A neural network approach to understanding implied volatility movements

The expected response of an index volatility surface to index movements is quite different in high and low volatility environments     

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.     

Toward a better expert system for auditor going concern opinions using Bayesian network inflation factors

We develop an analytical model intended as the first stage in the development of expert systems to improve auditor knowledge in, and assist in the decision process of, Going Concern Opinions (“GCOs”). Our approach is consistent with a design science approach to developing information systems, resulting in an initial artifact, the mathematical model, which can, through iterative design science and behavioral research, inform a technology-based expert system. Based on Bayesian networks, our model provides insights about auditors’ revision, or inflation, of the probability to issue a GCO based on the interrelationship that forms with the incremental existence of one, two, or three publicly observable financial statement risk factors – net operating loss, negative cash flows from operations, and negative working capital. We calculate the revised probabilities using empirical data of GCOs from 2004 to 2015. Results reveal that the incremental relationship (one, two, or three factors present) effectively models expert auditors’ decisions to issue a GCO, and suggests the existence of these measurable inflation factors that represent situational and auditor-specific factors. We also find that Non-Big Four auditors inflate these factors differently than Big Four auditors to arrive at a decision to issue a GCO.     

Time-varying neural network for stock return prediction

We consider the problem of neural network training in a time-varying context. Machine learning algorithms have excelled in problems that do not change over time. However, problems encountered in financial markets are often time varying. We propose the online early stopping algorithm and show that a neural network trained using this algorithm can track a function changing with unknown dynamics. We compare the proposed algorithm to current approaches on predicting monthly US stock returns and show its superiority. We also show that prominent factors (such as the size and momentum effects) and industry indicators exhibit time-varying predictive power on stock returns. We find that during market distress, industry indicators experience an increase in importance at the expense of firm level features. This indicates that industries play a role in explaining stock returns during periods of heightened risk.     

The horse has bolted: revisiting the market reaction to going concern modifications of audit reports

This paper examines what value is added by an audit report through an investigation of the information content for first‐time going concern modifications (GCMs). Consistent with prior research, we find no evidence of a short‐term market reaction to the public announcement of a first‐time GCM. We document a significant adverse medium‐term market reaction in the 12 months prior to a first‐time GCM announcement, but find no evidence of a persistent market underreaction in the 12 months following the announcement. These results are consistent with an audit opinion fulfilling an attestation function and confirming the deteriorating financial condition of a firm.     

Delays in reporting price-sensitive information: The case of going concern

Regulators require firms to disclose all price-sensitive information at the earliest possible date. The going-concern opinion constitutes a fundamental uncertainty for the firm and thus is likely to be of a price-sensitive nature. This paper explores whether going-concern uncertainty disclosures are price sensitive in the London market, and then tests whether managements report such audit report information to investors on a timely basis. We capitalize on a London Stock Exchange regulatory loophole which, in effect, allows financially-distressed firms to choose either to report a forthcoming going-concern at the preliminary results announcement stage, or to delay this crucial information to their annual report release. In line with the regulatory requirements, we expect that firms with more price-sensitive, i.e., more serious, adverse news will disclose their forthcoming going-concern opinion at the earliest stage i.e., in their preliminary announcement, rather than delay to their annual report.     

Determinants of going concern opinions and audit fees for development stage enterprises

Startup entities have been the focus of much political and academic interest recently. Development stage enterprises (DSEs), as defined by SFAS 7, are startup entities for which some publicly available information exists. New accounting standards have removed the DSE designation and related extra reporting requirements, and placed more responsibility on owners and managers to assess the ability of entities to continue as a going concern. We examined information from financial statements and audit reports of companies previously reporting as DSEs to investigate what increases the likelihood of receiving a going concern modification in auditors' opinions (GCO) and what affects audit fees. Our overall analyses indicate that the asset size of DSEs, negative working capital, and prior-year going concern modifications consistently influence going concern modifications to auditors' opinions. Managers should clearly consider these conditions when making their assessment of their companies' future going concern status. Our results indicate that the size of the audit firm did not influence the going concern modification decision, but Big4 auditors charge significantly higher fees than other auditors. Thus, managers/owners of DSEs should weigh the benefits of having a Big4 firm audit on their financial statements against the higher fees charged by those firms.     

Does the stock market underreact to going concern opinions? Evidence from the U.S. and Australia

We examine 12-month returns following disclosure of first-time going concern (GC) opinions in the U.S. and Australia. We find no evidence of significant negative abnormal returns associated with GC opinions in Australia. In the U.S., negative abnormal returns subsequent to GC opinions are sensitive to choice of expected returns—notably, there are no significant negative abnormal returns when using factor models or after controlling for momentum. Overall, contrary to Taffler, Lu, Kausar's [2004. In denial? Stock market underreaction to going-concern audit report disclosures. Journal of Accounting and Economics 38, 263–285.] U.K. results, we are unable to document a market anomaly in the U.S. or Australia associated with GC opinions.     

