Leveraging the Strengths of Choice Models and Neural Networks: A Multiproduct Comparative Analysis*

Choice models and neural networks are two approaches used in modeling selection decisions. Defining model performance as the out‐of‐sample prediction power of a model, we test two hypotheses: (i) choice models and neural network models are equal in performance, and (ii) hybrid models consisting of a combination of choice and neural network models perform better than each stand‐alone model. We perform statistical tests for two classes of linear and nonlinear hybrid models and compute the empirical integrated rank (EIR) indices to compare the overall performances of the models. We test the above hypotheses by using data for various brand and store choices for three consumer products. Extensive jackknifing and out‐of‐sample tests for four different model specifications are applied for increasing the external validity of the results. Our results show that using neural networks has a higher probability of resulting in a better performance. Our findings also indicate that hybrid models outperform stand‐alone models, in that using hybrid models guarantee overall results equal or better than the two stand‐alone models. The improvement is particularly significant in cases where neither of the two stand‐alone models is very accurate in prediction, indicating that the proposed hybrid models may capture aspects of predictive accuracy that neither stand‐alone model is capable of on their own. Our results are particularly important in brand management and customer relationship management, indicating that multiple technologies and mixture of technologies may yield more accurate and reliable outcomes than individual ones.     

Hybridizing logistic regression with product unit and RBF networks for accurate detection and prediction of banking crises

As the current crisis has painfully proved, the financial system plays a crucial role in economic development. Although the current crisis is being of an exceptional magnitude, financial crises are recurrent phenomena in modern financial systems. The literature offers several definitions of financial instability, but for our purposes we identity financial crisis with banking crisis as the most common example of financial instability. In this paper we introduce a novel model for detection and prediction of crises, based on the hybridization of a standard logistic regression with product unit (PU) neural networks and radial basis function (RBF) networks. These hybrid approaches are fully described in the paper, and applied to the detection and prediction of banking crises by using a large database of countries in the period 1981–1999. The proposed techniques are shown to perform better than other existing statistical and artificial intelligence methods in this problem.     

Redundancy in Designs

The author contends that a previous Risk Analysis article overemphasized the pitfalls of incorporating redundancy into designs. Relevant aspects of that article are reviewed and commented upon, then the potentials and pitfalls of redundancy in systems and procedures are more broadly discussed. To provide a solid foundation for that discussion, some definitions for systems risk analysis terminology are presented. It is shown that pairs and larger sets of related failures (the physical causes of shortfalls in redundancy effectiveness) can be divided into two types: (1) cascading/induced failures and (2) common-external-cause failures. Each type has its own physical characteristics and implications for mathematical modeling. Service experience with large-commercial-airplane jet-engine propulsion systems is used to illustrate the two types of related failures. Finally, an overview is provided of event-sequence analysis, an alternative approach to systems risk analysis. When the possibility of related failures of mutually-redundant system elements must be accounted for, event-sequence analysis can usually do that better than fault-tree analysis.     

FIRM DEFAULT AND AGGREGATE FLUCTUATIONS

This paper studies the relationship between macroeconomic fluctuations and corporate defaults while conditioning on industry affiliation and an extensive set of firm‐specific factors. By using a panel data set for virtually all incorporated Swedish businesses over 1990–2009, a period which includes a full‐scale banking crisis, we find strong evidence for a substantial and stable impact from aggregate fluctuations on business defaults. A standard logit model with financial ratios augmented with macroeconomic factors can account surprisingly well for the outburst in business defaults during the banking crisis, as well as the subsequent fluctuations in default frequencies. Moreover, the effects of macroeconomic variables differ across industries in an economically intuitive way. Out‐of‐sample evaluations show that our approach is superior to models that exclude macro information and standard well‐fitting time‐series models. Our analysis shows that firm‐specific factors are useful in ranking firms’ relative riskiness, but that macroeconomic factors are necessary to understand fluctuations in the absolute risk level.     

