An Applied Physicist Does Econometrics

The biggest problem those attempting to understand econometric data, via modeling, have is that economics has no \textbf{F }= m\textbf{a}. Without a theoretical underpinning, econometricians have no way to build a good model to fit observations to. Physicists do, and when \textbf{F} = m\textbf{a} failed, we knew it. Still desiring to comprehend econometric data, applied economists turn to mis-applying probability theory---especially with regard to the assumptions concerning random errors---and choosing extremely simplistic analytical formulations of inter-relationships. This introduces model bias to an unknown degree. An applied physicist, used to having to match observations to a numerical or analytical model with a firm theoretical basis, modify the model, re-perform the analysis, and then know why, and when, to delete ``outliers'', is at a considerable advantage when quantitatively analyzing econometric data. I treat two cases. One is to determine the household density distribution of total assets, annual income, age, level of education, race, and marital status. Each of these ``independent'' variables is highly correlated with every other but only current annual income and level of education follow a linear relationship. The other is to discover the functional dependence of total assets on the distribution of assets: total assets has an amazingly tight power law dependence on a quadratic function of portfolio composition. Who knew?