题 目：Quantile Regression Modeling in Finance and Economics

汇报人：Prof.Gib Bassett(Head of Finance Dept at UIC,USA)

时 间：2007年1月11日（周四）上午10：00－11：00

地 点：拉斯维加斯9888202室

Abstract: The conditional expectation model is standard in finance and economics. It expresses the expected value of a variable of interest as a function of independent variables. Expected return is a function of the market return, size, and value; the expected wage is a function of an individual's characteristics, and so on. This contrasts with the conditional quantile approach in which the distribution of a variable of interest is a function of independent variables, with different parts of the distribution being different functions of the independent variables. The quantile regression model includes the standard model as a special case while allowing for a richer set of relationships between variables. This added expressive power means a stock's return in the tails can be a different function of market return than at the expected value. It means that variables that have no affect on the expected wage can have important influences at other parts of the distribution. The quantile approach becomes especially relevant for portfolio and policy questions that ask about what happens away from the expected value; e.g., how does risk change with the market, how is the lower tail of the income distribution affected by a change in education.

A non technical overview of quantile regression will be presented. Examples will be used to contrast conditional quantile and expectation models. Also considered will be the distinction between the associated unconditional models. Since data requirements are identical, the recommendation is that the standard conditional expectation approach be replaced with the greater expressive power of quantile regression.