Do mutual fund managers exploit the Ramadan anomaly? Evidence from Turkey

Emerging Markets Review 15 (2013) 211–232

Contents lists available at SciVerse ScienceDirect

Emerging Markets Review journal homepage: www.elsevier.com/locate/emr

Do mutual fund managers exploit the Ramadan anomaly? Evidence from Turkey Jędrzej Białkowski a, 1, Martin T. Bohl b,⁎, Philipp Kaufmann b, 2, Tomasz P. Wisniewski c, 3 a Department of Economics and Finance, College of Business and Law, University of Canterbury, Private Bag 4800, Christchurch 8140, New Zealand b Department of Economics, Westfälische Wilhelms-University Münster, Am Stadtgraben 9, 48143 Münster, Germany c School of Management, University of Leicester, University Road, Leicester LE1 7RH, United Kingdom

a r t i c l e

i n f o

Article history: Received 4 July 2012 Received in revised form 15 November 2012 Accepted 12 February 2013 Available online 24 February 2013 JEL classification: G11 G14 G23 Keywords: Mutual fund performance Ramadan effect Calendar anomaly Investor sentiment Behavioral finance Emerging markets

a b s t r a c t Recent literature shows that the holy month of Ramadan exerts a positive influence on investor sentiment in predominantly Muslim countries. This anomaly has been found to be particularly pronounced in Turkey. We therefore examine whether mutual fund managers investing in Turkish stocks are able to benefit from the Ramadan effect. We find that risk-adjusted performance of domestic institutional funds, hybrid funds and foreign Turkish equity funds is substantially higher during Ramadan compared to the rest of the year. By contrast, domestic index funds fail to deliver higher abnormal returns as they are adversely affected by increased money inflows during Ramadan. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Portfolio managers of actively managed mutual funds are expected to use their expert knowledge and skills to identify wealth-enhancing investment opportunities. In order to fully justify their remuneration ⁎ Corresponding author. Tel.: +49 251 83 25005; fax: +49 251 83 22846. E-mail addresses: [email protected] (J. Białkowski), [email protected] (M.T. Bohl), [email protected] (P. Kaufmann), [email protected] (T.P. Wisniewski). 1 Tel.: +64 3 364 3316; fax: +64 3 364 2635. 2 Tel.: +49 251 83 28200; fax: +49 251 83 22846. 3 Tel.: +44 116 252 3958; fax: +44 116 252 3949. 1566-0141/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ememar.2013.02.003

212

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

packages, they should strive to find mispriced securities, successfully time the market and uncover any existing market anomalies. The overarching objective of this paper is to examine whether they are indeed able to achieve such ambitious goals in practice. More specifically, we consider the behavior of Turkish mutual funds in relation to the Ramadan stock market seasonality. Astute managers should have been able to spot the market inefficiency arising during the ninth month of the Islamic lunar calendar and to design trading strategies capable of exploiting it. In doing so, they would have arbitraged away this seasonal effect and made the market more efficient. The existence of the Ramadan effect is well-documented in the extant literature. By considering a large sample of predominantly Muslim countries, Białkowski et al. (2009, 2012) show that stock returns are significantly higher during the Muslim holy month and the subsequent Eid al-Fitr festival. At the same time, stock investments exhibit less risk, as the volatility of returns is significantly reduced (see also Seyyed et al., 2005). These results have been further confirmed in a recent study of Middle East stock markets conducted by Al-Hajieh et al. (2011). Rationalizations offered for the existence of the seasonality have been essentially behavioral in nature. The collective experience of Ramadan has the potential to affect the mood of its observers, to enhance their sense of belonging and to promote solidarity in the Muslim world (Al-Hajieh et al., 2011; Białkowski et al., 2009, 2012; Odabaşi and Argan, 2009). Not only do religious convictions affect the psychology of the public, but they may also have ramifications for tangible economic outcomes. Baele et al. (2012), for instance, report that default rates on Islamic loans in Pakistan are only half of those on conventional loans. Interestingly, the authors document that defaults on Shariah-compliant loans are less likely to occur during Ramadan. In their ethnographic study, Sandikci and Omeraki (2007) reflect on the increasing commercialization of the holy month in Turkey. By drawing parallels to Christmas, they note that consumption patterns and marketing communications are noticeably altered during Ramadan. In a similar vein, Odabaşi and Argan (2009) argue that Ramadan has a deep social, cultural and economic impact on the daily life of Muslims in Turkish society. They emphasize that Ramadan has changed from a religious ritual to a holiday marked by feelings of nostalgia, a strong sense of communality and significantly higher consumer spending. We therefore expect the holy month of Ramadan to have a pervasive effect on the decision-making of Muslims, who constitute 99.8% of the Turkish population (CIA World Factbook, 2011). As Turkey is usually seen as a collectivistic country (Ayçiçegi-Dinn and Caldwell-Harris, 2011; Chui et al., 2010), we assume that Turkish investors are more consensus-oriented and inclined to carefully consider the opinions of their peers rather than relying solely on their own private information. Cultural collectivism generally strengthens social interdependence and thus plays a crucial role in further enhancing the upbeat sentiment during Ramadan. This paper contributes to the existing literature by being the first to examine the response of mutual funds to the occurrence of the Muslim holy month.4 While previous research has merely documented the existence of the Ramadan anomaly for a number of Middle East stock markets, our study goes one step further and investigates whether mutual funds investing in Turkey are able to benefit from this seasonality. We chose Turkey as our sample country for several reasons. First, Turkey has the second largest stock market in the Middle East region after the Tadawul in Saudi Arabia (FEAS, 2010; WFE, 2010) and exhibits a particularly pronounced market upturn during Ramadan (Al-Hajieh et al., 2011; Białkowski et al., 2009, 2012). With reference to the Istanbul Stock Exchange, Oğuzsoy and Güven (2004) also report strikingly high returns prior to religious holidays, such as Eid al-Fitr. Second, the literature on Turkish mutual fund performance, especially with respect to risk-adjusted timing ability, is still relatively scarce, even though the first mutual fund was launched as early as 1987. Interestingly, the domestic mutual fund sector gained considerable momentum at the beginning of the 2000s and still has untapped growth potential in the coming decades. Third, due to the country's close ties with Europe and the Middle East, many Western European institutional investors focus on Turkey to gain access to the MENA capital markets and to benefit from the international portfolio diversification generally offered by emerging market countries. While there have been restrictions on foreign equity ownership in a number of Muslim countries, the Turkish securities market was extensively liberalized and fully opened to foreign investment in 1989. Last, the choice of Turkey is also motivated by data

4 Menkhoff and Schmidt (2005) and Wright et al. (2008) are other examples of studies that examine whether mutual fund managers trade on behavioral anomalies.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

213

availability. Unfortunately, most commercial data vendors only provide mutual fund data for Muslim countries at a monthly frequency, whereas survivorship-bias free daily mutual fund data are available from the Turkish regulatory authorities. Daily frequency is essential when analyzing a moving calendar anomaly, such as the Ramadan effect. Our results indicate that many of the actively managed funds tended to increase their equity exposure around the time of Ramadan and the Eid al-Fitr festival. The risk-adjusted performance of mutual funds, in particular domestic institutional funds, larger domestic hybrid funds and foreign Turkish equity funds, was substantially enhanced during Ramadan for the time period between 2000 and 2011. Domestic index funds were the exception to this rule as they experienced a significant increase in fund flows from investors, who presumably wanted to participate in the stock market rally. The index funds, however, were rather slow to convert the cash inflows into stock holdings, which proved detrimental to their timing performance. Nevertheless, index funds still managed to earn the highest average raw return of all domestic fund groups during Ramadan as they maintained the highest equity exposure. Besides, the risk-return profile of Turkish mutual funds noticeably improved during Ramadan, as fund returns tended to be less volatile compared to the rest of the year. Investors' apparent increasing awareness of the Ramadan anomaly in Turkey is consistent with our findings of higher stock market turnover during this period and a positive but weakening return effect over time. However, the economically still relevant rise in stock prices, along with a significant drop in return volatility in recent years, continues to be a conspicuous phenomenon. The fact that the Ramadan anomaly has persisted in the data for a prolonged period of time is a testament to the power of investors' emotions and an indictment of the efficient market hypothesis. However, as market participants become increasingly aware of the existence of an anomaly, their actions change to reflect this awareness. Some anomalies can be substantially attenuated and may even vanish when they become publicly known (Schwert, 2003). This process of learning on the part of investors attests to the fact that they are not completely irrational. The remainder of the paper is organized as follows. Section 2 elaborates on the data sources and characteristics of the time series. Section 3 outlines the methodological approaches used in the study. We present our empirical results and their interpretation in Section 4. The paper ends with concluding remarks and reflections in Section 5.

2. Data 2.1. Turkish mutual funds We study daily data of domestic Turkish mutual funds and foreign mutual funds with investment focus on Turkish equities in the time period from January 2000 to March 2011. According to Takasbank, the settlement and custody bank of the Istanbul Stock Exchange, the year 2000 marks the beginning of a systematic and reliable recording of Turkish mutual fund data. The reason for using daily data is twofold. First, we intend to precisely capture the effect of a moving holiday, such as Ramadan, and second, Bollen and Busse (2001, 2005) show that a frequency of mutual fund data higher than monthly leads to more powerful timing tests. Our data set is free of survivorship bias. Mutual funds that have discontinued operations during our sample period are included until they disappear. Since the interest of this paper lies in the Turkish stock market, we collect information on Type A mutual funds that are legally required to invest at least 25% of their portfolio holdings in equities issued by domestic corporations. This domestic sample is obtained from Takasbank and consists of 166 open-end mutual funds. The Takasbank database contains the mutual funds' style classification, daily time series of total net assets and net asset values per share as well as details on the funds' asset allocation. With respect to their investment objective, 27 of the 166 mutual funds are classified as equity funds, 24 as balanced funds, 70 as variable funds, 34 as index funds and 11 as private funds. The regulations of the Capital Markets Board of Turkey distinguish between mutual funds that continuously invest at least 51% of their portfolio holdings in stocks (equity funds), at least 20% in both stocks and bonds (balanced funds) and at least 80% in the constituents of a particular benchmark stock index (index funds). Variable funds are not subject to any further investment restrictions, whereas private funds represent investment vehicles exclusively dedicated to a predetermined group of individual or institutional investors.

