c Copyright 2015 Elsevier B.V. Notice Changes introduced ... · (Cuthbertson, Nitzsche, and...
Transcript of c Copyright 2015 Elsevier B.V. Notice Changes introduced ... · (Cuthbertson, Nitzsche, and...
This may be the author’s version of a work that was submitted/acceptedfor publication in the following source:
Basu, Anup & Huang-Jones, Jason(2015)The performance of diversified emerging market equity funds.Journal of International Financial Markets, Institutions and Money, 35, pp.116-131.
This file was downloaded from: https://eprints.qut.edu.au/82173/
c© Consult author(s) regarding copyright matters
This work is covered by copyright. Unless the document is being made available under aCreative Commons Licence, you must assume that re-use is limited to personal use andthat permission from the copyright owner must be obtained for all other uses. If the docu-ment is available under a Creative Commons License (or other specified license) then referto the Licence for details of permitted re-use. It is a condition of access that users recog-nise and abide by the legal requirements associated with these rights. If you believe thatthis work infringes copyright please provide details by email to [email protected]
Notice: Please note that this document may not be the Version of Record(i.e. published version) of the work. Author manuscript versions (as Sub-mitted for peer review or as Accepted for publication after peer review) canbe identified by an absence of publisher branding and/or typeset appear-ance. If there is any doubt, please refer to the published source.
https://doi.org/10.1016/j.intfin.2015.01.002
1
The Performance of Diversified Emerging Market Equity Funds
Abstract
We investigate the performance of globally diversified emerging market equity funds during
the first decade of the twenty-first century. A vast majority of these funds do not outperform
the market benchmark even before transaction costs. The systematic risk of most of the funds
is similar to that of the market benchmark portfolio, which may suggest that they aim to offer
diversification benefits rather than seeking superior risk-adjusted returns through active
management. We do not find any evidence of market timing ability among these funds.
Finally, whilst we detect persistence in performance, this result is driven mainly by the poorly
performing funds.
Keywords: emerging market, diversified equity funds, fund manager performance, market timing; persistence.
2
1. Introduction
Investing in emerging markets has been a major trend among investors well over last two
decades. Emerging markets reportedly offer investors in developed nations the potential of
higher returns as well as risk reduction benefits through portfolio diversification (see for
example, Ratner and Leal (2005)1. For investors in developed markets, mutual funds have
been one of the most important vehicles for investing in emerging markets. Most of these
funds are open-end equity funds (Kaminsky, Lyons, and Schmukler, 2001). In this paper, we
investigate the performance of globally diversified emerging market equity funds between
2000 and 2010. Specifically, we aim to address two important questions related to diversified
emerging market equity funds. First, do these funds produce superior risk-adjusted returns?
Second, is there any persistence in the performance of these funds?
Traditionally mutual funds have provided a low cost route to portfolio diversification.
However, with the spectacular growth of exchange-traded-funds (ETFs) in recent years,
investors now have an alternative vehicle to construct a low cost, well-diversified portfolio.
One important rationale for choosing to invest in traditional mutual funds over ETFs can be
the expectation of ‘abnormal’ returns resulting from the perceived informational advantages
or superior skills of fund managers. Hence it is important to evaluate whether these funds
deliver any (ex-post) positive abnormal performance. Second, from the investor’s
perspective, it is important to examine whether there is any evidence of persistence in the
performance of these funds. If there is predictability in performance of funds, investors may
1 The diversification benefits, arguably, have been less visible in recent years due to financial liberalization and integration of international markets. For example, Charitou, Makris, and Nishiotis (2006) find no significant international diversification benefits in the post 1993 period.
3
(ex-ante) reallocate their savings towards ‘winner’ funds and enhance abnormal returns
(Cuthbertson, Nitzsche, and O’Sullivan, 2008).
There has been a plethora of empirical research on performance of mutual funds in
United States (US), United Kingdom (UK) and other developed markets. Overall, the
evidence does not appear supportive of the claim of any superior performance. There is
limited evidence on persistence with most studies finding the phenomenon to be more
concentrated among the underperforming funds. These findings are generally consistent with
the ‘efficient market’ paradigm according to which asset prices quickly assimilate all
available information thereby making it impossible for any investor or group of investors to
systematically outperform the market on a risk-adjusted basis.
Yet whether the evidence gathered on fund performance from developed markets
applies to emerging market funds remains an open question. It is a commonly held notion that
many of the emerging markets have weaker regulatory environment relative to developed
markets and lack ‘informational efficiency’ due to poor information disclosure requirements
in emerging markets. If true, this may imply more opportunities for emerging market funds to
exploit these inefficiencies and deliver abnormal returns. On the other hand, even if
inefficiencies exist in these markets, the cost of collecting firm-specific information could be
sufficiently high to negate any attempt by funds to deliver superior returns (Chan and
Hameed, 2006).2
Despite the growth in diversified emerging markets funds, empirical studies on their
performance have been rather scarce. Past research conducted on performance of emerging
2 Empirical evidence on the issue of efficiency in emerging markets is divided. Many studies suggest inefficiency or a lower level of efficiency in emerging markets compared to developed markets (See for example, Fillis, 2006; Mobarek, Mollah, and Bhuyan, 2008, Risso, 2009) while others contradict this claim (see, for example, Karamera, Ojah, and Cole, 1999; Griffin, Kelly, and Nardari, 2010). However, there is agreement on the heterogeneous levels in efficiency across these markets.
4
market funds has focused on country specific (like Russia, Poland, India, Malaysia etc.) or
region specific (like Africa or Latin America) funds. Ours is among only a handful of studies
to analyse the performance of emerging market funds that are diversified across multiple
markets. The sample period of most of the past studies do not extend beyond 2006 and hence
they are almost a decade old. The exception is Eling and Faust (2010) whose sample period
extends to August 2008 but their study has an important shortcoming due to its choice of
inappropriate benchmark (S&P 500) in evaluating performance of emerging market funds.
Our study analyzes more recent data (2000-10) to investigate the performance of these funds
at an aggregate as well as individual level. To our knowledge, ours is the first study to
analyze the performance of emerging market equity funds since the onset of global financial
crisis (GFC). We find evidence that is different from some of the earlier findings on
performance of emerging market funds on several counts. Our study reveals greater
underperformance among emerging market funds compared to Eling and Faust (2010). In
contrast to Huij and Post (2011), we show that the persistence in performance is mainly
attributable to the underperforming funds as in developed markets. Also, we find the
abnormal returns spread between the best and the worst performing funds to be much smaller
than what they estimate.
The remainder of the paper is organized as follows. Section 2 briefly reviews the
literature on mutual fund performance in developed and emerging markets. Section 3
describes our data and methodology used in this study. Section 4 presents the empirical
results and Section 5 concludes.
2. Literature review
2.1 Mutual fund performance in developed markets
5
There is a long history of research in mutual fund performance in the US starting with Jensen
(1968) who samples 115 open-end mutual funds over a 20-year period (1945-64). Using the
single factor capital asset pricing model (CAPM), he reports an average excess performance
(alpha) of 0.4% and -1.1% gross and net of expenses respectively, which implies that on
average, the funds earned approximately 1.1% less per year than the benchmark index (S&P
500) on a risk adjusted basis. The results also find very little evidence of an individual fund’s
ability to outperform the market, except by mere chance. Of the 115 funds, only 3 funds show
significantly positive alphas.
Ippolito (1989) conducts a similar study of 143 mutual funds over a different 20-year
sample period (1965-84). Unlike Jensen (1968), he finds an average alpha of 0.81% for the
sample. However, only 12 out of the 143 funds show significantly positive alphas. The author
concludes that net of fees and expenses (excluding load fees), funds are able to outperform
the benchmark by a margin that is large enough to cover their load charges. However, Elton,
Gruber, Das, and Hlavka (1993) take issue with Ippolito’s benchmark and contest the
author’s results. After considering the mutual funds’ holding of non-S&P stocks, they find
the average alpha to be -1.49% with not a single fund producing significantly positive alpha.
Grinblatt and Titman (1989) compare gross and net returns of US mutual funds using
a single factor model over the 1975-85 period across four sets of benchmarks. For gross
returns, the funds outperform two indexes by 1.8% and 2.28% per annum and underperform
the other two indexes by 2.31% and 2.64%. However, outperformance diminishes when net
returns are used with no statistically significant alphas against any of the indexes.
Malkiel (1995) employs the CAPM model to evaluate performance of 239 mutual
funds in US between 1971 and 1991 and finds that these funds significantly underperform the
index both before and after expenses. The reported average alphas for gross and net returns
6
are -2.03% and -3.20% per annum respectively. As a result, he concludes that investors are
better off by investing in the market index rather than by investing with active managers.
Unlike Grinblatt and Titman (1989) who claim that the impact of survivorship bias is modest,
Malkiel finds that the potential upward bias in the estimated alpha due to non-inclusion of
non-surviving funds in the sample is quite severe.
