Principal Component Analysis and Factor Analysis in Accounting Research

Abstract

Principal component analysis (PCA) and factor analysis (FA) are variable reduction techniques used to represent a set of observed variables in terms of a smaller number of variables. While PCA and FA are similar along several dimensions (e.g., extraction of common components/factors), researchers often fail to recognize that these techniques achieve different goals and can produce significantly different results. We conduct a comprehensive review of the use of PCA and FA in accounting research. We offer guidelines on programming PCA and FA in SAS/Stata and emphasize the importance of implementation techniques as well as the disclosure choices made when utilizing these methodologies. Furthermore, we present intuitive, practical examples highlighting the differences between the techniques and provide suggestions for researchers considering the use of these procedures. Finally, based on our review, we provide recommendations, observations, notes, and citations to the literature, regarding the implementation of PCA and FA in accounting.

Pleiotropic Potential of Evernia prunastri Extracts and Their Main Compounds Evernic Acid and Atranorin: In Vitro and In Silico Studies

“…Principal component analysis (PCA) is an increasingly popular data analysis method used in many scientific disciplines [ 92 ]. PCA allows the observation of regularities among the examined variables [ 93 ]. The observed relationships are visualized in charts, which facilitates data interpretation.…”

A Simple Approach to Better Distinguish Real Earnings Manipulation from Strategy Changes*

“…We retain one component rather than multiple components because using 11. See Allee et al (2022) for an overview of the differences between PCA and factor analysis. 12.…”

“…This issue is less problematic for PCA than factor analysis as conceptually factor analysis assumes that a continuous latent factor is the underlying cause of the observed weightings. With PCA there is no conceptual problem; rather the problem only stems from the correlation matrix, which is the default option in most statistical packages (Allee et al 2022) and relies on the assumptions that variables are normally distributed. We use the standard PCA approach that uses the correlation matrix as it is the easiest to implement.…”

Analyzing the strategy–performance relationship in Germany – can we still use the common strategic frameworks?

“…PCA is a statistical technique that analyzes data patterns and identifies principal components based on correlations between variables. The principal components are a linear transformation of the variables and explain most of their variance (Allee et al., 2022). Thus, PCA allows us to identify competitive strategies based on strategically relevant variables and calculates their weights so that competitive strategies reflect the underlying variables as accurately as possible.…”

“…Due to the mentioned advantages, PCA has been employed in several survey-based studies (Allen and Helms, 2006;Allen et al, 2007;Koo et al, 2004). However, PCA is also suitable for archival data (Allee et al, 2022).…”