CHARACTERISTICS AND PERFORMANCE EVALUATION OF SELECTED MUTUAL FUNDS IN INDIA

Transcription

1 CHARACTERISTICS AND PERFORMANCE EVALUATION OF SELECTED MUTUAL FUNDS IN INDIA Sharad Panwar and Dr. R. Madhumathi Indian Institute of Technology, Madras ABSTRACT The study used samle of ublic-sector sonsored & rivate-sector sonsored mutual funds of varied net assets to investigate the differences in characteristics of assets held, ortfolio diversification, and variable effects of diversification on investment erformance for the eriod May, 2002 to May,2005. The study found that ublic-sector sonsored funds do not differ significantly from rivate-sector sonsored funds in terms of mean returns%. However, there is a significant difference between ublic-sector sonsored mutual funds and rivate-sector sonsored mutual funds in terms of average standard deviation, average variance and average coefficient of variation(cov).the study also found that there is a statistical difference between sonsorshi classes in terms of e SDAR(excess standard deviation adjusted returns)as a erformance measure. When residual variance (RV) is used as the measure of mutual fund ortfolio diversification characteristic, there is a statistical difference between ublic-sector sonsored mutual funds and rivate-sector sonsored mutual funds for the study eriod. The model built on testing the imact of diversification on fund erformance and found a statistical difference among sonsorshi classes when residual variance is used as a measure of ortfolio diversification and excess standard deviation adjusted returns as a erformance measure. RV,however, has a direct imact on Share fund erformance measure. Keywords: Mutual Funds, erformance evaluation, risk-return analysis, Net asset value, residual variance, fund return

2 1. INTRODUCTION Mutual Funds is a toic which is of enormous interest not only to researchers all over the world, but also to investors. Mutual funds as a medium-to-long term investment otion is referred as a suitable investment otion by investors. However, with several market entrants the question is the choice of mutual fund. The study focuses on this roblem of mutual fund selection by investors. Though the investment objectives define investors reference among fund tyes (balanced, growth, dividend etc.) the choice of fund based on a sonsor s reutation remains to be robed. Indian mutual fund industry has two distinct tyes of sonsors, ublic-sector and rivate-sector. The number of funds floated by ublicsector sonsors are minimal comared to rivate-sector layers. There is a hyothetical assumtion that rivate-sector outerforms ublic-sector due to several factors such as resonsibility, commitment and so on. We focus on testing this hyothesis on the mutual fund industry. Although many studies document the investment erformance of mutual funds irresective of whether they are ublic-sector sonsored or rivate-sector sonsored, researchers do not investigate the influence of ortfolio characteristics and the variable effect of diversification on mutual fund erformance. 2. OBJECTIVES OF THE STUDY To identify differences in characteristics of ublic-sector sonsored & rivate-sector sonsored mutual funds To find the extent of diversification in the ortfolio of securities of ublic-sector sonsored and rivate-sector sonsored mutual funds To comare the erformance of ublic-sector sonsored and rivate-sector sonsored mutual funds using traditional investment measures 3. LITERATURE REVIEW Literature on mutual fund erformance evaluation is enormous. A few research studies that have influenced the rearation of this aer substantially are discussed in this section. Share, William F. (1966) suggested a measure for the evaluation of ortfolio erformance. Drawing on results obtained in the field of ortfolio analysis, economist Jack L. Treynor has suggested a new redictor of mutual fund erformance, one that differs from virtually all those used reviously by incororating the volatility of a fund's return in a simle yet meaningful manner. Michael C. Jensen (1967) derived a risk-adjusted measure of ortfolio erformance (Jensen s alha) that estimates how much a manager s forecasting ability contributes to fund s returns. As indicated by Statman (2000), the e SDAR of a fund ortfolio is the excess return of the ortfolio over the return of the benchmark index, where the ortfolio is leveraged to have the benchmark index s standard deviation. S.Narayan Rao, et. al., evaluated erformance of Indian mutual funds in a bear market through relative erformance index, risk-return analysis, Treynor s ratio, Share s ratio, Share s measure, Jensen s measure, and Fama s measure. The study used 269 oen-ended schemes (out of total schemes of 433) for comuting relative erformance index. Then after excluding funds whose returns are less than risk-free returns, 58 schemes are finally used for further analysis. The results of erformance measures suggest that most of mutual fund schemes in the samle of 58 were able to satisfy investor s exectations by giving excess returns over exected returns based on both remium for systematic risk and total risk. Bijan Roy, et. al., conducted an emirical study on conditional erformance of Indian mutual funds. This aer uses a technique called conditional erformance evaluation on a samle of eighty-nine Indian mutual fund schemes.this aer measures the erformance of various mutual funds with both unconditional and conditional form of CAPM, Treynor- Mazuy model and Henriksson-Merton model. The effect of incororating lagged information variables into the evaluation of mutual fund managers erformance is examined in the Indian context. The results suggest that the use of conditioning lagged information variables imroves the erformance of mutual fund schemes, causing alhas to shift towards right and reducing the number of negative timing coefficients. Mishra, et al., (2002) measured mutual fund erformance using lower artial moment. In this aer, measures of evaluating ortfolio erformance based on lower artial moment are develoed. Risk from the lower artial moment is measured by taking into account only those states in which return is below a re-secified target rate like risk-free rate. Kshama Fernandes(2003) evaluated index fund imlementation in India. In this aer, tracking error of index funds in India is measured.the consistency and level of tracking errors obtained by some well-run index fund suggests that it is ossible to attain low levels of tracking error under Indian conditions. At the same time, there do seem to be eriods where certain index funds aear to deart from the disciline of indexation. K. Pendaraki et al.