Gruber (1996) employs both CAPM and a multifactor model with the inclusion of
‘small minus big’ (SMB) and ‘high minus low’ (HML) factors and also a bond index. The
funds in his study underperform by 1.56% and 0.65% p.a. against the single and the
multifactor model respectively. He also estimates the potential bias of performance
measurements when non-surviving funds are not included in the sample. The non-surviving
funds underperform the benchmark by 4.2% and 2.75% per annum against the single and
multifactor index respectively.
Daniel, Grinblatt, Titman, and Wermers (1997) use benchmarks based on
characteristics of stocks to evaluate whether active mutual funds outperform passive
strategies. They investigate a database of stock holdings covering 2,500 equity funds between
1975 and 1994. Their results show that whilst the average fund outperforms the passive
strategies, the magnitude of the outperformance (approximately 100 basis points) barely
outweighs the costs in the form of higher management fees. Kothari and Warner (2001) also
analyse simulated funds based on characteristics that mimic actual funds. They conclude that
past studies of mutual fund performance are prone to underestimate abnormal performance by
a large magnitude, particularly if a fund's style characteristics differ from those of the value-
weighted market portfolio.
Mutual fund performance has also been researched in other developed markets
outside the US. Cumby and Glen (1990) examine 15 internationally diversified mutual funds
7
with Morgan Stanley World Index as the benchmark. The authors find no evidence of fund
mangers outperforming the relevant benchmark. The average alpha of the funds in their study
is reported to be is -2.28% p.a. using net returns. Only 3 individual funds show postive alpha
but none of these are statistically significant. Cai, Chan and Yamada (1997) investigate the
performance of 64 Japanese mutual funds over a 10-year period (1981-92). They find the
alphas for the funds to be consistently negative. The alphas using the CAPM model vary
between -6.01% and -7.33% per annum whilst the alphas for the Fama and French three
factor (3F) model range from -5.53% to -6.19%. Blake and Timmermann (1998) study 2,300
UK mutual funds between 1972 and 1995 with the funds sub-divided into four categories:
growth, general, income and small companies. Their results show negative alphas across all
fund categories and estimated to be about -1.8% p.a. Cesari and Panetta (2002) examines
Italian mutual funds and finds positive alpha even after accounting for their non-equity
investments. 3 However, Matallin-Saez (2006) reports an average alpha of -1.2% among
Spanish mutual funds using CAPM. The alphas from other multifactor measures are
substantially lower but the estimates are not statistically significant.
One of the few studies that find mutual funds outperform benchmark net of expenses
is Otten and Bams (2002), who examine the performance of 506 mutual funds in various
European countries including France, Germany, Italy, Netherlands and the UK from 1991 to
1998. They employ Carhart (1997) four factor model and report positive alphas for funds in
all countries except Germany. However, only UK funds significantly outperform benchmark
by 1.33% per year after expenses. Their findings contradict those of Blake and Timmermann
(1998) who do not account for the momentum factor, the main driver for fund performance in
Otten and Bam’s study.
3 The alpha net of expenses, however, is not significantly different from zero.
8
2.2 Performance of funds in emerging markets
Empirical research studies conducted on the performance of emerging market fund managers
have been mainly country specific. For example, Lai and Lau (2010) examine performance of
311 mutual funds in Malaysia and find the average returns to be superior against market
benchmark, particularly during bull markets. Sehgal and Jhanwar (2008) investigate 60
growth and growth income mutual funds in India and find about 25% of the sample exhibit
significantly positive alphas. Swinkels and Rzezniczak (2009) explore security selection and
market timing ability for Polish mutual funds. Their results indicate selection skills but
negative market timing ability.
Among a handful of studies that investigate funds investing in multiple emerging
markets (rather than specific countries), Abel and Fletcher (2004) analyse the performance of
UK-based unit trusts with emerging market equity investments. They do not find any
evidence of superior performance. Gottesman and Morey (2007) examine the performance of
diversified emerging market equity funds over three consecutive periods: 1997-2000, 2000-
2003 and 2003-2005. The authors find some support for the claim that actively managed
funds generally underperform passive benchmarks with the proportion of funds with negative
alphas ranging from 45.03% to 80.90% for the different sample periods. Michelson,
Philipova, and Srotova (2008) examine the performance of 55 diversified emerging markets
funds between September 1999 and January 2005. Their findings suggest that these funds
outperform the MSCI Index and the S&P 500 Index, but underperform the emerging market
index. But since their sample includes only funds with data for the full sample period of 64
months, their results are clearly suffer from survivorship bias.
Eling and Faust (2010) study performance of mutual (and hedge) funds in emerging
9
markets between January 1995 and August 2008. They find that these funds outperform the
benchmark by 0.48% p.a. against the CAPM but underperform by 0.84% per annum against
the FF model. However, the alphas under both models are not statistically significant. As for
the distribution of statistically significant alphas, 14.15% of the funds have positive CAPM
alphas while 3.97% show negative CAPM alphas. The corresponding estimates using the FF
model are 10.33% and 3.97% respectively. However, the authors’ use of a broadly diversified
portfolio of US stocks (with a correlation of 0.97 with S&P 500) as the market factor in their
models to evaluate the performance of emerging market funds raises some concern about the
validity of their conclusions.
Hayat and Kraeussl (2011) analyzes the risk and return characteristics of a sample of 145
Islamic Equity Funds (IEF) that focus mainly on global as well as region/country specific
emerging markets over the period 2000 to 2009. Their results show that IEFs underperform
both Islamic as well as conventional equity benchmarks with underperformance increasing
during the recent financial crisis. They also find that IEF managers have poor market timing
ability.
Whilst studies examining the performance open-end equity funds are of utmost relevance
to our paper, there is also a section of literature that has looked at performance of closed-end
funds. But the conclusions have been similar. For example, Chang, Eun, and Kolodny (1995)
find closed-end emerging market funds fail to generate alphas for US-based investors. Lee
(2001) observes that closed-end country funds in emerging markets underperform their
benchmark IFCI indices. Movassaghi, Brahmandkar, and Shilkov (2004) also could not find
any evidence of superior performance of closed-end funds in any particular region or country
within the emerging markets.
ETFs offer an alternative route for investing in emerging markets through tracking of
passive benchmark indices in these markets. Blitz and Huij (2012) evaluate the performance
10
of such funds in global emerging markets (GEM). They find that, on average, GEM ETFs fall
short of their benchmark indexes by around 85 basis points per annum, which is consistent
with the expected drag on return due to fund expense ratios plus the impact of withholding
taxes on dividends.
Finally, while our study focusses on equity funds, when investing in emerging markets,
some investors also use hedge funds as investment vehicles. Strömqvist (2007) presents
evidence that emerging market hedge funds do not outperform their underlying benchmarks
during the period 1994-2004. This evidence is further supported by Peltomäki (2008) who
finds inferior performance of emerging market hedge funds at the index level. However, he
also finds that nearly 40% of emerging market hedge funds shows positive abnormal returns.
A further study by Abugri and Dutta (2009) also shows that emerging market hedge funds are
unable to consistently outperform their underlying benchmarks. Kotkatvuori-Örnberg,
Nikkinen, and Peltomäki (2011) hypothesize that the poor aggregate performance of
emerging market hedge funds may be due to lack of focus of these funds. Their results
suggest that a portfolio of emerging market hedge funds, which are geographically focused,
outperform their underlying stock market indices.
2.3 Performance persistence
The issue of persistence is of critical importance to investors. If there is evidence of
persistence in performance, they would allocate funds to those with superior past
performance and punish poorly performing funds through redemption. There are a vast
number of studies that examines persistence in performance of mutual funds.
The traditional approach to detect persistence involves splitting a fund’s performance
into two periods and utilizing the rank correlation coefficient to investigate the statistical
11
relationship between performances of the two periods. A positive correlation would indicate
persistence is evident. Sharpe (1966) rank performance based on Sharpe ratio for two decades
and reports a rank correlation of 0.36 but no statistically significance. Jensen (1968) finds
evidence of positive correlation in alphas but it is largely attributed to persistence of inferior
performance. Lehmann and Modest (1987) report some degree of persistence but find it to be
highly dependent on the performance measure employed. Elton et al. (1993) also provide
evidence of positive persistence but mostly among underperforming funds.
A second method to examine persistence in performance is by sorting winners and
losers over successive periods and determining the statistical significance of repeated winners
or losers. Goetzmann and Ibbotson (1994) employ this methodology to examine performance
persistence of funds between 1976 and 1988. Their results suggest that past performance do
entail some predictive power on future performance across various time periods tested.
Brown and Goetzmann (1995) find clear evidence of persistence but most of this
phenomenon is attributed to underperformers. Malkiel (1995) finds considerable persistence
in funds during the 1970s but not in the 1980s. Lastly, Droms and Walker (2001) test for long
term and short term persistence over two decades. They find evidence of short term but not
long term persistence.
