Topic: Risk Management and Governance in a Global EnvironmentAssignment must be worded and look exactly like the Annotated Bibliography template provided. The five aricles are provided also.

The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 40 Equity Risk Exposure:

A Case of Indian Banking Industry * Research Associate under Dr. Ruchi Sharma, Department of Humanities and Social Sciences, IIT Indore, Khandwa Road Simrol, Indore 453552, Madhya Pradesh, India. E-mail: [email protected] * * Research Scholar, School of Economics, University of Hyderabad, Hyderabad 500046, Telangana, India; and is the corresponding author. E-mail: [email protected] © 2018 IUP. All Rights Reserved. Md. Danish * and Aaqib Ahmad Bhat ** Global market integration and increase in trading activities have magnified the financial system complexities and increased the degree of riskiness. Value-at-Risk (VaR) has been universally accepted as a measure of market risk in financial institutions. In this study, using the data of NSE-Nifty Bank Index and indices of SBI and ICICI Bank over a sample period from January 3, 2005 to November 19, 2014, an attempt has been made to analyze the market exposure in Indian banking industry by employing various methods of VaR. The study reveals that there is greater market turbulence during the financial crisis period than the pre- and post-crisis periods through all the three techniques of VaR. Moreover, bifurcating the full sample into three sub-samples (pre-crisis, crisis and post-crisis periods) seems to assure robustness, thereby validating the applicability of VaR methods for the Indian banking sector. Further, backtesting through Kupiec test revealed that historical simulation approach accounts for less statistical noise than other methods of estimating VaR. Introduction Financial institutions have always been imperative for stimulating investment and other financial developments of an economy. Over the past two decades, the financial system has become more complex due to increase in trading activities, which in turn led to increase in the degree of riskiness. Indulging in more financial activity other than lending and depositing of money, although may be more beneficial, at the same time makes them more prone to market turbulences, resulting in high degree of risk in their daily trading activity. Therefore, with increasing complexities and turbulence in the financial system, the risk management becomes the most crucial strategic activity in any financial firm. In the recent past, major financial and non-financial corporations experienced insolvency as evidenced by the insolvency of many financial and non-financial corporations like Lehman Brothers, Washington Mutual, Royal Bank of Scotland, WorldCom, General Motor, and CIT due to the global financial crisis of 2008. All these bankruptcies and market turbulences in the world economy also bear significant impact on Indian economy and particularly on the financial and banking sector. 41 Equity Risk Exposure: A Case of Indian Banking Industry Before proceeding to risk management, it is imperative to first have a glimpse of what risk is? Risk can be understood as the uncertainty about the future, the possibility or chance that something wrong or unpleasant may happen in future. In finance, risk is the probability that actual return on an investment will be lower than expected. Therefore, we may define risk as the future uncertainty about expected return on any financial investment due to market fluctuations. Financial risk management has always been the crucial strategic activity of any firm from their inception. But in the last two-and-a-half decades, it has become the key issue in financial management of any organization due to many recent developments in financial system all across the globe like deregulation, globalization and financial innovation.

Indian economy, after the New Economic Policy in 1991, has experienced several policy and structural changes such as deregulation, removal of trade barriers, financial reforms, etc. The liberalization policy has opened various other sources of earning to the banks which include innovated new financial products, Internet banking, credit cards, mobile banking and many more. The other side of these developments of the banking sector is the introduction of new risk or increase in risk factor. This compelled banks to focus more on the risk management strategy.

Initially, the Indian banks used risk control systems that kept pace with the legal environment and Indian accounting standards. In India, the statutory regulation of commercial banks by Reserve Bank of India (RBI) until the early 1990s was mainly focused on licensing, administration of minimum capital requirements, pricing of services including administration of interest rates on deposits as well as credit, reserves and liquid asset requirements (Kannan and Aulbur, 2004). But with the growing pace of deregulation and associated changes in the customer’s behavior, banks are exposed to mark-to-market accounting. In order to maintain the regulatory framework in the country in response to market dynamics, banks have to follow certain risk management norms as suggested by both RBI and Bank for International Settlements.

The increased financial developments also heightened the risk and to realize the benefit of financial development, it is important to manage the associated risk professionally. In this context, the present study employs the most preferred risk management method: Value-at- Risk (VaR) method to empirically assess the risk associated with Indian banking industry.

The VaR method in the context of portfolio theory developed by Markowitz (1952) is widely applied to estimate market risk and exposure. Moreover, Basel I also suggested banks to use the VaR methodology as an internal tool to quantify and manage the market risk of any financial firm or bank.

The study focuses on the Indian banking industry due to the emerging importance and role in financial development in the Indian economy. Moreover, to the best of our knowledge, none of the studies earlier have analyzed the market exposure associated with Indian banking industry along with comparative analysis of different VaR techniques that fit and accurately measure the market exposure in Indian banking industry. Particularly, this study focuses on the NSE Nifty Bank Index, State Bank of India (SBI), and Industrial Credit and Investment Corporation of India (ICICI) indices. The study is organized as follows: following the The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 42 introduction, a brief literature review is presented. Subsequently, data and methodology used in the study are discussed, followed by a discussion of the results obtained. Finally, the conclusion is offered.