214

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

The index, balanced and variable Type A mutual funds represent by far the largest groups of our sample in terms of total net assets. As of 31/03/2011, the total size of the index fund group reached US$248.3 million, followed by balanced and variable funds having assets of US$243.2 million and US$231.6 million, respectively. Equity and private mutual funds represent a smaller group with total net assets of US $122.6 million and US$91.8 million, respectively. Overall, Type A mutual funds only account for a relatively small share of the Turkish mutual fund industry, whose total size amounted to US$20.11 billion by the end of 2010. According to the annual report of the Capital Markets Board of Turkey in 2010, total net assets of all Type A mutual funds reached US$1.02 billion, while Type B mutual funds, which primarily hold domestic fixed-income securities as they are not legally bound to invest a certain percentage of their portfolios in Turkish equities, managed assets worth US$19.09 billion. In comparison with developed capital markets, the Turkish mutual fund industry is still in its infancy. While in the U.S., for example, the size of the mutual fund market in 2010 corresponded to 81.1% of the Gross Domestic Product, the Turkish mutual fund industry represented only 2.7% in relation to Turkey's GDP in 2010.5 Nevertheless, the Turkish environment represents an excellent setting to investigate the exploitation of the Ramadan stock market rally by professional money managers, as relatively small mutual funds are expected to have greater flexibility in terms of changing their asset allocation. The sample of foreign equity funds investing in Turkey is collected from Thomson Reuters Lipper and comprises 30 open-end mutual funds domiciled in 16 countries, mainly on the European continent. For foreign mutual funds consisting of multiple share classes, we average returns across all tranches (see Comer et al., 2009). For ease of comparability, we express the performance of the foreign sample from the perspective of an investor residing in the euro area by converting the prices of all foreign equity funds into Euros. As of 31/03/2011, total net assets of the foreign Turkish equity fund sample amounted to US$1.89 billion, which is about twice the size of our domestic Type A fund sample. It is worth noting that foreign mutual fund investment is not hindered because the domestic securities market has been fully open to foreign institutional and individual investors since August 1989 (Bekaert and Harvey, 2000; Edison and Warnock, 2003).6

2.2. Benchmark factors and macroeconomic data Continuously compounded return data for benchmark indices as well as macroeconomic data are taken from Thomson Reuters Datastream. To compute daily excess returns, we employ two different interest rates. Since its introduction in August 2002, we use the Turkish Lira Reference Interest Rate (TRLIBOR) with maturity of one day provided by the Banks Association of Turkey. Prior to August 2002, we use the overnight middle rate from the Turkish interbank money market as announced by the Central Bank of the Republic of Turkey. For the foreign equity fund sample, the Euro Overnight Index Average (EONIA) serves as our proxy for the risk-free interest rate. The value-weighted market excess return and the factor-mimicking portfolios of our performance measurement models are constructed as described in Fama and French (1993) and Carhart (1997) as well as Schmidt et al. (2011) for pan-European stock markets. To calculate the market excess return Market, the size factor SMB, the book-to-market factor HML and the one-year price momentum factor PR1YR, we use all active and dead Turkish stocks traded on the Istanbul Stock Exchange, provided that return and balance sheet data are available on Thomson Reuters Datastream and Worldscope.7 The daily factor returns are converted into Euros for the foreign equity fund sample.

5 These figures are calculated based on information from the Capital Markets Board of Turkey, the Investment Company Institute and the World Bank. 6 We have no particular expectations regarding the performance of domestic Turkish fund managers versus fund managers located abroad. Informational disadvantages are generally held responsible for foreign investors' potential underperformance. However, Otten and Bams (2007) find no significant performance difference between domestic and foreign mutual funds investing in the U.S. equity market. Huij and Post (2011) even demonstrate that emerging market equity funds deliver stronger risk-adjusted returns than their domestic U.S. counterparts. Moreover, Borensztein and Gelos (2003) find evidence of herding behavior and positive feedback trading among managers of emerging market mutual funds. 7 We use constituents of the Thomson Reuters Datastream and Worldscope research lists FTURK, DEADTK and WSCOPETK. We eliminate any duplicates and closely follow the screening procedures suggested by Schmidt et al. (2011) to ensure high data quality.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

215

3. Methodology 3.1. Basic regression model First of all, we examine the magnitude of the effect that the holy month of Ramadan exerts on returns and conditional volatility in the Turkish stock market. To further account for a potential asymmetry in the conditional volatility of stock index returns, we employ a parsimonious GJR-GARCH(1,1) approach (Glosten et al., 1993) to estimate the following equations: r t ¼ β0 þ β1 Ramt þ β2 r World þ β3 r World t t−1 þ ε t 2 ht ¼ γ0 þ γ1 Ramt þ γ 2 εt−1 þ γ 3 ht−1 þ γ 4 ε2t−1 It ðεt−1 b0Þ :

ð1Þ

The continuously compounded percentage return of the ISE 100 Index, a capitalization-weighted price index considered as the benchmark for the Istanbul Stock Exchange, is denoted by rt. Furthermore, both the World contemporaneous and the lagged return of the MSCI All Country World Index, rtWorld and rt−1 , are used to control for global market movements. This equity index measures the performance of 45 developed and emerging markets, whose population is predominantly non-Muslim. It therefore represents an eventWorld independent benchmark. The lagged world market return rt−1 is included to account for time zone differences between the Istanbul Stock Exchange and other international stock markets. The residuals of the mean equation are denoted by εt ~ N(0, ht).8 Their conditional variance ht is assumed to depend on a Ramadan dummy, the lagged value of the squared residuals, its own lagged value and an asymmetric response to negative return innovations, where the indicator function It is equal to one if the lagged residual is negative (εt−1 b 0) and zero otherwise. To ensure that the conditional variance is stationary, the coefficient restriction γ2 + γ3 + 0.5γ4 b 1 must hold (Ling and McAleer, 2002). The coefficients β1 and γ1 measure the difference in mean return and conditional volatility during Ramadan compared to all other months of the Islamic calendar year. The dummy variable Ramt takes the value of one throughout the duration of the holy month and the days surrounding the festival that follows it. Abadir and Spierdijk (2005) and Białkowski et al. (2009, 2012) point out that stock returns in Muslim countries continue to rise after the three-and-a-half day celebration that marks the end of Ramadan. The Feast of Ramadan, known as Ramazan Bayramı or Eid al-Fitr in other Muslim countries, is a public holiday in Turkey during which the stock market is closed. In order to capture the positive post-festivity sentiment, the Ramadan dummy Ramt equals unity not only during the exact days of Ramadan, but also for a further seven days afterwards. The precise start and end dates of Ramadan are determined by the local lunar cycle. We reconstruct the exact lunar phases using an applet provided by the Astronomical Applications Department of the U.S. Naval Observatory, assuming that the astronomical observations are made from the geographical coordinates of the Turkish capital city of Ankara. This procedure is described in greater detail in Białkowski et al. (2009, 2012). To investigate whether the Ramadan effect has strengthened or declined over time, we also estimate a model similar to that presented in Chong et al. (2005): r t ¼ β0 þ β1 Ramt þ β2 Ramt T t þ β3 r World þ β4 r World t t−1 þ ε t ht ¼ γ0 þ γ1 Ramt þ γ 2 Ramt T t þ γ3 ε2t−1 þ γ4 ht−1 þ γ5 ε2t−1 It ðεt−1 b 0Þ :

ð2Þ

Tt is a time trend variable that captures the number of trading days that have elapsed between the beginning of the investigation in t = 1 and date t. Therefore, the signs of the coefficients β2 and γ2, respectively, indicate whether the effect of Ramadan on returns and conditional volatility has increased or declined over time. 8 Stock returns in emerging markets commonly exhibit heavy tails. Instead of assuming that the errors are conditionally normally distributed, we also experimented with Student's t-distribution and the Generalized Error Distribution (GED). Besides, we used the more local MSCI AC Europe & Middle East Index as our benchmark instead of the global MSCI AC World Index. The robustness checks yield quantitatively similar results and do not change our inferences.