Persistence can also be estimated by regressing current performance of a fund on its
lags. Hendricks, Patel and Zeckhauser (1993) follow this method to estimate quarterly
performance of mutual funds by incorporating up to eight lags. Their results show significant
positive persistence in the first four lags but insignificant negative persistence for the
following four lags suggesting persistence up to one year. Quigley and Sinquefield (2000) use
the FF model with yearly lags to detect persistence in the performance in UK equity funds.
12
They report evidence of persistence only among the losers and largely concentrated on small
cap funds.
Finally, another strand of literature uses the rank portfolio approach to examine
persistence. This is done by forming winner-loser portfolios of funds based on ranking
criteria such as past returns or alphas. Consequently, a time series of returns or alphas is
generated from the portfolios and persistence is determined by their ability to outperform the
benchmark in the post-formation period. Hendricks et al. (1993) sort mutual funds into
octiles based on past performance and find evidence short term persistence, particularly
among poor performers. Overall, they find stronger support for inferior persistence than
superior persistence. Elton, Gruber and Blake (1996) investigate persistence of funds by
forming deciles and affirm persistence over one year. Carhart (1997) estimate performance
using a four factor model and suggest that size and momentum factors can explain short-term
persistence in performance. Again, persistence is more visible amongst poor performers.
Blake and Timmermann’s (1998) study of UK mutual funds supports persistence in
performance. In contrast to most other studies, they observe persistence amongst both top and
bottom performing funds. Finally, Bollen and Busse (2005) find evidence of persistence
among top decile funds over one quarter post ranking but not over longer periods.
The question of persistence of performance among diversified emerging market funds
has been barely investigated. Among the two notable studies, Gottesman and Morey (2007)
find no evidence to support any relation between past and future performance.4 The other
study is by Huij and Post (2011) who examine persistence of 137 diversified emerging
market mutual funds between 1993 and 2006. The authors find evidence of short term
persistence in the return spread between winner and loser funds. Interestingly, they find that 4 The authors find expense ratios of funds as the only significant predictor of their future performance. Funds with lower expenses, on average, show better performance.
13
the contribution of winner funds to persistence is significantly higher in emerging markets
compared to US market.
3. Data and Methodology
3.1 Sample Data
The sample data comprises of all open-end5 (including SICAV and ICVC6) emerging market
mutual funds available on Morningstar Direct’s database over the period from August 2000
to July 2010. These mutual funds are equity funds with a small portion of assets invested in
cash (approximately 6%). Whilst these funds are classified under diversified emerging
markets covering Asia, Latin America, Eastern Europe, Africa, and Middle Eastern countries,
a small proportion of assets are allocated to developed markets.7 The domiciles of these funds
are also diverse including the US, UK, Luxembourg, Sweden, Denmark, Finland, Spain,
Italy, Ireland, Australia, Austria and Chile. For funds whose assets are denominated in
currencies other than US Dollar, we convert them into equivalent US Dollars using prevailing
exchange rates.
To tackle survivorship bias in our sample we include all funds that have at least 12
months of returns reported during the sample period with some exceptions. We omit funds
without any reported returns as well as those with fragmented return records. Regional or
country specific funds (for example, JP Morgan Middle East Africa or BNP Brazil Equity)
are also excluded. For funds under the same management with identical or near-identical
returns, only the fund with the oldest history is selected. The resultant sample consists of 498
5 Open-end refers to funds that continuously offers new shares to investors and redeems them on demand. Close-end funds, however, issues limited shares and does not redeem. Bekaert and Urias (1999) find open-end funds track the emerging market benchmark index better than closed-end funds. 6 SICAV and ICVC are Western European and UK equivalent of open-end funds. 7 We do not include funds that invest in specific regions such as Africa, the Middle East and Eastern Europe because it can produce alpha estimates resulting from model misspecification rather than managerial skill.
14
funds comprising 202 funds with full data over the sample period (i.e. surviving funds with
full data), 72 funds that became defunct (i.e. non surviving funds) and 224 funds that were
incepted during the sample period and are still in existence (i.e. new funds that survived).
Monthly net returns are calculated by taking the change in monthly net asset value for
the fund, reinvesting all income capital gains distribution during that month, less the expenses
incurred during that month such as operating, management and other asset related costs, and
then dividing by the starting net asset value. Monthly gross returns are then derived by adding
back the most recent expense component to net returns. The returns are not adjusted for any
sales charges such as loading fees and redemption fees.
Table 1
Descriptive statistics for fund returns during the sample period. This table provides descriptive statistics of gross returns for mutual funds during the sample period (August 2000-July 2010). The reported minimum (Min), maximum (Max), median, mean, standard deviation (Std. Dev.), are calculated from each funds’ average returns, while fund size and net expense ratios are also averaged over the sampled period.
Number of Funds
Returns Fund Size ($
Million)
Net Expense
Ratio (%) Min (%)
Max (%)
Median (%)
Mean (%)
Std. Dev. (%)
All Funds 498 -5.71 3.39 1.19 1.05 0.75 714.83 2.08
Surviving Funds (Full Data) 202 0.44 2.06 1.22 1.22 0.22 1,418.65 2.19
Surviving Funds (New) 224 -2.12 3.38 1.17 1.00 0.72 271.30 1.96
Non Surviving Funds 72 -5.71 3.39 1.06 0.66 1.38 82.98 2.16
Table 1 provides summary statistics of the funds in our sample. Monthly returns for
funds that are operational through the full sample period average 1.22% with a standard
deviation of 0.22%. All of these funds provide positive returns within the range of 0.44% and
2.06%, a much smaller spread compared to new funds and non-surviving funds. Furthermore,
these funds also represent largest fund size category with an average of $1.42 Trillion assets
under management. The average net expense ratio at 2.19% is slightly higher than the other
15
categories. A large number of funds in our sample are new funds i.e. funds that were incepted
during the sample period. For 224 such funds that survived till the end of the sample period,
the average return is slightly lower but standard deviation higher compared to the older
surviving funds. However, these newly established funds have a slightly lower average net
expense ratio of 1.96%. The funds that do not survive for the full period, not unexpectedly,
perform the worst. The average return for these funds is only 0.66% with a high standard
deviation of 1.38%. The range of returns for this category is from 3.39% to -5.71%.
3.2 Methodology
We use two performance measurement models in our empirical analysis. The first is the well-
known Jensen’s alpha based on the single factor Capital Asset Pricing Model (CAPM).
Initially the following regression is performed following Jensen (1968).
(1)
Where is the monthly return of the fund i in the month t, is the one month is the risk-
free rate of return and is the monthly return on the benchmark index. The Jensen’s alpha
is given by which gives the excess return of the fund that is unexplained by the excess
return of the benchmark index.
Second, we apply the FF three factor model of Fama and French (1993) to regress
returns of the fund to allow for risk adjustment after controlling for ‘size’ and ‘value’ factors.
. . (2)
In addition to the excess return of market index as in (1), the FF model captures the excess
return of small stocks over large stocks (SMB) and that of value stocks over growth stocks
(HML).
16
If emerging market fund managers have market timing ability, they will switch
portfolios to high or low beta securities depending on whether future returns are expected to
be high or low respectively. Unless this timing ability is explicitly accounted for in our
performance measurement model, the beta coefficient will increase (decrease) in case of
positive (negative) forecast of excess market return and as a result systematically downward
bias the alpha estimate. In order to avoid the possibility of such bias, we also conduct
nonlinear regressions of realized returns of the fund against contemporaneous market returns
following Henriksson and Merton (1981). This also enables us to explicitly test for market
timing ability of the funds.8 The Henriksson-Merton market-timing measure allows for the
beta risk to be different in rising and falling markets. Specifically,
∗ (3)
where D is a dummy variable that equals 1 for (RMt - Rft) > 0 and zero otherwise, and and
are the selectivity and market timing coefficients respectively. Under the null hypothesis of
no market timing is expected to be zero, whereas for a successful market timer it should
exhibit a positive value.
Finally, to investigate persistence in fund performance, we employ the rank portfolio
approach. Starting November 2000, we rank individual funds every month based on their
average cumulative excess returns over the previous quarter and group them into deciles. The
portfolios are equally weighted comprising the winners (decile 10) through to the losers
(decile 1), and these are held for 3 months and 6 months following formation. The portfolios
are revised every month as fund rankings change thus generating a time series of monthly
8 An alternative test of market timing ability is given by the quadratic regression model of Treynor and Mazuy (1966).
17
excess returns for each decile portfolio from. Funds that do not survive till the end of our
sample period are included until they disappear. We compute the alphas for the decile
portfolios by regressing their excess returns against excess returns of the market index as per
(1). A positive and statistically significant difference between the top and bottom decile alpha
estimates would suggest persistence.