3 studied construction of mutual fund ortfolios, develoed a multi-criteria methodology and alied it to the Greek market of equity mutual funds. The methodology is based on the combination of discrete and continuous multi-criteria decision aid methods for mutual fund selection and comosition. UTADIS multi-criteria decision aid method is emloyed in order to develo mutual fund s erformance models. Goal rogramming model is emloyed to determine roortion of selected mutual funds in the final ortfolios. Zakri Y.Bello (2005) matched a samle of socially resonsible stock mutual funds matched to randomly selected conventional funds of similar net assets to investigate differences in characteristics of assets held, degree of ortfolio diversification and variable effects of diversification on investment erformance. The study found that socially resonsible funds do not differ significantly from conventional funds in terms of any of these attributes. Moreover, the effect of diversification on investment erformance is not different between the two grous. Both grous undererformed the Domini 400 Social Index and S & P 500 during the study eriod. 4. METHODOLOGY USED These three traditional measures of investment erformance are used to comare the ublic-sector sonsored & rivate-sector sonsored mutual funds. These are Jensen s alha, ; Share information ratio, S; and excess standard deviation adjusted return, e SDAR. Jensen s alha relies on beta as a measure of the risk of a mutual fund ortfolio whereas the two other erformance measures rely on total variability of returns. Jensen s alha is estimated as: r = + β r.(1) t mt Where r t ortfolio in month t, is the excess return (i.e., the observed return minus the risk-free rate) on mutual fund r mt is the excess return on the benchmark index in month t(i.e., the observed return on the benchmark index minus the risk-free rate), β is mutual fund ortfolio s beta. The riskfree rate is reresented by the monthly return on three-month(91-day) Treasury bills. If Dt is the difference in return between the fund ortfolio and the benchmark ( Rt Rmt ) in eriod t, then Share information ratio is comuted as: S D / σ D Where =..(2) D benchmark, ( is the average value of the monthly differences in return between the fund ortfolio and the n D t )/n, and D t = 1 σ is the standard deviation of the differential return. As with Jensen s alha, this measure indicates ortfolio erformance relative to the benchmark ortfolio and lends itself to statistical tests of significance. However, unlike the Jensen s alha, the Share erformance measure adjusts for total risk rather than just systematic risk. The third measure of investment erformance is e SDAR (Statman 2000), measured as follows: esdar = rf + ( r rf)*( Sm/ S) r m (3) where r f =monthly return on three-month Treasury bills, r r m S S m =monthly return on fund ortfolio, =monthly return on the benchmark index, =standard deviation of ortfolio s return, and = standard deviation of return on the benchmark index. Residual variance is also used to evaluate mutual fund erformance. It is also called unexlained variance. In general, it is known as the variance of any residual. In articular, it is the variance 2 (y - Y) of the difference between any variate y and its regression function Y. Residual variance tends to decrease as the number of shares held by the mutual fund increases. Therefore, the higher the residual variance, the less diversified the mutual fund is.

4 Mutual fund ortfolio s residual variance, normalized by the total variance of the fund ortfolio (or RV) is estimated as: RV = 1 ( β σm/ σ )...(4) This estimated residual variance is used to comare the levels of unsystematic risk in ortfolios of ublic-sector sonsored and rivate-sector sonsored mutual funds. 5. TESTABLE HYPOTHESES Mutual funds based on the sonsors have been differentiated into three classes. The funds were initially categorized as ublic-sector sonsored and rivate-sector sonsored funds. Private-sector sonsored funds were further subdivided into Indian and foreign sonsors. Mutual funds could be defined in terms of the following characteristics: net assets, stock%, holdings, to ten%, market caitalization where stock% is common stock investments as ercentage of the fund s assets, holdings is the total number of comanies held by the fund, to ten % is the ercentage of net assets invested in the fund s to ten holdings which is calculated by summing ercentage of net assets in to ten holdings. Ca is the median market caitalization of the comanies/securities held by the fund. Median market caitalization is calculated from the sorted market caitalization of the comanies held by the fund. Net assets and market caitalization are in crores of Indian Ruees. The fund characteristics that can be used to measure ortfolio diversification are caitalization, holdings and to ten %. Besides, residual variance is also an imortant measure of fund diversification. Number of comanies held by the mutual fund(holdings) and the ercentage of assets in to ten holdings can rove to be very useful in gaining insight into mutual fund ortfolio diversification because when the number of comanies held by the mutual fund is lower or the ercentage of assets invested in the to ten holdings is higher, the mutual fund is more concentrated in a few comanies and the mutual fund is more suscetible to market fluctuations in these holdings. To investigate the diversification roerties and investment erformance of selected mutual funds, as well as the effect of diversification on investment erformance, both arametric and nonarametric statistical methods like Wilcoxon-two samle rank sum test, k-samle Jonckheere- Terstra (J-T)test, correlation and analysis of variance were used. The statistical testing of difference in fund classes are through the non-arametric J-T Test. The Jonckheere-Terstra test is a nonarametric test for ordered differences among classes. It tests the null hyothesis that the distribution of the resonse variable does not differ among classes. It is designed to detect alternatives of ordered class differences, which can be exressed as (or ), with at least one of the inequalities being strict, where denotes the effect of class i. For such ordered alternatives, the Jonckheere-Terstra test can be referable to tests of more general class difference alternatives, such as the Kruskal - Wallis test.the Jonckheere-Terstra test statistic is comuted by first forming R(R-1)/2 Mann-Whitney counts M i,i ', where i < i', for airs of rows in the contingency table,.(5) where X i,j is resonse j in row i. Then the Jonckheere-Terstra test statistic is comuted as.(6) This test rejects the null hyothesis of no difference among classes for large values of J. Asymtotic - values for the Jonckheere-Terstra test are obtained by using the normal aroximation for the distribution of the standardized test statistic.

