The Week That Was
One word: Wow!
The human tendency to perceive meaningful patterns within random data.
….nothing is so alien to the human mind as the idea of randomness.” –John Cohen
FinTwits is just full of things like:
When A has occurred the S&P500 has been higher/lower at: 1 day, 1 week, 1 month, 3 months, 6 months, etc.
Where “A” is some esoteric metric far from the main in technical analysis.
These observations are an attempt to be constructive either by demonstrating a correlation or a lack thereof. I can’t say with any certainty that any of those correlations are wrong. But neither can most of these people prove they’re predictive.
For the most part, though, I see this stuff as an effort at legitimate quantitative analysis. I’m sure the people doing this work think they are. It’s not and, in fact, it’s usually just a terribly bad misuse of statistics.
Absent in most of these data sets is any substantive regression analysis, confidence intervals or even a mention of a standard deviation. You know, the stuff taught in Intro to Statistics classes.
The larger error in my opinion is that quantitative analysis most often demands multivariate correlations for predictions in complex systems.
I double-dog dare you to argue that the capital markets are not complex systems.
Missing are things like time-varying covariates, correlation to an alternate principal component, and confounding and/or interacting variables – just to name a few.
Don’t get me wrong. There are some simple models that show higher correlations to future market movements. Most of these are well known such as extremes in the put/call ratio or tested indicators like the McClellan Oscillator. For the most part, though, they’re only predictive of short term moves and all of them fail from time to time.
Good quant is very hard. Why do you think Wall Street hires physicists, mathematicians and engineers? I’m pretty sure it’s not from their prowess in single variable regression analysis.
The really smart people would even dismiss the multivariate regression approach and use Chaos Theory. They tell us that the market is too dynamic for anything but “chaos” to prevail and forecasting needs to be approached by models used in quantum physics. Really heady stuff.
Chaos: When the present determines the future, but the approximate present does not approximately determine the future.
Think of this graphic when looking at how co-dependent variables interact and the randomness it generates. Now, complicate that with a hundred variables. (h/t Wikipedia.)
When a butterfly flaps its wings…
For the record, I have a bare minimum familiarity with Chaos Theory. I do have, however, significant training in statistically based quantitative analysis. And the only certainty one gets from post graduate study of QA is how error prone – or perhaps I should say “imprecise” – it is; as in “There is an X% probability of Y happening with a Z% confidence interval. Or, as Nate Silver might explain “Why polls fail.”
That said, it’s much better to have some model especially if you’re watching for when it fails.
But having a single variable correlation model might be worse than having nothing.
There, I said it. You can all commence with telling me how I’m wrong. But I’m not. Unfollow me. Vote me off the island. Throw me out of the band. Whatever.
I just calls it as I sees it. And I call most of this worthless. It’s just throwing chicken bones.
What Value Is Studying Economics?
Let’s start with the fact that I am not now nor ever have been an economist (imagine my serious voice.) Let me also say that I have been seriously studying economic theory most of my adult life and have, for perhaps 30 years, spent on average an hour a day reading economic research and academic papers. I have found that there are very few subjects at the Masters degree level that I’m unfamiliar in macro economics. The outcome has been: the more I learn the less I know. So why do I do it?
1 – Simplification. Thinking like an economist helps me reduce complex problems into manageable models. We can see correlations much better if we properly isolate variables. It helps us build a context for how we imagine how things work and gives us a baseline from which to test our understanding.
2 – Beauty. I came to the conclusion at a young age that I would never be a giant thinker. I have, though, always been in awe of minds that have countered and changed popular conventions (my older brother was force feeding me Bertrand Russell when I was 12.) The elegance of thought from history’s great economists, as with the great philosophers and writers, is just simply beautiful.
3 – Dear Old Dad. My father has minors in economics at the Bachelors, Masters and Doctorate levels. While I was in college he constantly poked at my economic and financial thinking with a continuous (and most annoying) Socratic dialogue. I’m sure that he enjoyed having at least one of his three sons who aspired to beat him in a debate. He played my ego and that caused me to learn – and then caused me to question what I thought I had learned. He still does it and is the most effective devil’s advocate I’ve ever known. I use him to bounce investment theses off regularly. It is one of the few intellectual fortunes I have in my life. And he gave me a life-long curiosity for economics. Maybe it’s become a character flaw. It’s too soon to tell.
Buy the way, if you’re in the investment arena, nothing will serve you better than an honest and qualified devil’s advocate. If you don’t have one, find one – or ten. This is Ray Dalio’s approach.
What Economics Can’t Do
1 – Predict the near term capital markets. Although the markets respond to economic data they are, in fact, looking backward. Sure, certain data reinforces assumptions about longer term trends but those movements are reactive instead of predictive. I’ll get to the problems with economic data later.
