# Credit Risk Rating and Financial Ratios in Pakistan

### Chapter - I: Introduction

The credit rating is a formal evaluation of a firm, governments, individuals, or even a country. Credit rating is assessed on the basis of previously carried out financial transactions and recent assets and liabilities. Credit rating eases the life of a lender to assess the capability of the borrower to repay a loan. Whenever a borrower expects to use future cash flows to pay a current debt, credit risk is there. the borrower or issuer of a debt obligation uses interest payments as the way to compensate the investors for taking credit risk. The higher perceived credit risk for lending the capital will demand higher rate of interest from the investors

.

In order to assess the credit risk of any person or entity various credit risk models are developed by taking numerous considerations into account depending on the nature of target for which credit risk is meant to be judged. Various risks assessed during the process could also include the general business risk through assessing the economic enviornment, legislative and regulatory enviornment, competitive enviornment, country risk, management and corporate governance. Similarly industry specific factors that can be assessed may also include primary factors such as market position, production efficiency and financial volatility or the seconday factors such as barriers to entry, maturity of markets etc. An important factor in assessing the risk is the analysis of financial side of the subject to be evaluated. For example, if some organization or company is being evaluated, one would analyze its current and prior financial position, reputation and operating performance in the market.

Conducting such a detailed analysis may not be possible for every single investor specifically in the present time where the globalization has not only opened the wide range of opportunities for potential investors to invest in but, at the same time, not all the opportunities of investment are safe and secure where a good return could be earned. In fact, very less opportunities exist in the market where a safe investment could be made ensuring some reliable and feasible returns to the investors; such investment opportunities need to be identified by one but evaluation of every opportunity is not in the reach of each individual. In order to facilitate the investors, a trend of credit rating has become quite popular at the present time where various credit rating agencies evaluate and assess number of organizations suggesting an outlook of those organizations as per their judgement. This definitely does not ensures the good rated organizations by such agencies to be safe and investable but some how ease up the confidence of an investor to make appropriate investment decision.

Moody's and Standard and Poor's (S&P) are the famous and intenrationally known rating agencies to name the well known and important rating agencies of the world. Pakistan Credit Risk Agency Limited (PACRA) is a reputable and important credit rating agency operating in Pakistan and this is the one chosen in present study.

As the financial aspect of any organization plays an important role in the evlauation of risk, financial ratios act as important factors while analyzing the financial position of any company. Financial ratio analysis is the computation and evaluation of ratios derived from a company's financial statements. The conclusion about the financial condition, operations and investment attractiveness of a company is possible using the historical trends of these ratios. It will not be wrong to state ratio analysis as a diagnostic tool that helps to recognize problem areas and opportunities within a company. These ratios do not only expose the real financial position of any firm but also enable one to easily compare the position of that firm with any other in the industry. Ratios are helpful to compare various aspects of a company's performance with regard to its history or even within the industry or in a region. They disclose very essential information such as whether too much debt is accumulated, heavy inventory is stockpiling or collection of receivables is not fast enough.

(Beaver, 1966) and (Altman, 1968) showed financial ratios as significant predictors of bankruptcy. Deakin (1972) and Ohlson (1980) also worked on financial ratios to determine their predictive power for bankruptcy. Adams, Burton, & Hardwick ( 2003) and Grey, Mirkovic, & Ragunathan (2006) worked to predict the credit rating of insurers using the financial ratios. Thomson (1991) used financial ratios in predicting bank failures whereas Sinkey (1975) conducted a study to disriminate between problem and non- problem banks using the financial charachterstics of the banks.

(Sinkey, 1975); (Meyer & Pifer, 1970) and (Kumar & Arora, 1995) used liquidity ratios in their study to distinguish between problem and non- problem banks and the same ratios is incorporated as independent variables in this study to test their predictive accuracy in relation to the credit risk rating. Other ratios used as independent variables in the study are asset quality used by Kumar & Arora (1995) and Sarkar & Sriram (2001) and capital adequacy found significant by Kumar & Arora (1995) in the respective studies. In order to cover profitability, the net income and admin expenses growth ratios are also used as independent variables to observe their effect on credit risk rating.

Due to the significant importance in practice, liquidity, capital adequacy, asset quality and profitability ratios are analyzed in this study in order to predict the credit risk rating of banking sector, a key segment playing an important role in the economy of any country.

The initiating section of the study introduces to the importance of credit risk rating and financial ratios. In the second section, various evidences from previous studies are discussed. The third section covers the data collection procedure and methodology while the fourth section describes the variables used in the study. Fifth section explains the results and findings during the study. The last section presents the conclusion and limitations along with future implications of the study.

### Chapter - II: Literature Review

The recent financial crisis of 21st century has increased the sensitivity level of investors while applying his investments in any company. At the same time the declaration of large multinational organization for bankruptcy during the present crisis has drown a lot of investment from private sector. In such a situation an investor is more conscious while making his/her investment in a particular company and need the assurance for the risk level to be low for the company he/she is investing in, but every investor cannot check or analyze the level of risk for all the potential companies where such investments could be made. Therefore, the trend of analyzing Credit Risk Rating of firm is found to be an easier way for investor to make a decision.

The studies on financial ratios and credit rating have been the main focus of many researches. Many researchers have found several ratios to be quite important in determining and predicting the credit risk rating of any firm among which Adams, Burton and Hardwick (2003) and Gray, Mirkovic & Ragunathan (2006) are popular ones to be named. The literature of financial ratios significance is centered study of Beaver (1966) and Altman (1968).

(Beaver, 1966) gave a new introduction to financial ratios through defining their predictive ability for a firm failure. The data was obtained on which financial statement was available for the first year before failure. In order to keep the data balanced, for each failed firm, a non failed firm from the same industry and asset size was selected. Various financial ratios were used as variables in predicting the failure of the firm, to highlight; these ratios included cash flow ratios, profitability ratios, liquidity ratios and turnover ratios. A profile analysis was also done by comparing the mean values of ratios for each of the failed and non- failed firms in each of the years before failure.

(Beaver, 1966) applied dichotomous classification test in order to predict the failure status of the firm based upon the financial ratios. The findings suggested that financial ratios can be a source of predictability at least five years before failure. It also concluded that all the ratios do not have the same capability to predict the failure status, whereas, these ratios don't also correctly predict the failed and non failed firms with the same degree of success. The test applied in this paper also included some limitations which included the determination of probability of error as well. In the end Beaver (1966) also warned about the factor that prevented a measurement of the true predictive ability of ratios which also included the possibility of existence of some ill firms whose illness was detected before the failure and proper treatment was applied and the firms did not fail. Furthermore, a wide gap was also noted between the ratios of failed and non failed firms.

