Professional Documents
Culture Documents
ABSTRACT
This study assesses the classification accuracies of two statistical methods namely, multiple discriminant analysis and logistic
regression approach in default prediction. It uses a small sample of 32 India firms listed on Bombay Stock Exchange for the
sample period of six years over 2010-11 to 2015-16. Two models have been built using the two statistical methods. Results of the
study clearly indicate that there are no significant differences in the classification accuracies because of change in statistical
methods. Different statistical measures clearly suggest that the two developed models have comparative classification accuracies
in default prediction. However, the specification and robustness of logit model is found to be on lower side than that of the
discriminant model contrary to large numbers of studies. The weaker model specification of logistic model may be result of
small sample size.
INTRODUCTION of study two models have been built using the two statistical
methods.
Any form of external liability is risk to any firm. It is true for
every size of firm. It is quite possible that a firm may fail to LITERATURE REVIEW
meet its obligations in time. It can happen even with financially
For more than a century there have been efforts to predict
healthy firms. In today’s times when large numbers of the
probable defaults before those actually occur. However, prior
borrowers are highly leveraged, even the slightest negative
to the breaktrhough study by Beaver (1966), most of the
fluctuations in the rate of interests may prove very costly for
studies had been using univariate methods that could give
the firms (Bandyopadhyay A., 2006). In response to a number
multiple signals. Beaver exhibited the importance of diffrernt
of factors (external and internal) as well as Basel norms for
financial ratios in prediction of failure of the firms as well as
banking, RBI has been issuing guidelines to the banking and
role of credit extension to corporate by banking and financial
financial sectors to deal with increasing non-performing assets
sector (Beaver, 1966). Altman (1968) extended the work of
but these guidelines have not been able to prevent events of
Beaver (1966) by developing a single and directional credit
default by corporate borrowers.
scoring model. This model used multivariate method to a clear
Even after concentrated efforts to improve the situation on signal about possible default. This study used multiple
NPAs, the problem of NPA is alarming. The numbers of discriminant analysis to build a Z-score model using different
defaults as well as wilful defaults are increasing. As a result financial and market ratios to arrive at a single score. The
the banking and financial system is forced to take deep debt developed score(Z-score) could provide clear information
haircut to recover a portion of its NPA. As per RBI financial about the possible default by the firm in near future. The
stability report 2016, the NPA of scheduled banks has developed model has classification accuracy of 94 percent
increased significantly (RBI, 2016). In such circumstances, two years prior to default (Altman E. I., 1968). This model
prior information about probable defaults by corporate revolutionsed the whole arena of defualt prediction forever.
borrowers could be of very help for the whole banking and Even today, banking and financial sector across world is using
financial system for corrective measures, both at operative as this method for developing internal credit scoring models for
well as regulatory levels. evaluting loan applications along with other methods.
As per available literature, there are few studies on different Over time different approaches for default prediction have
methods and models for default prediction in Indian context. been developed. In theory when the value of assets is lower
Moreover these studies cannot be termed to be enough. The than that of the liabilities or there is lack of liquidity, firms
present study assesses the predictive ability of two statistical tend to default (Wilcox, 1971). Ohlson (1980) applied
methods namely multiple discriminant analysis and logistic conditional Logit model on a data set with 105 bankrupt and
regression approach in default prediction. It uses a small 2058 non-bankrupt firms from the US over the period of 1970-
sample of 32 India firms listed on Bombay Stock Exchange 1976 using 7 financial ratios and 2 categorical variables. The
for the sample period of 2010-11 to 2015-16. For the purpose classification accuracy was lower than that of the studies based
on MDA (Ohlson, 1980). Scott (1981) found that a few
1
Assistant Professor, Sri Aurobindo College (Evening), University of Delhi, E-mail : rajeevupadhyay@live.in
There is a need for more studies in the area of default Sales/TA = Sales to total assets ratio
prediction which could explore different aspects of default The collected data for the ratios after checking for outliers
prediction models and practices such as impact of size, sector has been processed in SPSS using multiple discriminant
and change in statistical methods and models on classification analysis and as a result, the analysis yields the following
accuracy.The present study aims to assess classification model:
accuracies of developed model with the change in statistical
method. For predicting default of bank loans by Indian Z = -1.099 – 0.488NWC/TA – 0.391RE/TA + 10.229EBIT/
corporates, this study uses two statistical methods namely TA + 0MVE/BVD + 0.334Sales/TA
For the logistic regression the study has used the same Part One: Multiple Discriminant Analysis
variables used by Ohlson (1980) model to build O-score model
for the sample on the lines of following model (Ohlson, 1980). Table 1.1 provides information about mean and standard
deviation of the ratios used in multiple discriminant analysis
O-Score = ± – ²1 WC/TA + ²2TL/TA + ²3CL/CA + ²4NI/TA + to predict default. From the table, it is clear that there is huge
²5FFO/TL + ²6CHIN + ²7Size + ²8X + ²9Y difference among the ratios between the two groups. The mean
Where of ratios of non-defaulting group are found to be more robust
than that of the defaulting group. Similarly standard deviation
WC/TA = Working capital to total assets ratio of ratios of the defaulting group is unfavourably high in
TL/TA = Total liabilities to total assets ratio comparison to non-defaulting group. On comparing the mean
and standard deviation of each ratio for both the groups, it
CL/CA = Current liabilities to current assets ratio signifies that these two groups are clearly different and may
NI/TA = Net income to total assets ratio be making two separate clusters. Though, this requires further
investigations. However, this pattern is more evident in case
FFO/TL = Funds from operations to total liabilities of market value of equity to book value of debt ratio as the
CHIN = (NIt - NIti)/(INItI + INIt-il), where NIt is net income mean of this ratio has deteriorated sharply from 101.63 to
for the most recent period. The denominator acts as a level 0.53. The findings of the study are in confirmation with the
indicator. The variable is thus intended to measure change in findings of Wilcox (1971) which finds that when a firm is
net income about to default its key ratios relating to liquidity and structure
start deteriorating over time (Wilcox, 1971).