Literature Review In the arena of volatility and risk, both market participants and the market managers need to adopt techniques to measure and manage market risk. The success story of risk management models mainly depends on the estimates of the volatility of underlying prices. VaR methods have been extensively used in forecasting future returns, therefore it becomes imperative to analyze which of the methods estimates market exposure with higher authenticity and accuracy.

To have a perspective on the comparative performance of the VaR methods, both the empirical and methodological studies have been reviewed thereon.

Dutta and Bhattacharya (2008) evaluated the predictive performance of VaR methods for the stock market of India. The VaR model, assuming linearity (variance-covariance approach), is consid ered an efficient candid ate for the financial retu rns ove r th e period under consideration. The authors found the technique quite satisfactory and more comprehensive as it takes into account the scarcity of inadequate data. Further, the technique was found to work even in nonlinearity as it can take care of the volatilities and the correlations in the data.

Rejeb et al. (2012) estimated the market exposure for three currencies (US dollar, euro, and Japanese yen) and four currency portfolios in the Tunisian currency market during the time period 1999-2007. Besides, the study analyzed and compared the empirical estimates of four VaR simulations, namely, the variance-covariance, historical simulation, bootstrapping and Monte Carlo. The study highlighted that market risk primarily depends on the degree of portfolio diversifications. Among the three currencies, Japanese yen was found to be the most risky currency, followed by the US dollar, while euro was found to be a less risky one.

By employing the backtesting technique, the study revealed that three methods underestimate the VaR. However, the least error is found by using the variance-covariance method, then the Monte Carlo and finally, the historical simulation method. Therefore, the authors concluded that traditional variance-covariance method predicts the foreign exchange risk better in the Tunisian exchange market than the other two methods.

In contrast to the above studies, Yang (2010) while examining the market exposure of Chinese stock market through various approaches of VaR, found that variance-covariance method is easily influenced by market turbulence, especially the effect of global financial crisis in 2008. The study also compared the market exposure of Chinese market to the US market and Japanese market. The study found that the behavior of Chinese market is similar to the Japanese market, however, VaR models are able to predict the market risk of Chinese market better than the other two markets.

Pritsker (1997) sketched down the methodologies of the VaR for the twin attributes of accuracy versus computational time. The author held that the recent development in the literature of the risk models reveals that the outcomes of the studies vary widely, thus forcing the researcher to choose among the appropriate models. The study found that the 43 Equity Risk Exposure: A Case of Indian Banking Industry method of delta-gamma Monte-Carlo simulations was the most satisfactory among the class of the techniques used for examining market exposure. Mentel (2013) also found Monte- Carlo simulation to perform well and being a representative of a group of nonparametric methods. However, Hull and White (1998) found that for investment in securities, simple historical simulation method is more feasible. Bohdalova (2007) compared the different VaR methods of risk measurement by taking into account a hypothetical data on government bonds that mature on monthly basis. The authors found that the authenticity of all the VaR models holds and recommended that one should compare them if drawn on the comparable assumptions.

In contrast to the above studies, Lambadiaris et al. (2003) assessed the performance of historical and Monte-Carlo simulation as alternative approaches to calculating VaR by using data from Greek stock and bond market. Based on various backtesting criteria, mixed results were obtained on the accuracy and adequ acy of differen t VaR approaches. Sarma et al. (2003) tried to devise the techniques and the ways to choose the best models among the most competing models. For the purpose, the author used two-stage model selection procedure (statistical accuracy test in terms of pre-specified failure rate, followed by comparing the loss functions) for the S&P 500 index and India’s NSE 50 index. The study found similar inconclusive results regarding the selection of the unique model for risk valuation.

In accordance with the above studies, Linsmeier and Pearson (1996) made a comparative discussion of various methodologies aimed to measure the VaR. The study found inconclusive results in prioritizing the VaR methods.

Van and Vlaar (1999), in their comparative evaluation of various VaR techniques on Dutch stock market index AEX and to the Dow Jones Industrial Average, came out with multiplicity of conclusions regarding the VaR modeling. The study found volatility clustering as the most important feature of stock market returns for VaR modeling and therefore advocated the use of GARCH models for precise modeling setup as this could effectively reduce the average failure rates and the fluctuations of failure rates over time. Similar analysis was done by Hull and White (1998) who in their treatment of VaR estimation proposed the use of volatility updating scheme like GARCH as a supplementary approach to the historical simulation method in order to arrive at a better measure of VaR. Varma (1999) tried to identify the best predictive model in the case of Indian financial system. The author found that the Generalized Autoregressive Heteroscedasticity with Generalized Error Distribution residuals (GARCH-GED) perform exceedingly better than any other model. The author recommended that in order to better judge the performance of the VaR models, we should also take into account the movements of foreign portfolio prices.

