This is part two of an exploration of the Lucas Critique and how patterns cease to be patterns once they are known. This work is inspired by a single page of the our featured book, Fooled By Randomness by Nassim Nicholas Taleb. In part one, we started with the following quote:
We are human and act according to our knowledge, which integrates past data. [Therefore] If rational traders detect a pattern of stocks rising on Mondays, then, immediately such a pattern becomes detectable, it would be ironed out by people buying on Friday in anticipation of such an effect.
Which is to say that reliable patterns are highly coveted in stock trading. Traders quickly seek out and capitalize on these patterns the instant they are known. Such adaptive responses are what make these markets “efficient” and “irons out” the anomaly. Information advantages do not last long.
But who cares, right? I’m not a trader and you probably aren’t one either.
I think the point, as better explained in the Lucas Critique, is this:
You can’t use history to accurately predict what a change today will do in the future.
Additionally, if you want to prevent something from happening, predict that it will happen. I write that with a mild bit of humor but there is real truth to this on a macro level.
Paranoia and Pandemics Prevented
In 1997, the H5N1 bird flu virus emerged in Hong Kong. Over time, 846 people from 16 countries were infected and more than half perished as a result. The terrifying mortality rate coupled with the virality of the disease was more than enough to warrant serious measures from the World Health Organization (WHO) and the Center for Disease Control (CDC). This included travel advisories, outright bans, mass destruction of poultry stocks, and response plans for first responders in the United States. Vaccines were distributed to emergency personnel across the country. A new, useful technology called the internet became a clearinghouse for information on the evolving situation. And, of course, the media picked up the story.
For weeks, the media monitored the death toll and made allusions to the 1918 Spanish Flu pandemic. It was reasonable to imagine the rapid spread of such a deadly disease in an age of increased air travel. A disaster was forthcoming. Be prepared.
But a disaster did not follow. Why?
Some might say there was no disaster because the H5N1 virus was never the threat it was made out to be. That the media trumped it up for attention. Just as they would do again with SARS and Y2K and other things.
Or, of course, it could be that the abundance of caution and immediate response by the WHO and the CDC stanched the spread. The media’s coverage probably led to an increase in healthy hand-washing for a week and an avoidance of all things poultry related. This surely helped, too.
A bit of paranoia built on the initial prognostication that this virus could lead to a pandemic on the scale of the 1918 Spanish Flu is evidence of the Lucas Critique in action. Once a prediction was made with strong correlation to the past, it increased the believability of the threat to such an extent that we humans became irrationally protective.
That seems bad. Even today, experts try to dispel this paranoia and misunderstanding on the disease. But to borrow from the title of another fine book by Andy Grove (last week’s featured author): Only The Paranoid Survive. Maybe this well-intended hyperbole is a good thing on the whole.
It makes for bad prediction but good behavior. This time.
Have We Seen This Before?
When disease, climate, econometric, and other such forecasts are built on historic data, we presume the forecaster must know every vital thing that happened last time and expect it to happen again. This forgets the dynamic nature of current events and shadows the fact that past events had been dynamic, too.
To return to meteorologists, we would never believe a forecaster who said, On January 23, 2018, the temperature was X. Therefore, for today’s forecast, the same day of the year 2019, the temperature will again be X.
But my initial take on the Ray Dalio prognostication from yesterday was just that. I thought he was saying the history is repeating itself:
The history never does. Not really. But prediction mistakes certainly do.
Could it be that Dalio’s prediction actually prevents a downturn a’la the media’s prediction of a pandemic? Could this sort of story be what changes behavior and lessens the impact?
I don’t know. You can google stock market trends right now and find people who say yes and those who say no. At this point, it’s a matter of belief more than fact. And as our author for this week’s book would say, there’s a certain amount of randomness to the whole thing.
Dentists versus Economists
In yesterday’s article, I channelled my inner Taleb to offer some unique distinctions on probability and distribution between areas that are either small-scale and skill-based or large-scale and variance-based. Dentists cannot scale their talents but rock musicians can. Thus, there are a lot of dentists making very good livings; they’re all winners because no one can take it all. And there are very few rock musicians making very good livings because popular music scales and a few bands get all the attention while the rest do not.
Scale changes everything.
Next, I tried to lay out the idea that a dentist can make marvelous predictions on your future cavities because they have seen thousands (millions?) of cavities before and the mouth, for all its complexity, is a much more confined system than, say, the planet. And the planet’s weather patterns. If a dentist can make oral health predictions with a 99% accuracy and a meteorologist can only hope for an 80% accuracy on weather predictions, it isn’t because the meteorologist is deficient. She is simply working in a much more dynamic field.