216

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

3.2. Performance measurement models Our primary goal is to examine whether, and to what extent, domestic and foreign mutual fund managers exploit the Ramadan anomaly in Turkey. If this is the case, they should be able to generate higher riskadjusted returns during the holy month, which may arise from both security selection and market timing skills. As we are interested in the performance of the mutual fund industry as a whole, we construct equaland value-weighted portfolios of funds.9 The portfolios display the average performance of all mutual funds, in which investors were able to invest throughout the sample period. Based on the funds' daily net asset values per share, we compute continuously compounded total returns that are net of expenses and, where applicable, include any distribution of capital gains and dividends. 10 One way to exploit the positive investor sentiment during Ramadan is to successfully time the market. In general, timing activities can be described as an increase (decrease) of a mutual fund's systematic risk during positive (negative) market conditions. We measure the mutual funds' market timing skills by running multi-factor extensions of the timing models developed by Treynor and Mazuy (henceforth TM, 1966) and Henriksson and Merton (henceforth HM, 1981).11 Besides these two popular return-based timing models, another strand of literature has employed holdings-based approaches to assess market timing abilities (e.g., Elton et al., 2012; Jiang et al., 2007). However, a holdings-based approach is not feasible for our purposes since the domestic and foreign Turkish mutual funds do not report their exact portfolio holdings on a regular basis. Our extended unconditional TM timing model is specified as follows: TM ′TM ′TM TM 2 Rpt −r ft ¼ α TM 1p þ α 2p Ramt þ β1p Ft þ β 2p ðRamt Ft Þ þ γ 1p Market t


 TM 2 þγ 2p Ramt Market t þ εpt ;

ð3Þ

where Rpt − rft is the fund portfolio's daily percentage excess return over the overnight interbank rate and Ramt is the Ramadan dummy variable defined in the previous subsection. Selection ability in excess of the TM funds' exposure to commonly used benchmarks is measured by α1p for the non-Ramadan period, while TM TM α2p captures the difference during the Ramadan period. The coefficient vector β1p contains the sensitivity parameters with respect to the Fama and French (1993) and Carhart (1997) factors. Therefore, Ft represents a (4 × 1) column vector comprising the market excess return Market as well as the factormimicking portfolios SMB, HML and PR1YR. The return difference between small-cap stocks and large-cap stocks is denoted by SMB, while HML represents the return difference between high book-to-market equity stocks and low book-to-market equity stocks. PR1YR captures the return spread between a portfolio TM of past winner stocks and a portfolio of past loser stocks. The parameter vector β2p measures deviations from the fund portfolio's average factor exposures during the Ramadan period. TM The coefficient on the squared market excess return γ1p indicates the fund managers' ability to correctly TM forecast market returns during the non-Ramadan period, while γ2p captures any incremental timing activities during Ramadan. A positive timing coefficient indicates successful market timing and the fund portfolio's excess return will resemble a convex function of the market excess return. A negative timing coefficient, however, results in a concave relationship between the fund portfolio's excess return and the market excess return. Hence, the specification of the TM model assumes that the funds' systematic risk is a function of both the sign and the magnitude of the market excess return. To account for volatility clustering in daily mutual fund returns, the variance of the error term εpt is modeled via a GARCH(1,1) process. 12 It is well established that mutual fund managers may alter their factor sensitivities in response to changing macroeconomic conditions (e.g., Christopherson et al., 1998; Ferson and Schadt, 1996). Security selection and 9 For the foreign mutual fund sample, we only compute equal-weighted portfolio returns due to the limited availability of daily total net assets. 10 According to Takasbank, domestic Turkish mutual funds do not distribute dividends to their investors. Instead, they reinvest their current income and capital gains and can therefore be regarded as accumulation units (see also İmişiker and Özlale, 2008). 11 To maintain a feasible performance measurement model, we only assume that fund managers time their exposure to the market factor and therefore do not consider possible style timing activities. Findings from previous research also indicate that the Ramadan effect impacts mainly the overall market. 12 We are only interested in the coefficients of the mean equation. Hence, to reduce the number of estimated parameters we do not include the Ramadan dummy and the asymmetric effect of negative return innovations in the conditional variance equation.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

217

successful market timing based on publicly available information is generally not considered to be a superior fund manager's skill. We control for this effect by allowing for time-varying alphas and factor exposures. Similar to Ferson and Warther (1996), we use the dividend yield of the Turkish stock market and the one-month interbank interest rate as our macroeconomic instruments. These conditioning variables are detrended by subtracting a trailing 12-month average in order to reduce their relatively high degree of persistence (Ferson et al., 2003). The (3× 1) column vector zt−1 of the conditional TM timing model contains the Ramadan dummy Ramt and the deviations of the two lagged instruments from their unconditional means: ′TM

′TM

′ Rpt −r ft ¼ α TM 1p þ A p zt−1 þ β 1p Ft þ B p ðFt ⊗zt−1 Þ þ γ 1p Market t
 TM 2 þ γ2p Ramt Market t þ εpt ; TM

TM

2

ð4Þ

where the elements of the (3× 1) and (12 × 1) vectors ApTM and BpTM, respectively, measure the response of the conditional alpha and betas to the public information contained in zt−1. The symbol ⊗ denotes the Kronecker product and stands for the multiplication of each factor contained in Ft by the entire vector of information variables zt−1. In contrast to the TM model, the HM approach assumes that the mutual fund manager can only forecast the direction of the market excess return, but not its magnitude. The manager is assumed to adjust the portfolio to a higher target beta during up markets, whereas a lower target beta is chosen during down markets. Our extended variant of the unconditional HM model takes the following form: þ

HM ′HM ′ Rpt −r ft ¼ α HM 1p þ α 2p Ramt þ β 1p Ft þ β 2p ðRamt Ft Þ þ γ 1p Market t
 HM þ þ γ 2p Ramt Market t þ ε pt : HM

HM

ð5Þ

HM measures managerial timing activities Markett+ equals max(0, Markett) and thus the coefficient γ1p that can be interpreted as a call option on the market index with an exercise price equal to the overnight HM interbank rate. Analogous to the TM model, the coefficient γ2p detects the incremental change of timing activities during the Ramadan period. The conditional HM timing model is given by:

′HM

′HM



′ zt−1 þ β 1p Ft þ B p ðFt ⊗zt−1 Þ þ γ1p Market t Rpt −r ft ¼ α HM 1p þ A p  HM ′  þγ 2p zt−1 Market t þ εpt ; HM

HM

ð6Þ

where Market⁎t denotes the product of the market excess return and an indicator function. The indicator function equals unity if the difference between the market excess return and its conditional mean with respect to public information is positive, and zero otherwise. The conditional mean is estimated by an OLS regression of the market excess return on the information variables contained in zt−1 (see Ferson and Schadt, 1996). 3.3. Total performance measures A number of performance studies have shown that timing activities cause a mutual fund's alpha to be biased downward, especially since the latter reflects the transaction costs of implementing a timing strategy (Bollen and Busse, 2005). More important, there is substantial evidence of artificial timing if fund managers engage in option-like strategies (Jagannathan and Korajczyk, 1986) or if relevant benchmark factors are omitted in the performance measurement model (Matallín-Sáez, 2003). Spurious negative market timing may result from mutual fund flows that are positively related to expected market returns and that tend to temporarily increase cash holdings during market upturns (Edelen, 1999; Ferson and Warther, 1996). Since security selection and market timing performance are usually negatively correlated in the cross-section of mutual funds, we base our inferences regarding the fund managers' ability to exploit the Ramadan anomaly on the combined effect of selection and timing skills. To this end, we calculate the ex-post risk-adjusted total performance similar to the abnormal return measure of Bollen and Busse (2005), which has also been

218

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

employed by Huij and Derwall (2008) and Comer et al. (2009). We compute our daily total performance measures (TPMs) of the multi-factor timing models separately for the Ramadan and the non-Ramadan period. More precisely, the values of the TPMs depend on the dummy variable Ramt, which is equal to either one or zero. For the unconditional and conditional TM model, the TPM of a mutual fund portfolio is defined as: TM TPMRamt ;p

¼

^ TM α 1p

þ

^ TM α 2p Ramt

þ




^ TM γ 1p

þ

^ TM γ 2p Ramt

" # T  1X 2 Market t : T t¼1

ð7Þ

^ TM ^ TM ^ TM ^ TM The selection skill coefficients, α 1p and α 2p , as well as the timing coefficients, γ 1p and γ 2p , are estimated by Eqs. (3) and (4), respectively. Markett2 denotes the squared market excess return of the TM model, which is then averaged over all trading days T of our sample period. The TPM can be regarded as a fund portfolio's average daily abnormal return that includes not only the return contribution of stock picking and market timing, but also management fees and trading costs arising from active portfolio management. Similarly, we compute the TPM for the unconditional HM timing model of Eq. (5) as: " # T
 1X HM HM HM HM þ ^ ^ ^ HM ^ TPMRamt ;p ¼ α þ α Ram þ γ þ γ Ram Market 1p 2p t ; t t 1p 2p T t¼1

ð8Þ

where Markett+ is replaced by Market⁎t in case of the conditional HM timing model in Eq. (6). 13 The above TPMs lend insight into the fund managers' selection and timing abilities during both the Ramadan and non-Ramadan period. A positive and significant TPM can be interpreted as superior fund manager skill adjusted for the exposure to the Fama and French (1993) and Carhart (1997) factors. A negative TPM, however, may not necessarily be the result of inferior skills. It could also indicate that fund managers are not able to recoup the trading costs entailed by active portfolio management. To facilitate comparability, we also estimate the TPMs for the TM and HM timing models of Eqs. (3) through (6) while not including the Ramadan dummy Ramt. As a result, the corresponding TPMs gauge the mutual funds' abnormal return for the entire sample period without distinguishing between the months of the Islamic lunar calendar. The TPM for the timing models without Ramadan dummy is computed by: " ^p ^p þ γ TPMp ¼ α

# T 1X f ðMarket t Þ ; T t¼1

ð9Þ

where f(Markett) is a function of the market excess return, which takes the form of Markett2 for the unconditional and conditional TM model, whereas for the unconditional and conditional HM model it equals Markett+ and Market⁎t , respectively. 4. Empirical results 4.1. Ramadan effect in stock returns and volatility Before analyzing the mutual fund performance, we briefly discuss the effect that the holy month of Ramadan exerts on both returns and volatility of the Turkish stock market. Panel A of Table 1 presents the estimation results of Eq. (1) for the time period from January 1988 to March 2011, indicating that the daily return of the ISE 100 is on average four times higher during Ramadan and the trading week after the Feast of Ramadan compared to the rest of the year. The daily return differential of about 0.22% is statistically significant at the 1% level. The effect of Ramadan on conditional volatility is slightly negative, but negligible. The same holds true for the asymmetric response of the conditional volatility to negative return innovations measured by γ4. However, dividing the sample into two roughly equal sub-periods running from 1988 to 1999 and from 2000 to 2011, respectively, reveals that the Ramadan effect on Turkish stock returns has 13 We compute the significance of the total performance measures by following the procedure outlined in Grinblatt and Titman (1994) and Cesari and Panetta (2002).

