Explaining qualifications in audit reports using a support vector machine methodology

The verification of whether the financial statements of a firm represent its actual position is of major importance for auditors, who should provide a qualified report if they conclude that the financial statements fail to meet this requirement. This paper implements support vector machines (SVMs) to develop models that may support auditors in this task. Linear and non‐linear models are developed and their performance is analysed using training samples of different size and out‐of‐sample/out‐of‐time data. The results show that all SVM models are capable of distinguishing between qualified and unqualified financial statements with satisfactory accuracy. The performance of the models over time is also explored. Copyright © 2005 John Wiley & Sons, Ltd.     

Feedforward neural networks in the classification of financial information

Financial research has given rise to numerous studies in which, on the basis of the information provided by financial statements, companies are classified into different groups. An example is that of the classification of companies into those that are solvent and those that are insolvent. Linear discriminant analysis (LDA) and logistic regression have been the most commonly used statistical models in this type of work. One feedforward neural network, known as the multilayer perceptron (MLP), performs the same task as LDA and logistic regression which, a priori, makes it appropriate for the treatment of financial information. In this paper, a practical case based on data from Spanish companies, shows, in an empirical form, the strengths and weaknesses of feedforward neural networks. The desirability of carrying out an exploratory data analysis of the financial ratios in order to study their statistical properties, with the aim of achieving an appropriate model selection, is made clear.     

Bankruptcy prediction of financially stressed firms: an examination of the predictive accuracy of artificial neural networks

This is an extension of prior studies that have used artificial neural networks to predict bankruptcy. The incremental contribution of this study is threefold. First, we use only financially stressed firms in our control sample. This enables the models to more closely approximate the actual decision processes of auditors and other interested parties. Second, we develop a more parsimonious model using qualitative ‘bad news’ variables that prior research indicates measure financial distress. Past research has focused on the ‘usefulness’ of accounting numbers and therefore often ignored non‐accounting variables that may contribute to the classification accuracy of the distress prediction models. In addition, rather than use multiple financial ratios, we include a single variable of financial distress using the Zmijewski distress score that incorporates ratios measuring profitability, liquidity, and solvency. Finally, we develop and test a genetic algorithm neural network model. We examine its predictive ability to that of a backpropagation neural network and a model using multiple discriminant analysis. The results indicate that the misclassification cost of the genetic algorithm‐based neural network was the lowest among the models. Copyright © 2001 John Wiley & Sons, Ltd.     

A neural network approach to the prediction of going concern status

The assessment of a firm's going concern status is not an easy task. To assist auditors, going concern prediction models based on statistical methods such as multiple discriminant analysis and logit/probit analysis have been explored with some success. This study attempts to look at a different and more recent approach—neural networks. In particular, a neural network model of the feedforward, backpropagation type was constructed to predict a firm's going concern status from six financial ratios, using a data set containing 165 non-going concerns and 165 matched going concerns. On an evenly distributed hold-out sample, the trained network model correctly predicted all 30 test cases. The results suggest that neural networks can be a promising avenue of research and application in the going concern area.     

Deep networks for predicting direction of change in foreign exchange rates

Trillions of dollars are traded daily on the foreign exchange (forex) market, making it the largest financial market in the world. Accurate forecasting of forex rates is a necessary element in any effective hedging or speculation strategy in the forex market. Time series models and shallow neural networks provide acceptable point estimates for future rates but are poor at predicting the direction of change and, hence, are not very useful for supporting profitable trading strategies. Machine learning classifiers trained on input features crafted based on domain knowledge produce marginally better results. The recent success of deep networks is partially attributable to their ability to learn abstract features from raw data. This motivates us to investigate the ability of deep convolution neural networks to predict the direction of change in forex rates. Exchange rates for the currency pairs EUR/USD, GBP/USD and JPY/USD are used in experiments. Results demonstrate that trained deep networks achieve satisfactory out‐of‐sample prediction accuracy.     