(Altman, 1968) worked on the significance of financial ratios in predicting the bankruptcy using the multiple discriminant analysis technique beside the univariate analysis that was used in prior studies. Various ratios were used as variables for the evaluation and from this list five were selected as the best predictors of corporate bankruptcy. The only limitation of the study was that sample size was selected from the manufacturing industry from where the financial data of the companies was obtainable. Results of the study were found to be quite encouraging for Altman (1968) as model proved to be extremely accurate in predicting bankruptcy correctly in 94 per cent of the initial sample with 95 per cent of all firms in the bankrupt and non-bankrupt groups assigned to their actual group classification. It was suggested that this model is an accurate forecaster of failure up to two years prior to bankruptcy and that the accuracy starts diminishing as the lead time keeps on increasing and up to five years prior to bankruptcy the model becomes insignificant and inaccurate.

(Deakin, 1972) anlyzed the predictors of business failure with an aim to propose an alternative model for predicting failure. Initially the accounting ratios used by Beaver (1968) along with the methodology was replicated in the study and later on a decision rule was deviced to validate the cross-sectional sample of firms. The methodology employed in the study was the dichotomous classification test. The discriminant analysis was also used in the study in order to find the linear combination of ratios which best dicriminates between the groups classified. The cash flow, income, current ratio, quick ratio, and working capital are the famous ratios used in the study. The results suggested that discriminant analysis is sufficiently strong to be used for distributions of financial data.

(Ohlson, 1980) following the trend of predicting bankruptcy tried to come up with further progress in the study by employing a different methodology while considering the timing issue as an important factor in predicting bankruptcy. Besides choosing the popular Multiple Discriminant Analysis (MDA) technique of bankruptcy studies, the econometric methodology of logit analysis was opted by Ohlson (1980), in order to avoid the problems noted in MDA. The popular independent variables used in the study included size, current ratios, working capital ratios and income ratios. It was found that predictive power of any model depends upon when the information is assumed to be available which was not taken care of in some of previous studies and the estimation procedures or method used is also quite sensitive to the predicting power of financial ratios.

(Libby, 1975) designed the study to jointly evaluate the pridictive power of financial ratios and the ability of loan officers to evaluate that information in predicting the business failure. The ratios covered in the study were taken from the studies of Beaver (1966) and Deakin (1972) which reflected five independent sources labeled as profitability, activity, liquidity, asset balance, and cash position. Principal components analysis followed by varimax rotation was applied in order to reduce the number of ratios, remove redundancy and identify salient financial dimensions represented by the ratios. Binomial test was also employed to predict the individual's perofrmance. The prediction achivement was noticed to be seventy four percent. None of the other variables were found to be significantly correlated with prediction accuracy.

(Adams, Burton, & Hardwick, 2003) worked on the financial services sector in their study and investigated the insurance firm's likelihoods of being externally rated by A.M. Best or S&P and the determinants of the letter ratings. The independent variables considered in the study included leverage, , liquidity, growth, profitability, company size, organizational form, and business activity. To find out the rating likelihood a multinomial logit model was used. A sample of UK based insurers was taken in which the insurers in first group were rated either by S&P or A.M. Best between the period of 1993 and 1997 and another group of insurers were not rated by any credit rating agency over the preiod of five years. Various findings concluded that the likelihood of being rated is positively related to an insurer's profitability and negatively related to leverage, though some differences in the determinants of the likelihood of being rated by A.M. Best or by S&P existed there. The results also evident that higher A.M. Best ratings and S&P ratings could be reached through higher levels of profitability and liquidity. Furthermore, a negative relation between leverage and S&P ratings was found showing that lower levels financial leverage would more likely to be assigned a higher S&P rating.

(Grey, Mirkovic, & Ragunathan, 2006) worked on examination of association between Australian credit ratings and industry variables and a set of financial ratios. The sample consisted of firms rated by Standard and Poor's (S&P) was taken into account. Various financial ratios were taken into consideration as variables which included interest coverage, leverage, profitability and cash flow ratios. Other variables considered in the study reflected the industry concentration. The Ordered Probit Model approach was adopted in this paper. The ratings were denominated by numerical values being dependent variable in order to make the analysis more efficient. While evaluating the determinants of Australian credit ratings it was observed as a result that the most significant effect on the credit ratings was reflected by interest coverage and leverage ratios, whereas, profitability variables and industry concentration were also found to have an important impact on the credit ratings.

(Tanthanongsakkun & Treepongkaruna, 2008) compared the accounting-based model ( interest coverage and debt leverage ratios) with the market-based model (firm size and book to market (BM) ratio) and examined the likelihood of both in explaining the credit risk rating. The sample consisted of Australian listed companies rated by Standard and Poor's (S&P) during 1992 to 2003. While considering the accounting based model two variables i.e. Interest Coverage and debt leverage ratio were considered which were found to have most significant effect on credit risk rating during the study conducted by Gray, Mirkovic & Ragunathan (2006). The variables considered in market- based model were Default Liklihood Indication (DLI) computed from Metron Model, firm size and book to market value ratio. The methodology employed by Tanthanongsakkun & Treepongkaruna(2008) is ordered probit model. The results reflected the likely relationships between the credit ratings and all explanatory variables whereas the coefficients of market based model's variables were found to be more significant than that of accounting based model.

Unlike previous studies of insolvency of insurance companies operated in United States and developed economies Chen & Wong (2004) stepped ahead to work on the Asian insurance companies where the firm data and market factors were used seperately in analyzing seperately the financial strength of life and general insurers in Japan as well as Malaysia, Singapore, and Taiwan. The firm specific variables that were common in general and life insurers' insolvency included firm size, investment performance, and operating margin. Liquidity ratio, combined ratio, premium growth, and growth rate of surplus were other firm specific variables considered for general insurers whereas change in asset mix, change in product mix, and insurance leverage variables were taken in addition for life insurers' insolvency in firm specific variables.

(Chen & Wong, 2004) used the number of insurers, interest rate changes, absolute level of interest rate and inflation rate change as variables while considering the market/ economic factors on both General and life insurers. The various methodoligies included in analysis included The HMM Model which uses ratio-based methodology to provide an early warning of insurers in possible financial trouble. Another methodology used was an implication of logit model to estimate the equations onfirm specific factors whereas multiple regression was employed to estimate equation on market/economic factors for both life and general insurers' solvency. The logistic regression reflected the significant result consistent with majority of the hypotheses formulated on firm-specific factors for determinig an insurer's financial strength. None of the hypotheses were supported about the effects of market/ economic factors on the general insurers' financial health. The limitation identified in the study was the only use of financial data at the firm level and none of the qualitive information were taken into account which can be an important variable in assessing the insurers' financial health in a better way.

