Persistence in performance of actively managed equity mutual funds: New Indian evidence

Accepted Manuscript

Persistence of actively managed equity mutual fund performance: new Indian Evidence Soumya G. Deb Associate Professor PII: DOI: Reference:

S0970-3896(19)30164-8 https://doi.org/10.1016/j.iimb.2019.03.014 IIMB 330

To appear in:

IIMB Management Review

Received date: Revised date: Accepted date:

11 August 2016 25 May 2017 27 March 2019

Please cite this article as: Soumya G. Deb Associate Professor , Persistence of actively managed equity mutual fund performance: new Indian Evidence, IIMB Management Review (2019), doi: https://doi.org/10.1016/j.iimb.2019.03.014

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Persistence of actively managed equity mutual fund performance: new Indian Evidence By

Associate Professor,

CR IP T

Soumya G. Deb1

Indian Institute of Management Sambalpur, Jyoti Bihar, Burla Sambalpur: 768019, India Phone: +91 99388 53926 Email: [email protected]

AN US

Short Abstract:

This paper examines the persistence of actively managed equity mutual funds after controlling for market risk, size, value, momentum and expenses, and whether such performance persistence depends on investor holding period, fund size, age, style or expense-ratio. We find overall

M

evidence of persistence over shorter time horizons. We also find that larger, older and highexpense-ratio funds are more persistent. However, some of that persistence is due to the relatively

ED

poor performers in the group. We affirm the robustness of our results using a number of

AC

CE

PT

robustness tests.

1

Contact Author

ACCEPTED MANUSCRIPT

Persistence of actively managed equity mutual fund performance: new Indian Evidence

Abstract

This paper examines the persistence in performance of actively managed Indian equity mutual

CR IP T

funds over short and long time horizons after controlling for systematic factors like market risk, size, value, momentum and expenses. The paper also explores whether such performance persistence depends on investor holding period, fund size, age, style or expense-ratio.

Using monthly returns of a sample of 263 actively managed Indian equity mutual funds which continuously existed between January 2000 and December 2014, we generate their 4-factor-alpha

AN US

(Carhart,1997) and adjust it for fund-expenses scaled by recent most inflation-adjusted assets under management(AUM) as net risk adjusted performance measure of funds. We then test the persistence in fund performance using non-parametric „contingency table’ and „quartiletransition- matrix‟ approaches. We find overall evidence of persistence, especially over shorter time horizons. This pattern is more or less uniform and independent of the calendar years in our

M

study period. We find that larger, older and high-expense-ratio funds are marginally more persistent, although no discernible pattern is identifiable within style partitions. However, a

ED

considerable part of persistence observed is due to the relatively poor performers in the group. Only exceptions are the older funds, which exhibit short as well as long term persistence, and this comes mostly from the winners rather than the losers. We affirm the robustness of our results

PT

using a number of robustness tests. The findings indicate that past performance of managed

CE

portfolios, at least in emerging market like India, can have some useful information for investors.

KEYWORDS: Mutual funds, alpha, performance, persistence

AC

JEL Classification G11, G12

ACCEPTED MANUSCRIPT

Persistence of actively managed equity mutual fund performance: the new Indian Evidence

1. Introduction This paper examines the persistence in performance of actively managed Indian equity mutual

CR IP T

funds over short and long time horizons after controlling for systematic factors like market risk, size, value and momentum. We also check whether such performance persistence depends on investor holding period, fund size, age, style or expense-ratio. We find evidence of short term persistence, particularly for larger, older and high expense ratio funds, though much of that

AN US

persistence comes from relatively poor performers within the group.

Understanding mutual fund performance is an important field of research in finance for both investors and portfolio managers because of its obvious impact on wealth. While the issue of past

M

performance is important, that of whether it might persist in future is of even greater significance, the following reasons for which can be easily identified : i) the fund manager‟s reputation and

ED

remuneration are heavily influenced by her ability to achieve consistently superior performance ii) marketing of a fund is based on its performance track record iii) historical performance is an

PT

important criterion in the investor‟s choice of a fund and iv) from an academic perspective also,

CE

assessing the existence and persistence of mutual fund managerial ability is an important test of the efficient market hypothesis, since evidence of persistent ability would support a rejection of its

AC

semi- form ( Bollen and Busse,2005). Grossman and Stiglitz (1980) argue that one should not expect security prices to fully reflect information of informed individuals; otherwise, there would be no reward for the costly endeavor of seeking new information. In the context of mutual fund performance, one can thus expect some fund managers to possess an informational advantage and be outperforming the others on a consistent basis. Berk and Green (2004) show theoretically that a fund manager's informational

ACCEPTED MANUSCRIPT

advantage should be short-lived when investors direct their capital to recent winners because of diseconomies of scale. The current study derives its motivation from this background. Numerous studies test persistence of fund performance and their findings are generally mixed in nature across different samples and different time periods. Jensen(1968), Kritzman(1983), Dunn

CR IP T

and Theisen(1983), Elton et al.,.,(1990), Kahn and Rudd(1995), Barras et al., (2010), FamaFrench(2010), and Busse et al.,., (2010) show little to no evidence of persistence over long time horizons while Lehman and Modest(1987), Hendricks et al.,(1993), and Goetzmann and Ibbotson(1994), Bollen and Busse (2005) and Avramov and Wermers (2006) report evidence of

AN US

persistence. Although much has been clarified for the American market, little evidence is there for other major markets. Brown.et al., (1997), Blake and Timmermann(1998), worked with European equity mutual funds and report evidence of persistence. Robson (1986) and Vos. et al., (1995)

M

report lack of persistence for fund samples from Australia and New Zealand. In this paper, we use a sample of 263 actively managed equity mutual funds which existed

ED

continuously between January 2000 and December 2014 and explore persistence in fund performance. estimate their 4-factor-alpha (Carhart, 1997) as risk adjusted performance measure.

PT

This is to control the fund performance for systematic risk factors catering to market risk, size,

CE

value and momentum. 4-factor alpha is widely applied in the recent mutual fund literature, by Kosowski et al.,. (2006), Fama and French (2010), Busse et al., (2010), Huij and Derwall (2011),

AC

among others to measure risk adjusted performance of funds. We then apply the widely used „contingency table approach‟ (Brown et al., 1992; Goetzmann and Ibbotson, 1994; Malkiel, 1995, Kahn and Rudd, 1995, Babalos et. al. 2008; Elyasini and Jia, 2011) and „quartile transition matrix approach‟ (Brown et al.,1997) to test persistence in 4-factor alphas across short and long time horizons of ranging from 12 months to 36 months. Our results show evidence of short term persistence over 12-month horizon which reduces over longer horizons. We also find that larger,

ACCEPTED MANUSCRIPT

older and high-expense-ratio funds are marginally better performers and more persistent. However, a considerable part of this persistence is attributable to relatively poor performers within the group. Hence all is not well for the investors. To the best of our knowledge, not much work has been done in emerging markets, particularly in

CR IP T

India to address fund performance persistence. Roy and Deb (2004) work with a small sample of Indian mutual funds and report some evidence of persistence. Deb.et al., (2008), study a small sample of Indian equity mutual funds between 2000 to 2005 and report weak evidence of persistence. This paper contributes to the existing literature, particularly in an emerging market

