Bloomberg bashes quant investing . . .

My response to Bloomberg:

https://actiquant.com/2019/01/22/bloomberg-says-quants-arent-adding-up-i-say-bull-if-theyre-legit/

The Bloomberg link’s title:

“Black-box investing strategies supposedly based on rocket science are vulnerable to a host of inherent biases, which are now being exposed.”

Self contradictory. If they are truly black boxes then what exactly is being exposed? This would be okay if the article were to actually tell us what Renaissance Technologies is using: it does not. Instead, the article complains about the continued Black-Box nature. Throws out the names of a few potential biases that would be addressed by any serious quantitative investor (albeit not always successfully).

In the end, it gives us no factual information about what is actually being used by various funds or how those biases may actually be affecting any of the fund’s returns.

Not really rocket science either. Me, I will insist on 100% chance that Mars (or the moon) will be there when my spaceship arrives. Not the same kind of math for “Rocket Science” or physics. Who would use machine learning in rocket science? Maybe use a random forest to determine the probability that a satellite will reach orbit. Really?

The article only gets worse. The rest of the article has almost no real information.

More a comment on what people call Journalism now days than anything.

Marc is being kind by reacting.

BTW, P123 isn’t a black box like some quantitative strategies/algorithms can be. And it does not require a brain surgeon (let alone a rocket scientist).

-Jim

Enjoyed the article, Marc!

If I could I sum up Marc’s article in a few words, it would be “correlation versus causation”. I also feel that “pattern recognition versus rational expectations” is an appropriate description. Marc is basically arguing that many of the self and media-proclaimed quantitative investors are simply quantitative modelers without an appreciation for investing fundamentals that are based upon rational expectations. Marc adds that those fundamentals can all be derived from a simple annuity model (e.g., the dividend discount model, the Gordon Growth Model, etc.). While I would like to agree with Marc’s perspective (based on my deeply-held hope that the purpose of markets is to seek fair prices based on rational expectations for the discounted sums of the present value of future cash flows), I still believe that the Bloomberg article brings up several valid criticisms of quantitative investing.

The biggest issue I think is informational asymmetry. According to Marc, “All reputable quants approach this in their own respective ways and most typically produce white papers, etc. explaining their respective processes.” In my view, the rational expectations approach to investing depends on the existence of informational asymmetry-i.e., someone knows somethings the others do not. Whether that is unique information or an unique modeling approach is immaterial, because there is no advantage if everyone is acting on the same information and premises. For example, even if we just accept that the modeling approach is useful, common inputs to the perpetuity/annuity model are increasingly based on shrinking informational asymmetry (i.e., programmatic access to and analysis of standardized financial and estimate datasets is increasingly ubiquitous).

Furthermore, the rational expectations approach fails to explain at least some statistically persistent anomalies, namely momentum. While I understand that this criticism is susceptible to the joint-hypothesis problem (i.e., it is impossible to say whether anomalies refute the theory or are the result of spurious correlations), I think that the behavioral economists have increasingly been able to provide satisfactory explanations of the behavioral bases of many market anomalies which are altogether independent from explanations rooted in rational expectations theory.

On a more granular level, I think additional problems stem from over-simplifying the economic theory of net present value into a simple annuity model. This is nitpicking, of course. As they say, all models are wrong, but some are useful.

Hi David,

I cannot find anything to disagree with in your post or with your ideas from previous posts: certainly nothing that I could prove or would argue for.

But who is more supportive of the above statement? Marc is arguing in SUPPORT of quantitive investing. The Bloomberg article argues AGAINST quants.

Marc has his own ideas and ways of doing quantitative investing and I certainly am not going to argue for or against his particular methods. But HE HAS USED MOMENTUM in some of his models (I know this because I have copied some of his formulas).

So I see Marc supporting quants. If he prefers rational factors then that may make sense from a Bayesian perspective were we have different priors. Perhaps we should assign greater credibility to those things that make sense or that we learned in a graduate course.

As for the Bloomberg article: it is just a bunch of terms (including terms for biases and for problems faced by any investor). Hard to have any opinion—other than facts are missing—about this article.

That is not to mention the fact that Marc has helped developed the PREMIER QUANT SITE IN THE WORLD (that is open to everyone): P123.

Did I misunderstand the Bloomberg article? I do understand if you like different factors than the ones Marc does. I have my own preferred factors/formulas too and I doubt that Marc would use some of them. We are all here to pursue our own ideas so this I DEFINITELY understand.

But seriously, isn’t Marc the one arguing FOR quants? At least to a greater extent than the author of the Bloomberg article?

-Jim

I’d like to take issue with this. Information asymmetry is very different from “a unique modeling approach,” and that difference is not immaterial. If you favor companies with high profit margins and I favor companies with low profit margins, we are both acting on exactly the same information, but we have very different modeling approaches and are going to buy very different stocks. (There is, by the way, a good reason to favor companies with low profit margins, because they’re the ones that are most likely to experience earnings growth, if one assumes that sales drive earnings.) One quant investor can have an edge in how he interprets data, another can have an edge in the data she gets.

I also differ with you in your statement that the rational expectations approach fails to explain “statistically persistent anomalies.” Marc himself explained momentum well: it’s a sentiment indicator. I’m not quoting here, just summarizing from memory: a sound company that is growing and is well-run is more likely to experience a rise in price than an unsound company that is falling on hard times and is badly run; and companies in the first category are more likely to persist in exceeding expectations than companies in the second. Because the “momentum anomaly” is so weak, statistically speaking, and because mean reversion is actually more prevalent if you look at things probability-wise, if only five or ten percent of companies fulfill those two roles (sound and unsound), that would be explanation enough. If an “anomaly” (by which I mean a factor) is statistically significant and persistent, I would think that there is a rational explanation for it.

Just wanted to be clear that the only criticism I had of David’s post was I could not see why he supported the Bloomberg article. And I still cannot see how the article is not trying to make a point that opposes his (David’s). But in the end I think the article fails to make a good argument for any point so I might have just let it drop if I did not like Marc’s general support for Quant investing.

We can all come with examples of were observations came before theory and vise versa. Both in an out of finance.

I guess Kepler’s theories came before Galileo’s observations of the phases of Venus. But then again Alexander Flemming’s observations came before anyone figured out how penicillin worked or even had a question about antibiotics.

Einstein’s theories must have come from God himself and are separate from any discussion here. Actually, the Mickleson-Morley experiment and the unexplained anomaly (before Einstein came along) was first.

In finance you can find examples of both. And sometimes there is more than one theory that nobody is entirely certain about—often created to explain an observation.

Personally, I am going to stay away from absolutes and/or specific examples in finance. Because, honestly, I think both views (either view) can be correct depending on the example. Arguing for either point being absolutely correctly all of the time would be worse than a waste of time.

-Jim