AS THE ECONOMY TURNS

IN THE EYE OF THE BEHOLDER

Dr. William Shingleton

September 23, 2006

 

Sometimes students want to know why we spend so much time studying MODELS in economics.  The answer lies in the complexity of the real world. If we started off trying to explain everything at once, the answer would have to be so long and so convoluted that no one would be able to understand anything. In contrast, although a simple model is unrealistic, it should also be relatively easy to understand and it usually offers an approximation of how a part of the real world works.  We trade off the realism for the simplicity.  Then, as we draw students into the fascinating web of economics over the course of a semester, we keep making the models a little more realistic and a little more complex every day until, at the end of the semester, the models are so realistic and so complex that nobody has any idea what we are talking about any more.  Then we are done and they are ready to graduate.

 

A model is a SIMPLIFICATION that helps us to understand the basics of how the world works.  Actually, models are widely used in a number of other fields.  For example, many of our students are familiar with the clothing advertisements we see in the Sunday newspaper or in a fashion magazine. [NOTE: Economics faculty have never, to our knowledge, been accused of being otherwise aware of fashion guidelines, but that is a sad story for another day.]  When our students look at the clothes in these advertisements the clothes always fit just right on the people in the pictures.  However, when our students actually try on the clothes they don’t fit quite as well.  There is nothing wrong with economics students, they are all (or mostly all) fine, noble specimens of the human race; they are just not models. The fashion models in the magazines are used to offer APPROXIMATIONS of what the clothes would look like on normal people. Models work that way in economics as well. Models give us approximations of how things work in the real world.

 

If a model is going to be useful it needs to be able to PREDICT or EXPLAIN something that we see around us. Therefore, if we are going to have some confidence in the ability of a model to perform, it needs to be tested to see if it holds water. How do we do that? One answer comes from the field of statistical testing in economics, known as ECONOMETRICS. In econometrics, we use statistical tools to test for STATISTICAL INFERENCE, whether or not our models conflict with reality. The tests try to examine whether or not the patterns of data we see fit together in a way that is consistent with what our models say.

 

And how do we come up with these models in the first place?  Some of it is just brilliant insight, and we’ll leave that right there.  But a good deal of it comes about because someone noticed a pattern in some data, at the simplest level, a CORRELATION. [NOTE for the nerds reading this: A correlation can come in two forms. In a POSITIVE CORRELATION an increase in one event seems to happen together with an increase in a second event; while in a NEGATIVE CORRELATION, an increase in one event seems to happen together with a decrease in the second event.]  While a correlation, by itself, doesn’t actually prove anything and may just be a coincidence, it often makes us suspicious that there may be some sort of CAUSE AND EFFECT RELATIONSHIP going on.  However, to argue that a causal relationship exists, we need to develop a logical connection, a theory.

 

For example, for the first thirty-one SUPER BOWLS, the outcome of the football game matched with the ups and downs of the stock market twenty-eight times, a ninety percent correlation (If a team from the old NFL won the game the market would rise for the year; if another team won, the market would fall.). However, there was no logical connection between the outcome of the game and the performance of the market, it was just a coincidence, and we do not believe there were any models asserting that the pattern would continue. No one seemed overly surprised when the pattern fell apart over the last eight years, holding only twice in that span. Correlation, by itself, does not prove anything. Sometimes things just happen without any reason, like the Red Sox winning the pennant in 2004.

 

Sometimes a pattern shows up and it’s not so clear whether we are talking about a coincidence or not.  When we start poking around for a logical (causal) relationship, there are certain rules we need to follow.  For instance, if we are not sure which event is the cause and which one is the effect, we want to get the SEQUENCE right. If one event tends to happen first and the other tends to happen second then we are just about certain that the event that happens first is the cause and the one that happens second is the effect, unless you believe in some sort of time warp or something like that.  Unfortunately, when we accumulate data with recurring patterns over time the case isn’t that clear-cut; it becomes more of a chicken and the egg kind of problem. [NOTE: We call this kind of data TIME SERIES DATA to distinguish it from numbers that are collected from numerous sources at the same moment in time which are called CROSS-SECTIONAL DATA.] 