(Pinches & Mingo, 1973) worked to develop and test a model for predicting industrial bond ratings. Number of financial figures and ratios were used as variables which included total assets, working capital ratios, sales worth, earning ratios, debt and debt coverage ratios and means for percentage changes in sales, income, debt etc., which were then grouped into a total of six variables using factor analysis and a multiple discriminant analysis was conducted to predict industrial bond ratings. The model was enable to correctly predict 69.70 per cent of the actual ratings in the original sample, and predicted approximately 60 per cent of the ratings for a holdout sample and another sample of newly rated bonds. The variables relating to earning stability, financial leverage, size, debt and debt coverage stability, and return on investment simulated the best results of Moody's ratings.

(Carty, 2000) used companies' public credit ratings as proxy for default risk and identified whether there is any tendency for a company that maintains the same values for its accounting measures and equity risk measures over time to receive a lower rating today than in prior years. The methodology used was Ordered Probit Model which relates the rating categories observed explanatory variables through an unobserved continuous linking variable. The variables incorporated in the research includes interest coverage, opertaing margin, long term debt coverage, total debt coverage, and market value. The results explained that if it were not for the use of more stringent rating standards, the level of bond ratings might had actually been higher at the time of research as compare to past. Another explainantion concluded that the meanings of the firm variables used in the study have changed over time. The main conclusion of the study reflected that rating standard had become more stringent in terms of explanatory variables used in the study. Carty (2000) also found that accounting ratios and market-based risk measures were more informative for larger companies than smaller companies.

(Papoulias & Theodossiou, 1992) analysed the recent business failure of the time and presented models for detecting financialy distressed firms. Seven financial variables were examined as potential predictors of business failures which included the current ratio, working capital to total assets ratio, liquidity ratio, profitability ratios, and debt ratio. The statistical techiniques of logit, probit, regression analysis and Bayesian discriminant analysis were employed in order to test the purpose of study. An important finding in the study illustrated that regression analysis and the linear discriminant analysis produced identical results because the regression coefficient, excluding intercepts, are proportional to the discriminant coefficients by the same fixed constant and both the models classified firms identically. The coefficients of current ratio, working capital/ total assets, and income ratios had negative signs in all four models which was found consistent with the hypothesis that probability of failure of the firm will be negatively effected with an improvement in the liquidity position or improvement in the profitability of the firm. The coefficient of debt ratio had a positive sign justifying the hypothesis firm gets more vulnerable to failure with more leverage. In logit, the classification accuracy rate excedded ninety one percent in the samples. In short, all the models were found to be quite accurate in classifying initial and the holdout samples'firms correctly.

(Beaver, McNichols, & Rhie, 2005) examined the changes in the capability of financial statement data to foresee bankruptcy. The ability was predicted through identifying three trends in financial reporting: 'FASB standards, the precived increase in discretionary financial reporting behavior, and the increase in unrecognized assets and obligations'. The salient feature of the results was the potency of the projecting models over a forty-year period. The study predicted whether the predictive capability of financial ratios for bankruptcy has changed from the first to second sample period. The three variables capturing three key elements of financial strenght of a firm included in the study were return on assets reflecting profitability of the assets, Earning Before Interest, Tax, Deprciation and Amortization (EBITDA) divided by total liabilities reflecting cash flow avaibles to pay off liabilities and total liabilities divided by total assets (LTA) reflecting leverage. The market-based variables were also tested which reflected market capitalization, lagged cummulative security residual return and standard deviation of security returns. The statistical estimation model adopted in the study is Hazard Analysis. The results reflected that showing slight changes, robustness of the predictive models is strong over time and improvement in the incremental predictive ability of market-related variables offset the slight decline in the predictive ability of the financial ratios.

(Horrigan, 1966) attempted to test the efficacy of financial ratios in long-term credit administration decisions. The study included the financial ratios analysis of manufacturing firms with the bond ratings assigned by Moody's and/ or S&P. The ratings were taken as dependent variable in the study whereas the independent variables included various financial ratios which reflected the liquidity, solvency, capital turnover, profit margin, and return on investment of the firms. The study revolved around the methodology of multiple regressions analyzing the bond ratings on the range of combinations of independent variables. The regression coefficinent of the final model were later used in predicting the new sets of rating data. The results showes that around fifty eight percent of the Moody's new ratings and fifty two percent of the S&P's ratings were predicted correctly whereas approximately fifty four percent of Moody's and fifty seven percent of S&P's changed ratings were predicted correctly using the given independent variables. The limitation of the study is reflected in used and incorporation of the dummy variables part of the work.

(Pottier, 1998) checked the effectiveness of Best's rating and rating changes in comparision to financial ratios while predicting the life insurer insolvency. A number of financial ratios were incorporated in the study including liquidity, levergae and profitabiliuty ratios. Ratings and Rating changes were also incorporated as independent varaibles in the study. Three differenct models were developed using the given independent variables. The first model was purely based on financial ratios, the second one reflected ratings and rating changes whereas the last one was made by merging financial ratios with ratings and rating changes denoted as FINRATING. The methodology adopted in the study was Logistic Regression. The results reflected that combination of ratings and rating changes with financial ratios improves the predictive ability compared to financial ratios only for mostof the cost ratios. Another important finding suggests that rating changes should be incorporated in insolvency prediction models as these are important predictors of insurer collapse even when pooled with financial ratios.

(Yeh, 1996) evaluated the banks'performance through the appplication of Data Envelopment Analysis (DEA) in conjunction with financial ratios in order to distinguish the efficient banks from the inefficient ones as well as to understand the various financial dimensions that some how links to the bank's financial operational decisions. A wide variety of financial ratios were incorporated in the study to examince various characterstics of a bank's performance which are included in the categories of capital adequacy, asset utilization, profitability, financial leverage and liquidity ratios. The popular mathematical programming methodology used in the study is Data Envelopment Analysis and in line with its requirements the bank output used are interest income, non-interest income and total loans whereas, bank inputs include interest expense, non-interest expense and total deposits. The factor analysis was also conducted on the twelve financial ratios taken as variables in the study which produced four strong and consistent factors. The reflection of results concluded that banks which were more DEA efficient were less leveraged and more aggressive in employing their deposits and assets to generate revenues than those who were less DEA efficient while using financial intermediation as the input-output criteria for bank performance evaluation. It could be concluded from the results that the banks in the groups with higher DEA scores also have various higher ratios in capital adequacy, asset utilization and profitability, and lower ratios in financial leverage and liquidity, than those with lower DEA scores, except that the profit margin which didn't show enough acceptable results.