AN US

context, in several ways:

i) Larger sample over longer study period: We use a reasonably large (at least in the context of previous studies in India) continuous sample of 263 actively managed mutual fund schemes and

M

a time period of 15 years between 2000-2014 for our study.

ii) Analysis of sub-samples: For this reason, we consider different groups of mutual funds based

ED

on their size, age, style and expense ratio to estimate persistence. a. Sub sample based on the size of funds: It is expected that fund managers give a better

PT

service and put more effort into managing bigger funds, since if fees are based on the

CE

value of assets under management, the bigger mutual funds pay a higher fee to the fund manager. Also, a larger mutual fund with a larger corpus is expected to attract the best

AC

talent in the fund management industry. Whether this expectation is validated in Indian context is an interesting question. To test that, we track existence of persistence of actively managed equity mutual funds across different fund size partitions measured by their assets under management (AUM).

b. Sub samples based on the age of funds: It is also expected that the funds which survived the competition in the industry for some time should have been doing well.

ACCEPTED MANUSCRIPT

They are likely to attract better managers and charge higher fees. To test that we also track persistence of equity mutual funds across different fund age partitions within our sample, measured by no of years since inception at the point of analysis. c. Sub-samples based on the expense ratio charged by funds: Some recent literature

CR IP T

(Gottesman and Morey, 2007; Fama and French, 2010) attribute persistence to expense ratio of funds. To explore that possibility, we repeat the analysis for different fund expense-ratio partitions within our sample and check for differential patterns, if any, in performance persistence.

AN US

d. Sub-samples based on the fund style: Artificial evidence of persistence could be found where performance between stock classes is persistent over time (Matallín-Sáeza et. al., 2016). For any period, it is likely that funds managed according to same style

M

will perform similarly at least to some extent (Malkeil, 1995). For example, in a period when „growth‟ stocks do well it is likely that growth funds will also perform relatively

Robustness check of the findings and methodology: A common criticism in mutual fund

PT

iii)

ED

better. To test that we measure persistence of sub-samples based on fund styles.

CE

literature about the limitations of tests of persistence is the possibility of bias in the methodology (Matallín-Sáeza et. al. 2016). We, therefore, test for the robustness of the methodologies used by

AC

us through a series of robustness tests( 4 nos) as detailed in the following sections.

The remaining portion of our paper is organized as follows: the next section talks about the data, and detailed methodology, the third talks about the results, and the fourth concludes the discussion, followed by references and tables.

ACCEPTED MANUSCRIPT

2. Data and Methodology We use monthly return data of all actively managed equity mutual funds in India which existed between January 2000- December 2014. All such schemes, which are not actively managed, based on their stated investment objective and style (for example index funds) are excluded from

CR IP T

the sample. We drop all index funds from our sample, which by definition are passively managed and are not expected to exhibit any active management skills, including persistence. We also drop all such funds which did not exist continuously between 2000 to 2014 and/or have problems with data availability. This is to ensure a continuous data availability over these 15 years to facilitate

AN US

checking performance persistence, across short as well as long time horizons. We also drop all such funds whose recent most inflation-adjusted assets under management (AUM henceforth) is in the bottom quintile of sorted list of funds based on AUM. This is to weed out the funds, which are

M

too insignificant in size and impact on investor pool. We start with an initial sample size close to 2000 equity mutual fund schemes from the ACE mutual fund database,2 but finally, after applying

ED

the filtering mechanisms as mentioned above end up with a continuous sample of 263 actively managed funds. We collect the monthly return data and style specification (large cap growth,

PT

median cap-value etc.) for these funds from their published prospectus. We also collect the time

CE

series data of AUM of these funds, expense-ratio of these funds and the date of inception of these funds from the same source. Table 1 below shows the number of funds and principal

AC

characteristics of our overall sample and sub-samples.

2

[Table-1 here]

ACE mutual fund database is a popular data providing service in India which covers 45 AMCs and more than 8000 schemes which are currently prevalent in India. Data they provide includes all basic details of the scheme and portfolios, asset-wise and sector-wise details, fund manager's details, NAVs since Inception, ratios, returns, dividends, MF news etc.

ACCEPTED MANUSCRIPT

2.1 . Risk Adjusted net performance measure for the funds: We use Carhart (1997), 4-factor-alpha to find out the net risk adjusted performance (ALPHA henceforth) of the funds. This controls for the market risk, size, style, momentum and fundexpense related factors which are sometimes identified to be erroneously indicating persistence.

CR IP T

ALPHA is widely applied in the recent mutual fund literature, by Kosowski et al., (2006), Fama and French (2010), Busse et al., (2010), Huij and Derwall (2011), among others to measure risk adjusted performance of funds. The model estimated is as follows: 4 Factor Model:

AN US

(RP - Rf) t =  + 1 (Rm - Rf) t +  2 (SMB) t +  3 (HML)t +  4 (WML) t   t

....................

(1)

In this regression, RP(t) is the return on portfolio i for month t net of expenses (from gross fund return we deduct percentage fund expenses i.e fund expenses scaled by fund AUM) from the

M

gross alpha, Rf(t) is the risk-free rate, RM(t) is the market return. SMB(t) (Small minus big) is the

ED

factor capturing the size effect, and measured as the difference between the returns on diversified portfolios of small stocks and big stocks. HML(t) (high minus low) is the factor capturing the

PT

„value effect‟ and measured as the difference between the returns on diversified portfolios of high book-to-market(value)stocks and low book-to-market(growth)stocks. WML(t), captures the

CE

momentum factor measured as the difference between the month t returns on diversified

AC

portfolios of the winners and losers of the past year. Alpha estimated from this regression should capture the net abnormal performance of the portfolio after controlling for market, size, value and momentum effects (Fama and French, 1992,1993 and Carhart,1997). We collect the data on market return (Rm), risk free return (rf) and the standard factors SMB, HML and WML which are respectively

the

small-minus-big(size),

high-minus-low(value)

and

winners-minus-

ACCEPTED MANUSCRIPT

losers(momentum) factors, from Agarwalla et al., (2013)3 and estimate the model above4. This is done for all funds in our sample, and for all the sub-samples mentioned in the previous section. We generate ALPHAs for each group over time horizons of 12 months, 24 months and 36 months. The different choices of time horizon are primarily for testing the persistence in fund

of the time horizon on the persistence of a fund.

2.2 : Testing net fund performance persistence:

CR IP T

performance over shorter as well as longer time horizons, and to find the impact, if any, of length

AN US

After the ALPHAs of the funds are generated across time horizons of 12, 24 and 36 months respectively we assess the degree of persistence in fund performance. Carhart (1997) defines persistence as a positive relation between performance rankings in an initial ranking period and the

M

subsequent period. This requires the measurement of the fund performance in at least two consecutive time periods of similar lengths. We call the first period as the formation period (FP)

ED

and the second as the test period (TP). We use 3 lengths of FPs and TPs for forming various FPTP combinations: 12 months, 24 months and 36 months. Our total study period ranges over 15

PT

years or 180 months. Accordingly, we have 14 numbers 12- month, 6 numbers 24-month and 4

CE

numbers 36-month FP-TP combinations. For ease of referencing, from this point onwards we refer to FPs and TPs in terms of their sequential number notation rather than calendar year notation. The

AC

FP-TP combinations with corresponding calendar periods are detailed in Table-2 below.