We use the MSCI Emerging Markets Investable Index denominated in US
dollars as the benchmark for emerging market returns. The index aims to capture 85% of total
stock market capitalisation of global emerging markets. We use the one month US Treasury
bill rate as the proxy for the risk-free rate. Currently there is no available data for Fama
French factors for emerging markets as a whole. Cremers et al. (2010) suggest constructing
factors based on common and easily tradable size and style indices. Common style indices
like MSCI are not only convenient to use for ‘emerging markets’ (which cover many national
markets) but can also provide for construction of reasonably good proxies for the Fama
French factors. Dyck, Lins and Pomorski (2013) use local MSCI indices to calculate local
Fama French factors in their study of the international mutual fund industry. Cuthbertson and
Nitzsche (2013) also do the same in their study of the German equity mutual fund industry.
More recently, Bruckner, Lehmann, Schmidt, and Steale (2014) include SMB and HML time
series based on MSCI indices for Germany.
Morningstar provides four subsets of MSCI emerging market indices which replicate
size segment and style segment indices. The MSCI Emerging Markets Large Cap Index
includes large cap emerging market firms with approximately 70% of market capitalization
coverage (MSCI Barra, 2010). Similarly, the MSCI Emerging Markets Small Cap Index
offers exposure to small cap companies across the countries listed on the MSCI emerging
market index aggregating to 15% market capitalization coverage. We proxy the SMB return
18
premium using the difference between monthly returns of these two indices. As for the HML
premium, we derive the proxy by taking the difference in returns generated between the
MSCI Emerging Markets Value Index and MSCI Emerging Markets Growth Index.9
4. Empirical Results
4.1 Performance of All Funds
Table 2 provides the mean returns for all 498 mutual funds during the full sample period.
Monthly excess returns of an equally weighted portfolio of all funds are regressed against
monthly excess benchmark returns using the single factor CAPM (panel A1) and the three
factor FF model (panel A2). Under the two models, the average fund underperformed the
benchmark by 0.11% and 0.15% per month (1.33% and 1.74% per annum) respectively.
However, neither of the alphas is significantly different from zero. The market beta
coefficient under both models is in excess of 0.96 suggesting most of the funds closely track
the returns of the MSCI Emerging Markets index. The SMB and HML coefficients in the
three factor model are statistically insignificant. The R2 for both the models indicate that
more than 90% of the variation in the emerging market funds’ returns is explained by the
broad market movement. This may indicate emerging market funds generally are more
focused on achieving diversification rather than taking active bets in these markets. The
insignificant coefficients for SMB and HML factors also suggest that emerging market funds
in general do not have any significant size or value/growth tilt in their portfolios.
9 The variables to assemble the value characteristics of the companies listed in MSCI emerging markets index are book value to price ratio, price to earnings ratio and dividend yields. As for growth characteristics, the determinates are book value to price ratio, long term forward earnings per share, short term forward earnings per share growth rate and may also consider the current internal growth rate and long term historical earnings per share trend (MSCI Barra, 2007).
19
Panel B1 and B2 provides the regression results of fund quartiles sorted by alpha of
individual funds. Each quartile estimates represent those of an equally weighted portfolio of
funds within each quartile. The average first quartile fund outperforms the market but not by
a statistically significant margin. The outperformance is also not large enough to outweigh
the negative impact of the fourth quartile funds as these underperform the index by 0.64%
and 0.71% per month (7.71% to 8.57% per annum) under the two models. The spread in
alphas between the first and fourth quartile funds are strongly significant.
Table 2
Average gross returns based performance: all funds. This table reports the regression estimates for an equally weighted portfolio of 498 funds in our sample between August 2000 and July 2010. Panel A1 and A2 depicts the regression estimates under the single factor CAPM and Fama and French (FF) three factor model for the entire sample respectively. Panel B1 and B2 reports regression estimates for fund quartiles formed by ranking the alphas of individual funds from the two models. Q1, Q2, Q3 and Q4 represents first, second, third, and fourth quartiles respectively. The spread between the top and bottom quartiles is denoted by Q4 – Q1. ***, **, * indicates significance at 1%, 5%, and 10% level respectively.
α β(Rm - Rf) βSMB βHML Adjusted R2
Panel A1: Full Sample - CAPM Coefficient -0.1106 0.9662*** 0.9081 Std Error 0.2706 0.0328
T-Stat -0.2532 42.5174
Panel A2: Full Sample: FF Model Coefficient -0.1453 0.9601*** 0.0847 0.0132 0.9122 Std Error 0.2725 0.0339 0.1124 0.1995
T-Stat -0.3214 41.9454 0.5206 0.1167
Panel B1: Quartile Analysis: CAPM Q4 0.3110 1.0108*** 0.9074 Q3 0.0264 0.9829*** 0.9536 Q2 -0.1361 0.9609*** 0.9430 Q1 -0.6426 0.9104*** 0.8288
Q4-Q1 0.9536*** 0.1004*** 0.0786
Panel B2: Quartile Analysis: FF Model Q4 0.2659 1.0084*** 0.0458 -0.0065 0.9222 Q3 0.0088 0.9815*** 0.0396 0.0406 0.9526 Q2 -0.1414 0.9578*** 0.0678 0.0004 0.9453 Q1 -0.7134 0.8929*** 0.1849** 0.0182 0.8292
Q4-Q1 0.9793*** 0.1155*** -0.1391*** -0.0248* 0.0930
20
An interesting trend observed in table 2 is the reduction in the estimated market beta
from the first to fourth quartile. The lower quartile funds are slightly less volatile than the
benchmark index. The difference in market beta between the top and the bottom quartile
funds is significant. This could indicate that underperforming funds either overweight their
holdings of stocks in defensive sectors or hold higher proportion of cash relative to the top
performers. The latter may be caused by higher liquidity requirements if these funds confront
a higher redemption rate. The results also show significant difference between top and bottom
quartile funds in terms of their exposure to size factor. The bottom quartile funds demonstrate
a higher tilt towards large cap stocks.
Comparing our results with those of pervious studies on emerging market funds, Eling
and Faust (2010) report postive alpha with CAPM and a much lower underformance of -
0.48% p.a. for the FF model. However, their results may be subject to model misspecfication
due to their selection of S&P 500 as benchmark rather than any index representing emerging
market stocks. Huij and Post (2011) estimate CAPM alpha on the terciles formed from
ranked returns of the funds. The spread between top and bottom terciles are much smaller
(4.44% p.a.) compared with our results of top and bottom quartiles ranked by individual
fund’s alpha (11.44% p.a.). On the other hand, there are also some similarities between our
results and those of Huij and Post (2011). The size and value factor coefficients are
comparable with their results as they are insignificant. The R2 values are very similar with
their models explaining more than 90% of the variation in returns. The alphas for the bottom
tercile in their study are significant similar to our results for net returns (which we do not
provide here), where the bottom quartile alpha was statisitically significant.
4.2 Performance of surviving funds with full data
21
We separately evaluate the funds with full data to see whether these funds earn positive
abnormal returns and also to determine the magnitude of survivorship bias in the sample.
This approach to estimate survivorship bias by taking the difference between the alpha for
surviving funds with full data and that for the entire sample is consistent with previous
studies (see Malkiel, 1995; Elton et al., 1996; Brown and Goetzmann, 1995).
As shown in Table 3, the alphas are still negative but statistically insignificant. There
is also noticeable improvement in their estimates. The annualized alpha estimate is -0.07%
and -0.13% for the single factor and the three factor model respectively. All other coefficient
estimates remain consistent with table 2. Whilst, the top quartile alpha is close to its
corresponding estimate for all funds, the bottom quartile alpha improves considerably. The
spread between the top and the bottom quartiles is almost halved to around 0.55 (6.4% p.a.).
Even though the magnitude of underperformance is greatly reduced, the funds that survived
the entire sample period still on average do not outperform the market. It could be because of
the difficulty to exploit price anomalies as the older funds tend to be funds with large
capitalisation and therefore closely following the market index. Overall, our results show that
if our sample consisted of only these surviving funds, the performance would be overstated
by nearly 3% per year i.e. the results would be subject to severe survivorship bias.
Table 3
Average gross returns based performance: funds surviving full sample period. This table reports the regression estimates for an equally weighted portfolio of 202 funds that survived through the full sample period (August 2000 to July 2010). Panel A1 and A2 depicts the regression estimate under the single factor CAPM and Fama and French (FF) three factor model for the entire sample respectively. Panel B1 and B2 reports funds quartiles formed by ranking the alphas of individual funds from the two models. Q1, Q2, Q3 and Q4 represents first, second, third, and fourth quartiles respectively. The spread between the top and bottom quartiles are denoted as Q4 – Q1. ***, **, * indicates significance at 1%, 5%, and 10% level respectively.