5 The standardized test statistic is comuted as..(7) where E (J) 0 and var (J) 0 are the exected value and variance of the test statistic under the null hyothesis....(8) var 0 (J) = A / 72 + B / [ 36n(n-1)(n-2) ] + C / [ 8n(n-1) ]...(9) where..(10)..(11)..(12) When the standardized test statistic is greater than its null hyothesis exected value of zero, the rightsided -value is comuted, which is the robability of a larger value of the statistic occurring under the null hyothesis. A small right-sided -value suorts the alternative hyothesis of increasing order from row 1 to row R. When the standardized test statistic is less than or equal to zero, the left-sided -value is comuted. A small left-sided -value suorts the alternative of decreasing order from row 1 to row R. The one-sided -value P 1 can be exressed as The two-sided -value P 2 is comuted as

6 Following Null hyothesis is tested: H : 0 0 µ Pu RV µ Px RV =.. (13) Where RV is the Mutual fund s residual variance, also referred to as idiosyncratic or comany-secific variance; Pu is ublic-sector sonsored mutual fund; Px is rivate-sector sonsored mutual fund which includes both rivate-sector Indian sonsored, PvI and rivate-sector foreign sonsored mutual funds,pvf. That is, the degree of diversification of a ublic-sector sonsored mutual fund ortfolio is not different from that of a rivate-sector sonsored mutual fund ortfolio. The null hyothesis is tested against the alternative: H : µ µ > 0..(14) a Pu RV Px RV It is redicted that a significant correlation will exist between residual variance and measures of investment erformance. H : 0 0 µ Pu P µ Px P =..(15) Where P is the relevant measure of investment erformance. That is, the investment erformance of a ublic-sector sonsored mutual fund is not different from that of a rivate-sector sonsored mutual fund. It is exected that the investment erformance of ublic-sector sonsored mutual funds to be worse than that of rivate-sector sonsored mutual funds. Thus, the alternative hyothesis is H a: µ Pu P µ Px P< 0..(16) The investment erformance of the ublic-sector sonsored mutual funds and rivate-sector sonsored mutual funds, is comared using two alternative statistical methods. Using the first method, taking two samles at a time (ublic-sector sonsored and rivate-sector Indian sonsored mutual funds) and secondly,(ublic-sector and rivate-sector foreign sonsored mutual funds) and finally rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds, are comared using the Wilcoxon two-samle rank-sum test. Using the second method, the two samles are comared again using the analysis of covariance to investigate the differential imact of residual variance on investment erformance between the three classes of mutual funds. Analysis of covariance combines the characteristics of both analysis of variance and regression. This statistical method allows to test whether the means of ortfolio erformance measures are significantly different between ublic-sector sonsored, rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds and, simultaneously, to test whether differences in investment erformance are attributable to the difference in residual variance. The model that links investment erformance to the two indeendent variables is: y = ω + ω x + ω x + ω x x +ε (17) where y = investment erformance of the fund ortfolio; x 1 = residual variance (the covariate); and x 2 = grou(a=public-sector sonsored mutual funds, B=Private-sector Indian sonsored mutual funds and C=Private-sector foreign sonsored mutual funds ), where x 2 =1 if grou A, 2 if grou B and 3 if grou C. Because x 2 is equal to one, two and three if the mutual fund is ublic-sector sonsored, rivate-sector Indian sonsored, rivate-sector foreign sonsored mutual fund,resectively, then the exected values of investment erformance are E( y) = ( ω 0 + ω2) + ( ω1 + ω3) x 1..(18) for the ublic-sector sonsored funds, E( y) = ( ω0 + 3 ω2) + ( ω1 + 3 ω3) x 1..(19) for the rivate-sector Indian sonsored funds, and Ey ( ) = ( ω0+ 3 ω2) + ( ω1+ 3 ω3) x 1.(20)

7 Therefore, ω 2 is the difference in intercets and ω 3 is the difference between the sloes of the two analysis of covariance models. It can be tested whether there is a difference in the effect of residual variance on investment erformance as a function of grou by testing the hyothesis: H : ω = (Kleinbaum, Kuer, and Muller 1988). The difference in investment erformance between ublic-sector sonsored,rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds can be tested, after adjusting for the effects of residual variance, by testing the hyothesis: H : ω = 0. For rivate-sector Indian and foreign 0 2 sonsored mutual funds, a linear relation can be tested between residual variance and investment erformance by testing the null hyothesis : H 0 : ω 1 = DATA COLLECTION Net Asset Value(NAV) for the medium-term eriod May,2002 to May,2005 of selected mutual funds along with the index value of the two benchmark market indices, namely S &P CNX NIFTY and CRISIL Balanced Fund Index are taken from the following website: (1) S &P CNX NIFTY: htt:// (2) CRISIL Balanced Fund Index : htt:// (3) Net Asset Value(NAV): htt:// This website gave information on mutual fund s characteristics like net assets, stock %, holdings, to ten %, Market caitalization. The samle had six ublic-sector sonsored mutual funds oerating in India. These have been matched with twelve randomly selected rivate-sector sonsored mutual funds, of which seven were rivate-sector Indian sonsored and five rivate sector foreign sonsored mutual funds oerating in India. Indices: Following two are taken: S & P CNX NIFTY Index: It is a market index and is used by funds to benchmark their fund erformance. It is a well diversified 50 stock index accounting for 23 sectors of the economy. It is used for a variety of uroses such as benchmarking fund ortfolios, index based derivatives and index funds. It is owned and managed by India Index Services and Products Ltd. (IISL), which is a joint venture between NSE and CRISIL. Imact cost of the S&P CNX Nifty for a ortfolio size of Rs.5 million is 0.07%.It is rofessionally maintained and is ideal for derivatives trading CRISIL Balanced Fund Index: consists of tracking the returns on the constituents like the CRISIL Comosite Bond Fund Index and the Nifty Index. The Weighted Average Methodology is used to arrive at the returns for the Balanced Fund. The Index History is calculated from the base date of 31 st March, An index of this kind, generally serves as an indicator for all the market articiants in the category, to benchmark their erformance against the index, find out the attributes for the variation in their erformance vis-a-vis the index and reshuffle their ortfolio keeing in mind the risk/reward tradeoff. Since the resulting Index is a derived Index rather than a Primary Index, it also serves as a benchmark for non-diversified market articiants to evaluate their erformance against a diversified ortfolio containing a mix of all the instruments in the universe of non- equity instruments. Finally, it is a useful tool to track volatility, charting correlation and develoing hedging instruments. Daily Net Asset Values (NAV s) are obtained for each of these eighteen funds and also for each of the two indices taken. The returns are comuted using formula : return = ( NAV NAV )/ NAV (21) t t 1 t 1 NAVt NAVt 1 Where is Net asset value of a mutual fund or Index for a day t, is Net asset value of a mutual fund or Index for day (t-1). Returns on each of these eighteen mutual funds and also for each of the two indices is given in the following table. For the S & P Index, the returns are : return = ( Index Index )/ Index..(22) t t 1 t 1