(Mossman, Bell, Swartz, & Turtle, 1998) compared four prediction models and prepared significant contribution to model effectiveness. Four precition models were tested which included Z-score model in the basis of financial ratios, a cash flow based model of Aziz, Emanuel, & Lawson (1988), a model on market return by Clark & Weinstein (1983) and a market return variation model by Aharony, Jones, & Swary (1980). Mossman, Bell, Swartz, & Turtle (1998) found that the ratio model was most efficient in predicting probability of bankruptcy in the year earlier to bankruptcy. But the cash flow model frequently discriminated bankrupt and non-bankrupt firms over the period of three years prior to bankruptcy. These results imply different uses of the models, where cash flow variables could also be included by the stakeholders who want to add these variables when early warning is reuqired. Alternatively, a huge depressing shift in accounting ratio variables might be a helpful indicator of forthcoming financial collapse in upcoming year of the event.

(Sensarma & Jayadev, 2009) attempted to utilize the information contained in financial statements on the risk management competencies of banks and then determined the sensitivity of banks' stock to risk administration. A single variable representing four risk variables i.e. interest rate risk, natural hedging, credit risk, and capital adequacy was formed using the factor analysis technique in order to come up with a summary measure variable. This was done in various ways in order to conduct the robustness checks of the result. A simple mean (average), discriminant analysis and principle component analysis are the techniques used to come up with the required summary measure variable. Finally, the regression technique was applied in order to meausre the banks' stock sensitivity in relation to risk management.While using all four risk variables as discussed before, the results indicated that except for CAR, all other variables were statistically insignificant. Using the average score method it was found that the market returns coefficient was important which indicated that systematic risk is important in determining stock returns and the coefficient of average was also positive and significant. Results of principle components indicated that banks with good overall risk management score were enhancing shareholder value. In discriminant analysis technique the positive and significant attitude was found which indicated that investors are attracted to banks which signal superior risk management capabilities through their balance sheets.

(Metwally, 1997) worked on differentiating between the interest-free and conventional banks through analyzing the financial characterstics of banks in both the categories. Three methodologies were implemented in the study which included logit, probit and discirminant analysis. The independent variables were financial ratios reflecting the liquidity, leverage, credit risk, profitability and efficiency. The findings suggested that two groups of banks might not be differentiated using profitability and efficiency but could be differentiated in temr of other financial characteristics reflected through liquidity, leverage and credit risk.

(Kumar & Arora, 1995) developed a risk rating scheme for banks, based on more readily avialable performance data of banks i.e. financial statements. They referred this to as 'R risk rating'. The performance variables used in the study were set into various categories which include asset quality, earnings, liquidity, capital adequacy and management. The linear logit model and quadratic model were used as methodology in order to predict the failed and non failed firms. The results show that while testing classification perofrmance for the learning sample with linear logit model 96 percent of the failed banks were correctly classified whereas 70 percent of the non failed or safe banks were correctly classified, whereas, in testing the classification performance for the learning sample with quadratic model 95 percent of failed banks and 75 percent of the non failed firms were correctly classified.

(Thomson, 1991) conducted a study to model bank failures of all sizes and predicting the bank failures. The variables incorporated in the study includes solvency, liquidity, assetquality, and earnings ratios. The statistical methodology used in the study was logit regression which lead to the results showing that solvency and liquidity are the most important predictors of failure upto thirty months before failure, however, with the increase in time to failure asset quality , earnings and management quality starts gaining importance as predictors of failure.

(Sinkey, 1975) carried out a study to perform a multivariate statistical analysis of the characterstics of problem banks. The Multivariate Discriminant Analysis (MDA) technique was used to distinguish between the problem and non problem banks through there financial characterstics over the period of four years. These financial characterstics covered various financial ratios representing liquidity, operating efficiency, asset and deposit composition, profitability, capital adequacy and revenue . As expected the results indicated that the average problem bank was significantly less efficient with a relatively inadequate and deteriorating capital position and greater financialy burdened due to expenses than the average non problem (control) bank in each of the four year. The discriminant analysis tests reflected that both the mean group profile and group dispersion matrices were significantly different in all four years.

(Sarkar & Sriram, 2001) used probabilistic models to provide early warning of bank failures. Various financial ratios were taken as variables in the study which reflected a bank's perofrmance in four important areas i.e. asset quality, overhead risk, earnings risk, and capital adequacy. The naive Bayes classifica-tion model and a composite attributes (CA) model were the two probabilistic tests employed in the study. Both the models were found to be very sharp when recent predictive ratios were available. The composite attributes model correctly identified financially distressed firms 88% of the time, and healthy firms 93% of the time whereas the naive Bayes model was able to identify financially dis-tressed firms 80% of the time, and healthy firms 91% of the time which concluded that both the models had good discriminatory ability for both classes of bank but the composite attributes model was found somehow more powerful and accurate in identifying financial distressed banks.

(Meyer & Pifer, 1970) attempted to discriminate between bankrupt and solvent banks under the similar local and national market conditions. A large variety of financial ratios and balance sheet items were taken as independent variables in the study which covered most importantly the liquidity ratios, profitability, loans growth, and loan quality. The methodoloty adopted for the analysis was Regression where the dependent variable was dichotomous depending upon the independent variables showing the bank either as failed or solvent one. Under various conditions and cutoff points the prediction accuracy was quite encouraging where with a lead time of one or two years, approximately eighty percent of the observations were correctly classified and the R2 was about 0.70, quite high for a cross-section study. With a lead time of three year and more, the financial variables were unable to discriminate between the failing and viable banks.

(Reynolds, Fowles, Gander, Kunaporntham, & Ratanakomut, 2002) examined the financial capital structure of major financial companies over a fiive year period. The main focus was in estimating the probability of a financial company surviving to 1997 (the start of the crisis) and the 1993-96 economic determinants of that probability. The study dealt with the probability of survival of a company against the effect of various independent variables which included various balance sheet and income statements items as well as financial ratios covering capital structure ratios as a basic focus. The methodology used in the article included both probit and logistic binomial regression analyses. The commulative logistic model was also incorporated in the analysis. All three models were found to be good predictors of firm failure and survival, where the overall classification accuracy avergaed at sixty eight percent.

