Integrating corporate governance and financial variables for the identification of qualified audit opinions with neural networks

Abstract

Prior studies have contributed to the development of models that help predict audit opinions, and have applied several methodologies in the search of better predictions. Nevertheless, and even though the existing literature on the prediction of audit opinions is profuse, the results achieved by the existing modelling are still considerably far from having obtained high levels of prediction, and the results are not in excess of 80 % in terms of classification. In previous research, prediction models of audit opinions have used financial variables. The main contribution of this paper is to show that combining these financial variables with variables relating to corporate governance of companies, the predictive ability of the models is significantly higher. For this, a sample was selected from Spanish Listed companies during the financial years between 2008 and 2010. From this sample, financial and corporate governance information was obtained, enabling us to rely on a new set of variables, more complete than those used in prior research. The validity of the variables was studied by means of univariant tests, upon completion of the database. Then, the results of the models built upon said variables were compared through neural network tests that have proven to yield a higher level of prediction, according to prior literature, specifically, multilayer perceptron and probabilistic neural network.

Predicting the type of auditor opinion: Statistics, machine learning, or a combination of the two?

“…◆ Different predictors are used. Most researchers use only financial metrics, but some use market data (Dopuch et al, 1987;Francis and Krishnan, 1999;, nonfinancial data, or latent constructs (Fernández-Gámez et al, 2016) as well. Even when nominally using the very same predictors, the calculations are likely to differ due to the differences in national reporting frameworks.…”

“…As a result, over the past decades, numerous models for predicting the type of audit opinion have been proposed in the literature (Kirkos et al, 2007). Earlier studies have mainly relied on the use of classical statistical modeling techniques, mainly probit and logistic regression models (Dopuch et al, 1987;Francis and Krishnan, 1999;, while more recent ones have shown an increased preference for modern machine learning techniques, such as decision trees and artificial neural networks (Fernández-Gámez et al, 2016;Pourheydari et al, 2012;Saif et al, 2013;Yasar et al, 2015). 1 Although the abundance of empirical research, conducted on the topic, might be expected to provide a solid basis for practitioners to identify a predictive technique that is best suited for predicting audit qualifications, we argue that this is not possible for two main reasons.…”

Predicting Audit Opinion in Consolidated Financial Statements with Artificial Neural Networks

“…The present study uses 35 independent variables obtained from previous studies on audit opinion prediction models [1,2]. These independent variables are presented in Table 2.…”

“…Previous studies have developed models to predict audit opinion. Auditors can use these models to plan audit procedures, as a quality control tool in the stage of review of a commitment, and for the analysis of the variables that affect the probability of obtaining a qualified opinion [1,2]. The existing audit opinion prediction models have only been developed in the context of individual financial statements, and none have referred to consolidated financial statements.…”

“…The results show that retained earnings/total assets is the most predictive variable. The authors of [2] improved the adjustments of these models by introducing corporate governance variables. This study used a multilayer perceptron and probabilistic neural network, obtaining a 98.10% accuracy for the testing sample.…”