8 TABLE 1-Mutual Funds return, Standard Deviation, Variance, Coefficient of Variation, CRISIL Balanced Fund Index, S & P CNX NIFTY Index return (May, 2002 to May, 2005) CRISIL Balanced Fund Return Fund Name Tye Fund Return Standard Deviation Variance COV canbank index- growth lan Pu GIC Balanced Fund Pu LIC G Sec Fund - Dividend Pu UTI Balanced Fund - Growth Pu SBI Magnum Balanced Fund - Dividend Pu Escorts Balanced Fund - Dividend Pu Prudential ICICI Balanced - Growth PvI Birla Bond Index Fund - Growth PvI Chola Growth Fund -Growth PvI ING Vysya Liquid Fund - Growth PvI Reliance Growth - Growth PvI Sahara Taxgain - Growth PvI Sundaram Money Fund - Growth PvI JM MIP FUND-Monthly Dividend PvF Franklin FMCG Fund - Dividend PvF DSP Merrill Lynch Balanced Fund - Dividend PvF ABN AMRO Cash Fund - Growth PvF Grindlays Cash Fund - Growth PvF S & P CNX Nifty Return Note: Table 1 shows the six ublic-sector sonsored, seven rivate-sector Indian sonsored and five rivate-sector foreign sonsored mutual funds labeled as Pu, PvI, PvF resectively. Fund returns, Standard deviation of returns, CRISIL Balanced Fund Index returns, S & P CNX Nifty Index returns are also calculated using the formula deicted above. The square of standard deviation of returns gives the Variance. Coefficient of variation (COV) is found by dividing standard deviation by mean returns.

9 TABLE 2- Characteristics of Mutual Funds during May,2005 Fund Name Net Assets(in Rs crores)stock % Holdings(in numerals)to ten%market ca(in Rs crores) Canbank Index- Growth Plan GIC Balanced Fund LIC G Sec Fund - Dividend UTI Balanced Fund - Growth SBI Magnum Balanced Fund - Dividend Escorts Balanced Fund - Dividend Prudential ICICI Balanced - Growth Birla Bond Index Fund - Growth Chola Growth Fund -Growth ING Vysya Liquid Fund - Growth Reliance Growth - Growth Sahara Taxgain - Growth Sundaram Money Fund - Growth JM Auto Sector Fund - Dividend Franklin FMCG Fund - Dividend DSP Merrill Lynch Balanced Fund - Dividend ABN AMRO Cash Fund - Growth Grindlays Cash Fund - Growth Note: Net Assets is the mutual fund size,stock% is common stock investments as ercentage of the fund s assets, holdings is the total number of comanies held by the fund, to ten % is the ercentage of net assets invested in the fund s to ten holdings, and market ca is the median market caitalization of the comanies held by the fund.

10 TABLE 3-Fund return and characteristics differentiation among mutual fund classes (May,2002 to May,2005) Standard J-T samle mean statistic Public (Pu)Private (PvI)Private (PvF) Pu PvI - PvF sig. PANEL A : MONTHLY RETURN Mean return % Avg. standard deviation ** Avg. variance * Avg. COV ** PANEL B :PORTFOLIO CHARACTERISTICS Net Assets(Rs. Crores) Common Stock % Market Caitalization(Rs.Crores) Holdings(Numerals) To Ten% J-T is Jonckheere-Terstra test Note: COV is coefficient of variation. The Z-scores are from the three-samle Jonckheere -Terstra test. * * Jonckheere-Terstra test Z-score is statistically significant at the 5 % level. * Jonckheere-Terstra test Z-score is statistically significant at the 10 % level. 7. RESULTS Mutual Fund Portfolio Performance There is a wide variation among the samle funds ranging from a minimum net asset of Rs.2.08 crores to a maximum of Rs.3, crores. The funds selected have a mix of debt, equity and a combination of debt and equity. The holdings also varied from 2 to 61. To Ten % indicating nondiversification to the extent of % was reresented in the samle along with a diversified fund reresented by only 14.7 % held by to ten securities. The market ca also varied from Rs crores to Rs crores. Table 2 shows various ortfolio characteristics of ublic-sector sonsored funds, rivatesector Indian sonsored and rivate-sector foreign sonsored funds. Table 3 summarizes the results of a statistical test of the differences between the classes of funds. Panel A of Table 3 shows that the average return for the ublic-sector sonsored funds is 0.07 %, comared with 0.28% for the rivate-sector Indian sonsored funds, 0.05% for the rivate-sector foreign sonsored mutual funds for the study eriod. The three-samle Jonckheere-Terstra test indicates that the mean returns are not significantly different at the 5% level. However, the standard deviations of the 3-year returns are significant, as indicated by a Z-score(significance=0.047) from the three-samle Jonckheere-Terstra test. The variance (Zscore =-1.848, sig.=0.065) and coefficient of variation(cov) are also significant (Z-score= ,sig.=0.021),indicating that though in terms of mean returns, there is no statistical difference between sonsored classes though there is a statistical difference in terms of fund risk over the study eriod. Also, considering ortfolio characteristics like net assets, common stock %, caitalization, holdings, to ten% ; using the three-samle Jonckheere-Terstra test, it is found that since the Z-score of all of these characteristics is negative and non-significant at 5% level of significance, hence, there is no statistical difference between ublic-sector sonsored and rivate-sector sonsored mutual funds in terms of ortfolio characteristics. Table 4 shows the estimated traditional measures of investment erformance of the mutual funds like Jensen s alha, ortfolio beta which are estimated using the two alternative benchmark indices, S & P CNX NIFTY Index and CRISIL Balanced Fund Index. It can be seen that S & P CNX NIFTY Index is a better measure comared to CRISIL Balanced Fund Index. S & P CNX NIFTY Index is able to give more statistical significance in terms of ortfolio alha and beta as comared to CRISIL Balanced Fund Index. Generally, S & P CNX NIFTY Index is used as the benchmark index by the Indian mutual fund managers to evaluate the erformance of a mutual fund oerating in India.