3

4

[Table-2 here]

http://www.iimahd.ernet.in/~iffm/Indian-Fama-French-Momentum/

Before running the models, checks w.r.t to standard OLS assumptions are conducted and no serious problems were noticed.

ACCEPTED MANUSCRIPT

We then use a couple of non-parametric approaches namely: “contingency table” and a “quartile transition matrix’ approach widely used in mutual fund research literature to test persistence in fund performance measured by ALPHA. We discuss the approaches hereunder: 2.2.1: 2 x 2 Winner –Loser contingency table approach

CR IP T

The contingency table approach is a non-parametric statistical approach adopted to test persistence in performance in many previous studies (Brown et al., 1992; Goetzmann and Ibbotson, 1994; Malkiel, 1995, Kahn and Rudd, 1995, Babalos et. al. 2008, Elyasiani and Jia, 2011). In this approach for each formation period we divide the funds into two categories: i) Winners(W): funds

AN US

with performance (ALPHAs) above the median, and ii) Losers(L): those with below median performance. Since we look at two consecutive periods, the FP and the TP, some funds belong to winners in both periods (WW), some are losers in both (LL). The rest are winners in one, losers in

M

the other period (WL or LW respectively). For each combination of FP-TP, we obtain the number of funds belonging to each category as above. With persistence of performance, the frequency

ED

distribution in the four quadrants of a 2 X 2 winner-loser contingency table will be uneven and the diagonal quadrants will show higher numbers (WW and LL). Evidence for such persistence is

PT

statistically tested using a Chi Square test for a null hypothesis of no persistence.

(O  Ei ) We calculate the test statistic as:    i Ei

2

2

CE

(2)

AC

where, Oi is the observed number in each quadrant and Ei is the expected number in each quadrant (=n/4 for no persistence), where n is the total number of observations. χ2 follows a chi-square distribution with [(R-1) x (C-1)] degrees of freedom in a R x C contingency table. We check whether χ2calculated >= χ2

α,df

where α is the level of significance and

χ2 α,df is the critical value of χ2 for df = (R-1)*(C-1) . A χ2calculated >= χ2 α,df case indicates rejection of the null , thus implying persistence.

ACCEPTED MANUSCRIPT

It may be noted that in this set up, a significant chi-square may also result from performance reversals (WL and LW proportions being higher than WW and LL proportions). As a result, we look into the relative proportions of the WW and LL between every FP-TP combination, and mark the occurrence of persistence only in cases where proportion of „stayers‟ (WW+LL) is more than

CR IP T

the proportion of „movers‟ (LW + WL). The cases where the proportion of „movers‟ is more than the proportion of stayers, are marked as cases of „reversals‟. Within the cases of persistence, we also note the cases where persistence observed is primarily due to losers rather than winners (LL > WW).

AN US

2.2.2: Quartile transition matrix approach:

As investors often focus on top quartile performers, we extend the 2 X 2 contingency table approach to a 4 x 4 quartile level analysis which is also sometimes termed as a transition matrix

M

approach. We find application of this methodology in Brown et al., (1997). Here for each FP we divide the funds into four quartiles based on their performance in that period: top quartile (1),

ED

second quartile (2), third quartile (3) and bottom quartile (4). We then repeat the procedure for the TP as well and track the quartile position of each fund during the two periods. Funds that remain

PT

in the top quartile in the FP as well as the TP are included under the 1-1 category, while those that

CE

remain in the bottom quartile in both the periods are included in the 4-4 category. All other categories, 1-2, 2-3, 4-3 etc., indicate a shift in the quartile position of the funds in the FP and TP.

AC

For each FP-TP combination, we obtain the number of funds belonging to each category as above. We then observe the actual and the expected frequencies of funds and carry out a Chi-Square test similar to the 2 X 2 set up. As in the case of winner-loser contingency table, in a 4x 4 quartile analysis set up also, a significant chi-square may result from performance reversals or stray movements in between the quartiles. We thus once again look into the conditional proportions of "stayers" (maintaining their

ACCEPTED MANUSCRIPT

position within top half or bottom half i.e (1-1+1-2+2-1+2-2+3-3+3-4+4-3+4-4) vis-a vis that of "movers"(1-3+1-4+2-3+2-4+3-1+3-2+4-1+4-2) and consider only those cases, where stayers are proportionally more than movers and the chi-square is significant as the cases of persistence. The other cases are marked as cases of reversals. Like the contingency table approach, we also note the

CR IP T

cases where persistence is caused primarily by losers (stayers in the bottom half) rather than the winners (stayers in the top half).

2.3: Analysis of Persistence within size, style, age and expense-ratio partitions:

There may be differential patterns of persistence within sub-samples based on criteria of fund-

AN US

size, fund-style, fund-age and fund-expense-ratios. Identifying such differential patterns is as important as identifying persistence per-se. To explore that, we divide our sample of 263 funds based on four classification schemes:

Classification based on size of the funds: We measure fund size by AUM 5 at the point of

M

i)

analysis. We sort the sample funds every year in a descending order of AUM, and then divide

ED

the sample into top 40 %( 105 nos), median 20% (53 nos) and bottom 40% (105 nos)6 based on their reported AUM that year. We consider the first group of funds as large funds, second

PT

group as median funds and the last group as small funds.

CE

ii) Classification based on age: We measure fund age by the number of elapsed years since inception at the point of analysis. We calculate the age of the funds every year as the

AC

difference from that year to the year of inception and sort the funds based on their age again

5

As mentioned before AUM refers to recent most inflation adjusted assets under management of the funds reported at the point of analysis. Inflation adjustment is done to control for artificial increase in AUM across time merely due to inflation. This is done by deflating the AUM every year by the latest reported CPI at the point of estimate. 6

The choice of 40%, 20% and 40% is made to have a clear distinction between large and small funds and simultaneously have a good representation within both large and small, as they are our principal focus of analysis rather than the median group.

ACCEPTED MANUSCRIPT

in a descending order based on their age, and again divide our sample as above, into old, median and young funds 7. iii) Classification based on expense-ratio of the funds : We use the reported expense-ratio of the funds each year and follow an exactly similar strategy as in i) and ii) above and divide our

CR IP T

sample into high expense-ratio, median expense-ratio and low expense-ratio funds.

iv) Classification based on fund styles: For each sample fund, we find out its style from their stated objectives in its published prospectus and accordingly divide the sample funds into 9 groups based on a 3 x 3 style box position of the funds based on their size (large, median and

AN US

small) and value (value, blend and growth) dimensions. After all the filtering in our final sample we find representations only from the following groups: Large-Blend (LB), LargeGrowth (LG), Large-Value (LV), Median-Blend (MB), Median-growth(MG). We use these

M

sub-groups based on fund investment style for further analysis.8

We repeat all our analysis as detailed above for the full sample as well as size partitions (large,

ED

median and small funds), age partitions (old, median and young funds), expense-ratio partitions

PT

(high, median and low expense-ratio funds) and style partitions (LB, LG, LV, MG and MB).

