A Multilevel Factor Model: Identification, Asymptotic Theory and Applications

This paper studies a multilevel factor model with global and country factors. The global factors affect all individuals while the country factors affect only those within each specific country. A sequential procedure to identify the global and country factors separately is proposed. In the initial step, the global factors are estimated by canonical correlation analysis. Using this initial estimator, the principal component estimators (PCEs) of the global and country factors are constructed. It is shown that the PCEs estimate the spaces of the global and country factors consistently and are normally distributed in the limit. Several information criteria that can estimate the numbers of the country factors are proposed. The number of the global factors is assumed to be known. Extensive simulation results demonstrate that the sequential procedure and the information criteria work well in finite samples. The method of this paper is applied to 25 OECD countries to identify international business cycle. It is reported that the method extracts a global factor reasonably well.