11 When CRISIL Balanced Fund Index is used as the benchmark index, for ublic-sector sonsored mutual funds, the Jensen s alha is significant and negative for two out of six funds, i.e., % of the samle funds in this class(undererformers). However, for Indian rivate-sector sonsored funds, Jensen s alha is significant and ositive for two out of seven funds, i.e., % of samle(overerformers), and also significant and negative for three out of seven funds, i.e.,42.86 % of the samle(undererformers). Private-sector foreign sonsored mutual funds have statistical under erformance for three out of five funds,i.e.,60 % of the samle and over erformance for one fund, i.e., 20 % of the samle. When S & P CNX NIFTY Index is used as the benchmark index, then for ublic-sector sonsored mutual funds, Jensen s alha is significant and negative(undererformance) for five out of six funds, i.e.,83.33% of the samle at 5 % level of significance. However, Jensen s alha is significant and ositive(overerformance) for one out of six funds, i.e.,16.67 % of the samle.for rivate -sector Indian sonsored mutual funds, Jensen s alha is significant and negative (undererformance) for four out of seven funds,i.e.,57.14 % of the samle. For rivate-sector foreign sonsored mutual funds, the entire samle has shown statistical undererformance in terms of Jensen s alha. From Table 5, it is found that in terms of e SDAR (excess standard deviation adjusted return), the three classes of funds are statistically different. Also, rivate-sector Indian sonsored and rivatesector foreign sonsored mutual funds are statistically different in terms of e SDAR (excess standard deviation adjusted return) ( Two samle Wilcoxon rank-sum test Z-score= 0.028,significance=0.03)at 5% level. In terms of ortfolio diversification (residual variance, RV), ublic-sector sonsored and rivate-sector Indian sonsored mutual funds are statistically different (Two samle Wilcoxon rank-sum test Z-score= , significance=0.063) at 10% level. It is concluded that during May,2002 to May,2005 eriod, neither the investment erformance reresented by Jensen s alha, Share information ratio, nor the level of mutual fund ortfolio diversification(ortfolio beta) of ublic-sector sonsored mutual funds differs from those of rivatesector Indian and rivate-sector foreign sonsored mutual funds of varied fund characteristics. However, in terms of e SDAR, there is a statistical difference between ublic-sector sonsored, rivate-sector Indian and rivate-sector foreign-sonsored mutual funds of varied fund characteristics. In terms of ortfolio diversification(residual variance, RV), it is found that there is a statistical difference between ublic-sector sonsored and rivate-sector Indian sonsored mutual funds when S & P CNX NIFTY Index is used as a benchmark index. Results of Pearson correlation between investment erformance and measures of mutual fund ortfolio diversification show that when CRISIL Balanced Fund Index is used as the benchmark index, then for the ublic-sector sonsored funds, the erformance measure e SDAR alone has a significant correlation with caitalization. In the rivate- sector sonsored funds( both Indian and foreign), Jensen s alha has a significant correlation with holdings while Share s measure has a significant correlation with RV. For rivate-sector foreign sonsored funds, Share information ratio has a significant correlation with residual variance. The rivate-sector Indian sonsored funds show significant correlation between holdings with both Share information ratio and e SDAR as well as to ten % with both Share information ratio and e SDAR. However, to ten % has a negative correlation with holdings imlying that the diversification erformance is oor. When S & P NIFTY Index is used as the benchmark index, for ublic-sector sonsored mutual funds, e SDAR is significantly correlated with caitalization, holdings, to ten %. Holdings is correlated with residual variance (RV), caitalization and to ten. For rivate-sector sonsored funds, Jensen s alha is correlated with residual variance. To ten % is correlated with holdings. For rivate-sector Indian sonsored funds, Jensen s alha is correlated with residual variance. Share information ratio is correlated with to ten %. Share information ratio, e SDAR and to ten % are correlated with holdings. For rivate-sector foreign sonsored funds, Jensen s alha and Share information ratio are correlated with residual variance. Caitalization and holdings are correlated with to ten %. For combined samle, to ten% is correlated with holdings. So, wherever Pearson correlation is negative, it imlies that diversification erformance is oor. Conversely, wherever Pearson correlation is ositive, it imlies that diversification erformance is good.

