Detecting Accounting Fraud in Publicly Traded U.S. Firms Using a Machine Learning Approach

We develop a state-of-the-art fraud prediction model using a machine learning approach. We demonstrate the value of combining domain knowledge and machine learning methods in model building. We select our model input based on existing accounting theories, but we differ from prior accounting research by using raw accounting numbers rather than financial ratios. We employ one of the most powerful machine learning methods, ensemble learning, rather than the commonly used method of logistic regression. To assess the performance of fraud prediction models, we introduce a new performance evaluation metric commonly used in ranking problems that is more appropriate for the fraud prediction task. Starting with an identical set of theory-motivated raw accounting numbers, we show that our new fraud prediction model outperforms two benchmark models by a large margin: the Dechow et al. logistic regression model based on financial ratios, and the Cecchini et al. support-vector-machine model with a financial kernel that maps raw accounting numbers into a broader set of ratios.     

Audit committee quality indices,reporting quality and firm value

We investigate the relation between audit committee (AC) quality indices, financial reporting, internal control quality and firm value using a US dataset for the period 2002–12. The indices are developed by linking firm value with principal component analysis (PCA) factors based on a broad set of 82 AC variables, some of which influence the quality of the AC, but are not addressed in prior literature. Significant AC factors include ‘overlapping directors’, ‘busyness’ and ‘foreign director’, and we use these factors to develop ‘high’ and ‘low’ AC quality indices. We show that low AC quality firms are more likely to manage earnings, be external auditor dependent with respect to non‐audit tax services, and switch to a lower quality auditor. Low AC quality firms are also more likely to have internal control concerns disclosed by predecessor auditors, including accounting issues, financial restatements, audit opinion concerns and deficiencies that undermine internal control effectiveness. Further, they are more likely to receive an audit report containing additional explanatory notes. Conversely, high AC quality firms are significantly less likely to have these concerns. Our findings highlight the value of using AC quality indices in delivering greater oversight of the financial reporting process.     

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.     

INSOLVENCY PREDICTION IN THE PRESENCE OF DATA INCONSISTENCIES

In this paper we use data inconsistencies as an indicator of financial distress. Traditional models for insolvency prediction normally ignore inconsistent data, either by removing or replacing it. Instead of removing that information, we propose a new variable to capture it; using it together with traditional accounting variables (based on financial ratios) for the purpose of insolvency prediction. Computational tests use three datasets based on the financial results of 2033 Brazilian Health Maintenance Organizations over 7 years (2001 to 2007). Sixteen classification methods were used to evaluate whether or not the new variable impacted solvency prediction. Tests show a statistically significant improvement in classification accuracy – average results improve 1.3 (p = 0.003) and 1.8 (p = 0.006) percentage points, for 10‐fold and leave‐one‐out cross‐validations respectively. In addition, the analysis of false positives and false negatives shows that the new variable reduces the potentially harmful misclassification of false negatives (i.e. financially distressed companies being classified as financially healthy) and also reduces the estimated overall error rate. Regarding the extensibility of the results, even though this work uses data from Brazilian companies only, the calculation of the financial ratios variables, as well as the inconsistencies, could be extended to most companies worldwide subject to governmental accounting regulations aligned with the International Financial Reporting Standards. Copyright © 2014 John Wiley & Sons, Ltd.     

Financial distress and bankruptcy prediction among listed companies using accounting,market and macroeconomic variables

Using a sample of 23,218 company-year observations of listed companies during the period 1980–2011, the paper investigates empirically the utility of combining accounting, market-based and macro-economic data to explain corporate credit risk. The paper develops risk models for listed companies that predict financial distress and bankruptcy. The estimated models use a combination of accounting data, stock market information and proxies for changes in the macro-economic environment. The purpose is to produce models with predictive accuracy, practical value and macro dependent dynamics that have relevance for stress testing. The results show the utility of combining accounting, market and macro-economic data in financial distress prediction models for listed companies. The performance of the estimated models is benchmarked against models built using a neural network (MLP) and against Altman's (1968) original Z-score specification.     

Detecting financial statement fraud using dynamic ensemble machine learning

Our study uses Machine learning to develop an advanced fraud detection model that can detect fraudulent firms. We build our model using raw financial and non-financial variables following prior literature. In addition, we introduce the Dynamic Ensemble Selection algorithm to the fraud detection literature, which combines individual classifiers dynamically to make a final prediction. Using several performance evaluation metrics, we find that our model can outperform several machine learning models used in recent studies.     

