An Empirical Assessment of the Accounting Semi-Identity Problem's Pervasiveness and Severity

TL;DR

This paper uncovers a pervasive and severe methodological flaw in a widely used investment-cash flow sensitivity model, showing it often produces biased and spurious results across multiple datasets.

Contribution

It identifies the Accounting Semi-Identity problem, proposes an augmented model to correct it, and demonstrates its universal presence and impact on prior empirical research.

Findings

ASI distortion present in 100% of datasets

ASI explains over 83% of variance in investment models

Standard model accounts for only 17% of variance

Abstract

This paper investigates a fundamental methodological flaw in the investment-cash flow sensitivity model of Fazzari, Hubbard, and Petersen (1988). The model comes from a full accounting identity in which some components are missing, generating what I term an Accounting Semi-Identity, that mechanically links investment and cash flow, and this could bias coefficients, making the estimation difficult if not impossible. I propose an augmented specification including a variable that captures this arithmetic bias and test it across multiple firm-level databases. Results show that the ASI distortion is universal and severe: the ASI issue is present in 100% of the databases and explains more than 83% of the total explained variance, while the standard Fazzari, Hubbard, and Petersen (1988) model only accounts for approximately 17%. These findings suggest that a substantial body of prior empirical research based on this model may have reported spurious results rather than evidence of underlying economic behavior. This finding provides a compelling explanation for the substantial body of literature surrounding this model that has reported anomalous and incoherent results.