12 TABLE 4-Performance of Mutual Funds: May 2002 to May 2005 Fund Name CRISIL Balanced Fund Index S & P CNX NIFTY Index β sig β sig Share,S e SDAR sig sig Share,S e SDAR Canbank Index-Growth Plan *** *** *** *** GIC Balanced Fund *** *** *** 1.21E LIC G Sec Fund - Dividend *** ** *** ** UTI Balanced Fund - Growth *** *** *** SBI Magnum Balanced Fund - Dividend *** ** *** Escorts Balanced Fund - Dividend ** *** ** *** Prudential ICICI Balanced - Growth ** *** *** *** Birla Bond Index Fund - Growth *** ** *** * Chola Growth Fund -Growth ** *** E *** ING Vysya Liquid Fund - Growth *** *** Reliance Growth - Growth Sahara Taxgain - Growth Sundaram Money Fund - Growth *** *** * JM MIP Fund-Monthly Dividend *** *** -2E Franklin FMCG Fund - Dividend *** *** *** DSP Merrill Lynch Balanced Fund - Dividend *** *** *** ABN AMRO Cash Fund - Growth *** *** Grindlays Cash Fund - Growth *** *** * Table 4: Note: Jensen s alha,, is the measure of mutual fund ortfolio s investment erformance, using the CRISIL Balanced Fund Index and S & P CNX NIFTY Index as the two benchmark indices. β is mutual fund ortfolio s beta. S is Share information measure. e SDAR is excess standard deviation adjusted return. Statistical significance relative to and β is indicated in this table. Jensen s alha and ortfolio beta are significant for Bold entries. *** significance at 1 % level ** significance at 5 % level *significance at 10 % level

13 TABLE 5-Differences in Performance of Mutual Funds: May,2002 to May,2005 samle mean Pu Pv PvI PvF J-T Test(Pu-PvI-PvF) sig(2-tailed) Wilcoxon Z-score(Pu-PvI) sig(2-tailed) Performance relative to CRISIL Balanced Fund Index Wilcoxon Z-score(Pu-PvF) sig(2-tailed) Wilcoxon Z-score(PuI-PvF) sig(2-tailed) S esdar ** ** * β RV Wilcoxon Wilcoxon Wilcoxon Pu Pv PvI PvF J-T Test(Pu-PvI-PvF) sig(2-tailed) Z-score(Pu-PvI) sig(2-tailed) Z-score(Pu-PvF) sig(2-tailed) Z-score(PvI-PvF) sig(2-tailed) Performance relative to S & P CNX NIFTY Index S esdar * β RV * Note: J-T stands for Jonckheere-Terstra test ** significant at 5 % level * significant at 10% level

14 TABLE 6 Relationshi between Investment Performance and Portfolio Diversification CRISIL Balanced Fund Index S & P CNX NIFTY Index Measure of Diversification Share, S esdar Share, S esdar Public-sector sonsored funds RV significance Caitalization significance *** ** Holdings significance * To ten% significance * Private-sector sonsored Funds Share, S esdar Share, S esdar RV significance ** *** Caitalization significance Holdings significance 0.094* To ten% significance Private-sector Indian sonsored Funds Share, S esdar Share, S esdar RV significance ** Caitalization significance Holdings significance *** 0.022** *** 0.043** To ten% significance ** 0.088* ** Private-sector Foreign sonsored Funds Share, S esdar Share, S esdar RV significance ** *** 0.019** Caitalization significance Holdings significance To ten% significance Combined Samle Share, S esdar Share, S esdar RV significance Caitalization significance Holdings significance To ten% significance Note: *** Significant at 1 % level, ** Significant at 5 % level, * Significant at 10 % level

15 Source Panel A: Deendent Variable: CRISIL Balanced Fund Index TABLE 7-Covariance Analysis Degrees of a freedom F-value Prob.>F Source Panel A: Deendent Variable: S & P CNX NIFTY Index Degrees of freedom F-value Prob>F RV RV Grou Grou RV RV Grou Grou RV * Grou RV * Grou Panel C: Deendent Variable: Panel C: Deendent Variable: RV * RV * Grou Grou RV * Grou RV * Grou Panel D: Deendent Variable: e SDAR Panel D: Deendent Variable: e SDAR RV RV Grou Grou ** RV * Grou RV * Grou ** Note: This table shows the results of covariance analysis using a general linear model. The resonse variable is investment erformance measured by Jensen s alha, ;Share erformance measure, S ; or excess standard deviation adjusted return, e SDAR, all calculated using CRISIL Balanced Fund Index and S & P CNX NIFTY Index as the benchmark indices. Residual variance, RV, is the covariate. Grou is equal to 1 if ublic-sector sonsored mutual fund, 2 if rivate-sector Indian sonsored mutual fund, 3 if rivate-sector foreign sonsored mutual fund. The RV * Grou interaction term indicates difference in the relation between residual variance and erformance as a function of grou. a RV * Grou RV * Grou Panel B: Deendent Variable: Panel B: Deendent Variable: β β Tye III sum of squares (SS) for Panel B, and Tye I SS for Panels A, C, and D. ** Significant at the 5 % level * Significant at the 10 % level S SS S a SS

