First Principles or Bust

So over the weekend, I got into a nice little... discussion with someone that had a very different view of the world that I do. They were a progressive, so most of our disagreement was about economics and the governmental involvement in education, and things like that. This is not particularly unusual. The problem was that we would approach historical events from different perspectives, which would, of course, change our analysis.

History is an important empirical test for theories of economics and other social sciences, but there are some insurmountable problems with the study of history. Mises pointed out that historical circumstances are always different from case to case, there are no fixed values (Human Action). Because there are no fixed parameters, it is impossible to parse the historical record perfectly. Instead, we have to approach from the other direction. We develop a theory and we look at history through that lens or 'understanding' as Mises terms it.

Some people might disagree that this is how history should be done, and that history needs to capture all of the relevant facts, and then form judgments from there. But what facts are relevant?  When we are choosing what facts to include in the analysis, we are using some theory already to decide if a fact might be useful. History is messy, which means that unraveling the true cause and effect is difficult. Discarding a fact you think is irrelevant might make deciphering the events impossible.

This means that we need a good way to determine what actions, events, and people are important before looking at the data. We have to have an a priori theory to give a proper understanding of history, rather than having a theory that twists in the wind and cannot bring insight to the facts of the present, or worse, a theory that can bring any insight to the present. A purely empirical theory will not know what data needs to be gathered and what data can be safely ignored.

So how do we get around this issue? All the data that we look at will be cherry picked to some extent, since we don't have the time to go through all of the facts of any specific event, and so the data we get will have a selection bias. I think the solution is as Mises presented in Human Action. We have to develop an a priori theory that can explain the phenomena that we are interested in, and then look at the data in that light. It is also important to understand that this theory has to be logically consistent, otherwise it is useless. No a priori theory will be completely falsifiable, since there will always be confounding factors to various degrees. Conflicted data with the theory would highlight these areas where additional explanation is required.

We need to make sure that we correctly select the axioms and starting points of the theories we end up using. It is extremely important that the first principles are solid, otherwise any theory will fall apart, or worse, be correct in some cases for all the wrong reasons.

The way I see it, historical theories can either be specific enough to be falsifiable, or general enough to be somewhat predictive. Some mix of the two, and specific parts of the theory can be falsifiable, but the overall structure will not be. As an example, we could say that in general, higher taxes will lower the growth of GDP (or something similar). Specific test of the theory will be if we raise taxes, the rate of growth of GDP might go up, which would mean that in this specific case, the theory didn't hold. But there are other confounding factors that would make that a bad conclusion, and we can't go back in time and try holding taxes at the same level.

I guess the main thing I'm trying to get at is that when it comes to history, it is impossible to get all the facts, and the facts you get have a selection bias. Therefore it is important to get the theory explicit and established beforehand so that others can challenge it directly. If people are operating with different theories, this needs to be discussed both before and after discussion of the historical facts.