Risk assessment in internal auditing: a neural network approach

Risk assessment is a systematic process for integrating professional judgments about relevant risk factors, their relative significance and probable adverse conditions and/or events leading to identification of auditable activities (IIA, 1995, SIAS No. 9). Internal auditors utilize risk measures to allocate critical audit resources to compliance, operational, or financial activities within the organization (Colbert, 1995). In information rich environments, risk assessment involves recognizing patterns in the data, such as complex data anomalies and discrepancies, that perhaps conceal one or more error or hazard conditions (e.g. Coakley and Brown, 1996; Bedard and Biggs, 1991; Libby, 1985). This research investigates whether neural networks can help enhance auditors’ risk assessments. Neural networks, an emerging artificial intelligence technology, are a powerful non‐linear optimization and pattern recognition tool (Haykin, 1994; Bishop, 1995). Several successful, real‐world business neural network application decision aids have already been built (Burger and Traver, 1996). Neural network modeling may prove invaluable in directing internal auditor attention to those aspects of financial, operating, and compliance data most informative of high‐risk audit areas, thus enhancing audit efficiency and effectiveness. This paper defines risk in an internal auditing context, describes contemporary approaches to performing risk assessments, provides an overview of the backpropagation neural network architecture, outlines the methodology adopted for conducting this research project including a Delphi study and comparison with statistical approaches, and presents preliminary results, which indicate that internal auditors could benefit from using neural network technology for assessing risk. Copyright © 1999 John Wiley & Sons, Ltd.     

Calibrating rough volatility models: a convolutional neural network approach

In this paper, we use convolutional neural networks to find the Hölder exponent of simulated sample paths of the rBergomi model, a recently proposed stock price model used in mathematical finance. We contextualise this as a calibration problem, thereby providing a very practical and useful application.     

An options-pricing approach to election prediction

Election forecasting errors appear chiefly due to the mode of extracting outcomes from the polled share of the vote     

A neural network enhanced volatility component model

Volatility prediction, a central issue in financial econometrics, attracts increasing attention in the data science literature as advances in computational methods enable us to develop models with great forecasting precision. In this paper, we draw upon both strands of the literature and develop a novel two-component volatility model. The realized volatility is decomposed by a nonparametric filter into long- and short-run components, which are modeled by an artificial neural network and an ARMA process, respectively. We use intraday data on four major exchange rates and a Chinese stock index to construct daily realized volatility and perform out-of-sample evaluation of volatility forecasts generated by our model and well-established alternatives. Empirical results show that our model outperforms alternative models across all statistical metrics and over different forecasting horizons. Furthermore, volatility forecasts from our model offer economic gain to a mean-variance utility investor with higher portfolio returns and Sharpe ratio.     

A structured approach to conflict prediction in information systems strategy implementation

Organizational power and politics are the central issues of this paper. By developing a model of organizational power it is possible to determine whether an organizational change initiative is likely to be politically feasible. The formal model described has been derived largely from research reported in the social sciences. The modelling process involved using formal methods, in logic and entity relationship analysis, to discover an effective and consistent means of representing key organizational power concepts. The result is an advisory expert system called MP/L1 that can be employed by change agents to predict likely sources of potential resistance to major change initiatives and to suggest tactics that might be effective in combating anticipated resistance. Industrial experience with MP/L1 to date indicates that it has significant potential as a change management tool within the IS strategy-implementation domain. © 1998 John Wiley & Sons, Ltd.     

Financial innovation and Divisia monetary indices in Taiwan: a neural network approach

In this paper a weighted index measure of money using the ‘Divisia’ formulation is constructed for the Taiwan economy and its inflation forecasting potential is compared with that of its traditional simple sum counterpart. This research extends an earlier study by Gazely and Binner by examining the theory that rapid financial innovation, particularly during the financial liberalization of the 1980s, has been responsible for the poor performance of conventional simple sum monetary aggregates. The Divisia index is adjusted in two ways to allow for the major financial innovations that Taiwan has experienced since the 1970s. The technique of neural networks is used to allow a completely flexible mapping of the variables and a greater variety of functional form than is currently achievable using conventional econometric techniques. Results suggest that superior tracking of inflation is possible for networks that employ a Divisia M2 measure of money that has been adjusted to incorporate a learning mechanism to allow individuals to gradually alter their perceptions of the increased productivity of money. Divisia measures of money appear to offer advantages over their simple sum counter parts as macroeconomic indicators.     