Patr a an d Pad h i (2 0 1 5 ) an alyzed th e p res en ce o f Au to regres sive C o n d iti o n al Heteroscedastic (ARCH) and long-memory effects in the daily closing prices of the Bombay Stock Exchange (BSE)-BANKEX return series of India. The study employed different methods of VaR calculation such as Asymmetric Power ARCH (APARCH), Fractionally Integrated Exponential Generalized ARCH (FIEGARCH), Hyperbolic Generalized GARCH (HYGARCH) and risk metrics to check the accuracy of these methods in predicting the bank return series The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 44 in India. Furthermore, the study also examined the forecasting capabilities of these VaR methods through backtesting techniques such as Kupiec Likelihood Ratio (LR) test and dynamic quantile test.

The study found that banking shares in India have both long memory and asymmetry effects. The results based on Kupiec LR test and dynamic quantile test show that both asymmetries and long memory play an important role in market risk evaluation and its forecasting. A comparative analysis indicated that the results varied across different methods of estimating risk exposure, however HYGARCH model was found to perform well. Giot and Laurents (2003) also used VaR models to analyze long and short trading positions in commodity markets. An out-of-sample analysis is carried on various metals, energy oils and agricultural commodities to assess the performance of the Risk Metrics, skewed Student APARCH and skewed Student ARCH models. The study found skewed Student APARCH model performed best in all cases as its calculation does not require nonlinear optimization procedures.

Roynstrand et al. (2012) examined the statistical power of different methods of VaR through backtesting technique. The results of the study indicated that geometric conditional coverage test given by Berkowitz et al. (2011) performs best over all other methods of VaR.

Moreover, to have a satisfactory power of various tests, sample sizes of 1,000, 750 and 500 data points are found to be the minimum requirement when testing 1%, 5% and 10% VaR, respectively. Therefore, the belief of sample size of 250 data points, which is the minimum requirement set by the Basel Committee on Banking Supervision (2011), seems to be misspecified.

It may be summed up that over the years, accuracy, prediction and computational procedures of the VaR models have been put on the screen of methodological debate to assess the authenticity and validity of various approaches. However, the literature shows no consensus. Some studies have advocated the supremacy of variance-covariance approach, while others held the predominance of Monte-Carlo and historical simulations as a method for examining market exposure. Moreover, the validity of different approaches of VaR also varies across sectors and for different types of assets that are being examined. Therefore, the present study has adopted a holistic approach to analyzing market risk through widely accepted and recommended measures of VaR. Data and Methodology Data To carry out the study, data has been collected for daily closing prices of NSE-Nifty Bank Index, State Bank of India (SBI) and ICICI Bank from the NSE-Nifty database for the period from January 3, 2005 to November 19, 2014. Nifty Bank Index exhibits the behavior and performance of the highly liquid 12 commercial banks listed on NSE which represent 93.43% of total market capitalization of overall banking sector in Nifty Bank Index. Nifty banking index represents the financial health of Indian banking sector, whereas SBI and ICICI Bank are the 45 Equity Risk Exposure: A Case of Indian Banking Industry largest public and private sector banks respectively in terms of market share and capital in Indian banking industry. Indian banking sector is growing rapidly in the last two decades with a vital role in growth, development as well as the financial stability in Indian economy. The end point of the sample period is selected as stock split has been adopted by various banks thereafter, which led to structural shift in the stock prices of the respective banks.

As can be seen in Figures 1 and 2, due to stock split in SBI and ICICI Bank, there has been a sharp and sudden decline in their stock prices on November 20 and December 4, 2014. Other banks like Canara, PNB, and Axis Bank also did stock split in order to be compatible and increase market share during the same year. The share split should be differentiated from a structural break as stock split is a deliberate action on the part of bankers. It was done to increase capital and market share.

