No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Forecasting credit ratings Using ANN and statistical techniques
This study is an empirical analysis of sovereign credit ratings (SCR) in South Africa (SA) using Logistic Regression (LR) to identify their determinants and forecast SCRs. Data of macroeconomic indicators including SCRs from 1999 to 2020 in quarterly format were classified and analyzed to identify indicators utilized by Credit Rating Agencies (CRAs) and then predict future ratings CRAs take various information from political, infrastructure, financial, economic, regional, local, and other factors pertaining to a country and assess the ability of that country to pay its debt. This information is then presented through a grading scale termed rating, with the highest rating country being highly creditworthy and lowest rating likely to default. There are three major CRAs, namely, Fitch, Moodys and Standard and Poors. The study identified the use of different macroeconomic indicators by CRAs as well as different techniques in assessing and assigning sovereign credit ratings. The study points out that Household Debt to Disposable Income Ratio (HDDIR) was the most influential variable on SCRs. HDDIR, exchange rates and the inflation rate were the most crucial variables for guessing credit ratings. Policymakers should aim to reduce household debt in relation to disposable income, implement policies that strengthen the local currency and stabilize as well as lower inflation. Investors should watch out on nations that have high household debt levels as this may spill over into credit risk.
Three models based on the principle of Artificial Neural Network (ANN), predicting the lattice constants of full-Heusler alloys from the ionic radii, symbols of the constituent elements, and the combination of ionic radii and symbols are reported. The model predicted values of the lattice constants are compared with the theoretically simulated values using the Density Functional Theory (DFT) based commercial software WIEN2k. A statistical linear regression model is also employed for comparison. The average absolute error in the regression model is equal to 7.33%, which reduces to about 4.67% in the ANN model using ionic radii, 3.00% for the model using symbols, and 3.67% for the combined model. These models are simple and efficient and can be easily extended to compute parameters other than the lattice constant of not only full-Heusler alloys but also half-Heusler, Perovskite, and other crystalline materials.
This paper reports stock return changes induced by credit rating revisions disclosed by credit-rating agency in one stock market, where there is a list of special stocks. Do stock prices of Shariah-compliant special firms react to credit rating disclosures? There is a rich literature on credit rating revisions affecting rated firm's stock prices, not the special stocks. The findings reported in this paper suggest stock prices react significantly to credit change disclosures. We also find the average change in the special stock price returns significantly correlated with four key firm-specific variables: two working capital ratios, leverage and profit margin. These findings offer evidence that the special common stocks traded in several key world stock markets could have similar price reactions to credit rating change disclosures in line with the effects known for ordinary common stocks studied to-date. The findings reported here may prompt further studies on other special stocks in other markets.
Although Asia is at the forefront of global economic growth, its investment environment is very risky and uncertain. Credit ratings are objective opinions about credit worthiness, investment risk, and default probabilities of issues or issuers. To classify credit ratings, analyze their determinants, and provide meaningful decision rules for interested parties, this work proposes an integrated procedure. First, this work adopts an integrated feature-selection approach to select key attributes, and then adopts an objective cumulative probability distribution approach (CPDA) to partition selected condition attributes by applying rough sets local-discretization cuts. This work then applies the rough sets LEM2 algorithm to generate a comprehensible set of decision rules. Finally, this work utilizes a rule filter to eliminate rules with poor support and thereby improve rule quality. The experimental focus was the Asian banking industry. Data were retrieved from a BankScope database that covers 1327 Asian banks. Experimental results demonstrate that the proposed procedure is an effective method of removing irrelevant attributes and achieving increased accuracy, providing a knowledge-based system for classification of rules for solving credit-rating problems encountered by banks, thereby benefiting interested parties.
Outlines previous research on company failure prediction and discusses some of the methodological issues involved. Extends an earlier study (Tan 1997) using artificial neural networks (ANN) to predict financial distress in Australian credit unions by extending the forecast period of the models, presents the results and compares them with probit model results. Finds the ANN models generally at least as good as the probit, although both types improved their accuracy rates (for Type I and Type II errors) when early warning signals were included. Believes ANN “is a promising technique” although more research is required, and suggests some avenues for this.