Critical Infrastructure Vulnerability to Spatially Localized Failures with Applications to Chinese Railway System

This article studies a general type of initiating events in critical infrastructures, called spatially localized failures (SLFs), which are defined as the failure of a set of infrastructure components distributed in a spatially localized area due to damage sustained, while other components outside the area do not directly fail. These failures can be regarded as a special type of intentional attack, such as bomb or explosive assault, or a generalized modeling of the impact of localized natural hazards on large‐scale systems. This article introduces three SLFs models: node centered SLFs, district‐based SLFs, and circle‐shaped SLFs, and proposes a SLFs‐induced vulnerability analysis method from three aspects: identification of critical locations, comparisons of infrastructure vulnerability to random failures, topologically localized failures and SLFs, and quantification of infrastructure information value. The proposed SLFs‐induced vulnerability analysis method is finally applied to the Chinese railway system and can be also easily adapted to analyze other critical infrastructures for valuable protection suggestions.     

DEA as a tool for predicting corporate failure and success: A case of bankruptcy assessment

Using an additive super-efficiency data envelopment analysis (DEA) model, this paper develops a new assessment index based on two frontiers for predicting corporate failure and success. The proposed approach is applied to a random sample of 1001 firms, which is composed of 50 large US bankrupt firms randomly selected from Altman's bankruptcy database and 901 healthy matching firms. This sample represents the largest firms that went bankrupt over the period 1991–2004 and represents a full spectrum of industries. Our findings demonstrate that the DEA model is relatively weak in predicting corporate failures compared to healthy firm predictions, and the assessment index improves this weakness by giving the decision maker various options to achieve different precision levels of bankrupt, non-bankrupt, and total predictions.     

Optimization of Cascade‐Resilient Electrical Infrastructures and its Validation by Power Flow Modeling

Large‐scale outages on real‐world critical infrastructures, although infrequent, are increasingly disastrous to our society. In this article, we are primarily concerned with power transmission networks and we consider the problem of allocation of generation to distributors by rewiring links under the objectives of maximizing network resilience to cascading failure and minimizing investment costs. The combinatorial multiobjective optimization is carried out by a nondominated sorting binary differential evolution (NSBDE) algorithm. For each generators–distributors connection pattern considered in the NSBDE search, a computationally cheap, topological model of failure cascading in a complex network (named the Motter‐Lai [ML] model) is used to simulate and quantify network resilience to cascading failures initiated by targeted attacks. The results on the 400 kV French power transmission network case study show that the proposed method allows us to identify optimal patterns of generators–distributors connection that improve cascading resilience at an acceptable cost. To verify the realistic character of the results obtained by the NSBDE with the embedded ML topological model, a more realistic but also more computationally expensive model of cascading failures is adopted, based on optimal power flow (namely, the ORNL‐Pserc‐Alaska) model). The consistent results between the two models provide impetus for the use of topological, complex network theory models for analysis and optimization of large infrastructures against cascading failure with the advantages of simplicity, scalability, and low computational cost.     

Biotechnology collaborations: does business model matter?

This study investigates how a biotechnology firm??s collaboration incidence is affected by the business model it adopts. Specifically, we compare interfirm collaboration conducted by biopharmaceutical firms adopting the hybrid business model with those using the product-focused business model. The analysis based on 1,820 collaborations conducted by 87 dedicated biopharmaceutical firms suggests that firms adopting the hybrid business model generally engage in more collaboration. They also establish a greater proportion of exploration collaboration. These findings have implications for firm??s positioning using business models.     

Location models for ceding market share and shrinking services

New location models are presented here for exploring the reduction of facilities in a region. The first of these models considers firms ceding market share to competitors under situations of financial exigency. The goal of this model is to cede the least market share, i.e., retain as much of the customer base as possible while shedding costly outlets. The second model considers a firm essentially without competition that must shrink it services for economic reasons. This firm is assumed to close outlets so that the degradation of service is limited. An example is offered within a competitive environment to demonstrate the usefulness of this modeling approach.     

