Target PCA: Transfer Learning Large Dimensional Panel Data

TL;DR

This paper introduces target-PCA, a transfer learning method for estimating latent factors in large, incomplete panel data, improving efficiency and ability to detect weak signals compared to traditional approaches.

Contribution

The paper proposes a novel target-PCA estimator that leverages auxiliary data for better estimation in large, missing data panels, with theoretical guarantees and empirical validation.

Findings

Target-PCA outperforms benchmark methods in macroeconomic data imputation.

It can consistently estimate weak factors that traditional methods miss.

The method provides a new approach for handling missing data in large panels.

Abstract

This paper develops a novel method to estimate a latent factor model for a large target panel with missing observations by optimally using the information from auxiliary panel data sets. We refer to our estimator as target-PCA. Transfer learning from auxiliary panel data allows us to deal with a large fraction of missing observations and weak signals in the target panel. We show that our estimator is more efficient and can consistently estimate weak factors, which are not identifiable with conventional methods. We provide the asymptotic inferential theory for target-PCA under very general assumptions on the approximate factor model and missing patterns. In an empirical study of imputing data in a mixed-frequency macroeconomic panel, we demonstrate that target-PCA significantly outperforms all benchmark methods.