Steven Dutch, Natural and Applied Sciences, Universityof Wisconsin - Green Bay
First-time Visitors: Please visit Site Map and Disclaimer. Use"Back" to return here.
I was inspired to write this page after getting a call from a businessman who was concerned about his colleagues passing around a bogus report on global warming. After answering some of his questions and trying to come up with analogies that would make sense to business people, it occurred to me that science, business, and many other professions actually face some very similar fundamental problems.
With apologies to Music Man.
Every ad for any investment carries the disclaimer "past performance is not necessarily indicative of future performance." Why? If past performance is totally unrelated to future performance, why mention it at all? After all, who cares about last month's winning lottery numbers? Obviously everybody believes there is some correlation between past and future performance. But the connection is very complex. In a boom, even a real mutt of an investment might grow. And in a recession, even the very best companies might take a loss. A poorly run company might face a shareholder revolt, get a new CEO and Board of Directors, and start performing well (Chrysler under Lee Iacocca). A well run and respected company may make some bad decisions and start performing badly (Kodak missing the boat in the early years of digital photography?).
In short, there is a connection between past and future performance, but you have to be very careful about applying it. It takes skill and awareness of all kinds of potentially relevant factors to pull it off, and even the best investor will be wrong some of the time. Maybe even most of the time. After all, a .400 hitter in baseball, something not seen in over half a century, misses 60% of the time, and Brett Favre holds the NFL records for pass completions - to both teams.
Surprisingly, there are a lot of academic fields that face very similar problems. History is the prime example. Historians say "history has no predictive value," but if it had absolutely no predictive value, there wouldn't be any point in studying it. History has no simple predictive value. You can predict that Germany might do certain things in a crisis that France wouldn't, but you can't predict that because ancient Rome went a certain way, the United States will inevitably do the same thing. (Anyway, most of the folks who compare America to ancient Rome make it crystal clear they have never actually read Decline and Fall of the Roman Empire. I know, because I have read it.)
Once you get beyond physics and chemistry, a lot of the same problems turn up in science. When you can run an experiment and control all the variables, life is comparatively simple. Once you get out of the lab and into the field, science gets a lot more complex and a lot less clear cut. We can predict, based on long experience, that if cats get loose on a previously pristine island, they'll demolish the local ecology. We can predict, maybe, that certain ground-dwelling species will be more vulnerable. But we won't necessarily be able to predict that some rare orchid will go extinct because the moth that pollinates it went extinct because the plant its caterpillar feeds on was displaced by another plant that ran out of control because the bird that normally ate most of its seeds went extinct because of the cats.
In geology, my own field, I can tell you with overwhelming certainty there will be a major earthquake on the San Andreas Fault. I can tell you the odds strongly favor the quake occurring in Southern California (last major quake: 1857) over Northern California (1906). I can say the smart money would be on it happening by mid-century, and I wouldn't be surprised if it happened today (actually, the smart money is in earthquake-resistant construction). But I can't tell you exactly when. When I was a grad student, it looked like we were this close to having a viable earthquake prediction technology. But it turned out to be a far more difficult problem than it first appeared.
In all these examples, we can make general broad predictions, and we can make very narrow, specific predictions, but in that in-between zone, we can't make predictions that are both broad and accurate. For example, we can say that the French will or won't support some resolution in the U.N., and we can say that when the French feel their deepest interests are at stake - and they can win - they'll fight nasty as wolverines (don't believe those stereotypes of the French being cowards for a second). But we can't say what party they'll vote into office a year from now.
Weather and climate is another example of a field where we can make general broad predictions, and we can make very narrow, specific predictions, but in that in-between zone, we can't make predictions that are both broad and accurate. We can say that a blocking high will divert a hurricane in a certain direction, and we can say that hurricanes are more frequent in summer than other seasons, but we can't predict how many hurricanes there will be next October.
Business, of course, is a perfect example of a field where we can make general broad predictions, and we can make very narrow, specific predictions, but in that in-between zone, we can't make predictions that are both broad and accurate.
Note: I started on this discussion before the turmoil of the Fall of 2008 and the election of Barack Obama. The illustrations used here are hypothetical.