α β(Rm - Rf) βSMB βHML Adj R2
Panel A1: CAPM
22
Coefficient -0.0061 0.9704*** 0.9337 Std Error 0.1658 0.0217
T-Stat -0.0313 53.9328
Panel A2: FF Model Coefficient -0.0107 0.9698*** 0.0329 0.0002 0.9368 Std Error 0.1655 0.0215 0.0709 0.1108
T-Stat -0.0536 54.4810 0.2942 0.1200
Panel B1: Quartile Analysis: CAPM Q4 0.2726 0.9765*** 0.8944 Q3 0.0518 0.9949*** 0.9578 Q2 -0.0673 0.9701*** 0.9550 Q1 -0.2816 0.9407*** 0.9285
Q4-Q1 0.5542*** 0.0358** -0.0341
Panel B2: Quartile Analysis: FF Model Q4 0.2415 0.9908*** 0.0127 0.0221 0.9139 Q3 0.0478 0.9750*** 0.0274 0.0383 0.9387 Q2 -0.0587 0.9781*** 0.0109 -0.0184 0.9606 Q1 -0.2731 0.9356*** 0.0802 -0.0408 0.9345
Q4-Q1 0.5145*** 0.0553*** -0.0675** 0.0628* -0.0207
4.3 Non-Surviving Funds
Table 4 summarises the performance of non-surviving funds i.e. funds that do not exist at the
end of our sample period. These funds underperform the benchmark by as much as 4.25% per
annum before becoming defunct. It is also obvious that the distribution of returns is highly
skewed towards the left with an alpha below -13% per annum. Another interesting point
concerns the beta coefficient spread between top and bottom quartiles. Comparing the results
with the survivor funds, the spread is much larger for non-surviving funds due to the much
lower systematic risk of the bottom quartile funds.
Table 4
Average gross returns based performance: non-surviving funds. This table reports the regression estimates for an equally weighted portfolio of 72 funds that were deceased during the sample period. Panel A1 and A2 depicts the regression estimate under the single factor CAPM and Fama and French (FF) three factor model respectively. Panel B1 and B2 reports funds quartiles formed by ranking the alphas of individual funds from the two models. Q1, Q2, Q3 and Q4 represents first, second, third,
23
and fourth quartiles respectively. The spread between the top and bottom quartiles are denoted as Q4 – Q1. ***, **, * indicate significance at 1%, 5%, and 10% level respectively.
α β(Rm - Rf) βSMB βHML Adj R2
Panel A1: CAPM Coefficient -0.3299 0.9424*** 0.8867 Std Error 0.3288 0.0408
T-Stat -0.8060 32.7504
Panel A2: FF Model Coefficient -0.3545 0.9372*** 0.0734 0.0054 0.8914 Std Error 0.3327 0.0425 0.1331 0.2254
T-Stat -0.7801 31.6340 0.3331 0.0026
Panel B1: Quartile Analysis. CAPM Q4 0.2238 1.0111*** 0.9367 Q3 -0.1292 0.9814*** 0.9508 Q2 -0.2867 0.9527*** 0.9293 Q1 -1.1273 0.8243*** 0.7298
Q4-Q1 1.3512*** 0.1868*** 0.2069
Panel B2: Quartile Analysis. FF Model Q4 0.2136 1.0200*** 0.0626 -0.0198 0.9534 Q3 -0.1333 0.9621*** 0.0528 -0.0143 0.9449 Q2 -0.3091 0.9304*** 0.0041 -0.0113 0.9329 Q1 -1.1891 0.8363*** 0.1742 0.0669 0.7344
Q4-Q1 1.4027*** 0.1837*** -0.1116** -0.0867* 0.2189
4.3 Performance during the financial crisis sub-period
We measure the performance of the funds exclusively for the 2-year sub-period (August 2008
to July 2010) following the onset of the global financial crisis (GFC), a period marked with
high volatility in most equity markets. The selection of this sub-period is important for two
reasons. First, it enables us to observe the impact of the crisis and associated volatile market
conditions on the performance of emerging market funds. If fund managers possess market
timing skill, they will correctly adjust the portfolio risk anticipating the market movements.
Second, given the findings by Moskowitz (2000) who documented that during recessionary
periods US funds perform better due to information asymmetries as compared to poor
24
performance in expansionary periods, it would be of interest to find out whether emerging
market funds’ performance is consistent with this phenomenon. 62 funds are omitted from the
sample as they are either defunct or are due to defunct without at least 18 months of returns
reported.
Table 5
Average gross returns based performance: sub-period August 2008 to July 2010. This table reports the regression estimates for an equally weighted portfolio for 437 funds that had 18 months of returns data reported during the sub-period August 2008 to July 2010. Panel A1 and A2 depicts the regression estimates under the single factor CAPM and Fama and French (FF) three factor model respectively. Panel B1 and B2 reports funds quartiles formed by ranking the alphas of individual funds from the two models where the reported coefficients and adjusted R2 are based on quartile averages. Q1, Q2, Q3 and Q4 represents first, second, third, and fourth quartiles respectively. The spread between the top and bottom quartiles are denoted as Q4 – Q1. ***, **, * indicates significance at 1%, 5%, and 10% level respectively.
α β(Rm - Rf) βSMB βHML Adj R2
Panel A1: All Funds: CAPM Coefficient -0.2716 1.0015*** 0.9443 Std Error 0.5122 0.0464
T-Stat -0.3741 34.5738
Panel A2: All Funds: FF Model Coefficient -0.3917 0.9819*** 0.1287 -0.0838 0.9446 Std Error 0.6431 0.0646 0.2864 0.4625
T-Stat -0.4312 26.6638 0.4713 -0.4349
Panel B1: Quartile Analysis: CAPM Q4 0.4316 0.9860*** 0.9632 Q3 -0.0216 0.9827*** 0.9743 Q2 -0.2873 0.9921*** 0.9726 Q1 -1.2070 1.0450*** 0.8675
Q4-Q1 1.6386*** -0.0590*** 0.0956
Panel B2: Quartile Analysis: FF Model Q4 0.3279 0.9875*** -0.0142 -0.0346 0.9640 Q3 -0.0489 0.9850*** 0.0812 -0.1476 0.9782 Q2 -0.3501 0.9663*** 0.1855* -0.2773 0.9673 Q1 -1.4925 0.9887*** 0.2623** 0.1220 0.8696
Q4-Q1 1.8205*** -0.0012 -0.2765*** -0.1566* 0.0944
25
Table 5 presents results for the sub-period from August 2008 to July 2010. On
average, funds seem to underperform the benchmark for both the models but the negative
alphas are not statistically significant. The results are comparable with Eling and Faust’s
(2010) analysis of emerging market funds during the Asian crisis in 1997. Their results
showed an annual underperformance of 2.06% and 1.95% for the CAPM and the FF models
respectively. A possible reason for these results is that emerging market funds can face higher
redemptions during crisis periods. Given the risky nature of emerging markets, investors are
more likely to switch, by selling off assets and switch to more stable and liquid assets in
developed markets. In contrast, money only starts to trickle in emerging markets after
conditions stabilize and returns improve. This asymmetric nature of the fund flow may prove
to be a hindrance to the managers’ ability to time the markets and outperform the benchmark
index.
The performance of the different fund quartiles during the crisis period is revealing.
The funds in the top quartile have a slightly higher alpha during the crisis relative to that for
the full sample period under both models. The rest of the quartiles, however, have inferior
performance. The bottom quartile funds appear to experience a particularly rough period with
the magnitude of their underperformance almost doubling during the financial crisis
compared the full sample period. As a result, the difference between the alphas of the top and
bottom quartiles increases almost two-fold.
Furthermore, the beta estimates across the quartiles show an opposing trend compared
to the full period analysis. The beta coefficients for the top quartile funds are slightly lower
than the market whereas those for the bottom quartile funds are slightly higher. Therefore, we
can only conclude that the top performers seem to have adjusted their portfolio risk perhaps
26
by shifting allocations into safer assets like cash whereas the bottom performers were not able
appropriately adjust their funds’ beta during market downturns.
4.5 Performance of individual funds
Apart from the equally weighted portfolio of funds constructed for all of the above analyses,
we replicate the analysis by conducting time series regressions for individual funds. Figure 1
illustrates the distribution of monthly alphas for all funds under the two models that depict
very similar distributions. The peak of the distribution shows that the majority of alphas are
clustered around zero. The alphas for most funds lie within the -0.25 and 0.25 range. The
distribution is also skewed towards the left suggesting there are more underperforming funds
than there are outperforming funds relative to their benchmark. While we do not report the
full descriptive statistics for the distribution, the median alpha estimate is negative (-0.28%).
When we look at the distribution of funds which survive the full period as well as the
distribution of funds that do not survive, the median alpha estimates turns out to be negative
(-0.05% and -0.45% respectively).
27
a) CAPM Alpha
b) FF Alpha
Fig.1. (a) Distribution of CAPM alphas. This figure shows the frequency distribution of the estimated regression intercepts using Capital Asset Pricing Model. (b) Distribution of FF alphas. This figure shows the frequency distribution of the estimated regression intercepts using Fama and French three factor model.
