Working Paper: CEPR ID: DP1285
Authors: Mario Forni; Lucrezia Reichlin
Abstract: This paper develops a method to analyse large cross-sections with non-trivial time dimensions. The method: (i) identifies the number of common shocks in a factor analytic model; (ii) estimates the unobserved common dynamic component; (iii) shows how to test for fundamentality of the common shocks; and (iv) quantifies positive and negative comovements at each frequency. We illustrate how the proposed techniques can be used for analysing features of the business cycle and economic growth.
Keywords: business cycle; sectoral comovements; factor analysis; principal components
JEL Codes: C51; E32
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| traditional VAR methods (C32) | inadequacy for large cross-sectional datasets (C55) |
| factor analytic models (C38) | identification of common shocks (E32) |
| common shocks identified as fundamental (E32) | belong to the space spanned by present and past values of variables (C32) |
| rejection of fundamentalness (B41) | distinction between common and idiosyncratic components (L15) |
| OLS methods (C67) | estimation of common shocks (C51) |
| approach facilitates measurement of negative versus positive comovements (C10) | understanding economic dynamics (E32) |