Explanators of Local Government Distress

This article develops a statistical model to explain sources of distress in local government. Whereas ‘financial distress’ in the private sector has been equated with a failure to meet financial commitments, here ‘distress’ is interpreted as an inability to maintain pre-existing levels of services to the community. Since the late 1990s local councils in an Australian state (New South Wales) have been required to estimate the cost of restoring infrastructure assets to a satisfactory condition (a requirement which predates that form of reporting on infrastructure condition introduced as an option in U.S. GASB 34). Information regarding the cost of restoring infrastructure is used in this study as a proxy for levels of distress (in contrast to the binary classification that characterizes much of prior private sector financial distress research). Data regarding service levels for a sample of 161 councils for 2001 and 2002 were used and a multiple regression model was estimated and interpreted. The main findings were that the degree of distress in local councils is positively associated with the size of the population they serve and the size and composition of their revenues. Road maintenance costs featured prominently in results, as higher road program costs were associated with higher levels of distress (particularly when interacted with other variables). However, the revenue generating capacity of councils had the strongest statistical impact on local government distress. Councils with lower percentages of rates revenue to total revenue and lower ordinary revenue levels to total assets were typically identified as more distressed. However, no systematic evidence was found that rural councils have higher distress levels than urban councils (i.e., both rural and urban councils serving larger populations were relatively more distressed than councils serving smaller populations). It is suggested that the model (or modifications thereof) may serve as an early warning system for those monitoring the circumstances and performance of local governments.     

Accounting-based versus market-based cross-sectional models of CDS spreads

Models of financial distress rely primarily on accounting-based information (e.g. [Altman, E., 1968. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance 23, 589–609; Ohlson, J., 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research 19, 109–131]) or market-based information (e.g. [Merton, R.C., 1974. On the pricing of corporate debt: The risk structure of interest rates. Journal of Finance 29, 449–470]). In this paper, we provide evidence on the relative performance of these two classes of models. Using a sample of 2860 quarterly CDS spreads we find that a model of distress using accounting metrics performs comparably to market-based structural models of default. Moreover, a model using both sources of information performs better than either of the two models. Overall, our results suggest that both sources of information (accounting- and market-based) are complementary in pricing distress.     

Accounting for the financialized UK and US national business model

In this paper we adopt a ‘business model’ conceptual framework grounded in accounting to describe the processes and mechanisms of national economic development and transformation. We locate national business models within a broad econo-sphere where they evolve and adapt to information arising out of stakeholder/institutional interactions. These interactions congeal into reported financial numbers that are presented as current income flows (income, expenditure), balance sheet accumulations and changes in net worth (assets and liabilities outstanding). We employ financial data from national accounts to specifically describe how the US and UK national business models have become financialized as ongoing capitalizations run ahead of earnings capacity. This process of interminable re-capitalization is conditioned by variable institutional and sub-institutional sector characteristics. However, in financialized national business models the system of accounting takes on added analytical significance because it ‘transmits rather than contains’ and ‘amplifies rather than dampens’ adverse financial disturbance as capitalizations are recalibrated up or down in secondary markets.     

Predicting default of Russian SMEs on the basis of financial and non-financial variables

This article develops two models for predicting the default of Russian Small and Medium-sized Enterprises (SMEs). The most general questions that the article attempts to answer are ‘Can the default risk of Russian SMEs be assessed with a statistical model?’ and ‘Would it sufficiently demonstrate high predictive accuracy?’ The article uses a relatively large data set of financial statements and employs discriminant analysis as a statistical methodology. Default is defined as legal bankruptcy. The basic model contains only financial ratios; it is extended by adding size and age variables. Liquidity and profitability turned out to be the key factors in predicting default. The resulting models have high predictive accuracy and have the potential to be of practical use in Russian SME lending.     

Predicting the outcome following bankruptcy filing: a three-state classification using neural networks

This paper uses artificial neural networks (ANNs), multi-state ordered logit and nonparametric multiple discriminant analysis (NPDA) for predicting the three-state outcome of bankruptcy filing. The study compares the classification accuracy of these procedures. It differs from previous studies on predicting financial distress by focusing on the firm after the filing of bankruptcy using accounting data, market data, and court-related information. Following the filing and through court approval the bankruptcy is resolved as firms are either acquired by other firms, emerging as independent operating entities, or liquidated. Distinguishing this three-state outcome is more complex than discriminating between healthy and financially distressed firms. Models suggested in previous studies for predicting the two-group financial distress perform poorly for our three-state scenario. Therefore, we develop models which focus on characteristics relevant for the bankruptcy resolution. We use a sample of 237 publicly traded firms which have complete data. For the entire sample and estimation samples, ANNs provide significantly better three-state classification than logit and NPDA. However, for some holdout samples the differences in classification accuracies are statistically insignificant. © 1997 John Wiley & Sons, Ltd.     