0
20
40
60
80
100
120
140
160
180
200
‐4 ‐3.5 ‐3 ‐2.5 ‐2 ‐1.5 ‐1 ‐0.5 0 0.5 1 1.5 2 2.5 3 3.5 4
0
20
40
60
80
100
120
140
160
180
200
‐4 ‐3.5 ‐3 ‐2.5 ‐2 ‐1.5 ‐1 ‐0.5 0 0.5 1 1.5 2 2.5 3 3.5 4
Monthly alpha estimate
Monthly alpha estimate
No.
of
fun
ds
No.
of
fun
ds
28
Table 6 summarises the percentage of individual funds with alphas that are
insignificant as well as the percentage that are significantly positive and negative under the
two models. The results illustrate that the largest proportion of funds in our sample have
insignificant alphas. For funds with significant alphas, the proportion of funds with
significantly positive alphas is less compared to the proportion of those with significantly
negative alphas. 10 However, Eling and Faust (2010) reported a higher percentage of
statistically positive alphas than negative alphas for both performance measures on emerging
market funds. As discussed before, their results may be influenced by their selection of a
portfolio of US stocks as the market benchmark index.
As an exception to the general trend in the results, among funds that survive the full
sample period, the proportion of significantly positive alphas is greater than that of
significantly negative alphas. The proportion of funds with positive (negative) alphas was
8.91% (7.92%) for CAPM and 8.42% (6.93%) for the FF model. This may suggest that funds
with long tenure, to some extent, may have the expertise in capturing postive risk-adjusted
returns at the gross level.
Table 6
Alphas for Individual Funds. This table presents the percentage of individual funds with alphas that are statistically (i) insignificant (=0), (ii) significantly positive (> 0) and (iii) significantly negative (< 0) at 5% level. Panel A and B report alphas based on CAPM and Fama and French (FF) three factor model respectively.
Percentage of Funds with Alpha
Total =0 > 0) < 0
Panel A: CAPM
All Mutual Funds 498 85.35 5.42 9.23
10 Although we do not present the results for alphas net of expenses, for majority of the funds that exhibit positive alphas, we find they turn insignificant when expenses are factored in. This is consistent with Grinblatt and Titman (1989), Malkiel (1995) and Wermers (2000) who all reported an increase in statistically significant negative alphas under net returns evaluation.
29
Full Data Funds 202 83.17 8.91 7.92 Non-Surviving Funds 72 79.16 4.17 16.67 Sub-Period 437 90.85 3.20 5.95
Panel B: FF Model
All Mutual Funds 498 84.94 5.62 9.44 Full Data Funds 202 84.65 8.42 6.93 Non-Surviving Funds 72 75.00 6.94 18.06 Sub-Period 437 94.28 1.60 4.12
For non-surving funds, surprisingly a small proportion has positive alphas. A possible
explanation could be that these funds do not represent those that are forced to close down due
to bad performance but in fact, are mereged or taken over, possibly because of their strong
performanc and consequent attractiveness. For the financial crisis sub-period, we have the
largest porporption of statistically insignficant alphas under both models.
4.6 Nonlinear Regression Estimates
Table 7 reports the estimates for nonlinear regression model based on Henriksson and Merton
(1981) model as represented by equation (3). For the aggregated sample of 498 funds shown
in panel A, we do not find any evidence of selectivity or market timing skill as both the alpha
and lambda estimates are not significant. When we conduct the same analysis based on
quartiles (panel B), the results do not change. None of the alphas or lambdas is statistically
significant although the difference between the alphas of top and bottom quartile funds
remain significantly positive for this model. However, the difference in lambdas between the
two quartiles is negative (although not significant) and that somewhat counterbalances the
positive difference in alphas.
Table 7
Market timing performance: 498 mutual funds for the period August 2000 to July 2010. This table reports the regression estimates based on equation (3). Panel A1 shows the regression estimates for an equally weighted portfolio of all 498 funds. Panel A2 shows the regression estimates for equally
30
weighted portfolio of fund quartiles formed by ranking the alphas of individual funds given by equation (1). Q1, Q2, Q3 and Q4 represents first, second, third, and fourth quartiles respectively. The spread between the top and bottom quartiles are denoted as Q4 – Q1. ***, **, * indicate significance at 1%, 5%, and 10% level respectively.
α β(Rm – Rf) Adj R2
Panel A1: Full Sample Coefficient 0.1595 1.0019*** -0.0925 0.9100 Std Error 0.4191 0.0529 0.1070
T-Stat 0.3989 27.1912 -0.7615
Panel B1: Quartile Analysis Q4 0.8463 1.0966*** -0.2796 0.8457 Q3 0.2465 1.0215*** -0.0798 0.9497 Q2 0.0194 0.9782*** -0.0255 0.9497 Q1 -0.4739 0.9113*** 0.0143 0.8943
Q4-Q1 1.3202*** 0.1852*** -0.2938 -0.0486
Turning to the market timing coefficients of individual funds, Table 8 reports the
percentages of funds with lambdas that are insignificant as well as of those that have
significantly positive and negative lambdas. Less than 2% of the funds show evidence of
market timing ability with significantly positive lambdas. On the other hand, more than 16%
of the funds report negative lambdas.11 For funds that survive the full sample period, the
estimated percentage of funds with positive lambdas barely increase but that of funds with
negative lambdas decrease by about 25%. The trends are opposite for the non-surviving funds
with decrease (increase) in percentage of funds with positive (negative) lambdas. A very
interesting result is observed for the estimates during the GFC sub-period. The percentage of
funds with positive lambda during this 18-month period is 8.30% i.e. more than 4 times of
that during the full sample period. There is corresponding fall in the number of funds with
insignificant lambdas. It suggests that some emerging market funds may have been more
active in their market timing activities during the financial crisis.
11 Adequate caution should be exercised in interpreting such a perverse outcome. One explanation is that higher cash inflows (outflows) from investors during rising (falling) markets bias market timing coefficients downwards (See, for example, Bollen and Busse, 2001)
31
Table 8
Market timing coefficients for individual funds. This table exhibits the percentage of individual funds with market timing coefficient (lambda) that is statistically (i) insignificant (=0), (ii) significantly positive (> 0) and (iii) significantly negative (< 0) at 5% level.
No. of funds
Percentage of Funds with Lambda
=0 > 0 < 0 All Mutual Funds 498 81.93 1.81 16.26 Full Data Funds 202 85.64 1.98 12.38 Non-Surviving Funds 72 79.17 1.38 19.45 Sub-Period 470 74.90 8.30 16.80
Overall, the absence of market timing ability for emerging market funds is not entirely
unexpected. As discussed previously, most of the funds may just seek to have exposure in
these markets by simply mimicking the index rather than generate excess returns by timing
the market. Our evidence of only a small proportion of funds having market timing ability is
consistent with the findings of bulk of empirical studies in the developed markets (see for
example, Wermers, 2000; Cuthbertson, Nitzsche, and O’Sullivan, 2010).12
4.7 Persistence in Performance
For measuring persistence in performance of emerging market funds, we form decile
portfolios based on past quarter’s performance as described in section 3.2. Table 9 reports the
average alphas for the decile portfolios over holding periods of 3 and 6 months after portfolio
formation. The results show persistence for both the best and the worst performers as they
continue to remain in the top and bottom deciles respectively. The annualised average alpha
spread between the top and the bottom decile portfolios is 4.5% for the 3-months holding
12 Borensztein and Gelos (2000) examine the behaviour of emerging market funds with respect to movement of funds between different individual markets. They find evidence that funds withdrew large sums from the affected country in the month prior to the crisis. However, in many cases, the withdrawn money was invested in other countries that were seen as suffering from contagion effects. The study did not examine how fund flow behaviour affected performance.
32
period which is statistically significant at 1% level. For the 6-months holding period, the
annualized spread of 2.96% exhibits weaker significance at 10% level.
We find that approximately 35% of the 3-month holding periods exhibit
statistically significant positive persistence. This decreases to 29% for holding periods of 6
months. Conversely, there are also some holding periods periods where bottom performers
significantly outperform top performers. The proportions of such periods of negative
persistence are 14% and 19% for the 3 and 6 months’ holding period respectively.
Table 9
Performance of mutual fund decile portfolios formed based on performance over past quarter. Funds are sorted every month between November 2000 to April 2010 into equally weighted decile portfolios based on their previous quarter’s average cumulative excess return. Funds with the highest and lowest returns over past 3 months comprise Decile 10 and Decile 1 respectively. The average monthly excess returns of each decile for next 3 months and 6 months are shown below. The spread between Decile 10 and Decile 1 is denoted ‘Highest – Lowest’ with t-statistics indicated in parenthesis. The last two rows represent the proportion of persistence during the individual months of the sample period. ***, **, * indicates significance at 1%, 5%, and 10% level respectively.