Table 1 Performance results for the ISE 100. Panel A: The effect of Ramadan during different sub-periods β1

β2

β3

γ0

γ1

γ2

γ3

γ4

0.738⁎⁎⁎ (0.00)

0.213⁎⁎⁎ (0.00)

0.098⁎⁎⁎ (0.00)

−0.017 (0.50)

0.110⁎⁎⁎ (0.00)

0.881⁎⁎⁎ (0.00)

−0.001 (0.87)

(0.04)

0.253⁎⁎⁎ (0.00)

0.233⁎⁎⁎ (0.00)

0.500⁎⁎⁎ (0.00)

0.115 (0.29)

0.196⁎⁎⁎ (0.00)

0.764⁎⁎⁎ (0.00)

−0.006 (0.78)

Sub-period: 01/2000–03/2011 0.037 0.122 (0.30) (0.21)

0.866⁎⁎⁎ (0.00)

0.220⁎⁎⁎ (0.00)

0.044⁎⁎⁎ (0.00)

−0.036⁎⁎ (0.05)

0.048⁎⁎⁎ (0.00)

0.929⁎⁎⁎ (0.00)

0.026⁎⁎ (0.05)

0.561⁎⁎⁎ (0.00)

0.299⁎⁎⁎ (0.00)

0.151⁎⁎⁎ (0.00)

0.022 (0.58)

0.131⁎⁎⁎ (0.00)

0.858⁎⁎⁎ (0.00)

−0.001 (0.96)

Full sample period: 01/1988–03/2011 0.070⁎⁎ 0.216⁎⁎⁎ (0.02) (0.01) Sub-period: 01/1988–12/1999 0.130⁎⁎⁎ 0.371⁎⁎ (0.01)

Sub-period: 01/1988–12/2007 0.083⁎⁎ 0.280⁎⁎⁎ (0.02)

(0.01)

Panel B: The effect of Ramadan over time (01/1988–03/2011) β0

β1

0.070⁎⁎

0.471⁎⁎

(0.02)

(0.03)

β2

β3

β4

γ0

γ1

γ2

γ3

γ4

γ5

−6.316 (0.20)

0.738⁎⁎⁎ (0.00)

0.212⁎⁎⁎

0.096⁎⁎⁎

−5.270⁎⁎⁎

0.108⁎⁎⁎

(0.00)

(0.00)

0.203⁎⁎⁎ (0.01)

(0.00)

(0.00)

0.882⁎⁎⁎ (0.00)

0.001 (0.86)

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

β0

Panel A of this table reports coefficient estimates of Eq. (1) for the Turkish stock market index ISE 100, while Panel B shows the results of the estimation of Eq. (2). In Panel B, the coefficients β2 and γ2 are multiplied by 105. P-values in parentheses are based on robust Bollerslev and Wooldridge (1992) standard errors. ⁎, ⁎⁎ and ⁎⁎⁎ denote statistical significance at the 10%, 5% and 1% level, respectively.

219

220

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

declined over time. Even though the effect is no longer statistically significant for the later sub-period, it still maintains its economic relevance. In addition, while the Muslim holy month initially caused heightened stock market volatility, it now exhibits a dampening effect. The latter finding is consistent with the existing literature (e.g., Seyyed et al., 2005). Also, the asymmetric response coefficient to negative return shocks is positive and significant for the later sub-period, indicating that market declines tend to have larger impacts on volatility than market advances. Moreover, the sensitivities to the contemporaneous and lagged returns of the MSCI All Country World Index during both sub-periods suggest that the degree of integration of Turkey with global stock markets has considerably increased in recent years. With respect to the declining Ramadan effect in more recent years, it should be noted that the collapse of the U.S. investment bank Lehman Brothers in September 2008 falls within the Ramadan period of 2008. The ISE 100 lost 9.52% of its value during that period of time (even 20.76% including the trading week after the end of the holy month), indicating that the positive Ramadan effect was substantially obscured by contagion and spillover effects associated with the global financial crisis. We therefore also examine the period from 1988 to 2007, where the final year corresponds to the end of the sample period studied in Białkowski et al. (2009, 2012). Compared to the full sample period from 1988 to 2011, the results presented in the last row of Panel A in Table 1 underline that the weakening of the Ramadan effect primarily arises from the years after the beginning of the global financial crisis in 2008. In addition, the higher contemporaneous comovement with global equity markets, the vast media coverage of Ramadan's impact on stock returns following the publication of recent academic research and investors' attempts to exploit the anomaly may also have played a significant role in reducing the magnitude of abnormal returns in the last few years. Panel B of Table 1 displays the results of the estimation of Eq. (2) and confirms our general findings of a positive but weakening Ramadan effect on Turkish stock returns and an impact on conditional volatility that started out positive, but has substantially declined over time.

4.2. Asset allocation and fund flows The results of the previous subsection suggest that the positive investor sentiment during Ramadan translates into higher stock returns. To investigate whether mutual fund managers are aware of this calendar anomaly, we first analyze the funds' asset allocation, especially the percentage of portfolio holdings made up of equities and cash-like instruments. If fund managers actively tried to exploit the Ramadan effect, they would increase their exposure to domestic equities. Table 2 displays the average percentage of daily equity and cash holdings of the domestic mutual fund groups over the 2000 to 2011 period.14 The average percentage holdings are computed for the non-Ramadan period of the Islamic calendar year, for the month of Ramadan as well as for one trading week before and thereafter. The latter partition enables us to examine whether fund managers alter their asset allocation well in advance of the holy month and whether they maintain their equity exposure beyond the celebrations of Ramazan Bayramı to benefit from the post-holiday effect. In order to strengthen our statistical inferences for the mean differences between the fund holdings during the Ramadan and non-Ramadan periods, we perform a two-sided bootstrap t-test based on 1,000,000 replications as described in MacKinnon (2002). For a single bootstrap replication, fund holdings are randomly drawn with replacement so as to match the number of days during the Ramadan and non-Ramadan period in the original sample. As shown in column 2 of Panel A in Table 2, most of the domestic fund groups have rather moderate equity holdings, which reflects Type A fund managers' preference for investing in high-yield public debt securities or for engaging in reverse repo transactions. As argued by Bildik and Gülay (2007), the slight distaste for stocks of some of the fund groups could still be a result of the myopic behavior of Turkish investors, which was particularly prevalent during the 1990s and early 2000s. Persistently high levels of inflation and market volatility, as well as political and economic instability during that period of time, have caused domestic

14 Daily data on asset allocation and mutual fund flows are only available for the domestic fund groups, but not for the foreign mutual funds.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

221

Table 2 Mutual funds' asset allocation. Fund type

Non-Ram

Ram − 1

Difference

Panel A: Equity holdings during Ramadan and non-Ramadan Equity EW 74.13% 75.40% +1.27%⁎⁎ (0.03) Equity VW 72.23% 72.59% +0.37% (0.68) Balanced EW 43.19% 43.01% −0.18% (0.75) Balanced VW 38.07% 38.53% +0.46% (0.58) Variable EW 53.58% 54.24% +0.67% (0.28) Variable VW 53.41% 54.21% +0.79% (0.51) Index EW 91.74% 91.09% −0.65% (0.23) Index VW 91.80% 91.22% −0.58% (0.16) Private EW 62.00% 63.61% +1.62%⁎⁎⁎ Private VW

52.46%

53.07%

(0.00) +0.61% (0.33)

Panel B: Cash holdings during Ramadan and non-Ramadan Equity EW 22.62% 21.69% −0.93% (0.17) Equity VW 22.55% 22.84% +0.29% (0.77) Balanced EW 35.37% 35.97% +0.60% (0.64) Balanced VW 32.15% 31.61% −0.54% (0.78) Variable EW 34.11% 33.76% −0.35% (0.77) Variable VW 30.61% 29.75% −0.86% (0.55) Index EW 7.29% 7.17% −0.13% (0.65) Index VW 7.18% 6.90% −0.28% (0.40) Private EW 28.37% 26.34% −2.02%⁎⁎ (0.04) Private VW 29.54% 28.07% −1.47% (0.29)

Ram

Difference

Ram + 1

Difference

74.89%

+0.76%⁎⁎ (0.02) +0.40% (0.45) +0.11% (0.75) +0.27% (0.54) +0.71%⁎

75.84%

+1.71%⁎⁎⁎ (0.01) +1.99% (0.11) +1.59%⁎

72.63% 43.30% 38.34% 54.29% 55.06% 90.37% 90.81% 62.66% 53.01%

21.91% 22.88% 35.27% 32.07% 33.76% 29.37% 7.88% 7.94% 27.61% 28.62%

(0.06) +1.65%⁎⁎⁎ (0.00) −1.37%⁎⁎⁎ (0.00) −0.99%⁎⁎⁎ (0.00) +0.66%⁎⁎⁎ (0.01) +0.55% (0.14)

−0.71%⁎ (0.07) +0.33% (0.54) −0.10% (0.88) −0.08% (0.92) −0.35% (0.57) −1.24%⁎ (0.08) +0.59%⁎⁎⁎ (0.00) +0.75%⁎⁎⁎ (0.00) −0.75% (0.12) −0.92% (0.17)

74.22% 44.78% 40.94% 54.71% 55.86% 92.51% 91.99% 61.99% 55.14%

21.04% 20.51% 33.53% 28.70% 33.68% 28.21% 6.90% 6.99% 28.44% 27.48%

(0.08) +2.87%⁎⁎ (0.05) +1.13% (0.24) +2.45%⁎⁎ (0.05) +0.78% (0.29) +0.19% (0.79) +0.00% (0.99) +2.68%⁎⁎ (0.02)