Global financial crisis 2008 badly affected the financial condition of economy across the globe, the main sufferers being the stock markets and the financial firms. Global financial crisis that hit Indian stock market on October 13, 2008 caused benchmark indices of the economy to fall by 50% from the highest record that they scaled in January 2008. ICICI Bank, which is the second largest lender of Indian economy, faced sudden fall in its share price following rumors that it is exposed to toxic USA and UK assets. In order to analyze and segregate the impact of global financial crisis on the financial stability and performance in Figure 1: SBI Closing Price 4000 3500 3000 2500 2000 1500 1000 500 0 Clo sin g P ric e ( ) Date 1 0-M ay -0 5 1 3-S ep t- 0 5 1 9 -J a n -0 6 3 1-M ay -0 6 0 3-O ct- 0 6 0 9-F eb -0 7 2 0-J u n-0 7 2 4-O ct- 0 7 2 7-F eb -0 8 0 9-J u l- 0 8 1 8 -N ov-0 8 0 2 -A pr-0 9 1 2 -A ug-0 9 2 1-D ec-0 9 0 4-M ay -1 0 0 3-S ep -1 0 1 0 -J a n -1 1 1 9-M ay -1 1 2 3-S ep t- 1 1 0 1-F eb -1 2 0 7-J u n-1 2 1 2-O ct- 1 2 1 9-F eb -1 3 2 7-J u n-1 3 0 5 -N ov-1 3 1 2-M ar-1 4 2 1-J u l- 1 4 0 4-D ec-1 4 1 5 -A pr-1 5 1 8 -A ug-1 5 2 9-D ec-1 5 The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 46 Figure 2: ICICI Bank Closing Price 600 400 200 0 Clo sin g P ric e ( ) Date 0 5-M ay -0 5 0 2-S ep t- 0 5 0 5 -J a n -0 6 1 5-M ay -0 6 1 1-S ep t- 0 6 1 2 -J a n -0 7 1 9-S ep t- 0 7 2 1 -J a n -0 8 2 8-M ay -0 8 2 6-S ep t- 0 8 0 6-F eb -0 9 1 8-J u n-0 9 2 1-O ct- 0 9 2 4-F eb -1 0 2 9-J u n-1 0 2 7-O ct- 1 0 2 8-F eb -1 1 0 1-J u l- 1 1 0 4 -N ov-1 1 0 7-M ar-1 2 0 9-J u l- 1 2 0 9 -N ov-1 2 1 3-M ar-1 3 1 6-J u l- 1 3 2 0 -N ov-1 3 2 2-M ar-1 4 2 5-J u l- 1 4 0 5-D ec-1 4 1 0 -A pr-1 5 1 1 -A ug-1 5 1 6-D ec-1 5 800 1,000 1,200 1,400 1,600 1,800 2,000 2 2-M ay -0 7 India, by observing the trend of closing prices of the NSE-Nifty Bank Index, the analysis has been performed on the sub-sample periods along with full sample period. Accordingly, the full sample has been bifurcated into three sub-periods as:

i. Pre-crisis period (January 3, 2008 to January 14, 2008); ii. Crisis period (January 15, 2008 to March 19, 2009); and iii. Post-crisis period (March 20, 2009 to November 19, 2014).

The first phase includes the time period of the rising phase of Indian economy; the second phase comprises periods of monetary contraction and great depression in Indian economy and, finally, the third and last phase is recovery phase in Indian economy after the global financial crisis. Since VaR for different time periods will be different, to ensure comparability, data regarding other two indices is also restricted to this time period and similar bifurcation has been done.

Methods A number of methods have been used in the literature to quantify the financial risk of the stock prices, but the most advanced and commonly used method to calculate the financial risk of any financial or non-financial institution is the VaR methodology. VaR tells us about 47 Equity Risk Exposure: A Case of Indian Banking Industry the maximum expected loss of a firm’s portfolio at a given confidence level for a given period of time so as to ensure that the firm has sufficient capital reserve to cover the losses.

The Bank for International Settlements (BIS) Amendment 1996 prescribed VaR for Internal Model Approach (IMA) to be computed as:

•Daily basis with 99% confidence level.

• A horizon of 10-day trading days.

• Observation period must have a historical data of at least one year or nearly 250 days.

Following the literature and based on the objectives of the study, both parametric and nonparametric methods of estimation are employed. More precisely, the study uses variance- covariance approach (parametric), historical simulation (nonparametric) and Monte-Carlo simulation for the estimation of market exposure of NSE-Nifty Bank, SBI and ICICI Bank indices.

Backtesting quantifies the effectiveness and accuracy of VaR models by comparing the actual Profit and Loss (P&L) with the corresponding VaR estimates. If the estimated VaR figures are similar to the real return P&L, then it is said that the model is accurate and the validity of the risk model should be accepted.

Backtesting VAR Methodologies There are a wide range of methods available to backtest the VaR methodologies, but the most frequently used and suggested methods by the Basel Committee are mainly two:

•Kupiec (1995) LR Test • Basel Rule ‘Traffic Light’ Approach (1996) This study employs Kupiec LR test, which is unconditional coverage test. Kupiec test, also known as Probability of Failure (PoF) test, is the frequently used method to test the failure rates. It measures whether the number of exceeding or exception is consistent with the confidence level or not. To check the accuracy of the model in Kupiec LR test, the null hypothesis that the mo del is correct and alternate hypothesis that model is inaccurate are tested. :

ˆ : 0 N X P P H  Model is correct.

where P^ is the observed failure rate, X is the number of exceptions or failure, N is the total number of observations, and X/N is the failure rate. The idea behind setting the null hypothesis is to check whether the observed failure rate ( P^) (VaR value) is significantly different from predicted probability ( P).

Kupiec PoF test is best conducted in the form of LR test which is the likelihood ratio. The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 48 The formula of Kupiec LR test is as follows:                              x x n x x n pof N X N X P P LR 1 ln 2 } ) 1 ln{( 2 LR test follows chi-square ( 2) distribution with one degree of freedom. If the calculated value of LR test exceeds the critical value of chi-square distribution, then we reject the null hypothesis H0that the model is accurate. Results and Discussion First of all, descriptive statistics of the three banking indices (Nifty Bank Index, SBI, and ICICI Bank) over the three phases are given in Table 1. It is observed from the table that quite often for all the three banking indices, the value of mean is positive but close to zero, thus following the standard assumption of zero mean of VaR methodology. Besides, it signifies that there is an increasing trend in the price of all indices during pre- and post-crisis period.

