The predictive limit of Beta in the corporate default risk
Abstract
This study investigates the relationship between market beta and corporate default risk, measured using Altman's Z-Score. The objective is to determine whether beta—commonly used as a proxy for systemic risk—can serve as a reliable predictor of financial fragility at the firm level.
The analysis is based on a panel of 510 observations of Italian companies over the period 2019–2023. Several econometric approaches are employed, including panel data models (Pooled OLS, Fixed Effects, First Differences), parametric and non-parametric Generalised Additive Models (GAM), and LASSO regression for automatic variable selection. The independent variables include beta and a set of accounting indicators (ROE, leverage, interest coverage, net margin, asset turnover, current ratio). The Z-Score is used as a continuous dependent variable to capture default risk.
Across all model specifications, beta is consistently found to be statistically insignificant. Its contribution remains null or negligible even in non-linear and penalised models. In contrast, all accounting variables display strong, stable, and significant relationships with the Z-Score, underscoring their superior predictive power.
The findings indicate that market beta is not a useful metric for estimating corporate default risk. This suggests a clear empirical separation between systemic risk and firm-level financial fragility. The evidence reinforces the reliability of accounting-based indicators in assessing default probability, and highlights the limitations of using beta outside the theoretical scope of the CAPM.
Full Text:
PDFDOI: http://dx.doi.org/10.13132/2038-5498/16.3.921-932
_____________________________________________________________________________________________________________________________________________________________
Registered by the Cancelleria del Tribunale di Pavia N. 685/2007 R.S.P. – electronic ISSN 2038-5498
_____________________________________________________________________________________________________________________________________________________________
Privacy e cookies