Reliability Analysis Based on the Losses from Failures

The conventional reliability analysis is based on the premise that increasing the reliability of a system will decrease the losses from failures. On the basis of counterexamples, it is demonstrated that this is valid only if all failures are associated with the same losses. In case of failures associated with different losses, a system with larger reliability is not necessarily characterized by smaller losses from failures. Consequently, a theoretical framework and models are proposed for a reliability analysis, linking reliability and the losses from failures. Equations related to the distributions of the potential losses from failure have been derived. It is argued that the classical risk equation only estimates the average value of the potential losses from failure and does not provide insight into the variability associated with the potential losses. Equations have also been derived for determining the potential and the expected losses from failures for nonrepairable and repairable systems with components arranged in series, with arbitrary life distributions. The equations are also valid for systems/components with multiple mutually exclusive failure modes. The expected losses given failure is a linear combination of the expected losses from failure associated with the separate failure modes scaled by the conditional probabilities with which the failure modes initiate failure. On this basis, an efficient method for simplifying complex reliability block diagrams has been developed. Branches of components arranged in series whose failures are mutually exclusive can be reduced to single components with equivalent hazard rate, downtime, and expected costs associated with intervention and repair. A model for estimating the expected losses from early-life failures has also been developed. For a specified time interval, the expected losses from early-life failures are a sum of the products of the expected number of failures in the specified time intervals covering the early-life failures region and the expected losses given failure characterizing the corresponding time intervals. For complex systems whose components are not logically arranged in series, discrete simulation algorithms and software have been created for determining the losses from failures in terms of expected lost production time, cost of intervention, and cost of replacement. Different system topologies are assessed to determine the effect of modifications of the system topology on the expected losses from failures. It is argued that the reliability allocation in a production system should be done to maximize the profit/value associated with the system. Consequently, a method for setting reliability requirements and reliability allocation maximizing the profit by minimizing the total cost has been developed. Reliability allocation that maximizes the profit in case of a system consisting of blocks arranged in series is achieved by determining for each block individually the reliabilities of the components in the block that minimize the sum of the capital, operation costs, and the expected losses from failures. A Monte Carlo simulation based net present value (NPV) cash-flow model has also been proposed, which has significant advantages to cash-flow models based on the expected value of the losses from failures per time interval. Unlike these models, the proposed model has the capability to reveal the variation of the NPV due to different number of failures occurring during a specified time interval (e.g., during one year). The model also permits tracking the impact of the distribution pattern of failure occurrences and the time dependence of the losses from failures.     

中国股票市场技术分析非线性预测能力的实证检验

运用前向人工神经网络方法对我国股票市场技术分析非线性预测能力进行了实证检验.发现基于移动平均规则的人工神经网络模型具有明显高于AR模型和各种移动平均规则线性模型的样本外预测能力.为解释技术分析方法具有非线性预测能力的原因,本文构建了一个基于异质市场假说的移动平均规则非线性模型,发现该模型的预测能力远高于其它非线性模型.表明我国股票市场存在异质性特征,技术分析方法能捕捉到不同类型投资者之间非线性的相互作用关系可能正是其具有非线性预测能力的原因.     

Genesis of a research field: district,network, strategic network

This article summarizes the changes since the beginning of the 1980s in the scholarly approach to organizational and economic research on Italian firms. Beginning with the study of industrial districts, which sparked a major reconsideration of the conventional wisdom, most scholars focused primarily on the significance of firm geographical proximity while marginalizing issues related to firm structure and strategy. Nevertheless, industrial district research eventually led to the question of firm networks, in particular how to manage relational capabilities and cooperation, both of which affect a firm’s competitive position. This new analytical framework no longer dependent on either the single firm or an industrial sector has opened up new research perspectives that promise rich insights into socio-economic studies.     

沪深300股指期货的波动率预测模型研究

以沪深300股指期货仿真交易的5分钟高频数据为例,运用滚动时间窗的样本外预测和具有Bootstrap特性的SPA检验法,全面对比了基于日收益数据的历史波动率(historical volatility)模型和基于高频数据的已实现波动率(realized volatility)模型对波动率的刻画和预测能力.主要实证结果显示,已实现波动率模型以及加入附加解释变量的扩展随机波动模型是预测精度较高的波动模型,而在学术界和实务界常用的GARCH及其扩展模型对沪深300股指期货的波动率预测能力最弱.     

Two-Group Classification Using Neural Networks*

Artificial neural networks are new methods for classification. We investigate two important issues in building neural network models; network architecture and size of training samples. Experiments were designed and carried out on two-group classification problems to find answers to these model building questions. The first experiment deals with selection of architecture and sample size for different classification problems. Results show that choice of architecture and choice of sample size depend on the objective: to maximize the classification rate of training samples, or to maximize the generalizability of neural networks. The second experiment compares neural network models with classical models such as linear discriminant analysis and quadratic discriminant analysis, and nonparametric methods such as k-nearest-neighbor and linear programming. Results show that neural networks are comparable to, if not better than, these other methods in terms of classification rates in the training samples but not in the test samples.     