NEURO‐GENETIC PREDICTIONS OF CURRENCY CRISES

We create a neuro‐genetic (NG) model for predicting currency crises by using a genetic algorithm for specifying (1) the combination of inputs, (2) the network configuration and (3) the training parameters for a back‐propagation artificial neural network (ANN). The performance of the NG model is evaluated by comparing it with standalone probit and ANN models in terms of utility for a policy decision‐maker. We show that the NG model provides better in‐sample and out‐of‐sample performance, as well as provides an automatic and more objective calibration of a predictive ANN model. We show that using a genetic algorithm for finding an optimal model specification for an ANN is not only less laborious for the analyst, but also more accurate in terms of classifying in‐sample and predicting out‐of‐sample crises. For a sufficiently parsimonious, but still nonlinear, model for generalized processing of out‐of‐sample data, the creation and evaluation of models is performed objectively using only in‐sample information as well as an early stopping procedure. Copyright © 2011 John Wiley & Sons, Ltd.     

Valuation Weights,Linear Dynamics and Accounting Conservatism: An Empirical Analysis

Residual income models provide an important theoretical link between equity valuation and financial statement variables. While various researchers have developed models of how accounting policy impacts on the structure of these models, empirical support for these models is at best weak and frequently contradictory. In this paper, we develop an analytical model, which identifies the dependency between valuation weights in residual income models and the associated structure of earnings information dynamics and accounting conservatism. In contrast to many earlier studies, we find strong evidence of conservatism in our reformulation of the linear dynamics. We proceed to test our predictions of the dependency of the weights on two measures of conservatism, the conventional measure of price‐to‐book ratio and the recent measure of a C‐Score index developed by Khan and Watts (2009) and find that the empirical results accord well with our theoretical predictions in the case of the former but not the latter measure.     

An additional analysis of estimation techniques for the degree of financial leverage

This study compares three different empirical proxies for the financial leverage component of a systematic risk‐composition model employed in prior financial research. We consider one static accounting measure and two elasticity‐based measures. We find that the traditional static accounting measure of financial leverage provides statistically different estimates of financial leverage when compared to estimates from elasticity‐based measures of the degree of financial leverage. The findings are important because the elasticity‐based models for the degree of financial leverage have clear theoretical links to market‐based models of systematic risk, while the static accounting measure of financial leverage does not. Practitioners and researchers should carefully consider why they are estimating financial leverage and choose the appropriate method for doing so given the goals and potential consequences for biased estimation.     

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.     

A Note on Selecting a Response Measure for Financial Distress

Since 1966, researchers have examined financial distress prediction models to determine the usefulness of accounting information to lenders. These researchers primarily used legal bankruptcy as the response variable for economic financial distress, or included legal bankruptcy with other events in dichotomous prediction models. However, theoretical models of financial distress normally define financial distress as an economic event, the inability to pay debts when due (insolvency). This study uses a loan default/accommodation response variable as a proxy for the inability to pay debts when due. The purpose of this note is to empirically test whether or not using the inability of a firm to pay debts when due, loan default/accommodation, as a response measure produces different results than using legal bankruptcy as the response measure. The study's empirical results show that legal bankruptcy and loan default/accommodation financial distress prediction models produce different statistical results, thus suggesting that the responses measure different constructs. A loan default/accommodation model also fits the data better than a bankrupt model. Our results suggest that a loan default/accommodation response may be a more appropriate measure to determine which accounting information is most useful to lenders in evaluating a firm's credit risk.     

Generative Bayesian neural network model for risk-neutral pricing of American index options

Financial models with stochastic volatility or jumps play a critical role as alternative option pricing models for the classical Black–Scholes model, which have the ability to fit different market volatility structures. Recently, machine learning models have elicited considerable attention from researchers because of their improved prediction accuracy in pricing financial derivatives. We propose a generative Bayesian learning model that incorporates a prior reflecting a risk-neutral pricing structure to provide fair prices for the deep ITM and the deep OTM options that are rarely traded. We conduct a comprehensive empirical study to compare classical financial option models with machine learning models in terms of model estimation and prediction using S&P 100 American put options from 2003 to 2012. Results indicate that machine learning models demonstrate better prediction performance than the classical financial option models. Especially, we observe that the generative Bayesian neural network model demonstrates the best overall prediction performance.     

A RADIAL BASIS FUNCTION APPROACH TO EARNINGS FORECAST

The fundamental management problem of decision making in a climate where future values of important variables are unknown and can at best be estimated using traditional statistical techniques is addressed. The incorporation of forecast models into management decision‐support systems is critical for the overall success of organizational accounting information systems, where managers require confidence in the information that they use. The neural network paradigm has been described as a promising nonparametric approach, negating the required, and sometimes restrictive, statistical assumptions. The application of the neural network paradigm to the area of earnings forecasting is presented. A radial basis function (RBF) approach is developed and tested empirically using data from the Hong Kong Hang Seng 100 Index and macroeconomic data, mimicking an actual business valuation/forecast exercise. Results show that the RBF approach is superior to regression and financial analysts in earnings forecast. Copyright © 2012 John Wiley & Sons, Ltd.