Now consider the economist. There is a marvelous quote from one of the greatest, Paul Samuelsson, that reads:
The stock market has forecast nine of the last five recessions.
This is so funny to me. And why does this joke ring true? I think partly because of cycle times, limited observations, and complexity of open systems compared to closed ones. But it also has so much to do with this Lucas Critique in that the stock market prices are often, invariably, built on expectation which is built on history. Which is not good.
The history doesn’t repeat. The mistakes do.
We’ve Never Seen Anything Like This Before
Speaking of the economy, I listened to an interview this weekend where someone observed: this recovery is a bit long in the tooth. This is in relation to the economic recovery from the Great Recession. Consider the idea …
This recovery is long in the tooth? On what basis? History? History is a trap in this instance. Basing future expectations on past events is a higher form of nostalgia, a taste of the sweet, familiar “member berries”. It makes for fine conversation. That’s about it.
What if we adopted an attitude that we truly haven’t seen anything like any of this before? That your Wednesday is unlike any Wednesday that came before it? That this economic period is solely what it is and nothing like what happened in the past? That the next bird flu will be completely untethered to the presumed characteristics of past bird flus?
What if every instance was treated like something new and wholly individual?
Well, someone would offer the classic line from George Santayana: “Those who cannot remember the past are condemned to repeat it.”
And our author, Nassim Nicholas Taleb, would wholeheartedly agree. We cannot forget our past. We just need to use our knowledge of the past in a more effective way. Lest we be condemned to thinking we will repeat it.
To return to the Lucas Critique, the big idea was to make a shift to small behaviors as a basis of analysis. To not longer treat “The Past” as this big, singular thing that just walks in perfectly-timed, identical circles, bringing the next boom/bust on a regularly-appointed schedule just as we predicted. Large systems like the economy are simply too complex to have a reliable history of this sort.
That scale is far too difficult to wrangle. The randomness is far too high.
But small systems? Ones with limited variables and shorter cycles and greater numbers of observations? Those are the building blocks of the macro picture and those can be better understood, extrapolated, and aggregated into a larger whole. Smaller behaviors modeled together like bricks being cobbled for a bridge are stronger.
Lucas called them microfoundations.
Think of them as the smaller, closed systems that operate in patterns we can easily see. Like the emerging cavities a dentist can observe in my teeth. Or the heuristics behind loss aversion. Or the relationship between commodity prices and imports. There is less noise, less randomness, in those significantly-smaller relationships.
So whether it is to test future policy or make a prediction of where we’re heading, the key appears to be that we should assess proposed change on the way it affects a hundred small things rather than a singular narrative arc based on what happened last time. This is where I think Ray Dalio can be easily misunderstood by folks like me. Maybe.
I’ve read Principles (a personal favorite) and I’m reading the next book on debt cycles. In both books, Dalio establishes a beautiful array of microfoundations based on what I suspect is the maximum amount of observations that any one group or person can accumulate in their work. He calls them principles. Think of them as rules of thumb. Or, in the Charlie Munger style, mental models.
When Dalio says we’re two years from a downturn because we’re seeing something like 1937, I like to think that what he’s really saying is that the microfoundations at this time are showing patterns that are very similar to 1937.
Or maybe he’s just falling for the narrative fallacy. Weaving a story that sounds good and thus must be (but isn’t) true. I don’t know. He knows the probabilities are low for calling your shot two years ahead. It’s like deciding who will win the Super Bowl in 2021. Sure, you could say it’s the Patriots. Are you sure? How much are you going to bet on that today?
In instances of forewarning like this, if the downturn doesn’t occur, is it because—a’la the media and bird flu—he raised the awareness to such a degree that preventative actions ensued? How would we ever know?
Maybe it’s all randomness. That feels much more likely.
That feels disappointing, too. Demoralizing. But Taleb offers hope. Because the minute we understand the nature of probabilities and distribution, we get a better feel for how to operate. We begin to appreciate the microfoundations that inform economic analysis today and the safe/risky work of a dentist who makes music albums on the side. We begin to see how we can better dance with uncertainty by not only hedging against randomness but actually managing for it in the name of upside risk. Regardless of what the future may bring.
That might sound silly. But I predict it will be more sensible tomorrow.