Let's look at a hypothetical week on Wall Street:
It really doesn't take an MBA from Wharton to predict what the Dow is going to do in response to these developments: up on Monday, down on Tuesday and Wednesday, up on Thursday and Friday. The unexpected can happen, but those are the most likely outcomes. We have lots of cause and effect experience to tell us. Granted, no day is exactly like any other, but anything that, say, threatens oil supplies, is likely to make investors hesitant.
Now suppose the incoming Administration and Congress announce a bold program of doubling the minimum wage, sharply increasing capital gains and income taxes, and slashing interest rates. Financial analysts predict steep inflation along with a severe recession. In response to these fears, Congressional backers of these moves say:
Nobody believes a prediction of the Dow Jones Average twelve hours ahead. Now we're asked to believe a prediction that goes out a year into the future?
I think "economic illiteracy" is maybe the kindest thing people on Wall Street would say about that. The day to day fluctuations and random events on Wall Street make it very difficult to predict stock prices more than a short time into the future. Clear cut events are easiest to predict, as in the hypothetical week above, but what if OmniMaxiMegaCorp had taken a huge earnings hit and Congress had voted to cut the capital gains tax on the same day? It would be awfully hard to predict how those two events would interact.
Yet somebody did make a virtually identical comment on climate change:
Nobody believes a weather prediction twelve hours ahead. Now we're asked to believe a prediction that goes out 100 years into the future?"Aliens Cause Global Warming," a lecture by Michael Crichton at the California Institute of Technology, Pasadena, CA, January 17, 2003.
The day to day random events that drive prices up and down on Wall Street are like weather. We can make short-term cause and effect predictions. Even if we don't believe the Iranian military exercise is anything to worry about, we know some people will worry, and that will affect their behavior. So maybe we sell off things that will be perceived as safer because demand increases.
Long term economic trends are like climate. There's a reason why people use the term "business climate." Artificially giving people more money creates more dollars chasing the same amount of goods and services and that spells inflation. And tax hikes are a classic negative incentive to invest. We can't say what the Dow will be six months from now but all odds are that it will be lower and a whole raft of economic indicators will be sour.
So suppose people in Congress say a year or so after their new economic policies:
You predicted disaster when we announced our policies, but yesterday the Dow closed at a record high. So all those doom predictions were nonsense. Tax hikes don't cause economic problems.
You'd have to look at the big picture, right? And when you do, you might find unemployment is 8%, annual inflation is 12%, and the Dow is posting record highs because people are buying stocks with inflated dollars.
Or suppose somebody were to say:
There's nothing wrong with the stock market. Why only last week, Galactic Whoopie Cushion, Inc. posted a record high.
Uh, so what? Even in a boom, some companies will fail, and even in a crash, a few companies will defy the odds, or just be lucky. So the fact that a company posted a record high tells us just about nothing about the broader stock market. Yet how often do you see someone point to a record low temperature as evidence against global warming?
One of the most widely used but unreliable kinds of information is anecdotal evidence. Anecdotal evidence is some isolated example used to prove a point. Anecdotal evidence is valid if and only if:
The perfect example of anecdotal evidence for business professionals is the story of the millionaire who pays no income tax. Yes, the stories are true: if you have a big enough loss it may offset a million dollar income and result in no income tax. But easily available IRS data show that the top few per cent of taxpayers pay more than the bottom half of all taxpayers. So while the anecdote may be true, it is not representative.
Bogus anecdotes proliferate in this age of the Internet. As a general rule, any time you see an outrageous news bulletin on line, let it simmer for two or three days, because there's a very good chance it will be refuted before very long.
In early September, 2008, it snowed in the highlands of Kenya. Clear evidence that global warming is a crock, right? Except it turns out the "snow" was hail. Hail is actually fairly common in Kenya, but thanks to cool weather, the hail didn't melt as quickly as it normally does. So right off the bat, it turns out the anecdote was false.
Then there are anecdotes that are meaningless. In October, 2008, Boise, Idaho got its earliest snowfall on record, an event trumpeted by climate change skeptics as further proof that climate change was a crock. Except who, really, is surprised by October snow in the Rockies? This particular snow came a bit earlier than any previous snowfall, but so what?
Return to Pseudoscience Index
Return to Professor Dutch's Home Page
Created 18 July 2008; Last Update 24 May, 2020
Not an official UW Green Bay site