(Estrella, Park, & Peristiani, 2000) found the signifiacance of financial ratios in relating it with the bank failure. The ratios incorporated in the study were three types of capital ratios i.e. risk weighted, levergae and gross revenue ratios. A simple frequency distribution analysis was conducted which resulted in acceptable performance of all three capital adequcy ratios in identifying failure. Later on logit regression model was employed in the study. When all three variables were entered in the model seperately, they proved to be significant in predicting the failure fairly whiel entering all three variables together in the model, the gross revenue ratio appeared to had the highest significance overall. The findings suggested that bank regulators may find a valuable role of the simple ratios in order to design regulatory capital frameworks, particularly as indicators of the need for timely supervisory act. Riskweighted ratios, in contrast, were tend to perform better over longer horizons.

(Wheelock & Wilson, 2000) estimated a comprehensive model relating to the characteristics of bank failure probability with special emphasis on management quality. Management quality was considered as one of the alternative measurements of productive efficiency done in a framework that 'permits a bank to disappear either by failing or being acquired by another bank'. Management quality is difficult to measure directly as it can it can take several forms. Productive efficiency measurement is one of ways through which management quality is measureable. Many research papers deal with productive efficiency banking measurements. In addition, Wheelock and Wilson (2000) have examined the risks of banks disappearing due to acquisition or failure. A whole series of hazard models were used in order to estimate bank failure risk. These hazard models took account for capital adequacy, management quality, asset quality, earnings, liquidity and miscellaneous factors.

### Chapter - III: Research Method

In this thesis researcher focused on the banking sector of Pakistan. The total population of the industry operating in the country is 39 of which thirty three are pakistani banks and the remaining six are foreign banks. The actual target was to cover the complete population in the industry but relying on the availability of the financial data, the initial sample consisted of 30 banks. As all of the banks were not rated by PACRA as per the requirement of study thus in order to keep the number of observations viable, the analysis was needed to be done over the number of years, therefore, the analysis was conducted by collecting the sample of the banks which were being regularly rated by PACRA. This resulted to extend the analysis over the four year period in order to keep the sample size feasible, which was available for the 11 banks of the industry.

The finanical data for all of these banks was collected for last four years i.e. from 2005 till the most recent one 2008 through the annual reports of these banks. The credit risk rating was taken from the Pakistan Credit Risk Rating Agency (PACRA) website. In Pakistan, the financial year of banks ends on December 31st of the respected year and normally the ratings for the banks are announced after six months of the year end. Therefore, the financials of a particular year of a bank is compared with the next following rating announced by PACRA for that bank.

The technique employed in the study is Mulltiple Discriminant Analysis (MDA) due to its usefulness in understanding the power of financial ratios to discriminate the distinguishing groups which is ratings in the case of present study. Altman (1968), Deakin (1972) and Ohlson (1980) used dicriminant analysis to predict bankruptcy in their respective studies. Pinches & Mingo(1973) employed discriminant analysis in predicting industrial bond ratings whereas Sinkey (1975) made the effective use of MDA is distinguishing between the problem and non-problem banks.

### Chapter - IV: Variables

The data for variables is extracted from the publicly available information of banks in Pakistan which includes financial statemnts of the banks which is available in printed format on annual basis in the market as well as the website of State Bank of Pakistan. Now the dependent and independent variables taken in the study are being discussed in this thesis.

### Dependent Variable:

The Credit Risk Rating of the banks assigned by Pakistan Credit Rating Agency (PACRA) was obtained from the PACRA web site which is taken as dependent variable in the study. PACRA assignes various kinds of rating to the firms which includes entity ratings, instruments ratings, mutual fund rankings and funds stability ratings. The wholesome interest of this study is in entity ratings assigned by PACRA. This rating is further divided into long term and short term ratings by PACRA. The ratings considered in the current study are long term ratings keeping in mind the long term stablization of a firm. The standard rating scale as defined by PACRA is relfected in Table - A :

Table - A: Ratings and Description | |

Long term Ratings: |
Description: |

AAA |
Highest Credit Quality |

AA |
Very High Credit Quality |

A |
High Credit Quality |

BBB |
Good Credit Quality |

BB |
Speculative - Possibility of credit risk developing |

B |
Highly Speculative |

CCC, CC, C |
High default risk |

Source: (The Pakistan Credit Rating Agency Limited)

As the above given ratings are represented using string variable, a need to convert this into numeric scale existsin order to run the required statistical tool and generate reliable results. Therefore, the procedure attempted by Adams, Burton, & Hardwick (2003) would be followed in this study to code the ratings into four different numeric values as shown in Table - B:

Table - B: PACRA ratings coded into numeric values | |

PACRA Ratings: |
Coded Rating: |

AAA/ AA |
1 |

A/BBB |
2 |

BB/B |
3 |

CCC, CC, C |
4 |

(Adams, Burton, & Hardwick, 2003) assigned the numeric codes in a manner that better credit ratings had a higher value and the lower ratings had a lower value. As the purpose here is not any calculation or computation but to check the preciting power of independent variables so the order of assigning numeric codes would not effect any aspect of analysis in the study. As the seven categories show a wide range of ratings, the ratings with similar descriptions were combined together in order to reduce the number of categories. Moreover, none of the banks in sample are assigned the rating less than BBB in last four years, therefore, assigned a different code to every rating category was not needed. Therefore, ratings reflecting very high credit quality were assigned with numeric code '1' and ratings with good and better credit quality were given the numeric code '2'. Similarly, the ratings describing the speculative outlook of firms were denoted by numeric code '3', whereas numeric code '4' was assigned to the rating showing high default risk. As mentioned earlier, the numeric codes of 3 and 4 were yet not seen against any obeservation, which shows that the samples collected from the industry were having a better outlook as per PACRA's ratings.

### Independent Variables:

The independent variables are the financial ratios calculated using the information from financial statements of the bank. The following information has been derived from the financial statement in order to be used in calculating financial ratios:

### Liquid Assets:

Liquid assets include the cash, balances with other banks, lendings to other fiancial institutions and governent securities. The basic idea of liquid assets is to reflect the fastest available cash or cashable assets of the banks. Cash itself represents the amount available with bank in local currency in hands. Balances with other banks is the amount of money of the bank in other bank account(s). The head of lendings to other financial institutions shows the amount a bank has lended to other banks or financial institutions for various purposes. Government securities are part of investments which a bank makes into various securities or certificates or its own subsidiaries or stocks etc in order to keep on earning for short or long-term on the money available with it in excess of needed in a normal situation.