Bankruptcy prediction using fuzzy convolutional neural networks

We propose a combined method for bankruptcy prediction based on fuzzy set qualitative comparative analysis (fsQCA) and convolutional neural networks (CNN). Currently, CNNs are being applied to various fields, and in some areas are providing higher performance than traditional models. In our proposed method, a CNN uses calibrated variables from fuzzy sets to improve performance accuracy. In addition, there are no published studies on the effect of feature selection at the input level of convolutional neural networks. Therefore, this study compares four well-known feature selection methods used in financial distress prediction, (t-test, stepdisc discriminant analysis, stepwise logistic regression and partial least square discriminant analysis) to investigate their effect on classification performance. The results show that fuzzy convolutional neural networks (FCNN) lead to better performance than when using traditional methods.     

DETERMINANTS OF CAPITAL STRUCTURE IN THE UK RETAIL INDUSTRY: A COMPARISON OF MULTIPLE REGRESSION AND GENERALIZED REGRESSION NEURAL NETWORK

Firms need to rely on different financing sources, but the question is how capital structure is determined for a particular industry. Our aim is to undertake an investigation into the factors which determine capital structure in the UK retail industry. Our initial sample consists of 163 (final sample: 100) UK retail companies, using data from 2000 in order to analyse capital structure from 2002 to 2006. Nonlinear models tend to be unduly neglected in capital structure research, and so we apply generalized regression neural networks (GRNNs), which are compared with conventional multiple regressions. We utilize a hold‐out sample for the multiple regressions to make them comparable with the GRNNs. Stability of the data is also confirmed. Our main findings are: net profitability and the depreciation‐to‐sales ratio are key determinants of capital structure based on GRNNs, while two more variables are added in the multiple regressions, namely size and quick ratio; there is strong support for the pecking‐order theory; both root‐mean‐square errors and mean absolute errors are much lower for the GRNNs than those for the multiple regressions for overall, training and testing datasets. The potential benefit of this research to financial managers and investors in the UK retail sector is the identification of the overriding role of net profitability in reducing the financial risk from high levels of gearing. Copyright © 2012 John Wiley & Sons, Ltd.     

Profitability of alternative methods of combining the signals from technical trading systems

Past efforts determining the profitability of technical analysis reached varied conclusions. We test the profitability of a composite prediction that uses buy and sell signals from technical indicators as inputs. Both machine learning methods, like neural networks, and statistical methods, like logistic regression, are used to get predictions. Inputs are signals from trend‐following and mean‐reversal technical indicators in addition to the variance of prices. Four representative commodities from agricultural, livestock, financial, and foreign exchange futures markets are selected to determine profitability. Special care is taken to avoid data snooping error. Both neural networks and statistical methods did not show consistent profitability.     

A novel fuzzy neural approach to data reconstruction and failure prediction

Bank failure prediction is of great importance to a bank's clients, policy-makers and regulators. Various traditional models have been employed to study bank failures. Unfortunately, their performances are unsatisfactory. In this paper, the pseudo-outer product fuzzy neural network using the compositional rule of inference and singleton fuzzifier (POPFNN-CRI(S))-based bank failure prediction model is proposed. It employs computational bank failure analysis techniques coupled with reconstruction of missing financial data in financial covariates that are available from publicly available financial statements as inputs. The performance of the proposed model is assessed through the classification rate of 3636 US banks observed over a 21-year period. The effects of missing data reconstruction are investigated. Copyright © 2009 John Wiley & Sons, Ltd.     

Corporate culture and financial statement comparability

In this paper, I empirically examine the influence of corporate culture on the comparability of financial statements. I predict that firms with strong corporate cultures have less-opportunistic managers, who make homogenous decisions when faced with similar economic events, resulting in greater accounting comparability. For a sample of U.S. companies, I find empirical evidence consistent with this prediction: firms with strong corporate cultures have greater peer- and industry-level comparability. These results are robust to using an entropy-balanced sample, correcting for sample selection bias using Heckman's two-step procedure, and employing different measures of corporate culture strength. Further analysis reveals that sudden CEO turnovers that move firms towards (away from) a stronger corporate culture positively (negatively) influence post-turnover accounting comparability. My results provide new insights on the role of corporate culture for financial reporting.     

