A modified version of the EGZ (2019) model for equity valuation

: potentials and drawbacks

Abstract

This study investigates the performance of a modified version of the structural model for equity valuation of Eisdorfer, Goyal and Zhdanov (2019). The objective is understanding how the model performs in comparison with observed market values, whether it gives trading signals similar to those of the analysts and if some inputs might be misjudged or ignored by the model. This is achieved through the implementation of the adapted model on a cross-section of U.S. stocks from 2012 to 2019, retrieved from CRSP/Compustat Merged. The results obtained suggest that the model might be slightly overvaluing stocks on average, with a median and average Pricing Error on the entire time series of 3.15% and 10.84%. Moreover, the model’s median Root-Mean-Square Error and Absolute Pricing Error are respectively 41.34% and 36.96%, suggesting the presence of some outliers. In comparison with the analysts’ recommendations, this model conveys more mild signals, without a clear positive bias. This contrasts with analysts’ recommendations, which seem often too optimistic. The analysis of the main mispricing drivers indicates that some inputs, as the gross margin ratio, the dividend yield and leverage might not be properly taken into account, leading to overvaluation, while cash holdings, which is not considered in the model, may lead to undervaluation. Eventually two robustness checks are performed on the volatility input and on the model structure, confirming the model’s robustness.

Date of Award	16 Oct 2020
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Nuno Ricardo Raimundo Rodrigues Marques da Silva (Supervisor)