Terry Harris

Credit scoring using the clustered support vector machine

This study introduces the use of the clustered support vector machine (CSVM) for credit scoring.The CSVM has been shown to relax size constraints while remaining highly accurate.The results suggest that the CSVM is a useful alternative to kernel SVM approaches when training datasets get large. This work investigates the practice of credit scoring and introduces the use of the clustered support vector machine (CSVM) for credit scorecard development. This recently designed algorithm addresses some of the limitations noted in the literature that is associated with traditional nonlinear support vector machine (SVM) based methods for classification. Specifically, it is well known that as historical credit scoring datasets get large, these nonlinear approaches while highly accurate become computationally expensive. Accordingly, this study compares the CSVM with other nonlinear SVM based techniques and shows that the CSVM can achieve comparable levels of classification performance while remaining relatively cheap computationally.

Quantitative credit risk assessment using support vector machines: Broad versus Narrow default definitions

This paper compares support vector machine (SVM) based credit-scoring models built using Broad (less than 90days past due) and Narrow (greater than 90days past due) default definitions. When contrasting these two types of models, it was shown that models built using a Broad definition of default can outperform models developed using a Narrow default definition. In addition, this paper sought to create accurate credit-scoring models for a Barbados based credit union. Here, the results of empirical testing reveal that credit risk evaluation at the Barbados based institution can be improved if quantitative credit risk models are used as opposed to the current judgmental approach.

Antecedents of taxpayers’ intentions to engage in tax evasion: evidence from Barbados

Purpose Tax evasion has been a major problem for governments around the world, with innovative and ever-changing schemes making the practice increasingly difficult to regulate. In light of this, this study aims to use the extended version of the theory of planned behaviour (TPB) (Beck and Ajzen, 1991) to predict individuals’ intentions to engage in tax evasion. Design/methodology/approach The research adopts a two-stage approach for data collection and analysis. First, the authors obtained survey data from 150 taxpayers in Barbados to conduct multivariate analyses to test the validity of the study’s hypotheses. The authors also used several open-ended questions on the survey instrument to conduct thematic analyses to further explore the influence of the antecedents of intentions to engage in tax evasion. Second, the authors conducted a focus group with two tax officials and three tax advisors. Findings The authors find that attitudes toward the behaviour, perceived behavioural control and moral obligation are significant predictors of intentions to engage in tax evasion. Factors cited as encouraging tax evasion are perceived fairness, tax authorities’ institutional infrastructure and responses, potential financial benefit, perceptions of inequality, low level of trust in tax authorities, perceived poor use of tax revenues and poor treatment of taxpayers. Conversely, factors cited as discouraging tax evasion include fear of prosecution, high morals and potential adequate governmental regulation. Research limitations/implications The study measures intentions to engage in tax evasion rather than actual behaviour. The study does not measure social desirability bias. Originality/value This paper tests the applicability of variables used in the extended version of the TPB to predict intentions to engage in tax evasion in a Caribbean-based emerging economy. It also applies a mixed-methods approach of collecting data from taxpayers, tax advisors and tax officials.