AC

CE

2.4: Robustness check of the findings:

7

Of course, the age at any point of analysis of any fund should be proportional to their ages at the beginning of our study period. 8 Since we are using a constant sample of 263 funds throughout the entire study period, the group size of each category as detailed above remains constant for all years but the group composition changes from year to year. This is because of the fact that both AUM as well as expense-ratios of the funds will change across years. However, the age at any point of time will be proportional to the age at the first point of analysis. Again, the stated objective of the funds along value and size dimensions remains constant through all years. Hence the sorted list of funds based on age as well as fund style remains constant for all years. But their relative ranking in terms of performance within the category changes every year, making the sub-sample analysis w.r.t these classification schemes relevant.

ACCEPTED MANUSCRIPT

2.4.1: Robustness test 1: Controlling for survivorship bias: repeating the analysis over yearwise samples: We use a constant sample of 263 funds throughout the entire study period, to ensure a continuous data availability over these 15 years to facilitate checking performance persistence, across short as

CR IP T

well as long time horizons. However, an obvious concern arising because of that could be possibility of survivorship bias. A fund house would discontinue a fund that did not perform well. Which could affect the observed persistence in performance of mutual funds. To check whether such possibility is significant during our study period, we repeat the entre analysis based on a

AN US

yearwise sample of funds (i.e. we take all funds that are present in each combination of formation period and test period) and generate the number of cases of persistence or reversals based on the

M

same tests and compare those results with the continuous sample results.

2.4.2: Robustness test 2: Benchmarking Fund Persistence Against Passive Funds:

ED

Cortez, Paxson, and Armada (1999), Cuthbertson. et al., (2010), Matallín-Sáez et al., (2016) point out some limitations of contingency table and quartile transition matrix approach for testing

PT

persistence. Matallín-Sáez et al.,.(2016) state that these methodologies seem to be biased towards

CE

finding evidence to support mutual fund performance persistence, because, when they apply contingency tables and transition matrices approach to a set of passive funds, which are expected

AC

to exhibit no evidence of active management skills including persistence, they still find some evidence of persistence. To check whether our findings also suffer from similar spuriousness, we apply the tests of persistence to a group of completely “passive funds” which are expected to show no evidence of any active management skills, including persistence in performance. apply a similar approach. This approach of robustness test using passive portfolios is documented in previous literature such as Bollen and Busse (2001), Bollen and Busse (2005), among others.

ACCEPTED MANUSCRIPT

While many of these have used Sharpe (1992) return based quadratic programming approach to generate simulated passive portfolios mimicking the style of sample funds we use an actual sample of index funds (97 nos) which existed simultaneously with our sample funds, as the group of passively managed funds. These funds are originally excluded from our sample because of their

CR IP T

intrinsic passive nature. We then apply the same contingency-table and quartile- transition-matrix approach to this set of completely passive funds and check for evidence of persistence.

2.4.3: Robustness test 3: Regression model:

AN US

We regress the test period performance of sample funds with formation period performance

α TP,i  α  β * α FP,i  ε

…………………….(3)

where  TP,i is the TP-ALPHA of fund i, and  FP,i is its FP-ALPHA. Positive and significant „‟

M

should imply a positive association of performance over one period with that in the subsequent period and should be evidence of persistence of performance. A negative and significant „‟

ED

should on the other hand imply evidence of reversal in performance while an insignificant

PT

coefficient should imply no association across the formation and test periods.

CE

2.4.4: Robustness test 4: Using Spearmann rank Correlation coefficient (SRCC) test to check persistence across time:

AC

SRCC is calculated as:

Rs  1 

6 d 2

2

n(n  1)

…………………………………………….(2)

where Rs = Spearman-Rank-Correlation-Coefficient d = difference between rank of a fund, based on a performance indicator during FP and TP n = no of observations, i e, number of funds under consideration.

ACCEPTED MANUSCRIPT

One can test for significance of Rs using a t–test where the t-statistic is calculated as :

tr

n2 1 r2

which is distributed as t-distribution with n − 2 degrees of freedom under the null hypothesis of no

CR IP T

significance. Positive and significant Rs shall imply positive association between performance in FP and TP and hence a case of persistence while a negative and significant Rs shall imply a reversal. Insignificant Rs implies no association.

AN US

3. Results 3.1 : Main results:

Table-3 shows the mean ALPHAs generated across all funds in the full sample and across different sub-samples.

M

[Table-3 here]

ED

We can clearly see that actively managed equity mutual funds on the whole have done reasonably well in India, at least over the period under consideration. The mean monthly ALPHAs range

PT

between 0.28% to 0.32% for all funds which translates into abnormal returns around 3.5-4% on an

CE

annualized basis after controlling for size, value, momentum and expense ratio. The larger and older funds have done marginally better on the average, compared to smaller and younger funds.

AC

A similar observation is also noticed for funds with higher expense-ratios w.r.t their low expenseratio counterparts, although not much difference is visible between high and median-expense-ratio funds. Now whether they are consistent or not is the main question we intend to address next. Within the style partitions, we find that the Large Blend (LB) funds have performed poorly for no apparent reasons, which is in total contrast to the other two partitions within Large cap funds (LG and LV). It seems that funds belonging to LG and LV category, having a clear strategy to

ACCEPTED MANUSCRIPT

investing in either value or growth stocks are generating better performance than the fund managers caught in between. The ALPHAs generated by the mid cap funds (MB and MG) are lesser compared to LG and LV but better than LB. Panels A and B of Table 4, show summary results of the contingency table approach, across the

CR IP T

full sample, expense-ratio, style, size and age partitions. The table cells indicate the number of cases for persistence, reversals and no inference. The cases where chi-square statistics are significant at less than 10% indicate cases of either persistence (WW+LL>WL+LW i.e where "stayers" are proportionally more than “movers") or reversals (WW+LL
AN US

"movers" are proportionally greater than “stayers"). In the other cases where Chi-squares are not significant indicates lack of specific inference about persistence or reversal. Within cases of persistence, we also note and separately report persistence due to winner funds(WW>LL) or that

M

due to loser funds(LL>WW).

[Table-4 here]

i)

ED

From tables 4, following are our principal observations: We find evidence of persistence overall, particularly over short term (11 out of 14 FP-TP

PT

combinations within 12 months‟ horizon), which tend to reduce over longer horizons of

CE

24(3 out of 6 FP-TP combinations) and 36 months (2 out of 4 FP-TP combinations). However, a significant portion of this persistence comes from loser funds which are

AC

consistently poor performers within the group. It may be mentioned that; we find no distinct pattern in persistence visible across calendar years, although we do not report calendar yearwise results here for the sake of brevity. The cases of persistence are almost uniformly distributed across calendar years. This is true for short term as well as long term horizons.