However, in Phase 2 (crisis phase), the value of mean is negative and quite different from zero and the standard deviation is also high, thus revealing the severity of crisis period where the chances of losses in investing in the stocks are greater with high variability. The value of skewness and kurtosis are different from zero, thus exhibiting that the underlying series does not follow standard normal distribution. Similar kind of conclusion regarding the normality has been confirmed from the JB test and its associated p-value. One important point to be noted is that in the third phase SBI shows a very high negative skewness in return. This might be as a result of the greater role for SBI in neutralizing the impact of financial crisis of 2008 and to achieve sustained increase in growth in the post-crisis era. 1 231 2 312 3 Count 7592871408 759286 14087592861408 Mean 0.001–0.004 0.0010.002–0.003 –0.001 0.002 –0.005 0.001 SD 0.0190.034 0.0170.021 0.036 0.065 0.023 0.0490.023 Kurtosis 4.4213.459 10.774.518 3.6491104.63 3.969 4.4938.720 Skewness –0.2290.0660.684–0.206 0.112–31.250 0.028 –0.019 0.613 J-B 70.512.722 365578.25 5.62371424498 29.83 28.25 2009.59 Prob. 0.0000.256 0.0000.000 0.060 0.000 0.000 0.0000.000 Table 1: Summary Statistics of Nifty, SBI and ICICI Bank Indices Phase Statistic Banking Index Nifty Bank Index SBI Index ICICI Bank Index 49 Equity Risk Exposure: A Case of Indian Banking Industry Table 2 represents the VaR figures for NSE-Nifty Bank Index, SBI and ICICI Bank during three different time periods at 1% and 5% level of significance by applying the variance-covariance approach. In accordance with the financial theory, the results indicate an inverse relation between the value of VaR and the level of significance. Higher level of significance (lower confidence interval) was found to be associated with lower value of VaR. Most of the results on the market exposure in the Indian banking industry came as expected. In the second phase of global financial crisis, the VaR figures are quite high at both levels than the other two time periods. High VaR figure reveals that in the second phase, the magnitude of getting a loss on holding a stock was relatively more. Global financial crisis affected almost all economies of the world in all sectors, especially the financial sector.

Similar findings regarding the severity of financial crisis have been found in India and it can be seen from Table 2 where the value of VaR of Nifty Bank Index, SBI and ICICI surged and almost doubled during the time of crisis. Greater VaR values during all the three phases, especially during the crisis period, were found to be associated more with ICICI Bank as compared to SBI and Nifty Bank Index, hence signifying that the stock price of ICICI is more risky and the investors investing in ICICI stocks have greater chances of bearing a loss than investors who invest in SBI and Nifty Bank Index. Further, Table 2 reveals that VaR figures in the first and third phases are close to each other for all three stocks at both confidence levels, showing that the stock price was stable and less risky before and after the financial crisis.

Banking Index Nifty Bank Index SBI Index ICICI Bank Index 1515 15 Phase Phase 1 –4.207–2.932–4.823–3.359 –5.125–3.572 Phase 2 –8.378–6.032–8.653–6.214 –11.96–8.603 Phase 3 –4.005–2.799 –4.992 –3.507 –5.353–3.751 Table 2: VaR Through Variance-Covariance Approach (in %) Level Next, market risk and exposure for the three indices were calculated by the other two approaches of VaR methodology to compare the results with that of variance-covariance approach.

In accordance with variance-covariance approach, VaR figures obtained from historical simulation methods show that the magnitude of market exposure in the second phase (financial crisis period) was more than that in other two phases (Table 3). Similarly, the results of Monte-Carlo simulation as presented in Table 4 provide evidence for greater market turbulence in crisis period for all three stocks in comparison to the other two time periods, thus, again The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 50 Banking Index Nifty Bank Index SBI Index ICICI Bank Index Phase 1515 15 Phase 1 –4.812–3.007 –5.85–3.362 –5.898–3.402 Phase 2 –8.153–5.598–9.499–6.438 –14.7–8.244 Phase 3 –4.229–2.642–4.845–3.463 –5.367 –3.537 Table 3: VaR Through Historical Simulation (in %) indicating that in the financial crisis period, stock prices are more volatile and risky than in the pre-crisis and post-crisis periods. Further, as observed from Tables 2 and 3, VaR figures for ICICI Bank Index in both the approaches where found to exhibit greater chances of facing market turbulence than SBI and Nifty Bank Index. Therefore, ICICI Bank seems to be relatively more risky and prone to market failure than SBI and Nifty Bank Index.

It is worth noting that the magnitude of market risk for SBI in the third phase is larger than the financial crisis period under Monte-Carlo simulation which is probably as a result of using higher initial seed value for SBI data in the post-financial crisis period. Moreover, high negative skewness and kurtosis in SBI returns during the third phase are the other possible reasons. Therefore, the left side of the tail of probability density function for SBI is flatter than right side of the distribution and thus, there is an indication for chances of higher market turbulence in this phase.