### Gross Advances:

Gross Advances include the loans, cash credits, running finances, investment in finance lease, bills discounted and purchased and other credit facilities bank provides to its customers against some interest. Excluding the provisions (allowed) by state bank would lead to the number of Net Advances.

### Total Assets:

Total Assets of the bank are calculated by adding up cash, balances with other banks, lendings to other financial institutions, investments, advances, operating fixed assets, deffered tax assets and other assets in the balance sheet.

### Deposits:

Deposits are basically the liability part of a balance sheet which shows the sum of amount a bank has in its holding from various customers either consumers or corporates or could also be another financial institution up on which the depositors are paid various interest amount (if any applicable) for keeping their money with the bank.

### Total Equity:

Total Equityof the bank is represented by adding up the share capital (i.e. issued, subscribes and paid up capital), reserves, unappropriated profits (if any) and amount (surplus or deficit) on revaluation of securities (if any).

### Non-performing Loan:

The loans or advances given by the banks against which no interest and principle is being recovered are considered under non- performing status and knowns as Non-performing loans or NPLs.

Financial Ratios incorporated:

Liquidity ratios:

(Beaver, 1966), (Altman, 1968), (Libby, 1975) used the liquidity ratios in order to predict the bankruptcy and found significance impact of liquidity ratios to the subject. Adams, Burton And Hardwick (2003) found that higher level of liquidity leads to higher A.M. Best and S&P ratings. Yeh (1996) found that the banks with higher Data Envelopment Analysis (DEA) scores have lowers liqidity ratios than those with lower DEA scores. Kumar & Arora (1995) were able to correctly classify the major percentage of sample into failed or safe banks through incorporating liquidity as one of the variables in the study. Sinkey (1975), Meyer & Pifer (1970) used liquidty ratio in order to distinguish between problem and non problem banks and a significant difference was noticed between the liquidity of both the groups. This can be noted that from the prediction powers to the impact powers liquidity ratios has been playing an important role in the literature, therefore, liquidity ratio will be incorporated in the present study playing an important independent variables in finding out its relationship with credit risk rating in Pakistani banks. Hence the hypothesis(H1), (H2) and (H3) cosidered here would be estimating that the credit risk rating is predictable through liquidity ratios of the banks.

Liquid Assets/ Total Assets was used by Sinkey (1975) and Meyer & Pifer (1970) as liquidity ratio in their study. Kumar & Arora (1995) used Net Advances/ Total Deposits and Liquid Assets/ Total Liabilities to calculate liquidity ratios in their study.

### Capital Adequacy:

Capital adequacy ratio basically reflects the banks equity or capital against its risk containing assets and determines a bank's ability to meet its liabilities and risks. (Yeh, 1996) found that firms with higher DEA score had a higher capital adequacy ratio showing a better utilization of its assets. Sensarma & Jayadev (2009) found Capital Adequacy Ratio (CAR) as the only significant ratio in determining the stock returns. Kumar & Arora (1995) found capital adequacy playing an important role in predicting failed and safe banks while developing a risk rating scheme for banks. Sinkey (1975) noticed an inadequate and deteriorating capital position of the average problem bank as compare to the average non problem bank. Thus, capital adequacy would also be incorporated as an important independent variable in this study hypothesising(H4) that banks credit risk rating are predictable using capital adequacy ratio.

Total Equity/ Total Assets was taken as capital adequacy by Kumar & Arora (1995) in their study.

### Profitability:

(Beaver, 1966), Altman (1968), Ohlson (1980), Adams, Burton And Hardwick (2003), Pinches & Mingo (1973) and Libby (1975) in their analysis concluded that profitability/ income ratios play a very important role in prediction of bankcruptcy. Gray, Mirkovic & Ragunathan (2006) found profitability ratios to have a significant impact on credit risk rating of firms. Yeh (1996) also found a positive impact of profitability on theDEA scores calculated while distinguishing efficient and inefficient banks where higher DEA scored banks had a higher rate of profitability ratios. An important role of profitabilitywas also noticed when a study conducted by Kumar & Arora (1995) successfully predicted failed and safe banks. The simple growth trend in net income would be identified here expecting as a hypothesis(H5) that growth in income can predict redit risk rating and another hypothesis (H6) that growth in admin expenses is a predictor of credit risk rating.

### Asset Quality:

(Kumar & Arora, 1995) found a significant role of asset quality while developing the risk schemefor banks and analyzing the predictive ability of ratios in classifying failed and safe banks. Thomson (1991) observed asset quality gaining importance as predictor of failure with the increase of time. Asset quality will aslo be an important part of independent variables in this study with ahypothesis that it is directly proportional to credit risk rating. As the ratio calculated under this head represents the non performing advances as a percentage of total assets, so lower the value observed, better the asset quality would be and under the hypothesis (H7) of this head credit risk rating would be expected tobe predicted on the basis of asset quality.

(Kumar & Arora, 1995) and (Sarkar & Sriram, 2001) measured asset quality using NPLs/ Total Assets in their study.

### Chapter - V: Results

(Hair, Black, Babin, Anderson, & Tatham, 2006) explains discriminant analysis as a multivariate dependence technique used to predict and explain nonmetric variables as in the case of present study. The independent variable used in the study is a categorical variable i.e. credit risk rating reflected in nonmetric nature of variable. As multiple regression deals primarily with metric variables, multiple discriminant analysis is a technique used for nonmetric ones. The primary purpose of MDA is to spot the group to which the object belongs.

Out of the total observations 75% of the observations reflected the first three year 2005-2007 whereas the other 25% reflected the observations of year 2008. Therefore, the first three year data was taken as selection cases in the study, analysis of which was then conducted on hold out sample, other 25%, to predict the credit risk rating of the bank in the recent year i.e. 2008.

Table - I: Variables Entered/Removed | ||||
---|---|---|---|---|

Step |
Entered |
Wilks' Lambda | ||

Statistic |
Sig. | |||

1 |
Total Equity / Total Assets (R4) |
.766 |
.004 | |

At each step, the variable that minimizes the overall Wilks' Lambda is entered. |

The results in Table - I reflects the only one independent variable found significant while predicting the credit risk rating of the banks in Pakistan. This one variable is R4 i.e. the ratio used in this study for the capital adequacy denoted by Total Equity/ Total Assets. The significance of only one variable was an unexpected coverage but the prediction level of this variable was in line with the expectations while conducting the study. The results in Table - I also shows the value of Wilks' Lambda i.e. 0.766 which is lowest value of Wilks' Lambda in all the variables used in the analysis. While entering any other variables in the model, the value of Wilks' Lambda increased, therefore, the lowest possible Wilks' Lambda reflects only one variable to be significant.