 

As an example, consider that in the 19th century it was well known that there seemed to be a correlation between economic activity and SUNSPOTS. Of course, in the case of sunspots and economic activity, at least one piece of the puzzle is intuitively obvious.  While there may be a number of logical explanations for why the sunspots affect the Earth, and maybe even economic activity on Earth, no one, not even an economist, is likely to even try to argue that economic activity on Earth affects the probability of sunspots on the surface of the sun.  If there is a cause and effect relationship, then it is going to run from the sun to the Earth.  In the 19th century they came up with two remarkably different theories and both seemed to fit the facts, which seemed to indicate that the sunspots were evidence of unusual activity on the surface of the sun and they resulted in unusual radiation patterns coming here to Earth.

 

One model was a psychological model, or at least what passed for a psychological model at that time.  It was built around the idea that the radiation affected men’s minds, making them lazy and irritable and causing the economy to fall into a RECESSION. When the sunspots went away men would go back to being their sweet and lovable selves and the economy would improve. While any reasonable person should have known that the premise was completely wrong (Everyone knows that men are sweet, lovable, and hard-working all of the time. And we are never, never irritable!), the idea had some adherents who concluded that the sunspots were the cause of the ups and downs of the economy (BUSINESS CYCLES).

 

It took a dentist who became an economist, WILLIAM STANLEY JEVONS, to offer a more logical model, which was based on the idea that the radiation affected crop yields, and therefore agricultural income [See W.S. Jevons, “Commercial Crises and Sunspots” in INVESTIGATIONS IN CURRENCY AND FINANCE, 1884.] Since the 19th century was still a time when agriculture was one of the most important elements for all of the nations on the planet, the theory had more logical substance to it, even if the conclusion, that the sunspots were the cause of the ups and downs of the economy, was identical to that of its psychologically unbalanced rival, the one which had asserted that men could ever be irritable.

 

In modern economics we use advanced statistical tools to sort out cause and effect relationships, although sometimes you have to wonder how far we have really come.  Consider the case of the sunspot theory.  In 1982, RICHARD SHEEHAN and ROBIN GRIEVES used something called a GRANGER CAUSALITY TEST to go back and look at Jevons’ sunspot theory of the business cycle. [NOTE: The Granger test is a fancy statistical test to sort out the chicken and the end problem from two streams of time-series data. They did not waste any time going back to look at the psychological theory because every reasonable person knows that men are never irritable. We don’t know why you have to keep bringing that up! See: "Sunspots and Cycles: A Test of Causation”, SOUTHERN ECONOMIC JOURNAL, 1982]. The Granger Causality test is one of those complex statistical techniques that economists use when we are trying to impress everyone with our mathematical abilities, so we use it quite a bit when we are showing off in a publication. Anyway, they had too much time on their hands one night and ran the Granger Test on Jevons’ model.  The reason we bring it up is that they found that the economic activity in the United States has had a significant impact on the number of sunspots on the surface of the sun.

 

Of course, more than a few economists were surprised at those results. NICHOLAS NOBLE and WINDSOR FIELDS wrote a comment in the same journal, which essentially said that the reason the ridiculous result came up was that the original authors did not structure their tests correctly. The correlation was there, but it was spurious, and they should have seen that.  Not to be outdone, the original authors then wrote a REPLY, also in the same journal, which made the corrections and still supported the conclusion that economic activity in the United States has had a significant impact on the number of sunspots on the surface of the sun.

 

There are a couple of points to the story. First, as Sheehan and Grieves say in their Reply, if we already know which way the causation runs and we still have to twist and turn our results to get them to come out right then maybe our statistical tests don’t hold quite as much water as we had hoped. Our point for our classes would be that just because a model is complicated to the extent of being almost incomprehensible doesn’t necessarily make it correct.  If the conclusions violate your common sense then you should be suspicious of the model.  Economics, to a very large extent, is the science of common sense. Without a solid grounding in logic, the statistical patterns that some may use to promote ideas in economics, don’t really mean a great deal. While people believe a great deal in economics, models and econometrics help us find out what we know. And don’t know. The difference is important but sometimes the distinction is in the eye of the beholder.

 

Final notes:  If you would like to be removed from the distribution list just send a reply on this email.  Back issues are available on my website <http://www.oocities.org/wsirius30/2cents.html>. The opinions expressed in these newsletters are those of the author.  Comments, including suggestions for future newsletters, are always welcome.