The Basis for Predictions

[ad]In a previous post I discussed the well-known fact that economists’ predictions are always wrong, and why they always are. But one obvious problem with predictions was left out of the discussion, and I would like to discuss this problem in a separate post. In contrast to the previous post, which was quite general in tone and content, this issue is mainly methodological and somewhat philosophical.

The previous post discussed the problems of measurement and the very problematic assumption that “people are like rocks,” i.e. that individuals share a fixed and observable nature in the same way that rocks have common simple properties. I also stretched the discussion to cover the ever present tension between the Weberian concepts of erklären and verstehen.

The former kind of science strictly emphasizes explaining facts and establishing simple causal relationships that can be derived from the observable properties of the entity. The latter stresses the subjective understanding of what is going on, and finding a way of rationally establishing a way to “see” how things work and are related. Weber explicitly states that erklären is the purpose and method unique for the natural sciences whereas the social sciences need to have a verstehen-based perspective. Predictions, hence, are possible only in sciences based on the erklären methodology and this is the conflict in economics: a fundamentally social science attempting to make use of primarily (only?) the methods and methodology of the natural sciences.

But predictions are problematic in and of themselves even if we ignore the tension arising from using erklären methodology studying verstehen phenomena. The very nature of predictions imply the usage of historic data to say something about the future. As we know, and have known at least since the days of the Ancient Greeks, it does not follow from the fact that the sun has risen every morning for centuries that it will continue to do so. History and future are not the same and may even be very different. What makes the future so troublesome is that it is fundamentally uncertain and we cannot use the certain facts of history to create knowledge about it.

As was stressed in the previous post, extrapolating doesn’t necessarily make sense. Doing the same maneuver for predictions about the future from data about historical events makes even less sense. Tomorrow will not be exactly like yesterday, which is a fact everybody knows and should know. This fact is true for details as well. That a rock falls to the ground if dropped today does not mean it will do so tomorrow.

However, we can conclude that a rock will fall to the ground if dropped tomorrow if we can show what makes it drop and we can rely on the properties of these causes being the same tomorrow. A rock has a fixed nature with certain properties and these do not change. We have been able to establish that a rock is dead matter that responds to exogenous forces in a very reliable and predictable way – we know that a rock is a rock is a rock and that this means something in terms of its nature.

It may be the case that tomorrow does not have gravity or that all rocks have turned into lollipops, but that doesn’t change the fact that rocks, according to our defintion, are rocks and that they respond to different forces in certain ways. We cannot with complete certainty say that everything will be the same tomorrow, but we can make general statements that will hold true for the things, forces, and properties we have specified (if we have done a good job specifying them). 

Now try the same thing with a human being. An individual is an individual is an individual. If this is true in the same sense as a rock is a rock, then we should be able to establish if one and every individual likes ice cream, responds the same way to stimuli like heat and cold, reacts to a certain situation the same way with a high level of certainty. 

Try the latter and compare a rock with an individual. Expose the rock to exogenous forces and observe its “behavior” and what happens to it. Then expose an individual to some stimuli and observe the behavior. Repeat it and observe the behavior – is it exactly the same? You will find that different individuals react in different ways to stimuli – and that one individual’s reactions will change over time as he or she learns. The rock never learns.

So even if the way a rock is affected by certain experiments is not purely certain for the future, it is very much predictable. The way John Doe reacts to, e.g., a speeding car about to hit him is different every time – and may not [ever] be the same as how Jane Doe reacts. It is not predictable; we cannot know what will happen (i.e. how the individual will react). 

So how will people react to lower prices in a certain good? We can attempt to predict that tomorrow, if the price for widgets is 10% lower, people will purchase 500,000 more widgets. But that doesn’t make sense. If the price is indeed lower it does not follow that the people who bought a widget yesterday at the higher price are more likely to buy a widget again. It also doesn’t follow that people in general value the widget in the same way. 

The only thing we can say is that ceteris paribus people will tend to purchase more of the cheaper good, at least for as long as they subjectively expect to be better off through purchasing one [more]. People want to be better off (which follows from the definition of better) and therefore make choices to improve their situation – to the best of their ability. But their preferences change and their ranking of those preferences change – as do their needs, perspectives, experience, knowledge, etc. An individual is not an individual is not an individual, at least not the same way a rock is a rock is a rock.

The problem of induction is problematic in natural science where dead matter is studied, even though the deathness of matter makes its properties reliable and effects predictable. Add life to the equation and the problem of induction becomes insurmountable and obviously so. 

Some things do seem to be repeated over time and the saying that “history repeats itself” may be thought to disprove the point I am making. But it doesn’t. It may be true that history tends to repeat itself if we do not learn from it, but the problem is that there is no “we” in the sense that there is a “rocks.” Individuals are different from each other and they change over time; humankind may not learn from the lessons of history, but it is equally true that situations do not repeat themselves – only man-made abstractions of them do. It is rational to learn from the essence of a situation not to repeat it or its negative consequences, but it is equally rational to say that things have changed and therefore the outcomes may do so too.

The lesson to be learned is that collectivism doesn’t work when we speak of human behavior simply because human behavior is not as tightly bound to the properties of “human” as the effects on a rock are to its properties. The reason is that human consciousness is not necessarily the same as the human body – one could possibly predict the effects of stimuli in medicine, but not in economics. Medicine works with the properties of the human body, i.e. its constitution and chemical and biological relationships (however complex); economics studies human behavior, where one individual’s choice to act is not based on the same facts as another’s, and a specific individual tends to learn – and change – from experience.