16 Residual variance has a significant correlation for rivate-sector sonsored mutual funds. The effect of diversification on investment erformance have fund as control grou can be tested through analysis of covariance. In Table 7, the effect of diversification on investment erformance is H : ω = 0 tested as a function of fund sonsorshi class using 0 3. The Hyothesis of the relation between fund sonsorshi class and investment erformance is tested using H0 : ω 2 = 0.For CRISIL Balanced Fund Index, when Jensen s alha is used as a measure of investment erformance, the RV * Grou interaction term shown in Panel A of Table 7 indicates that there is no significant difference in the effect of diversification on investment erformance as a function of grou. Similarly, when e SDAR is used as measure of investment erformance, the RV * Grou interaction term does not indicate any significant effect of diversification as a function of grou. However, when Share information ratio, S, is used as a deendent variable in covariance analysis in Panel C, then residual variance is significant at 10 % level imlying that there is an imact of residual variance on Share information measure. However, there sums to be no sonsorshi class effect on fund erformance. However, RV as a function of fund sonsorshi class influences certain ortfolio erformance measures. Linking erformance fund diversification in terms of S &P CNX NIFTY Index, there is no statistical model of Jensen s alha and ortfolio beta. There is a statistical imact in terms of Share information ratio, S when RV is taken as covariate. When e SDAR is taken as deendent variable, in terms of grou and in terms of RV*Grou, there is a statistical imact at 5 % level of significance. To determine the relation between residual variance and investment erformance for rivate-sector H : ω = 0 is tested. Residual variance is not linearly related sonsored mutual funds, the Hyothesis 0 1 to investment erformance in terms of Jensen s alha and ortfolio beta, regardless of the benchmark index used. 8. FINDINGS OF THE STUDY There is no statistical difference between ublic-sector sonsored, rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds in terms of mean return ercentage. There is a statistical difference between ublic-sector sonsored, rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds in terms of average standard deviation, average variance and average coefficient of variation(cov) Public-sector sonsored mutual funds do not differ from rivate-sector Indian sonsored and rivatesector foreign sonsored mutual funds in terms of ortfolio characteristics like net assets (fund size in crores in Indian ruees), common stock %(common stock investments as ercentage of the fund s assets),to ten %(ercentage of net assets invested in fund s to ten holdings),market caitalization(median market caitalization of the comanies held by the fund), holdings (total number of securities held by the mutual fund). When considering CRISIL Balanced Fund Index as the benchmark index, for ublic-sector sonsored mutual funds, the Jensen s alha is significant and negative for two funds, significant and ositive for one fund. For rivate-sector sonsored mutual funds, Jensen s alha is significant and ositive for two rivate-sector Indian sonsored mutual funds, significant and negative for three rivate-sector Indian sonsored mutual funds, significant and negative for three rivate-sector foreign sonsored mutual funds. In terms of ublic-sector sonsored mutual funds, degree of ortfolio diversification (fund ortfolio beta) is significant and negative for one fund, significant and ositive for five funds. In terms of rivate-sector Indian sonsored mutual funds, ortfolio beta is significant and ositive for three funds and for two rivate-sector foreign sonsored funds, ortfolio beta is significant and ositive. When S & P CNX NIFTY Index is used as the benchmark index, then for ublic-sector sonsored mutual funds, Jensen s alha is significant and negative for five funds, significant and ositive for one fund. However, for rivate -sector Indian sonsored mutual funds, Jensen s alha is significant and negative for one fund, significant and negative for three funds. For rivate-sector foreign sonsored mutual funds, ortfolio beta is significant and negative for four funds, significant and negative for one fund. However, in terms of ublic-sector sonsored mutual funds, degree of ortfolio diversification (fund ortfolio beta) is significant and negative for one fund, significant and ositive for four funds. For rivate-sector Indian sonsored mutual funds, degree of ortfolio diversification (fund ortfolio beta) is significant and ositive for four funds. For rivate-sector foreign sonsored mutual funds, degree of ortfolio diversification (fund ortfolio beta) is significant and ositive for three funds.

17 Significant correlation is identified between fund erformance measures esecially excess standard deviation adjusted return (e SDAR), Jensen s alha and Share s information ratio. They vary among the various classes of mutual funds when Pearson correlation is used to analyze between investment erformance and measures of mutual fund ortfolio diversification using CRISIL Balanced Fund Index as the benchmark index, it is found that for ublic-sector sonsored mutual funds, residual variance (RV) is found to be correlated with holdings; market caitalization(ca) is found to be correlated with e SDAR(excess standard deviation adjusted return) and holdings(total number of comanies/securities held by the fund). For rivate-sector sonsored mutual funds, RV is found to be correlated to Share information ratio, S; market caitalization is found to be correlated with to ten%(ercentage of net assets invested in the fund s to ten holdings);holdings is found to be correlated with Jensen s alha and to ten%. For rivate-sector Indian sonsored mutual funds, holdings is found to be correlated with Share information ratio(s),e SDAR and To Ten %;To Ten % is found to be correlated with Share information ratio (S) and e SDAR. For rivate-sector foreign sonsored mutual funds, residual variance (RV) is found to be correlated with Share information ratio(s); market caitalization is correlated with To Ten %; holdings is correlated with To Ten %. For combined samle consisting of ublic-sector sonsored, rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds, holdings is correlated with To Ten %. When S & P CNX NIFTY Index is used as the benchmark index, then for ublic-sector sonsored mutual funds, RV is correlated with holdings; Caitalization is correlated with e SDAR and holdings; Holdings is correlated with e SDAR and To Ten %; To Ten % is correlated with e SDAR.For rivate-sector sonsored mutual funds, RV is correlated with Jensen s alha; Caitalization and Holdings are correlated with To Ten %. For rivate-sector Indian sonsored mutual funds, RV is correlated with Jensen s alha; Holdings is correlated with Share information ratio(s), e SDAR and To Ten %; To Ten % is correlated with Share information ratio, S; For rivate-sector foreign sonsored funds, RV is correlated with Jensen s alha and Share information ratio, S; Caitalization and Holdings are correlated with To Ten %; For combined samle consisting of six ublic-sector sonsored mutual funds, seven rivate-sector Indian sonsored and five rivate-sector foreign sonsored mutual funds; holdings is found to be correlated with To Ten %. Mutual fund erformance index measured in terms of excess standard deviation adjusted return has a statistically significant imact through diversification and residual variance, RV, as a function of fund sonsorshi class with the hel of following equation: X = WRV + ( * ) 1 WG + 2 W3 RV G. For CRISIL Balanced Fund Index, when Jensen s alha is used as a measure of investment erformance, the RV * Grou interaction term indicates that there is no significant difference in the effect of diversification on investment erformance as a function of grou. When e SDAR is used as the measure of investment erformance, the RV * Grou interaction term do not indicate any significant effect of diversification as a function of grou. When Share information ratio, S, is used as a deendent variable in covariance analysis, then residual variance is significant at 10 % level imlying that residual variance can be used by mutual fund managers for evaluating mutual fund erformance. When S & P CNX NIFTY Index is used as the benchmark index, then when Share information ratio, S, is taken as the deendent variable, it is found that residual variance is significant when RV is taken as covariate; when e SDAR is taken as deendent variable, Grou and RV*Grou are significant Residual variance is not linearly related to investment erformance in terms of Jensen s alha and ortfolio beta, regardless of the benchmark index used.