ACCEPTED MANUSCRIPT

ii) Across expense-ratio partitions we find that reasonable amount of short term persistence, which is highest for the high-expense-ratio group. But this short term persistence is mostly contributed by the loser funds, rather than the winners. This persistence again tapers down across longer horizons, except for the high-expense-ratio group which is pretty over longer

CR IP T

horizons also. The long term persistence for high-expense-ratio funds comes mostly from the winner funds.

iii) Not much pattern is visible across style partitions, except that persistence reduces across longer time horizons, almost equally contributed by winner and loser funds and is

AN US

marginally higher for large style funds (LB, LG and LV) than median style funds (MB and MG).

iv) Across size partitions we find evidence of short term persistence which reduces for longer

M

horizons. For large funds the persistence is contributed mostly by winner funds, while for small and median funds the losers are major contributors in short term persistence. Both

ED

persistence, as well as proportions of losers contributing to it, decrease for longer time horizons.

Across age partitions we find that old funds are much more persistent than the median and

PT

v)

CE

young age group. This is true for all time horizons. Interestingly, this is one group (old funds) where persistence is significantly due to winners only across shorter and longer

AC

horizons.

Panels A, B of Table-5, show the results of quartile-transition-matrix following a similar order as in tables 4. The table cells indicate the number of cases of persistence, reversals, persistence due to winners, persistence due to losers and no inference cases.

ACCEPTED MANUSCRIPT

[Table-5 here] Our principal findings from the transition-matrix test, are more or less along similar lines with the contingency table approach. Overall we find evidence of persistence is observed, particularly over short term which reduces with increase in length of time horizon. No significant visible patterns

CR IP T

are there across calendar years or style partitions. High-expense-ratio funds are more persistent than the median and low-expense-ratio funds but a significant portion of this persistence is due to loser funds, particularly over longer time horizons. Across size and age partitions, like before, we find the larger and older funds are marginally more persistent, but once again a major portion of

AN US

the persistence is contributed by relatively poor performers within the groups, particularly over longer horizons.

M

3.2 : Robustness test results:

Tables 6,7,8 and 9 indicate the results of the robustness test. Table 6 shows the results of the

ED

persistence tests on the yearwise sample to control for survivorship bias. Here we consider all funds that are present in each combination of formation period and test period, and take a time

PT

series average of cases of persistence or reversals based on the same tests. It can be observed

CE

that even for the full sample of funds we find evidence of persistence in line with our principal results from the constant sample.

AC

[Table-6 here]

Table 7 shows the results of the passive-index-fund approach. Our results are in sharp contrast to the findings of Matallín-Sáez et al., (2016) who find evidence of persistence when they apply contingency-table and transition- matrix approach to a set of passive funds, and posit that these methodologies seem to be biased towards finding evidence to support mutual fund performance persistence. We however, find no evidence of persistence here, in short as well

ACCEPTED MANUSCRIPT

as long time horizons for the passive- index-funds, thus negating the possibility of bias in the persistence test methods. This is true over 12-months, 24-months as well as 36-months horizon. [Table-7 here]

CR IP T

Tables 8 and 9 show the results of robustness tests 3 and 4 respectively vide the regression model and SRCC tests elaborated in the previous section. We find that the results here are in line with our principal findings from the main analysis i.e evidence of persistence is visible

AN US

particularly over shorter time horizons under both tests.

[Tables-8 & 9 here]

Based on the robustness tests, we may thus safely conclude that the results of the tests of

M

persistence we obtain are not spurious or biased. The findings are in line with some previous research on the issue in developed markets (Brown, et al.,1992; Goetzmann and Ibbotson ,1994;

ED

Brown and Goetzmann,1995; Malkiel,1995; Kahn and Rudd,1995; Babalos, et al.,2008; and Elyasiani and Jia,2011). On the whole we may summarize, that there is evidence of persistence in

PT

performance of actively managed equity mutual funds in India, particularly over short term (12

CE

months) which tend to reduce over longer horizons of 24 and 36 months. This pattern is more or less uniform and independent of the calendar years in our study period. The larger, high-expense-

AC

ratio and older funds tend to be more persistent than the other funds. However, a significant portion of this persistence comes from relatively poor performers within the group, particularly for shorter periods. An exception are the older funds, which exhibit short as well as long term persistence, and this comes mostly from the winners rather than the losers.

ACCEPTED MANUSCRIPT

4. Conclusion In this paper we explore whether a good past performance is indicative to any degree of the fund‟s subsequent performance for actively managed equity mutual funds in India. Using a sample of 263 actively managed equity mutual funds which existed between January 2000 and December 2014,

CR IP T

we apply a number of parametric and non-parametric approaches and find evidence of short term persistence overall and more so for larger, older and high expense-ratio funds.

We use 4-factor-alpha (Carhart,1997) and adjust it for fund-expenses scaled by recent most inflation-adjusted assets under management as net risk adjusted performance measure of funds (we

AN US

refer that as ALPHA throughout the text) and observe that the larger, older and high-expense-ratio funds exhibit higher ALPHAs, compared to smaller, younger and low-expense-ratio funds. We then test the consistency of fund performance w.r.t these ALPHAs across consecutive

M

measurement periods using a contingency-table-approach and a quartile-transition-matrix approach.

ED

We find evidence of persistence overall, particularly over short term (12 months), which tend to reduce over longer time horizons of 24 and 36 months. This pattern is uniform and independent of

PT

the calendar years in our study period. Some previous studies posit that persistence may actually

CE

be driven by some underlying factors like momentum(Carhart,1997), manager‟s costs (Gottesman and Morey,2007; Fama and French,2010) or persistence in performance of stock style classes

AC

(Matallín-Sáeza et. al., 2016) rather than true active management skills. To control for such possibilities, we repeat our analysis across a number of sub-samples based on size, age, style and expense ratio of funds. We still find evidence of persistence in general, with the larger, highexpense-ratio and older funds manifesting more persistence than the other funds. However a significant portion of this persistence comes from loser funds, particularly for shorter periods. An exception are the older funds where we observe short as well as long term persistence and this

ACCEPTED MANUSCRIPT

comes mostly from the winners rather than the losers. Combining these findings with the observation that the ALPHA per-se generated by the older funds are also higher in magnitude, we may conclude that, funds which have survived the competition of the industry for a substantial length of time in India, seem to be justifying their reputation through superior and persistent

CR IP T

performance.

We affirm the robustness of our results through several approaches, namely, using a period wise „full sample‟ instead of a constant sample existing throughout the study period, benchmarking persistence observed for our sample against an investment opportunity set faced by a passive

AN US

investor, in this case by a set of completely passive index funds, a regression model and a rank correlation test. The robustness tests entirely support our principal finding of existence of persistence, at least in the short run and negate the possibility of any bias in the methodology or

M

time period adopted.

From an academic perspective, the comforting observation is disappearance of persistence over

ED

longer time horizons, putting up a case for market efficiency, at least in the long run. Our findings indicate that markets in emerging economies like India are not fully efficient in the short run and

PT

past performance of managed portfolios can have some useful information for investors, to assess

CE

possible future performance. These findings, we believe, can have significant implications for all

AC

stakeholders in the mutual fund industry for India in particular, and emerging markets in general.