A comparison of nearest neighbours,discriminant and logit models for auditing decisions

This study investigates the efficiency of k-nearest neighbours (k-NN) in developing models for estimating auditors' opinions, as opposed to models developed with discriminant and logit analyses. The sample consists of 5276 financial statements, out of which 980 received a qualified audit opinion, obtained from 1455 private and public UK companies operating in the manufacturing and trade sectors. We develop two industry-specific models and a general one using data from the period 1998–2001, which are then tested over the period 2002–2003. In each case, two versions of the models are developed. The first includes only financial variables. The second includes both financial and non-financial variables. The results indicate that the inclusion of credit rating in the models results in a considerable increase both in terms of goodness of fit and classification accuracies. The comparison of the methods reveals that the k-NN models can be more efficient, in terms of average classification accuracy, than the discriminant and logit models. Finally, the results are mixed concerning the development of industry-specific models, as opposed to general models. Copyright © 2007 John Wiley & Sons, Ltd.     

Assessing the Impact of Fair‐Value Accounting on Financial Statement Analysis: A Data Envelopment Analysis Approach

The historical‐cost and prudence principles have guided accounting for financial investments and tangible fixed assets in many jurisdictions around the globe. This situation might change as a consequence of the increasing number of countries adopting International Financial Reporting Standards (IFRS), which, to some extent, permit accounting on a fair‐value basis. It is unclear how such a change would affect the analysis of financial statements and to what extent it could modify analysts' perceptions of companies' condition and performance. This paper attempts to shed some light on this issue by restating the financial investments and tangible fixed assets of a sample of 85 Spanish insurance companies, applying fair value instead of historical‐cost‐based valuations and by simulating analyst perception of these companies' efficiency and profitability for both sets of data using data envelopment analysis (DEA). We find that the numbers on the face of the financial statements change considerably and observe that the magnitude of these changes varies between companies and classes of assets. However, only in a few cases does a change in the valuation basis lead to a relevant change in DEA scores; within our sample, the overall assessment of companies with regard to efficiency and profitability remains largely the same under both valuation bases. These findings seem to indicate that a change from historical‐cost to fair‐value accounting could alter analyst perceptions of a limited number of companies but likely will not have a major impact on the appraisal of the majority of them.     

Predicting mispricing of initial public offerings

This article investigates the ability of neural network models to predict mispricing of initial public offerings (IPOs). The aim is to improve the modest explanatory power of existing models that are based on the theory of asymmetrically informed economic agents surrounding post‐issue market value of IPOs. This study develops and compares linear regression and neural network models. The results show that modelling variable interactions and non‐linearity allows a potentially fruitful approach for stagging in IPOs. Neural networks have been criticized for being a black box; however, this paper shows that, by using sensitivity analysis, neural networks can provide a reasonable explanation of their predictive behaviour and direction of association between variables. Copyright © 2005 John Wiley & Sons, Ltd.     

Making trading decisions for financial‐engineered derivatives: a novel ensemble of neural networks using information content

Over the last decades, there has been a growing interest in applying artificial intelligence techniques to solve a spectrum of financial problems. A number of studies have shown promising results in using artificial neural networks (ANNs) to guide investment trading. Given the expanding role of ANNs in financial trading, this paper proposes the use of a hybrid neural network, which consists of two independent ANN architectures, and comparatively evaluates its performance against independent ANNs and econometric models in the trading of a financial‐engineered (synthetic) derivative composed of options on foreign exchange futures. We examine the financial profitability and the market timing ability of the competing neural network models and statistically compare their attributes with those based on linear and nonlinear statistical projections. A random walk model and the option pricing method are also included as benchmarks for comparison. Our empirical investigation finds that, for each of the currencies analysed, trading strategies guided by the proposed dual network are financially profitable and yield a more stable stream of investment returns than the other models. Statistical results strengthen the notion that diffusion of information contents and cross‐validation between the independent components within the dual network are able to reduce bias and extreme decision making over the long run. Moreover, the results are robust with respect to different levels of transaction costs. Copyright © 2009 John Wiley & Sons, Ltd.     

