Quantitative and Technical Analysis

Shhh! Don’t tell anyone. Late at night, when nobody is looking... I’ve secretly been reading books on Technical Analysis! Don’t tell the CFA Institute, or they might revoke my standing! (disclaimer: that last point was a joke)

It is interesting to come out of the closet as an agnostic about technical analysis, since many of my colleagues who practice quantitative methods or do the CFA curriculum point out that numerous academic studies have found questionable value in TA. The CFA level I curriculum has approximately 1000 pages dedicated to financial statements analysis and another 500 devoted to fundamental equity valuation techniques. These are no doubt extremely useful and much needed in the curriculum, but the attitude of the CFA Institute to TA is indicated by its paltry 100 pages or so dedicated to technical analysis, complete with the conclusion that it probably doesn’t work. Also, I don’t recall a single question related to TA out of 240 questions given on the actual Level I exam. When I mention that I’ve been investigating TA, I can feel my friends and colleagues are looking at me as if I’ve just lost 20 IQ points. It’s as if Isaac Newton’s investigation of alchemy meant that you couldn’t take the theory of calculus seriously (note: Isaac Newton did study alchemy).

My interest in Technical Analysis (henceforth TA) has several drivers. The immediate issue had to do with my Global Macro view that gold was a good investment. This was a fairly clear choice: immense pressure on the dollar from budget and current account deficits, concerns about inflation and energy prices, the concentration of wealth in the United States, and increasing security problems around the globe. When I lived in Brazil, if I saw these things happening there, I knew to get my assets out of the local currency. The US Dollar is a special case, as the world’s reserve currency, but the current administration is also in a class (of irresponsibility) of its own, which requires reviewing this special case more critically. These are the types of moments when fundamental assumptions about the way the world works tend to break, and so the macro opportunities are potentially quite good. So gold was (and probably still is) an excellent investment choice for a US based investor, but the main problem was that I had few tools to determine a target exit price for the commodity. Typical quantitative analysis demands a target price (i.e. expected value), and fundamental tools attempt to compute one, but are designed to work with equities or debt instruments. Technical trading rules allow one to base an exit on changes in price movements without necessarily predetermining an exit price (exits being different from stops, which exist for risk control).

I realized that the exit price problem is a common macro decision dilemma, and that for many people, placing global macro positions was really more like a trader’s framework than the mindset of a fundamental or quantitative investor. I had never considered the trading mentality myself, because - in my mind’s eye - I imagined traders as type-A personalities glued to a trading desk and yelling into phones on an exchange floor. This was not a good fit for my personality, which I see as more of the analyst-observer-actor type than the get-in-your-face-and-force-that-trade type. After reading Steven Drobny’s Inside the House of Money, however, I realized that much global macro work is done within a trader’s framework more than a portfolio manager’s framework; the main difference is that position cycles are usually executed over longer intervals, and therefore require less yelling over telephones.

...I should point out that there is a more quantitative approach to global macro strategy, called tactical asset allocation (TAA or GTAA). This strategy involves having a model that can predict short term deviations from long term expected returns (predict is the key term here), and therefore inform asset class or sector over- or under-weightings with respect to the long term strategic allocation, usually with the use of an optimizer. But we didn’t have a model that could output a target price for gold, so the GTAA approach was not viable....

Another motivation was my sense that large portions of quantitative community had gotten a little too smug with the superior performance of their models. On the one hand, many simply point out that “TA can’t be shown to work.” These guys obviously managed to get through the 100 pages of CFA material on TA and absorbed the conclusions. On the other hand, others have said “yes, we use technical indicators: we find that momentum helps our quantitative models.” But as far as I could find, momentum was the only technical indicator that quants had decided might have some predictive value. Now admittedly, funds don’t tend to publish their models as widely as Nobel Prize winners like Eugene Fama and Kenneth French, so if there were more technical indicators in quantitative models, they would be hard to find. (Goldman Sachs’ Mark Carhart is known for adding momentum to the Fama-French model successfully).

Now, one of the reasons that technical analysis isn’t supposed to work in quant-land and CFA-land is the “efficient markets hypothesis” or EMH. There are several versions of the EMH that range from weak to strong. The weakest form is a form that says “any information that past price data can offer you is already included in the current price of an asset (else it would have been acted upon already).” This effectively translates to “don’t bother looking at past price data (other than to estimate volatility),” and “charting techniques don’t add value.” It basically asserts something called Markov Property in mathematics: if a process has the Markov Property, only the most recent element in a sequence contains any useful information about the future. The Markov Property does help make statistical models easier to specify. Applied to asset prices, it also sounds like the traditional economist’s joke that there can’t be a ten dollar bill lying on the ground, else someone would have already gotten it: therefore no ten dollar bills can ever be found on the ground (consequently, nobody bothers looking, and it soon becomes obvious that there are no ten dollar bills on the ground, because no one has ever seen one... “excuse me, sir, how will you be taking your Black Swan tonight?”).

