Quantile Factor Models

Quantile factor models (QFM) represent a new class of for high?dimensional panel data. Unlike approximate (AFM), which only extract mean factors, QFM also allow unobserved factors to shift other relevant parts the distributions observables. We propose quantile regression approach, labeled Factor Analysis (QFA), consistently estimate all quantile?dependent and loadings. Their asymptotic are established using kernel?smoothed version QFA estimators. Two consistent model selection criteria, based on information criteria rank minimization, developed determine number at each quantile. estimation remains valid even when idiosyncratic errors exhibit heavy?tailed distributions. An empirical application illustrates usefulness by highlighting role extra in forecasts U.S. GDP growth inflation rates large set predictors.