−1.57%⁎ (0.09) −2.04% (0.19) −1.84% (0.27) −3.45% (0.13) −0.43% (0.79) −2.40% (0.17) −0.40% (0.40) −0.20% (0.78) +0.07% (0.95) −2.06% (0.19)

This table reports the average percentage of equity holdings (Panel A) and the average percentage of cash holdings (Panel B) for the domestic Turkish mutual funds around Ramadan (Ram) and the remaining months of the Islamic calendar year (Non-Ram) from 01/2000 to 03/2011. Cash holdings include T-bills, reverse repo transactions and money market instruments. Ram−1 denotes one trading week before the beginning of Ramadan, while Ram+1 represents one trading week after the holy month. Mean differences and their associated two-tailed bootstrap-t p-values in parentheses are computed in comparison to the non-Ramadan period (Non-Ram). ⁎, ⁎⁎ and ⁎⁎⁎ denote statistical significance at the 10%, 5% and 1% level, respectively.

investors to reduce their investment horizons and tilt their portfolios towards short-term deposits. In the recent past, however, the appetite for equity investment in Turkey has increased again. All fund managers, with the exception of those running index funds, tend to increase their stock holdings one trading week before the start of Ramadan as well as during the month of Ramadan. These rebalancing actions are particularly evident and significant for the equal-weighted equity and private fund groups as well as for the value-weighted variable fund group. The average equity holdings have the tendency to be even higher one trading week after the end of Ramadan, especially for the larger balanced, variable and private

222

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

Table 3 Daily mutual fund flows. Fund type

Ramadan

Difference

Equity EW

Non-Ramadan 0.02%

0.02%

Equity VW

−0.01%

−0.04%

Balanced EW

−0.01%

0.05%

Balanced VW

0.01%

0.03%

Variable EW

0.01%

0.06%

Variable VW

−0.05%

−0.02%

Index EW

0.06%

0.31%

Index VW

0.01%

0.23%

Private EW

0.01%

0.07%

Private VW

−0.01%

0.05%

0.00% (0.99) −0.03% (0.64) +0.06% (0.26) +0.02% (0.56) +0.05% (0.46) +0.03% (0.57) +0.25%⁎⁎ (0.04) +0.23%⁎ (0.09) +0.06% (0.36) +0.06% (0.34)

This table reports average daily mutual fund flows assumed to occur at the end of the day for the domestic Turkish mutual funds during Ramadan and the remaining months of the Islamic calendar year from 02/2000 to 03/2011. Mean differences and their associated two-tailed bootstrap-t p-values in parentheses are computed in comparison to the non-Ramadan period. ⁎, ⁎⁎ and ⁎⁎⁎ denote statistical significance at the 10%, 5% and 1% level, respectively.

funds. Apparently, fund managers are familiar with the strong investor sentiment promoted by the Feast of Ramadan and the tendency of stock prices to rise even further. With respect to fund size, the equal- and value-weighted statistics attest to the fact that it is not only the smaller funds that are able to considerably increase their stock holdings during the holy month. The larger funds within particular groups exhibit such behavior as well, even though they may have less flexibility in terms of changing their existing asset allocation. By contrast, the significant decrease in equity holdings for index funds is rather puzzling. This result is again mirrored in the significantly increased cash holdings during Ramadan, as presented in Panel B of Table 2. We measure cash holdings as the aggregate of the most liquid instruments in the fund portfolios, including Turkish treasury bills, reverse repo transactions and money market instruments. Domestic index funds are obviously not able to augment their equity exposure until after the Feast of Ramadan, as indicated by the level of equity holdings during the week after Ramadan in Panel A of Table 2. In order to interpret the surprising findings for index funds, it is worth taking a closer look at daily mutual fund flows during the Ramadan period. 15 As shown in Table 3, index funds are the only domestic fund group that experiences significantly higher money inflows during Ramadan. In general, inflows add to the funds' cash holdings, at least temporarily, if portfolio managers are faced with limited investment opportunities. The higher inflows suggest that individual investors in Turkey have become aware of the positive Ramadan effect in stock returns. They obviously opt for the most convenient way to participate in the temporary stock market rally by purchasing index funds, which aim to replicate the market movements of the Istanbul Stock Exchange as closely as possible and thus maintain the highest equity exposure of all Type A mutual funds. However, the increased inflows to index funds come at the cost of higher cash holdings that have a negative impact on the index funds' market beta and their potential timing ability (Bollen and Busse, 2001; Edelen, 1999; Ferson and

15 For a single mutual fund i, flows assumed to occur at the end of day t are computed as in Sirri and Tufano (1998) and Bollen (2007): FLOWi,t = [TNAi,t − TNAi,t−1 × (1 + Ri,t)] / TNAi,t−1, where TNAi,t denotes total net assets and Ri,t is the fund's discrete total return. Results are virtually identical if we assume the flows to occur at the beginning of the day. To reduce the impact of potential outliers, we winsorize individual fund flows at the 1st and 99th percentiles of the daily fund group's cross-sectional distribution, before averaging them across funds. We also exclude January 2000, the first month of our sample period, as the first days of this month seem to be affected by the initial compilation of the Takasbank mutual fund database.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

223

Warther, 1996). The reader should note that index funds have become one of the largest Type A fund groups in terms of total net assets, which might have made their portfolios less versatile when faced with new money inflows.

4.3. Multi-factor performance results The changes in the funds' asset allocation during Ramadan indicate that domestic mutual fund managers try to engage in market timing. In what follows, we investigate whether these activities translate into positive abnormal returns for investors. Given the results discussed above, we hypothesize that the index funds' attempts to time the market during Ramadan are thwarted by increased money inflows, leading to a negative risk-adjusted performance. Note that index funds are passive investment vehicles that generally do not engage in market timing. However, as reported in Panel B of Table 2, cash reserves of around 7% would theoretically provide some leeway for a temporarily higher equity exposure to boost the risk-adjusted performance during Ramadan. In addition, the average percentage of equity holdings of more than 90% could still enable index funds to earn at least higher raw returns. As Table 4 shows, this is indeed the case since the index fund group delivers the largest average raw return of all domestic mutual fund groups during the extended Ramadan period. Table 4 also reveals that the raw return difference compared to the non-Ramadan period is strongest for the index funds. Not surprisingly, it appears that the higher the average equity exposure, the more a mutual fund can benefit from the Ramadan stock market rally. In addition, the last column of Table 4 shows a mostly significant decrease in the conditional variance of daily mutual fund returns. This is consistent with our previous finding that stock market volatility declines during Ramadan.

Table 4 Average mutual fund returns and changes in volatility during Ramadan and non-Ramadan. Fund type

Mean return

Conditional variance

Non-Ramadan

Ramadan

Difference

Difference

0.094⁎⁎⁎ (0.00) 0.083⁎⁎⁎

0.265⁎⁎⁎ (0.00) 0.262⁎⁎⁎

+0.170⁎⁎⁎ (0.01) +0.179⁎⁎

Balanced VW

(0.00) 0.079⁎⁎⁎ (0.00) 0.074⁎⁎⁎

(0.00) 0.161⁎⁎⁎ (0.00) 0.163⁎⁎⁎

(0.02) +0.082⁎⁎ (0.04) +0.089⁎⁎

−0.032⁎⁎ (0.05) −0.020 (0.32) −0.010⁎⁎ (0.02) −0.009⁎⁎⁎

Variable EW

(0.00) 0.082⁎⁎⁎

Index VW

(0.00) 0.081⁎⁎⁎ (0.00) 0.083⁎⁎ (0.02) 0.085⁎⁎⁎

(0.00) 0.192⁎⁎⁎ (0.00) 0.214⁎⁎⁎ (0.00) 0.290⁎⁎⁎ (0.00) 0.304⁎⁎⁎

(0.02) +0.111⁎⁎ (0.04) +0.133⁎⁎ (0.02) +0.208⁎⁎ (0.03) +0.219⁎⁎

Private EW

(0.00) 0.099⁎⁎⁎

(0.00) 0.237⁎⁎⁎

(0.00) 0.101⁎⁎⁎ (0.00) 0.163⁎⁎⁎

(0.00) 0.207⁎⁎⁎ (0.00) 0.348⁎⁎⁎

(0.04) +0.138⁎⁎⁎ (0.01) +0.106⁎⁎ (0.03) +0.185⁎

(0.00)

(0.00)

(0.07)

Equity EW Equity VW Balanced EW

Variable VW Index EW

Private VW Foreign EW

(0.00) −0.017⁎ (0.07) −0.019 (0.17) −0.043⁎ (0.09) −0.044⁎ (0.10) −0.021⁎ (0.06) −0.013⁎ (0.08) −0.079 (0.21)

This table reports average daily mutual fund returns expressed as a percentage during the non-Ramadan and extended Ramadan period along with their corresponding differences. The mean returns are estimated by using a dummy variable regression whose error terms are assumed to follow a GARCH(1,1) process. The last column shows the average differences in the conditional variance of daily mutual fund returns between the extended Ramadan period and the non-Ramadan period. P-values in parentheses are based on robust Bollerslev and Wooldridge (1992) standard errors. ⁎, ⁎⁎ and ⁎⁎⁎ denote statistical significance at the 10%, 5% and 1% level, respectively.