From the above results, it can be concluded that all the three approaches confirm that financial crisis is found to be associated with greater market exposure, and therefore the chances of turbulence of Indian banking sector were more during the financial crisis period.

Also, all the three techniques of estimating market risk reveal that ICICI Bank was more risky than SBI and Nifty Bank Index. Least chances of having maximum expected shortfall were found to be associated with the aggregate Nifty Bank Index. Level Banking Index Nifty Bank Index SBI Index ICICI Bank Index Phase 1515 15 Phase 1 –4.501–3.057–4.789–3.053 –5.279 –3.58 Phase 2 –9.022–6.519–9.906–7.692–12.694 –8.897 Phase 3 –4.229–2.815–16.02–10.863 –5.079–3.821 Table 4: VaR Through Monte-Carlo Simulation (in %) Level 51 Equity Risk Exposure: A Case of Indian Banking Industry Backtesting VaR is the modern risk management model which measures the worst expected loss that a firm can face in a given period at a specific confidence level. However, there are various techniques of VaR to quantify the financial risk of any stock price, but we do not know which of the technique is applicable to which dataset and which sector of the economy.

As mentioned earlier, Kupiec LR test is most commonly applied and used frequently to test the PoF. Therefore accordingly, Kupiec test was employed to check the validity of different VaR methods in the context of the Indian banking industry.

Table 5 presents the results of backtesting. It is observed from the results of the full sample period that the variance-covariance approach and Monte-Carlo simulations do not fit the Nifty Bank Index data well. The LR value for both the methods is greater than the tabulated value at 1% level of significance. 1 Similar conclusions of inadequacy and inaccuracy of VaR techniques for fitting the other two (SBI and ICICI) banking indices were found in both variance-covariance and Monte-Carlo simulation techniques. However, in the case of historical simulation technique, the results are insignificant for all the three indices, thereby exhibiting that historical simulation technique of VaR fits and accurately measures the market risk associated with Indian banking industry. The results provide clear evidence that least statistical noise in terms of applicability for Indian banking industry is found with historical simulations rather than the other two techniques. This is due to the fact that unlike variance- covariance approach and Monte-Carlo simulations, historical simulation technique is free from any restrictive assumptions like normality in returns which is not found for the Indian banking sector.

Next, the Kupiec’s backtesting technique is applied on the three phases of all the three indices. This analysis on backtesting for the full sample period was carried out to observe 1 One point to note in backtesting estimation is that estimations in this study were made only at 1% level of significance, the main reason for this being that at both 1% and 5% levels, same results were obtained.

X 40 241 252424 13 245 LR- 8.277*35.24*9.29* 0.0090.0120.0126.606* 35.24* 13.842* Statistic Table 5: Backtesting Aggregate Indices – Kupiec Test Results Note: N = No. of trading days, X is the no. of exceptions, 2 0 .01 = 6.635, * indicates significance at 1% level of significance. Variance-Covariance Historical Simulation Monte-Carlo Simulation Nifty Bank Index ICICI Bank Index N = 2453 SBI Index Nifty Bank Index SBI Index ICICI Bank Index Nifty Bank Index SBI Index ICICI Bank Index The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 52 whether the bifurcation of sample period into three phases (viz., pre-crisis, crisis and post-crisis) has improved the validity of the VaR techniques to fit the Indian banking industry or not.

Table 6 presents the results for checking the validity of variance-covariance approach to fit the Nifty Bank, SBI and ICICI indices. The results found that after bifurcating the sample period into crisis, pre-crisis and post-crisis time periods, variance-covariance method of VaR is found to become viable for all the three indices. From Table 6, it is observed that in all the three indices, the null hypothesis of no difference between the probability value and the observed ratio of number of exceptions to total observation is not statistically significant.

Therefore, by dividing the full sample period into three phases (pre-crisis, crisis and post- crisis), the variance-covariance approach for estimating VaR seems to correctly estimate the market risk. The main reason for this is that the full sample period seems to reduce the VaR value by smoothening the aggregate fluctuations. Therefore, the number of worst exceptions due to financial crisis was more than the predicted probability value; as a result, the model was not found to fit the indices perfectly. By dividing the full sample period into three time periods according to the market conditions and market turbulences, in each phase, we have separate VaR figures and therefore the null hypothesis of accuracy and applicability of the variance-covariance model for the Indian banking sector is accepted.