An Evaluation of the Corporate Takeover Model Using Neural Networks

Neural networks have been found to be promising in financial prediction tasks like bankruptcy and loan defaults. Their use in the capital markets is relatively new, although they have been used with some success in picking undervalued stocks. Accurate prediction of corporate takeover targets results in high financial payoffs. Researchers have used statistical procedures like logistic regression with little success in predicting corporate takeover targets. We use neural networks that are capable of producing complex mapping functions to predict mergers. We develop several neural network models carefully controlling for overfitting. Our results indicate that although neural networks map the data very well, they do not predict merger targets significantly better than logistic regression. This strongly suggests that the financial models used to predict mergers are inadequate. Firms should approach the development of merger prediction models cautiously and identify other factors that are more likely to predict mergers. Attempts to apply better analysis techniques to existing models will most likely produce similar results.     

Words that shake traders : The stock market's reaction to central bank communication in real time

This paper investigates the effects of Federal Reserve's decisions and statements on U.S. stock and volatility indices (Dow Jones Industrial Average, NASDAQ 100, S&P 500, and VIX) using a high-frequency event-study analysis. I find that both the surprise component of policy actions and official communication have statistically significant and economically relevant effects on equity indices, with statements having a much greater explanatory power of the reaction of stock prices to monetary policy. For instance, around 90% of the explainable variation in S&P 500 is due to the surprise component of Fed's statements. This paper also shows that equity indices tend to incorporate FOMC monetary surprises within 40 min from the announcement release. Finally, I find that these results are robust along several dimensions. In particular, I consider different estimators, such as the Generalized Empirical Likelihood, and I extend the sample to include the recent period of heightened financial stress. This sensitivity analysis corroborates that central bank communication about its future policy intentions is a key driver of stock returns.     

Is Polaris a HOG? A Case Study*

The main usefulness of a general purpose financial statement centers on its comparability to the financials produced by an entity's competition. This case works for both undergraduate and graduate students because it offers comparisons between two well known recreation industry companies. Financial statement analysis is a lesson which serves as an appropriate capstone to financial accounting education. Analyzing the financial statements of competing entities explains “why” accountants must implement the intricate “how” which produces the statements and the resulting decision models such as ratios and common‐sized statements. Compared to previous generations, current students will be more responsible for managing their own retirement funds. This case can help students start to appreciate investment analysis by providing enough detail for any level student to conduct financial statement evaluations that make comparisons to find successful fundamental business strategies.     

Using client performance measures to identify pre-engagement factors associated with qualified audit reports in Greece

The purpose of this paper is to test the extent to which client (corporate) performance measures can be used to enhance the ability to discriminate between the choice of a qualified or unqualified (clean) audit report. Audit firms face the risk of losing the client if they issue a qualification. On the other hand, failing to qualify exposes the auditor to potential lawsuits and loss of reputation. We examined the financial statements, auditors' opinions, and financial statements notes for companies in Greece that received a qualified audit report and for those that received an unqualified audit report. We modeled the auditor's qualification using a multicriteria decision aid classification method (UTADIS—UTilités Additives Discriminates) and compared it with other multivariate statistical techniques such as discriminant and logit analysis. The qualification decision is explained by financial ratios and by nonfinancial information such as the client litigation. The developed models are accurate in classifying the total sample correctly with rates of almost 80%.     

Predicting Private Company Failures in Italy Using Financial and Non‐financial Information

The study assesses the use of non‐financial information in predicting financial distress in private companies by developing credit risk models tailored to Italian private companies. The in‐sample and out‐of‐sample prediction test results are indicative of the incremental predictive ability of the two new non‐financial variables, that is, number of shareholders and number of subsidiaries, over accounting ratios and other widely used non‐financial information, including firm age and industry dummies. To be more specific, number of shareholders and number of subsidiaries are negatively associated with private company failures, and the models augmented by the two non‐financial variables improve forecasting performance from acceptable discrimination to excellent discrimination over one‐ to three‐year time horizons.