224

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

Table 5 Summary statistics and cross-correlations for factor-mimicking portfolios. Factor Portfolio

Mean Return

Standard Deviation

t-Statistic for Mean = 0

Market SMB HML PR1YR

0.013% 0.001% 0.052%⁎⁎⁎ −0.003%

2.39% 1.23% 0.96% 0.90%

0.29 0.05 2.86 −0.18

Correlation Matrix Market

SMB

HML

PR1YR

1 −0.43 0.00 −0.08

1 −0.21 −0.01

1 −0.03

1

This table shows daily descriptive statistics for the Fama and French (1993) and Carhart (1997) factor-mimicking portfolios of the Turkish stock market during the period from 01/2000 to 03/2011. Market denotes the market excess return over the daily interbank interest rate, SMB and HML are proxies for the size and value effect, and PR1YR is a zero-investment portfolio capturing prior one-year price momentum. ⁎⁎⁎ Denotes statistical significance at the 1% level.

Table 5 reports descriptive statistics for the Turkish factor-mimicking portfolios used in the performance models. The average market excess return is positive but insignificant, reflecting the high level of short-term interest rates in Turkey. In line with van Dijk (2011), who documents the apparent disappearance of the size premium in international equity markets, we do not find a significant size effect for the Turkish stock market in our sample period. Akdeniz et al. (2000) have already pointed out that the negative and significant effect of firm size on average Turkish stock returns during the 1992 to 1995 period became insignificant in the late 1990s. Similarly, Gonenc and Karan (2003) show that large-cap stocks tend to outperform small-cap stocks in the 1993 to 1998 period. Consistent with Griffin et al. (2003), Bildik and Gülay (2007) and Chui et al. (2010), who all find a contrarian effect in the short run, our results reveal that a positive momentum effect does not exist in the Turkish stock market. The absence of short-term return continuation could be a consequence of the rather collectivistic culture in Turkey. As Chui et al. (2010) show, the magnitude of a country's momentum profits is positively related to the degree of individualism. Only the value premium turns out to be positive and highly significant, corroborating earlier findings of Akdeniz et al. (2000) and Bildik and Gülay (2007). Apparently at odds, Gonenc and Karan (2003) report that the average returns on value stocks are lower than those on growth stocks during the 1990s. Panel A of Table 6 reports the estimation results for the unconditional TM timing model. The relatively low market betas of balanced, variable and private funds as well as their adjusted R2 reflect the non-negligible portion of total net assets invested in short-term fixed-income instruments. By contrast, index funds seem to closely track the performance of the overall Turkish stock market. Furthermore, they load negatively on the SMB factor, indicating that their portfolios are primarily composed of large-cap stocks. All other domestic fund groups, however, significantly tilt their holdings towards small-cap stocks. The sensitivity to the HML factor is positive and significant for all domestic fund groups, except for the larger index funds. Given the positive value premium in the Turkish stock market (see Table 5), the slight tilt towards value stocks contributes positively to the performance of domestic Type A mutual funds. Moreover, domestic fund managers seem to be aware of the contrarian effect, as the exposure to the momentum factor PR1YR is negative and significant for most of the fund groups. The factor exposures of foreign Turkish equity funds are similar to those of the domestic fund groups. However, their portfolio holdings are neutral to the value and momentum factors, as displayed in the insignificant coefficients on HML and PR1YR. For the foreign mutual fund sample, we follow Otten and Bams (2007) and additionally control for currency effects. Fund managers could either hedge their foreign exchange rate exposure to the Turkish Lira or deliberately leave it uncovered in order to benefit from favorable exchange rate movements. To unveil these effects, all timing models incorporate both contemporaneous and lagged changes of the official ECB EUR/TRY rate as a proxy for overall movements of the Turkish Lira against other major currencies. The coefficient on the lagged changes is particularly negative and significant. For example, the unconditional TM model yields a coefficient of −0.54 (p-value is close to zero), indicating that the foreign mutual funds are on average substantially exposed to currency risks. This means that they tend to suffer from a depreciation of the Turkish Lira and benefit from an appreciation. The lower adjusted R2 also suggests that exchange rate effects might play a considerable role in the return variation of foreign equity funds investing in Turkey.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

225

The factor sensitivities during Ramadan do not reveal any substantial changes compared to the non-Ramadan period. However, most of the fund groups tend to load more strongly on the SMB factor, while their positive loadings on HML are somewhat reduced. The results for the conditional TM model in Panel B of Table 6 are mostly similar to those of the unconditional version in Panel A. 16 The combined effect of selection and market timing activities, denoted by the alpha and gamma coefficients, is measured as the total performance in Table 7. For the full sample period from January 2000 to March 2011, the total performance measures of both the unconditional and conditional timing models show that domestic fund managers are largely unable to outperform their benchmarks after costs. The TPMs for the equity, balanced and variable fund groups are insignificant and close to zero. Not surprisingly, index funds yield negative and significant abnormal returns, reflecting the expenses incurred in their passive investment strategies. Only private funds deliver positive and significant abnormal returns for their select group of institutional investors, which could be considered as evidence of superior skills of institutional money managers compared to managers of retail mutual funds. Our results are broadly consistent with İmişiker and Özlale (2008), who also find rather limited evidence for selection and market timing abilities of Type A mutual fund managers during the 2000 to 2003 period. The findings may therefore also be indicative of some degree of market efficiency in the Turkish mutual fund market. By contrast, the strongly positive and significant TPMs of foreign equity funds reveal that euro-based investors can earn sizeable risk-adjusted returns when investing in the Turkish stock market through these investment vehicles. As discussed earlier, a substantial part of the abnormal return might be due to favorable fluctuations between the Turkish Lira and other currencies. When we split the total performance evaluation into a Ramadan and non-Ramadan period, some striking patterns emerge. While the abnormal returns earned during the non-Ramadan period are mostly insignificant and negative for the majority of domestic funds, they are considerably higher – at least economically – during the holy month. The daily TPMs of the unconditional TM model of the larger balanced and variable funds, for instance, amount to a significant figure of 0.038% and 0.049%, respectively, while they are practically equal to zero during the non-Ramadan time. The corresponding return spreads that are gained during Ramadan are quite substantial for the two larger domestic hybrid fund groups. As reported in Table 8, they add up to significant +0.038% and +0.055%, respectively, and are even larger when we consider the unconditional HM timing model. The abnormal return during Ramadan is particularly pronounced for private mutual funds, indicating that their relatively small size increases their flexibility in terms of market timing. While the overall results reveal a rather moderate statistical significance, their economic relevance is quite compelling. Assuming 250 trading days per year, the equal-weighted equity fund group's daily Ramadan total performance of 0.044% (unconditional TM model) corresponds to an annualized abnormal return of 11.02%. With respect to the non-Ramadan period, this implies a return spread of 0.062% per day and 15.41% per year, which is already net of management fees and adjusted for exposures to commonly employed benchmarks. Note that the TPMs for the Ramadan period are slightly lower and often lose their statistical significance when we use the conditional timing models. This finding suggests that publicly available information plays a non-negligible role in assessing the performance of Turkish mutual fund managers. With respect to foreign Turkish equity funds, we find abnormal returns that are more than three times larger during Ramadan than during the remaining lunar months. Thus, the exploitation of this calendar anomaly is not only limited to domestic Turkish investors, but it also appears that European investors can take advantage of the market upturn. However, the lack of daily mutual fund holdings and asset allocation data for foreign mutual funds prevents us from exploring the importance that non-local fund managers attach to the anomalous return behavior during and after the Muslim holy month. Consistent with our prior conjectures, results obtained for the domestic index fund group differ from those obtained for other funds. Their abnormal returns, in particular those of the equal-weighted index fund group, are inferior during Ramadan. Not surprisingly, higher money inflows reported in the previous subsection deteriorate the timing performance of index fund managers, as they are faced with involuntarily increased cash holdings. However, the abnormal return differences with respect to the non-Ramadan period are not

16 The results for the unconditional and conditional Henriksson and Merton (1981) timing model are quantitatively similar to those obtained for the Treynor and Mazuy (1966) model. To conserve space, we do not report them, but they are available upon request.

226

Panel A: Unconditional Treynor–Mazuy timing model Fund type Equity EW Equity VW Balanced EW Balanced VW Variable EW Variable VW Index EW Index VW Private EW Private VW Foreign EW

α1TM −0.007 (0.51) −0.007 (0.55) −0.012⁎ (0.09) 0.002 (0.79) −0.014⁎ (0.08) −0.003 (0.72) −0.032⁎⁎⁎ (0.00) −0.023⁎⁎⁎ (0.00) −0.014 (0.12) 0.006 (0.36) 0.048⁎⁎⁎ (0.00)

α2TM

Market

SMB

HML

PR1YR

0.001 (0.97) 0.026 (0.51) 0.010 (0.71) 0.024 (0.33) 0.015 (0.63) 0.028 (0.39) −0.008 (0.61) 0.003 (0.86) 0.045 (0.17) 0.017 (0.54) 0.077 (0.21)

0.635⁎⁎⁎

0.073⁎⁎⁎

−0.047⁎⁎⁎

(0.00) 0.617⁎⁎⁎ (0.00) 0.362⁎⁎⁎

0.139⁎⁎⁎ (0.00) 0.086⁎⁎⁎ (0.00) 0.106⁎⁎⁎

(0.00) 0.038⁎⁎⁎ (0.00) 0.067⁎⁎⁎

(0.00) −0.025⁎ (0.06) −0.028⁎⁎⁎

(0.00) 0.340⁎⁎⁎

(0.00) 0.047⁎⁎⁎

(0.00) 0.032⁎⁎⁎

(0.00) 0.469⁎⁎⁎ (0.00) 0.465⁎⁎⁎

(0.00) 0.101⁎⁎⁎ (0.00) 0.092⁎⁎⁎

(0.00) 0.059⁎⁎⁎ (0.00) 0.054⁎⁎⁎

(0.00) −0.014⁎⁎ (0.05) −0.033⁎⁎⁎ (0.00) −0.020⁎⁎

(0.00) 0.975⁎⁎⁎

(0.00) 0.021⁎⁎⁎

(0.00) 0.978⁎⁎⁎ (0.00) 0.555⁎⁎⁎ (0.00) 0.463⁎⁎⁎

(0.00) −0.056⁎⁎⁎ (0.00) −0.123⁎⁎⁎ (0.00) 0.117⁎⁎⁎ (0.00) 0.106⁎⁎⁎

(0.00) 0.522⁎⁎⁎

(0.00) 0.074⁎⁎⁎

(0.00)

