Stylized fact

From Hobson's Choice

A "stylized fact" is a simplified expression of a statistical observation. For example, if one runs a regression analysis of economic data for OECD countries, one will observe a strong correlation between taxes (as a share of GDP) and trade balance. You could say "if taxes are increased by 1% of GDP, then the trade balance increases (or becomes less negative) by 0.25% of GDP."

To make this interesting, let's point out that the results are statistically significant.

This is not actually true, although it is an intuitive explanation of the 0.25 coefficient on T (where T is taxes/GDP). It is not true because you cannot conclude that the country exports more, or imports less, because taxes are higher. There may be some attribute of the sample, that leads to high taxes occurring whenever the trade balance is relatively favorable (in other words, something else might have caused the other two things). Or the cause and effect might be reversed.

Hypotheses need to explain stylized facts, but it is not accurate to infer that stylized facts are explanatory elements. Suppose you are studying breast cancer in two regions. Region A has a semiconductor industry, while region B has a textile facility. In both regions, breast cancer is higher among the more polluted, poorer neighborhoods. But the money income of the poorer neighborhoods in region A will be higher than that of the poorer neighborhood in B, while breast cancer incidence will be higher in the poor neighborhoods of A than in lower-income neighborhoods B. Most of the cases will be in the poor neighborhoods of either A or B, so a stylized fact would be that higher income leads to a higher likelihood of breast cancer.^[1] That's not true, but it's a stylized fact.

One of the more controversial uses of stylized facts in the social sciences is in economics, where the dynamic general equilibrium (DGE) methodology has taken over. In this system, the economy is modeled as the sum of millions of identical households, or rather, as one household. Since there is only one household, matters such as liquidity constraints or involuntary unemployment are not represented. Please see positive economics.

[edit]