In Tables 7 and 8, the adequacy and validity of historical and Monte-Carlo methods of VaR are checked after dividing the indices according to three phases. Table 6: Backtesting Variance-Covariance Method – Kupiec Test Results Note: N = No. of trading days, X is the no. of exceptions, 2 0 .01 = 6.635. Phase Phase 1 Phase 2Phase 3 Banking Index Statistic N 7592861408 Nifty Bank Index X 15 319 LR 5.69 0.0071.565 SBI Index X 14 313 LR 4.377 0.00680.0858 ICICI Bank Index X 15 414 LR 5.689 0.4080.0005 53 Equity Risk Exposure: A Case of Indian Banking Industry In Table 7, like the counterpart results of backtesting of variance-covariance method, the value of LR statistic is less than the chi-square value at 1% level of significance for all the three banking indices. Therefore, the null hypothesis is accepted and it is inferred that Table 7: Backtesting Historical Simulation Technique – Kupiec Test Results Note: N = No. of trading days, X is the no. of exceptions, 2 0 .01 = 6.635. Phase Phase 1 Phase 2Phase 3 Banking Index Statistic N 7592861408 Nifty Bank Index X 7 214 LR 0.0476 0.2920.0005 SBI Index X 7 315 LR 0.0476 0.00670.0595 ICICI Bank Index X 7 314 LR 0.0476 0.00680.0005 Table 8: Backtesting Monte-Carlo Simulation Technique – Kupiec Test Results Note: N = No. of trading days, X is the no. of exceptions, 2 0 .01 = 6.635, * indicates significant at 1% level of significance. Phase Phase 1 Phase 2Phase 3 Banking Index Statistic N 7592861408 Nifty Bank Index X 11 114 LR 1.358 1.630.0004 SBI Index X 15 21 LR 5.689 0.29220.99* ICICI Bank Index X 13 421 LR 3.21 0.4084 2.985 The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 54 historical simulation too after bifurcating the full sample period into three separate phases is applicable for the Indian banking indices.

Similar results were found for the Monte-Carlo simulation which is based on Geometrical Brownian Motion for generating the return series. In Table 8, the results of LR test for Nifty Bank and ICICI indices are found to be statistically insignificant, thereby signifying that Monte-Carlo simulations work well for Nifty Bank and ICICI indices. Similar results are obtained for SBI index in the pre-crisis and crisis period. However, in the post-crisis period, different result is obtained. This is mainly due to the fact that in the third phase, SBI index showed greater negative skewness and the kurtosis was also quite high. Therefore, the resulting distribution will be highly negatively skewed.

It is worth noting that although all the methods of VaR were found to accurately measure market exposure after dividing the sample into three phases, the number of exceptions was found to be least for the historical simulation technique than the other two methods.

Therefore, the least statistical noise in applicability for the Indian banking sector seems to be associated with the historical simulations of VaR methodology both for full sample and sub-sample indices.

Therefore, dividing the full sample to analyze the severity of financial crisis was found to improve the validity of the different VaR approaches for examining the market risk associated with Indian banking industry. It may be inferred that all the three approaches of VaR were found to predict the market exposure associated with Indian banking industry accurately. However, least chances of errors were found to be associated with the historical simulation method as compared to the other two methods. Therefore, the results of the study support the recommendation of the Basel Committee to use VaR models as an internal risk management tool.

Conclusion Market risk and the associated exposure is a universal phenomenon in the world, however managing risk is purely an art. The present study tried to examine the market risk and exposure of NSE-Nifty Bank Index. The main reason for analyzing the Nifty Bank Index is its emerging importance in the Indian banking sector, especially after the 1990s. Although there are a few studies on BSE banking index, to the best of authors’ knowledge, no study is available till date that has analyzed the market risk and exposure of Nifty banking index.

Additionally to have a robust analysis, the present study also compared the market risk of Nifty Bank Index with two largest banks, SBI and ICICI Bank (one from public sector and other from private sector) of India. In order to analyze the impact of financial crisis of 2008 on the Indian banking sector, the study divided the data for Nifty, SBI and ICICI Bank indices into three time periods (pre-crisis, crisis and post-crisis periods). Given the applicability 55 Equity Risk Exposure: A Case of Indian Banking Industry and authenticity of VaR methods, the present study used VaR methodology to quantify the market exposure associated with the Indian banking industry. Further, by using the backtesting technique via Kupiec test, the study tried to investigate which of the VaR methods fits the Indian banking index well.

In accordance with the financial theory, the study found inverse relation between the VaR figures and level of significance. As expected, the results of all the three techniques of VaR methodology (variance-covariance, historical simulation and Monte-Carlo simulation) provide evidence of greater market risk and exposure of Indian banking industry in the financial crisis period of 2008, thus signifying the severity of the financial crisis of 2008.

The crisis period was found to be associated with double chances of having a loss than the pre-crisis and post-crisis periods. The study found similar results for all the three banking indices through all the three approaches of VaR. Therefore, the presence of uncertainties during the financial crisis of 2008 was found to hamper the economic performance of Indian banking sector. Among the three indices, ICICI was found to be more prone to market risk than the SBI and Nifty Bank Index. Aggregate Nifty banking index was found to be relatively safer index than the ICICI and SBI indices.

The Kupiec’s backtesting test results revealed that variance-covariance and Monte-Carlo simulation could not perform well for the whole aggregate data for all the three indices.

