While the end of 2013 has come and gone, and the New Year has been ushered in, I have found my self once again pondering over the complexities involved in forecasting. I have been thinking more specifically about how those forecasts/predictions/estimates may be both beneficial and toxic to the perception of the interpreter. There have been more forecasts than I can count in the last few weeks, ranging from individual stocks picks, to the North American markets as a whole, followed by the macro picture in the emerging markets and China, U.S Treasuries, 3-D printing, gold, bitcoin, USDJPY, USDCAD and pretty much everything else under the sun.
This year I have (hopefully) learned more from the biases that manifest within predictions, from what I will call the “utility of predictions” and the self-cancelling/fulfilling (asymmetric) affects of the predictions (the premise of the article). The purpose will be much more philosophical than scientific, as much remains unknown regarding how predictions affect our decisions, or the “utility of the prediction” and how it can be quantified.
We will start off with a few failed predictions over the last quarter century that may have had an influence on economic outcomes, consisting of recessions, the 1981 Peru earthquake, farmers and weather.
Most people have heard of the old saying that economists will have predicted nine of the last six recessions. Or maybe the saying, even a broken watch is right twice a day.
What those statements don’t say are the affects of having the wrong time for the other 1,438 minutes in the day or how market participants’ decisions will be affected by the other three recession forecasts.
“Charlie and I believe it’s a terrible mistake to try to dance in and out of it based upon the turn of tarot cards, the predictions of ‘experts,’ or the ebb and flow of business activity. The risks of being out of the game are huge compared to the risks of being in it.” –Warren Buffett (Trades,Portfolio)
Clearly economists do a terrible job (outlined above), and it is sometimes argued they actually add negative value. The question I am trying to answer here is how do economists (or other) predictors influence our actions and to what degree? I read an interesting study recently on weather and farmers’ decisions and found a few key things peculiar: how much they were influened and by what.
Weather and Farmers
The table above shows current and recent past conditions (CRPC), the short-term forecast (STF) and the long-term forecast (LTF). The numbers below are on a scale of 0 to 7, representing the utility of the forecast to farmers. Farming decisions were categorized into five groups corresponding to different stages of crop production: 1) agronomic decisions through planting (including choice of crop type, seed variety, tillage method, planting density, and date), 2) summer growing-season decisions (including applications of pesticides, herbicides, fertilizers, and irrigation), 3) harvest and postharvest decisions (including harvest date, autumn irrigation, and tillage), 4) crop insurance, purchased before the March 15 federal deadline of each year, and 5) marketing choices made throughout the year. The study characterized this outcome of utility by the following grouping of 10 descriptions,. i.e the value of each to the farmer:
- Planting the best crop and variety; optimum spring tillage; best planting density and planting date.
- Right amount of crop insurance.
- Optimal amount of spraying, fertilizing, and water applied (used); best harvest date.
- Maximizing crop revenue from marketing.
- Lowest possible costs of production.
- Reducing financial risk.
- Sharing limited sources of irrigation water with others.
- Reducing fertilizer and pesticides in runoff/ground water.
- Sustaining rural communities.
- Others (please specify).
As we can see from the study (the two pictures provided above), forecasting and past trends were clearly influencing the utility of the decision for the farmers from various sources, although some more than others (like the spouse or television and radio). There are also other various social norm weightings shown above; we can infer how influential a spouse and television/radio can be compared to others. (The scale is 0-49.)
The stronger the signal is perceived (or the more the forecaster is trusted to be accurate) the longer the signal will last. Attraction and repulsion laws will run there course and as more people are influenced by a particular forecast, the more likely they are to add to it, alter it or simply claim it as their own, causing an exponential function to the forecasts growth, or in a sense, self-fulfilling.
You may be asking, well, why does any of this matter? Why do I care? How can it benefit me? Well, think of how these simple everyday predictions are correlated with psychological sentiment (pessimism and optimism) the higher the emotional function, the more attraction or repulsion; the stronger the signal, the higher the utility of the forecasts. These are what I like to mentally visualize as the extremes of the investor emotional pendulum that is forever (albeit asymmetrically) swaying back and forth. If we want to exploit these extremes of emotion we must become students of our own (and others) pysche/emotion, conducting metacognitive thinking excersises.
We should ask and think probabilistically, do all the inputs create a signal or are they simply noise? Do the inputs affect my Bayesian prior in a material manner? Take the prediction by Bailey Willis of the 1981 earthquake in Peru that was supposed to be 9.9 in magnitude. This prediction ended up causing mass hysteria and real-estate prices to plummet simultaneously across the country although no serious earthquake came. These types of forecasts can have such strong affects of attraction that they become self-fulfilling, or so repulsive that they become self-defeating, hence the term “self-fulfilling prophesy.”
In the case of the earthquake, nature does not care about our perception or beliefs of an event, and nature produces no effect from the causation of our thoughts. Humans are quite different as we can/will influence others’ perceptions through our own narratives, predictions and beliefs. This of course is in the present or future tense and our confirmation bias, hindsight bias and narrative fallacy, all help to dilute the objective event as it passes into our own subjective perception. Mirror neurons further complicate what we believe, percieve and learn as we subconsciously/implicitly absorb others’ language, choices and actions.
I fee like I am rambling, and I will move to wrap up the conversation with a summary of a few main points.
- Crowd wisdom may be flawed when the collective judgment is weighted inconsistently, giving certain variables higher or lower weights than others.
- Forecasts can be self-fulfilling and self-defeating; what is the incentive of the forecast, is there any skin-in-the-game?
- We all deliver some utility in what we believe and care to share about our beliefs; there isperceived value to a forecast
- Hindsight bias, confirmation bias and narrative fallacy cloud our judgment about past events and how we perceive them.
- We need to weigh forecasts probabilistically against our prior perceptions.
- Do not make forecasts if you do not understand who you may be affecting and how their decisions may be impaired as a result; a forecast is not something that should be taken lightly.
Finally, we should not forget that 90% of all communication is non-verbal and our actions can speak profusely louder than our words. When a certain threshold of forecast-correlation becomes (implicitly) present, we will likely be affected. It takes much practice and experience to become a stoic contrarian.
A great book I read recently covering forecasts and our inability/fallibility of predicting was, “The Signal and The Noise: Why So Many Predictions Fail – but Some Don’t” by Nate Silver.
Another great article from National Geographic I read recently and would recommend was about collective behavior in ants, bees and flocks of birds.