An investigation of recent changes in going concern reporting decisions among Big N and non-Big N auditors

Corporate accounting failures and regulatory proceedings that led to the enactment of the Sarbanes–Oxley Act of 2002 increased the scrutiny of auditors. We investigate whether these events resulted in a change in auditor behavior with respect to going concern reporting. Generally speaking, we find that non-Big N auditors became more conservative while Big N auditors became more accurate. Specifically, non-Big N auditors issued more going concern opinions to both failing and non-failing clients post-2001, reducing their Type II misclassifications at the expense of increased Type I misclassifications. However, Big N auditors decreased their Type I misclassifications with no corresponding increase in Type II misclassifications. Thus, our findings suggest that increased auditor scrutiny resulted in performance improvements in the area of going concern reporting primarily for larger auditors. For smaller auditors, improved going concern accuracy for subsequently bankrupt clients came at the cost of more going concern opinions being issued to subsequently non-failing clients.     

Sovereign risk spillovers: A network approach

Understanding the global financial network for sovereign debt, particularly with a focus on interaction and spillover effects of sovereign risk, has become important for policy makers as they look to protect the stability of their economies. Using high dimensional Vector Autoregression techniques and network simulation on Sovereign Credit Default Swaps (CDS)’ data of 57 countries, we identify that the global sovereign CDS network is fully integrated as there is virtually no country without any connection to at least one specific node in the system. However, each country has a unique attribute in the network, as a risk exporter or importer and/or risk transmitter. Among developed countries, the US (unsurprisingly) holds the dominant position as a risk exporter while Germany is identified as a connecting country that transmits shocks. The most connected countries in the sovereign CDS network belong to the new European Union members. We examine possible drivers of the network relationships observed, in order to better understand the risk transmission process, and find that connections in the sovereign risk network are stronger within regional groups and countries with the same level of economic development. Central and Eastern Europe and Middle East and Africa have more interactive networks than Northern Western Europe, Asia Pacific and Latin America. We also identify that financial volatility and economic policy uncertainty increase the interactions in market-based default risk assessment.     

<Emphasis Type="Bold">Measuring the market response to going concern modifications: the importance of disclosure timing</Emphasis>

Auditor going concern modifications (GCMs) are intended to provide market participants with information related to financial distress, and prior research suggests that the disclosure of a GCM elicits a substantial negative market reaction from investors. In this study, we investigate the market reaction to GCMs in a contemporary disclosure regime and consider whether the observed market reaction is confounded by other material disclosures. We find that the majority of GCMs are issued concurrently with earnings announcements (EAs) and that EAs in the year of new GCMs elicit large negative cumulative abnormal returns (CARs). We also find that CARs surrounding GCMs are significantly more negative when GCMs are disclosed with EAs versus following EAs. We then evaluate whether GCMs convey distress that is incremental to EA disclosures by measuring i) the market reaction to GCMs disclosed following EAs, and ii) whether EA CARs are substantially more negative for companies disclosing GCMs with EAs as opposed to after EAs. In both cases, we find that the incremental market response to GCMs is statistically weak and much smaller in economic magnitude than is suggested by prior research. Finally, we find that management disclosures in EAs, rather than the presence of a GCM, appear to convey information that investors use to anticipate bankruptcy. Taken together, these findings suggest that GCMs are confounded by other significant disclosures and that the informational benefits of GCM reporting are significantly smaller than previously thought.     

A continuous time Bayesian network classifier for intraday FX prediction

Abstract

Prediction of foreign exchange (FX) rates is addressed as a binary classification problem in which a continuous time Bayesian network classifier (CTBNC) is developed and used to solve it. An exact algorithm for inference on CTBNC is introduced. The performance of an instance of these classifiers is analysed and compared to that of dynamic Bayesian network by using real tick by tick FX rates. Performance analysis and comparison, based on different metrics such as accuracy, precision, recall and Brier score, evince a predictive power of these models for FX rates at high frequencies. The achieved results also show that the proposed CTBNC is more effective and more efficient than dynamic Bayesian network classifier. In particular, it allows to perform high frequency prediction of FX rates in cases where dynamic Bayesian networks-based models are computationally intractable.     

A comparison of statistical tests for the adequacy of a neural network regression model

An integral part of econometric practice is to test the adequacy of model specifications. If a model is adequately specified, it should not leave interesting features of the data-generating process in the errors. Despite the common tradition, the importance of diagnostic checking as a safeguard against mis-specification has only recently been recognized by neural network (NN) practitioners, possibly because this type of semi-parametric methodology was not originally designed for economic and financial applications. The purpose of this paper is to compare a number of analytical statistical testing procedures suitable to diagnostic checking on a neural network regression model. We present the standard Lagrange multiplier (LM) testing framework designed under the assumption of identically distributed disturbances and also examine two modifications that are robust to heteroskedasticity in errors. One modification also gives the researcher an opportunity to incorporate information concerning the volatility structure of the data-generating process in the testing procedure. By means of a Monte Carlo simulation, we investigate the performance of these tests under GARCH-type heteroskedasticity in errors and various distributional assumptions. The results show that although the primary concern of the researcher may be to design a regression model that accurately captures relations in the mean of the conditional distribution, developing a good approximation of the underlying volatility structure generally increases the efficiency of tests in detecting non-adequacy of a NN model.