Stylized facts and over aggregation.

Back to science.


Recently, I've been coming across similar things in the more informal side of economic discussion (and, unfortunately, the formal). One is "stylized facts", and the other is, as I call it, over aggregation (many different ways of putting this one, but I like to use this).

Wiki defines "stylized facts" thusly:
 

"A stylized fact is often a broad generalization that summarizes some complicated statistical calculations, which although essentially true may have inaccuracies in the detail."

Well, you'd think that facts that are "stylized" are, by definition, not facts. Even if generally true, that does not make them "true", per se. However, it's surprising how often you see this in even formal and rigorous economic studies. Even if it's only used for convenience it can be detrimental. Even if it plays a minor role in your summary of a body of literature, it doesn't mean it won't have adverse effects.

This is all very related to another problem which, thankfully, is much more closely related to the informal side, but, alas, still found in even the most prestigious (pretentious?) journals.

Now, statistics is used to find causality and correlation in complex systems such as an economy. Good studies take into account everything about how their models and controls work. For instance, does increasing police officers decrease crime? The first wave of research suffered from an "endogeneity" problem. It just so happens that places that have more police officers are also more crime ridden. Can't get far with that. It'll tell you nonsense. So you have to go to more rigorous statistical methods to correct for this bias. That's what they did. They found that increasing police officers has an inverse relationship with crime.

That's how science is done. Want to see how it's not done?

I'm taking this from a review named "Real World Economics Review". They seem to be overcompensating with the name, right? Well, it's from a movement named "Post Autistic Economics", which, apparently, is aimed at people who use mathematical and statistical methods they don't know how to use.

Anyways, I didn't take it seriously. It was formed against the Neo-Classical side of economics which I don't think anyone fully subscribes to anyways, although they might find a great deal to agree with. Among other reasons it didn't seem serious. I think it could be easily countered with a movement of "Post-prepubescent-angst-economics", but I'll let you judge. 

I took this from the first article I grabbed, which was provocatively named "Neo-classical economics:A trail of economic destruction since the 1970s." Supposedly, this is "good" science from the next paradigm in economics:

This is the GDP of both Korea and Somalia. Well, it's obvious that they diverged. What could be the multitude of causes? And where's the sound empirical work in support? This is what the author had to say:

"Figure 1 illustrates the explosive growth of South Korea, starting only in the very late 1960s, as that nation diversified its economy away from agriculture and raw materials and into manufacturing industry. Through very heavy-handed industrial policy, Korea broke away from its ‘comparative advantage’ in agriculture. By comparison, Somalia – being richer that Korea until the mid-1960s – did not, and instead continued to specialize according to its comparative advantage in being poor."

This is awful science, and I will explain what I mean. When you step back and ignore nearly all the different causes and actions within an economy, you can get nearly anything to support what you claim. In this piece of nonsense, this is supposedly condemning evidence against "comparative advantage", which, anyone with at least a little bit of honesty would notice says nearly nothing about it (and more about nearly EVERYTHING). 

What are the effects of minimum wages on employment? Don't run around looking at aggregate data of different countries and their respective minimum wages. That's a small fraction of the economy that's absorbed by your over-aggregated data. You're basically absorbing every unrelated thing into a gigantic mess of data, and then focusing on two things going on in that mess. You need to do better.

To be fair, from what I've seen outside of this, most of the stuff in their reviews falls short of most peer-reviewed journals, which is by no means an insult. Some of this is completely helpful stuff. Such as his comment about increasing returns to scale which, if he had looked into the evidence, would notice economists do, in fact, study it. 

But outside of that it's essentially all of the teenage angst regarding the financial crisis and and an overbearing urge to revolt against the powers that be, but really aren't. This is glorified sociology, that reminds me of the half-assed historical studies of Marxists that were nothing more than glorified sociology. Not science. Politics. 

I have a feeling I'll be talking about them again, or at least another segment on bad statistics.

To sum up, stylized facts are exactly that, stylized. Looking at gigantic economies and using very aggregated data to prove, well, anything you want, isn't science.

Now sleep.