Using High-Order Moments to Estimate Linear Independent Factor Models
Stéphane Bonhomme, Madrid Jean-Marc Robin · 2006
We study the identification and estimation of linear factor models under the assumptions that factors and errors are independent and that factors are not normally distributed. High-order moments are shown to yield full identification of the matrix of factor loadings if factor distributions are sufficiently skewed or kurtotic. We develop simple algorithms to estimate the matrix of factor loadings from the second, third and fourth-order moments of the data. We run Monte Carlo simulations and apply our methodology to microdata on wages and education, and to financial data on stock returns. JEL codes: C13.