Canonical discriminant functions were also used in the analysis. The value of canonical correlation 0.484 shows that neither a very weak nor a very strong relationship exists between the dependent and independent variables found significant in the study, on the other hand, this value is a sign of satisfactory relationship existing between the dependent and independent variable. The value of R close to zero shows that very weak relationship exist between two variable whereas the value of R close to one reflects a very strong relationship among two variables. Since R is 0.484 here, therefore, R2 being coefficient of determination is found to be (0.484)2 = 0.234 which shows that 23.4% of the variance in dependent variable is due to this model. As the model reflects only one significant independent variable, therefore, it concludes that capital adequacy explains 23.4% of the credit risk rating assigned by PACRA to banks in Pakistan.

Table-II shows that capital adequacy is powerful enough to predict credit risk rating of 72.7% of the selected cases (i.e. 2005-2007) correctly whereas it was more efficient in predicting the credit risk rating of the hold out (unselected) cases (i.e. 2008) through predicting 81.8% of the hold out sample correctly. The results show that capital adequacy is a valid predictor of credit risk rating and helps in predicting at least 72.7% of cases for the period of three years whereas 81.8% of cases for less than one year. Thus, our hypothesis (H4) is accepted showing the good predictive accuracy as observed in the results. As none of the other variables were found to be significantly predicting the credit risk rating, all other hypotheses i.e. (H1), (H2), (H3), (H5), (H6) and (H7) are rejected in line with the results showing that liquidity ratios, profitability and asset quality are found unable to predict the credit risk rating of the banks in Pakistan.

Table - II: Classification Resultsa,b | |||||||

Rating (PACRA) |
Predicted Group Membership |
Total | |||||

1 |
2 | ||||||

Cases Selected |
Count |
1 |
19 |
3 |
22 | ||

2 |
6 |
5 |
11 | ||||

% |
1 |
86.4% |
13.6% |
100.0 | |||

2 |
54.5% |
45.5% |
100.0 | ||||

Cases Not Selected |
Count |
1 |
6 |
2 |
8 | ||

2 |
0 |
3 |
3 | ||||

% |
1 |
75.0% |
25.0% |
100.0 | |||

2 |
00.0% |
100.0% |
100.0 | ||||

a. 72.7% of selected original grouped cases correctly classified. b. 81.8% of unselected original grouped cases correctly classified. |

As discussed earlier in the study, the financial ratios for the number of banks other than those taken in the sample were also calculated. The various descriptive of twenty-nine banks for year 2008 are in Table - III that reflect some interesting figure. The minimum liquid assets noted were 7% of total assets (R1) where the maximum for the same stood at 91% in the industry. The average liquid assets were lying around 33% of the total assets in the industry. Net loans as a percentage of total deposits were at a minimum of 51% and a maximum of 1.10% whereas the average of the industry reflected around 72%. A minimum of as low as 8% liquid assets were found available against the total liabilities against the as good as 4.29% of the maximum. The average liquid assets available against total liabilities of the industry were not satisfactory enough at the level of 49%.

Capital adequacy was at a minimum of 5% in the industry in comparison to a peak of 79%. The average capital adequacy of the industry stood around 49%. The average income growth of the industry was around 26% whereas the administrative expenses grew on an average of 50%. The NPLs against total assets were from the minimum of zero percent to the maximum of 23% whereas average asset quality of the industry was around 6%.

Table - III: Descriptive Statistics for the 29 banks on basis of data availability (2008) | |||

Minimum |
Maximum |
Mean | |

Liquid Assets / Total Assets (R1) |
.07 |
.91 |
.3283 |

Net Loans / Total Deposits (R2) |
.51 |
1.10 |
.7193 |

Liquid Assets / Total Liabilities (R3) |
.08 |
4.29 |
.4923 |

Total Equity / Total Assets (R4) |
.05 |
.79 |
.1573 |

Net Income Growth (R5) |
-5.92 |
20.32 |
.2567 |

Admin. Expenses Growth (R6) |
.03 |
2.21 |
.4930 |

NPLs / Total Assets (R7) |
.00 |
.23 |
.0570 |

The other findings during analyzing twenty-nine banks out of thirty-nine banks in the industry, which reflects more than seventy percent of the industry, are also remarkable for discussion. The total assets of the industry had a worth of more than 5 billion rupees whereas the liquid assets consisted of more than 1.5 billion rupees including a sum of more than 400 million rupees. Net investments valued to be more than 250 million rupees whereas net advances given by the banks in Pakistan had a value of more than 2.8 billion rupees on December 31, 2008. More than 250 million rupees were under the non-performing status in the industry. With total deposits of more than 3.9 billion rupees, the total liabilities stood at 4.5 billion rupees. The total capital reflected in the industry had a worth of more than 500 million rupees in the industry. The net income of the industry summed up to more than 52 million rupees after paying off a taxation of around 25 million rupees.

Hypotheses Assessment Summary: | ||||

Independent Variables |
Measurement Ratio |
Hypothesis |
Findings | |

Liquidity |
Liquid Assets / Total Assets (R1) |
H1: Credit Risk Rating is predictable using R1. |
R1 was unable to predict the credit risk rating. |
Rejected |

Net Loans / Total Deposits (R2) |
H2: Credit Risk Rating is predictable using R2. |
R2 was unable to predict the credit risk rating. |
Rejected | |

Liquid Assets / Total Liabilities (R3) |
H3: Credit Risk Rating is predictable using R3. |
R3 was unable to predict the credit risk rating. |
Rejected | |

Capital Adequacy |
Total Equity / Total Assets (R4) |
H4: Credit Risk Rating is predictable through Capital Adequacy. |
Capital Adequacy was only independent variable that can predict Credit Risk Rating. |
Accepted |

Profitability |
Net Income Growth (R5) |
H5: Credit Risk Rating is predictable using R5. |
R5 was unable to predict the credit risk rating. |
Rejected |

Admin. Expenses Growth (R6) |
H6: Credit Risk Rating is predictable using R6. |
R6 was unable to predict the credit risk rating. |
Rejected | |

Asset Quality |
NPLs / Total Assets (R7) |
H7: Credit Risk Rating is predictable through Asset Quality. |
R7 was unable to predict the credit risk rating. |
Rejected |