References:

Agarwala, S. K., Jacob, J., & Varma , J. R. (2013). Four factor model in Indian equities market. Working Paper W.P. No. 2013-09-05, Indian Institute of Management, Ahmedabad. Avramov, D., & Wermers, R. (2006). Investing in mutual funds when returns are predictable. Journal of Financial Economics,81, 339-377.

ACCEPTED MANUSCRIPT

Babalos, V. C., Kostakis, G. M., & Philippas, N. (2008). Testing for persistence in mutual fund performance and the ex-post verification problem: evidence from the Greek market. The European Journal of Finance, 14(8), 735-753. Barras, L. B., Scaillet, O., & Wermers, R. (2010). False discoveries in mutual fund performance:

CR IP T

measuring luck in estimated alphas. Journal of Finance,65, 179-216.

Berk, J. B., & Green , R. C. (2004). Mutual Fund Flows and Performance in Rational Markets. Journal of Political Economy, 112(6), 1269-1295.

Blake, D., & Timmermann, A. (1998). Mutual Fund Performance: Evidence for the UK. Review of

AN US

Finance,2(1), 57-77.

Bollen , N., & Busse, J. (2001). On the timing ability of mutual fund managers. Journal of Finance , 56(3), 1075-1094.

M

Bollen, N. P., & Busse, J. A. (2005). Short-term persistence in mutual fund performance. Review of Financial Studies,18(2), 569-597.

ED

Brown, G., Draper, P., & McKenzie, E. (1997). Consistency of UK Pension Funds Investment Performance. Journal of Business Finance & Accounting,24(2), 155-178.

PT

Brown, S. J., Goetzmann, W., Ibbotson, R. J., & Ross, S. A. (1992). Survivorship bias in

CE

performance studies. Review of Financial Studies, 5(4), 553-580. Busse, J. A., Goyal, a., & Wahal, s. (2010). Performance and persistence in institutional

AC

investment management. Journal of Finance,65(2), 765-790. Carhart, M. M. (1997). On Persistence in Mutual Fund Performance. Journal of Finance,52(1), 57-82.

Cortez, M. D., Paxson, C. A., & Armada, M. J. (1999). Persistence in portuguese mutual fund performance. European Journal of Finance, 5(4), 342-365.

ACCEPTED MANUSCRIPT

Cuthbertson, K., Nitzshe, D., & O'Sullivan, N. (2010). Mutual fund performance: Measurement and evidence. Financial Markets,Institutions and Instruments, 19(2), 95-187. Deb, S. G., Banerjee, A., & Chakrabarty, B. B. (2008). Persistence in performance of Indian equity mutual funds: an empirical investigation. IIMB Management Review,20(2), 172-

CR IP T

187.

Dunn, P. C., & Thiesen, R. D. (1983). How Consistently Do Active Managers Win? Journal of Portfolio Management,9(4), 47-50.

Elton, E. J., Gruber, M. J., & Rentzler, J. (1990). The Performance of Publicly Offered

AN US

Commodity Funds. Financial Analysts Journal,46(4), 23-30.

Elyasiani, E., & Jia, J. (2011). Performance persistence of closed-end funds. Review of Quantitative Finance and Accounting, 37(3), 381-408.

Finance, 46, 427-466.

M

Fama, E. F., & French, K. R. (1992). The cross section of expected stock returns. Journal of

ED

Fama, E. F., & French, K. R. (2010). Luck versus skill in the cross section of mutual fund returns. Journal of Finance,65(5), 1915-1947.

PT

Fama, E., & French, K. (1993). Common Risk Factors in Returns on Stock and Bonds. Journal of

CE

Financail Economics, 33(1), 3-56. Goetzmann, W. N., & Ibbotson, R. G. (1994). Do Winners Repeat? Journal of Portfolio

AC

Management,20(2), 9-18. Gottesman, A. A., & Morey, M. R. (2007). Predicting emerging market mutual fund performance. The Journal of Investing, 16(3), 111-122.

Grinblatt, M., & Titman, S. (1992). Performance measurement without benchmarks: An examination of mutual fund returns. Journal of business, 47-68.

ACCEPTED MANUSCRIPT

Grossman, S. J., & Stiglit, J. E. (1980). On the impossibility of informationally efficient markets. The American economic review, 70(3), 393-408. Hendricks, D., Patel, J., & Zeckhauser, R. (1993). Hot Hands in Mutual Funds: Short-Run Persistence of Relative Performance,1974-1988. Journal of Finance , 48(1), 93-130.

CR IP T

Huij, J., & Derwall, J. (2011). Global equity fund performance, portfolio concentration, and the fundamental law of active management. Journal of Banking & Finance, 35(1), 155-165. Jensen, M. C. (1968). The performance of mutual funds in the period 1945–1964. Journal of Finance , 23(2), 389-416.

AN US

Kahn, R. N., & A, R. (1997). The Persistence of Equity Style Performance: Evidence from Mutual Fund Data. In T. D. Coggin, & F. J. Fabozzi, Handbook of Equity Style Management (pp. 257-267). New Jersey: Wiley & Sons.

M

Kahn, R. N., & Rudd, A. (1995). Does Historical Performance Predict Future Performance? Financial Analysts Journal,51(6), 43-52.

ED

Kosowski, R., Timmermann, A., Wermers, R., & White, H. (2006). Can mutual fund “stars” really

2595.

PT

pick stocks? New evidence from a bootstrap analysis. The Journal of finance, 61(6), 2551-

CE

Kritzman, M. (1983). Can Bond Managers Perform Consistently? Journal of Portfolio Management,9(4), 54-56.

AC

Lehmann, B. N., & Modest, D. M. (1987). Mutual Fund Performance Evaluation: A Comparison of Benchmarks and Benchmark Comparisons. Journal of Finance,42(2), 233-265.

Malkiel, B. G. (1995). Returns from Investing in Equity Mutual Funds 1971-1991. Journal of Finance,50(2), 549-572.

ACCEPTED MANUSCRIPT

Matallín-Sáez, J. C., Soler-Domínguez, A., & Tortosa-Ausina, E. (016). On the robustness of persistence in mutual fund performance. The North American Journal of Economics and Finance, 36(1), 192-231. Robson, G. (1986). The Investment Performance of Unit Trusts and Mutual Funds in Australia for

CR IP T

the Period 1969 to 1978. Accounting & Finance,26(2), 55-79.

Roy, B., & Deb, S. S. (2004). Conditional alpha and performance persistence for Indian mutual funds: empirical evidence. ICFAI Journal of Applied Finance, 30-48.

Sharpe, W. F. (1992). Asset allocation: Management style and performance measurement?

AN US

Journal of Portfolio Management, 18(2), 7-19.

Vos, E., Brown, P., & Christie, S. (1995). A Test of Persistence in the Performance of New

AC

CE

PT

ED

M

Zealand and Australian Equity Mutual Funds. Accounting Research Journal,8(2), 19-34.