Average Excess Holding Period Alphas
Portfolio Decile 3 Months 6 Months
10 (Highest) 0.102 0.018 9 0.020 -0.093 8 -0.021 -0.089 7 -0.181 -0.083 6 -0.157 -0.100 5 -0.228 -0.132 4 -0.254 -0.168 3 -0.246 -0.216 2 -0.262 -0.202
1 (Lowest) -0.267 -0.226
Highest - Lowest 0.368*** (3.003) 0.244* (1.730) Proportion of Holding Periods
with Positive Persistence 35.09% (40 out of 114) 28.83% (32 out of 111) Proportion of Holding Periods
with Negative Persistence 14.06% (16 out of 114) 18.92% (21 out of 111)
33
Whilst our results support persistence in performance among emerging market mutual
funds observed by Huij and Post (2011), there are two important differences. Huij and Post
(2011) report the alpha spread between the top and bottom ninth of their fund sample to be
7.09% but we find the spread between the top and the bottom deciles to be significantly
smaller. One explanation of this difference could be our use of longer holding periods (3 and
6 months) for performance evaluation compared to the 1-month holding period in their study
as it is likely that the persistence would be more visible over shorter periods. This trend is
also apparent from our results as the return spread between the top and bottom decile
portfolios declines substantially over the 6-month holding period (relative to the 3-month
holding period) and is no longer statistically significant at 5% level.
The other point where we differ from Huij and Post (2011) is that we find that the
persistence in fund performance is driven more by the bottom performers than by the top
performers. Huij and Post (2011) claim the top ninth funds in their sample outperformed the
benchmark by 4.29% whereas the bottom ninth underperformed it by 2.8% on a risk-adjusted
basis. The corresponding annualized alphas for the top and bottom decile portfolios in our
study are 1.23% and -3.25% for the 3-month holding periods. Our result is consistent with the
evidence of persistence in the US like Carhart (1997) who find the lowest return funds to be
largely responsible for persistence in mutual fund performance and in UK where Cuthbertson
et al. (2010) find persistence among ‘loser’ funds but not among ‘winner’ funds.
34
Fig.2. Performance spread (cumulative) between best and worst performers. This figure shows the cumulative excess return spread between top and bottom decile funds ranked by past quarter’s performance.
Figure 2 demonstrates the cumulative excess return spread between top and bottom
decile funds ranked by their returns over the previous quarter throughout the entire sample
period. It is apparent that the cumulative excess return spread for the 3 months’ holding
period is generally higher compared to the 6 months’ holding period, indicating persistence is
stronger for shorter evaluation periods. The overall trend shows a gradual increase in the
spread over time, suggesting persistence in emerging market funds is not specific to a
particular sub-period. The exception to this is the period following the onset of the financial
crisis.13
5. Conclusions
13 One can conjecture that the spread net of fees and expenses would be smaller if the better performing ‘top decile’ funds increase their fees subsequently to reflect the value difference they create for investors compared to the underperformers. However, this is unlikely given the evidence of negative relationship between before-fee performance and fees documented in the mutual fund literature (see, for example, Gil-Bazo and Ruiz-Verdu, 2009).
‐5
0
5
10
15
20
25
30
35
40
Feb‐2001
Jul‐2001
Dec‐2001
May‐2002
Oct‐2002
Mar‐2003
Aug‐2003
Jan‐2004
Jun‐2004
Nov‐2004
Apr‐2005
Sep‐2005
Feb‐2006
Jul‐2006
Dec‐2006
May‐2007
Oct‐2007
Mar‐2008
Aug‐2008
Jan‐2009
Jun‐2009
Nov‐2009
Apr‐2010C
um
ula
tive
Ret
urn
Sp
read
(%
)
3MonthsHolding6MonthsHolding
35
The purpose of this study was to examine the ability of diversified emerging market equity
funds to produce risk-adjusted returns over and above a comparable market benchmark index.
Amongst studies of fund performance in developed markets, most have reported no superior
performance, which is consistent with one’s expectation in informationally efficient markets.
Emerging markets, on the other hand, are commonly considered to offer greater opportunities
for fund managers to exploit stock mispricing and information asymmetries. Our results do
not support this notion as we demonstrate that on an average diversified emerging market
funds do not outperform their benchmark index. Our analysis of individual funds shows that
nearly 95% of emerging market funds fail to outperform the benchmark index. Most of these
funds have market beta risk that is close to the benchmark, which may suggest that these
funds mainly aim to offer diversification benefits to investors rather than seek superior risk-
adjusted returns through active fund management. We test for market timing skills of funds
but find that less than 2% of funds show evidence of superior ability at a statistically
significant level.
We conduct separate analysis of fund performance for the period since the onset of
financial crisis which marks a period of significant volatility. Emerging market funds,
expectedly, have inferior risk adjusted performance during this period. However,
surprisingly, the average alpha of the top quartile funds is higher during this sub-period than
that during the full sample period. The other quartiles, particularly the bottom quartile, is
however significantly poorer during this sub-period. Interestingly, the proportion of funds
showing superior market timing ability increase significantly during the sub-period, although
it is still less than one in ten.
We also look for the evidence of persistence in emerging market funds’ performance.
Consistent with past studies evaluating mutual fund performance, we find evidence of short
36
term persistence among emerging market funds. But unlike the recent findings on persistence
in emerging markets by Huij and Post (2011), we find that the poorly performing funds play a
greater role in driving persistence than the top performers. In this respect, our results are
more in agreement with the evidence of persistence in developed markets.
Why do most diversified emerging market funds fail to outperform the benchmark?
There are two possible explanations. First, emerging markets in recent times may be no less
informationally efficient than developed markets as shown by Griffin, Kelly, and Nardari
(2010). In that case, fund managers in emerging markets will find it almost as difficult to
outperform the benchmark index on a risk-adjusted basis as their counterparts in the
developed markets. If funds face higher transaction costs in operating in emerging markets,
they will find it even harder to beat the market on a net basis. Second, a body of research has
shown domestic fund managers to have informational advantages over their foreign
counterparts (Shukla and van Inwegen, 1995; Bialkowski and Otten, 2011). Since the
diversified equity funds in our study are generally domiciled in developed markets and
mainly run by foreign managers, they might be at a disadvantage in exploiting any potential
inefficiency in emerging markets.
The findings of our study have important implications for investors. Our evidence
suggests that attempts to earn superior risk-adjusted returns by investing in diversified
emerging market equity funds are likely to be disappointed. Whilst emerging markets can
offer potential diversification benefits to investors in developed markets, these benefits are
sensitive to the particular investment vehicle used by the investor (Bekaert and Urius, 1999).
In light of our evidence, it appears that equity funds that diversify across multiple emerging
markets may not be the best vehicle to access such opportunities. In absence of superior risk-
adjusted performance (alpha) by these funds, investors would be better off by allocating to
37
ETFs which can provide similar diversification opportunities through beta exposure at a
lower cost. The other alternative that investors can consider is to invest in funds that focus on
individual emerging markets as there is some evidence in the hedge fund literature showing
that funds with a geographical focus do better than those that invest across multiple emerging
markets. It is quite possible that funds will have informational advantages when they have a
clear geographical focus as suggested by Kotkatvuori-Örnberg et al. (2011).
One limitation of our study deserves particular attention. We have not considered the
heterogeneity across different geographical regions or markets comprising the ‘emerging
markets’. Whilst all funds evaluated in this study hold diversified portfolios across these
markets, there may be individual funds with greater (lesser) exposure to certain regional
markets. Broad emerging market indexes like MSCI may not capture the true systematic risk
of such funds. Future research could look more closely into the specific asset allocation of
emerging market funds and evaluate their performance by constructing more appropriate
benchmarks.
References
Abel, E., Fletcher, J., 2004. An empirical examination of UK emerging market unit trust
performance. Emerging Markets Review 5, 389-408.
Abugri, B. A., Dutta, S. 2009. Emerging market hedge funds: Do they perform like regular
hedge funds? Journal of International Financial Markets, Institutions & Money 19,
834-849.
Bekaert, G., Urias, M., 1999. Is there a free lunch in emerging market equities?, Journal of
Portfolio Management 25, 83–95.
38
Bialkowski, J. Otten, R., 2011. Emerging market mutual fund performance: Evidence for
Poland. North-American Journal of Economics and Finance 22, 118-130.
Blake, D., Timmermann, A., 1998. Mutual fund performance: Evidence from the UK.
European Finance Review 2, 55-77.
Blitz, D., Huij, J., 2012. Evaluating the performance of global emerging markets equity
exchange-traded funds. Emerging Markets Review 13, 149-158.
Bollen, N., Busse, J., 2001. On the timing ability of mutual fund managers. Journal of
Finance 56, 1075-1094.
Bollen, N., Busse, J., 2005. Short-term persistence in mutual fund performance. Review of
Financial Studies 18, 569-597.
Borensztein, E., Gelos, G., 2000, A panic-prone pack? The behavior of emerging market
mutual funds. IMF Working Paper No. 00/198.
Brown, S., Goetzmann, W., 1995. Performance persistence. Journal of Finance 50, 679-698.