(0.00)

(0.00) −0.021⁎⁎⁎ (0.00) 0.064⁎⁎⁎ (0.00) 0.055⁎⁎⁎

(0.02) −0.003 (0.64) 0.017⁎⁎⁎ (0.00) −0.040⁎⁎⁎ (0.00) −0.021⁎⁎

(0.00) 0.010 (0.62)

(0.02) −0.001 (0.97)

Ramt × Market

Ramt × SMB

Ramt × HML

Ramt × PR1YR

γ1TM

γ2TM

0.002 (0.87) 0.003 (0.88) 0.023 (0.11) −0.011 (0.78) 0.014 (0.40) −0.001 (0.96) −0.016⁎⁎

0.098⁎⁎⁎

−0.135⁎⁎⁎

−0.002⁎⁎⁎

(0.00) 0.076⁎ (0.07) 0.087⁎⁎⁎

(0.00) −0.127⁎⁎⁎ (0.00) −0.053⁎

0.011⁎⁎⁎ (0.00) 0.005⁎⁎⁎ (0.00) 0.004⁎⁎⁎

(0.00) 0.045 (0.49) 0.068⁎⁎⁎ (0.00) 0.046⁎⁎

(0.06) −0.064⁎ (0.06) −0.065⁎ (0.07) −0.090⁎⁎⁎

−0.085⁎⁎ (0.05) −0.064 (0.35) −0.010 (0.71) −0.047⁎

(0.02) −0.006 (0.83) 0.040⁎⁎⁎ (0.00) 0.002 (0.88) 0.075⁎⁎⁎

(0.02) −0.015 (0.38) −0.005 (0.90) 0.133⁎⁎⁎ (0.00) 0.034 (0.16) 0.140⁎⁎⁎

(0.00)

(0.01)

(0.00) −0.007 (0.72) −0.036 (0.11) −0.073⁎ (0.06) −0.047 (0.11) −0.074 (0.20)

(0.10) −0.020 (0.52) −0.044 (0.15) 0.039⁎ (0.07) 0.051⁎⁎ (0.03) −0.006 (0.87) −0.039 (0.22) −0.101⁎⁎ (0.02)

(0.00) −0.001⁎⁎⁎ (0.01) 0.001 (0.33) 0.000 (0.61) 0.000 (0.81) 0.000 (0.57) 0.001⁎⁎⁎ (0.00) 0.001⁎⁎⁎ (0.00) 0.005⁎⁎⁎ (0.00) 0.000 (0.73) 0.000 (0.87)

(0.00) 0.002⁎⁎⁎ (0.00) 0.004⁎⁎⁎ (0.00) 0.005⁎⁎⁎ (0.00) −0.001 (0.66) −0.001 (0.65) 0.001 (0.52) 0.004⁎⁎⁎ (0.00) 0.001 (0.82)

2 Radj.

0.79 0.77 0.76 0.73 0.79 0.76 0.95 0.95 0.80 0.79 0.37

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

Table 6 Estimation results for Treynor–Mazuy timing model.

Panel A: Unconditional Treynor–Mazuy timing model Fund type

α1TM

α2TM

SMB

HML

PR1YR

Ramt × Market

Ramt × SMB

Ramt × HML

Ramt × PR1YR

γ1TM

γ2TM

0.631⁎⁎⁎ (0.00) 0.620⁎⁎⁎ (0.00) 0.363⁎⁎⁎

0.144⁎⁎⁎ (0.00) 0.100⁎⁎⁎ (0.00) 0.105⁎⁎⁎

0.081⁎⁎⁎ (0.00) 0.047⁎⁎⁎ (0.00) 0.066⁎⁎⁎

−0.037⁎⁎⁎ (0.00) −0.014 (0.28) −0.030⁎⁎⁎

−0.005 (0.71) −0.021 (0.48) 0.018⁎

0.049 (0.16) 0.005 (0.93) 0.072⁎⁎⁎

−0.126⁎⁎⁎ (0.01) −0.115⁎⁎⁎ (0.01) −0.059⁎⁎

(0.00) 0.049⁎⁎⁎

(0.00) 0.033⁎⁎⁎

(0.00) −0.015⁎⁎

(0.00) 0.054⁎⁎⁎

(0.00) 0.471⁎⁎⁎ (0.00) 0.468⁎⁎⁎ (0.00) 0.973⁎⁎⁎

(0.00) 0.100⁎⁎⁎ (0.00) 0.099⁎⁎⁎ (0.00) −0.057⁎⁎⁎

(0.00) 0.062⁎⁎⁎ (0.00) 0.061⁎⁎⁎ (0.00) 0.020⁎⁎⁎

(0.00) 0.982⁎⁎⁎ (0.00) 0.551⁎⁎⁎ (0.00) 0.461⁎⁎⁎

(0.00) −0.121⁎⁎⁎ (0.00) 0.117⁎⁎⁎ (0.00) 0.110⁎⁎⁎

(0.00) −0.024⁎⁎⁎ (0.00) 0.067⁎⁎⁎ (0.00) 0.058⁎⁎⁎

(0.04) −0.034⁎⁎⁎ (0.00) −0.017⁎ (0.08) −0.001 (0.80) 0.019⁎⁎⁎

(0.00) 0.505⁎⁎⁎

(0.00) 0.071⁎⁎⁎

(0.00) 0.063⁎⁎⁎ (0.01) 0.025 (0.31) −0.009 (0.74) −0.005 (0.81) 0.092⁎⁎⁎ (0.00) 0.039 (0.16) 0.125⁎⁎

(0.00)

(0.00)

(0.00) 0.021 (0.21)

(0.08) −0.011 (0.36) 0.007 (0.50) −0.014 (0.20) −0.014 (0.21) −0.011 (0.19) 0.026⁎⁎ (0.04) 0.007 (0.62) 0.070⁎⁎⁎ (0.01)

(0.02)

(0.03) −0.060⁎⁎⁎ (0.01) −0.072⁎⁎ (0.02) −0.097⁎⁎⁎ (0.00) 0.003 (0.88) −0.033 (0.23) −0.064⁎ (0.07) −0.049 (0.26) −0.029 (0.64)

−0.002 (0.27) −0.001 (0.44) 0.000 (0.90) 0.000 (0.81) 0.000 (0.77) 0.001⁎ (0.06) 0.001⁎⁎⁎

0.009⁎⁎⁎ (0.01) 0.003 (0.19) 0.003⁎⁎⁎

(0.00) 0.342⁎⁎⁎

−0.098⁎⁎ (0.02) −0.095⁎⁎ (0.03) −0.011 (0.73) −0.039⁎

Market

2 Radj.

Panel B: Conditional Treynor–Mazuy timing model Equity EW Equity VW Balanced EW Balanced VW Variable EW

Index EW Index VW Private EW Private VW Foreign EW

(0.00) −0.023⁎⁎⁎ (0.00) 0.004 (0.71) 0.007 (0.38) 0.054⁎⁎⁎ (0.00)

−0.006 (0.88) 0.015 (0.70) 0.010 (0.68) 0.030 (0.25) 0.011 (0.72) 0.023 (0.49) −0.012 (0.43) 0.003 (0.85) 0.014 (0.69) 0.014 (0.61) 0.055 (0.30)

(0.00) −0.035⁎⁎⁎ (0.00) −0.019⁎⁎ (0.03) −0.001 (0.96)

(0.10) −0.018 (0.58) −0.056⁎ (0.09) 0.024 (0.24) 0.029 (0.14) −0.043 (0.24) −0.045 (0.17) −0.123⁎⁎ (0.02)

(0.00) 0.001⁎⁎⁎ (0.00) −0.001 (0.28) 0.001 (0.43) −0.002⁎⁎ (0.03)

(0.00) 0.002⁎⁎⁎ (0.01) 0.004⁎⁎⁎ (0.00) 0.002⁎ (0.06) 0.000 (0.81) −0.001 (0.68) 0.005⁎⁎⁎ (0.00) 0.004⁎⁎⁎ (0.01) 0.003 (0.23)

0.80 0.78 0.76 0.72 0.79 0.77 0.95 0.94 0.81 0.79 0.37

Panel A of this table reports coefficient estimates of the unconditional Treynor and Mazuy (1966) timing model as represented by Eq. (3), while Panel B shows the results of the conditional model of Eq. (4) in the period from 01/2000 to 03/2011. The dummy variable Ramt takes the value of one during the extended Ramadan period and zero in the remaining months of the Islamic calendar year. The regression error terms are assumed to follow a GARCH(1,1) process. P-values in parentheses are based on robust Bollerslev and Wooldridge (1992) standard errors. ⁎, ⁎⁎ and ⁎⁎⁎ denote statistical significance at the 10%, 5% and 1% level, respectively.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

Variable VW

−0.006 (0.59) −0.008 (0.51) −0.010 (0.16) 0.003 (0.72) −0.014⁎ (0.08) −0.005 (0.55) −0.032⁎⁎⁎

227

228

Table 7 Total performance measures. Fund type

Full sample

Non-Ramadan

Unconditional

Equity VW Balanced EW Balanced VW Variable EW Variable VW Index EW Index VW Private EW Private VW Foreign EW

Ramadan

Unconditional

Conditional

Unconditional

Conditional

TM

HM

TM

HM

TM

HM

TM

HM

TM

HM

TM

HM

−0.002 (0.85) −0.008 (0.42) −0.004 (0.51) 0.005 (0.57) −0.009 (0.20) 0.002 (0.77) −0.027⁎⁎⁎ (0.00) −0.018⁎⁎⁎