### Chapter - VI: Conclusion

The purpose of the research was to understand the relationship between the financial ratios and the credit risk rating in Pakistan. The previous studies showed the evidence that financial ratios are powerful enough to predict bankruptcy, credit risk and credit ratings. Therefore, the present study tested the predictive power of financial ratios in the financial sector of Pakistan. In order to serve the purpose, the banking industry was targeted to evaluate the financial characteristics with respect to credit risk rating assigned by PACRA. The methodology used in the study in order to check the discriminating and predictive power of financial ratios was Multiple Discriminant Analysis (MDA). The financial ratios included based on their importance in previous studies were liquidity ratios, capital adequacy, and profitability and asset quality ratios. The study carried out a trend of four years 2005-2008 where first three years were discriminated and then the model was tested on the fourth year. The prediction accuracy of original sample was around seventy-three percent whereas around eighty-three percent of the observations in hold out sample were predicted correctly. The only ratio found significant in predicting the credit risk rating was capital adequacy, which also showed that more than twenty-three percent of change in credit risk rating was due to the model developed. The limitations of the study is that only financial characteristics of the banks were evaluated and result also suggest the evaluation of market related or regulatory related variables for the future scope. Another limitation also lied in the small sample size due to non availability of data which can be waived off in the future with an increased sample size with the passage of years and popularity of the credit risk rating in the country.

### References:

Adams, M., Burton, B., & Hardwick, P. (2003). The Determinants of Credit Ratings in the United Kingdom Insurance Industry. Journal of Business Finance & Accounting , 30, 539-572.

Aharony, J., Jones, C., & Swary, I. (1980). An Analysis of Risk and Return Characteristics of Corporate Bankruptcy Using Capital Market Data. Journal of Finance , 1001-1016.

Altman, E. I. (1968, September). Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy. The Journal of Finance , 589-609.

Aziz, A., Emanuel, D., & Lawson, G. (1988). Bankruptcy Prediction-an Investigation of Cash Flow based Models. Journal of Management Studies , 419-437.

Beaver, W. H. (1968). Alternative Financial Ratios as Predictors of Failure. The Accounting Review , 71-111.

Beaver, W. H. (1966). Financial Ratios As Predictors of Failure. Journal of Accounting Research, Vol. 4, Empirical Research in Accounting: Selected , 71-111.

Beaver, W. H., McNichols, M. F., & Rhie, J.-W. (2005). Have Financial Statements Become Less Informative? Evidence from the Ability of Financial Ratios to Predict Bankruptcy. Review of Accounting Studies , 10 (1), 93-122.

Carty, L. V. (2000). Corporate Credit-Risk Dynamics. Financial Analysts Journal , 56 (4), 67-81.

Chen, R., & Wong, K. A. (2004). The Determinants of Financial Health of Asian Insurance Companies. The Journal of Risk and Insurance , 71, 469-499.

Clark, T., & Weinstein, M. (1983). The Behavior of The Common Stock of Bankrupt Firms. Journal of Finance , 489-504.

Deakin, E. B. (1972). A Discriminant Analysis of Predictors of Business Failure. Journal of Accounting Research , 167-179.

Estrella, A., Park, S., & Peristiani, S. (2000). Capital Ratios as Predictors of Bank Failure. Economic Policy Review , 6 (2).

Grey, S., Mirkovic, A., & Ragunathan, V. (2006). The Determinants of Credit Ratings: Australian Evidence. Australian Journal of Management , 31 (2), 333-354.

Hair, J. F., Black, B., Babin, B., Anderson, R. E., & Tatham, R. L. (2006). Multivariate Data Analysis. Prentice Hall.

Horrigan, J. O. (1966). The Determination of Long-Term Credit Standing with Financial Ratios. Journal of Accounting Research , 44-62.

Kumar, S., & Arora, S. (1995). A Model for Risk Classification of Banks. Managerial and Decision Economics , 16 (2), 155-165.

Libby, R. (1975). Accounting Ratios and the Prediction of Failure: Some Behavioral Evidence. Journal of Accounting Research , 13 (1), 150-161.

Metwally, M. (1997). Differences between the financial characteristics of interest-free banks and conventional banks. European Business Review , 97 (2), 92-98.

Meyer, P. A., & Pifer, H. W. (1970). Prediction of Bank Failures. The Journal of Finance , 25 (4), 853-868.

Mossman, C. E., Bell, G., Swartz, M., & Turtle, H. (1998). An Empirical Comparison of Bankruptcy Models. Financial Review , 35-54.

Ohlson, J. A. (1980). Financial Ratios and the Probabilistic Prediction of Bankruptcy. Journal of Accounting Research , 18, 109-131.

Papoulias, C., & Theodossiou, P. (1992). Analysis and Modeling of Recent Business Failures in Greece. Managerial and Decision Economics , 13 (2), 163-169.

Pinches, G. E., & Mingo, K. A. (1973). A Multivariate Analysis of Industrial Bond Ratings. The Journal of Finance , 28, 1-18.

Pottier, S. W. (1998). Life Insurer Financial Distress, Best's Ratings and Financial Ratios. The Journal of Risk and Insurance , 65 (2), 275-288.

Reynolds, S., Fowles, R., Gander, J., Kunaporntham, W., & Ratanakomut, S. (2002). Forecasting the Probability of Failure of Thailand's Financial Companies in the Asian Financial Crisis. Economic Development and Cultural Change , 51 (1), 237-246.

Sarkar, S., & Sriram, R. S. (2001). Bayesian Models for Early Warning of Bank Failures. Management Science , 47 (11), 1457-1475.

Sensarma, R., & Jayadev, M. (2009). Are bank stocks sensitive to risk management? The Journal of Risk Finance , 10 (1), 7-22.

Sinkey, J. F. (1975). A Multivariate Statistical Analysis of the Characteristics of Problem Banks. The Journal of Finance , 30 (1), 21-36.

Tanthanongsakkun, S., & Treepongkaruna, S. (2008). Explaining Credit Ratings of Australian Companies—An Application of the Merton Model. Australian Journal of Management , 33 (2), 261-276.

The Pakistan Credit Rating Agency Limited. (n.d.). Retrieved December 2009, from The Pakistan Credit Rating Agency Limited: http://www.pacra.com/

Thomson, J. B. (1991). Predicting Bank Failures in the 1980s. Economic Review , 9-20.

Wheelock, D. C., & Wilson, P. W. (2000). Why do Banks Disappear? The Determinants of U.S. Bank Failures and Acquisitions. Review of Economics and Statistics , 82 (1), 127-138.

Yeh, Q.-J. (1996). The Application of Data Envelopment Analysis in Conjunction with Financial Ratios for Bank Performance Evaluation. The Journal of the Operational Research Society , 47 (8), 980-988.