Size = Log (TA/GNP price-level index)
From the table 1.2, it is clear that there are weak correlations
X = 1 if TL>TA, 0 otherwise among most of the variables besides negative correlations
Y = 1 if a net loss for the last two years, 0 otherwise among a few. However, there are strong positive correlations
between NWC/TA and RE/TA and RE/TA EBIT/TA. Strong
The collected data for the ratios after checking for outliers positive correlation is undesirable in the analysis as weak and
has been processed in SPSS using logistic regression analysis negative correlation is favourable condition for predicting
and as a result, the analysis yields the following model: default (Cochran, 1964).
O-Score = -1.289-0.974WC/TA + 2.48TL/TA + 0.21CL/CA
- 4.307NI/TA + 0.614FFO/TL + 0.251CHIN - 0.302Log (TA/
GNP) + 17.54X + 2.617Y
The overall sample classification accuracy of the derived caused by the small sample size than the small sample period.
model is found to be 90.2 percent. This level of classification Moreover, it requires further investigations to confirm this
accuracy is comparable to results of many studies such Ohlson result.
(1980), Bandyopadhyay (2006) Acharya, Chatterjee, & Pal
CONCLUSION
(2003), Agarwal & Taffler (2008), Mishra & Singh (2016),
Gupta (2014), Sharma, Singh, & Upadhyay (2014) and The overall sample classification accuracies are found to be
Upadhyay (2018). The classification accuracy for the non- 89.1 percent with misclassification of 10.9 percent for
defaulting firms is 96.9 percent with Type I error of 3.1 discriminant model and 90.2 percent with misclassification
percent. The classification accuracy for the defaulting firms of 41.2 percent for logit model. So on the basis of accuracies,
is merely 58.8 percent with Type II error of 41.2 percent. it can be concluded that there are no significant differences
However the overall misclassification is 9.2 percent only. in the classification accuracies of developed default prediction
Considering the small sample size as well as the overall models (DPM) because of the change in statistical methods.
misclassifications, the accuracy of the specified model can Also the classification accuracies of the developed models
be termed satisfactory but the high level of type II error makes are comparable to many studies like Altman (1968), Ohlson
the specification of model weaker. (1980), Bandyopadhyay (2006) Acharya, Chatterjee, & Pal
Part Three: Comparison of Results (2003), Agarwal & Taffler (2008), Mishra & Singh (2016),
Gupta (2014), Sharma, Singh, & Upadhyay (2014) and
From the table 3.1, on comparing the two developed models, Upadhyay (2018). So it can be concluded that small sample
it is observed that there are no significant differences between size seems to have no negative impact on the classification
the two models on account of overall accuracy and Type I accuracies of default prediction models.
error but on the basis of Type II errors, the relative robustness
However the relative robustness as well as the specification
and specification of the developed discriminant model is found
of the developed discriminant model is found to be higher
to be stronger than that of the logit model. So it can be
than that of the developed logit model. So it can be said that
concluded that for a relatively small sample size and small
the robustness and specifications of the developed models is
sample period, there are no significant differences in
sensitive to change in statistical method.The weak model
classification accuracies of default prediction models.
robustness and specification may be the cause of higher type
However the robustness and specifications of the developed
II error in the developed logit model which may have been
logit models seems to be affected either by small sample size
caused by small sample size. Though, it requires further
or small sample period or both. No tool has been used to
investigation to confirm this result.
directly assess the impact of sample period in this study.
Though, the fact remains intact that there must be some impact For the defaulting firms, the key ratios relating to liquidity
of sample period as the changing values of ratios must be and structure start deteriorating than that of the non-defaulting
reflecting on the result. So vaguely it can be argued that the firms. From the mean and standard deviation of each ratio for
weak model specification and robustness might have been both the groups, it seems that these two groups are making