Having heard that technical analysis doesn’t work, that minimally efficient markets should preclude technical analysis from working, and then hearing many of the same analysts argue that momentum adds value to quantitative models makes me feel that someone isn’t “‘fessing up.“ For momentum to add value, the Markov Property can’t possibly hold, and therefore even the ”weak form“ of the EMH doesn’t seem to be weak enough. And, as someone who has experience working with market failures in environmental provision, political meddling in markets, and oligopolistic distortions, the idea that markets are efficient or ”only occasionally inefficient“ seems simply implausible to me. There have got to be inefficiencies, and these inefficiencies are potential sources of comparative investing advantage for those who understand them (not that I claim to understand all of them). And we know, instinctively, that when market conditions change, many people - if not most - need time to come to a decision about what that means for the price, and even institutions with professionally trained investors have organizational processes and constraints (e.g. risk management policies or staff) that preclude ”instantaneous simultaneous adjustments by all actors to incorporate new information.“ To me, this means that the main theoretical reason that technical analysis can’t possibly work is undone. The real question is, how long a window does the average investor (or professional investor) have to react to new market data and can this window permit any forms of TA to add value?

Now, I certainly think there may be good TA and bad TA, just as there is good statistical analysis and bad statistical analysis in quantitative work. Personally, I have always been suspicious of TA explanations that sound like ”Crouching tiger on lonely mountain scares market bull by morning mists,“ but what I have found from my reading of technical indicators is that there are genuine causal mechanisms being hypothesized for many technical formations in terms of what market participants are actually doing. Now, maybe these hypotheses are wrong, but it does bring TA out of the realm of astrology and witchcraft into the realm of social psychology, which is a field that has a scientific methodological tradition. Most of the popular quantitative techniques for valuation rely on equilibrium models, which have a compelling logic of their own, but they are surprisingly devoid of causal-structure more complex than ”participants desire to make a buck, and might use arbitrage to do it.“

It seems eminently plausible that a diversity of valuations and valuation techniques - not to mention varied institutional investing decision processes and constraints - could make buying and selling pressures ”lumpy“ over given price ranges, in the sense that the interaction between buyers and sellers is not a smooth or even a monotonic function of price. If supply and demand pressures are indeed lumpy, this would lead to non-linear effects on price changes, and these price changes would themselves affect future valuation. Consider this: the urgency with which one considers selling an asset may depend on how quickly its price has dropped. If others may also be feeling urgent needs to sell the asset, because they suddenly are not sure that they understand their investment or their risk, then the asset value at the next interval is likely to be smaller still, even though - in theory - the next price change should be independent of the last price change. Which scenario here has a more plausible cause-effect mechanism? Price changes are independent of each other because of equilibrium, or price changes alter people’s confidence in their investment process. Even in highly quantitative shops that trust their risk models, the difficulties that quant funds faced in august caused many PMs to dial down their tracking error because they no longer trusted the valuations. When they did that, portfolios were unable to recover as much as the market bounceback would suggest they had.

Where this all leads is that there may be non-linear determinants of price changes (i.e. the ”lumpiness of supply and demand“) that, in turn, affect future valuations, which, in turn, affect future price changes. This is a classic situation in which physicists and mathematicians would expect to find deterministic chaos. What then might TA charting patterns be? Very possibly, head-and-shoulders patterns, double-tops, double bottoms, wedges, etc. are the strange attractors that come out of this non-linear, recursive, lumpiness in prices subsequently affecting valuations. Different attractors correspond to different semi-stable states of the market, and very likely reflect different psychological states of market participants that affect how lumpy price ranges are. Over a large range, these prices probably are constrained by general equilibrium limits, which would explain why theories like CAPM or the APT methodology (note that APT is really a methodology, and not a true theory) appear to be minimally plausible (and also why they probably perform better at longer time frames that smooth out short-term disruptions).

In any case, I am relatively new to technical analysis techniques, and I am sure there are plenty of silly ones out there, but I am not convinced that this ground should be dismissed as quickly as the quantitative and fundamental communities do. There are very defensible reasons to think that at least some of TA could be valuable.

As I proceed further, I promise to tell you more about what I find (though I may not publish my TA model either). I may indeed eventually come to the conclusion that TA doesn’t have any value after all, but even that would be a valuable conclusion if that’s where this leads.