ACCEPTED MANUSCRIPT

Size Partitions Large funds Median Funds Small Funds

CR IP T

Table 1: Sample description This table shows the principal characteristics of our sample namely monthly average return(annualized), AUM in Rs. millions and expense ratios as percentage of AUM. Values are reported for the overall sample, and various sub-samples based on size, age and expense ratio, and fund styles separately. Number of Monthly AUM (Rs. Expense ratios (% funds (n) return(annualized) Millions) of AUM) 263 10.61% 4838.10 1.49% All Funds

5805.72 4644.58 3525.82

105 63 105

10.93% 10.19% 9.54%

Old Funds Median Funds Young Funds

105 63 105

11.35% 9.76% 9.37%

5321.91 4547.81 4067.69

1.42% 1.45% 1.44%

Exp. ratio Partitions High exp. ratio Median exp. ratio Low exp. ratio

105 63 105

12.41% 9.76% 8.31%

4354.29 4523.62 5049.82

1.71% 1.56% 1.08%

Style Partitions Large Blend (LB) Large Growth (LG) Large Value (LV) Median Blend (MB) Median growth(MG)

40 140 9 55 19

10.93% 11.26% 9.82% 9.65% 8.34%

4692.96 5631.55 3420.05 3782.95 3023.29

1.42% 1.57% 1.50% 1.37% 1.42%

M

ED

PT

CE AC

AN US

Age Partitions

1.53% 1.43% 1.34%

ACCEPTED MANUSCRIPT

FP month range

TP month range

1 - 12 13-24 25-36 37-48 49-60 61-72 73-84 85-96 97-108 109-120 121-132 133-144 145-156 157-168

13-24 25-36 37-48 49-60 61-72 73-84 85-96 97-108 109-120 121-132 133-144 145-156 157-168 169-180

2002-03 2004-05 2006-07 2008-09 2010-11 2012-13

1 - 24 25-48 49-72 73-96 97-120 121-144

25-48 49-72 73-96 97-120 121-144 145-168

2003-05 2006-08 2009-11 20012-14

1 - 36 37-72 73-108 109-144

37-72 73-108 109-144 145-180

FP

TP

12 months

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

24 months

2000-01 2002-03 2004-05 2006-07 2008-09 2010-11 2000-02 2003-05 2006-08 2009-11

M

ED

PT CE

AC

AN US

Time Horizon

36 months

CR IP T

Table 2: This table shows the various formation period(FP) and test period (TP) combinations used by us for testing persistence in fund performance. During analysis we use a predetermined serial number of the months in our study period from 1 (corresponding to January 2000) to 180(corresponding to December 2014). Given our sample period is from month 1 to month 180, we have 14 nos FP-TP combinations for 12 months‟ horizon, 6 nos FP-TP combination for 24 months‟ horizon and 4 nos FP-TP combinations for 36 months‟ horizon. The month serial numbers for various FPs and TPs and their respective calendar durations are indicated in the table. FP-TP combination Calendar period ( month from-to )

ACCEPTED MANUSCRIPT

Table 3: This table shows the mean 4-factor ALPHA values obtained by running the following 4 factor model:

(RP - Rf) t =  + 1 (Rm - Rf) t +  2 (SMB) t +  3 (HML)t +  4 (WML) t   t

CR IP T

across all funds in our sample and across different partition categories viz : size, age, style and expense ratio partitions . The number of funds in each sub category and the 4 factor ALPHAs obtained as above are indicated in the table cells. For size, age and expense ratio partitions, we divide the full sample into top 40%(105 nos), median 20%(53 nos) and bottom 40% (105 nos) respectively based on the criteria. However, the style partitions are based on the actual number of funds representing each category, determined from the stated objectives of the funds in their prospectus. The ALPHAs are estimated by the model above over different holding periods of 12 months, 24 months and 36 months. All cross sectional mean ALPHAs thus obtained across all categories are significant at 5% or 10% level of significance and thus the significance is not indicated separately. 4 factor ALPHA

All Funds

Number of funds (n) 263

Size Partitions Large funds Median Funds Small Funds

105 63 105

Age Partitions Old Funds Median Funds Young Funds

105 63 105

0.38% 0.34% 0.23%

0.34% 0.37% 0.22%

0.36% 0.31% 0.21%

105 63 105

0.32% 0.33% 0.30%

0.34% 0.30% 0.31%

0.30% 0.29% 0.28%

40 140 9 55 19

-0.36% 0.33% 0.40% 0.28% 0.29%

-0.34% 0.34% 0.39% 0.29% 0.29%

-0.32% 0.29% 0.38% 0.27% 0.29%

Style Partitions Large Blend (LB) Large Growth (LG) Large Value (LV) Median Blend (MB) Median growth(MG)

36 months 0.29%

0.36% 0.30% 0.28%

0.35% 0.27% 0.27%

0.34% 0.28% 0.24%

AN US

24 months 0.32%

M ED

PT

AC

CE

Exp ratio Partitions High exp ratio Median exp ratio Low exp ratio

12 months 0.32%

CR IP T

ACCEPTED MANUSCRIPT

AN US

Table 4: Panel A: Contingency Table Approach results: Number of cases of persistence (due to winners and losers) and reversals for all funds, expense ratio and style partitions. The cells in the table indicate the total number of cases in the contingency table approach for persistence, reversals and no inference cases. The way we identify such cases are elaborated in the text. The total number of formation period (FP) and test period (TP) is also indicated in the last column. All Expense ratio partitions Style partitions Funds Total no of Low Median High LB LG LV MB MG FP-TP combinations 12 months -do-do-do-do-

11 7 4 3 0

7 5 2 3 4

7 3 4 3 4

10 2 8 3 1

7 4 3 2 5

7 4 3 3 4

6 0 6 6 2

4 1 3 1 9

4 1 3 1 9

14 14 14 14 14

Persistence Persistence due to winners Persistence due to losers Reversal No inference

24 months -do-do-do-do-

3 2 1 2 1

1 1 0 2 3

3 3 0 2 1

5 5 0 0 1

1 1 0 1 4

2 1 1 2 2

1 0 1 5 0

2 1 1 1 3

1 1 0 0 5

6 6 6 6 6

Persistence Persistence due to winners Persistence due to losers Reversal No inference

36 months -do-do-do-do-

1 1 0 1 2

3 3 0 1 0

2 2 0 1 1

1 0 1 1 2

3 0 3 1 0

1 1 0 0 3

1 1 0 1 2

4 4 4 4 4

ED 2 0 2 2 0

PT

CE AC

M

Persistence Persistence due to winners Persistence due to losers Reversal No inference

1 1 0 1 2

Table 4: (--- contd---)

CR IP T

ACCEPTED MANUSCRIPT

Panel B: Summary of Contingency Table Approach results: Number of cases of persistence and reversals for all funds, Size and age partitions All Funds

Size partitions

12 months -do-do-do-

11 7 4 3

8 3 5 5

No inference

-do-

0

1

Persistence Persistence due to winners Persistence due to losers

24 months -do-do-

3 2 1

3 3 0

-do-

2

-do-

1

Persistence Persistence due to winners Persistence due to losers Reversal No inference