Branch banking in the year 2000

The topic of branch banking in the United States in the year 2000 is obviously one that has been in the minds of California bankers for some time because California is the largest state in the United States to permit virtually unlimited statewide branch banking. It is a topic that must also be in the minds of a large number of European and Asian bankers as well because several non-U.S. banking systems have already established large branch networks in the United States. It is a difficult topic because no one knows precisely what will be the status of branch banking in the year 1990, let alone the year 2000. This article makes some predictions.     

Innovation and internationalization in an emerging market context: Moderating effects of interpersonal and organizational social networks

This study draws upon the social network theory to understand under what conditions innovation influences firm internationalization in the context of Vietnamese small and medium-sized enterprises (SMEs). We theorize that different types of social networks play varying roles in moderating the relationship between innovation and firm internationalization. Using a panel dataset of more than 15,800 observations of SMEs, we found that high levels of interorganizational social networks positively moderate the relationship between innovation and firm internationalization. However, there is no such relationship for interpersonal social networks for the studied firms. Our findings will allow SME managers to better understand the crucial role that interorganizational social networks can play in their successful internationalization.     

The Internet and the Deconstruction of the Integrated Banking Model

The rapid development of information and communication technologies (ICTs), with the Internet being one of the most significant, is shaking the foundation of the banking industry. Simply deploying the Internet as a more efficient distribution channel will not bring sustainable strategic advantages. To compete effectively, banks may need to embrace a new set of strategic priorities, based on the 'unbundling' of banking services and processes, and the 'deconstruction' of the integrated banking model. Such a radical transformation threatens the profitability and survival of some existing incumbent banks, and at the same time, it brings a cohort of new opportunities and powerful new players to the market. This paper reviews existing studies on the use of the Internet in banking, and highlights two prevailing models and the strategic thinking behind them. Based upon recent empirical evidence gathered from in-depth case studies and other information sources in the UK, the paper discusses a number of emerging tendencies and explores possibilities to reconcile the discrepancies between the two prevailing models. The paper calls for a radical departure from existing strategies in the banking industry based on the concept of process integration and service bundling. A new framework is outlined for banks and other financial organizations in developing and evaluating their strategies. The latest evidence suggests that at least eight models of Internet banking, somewhere on the spectrum between the two prevailing models highlighted by previous studies, have emerged in the UK. These models are illustrated briefly in the paper. This is still a rapidly evolving area, and new research is clearly needed to understand the key dynamics of the banking industry in the networked economy. Finally, some themes for further research are highlighted.     

Should financial reporting reflect firms’ business models? What accounting can learn from the economic theory of the firm

The paper draws on the economic theory of the firm as developed by Ronald Coase and Oliver Williamson. The theory helps us to understand why firms exist and why market prices are available for some items in accounts and not for others that are part of in-firm processes. The paper argues that financial reporting already reflects firms’ business models and makes the case for an approach to measurement in financial reporting based on firms’ business models. This approach distinguishes between assets that are transformed by a firm’s in-firm processes and those that are not. Historical cost measurements would usually be appropriate for the former, market price measurements (fair value) for the latter. The paper identifies a number of problems with the business model approach to measurement, but suggests that none of them should lead to the conclusion that such an approach would be mistaken. It also suggests opportunities for further research.     

Does a firm’s exposure to ethical failures matter to financial markets? A governance perspective

This paper investigates if a firm’s ethical reputation, in conjunction with its governance, affects its standing within financial markets. A firm`s ethical reputation, as measured by ethical failures, arises from its involvement in ethical violations and incidents while a comprehensive index proxies for governance. We assess a firm’s standing within financial markets through two complementary perspectives, i.e., the level of information asymmetry between managers and investors, as inferred from analyst forecast dispersion and analyst forecast error, and the relation between a firm’s earnings and its stock market valuation or return (value relevance). Our results suggest that a firm`s ethical reputation affects financial analysts’ forecasts as well as the stock market value assigned to its reported earnings. Moreover, it appears that corporate governance moderates such relations, with strong (weak) governance compensating for a weak (strong) ethical reputation. Overall, our evidence shows that ethical failures do not seem to pay.