Bruckner, R., Lehmann, P., Schmidt, M., Steale, R., 2014. Fama/French factors for Germany:
Which set is best?. SSRN Working Paper
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2390063
Cai, J., Chan, K., Yamada, T., 1997. The performance of Japanese mutual funds. Review of
Financial Studies 10, 237-273
Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52, 57-82.
Cesari, R. Panetta, F., 2002. The performance of Italian equity funds, Journal of Banking &
Finance 26, 99-126.
Chan, K., Hameed, A., 2006. Stock price synchronicity and analyst coverage in emerging
markets. Journal of Financial Economics 80, 115-147.
Chang, E., Eun, C. S., Kolodny, R., 1995. International diversification through closed-end
country funds. Journal of Banking and Finance 19, 1237-1263.
39
Charitou, A., Makris, A., Nishiotis, G., 2006. Closed-end country funds and international
diversification. Multinational Finance Journal 10, 251-276.
Cremers, M., Petajisto, A., and Zitzewitz, E., 2010. "Should benchmark indices have alpha?
Revisiting performance evaluation", Working Paper.
Cumby, R., Glen, J., 1990. Evaluating the performance of international mutual funds. Journal
of Finance 45, 497-521
Cuthbertson, K., Nitzsche. D., O’Sullivan, N., 2008. UK mutual funds performance: Skill or
luck? Journal of Empirical Finance 15, 613-634.
Cuthbertson, K., Nitzsche. D., O’Sullivan, N., 2010. The market timing ability of UK mutual
funds. Journal of Business Finance and Accounting 37, 270-289.
Cuthbertson, K., Nitzsche, D., 2013. Performance, stock selection and market timing of the
German equity mutual fund industry. Journal of Empirical Finance 21, 86-101.
Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997. Measuring mutual fund
performance with characteristic-based benchmarks. Journal of Finance 52, 1035–
1058.
Droms W., Walker, D., 2001. Persistence of mutual fund operating characteristics: returns,
turnover rates, and expense ratios. Applied Financial Economics 11, 457-466.
Dyck, A., Lins, K. V., Pomorski, L. 2013. Does active management pay? New international
evidence. Review of Asset Pricing Studies 3, 200-228.
Eling, M., Faust, R., 2010. The performance of hedge funds and mutual funds in emerging
markets. Journal of Banking and Finance 34, 1993-2009.
Elton, E., Gruber, M., Das, S., Hlavka, M., 1993. Efficiency with costly information: A
reinterpretation of evidence from managed portfolios. Review of Financial Studies 6,
1-22.
40
Elton, E., Gruber, M., Blake, C., 1996. The persistence of risk-adjusted mutual fund
performance. Journal of Business 69, 133-157
Fama, E., French, K., 1993. Common risk factor in the return on stocks and bonds. Journal of
Financial and Economics 33, 3-56.
Fillis, G., 2006. Testing for market efficiency in emerging markets: Evidence from the
Athens Stock Exchange. Journal of Emerging Market Finance 5, 121-133
Gil-Bazo, J., Ruiz-Verdú, P., 2009. The relation between price and performance in the mutual
fund industry. Journal of Finance 64, 2153–2183.
Goetzmann, W., Ibbotson, R., 1994. Do winners repeat? Journal of Portfolio Management,
20, 9-18.
Gottesman, A., Morey, M., 2007. Predicting emerging market mutual fund performance.
Journal of Investing 16, 111-122.
Griffin, J., Kelly, P., Nardari, F., 2010. Do market efficiency measures yield correct
inferences? A comparison of developed and emerging markets, Review of Financial
Studies 23, 3225-3277.
Grinblatt, M., Titman, S., 1989. Mutual fund performance: An analysis of quarterly portfolio
holdings. Journal of Business 62, 393-416.
Gruber, M., 1996. Another puzzle: The growth in actively managed mutual funds. Journal of
Finance 51, 783-810.
Hayat, R., Kraeussl, R., 2011. Risk and return characteristics of Islamic equity funds.
Emerging Markets Review 12, 189–203.
Hendricks, D., Patel, J., Zeckhauser, R., 1993. Hot hands in mutual funds: Short-run
persistence of relative performance, 1974-1988. Journal of Finance 48, 93-130.
41
Henriksson, R., Merton, R., 1981. On market timing and investment performance II:
Statistical procedures for evaluating forecasting skills. Journal of Business 54, 513-
533.
Huij, J., Post, T., 2011. On the performance of emerging market equity mutual funds.
Emerging Markets Review 12, 238-249.
Ippolito, R., 1989. Efficiency with costly information: A study of mutual fund performance,
1965-1984. Quarterly Journal of Economics 104, 1-23
Jensen, M., 1968. The performance of mutual funds in the period 1945-1964. Journal of
Finance 23, 389-416.
Kaminsky, G., Lyons, R., Schmukler, S., 2001. Mutual fund investment in emerging markets:
An overview. The World Bank Economic Review 15, 315-340.
Karamera, D., Ojah, K., Cole, J., 1999. Random walks and market efficiency tests: evidence
from emerging equity markets, Review of Quantitative Finance and Accounting 13,
171-188.
Kothari, S., Warner, J., 2001. Evaluating Mutual Fund Performance. Journal of Finance
56, 1985–2010.
Kotkatvuori-Örnberg, J., Nikkinen, J., Peltomäki, J., 2011. Geographical focus in emerging
markets and hedge fund performance. Emerging Markets Review 12, 309-320.
Lee, S., 2001. Diversification benefits and investment performance of emerging market
funds: Evidence from closed-end country funds. PhD Dissertation, George
Washington University.
Lai, M., Lau, S., 2010. Evaluating mutual fund performance in an emerging Asian economy:
the Malaysian experience. Journal of Asian Economics 21, 378-390.
Lehmann, B., Modest, D., 1987. Mutual fund performance evaluation: a comparison of
benchmarks and benchmark comparisons. Journal of Finance 42, 233-265.
42
Malkiel, B., 1995. Returns from investing in equity mutual funds 1971 to 1991. Journal of
Finance 50, 549-572.
Matallin-Saez, J., 2006. Seasonality, market timing and performance amongst benchmarks
and mutual fund evaluation. Journal of Business Finance and Accounting 33, 1484-
1507.
Michelson, S., Philipova, E., Srotova, P., 2008. The case for emerging market funds. Journal
of Business and Economics Research 6, 81-88.
Mobarek, A., Mollah, A., Bhuyan, R., 2008. Market efficiency in emerging stock market:
Evidence from Bangladesh. Journal of Emerging Market Finance 7, 17-41.
Moskowitz, T., 2000. Mutual fund performance: an empirical decomposition into stock-
picking talent, style, transactions costs, and expenses: discussion, Journal of Finance
55, 1695-1703.
Movassaghi, H., Bramhandkar, A., Shikov, M., 2004. “Emerging” vs. “Developed” markets
closed-end funds: A comparative performance analysis. Managerial Finance 30, 51-
61.
MSCI Barra., 2007. MSCI Global Investable Market Value and Growth Indices.
http://www.msci.com/methodology/meth_docs/MSCI_May07_GIMIVGMethod.pdf
(Accessed on 08/08/2010)
MSCI Barra., 2010. MSCI Global Investable Market Indices Methodology.
http://www.mscibarra.com/eqb/methodology/meth_docs/MSCI_Sep10_GIMIMethod.
pdf (Accessed on 08/08/2010)
Otten, R., Bams, D., 2002. European mutual fund performance. European Financial
Management 8, 75-101.
Peltomäki, J., 2008. Emerging market hedge funds and the Yen carry trade. Emerging
Markets Review 9, 220–290.
43
Quigley, G., Sinquefield. R., 2000. Performance of UK equity unit trusts. Journal of Asset
Management 1, 72-92.
Ratner, M., Leal, R., 2005. Sector integration and the benefits of global diversification.
Multinational Finance Journal 9, 235-267.
Risso, W., 2009. The informational efficiency: the emerging markets versus the developed
markets, Applied Economics Letters 16, 485-487.
Sehgal, S., Jhanwar. M., 2008. On stock selection skills and market timing abilities of mutual
fund managers in India. International Research Journal of Finance and Economics,
15, 1-11.
Sharpe, W., 1966. Mutual fund performance. Journal of Business, 39, 119-138.
Shukla, R., van Inwegen, G., 1995. Do locals perform better than foreigners?: An analysis of
UK and US mutual fund managers. Journal of Economics and Business 47, 241–254.
Strömqvist, M., 2007. Should you invest in emerging market hedge funds? Working Paper.
Stockholm School of Economics.
Swinkels, L., Rzezniczak, P., 2009. Performance evaluation of Polish mutual fund managers.
International Journal of Emerging Markets 4, 26-42.
Treynor, J., Mazuy, F. 1966. Can mutual funds outguess the market? Harvard Business
Review 44, 131-136.
Wermers, R., 2000. Mutual fund performance: an empirical decomposition into stock-picking
talent, style, transaction costs, and expenses. Journal of Finance 55, 1655-1695.