36 months -do-do-do-do-

PT

CE AC

2 0 2 2 0

Young

Median

Old

Total no of FPTP combinations

12 7 5 1

s4 1 3 3

9 4 5 5

10 8 2 2

14 14 14 14

6

1

7

0

2

14

3 3 0

4 4 0

2 2 0

4 4 0

4 4 0

6 6 6

3

2

2

1

2

0

6

0

1

0

3

0

2

6

1 1 0 1 2

3 2 1 1 0

1 0

3 2

4 3

1 0 3

1 1 0

1 0 0

4 4 4 4 4

ED

Reversal No inference

Large

7 3 4 1

M

Persistence Persistence due to winners Persistence due to losers Reversal

Median

AN US

Small

Age partitions

1 1 0 0 3

CR IP T

ACCEPTED MANUSCRIPT

Low

12 months -do-do-do-do-

13 7 6 0 1

8 5 3 0 6

Persistence Persistence due to winners Persistence due to losers Reversal No inference

24 months -do-do-do-do-

3 1 2 2 1

2 0 2 2 2

Persistence Persistence due to winners Persistence due to losers Reversal No inference

36 months -do-do-do-do-

2 1 1 0 2

ED

PT

CE AC

Median

High

LB

LG

LV

MB

MG

Total no of FP-TP combinations

11 6 5 2 1

7 2 5 0 7

6 1 5 0 8

10 6 4 4 0

6 3 3 8 0

6 2 4 1 7

3 1 2 1 10

14 14 14 14 14

5 4 1 1 0

4 2 2 0 2

2 1 1 0 4

5 3 2 1 0

3 3 0 3 0

3 0 3 2 1

1 0 1 0 5

6 6 6 6 6

4 3 1 0 0

2 1 1 1 1

1 0 1 0 3

3 1 2 1 0

3 1 2 1 0

2 0 2 0 2

0 0 0 0 4

4 4 4 4 4

M

Persistence Persistence due to winners Persistence due to losers Reversal No inference

AN US

Table 5: Panel A: Transition Matrix Approach results: Number of cases of persistence and reversals for all funds, expense ratio and style partitions The cells in the table indicate the total number of cases in the contingency table approach for persistence, reversals and no inference cases. The way we identify such cases are elaborated in the text as well as in Panels A and B of Table 3. The total number of formation period (FP) and test period (TP) is also indicated in the last column. All Expense ratio partitions Style partitions Funds

1 0 1 0 3

Table 5: (---contd---)

CR IP T

ACCEPTED MANUSCRIPT

Panel B: Transition Matrix Approach results: Number of cases of persistence and reversals for all funds, Size and age partitions All Funds

Size partitions Median

Persistence

12 months

13

6

7

Persistence due to winners Persistence due to losers Reversal

-do-do-do-

7 6 0

3 3 1

3 4 1

No inference

-do-

1

7

6

Persistence Persistence due to winners Persistence due to losers Reversal

24 months -do-do-do-

3 1 2 2

2 2 0 1

2 1 1 3

No inference

-do-

1

3

Persistence Persistence due to winners Persistence due to losers Reversal No inference

36 months -do-do-do-do-

1 0 1 0 3

Young

Median

Old

Total no of FP-TP combinations

11

4

7

8

14

4 7 1

0 4 3

2 5 1

6 2 1

14 14 14

2

7

6

5

14

3 1 2 0

2 0 2 1

3 3 0 1

4 2 2 1

6 6 6 6

1

3

3

2

1

6

1 0 1 2 1

2 1 1 0 2

2 0 2 1 1

2 1 1 0 2

3 2 1 0 1

4 4 4 4 4

M

ED

PT

CE AC

2 1 1 0 2

Large

AN US

Small

Age partitions

ACCEPTED MANUSCRIPT

Table 6 Robustness test 1: Summary of Persistence test results for yearwise sample

Contingency Table

Transition Matrix

Total no of FPTP combinations

Persistence Reversal No inference

12 months -do-do-

6 3 5

5 4 5

14 14 14

Persistence Reversal No inference

24 months -do-do-

3 1 2

3 2 1

6 6 6

Persistence Reversal No inference

36 months -do-do-

1 1 2

4 4 4

AC

CE

PT

ED

M

AN US

Time Horizon

CR IP T

Here we consider all funds that are present in each combination of formation period and test period and generate number of cases of persistence or reversals based on the same tests. The cells in the table indicate the total number of cases in various approaches for persistence, reversals and no inference cases. The way we identify such cases are elaborated in the text as well as in Panels A and B of Table 3. The total number of formation period (FP) and test period (TP) is also indicated in the last column.

2 1 3

ACCEPTED MANUSCRIPT

Table 7 Robustness test 2: Summary of Persistence test results for the Passive Index Funds The cells in the table indicate the total number of cases in the contingency table approach for persistence, reversals and no inference cases. The way we identify such cases are elaborated in the text as well as in Panels A and B of Table 3. The total number of formation period (FP) and test period (TP) is also indicated in the last column. Contingency Table

Persistence Reversal No inference

12 months -do-do-

3 2 9

Persistence Reversal No inference

24 months -do-do-

1 3 2

Persistence Reversal No inference

36 months -do-do-

AN US

M ED PT CE AC

0 4 0

Transition Matrix

Total no of FPTP combinations

CR IP T

Time Horizon

5 8 1

14 14 14

2 3 1

6 6 6

1 2 1

4 4 4

ACCEPTED MANUSCRIPT

Table 8 Robustness test 3: Summary of Persistence test results using the regression model The cells in the table indicate the total number of cases using the regression model under robustness tests for persistence, reversals and no inference cases. The model used for the test is as follows:

α TP,i  α  β * α FP,i  ε

CR IP T

where αTP,i is the TP-ALPHA of fund i, and αFP,i is its FP-ALPHA. Positive and significant „β‟ should imply a positive association of performance over one period with that in the subsequent period and should be evidence of persistence of performance. A negative and significant „β‟ should on the other hand imply evidence of reversal in performance while an insignificant coefficient should imply no association across the formation and test periods. The total number of formation period (FP) and test period (TP) is also indicated in the last column. Regression model

12 months -do-do-

Persistence Reversal No inference

24 months -do-do-

Persistence Reversal No inference

36 months -do-do-

AC

CE

PT

ED

M

Persistence Reversal No inference

Total no of FPTP combinations

AN US

Time Horizon

7 4 3

14 14 14

3 1 2

6 6 6

1 1 2

4 4 4

ACCEPTED MANUSCRIPT

Table 9 Robustness test 4: Summary of Persistence test results using SRCC The cells in the table indicate the total number of cases using the SRCC test under robustness tests for persistence, reversals and no inference cases. The description of the test and the way the test results are interpreted are discussed in detail in the text. Total number of formation period (FP) and test period (TP) is also indicated in the last column. Regression model

Total no of FPTP combinations

Persistence Reversal No inference

12 months -do-do-

6 3 5

14 14 14

Persistence Reversal No inference

24 months -do-do-

2 2 2

Persistence Reversal No inference

36 months -do-do-

AN US

M ED PT CE AC

CR IP T

Time Horizon

2 1 1

6 6 6 4 4 4