Quantile Regression: A Robust Alternative to Least Squares
As is usual in applied econometric exercises of the present type, the workhorse of the analyses is least-squares estimation, which is so well-known that it and its properties need no discussion. However, in the course of estimating Engel and demand functions from the BLS-CES surveys, a rather surprising phenomenon has steadily revealed itself, namely, a marked tendency toward asymmetry in the distribution of residuals. The graph in Fig. 3.1 is typical. Quite clearly, the distribution is not symmetrical, but rather is left-skewed (i.e., has a longer tail on the left than on the right) and has peak density well to the right of its OLS mean of 0.