdom states that you can’t time the market; even if you exit in time before a crash, you will run the risk of missing the recovery. Yet you can also read about the investing greats who wait for their moments to take key positions, or in other words, they are timing. Seth Klarman, for example, is famous for keeping ample dry powder just for this purpose.

Having been investing for a while, I’ve noticed that some of my strategies have worked exceedingly well at some point, only to crash and burn 1 year later. This is enough evidence for me that there is something to timing your strategy(ies) to the market.

That said, I have gone down the rabbit hole of decomposing all of the P123 Core Ranking Systems by business cycle stage:
• Recovery
• Growth
• Overheating
• Recession/slowdown

I’m working on a new Seeking Alpha piece just on this, but it’s taking some time.

The objective? Can you just “buy and hold” a strategy and achieve market beating returns “over long periods of time” like conventional wisdom holds (and particularly over the last 20 years, as some research, like what you linked, is suggesting you cannot), or do certain factors/strategies work particularly well during a business cycle stage(s)?

This research is a long process, I am still collecting data, but the results so far are making some strong arguments for timing these factors, instead of just holding over long periods.

For example, over “long periods” small cap value does not perform very well. However, during the recovery stage of the business cycle? It outperforms every other factor. In realtime, my small cap value strategy has returned 175% during the most recent COVID recovery, since May of 2020 until today (IJS returned 122% in the same period, and SPY 60%). All of my other strategies are up YTD and are beating their benchmarks by a wide margin.

I’m still compiling and designing strategies for business cycles and more, but it is promising. Sneak peak - there seems to be specific periods during the cycle when low quality seems to outperform; I have a theory why, but still testing.

Will post my Seeking Alpha article once it’s published.

Cheers,
Ryan

10

I would not say Factor investing is dead but it is very hard. Which means your odds of success are very low. The vast majority of designer models fail or they are very high turnover. If it was easy the designer models would be beating the index.

Cheers,
MV

11

I started a micro-cap strategy based on this thread. So I must believe micro-caps can work.

But Mark has an important point I think. A point discussed in the book “Thinking Fast and Slow” and in a lot of books really. A point about the “Base Case”.

Often authors writing about the base case (or business schools) talk about restaurants. You may be fresh out of business school (or chef school) and think you have a great idea for a restaurant. Before you start construction on the restaurant these writer recommend that you look at the “base case.” In this case it would be how many restaurant start-up success there are out of all of the restaurant start ups.

You start with the base case and then add additional information like: everyone loves my recipes and they are bound to be a hit. Or my parents will back me up with additional funding if it takes longer than expected to start making a profit. Or experts on TV say we will have herd immunity after the Xth booster shot and I will not be shut down.

One can always find an example of a great/successful restaurant. That does not mean I should go into the restaurant business (probably I shouldn’t). Generally, the odds will be against you. Things like having great recipes does not change the odds as much as you would think.

For investing you can add your own personal factors that make you think you will do better (or worse) than the base case. For P123 the Designer Models can be the base case. But not always, I think.

One could consider getting away from the Designer models as your base case. You might consider not doing what most Designers have done, I think, which is endlessly optimize factors and perhaps overfitting the model.

What is the base case for Georg’s idea of taking an ETF and trying to improve upon it (i.e., piggybacking)? The base case would be the return of the ETF and not the average Designer Model results, I think. Maybe you will not improve on the ETF as much as you think you will out-of-sample. Maybe you will end up with the base case: the return of the ETF. Maybe that would not be so bad.

Does Steve Auger have a different base case sometimes. A base case that is different than the usual Designer Model at times for his models? Maybe I think. But I’ll say just maybe hoping Steve will tells us what he thinks about that.

So I am starting a micro-cap model. But I do not think the designer models are the appropriate base case for my model. I do not think anyone should be foolish enough to do what I will be doing but I do think Mark makes an important point about the base case. And perhaps we can change the base case from the designer models: each member doing it in their own way.

Obviously, if you have already built a great restaurant (trading model) you should pretty much ignore what I said above and keep doing what you have been doing. But then again, maybe the usual base case never applied to you in the first place.

Jim

12

First, Designer Models started off on the wrong foot. There was a huge subscriber base, feeding off of showcase simulations, each one more astounding than the previous, with uneducated subscribers not understanding that backtest is not the same as forward performance. The end result was quite predictable. With the first downturn everyone scattered with the wind, with the DM platform becoming a ghost town. That is what you need to understand when scrutinizing the track record of DMs.

Second, there is an underlying (incorrect) assumption that there are golden factors that work on any time frame, on any stock universe. This started originally with P/E ratio, then P/S ratio, now we are on to DDM, DCF, Piotroski, etc. The idea that DMs are set and forget is ill-conceived. Markets change along with the factors driving success. Portfolios/models have to change with the underlying currents.

I don’t need to justify what has been done by everyone else in the history of Designer Model offerings. I only have to justify what I am doing and that is available for everyone to follow along and watch. If it is not your cup of tea then don’t follow along. But I’m getting pretty tired of hearing the same regurgitation of “factors no longer work” while at the same time discounting what does work. You know the old saying: “I can bring a horse to water, but I can’t make it drink”. THis is a perfect example.

13

Steve, I think you know this. But I want to be sure that you understand that my comments were meant as positive. Or at least meant to say, that some (or perhaps all) of your models should not be put into the mold of your average Designer Model :slight_smile: Jim

14

I understand this Jim, thanks! The problem is that people think that there should be golden factors that work in all timeframes and markets. With today’s technology and market data, that is unlikely. We need to get beyond the “one size fits all” if we want to be successful.

15

This was a very good discussion, and enlightening.

Let me first say that I do not mind DM, but it is probably the desire to first and foremost learn this. And I think I do, by trying and testing and asking questions, and trying and testing again.

That said, I also agree, I do not think it’s a strategy or a factor that always works in all universes, but I’m probably still surprised how otherwise known factors have performed, and some have done poorly in long periods. Then it is probably correct as it is said, that we have to adjust certain elements in a factor, but then the question quickly arises whether we over-optimize

Finally, it was probably also a point for me to show that I was surprised at how well the sentiment factor works on different universes, time periods and Capsize and portfolio sizes. This seems to be supported by the fact that the best strategies on AAII and Zacks are variations of sentiment

16

Which sentiment factor are you referring to? Sentiment means different things to different people. For some, it is price-related. For me, it is short interest.

The thing about DMs s they are independently tracked on an out-of-sample basis. So you can’t have someone dazzling you with incredible simulations like occurred back in the early days.

17

I am thinking primarily of “Core investing styles - Sentiment” and “Basic: sentiment”.

18

Ryan -
You might find this review of recent literature on the topic helpful: Factor Timing Is Tempting -

19

When the Alpha Architect ETFs first came out, I did some backtest simulations to see how the ETFs would have performed prior to launch. Needless to say, I wasn’t impressed with the results. The timing was way too slow/ineffective. I believe that my conclusions were validated by the performance of the ETFs since then. Only QMOM is outperforming the S&P 500 and that is marginal and the price chart is highly volatile. The timing was not effective. The remaining ETFs have substandard performance. IN short, Alpha Architect “talks the talk” but that is about all.

AlphaArchitect.gif
AlphaArchitect.gif2168×821 294 KB

20

Thanks Yuval, good read. Not particularly promising results from Alpha Architect’s review, but still drilling down into my own research. Also breaking down each factor by size, and using a leading indicator rather than just looking at past regimes in hindsight. We’ll see.

Steve, agreed the Alpha Architects ETFs have not been very impressive; the link Yuval provided is their summary of research by others (timing factors by business cycle stage).