However, the value of LR for historical simulation test was found to be insignificant, thus indicating that it fits the data well. This is mainly due to the fact that the variance-covariance method and Monte-Carlo simulations are based on restrictive assumptions like normality which are not satisfied in the present case. However, after dividing the series into three phases to examine the impact of financial crisis of 2008, the results changed. After categorization into pre-crisis, crisis, post-crisis periods, all the approaches of VaR methods were found to give fruitful results and were found to fit the Nifty Bank Index and its associated indices. Therefore, dividing the aggregate indices into three phases not only assured authenticity of results but also validated the applicability of VaR methods to the Indian banking sector. The results of the study support the recommendations of the Basel committee to use VaR as an internal model for risk management.  References 1. Berkowitz J, Christoffersen P and Pelletier D (2011), “Evaluating Value-at-Risk Models with Desk-Level Data”, Management Science , Vol. 57, No. 12, pp. 2213-2227.

2. Bohdalová M (2007), “A Comparison of Value-at-Risk Methods for Measurement of the Financial Risk”, Faculty of Management, Comenius University, Bratislava, Slovakia, E-learning Working Paper (Online).

3. Dutta D and Bhattacharya B (2008), “A Bootstrapped Historical Simulation Value at Risk Approach to S&P CNX Nifty”, National Conference on Money and Banking, IGIDR, Mumbai, India. The IUP Journal of Applied Economics, Vol. XVII, No. 1, 2018 56 4. Giot P and Laurent S (2003), “Market Risk in Commodity Markets: A VaR Approach”, Energy Economics , Vol. 25, No. 5, pp. 435-457.

5. Hull J and White A (1998), “Incorporating Volatility Updating into the Historical Simulation Method for Value at Risk”, Journal of Risk, Vol. 1, No. 1, pp. 5-19.

6. Jorin P (2001), Value at Risk: The New Benchmark of Managing Financial Risk , McGraw-Hill, New York.

7. Kannan G and Aulbur W G (2004), “Intellectual Capital: Measurement Effectiveness”, Journal of Intellectual Capital , Vol. 5, No. 3, pp. 389-413.

8. Kupiec P H (1995), “Techniques for Verifying the Accuracy of Risk Measurement Models”, The Journal of Derivatives , Vol. 3, No. 2.

9. Lambadiaris G, Papadopoulou L, Skiadopoulos G and Zoulis Y (2003), “VAR: History or Simulation”, Risk , Vol. 16, No. 9, pp. 122-127.

10. Linsmeier T J and Pearson N D (1996), “Risk Measurement: An Introduction to Value at Risk”, Office for Futures and Options Research Working Paper No. 96-04, University of Illinois.

11. Markowitz H (1952), “Portfolio Selection”, The Journal of Finance , Vol. 7, No. 1, pp. 77-91.

12. Mentel G (2013), “Parametric or Non-Parametric Estimation of Value-at-Risk”, International Journal of Business and Management, Vol. 8, No. 11, p. 103.

13. Patra B and Padhi P (2015), “Backtesting of Value at Risk Methodology: Analysis of Banking Shares in India”, Margin: The Journal of Applied Economic Research , Vol. 9, No. 3, pp. 254-277.

14. Pritsker M (1997), “Evaluating Value at Risk Methodologies: Accuracy Versus C o mp u t a t i o n a l Ti me ” , J o u r n a l o f F i n a n c i a l S e r v i c e s R e s e a rc h , Vo l . 1 2 , Nos. 2-3, pp. 201-242.

15. Rejeb A B, Salha O B and Rejeb J B (2012), “Value-at-Risk Analysis for the Tunisian Currency Market: A Comparative Study”, International Journal of Economics and Financial Issues, Vol. 2, No. 2, p. 110.

16. Røynstrand T, Nordbø N P and Strat V K (2012), “Evaluating Power of Value-at-Risk Backtests”, Master’s Thesis, Institutt for industriell okonomi og teknologiledelse.

17. Sarma M, Thomas S and Shah A (2003), “Selection of Value-at-Risk Models”, Journal of Forecasting , Vol. 22, No. 4, pp. 337-358. 57 Equity Risk Exposure: A Case of Indian Banking Industry 18.

Van den Goorbergh R W and Vlaar P J (1999), “Value-at-Risk Analysis of Stock Returns Historical Simulation, Variance Techniques or Tail Index Estimation?”, De Nederlandsche Bank, NV.

19. Varma J R (1999), “Value at Risk Models in the Indian Stock Market”, Indian Institute of Management, Ahmedabad.

20. Woller J (1996), “The Basics of Monte Carlo Simulations”, Physical Chemistry Lab , Spring, University of Nebraska-Lincoln.

21. Yang N (2010), “Empirical Study of Value at Risk of Chinese Stock Market”, M.Sc.

Business Economics, Finance Track, Universiteit Van Amsterdam. Reference # 05J-2018-01-03-01 Copyright ofIUP Journal ofApplied Economics isthe property ofIUP Publications andits content maynotbecopied oremailed tomultiple sitesorposted toalistserv without the copyright holder'sexpresswrittenpermission. However,usersmayprint, download, oremail articles forindividual use.