What economic data primarily does to the markets is reinforce or upset sentiment. Some people can trade on that. I can’t. But investing on a single data point is just slightly less than completely insane as far as I’m concerned. If you’re a trend follower, apply that discipline to economic data. Understand too the glacial speed of economic trend changes outside of demand/supply shocks.
2 – Predict the economy. All we have to do is look at the best economists from the worlds central banks to prove this. The old joke among economists is “I’ve predicted 9 of the last 4 recessions.”
3 – Be correct. It can come close but it usually doesn’t. That is to say, like business forecasts, there is always a level of error.
Economics: It’s the data, stupid
Almost all government economic data is a statistical estimate. Real-time data can be exceptionally noisy. One standard deviation for the Non-farm Payroll report is roughly +/- 100,000 jobs. What does a beat or a miss of 50,000 jobs mean? Statistically nothing.
GDP estimates are a much more difficult task. The inputs for the GDP estimates are based on a set a surveys called the National Income and Product Accounts (NIPA.) The Bureau of Economic Analysis (BEA) understands the statistical error potential in their estimates. Observers are familiar with “advance”, “preliminary, and final quarterly GDP reports. But the BEA, being the home of world class statisticians, also does this:
Each quarterly estimate is subject to three successive annual revisions … The first annual revision incorporates further revisions in the monthly or quarterly source data and introduces some annual source data. The second and third annual revisions incorporate a broad range of annual source data. Each quarterly estimate is also subject to one or more comprehensive revisions, in which information from the economic and demographic censuses is incorporated.
In other words, we don’t have even the best guess until three years after the “final” quarterly report is delivered.
I don’t want to overstate the problem. It’s actually remarkable they do as well as they do. Nonetheless, it’s still statistics. So let’s revisit this:
Chaos: When the present determines the future, but the approximate present does not approximately determine the future.
We, and most of the developed economies, have reasonably well developed statistical processes. Thus, from a broad brush perspective, they are generally “directionally correct.” That said, they’re not very good at predicting trend changes. Most recessions are already in process by the time they are identified.
For the developing world the data is harder. China, India and Indonesia, for example, still have very underdeveloped and crude measuring processes. Consequently, all economic data from any of those countries is much more suspect than that of the developed world.
Human nature, of course, usually doesn’t lend itself to considering quality constructs of the data outside from where we live. This is the “hometown bias” in data interpretation. Ergo, there is a tendency to project the quality of one’s domestic data onto foreign data. That is a rather big mistake. Keep that in mind.
How I Use Economics In Investing
I think it’s important to stay on top of economic data as part of one’s investing discipline. But I think it equally important to understand its limitations. We get fooled by complex analysis thinking that complexity adds understanding. But I have to think that F.A Hayek was right that complex econometric models usually are just a pretense of knowledge. He referred to that as “Scientism“. The balance between over simplification (from above) and over-complexity is very hard to attain.
I had a finance professor who said “It’s an unfortunate fact that balance sheet balance because it can sometimes add a sense of legitimacy where none exists.” Economic models are much worse.
To be frank, I don’t rely on macro day to day but keep it in mind as a backdrop in my investment decisions. That said, when I use it I often get it wrong. I was convinced at the end of 2013 that industrial commodities had bottomed. Guess how that worked out.
But there are some things that it can be quite useful for given a long enough time horizon. For example, demographic trends will determine things like housing demand, generation specific goods and services consumption trend, etc.
I do rely on micro economics often. Understanding industry cost structures, price elasticity and regulatory cost shifting for example can help identify value opportunities.
The advice I would give to young investors is to focus on micro. It’s not the sexy stuff of which macro is made but it’s where your knowledge will actually give you an edge. You wanna look smart or do you wanna make money?
A lot of what we do in developing investing theses really are no better than a Zulu fortune teller casting chicken bones. That shouldn’t stop us from trying to see the future but we should understand that what we think we know may be imaginary. To paraphrase Socrates, wisdom is the understanding of our own ignorance.
That’s why in investing it makes sense to have the default position to be “I could be wrong”. That’s why keeping and eye on and managing to the risk of a permanent loss of capital is every bit as important as seeking opportunities. Some would say more so.
I’ve decided I will no longer share my trades on social media. There’s a handful of reasons:
1 – regulatory compliance issues
2 – trolls
3 – A client asked me why he’s paying me if I’m giving my trades away to strangers? Good question?
OK, that’s it. Troll away.
Keep in mind, though, no matter what you do, what process you have, what data you think is valuable, I really am pulling for you. As we all know, this business is hard and all of us make mistakes.
Trade ’em well.
ps – comments are open and always welcome but due to the high level of spam I screen all commenters. Once I have approve your first comment all future comments will be approved by default. So if you comment give me a little while to see it an approve you. I get to it sooner or later.