18 9. SUMMARY AND CONCLUSIONS The study found that ublic-sector sonsored, rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds do not differ statistically in terms of ortfolio characteristics such as net assets, common stock%, market caitalization, holdings, To Ten %. However, there is a statistical difference between three classes of ublic-sector sonsored, rivate-sector Indian sonsored and rivate-sector foreign sonsored mutual funds in terms of average standard deviation, average variance and average coefficient of variation. Portfolio risk characteristics measured through rivate-sector Indian sonsored mutual funds seems to have outerformed both Public- sector sonsored and Private-sector foreign sonsored mutual funds. Residual variance is not linearly related to investment erformance in terms of Jensen s alha and ortfolio beta, regardless of the benchmark index used. The general linear model of analysis of covariance establishes differences in erformance among the three classes of mutual funds in terms of ortfolio diversification. 10. SCOPE FOR FUTURE RESEARCH There is lot of scoe for imrovement in the research for evaluating mutual fund erformances. Various other multi-criteria decision models could be tested for evaluating mutual fund erformances. Testing of fund erformances in the long run can be done. Extended samle of ublic-sector sonsored, rivatesector Indian sonsored and rivate-sector foreign sonsored mutual funds can be taken for generating results. Portfolio risk through the measure of value at risk (VaR) can also be tested for differences in mutual fund classes. 11. REFERENCES & BIBLIOGRAPHY Journal Articles: Banikanta Mishra, Mahmud Rahman, Measuring mutual fund erformance using lower artial moment, Global Business Trends, Contemorary Readings, 2001 edition. Bijan Roy, Saiket Sovan Deb, The conditional erformance of Indian mutual funds: an emirical study, Working aer, htt://aers.ssrn.com/sol3/aers.cfm?abstract_id= Clow R., Money that grows on trees, Institutional Investor 33,1999, Diltz, J.D., The rivate cost of socially resonsible investing, Alied Financial Economics 5,1995, Goldreyer, E.F., P. Ahmed, and J.D. Diltz, The erformance of socially resonsible mutual funds: Incororating socioolitical information in ortfolio selection, Managerial Finance 25, 1999, Grossman,B.R. and W.F. Share, Financial imlications of South African divestment, Financial Analysts Journal 49, 1986, Kinder, P., S.D. Lydenberg, and A. L. Domini, Investing for Good (HarerCollins, New York),1993. Kleinbaum, D.G., L.L. Kuer, and K.E. Muller, Alied regression analysis and Other multivariate methods(pws-kent, Boston),1988. Kshama Fernandes, Evaluating index fund imlementation in India, Working aer, htt:// Michael C. Jensen, The Performance of Mutual Funds in the eriod , Journal of Finance, Vol. 23, No. 2,1967, Pendaraki, K., Zoounidis, C., Doumos, M., On the construction of mutual fund ortfolios: A Multicriteria methodology and an alication to the Greek market of equity mutual funds, Euroean Journal of Oerational Research, Jun2005, Vol. 163 Issue 2, Reilly, F.K. and E. A. Norton, Investments (South-Western, Mason, OH), Rudd, A., Social resonsibility and ortfolio erformance, California Management Review 23, 1981, Sauer, D.A., The imact of social-resonsibility screens on investment erformance: Evidence from the Domini 400 Social Index and Domini Equity Mutual Fund, Review of Financial Economics 6, 1997, Share, W.F., The Share ratio, Journal of Portfolio Management 21, 1994, S. Narayan Rao, M. Ravindran, Performance Evaluation of Indian Mutual Funds, Working aer, htt://aers.ssrn.com/sol3/aers.cfm?abstract_id= Statman, M., Socially resonsible mutual funds, Financial Analysts Journal 56,2000, Zakri Y. Bello, Socially Resonsible Investing and Portfolio Diversification, The Journal of Financial Research, Vol. XXVIII, No. 1,2005,41-57.

19 Books: H. Sadhak, Mutual Funds in India, Marketing Strategies and Investment Practices, Second Edition, Resonse Books, A division of Sage Publications India Pvt. Ltd, Internet: htt://ww w.amfiindia.com htt:// htt:// 12. AUTHOR PROFILE Sharad Panwar is B.Tech (Comuter Science & Information Technology), I.E.T, M.J.P. Rohilkhand University, Bareilly(U.P.), PGDIT(I.I.T Kharagur). He is currently doing M.S. (By research in Finance) from I.I.T. Madras, Chennai in Mutual Funds Performance Evaluation. He got 44 th Rank in Junior Mathematical Olymiad [JMO,DELHI 1995], conducted by National Board of Higher Mathematics, Deartment of Atomic Energy, Govt. of India. Dr. R. Madhumathi is Assistant Professor, Deartment of Management Studies, I.I.T. Madras, Chennai.