−0.007 (0.46) −0.007 (0.58) −0.008 (0.21) 0.004 (0.62) −0.010 (0.21) 0.001 (0.93) −0.028⁎⁎⁎ (0.00) −0.017⁎⁎⁎

−0.007 (0.52) −0.011 (0.30) −0.005 (0.48) 0.006 (0.36) −0.005 (0.53) 0.005 (0.51) −0.027⁎⁎⁎ (0.00) −0.017⁎⁎⁎

−0.012 (0.25) −0.012 (0.27) −0.009 (0.16) 0.000 (0.98) −0.008 (0.31) 0.003 (0.78) −0.028⁎⁎⁎

−0.018⁎ (0.06) −0.015 (0.16) −0.009 (0.17) 0.000 (0.98) −0.015⁎ (0.07) −0.006 (0.48) −0.026⁎⁎⁎

−0.015 (0.14) −0.014 (0.15) −0.011⁎ (0.10) 0.000 (0.99) −0.016⁎⁎ (0.04) −0.006 (0.45) −0.027⁎⁎⁎

−0.016 (0.12) −0.015 (0.16) −0.012⁎

0.044 (0.24) 0.039 (0.28) 0.022 (0.35) 0.038⁎

0.050 (0.18) 0.043 (0.23) 0.023 (0.21) 0.041⁎

0.027 (0.48) 0.019 (0.59) 0.020 (0.36) 0.045⁎⁎

0.019 (0.61) 0.010 (0.80) 0.017 (0.47) 0.039⁎

(0.09) 0.022 (0.43) 0.049⁎ (0.09) −0.037⁎⁎⁎

(0.00) 0.018⁎⁎ (0.03) 0.014⁎⁎ (0.04) 0.056⁎⁎⁎

(0.00) 0.013 (0.12) 0.012⁎ (0.08) 0.051⁎⁎⁎

(0.00) 0.015⁎ (0.07) 0.017⁎⁎ (0.02) 0.043⁎⁎⁎

(0.00) −0.018⁎⁎⁎ (0.00) 0.005 (0.59) 0.013⁎ (0.07) 0.046⁎⁎⁎

(0.00) −0.018⁎⁎⁎ (0.00) 0.014 (0.11) 0.008 (0.29) 0.045⁎⁎

(0.00) −0.016⁎⁎⁎ (0.00) 0.006 (0.53) 0.007 (0.38) 0.037⁎⁎

−0.016 (0.17) −0.015 (0.19) −0.009 (0.23) 0.001 (0.86) −0.013 (0.12) 0.001 (0.95) −0.026⁎⁎⁎ (0.00) −0.017⁎⁎⁎

(0.04) 0.018 (0.52) 0.038 (0.21) −0.039⁎⁎⁎ (0.01) −0.018 (0.35) 0.043 (0.17) 0.046⁎ (0.07) 0.113⁎⁎

(0.09) 0.019 (0.48) 0.035 (0.23) −0.044⁎⁎⁎ (0.00) −0.024 (0.13) 0.041 (0.20) 0.045⁎ (0.08) 0.126⁎⁎⁎

(0.00)

(0.00)

(0.01)

(0.00)

(0.05)

(0.02)

(0.02)

(0.00)

(0.09) −0.004 (0.54) −0.015⁎⁎ (0.05) −0.001 (0.87) −0.026⁎⁎⁎

(0.00) 0.000 (1.00) 0.011 (0.19) 0.034⁎

(0.00) −0.017⁎⁎⁎ (0.00) 0.001 (0.95) 0.008 (0.28) 0.036⁎⁎

(0.01) −0.018 (0.26) 0.064⁎⁎ (0.03) 0.050⁎⁎ (0.04) 0.127⁎⁎

(0.07) 0.025 (0.33) 0.054⁎⁎ (0.03) −0.039⁎⁎⁎ (0.01) −0.020 (0.23) 0.069⁎⁎ (0.02) 0.053⁎⁎ (0.04) 0.142⁎⁎⁎

(0.07)

(0.03)

(0.02)

(0.00)

This table reports the mutual funds' daily total performance measures (TPM) expressed as a percentage as estimated by Eqs. (7) through (9) in the period from 01/2000 to 03/2011. TPMs are computed for both the unconditional and conditional TM and HM timing models and are reported for the full sample period as well as the non-Ramadan and the extended Ramadan period. P-values in parentheses are calculated as described in Grinblatt and Titman (1994) and Cesari and Panetta (2002). ⁎, ⁎⁎ and ⁎⁎⁎ denote statistical significance at the 10%, 5% and 1% level, respectively.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

Equity EW

Conditional

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

229

Table 8 Differences between total performance measures during Ramadan and non-Ramadan. Fund type Equity EW Equity VW Balanced EW Balanced VW Variable EW Variable VW Index EW Index VW Private EW Private VW Foreign EW

Unconditional

Conditional

TM

HM

TM

HM

+0.062 (0.12) +0.054 (0.15) +0.030 (0.21) +0.038⁎ (0.10) +0.036 (0.19) +0.055⁎

+0.065⁎ (0.10) +0.057 (0.12) +0.034⁎ (0.08) +0.041⁎ (0.09) +0.041 (0.13) +0.060⁎⁎⁎

(0.07) −0.011 (0.44) 0.000 (0.99) +0.050⁎ (0.10) +0.043⁎ (0.10) +0.082 (0.23)

(0.01) −0.013 (0.41) −0.004 (0.83) +0.063⁎⁎ (0.04) +0.047⁎ (0.08) +0.105⁎⁎ (0.02)

+0.043 (0.29) +0.034 (0.37) +0.029 (0.20) +0.044⁎ (0.07) +0.031 (0.27) +0.037 (0.24) −0.014 (0.38) −0.001 (0.95) +0.043 (0.19) +0.035 (0.20) +0.080⁎ (0.10)

+0.035 (0.38) +0.025 (0.54) +0.029 (0.23) +0.044⁎ (0.08) +0.034 (0.22) +0.036 (0.24) −0.018 (0.22) −0.008 (0.65) +0.040 (0.23) +0.037 (0.17) +0.090⁎⁎ (0.04)

This table reports the differences between mutual funds' daily total performance measures (TPM) during the extended Ramadan and non-Ramadan period from 01/2000 to 03/2011. The differences are computed for both the unconditional and conditional TM and HM timing models and are expressed as a percentage. P-values in parentheses are calculated as described in Grinblatt and Titman (1994) and Cesari and Panetta (2002). ⁎, ⁎⁎ and ⁎⁎⁎ denote statistical significance at the 10%, 5% and 1% level, respectively.

statistically significant, indicating that index fund managers make an effort to offset the negative impact of new money inflows. Overall, our findings suggest that the lion's share of the Turkish mutual funds' raw and abnormal returns throughout the year is earned during Ramadan. Similarly, Białkowski et al. (2009, 2012) report that between 1994 and 2007 the annualized returns of the MSCI Turkey Index during Ramadan far surpassed the gains obtainable throughout the rest of the year. 17

5. Summary and conclusions Intriguing results reported in recent literature show that the holy month of Ramadan has a strong impact on stock market returns and their volatility, particularly in countries where Muslims constitute a significant fraction of society. Ramadan is not only a religious month marked by fasting, prayer and self-reflection, but also a time in which obedient Muslims should seek to improve their human relationships. The Feast of Ramadan celebrated at the end of the fasting period further strengthens family ties and promotes feelings of solidarity and social identity. The positive psychological effects that these religious rituals have on the general public lead to optimistic beliefs as well as improved investor sentiment. Consistent with this notion, we confirm the results of previous studies and document higher returns during Ramadan for the Istanbul Stock Exchange, which is the second largest stock market in the Middle East in terms of market capitalization after the Tadawul in Saudi Arabia. However, the effect has gradually 17 We also test whether higher stock returns during Ramadan are driven by lower market liquidity. Similar to Haugen and Baker (1996) and Białkowski et al. (2009, 2012), we define liquidity as the turnover ratio, which is the daily trading volume of the Turkish stock market measured in Turkish Lira divided by the total market capitalization of the Turkish stock market. Surprisingly, the daily turnover ratio over the period January 2000 to March 2011 is 0.69% during Ramadan and 0.55% during the rest of the year. The difference of 0.14% is statistically significant at the 1% level. Our results are in line with Białkowski et al. (2009, 2012), who also find a noticeable increase in daily turnover during Ramadan for 14 predominantly Muslim countries.

230

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

decreased over recent years, reflecting both investors' awareness of the anomaly and an increasing integration of the Turkish stock market into the global financial landscape. In terms of volatility, the impact of Ramadan has reversed over time from a heightening to a dampening influence on stock price fluctuations. As prior research shows that the Ramadan effect has been particularly evident in Turkey, we investigate whether mutual fund managers investing in the Istanbul stock market are able to benefit from the calendar anomaly. Our main findings suggest that fund managers and investors alike have discovered the profit opportunities during Ramadan and the days subsequent to the month of fasting. First, we show that domestic institutional funds and larger domestic hybrid funds earn positive and significant risk-adjusted returns during Ramadan over the 2000 to 2011 period, while their risk-adjusted performance throughout the rest of the year is insignificant. For other domestic fund groups the abnormal returns are less pronounced, albeit economically relevant. However, it remains a subject for future research whether all mutual fund managers engage in deliberate market timing and do not just follow the herd during the Ramadan market rally. On the one hand, we can rule out the possibility that mutual funds simply earn higher abnormal returns during Ramadan because they pursue a momentum strategy. We control for this effect by using the Carhart (1997) four-factor model and allowing for differing factor exposures during Ramadan and the non-Ramadan period. Our results even indicate that domestic fund managers tend to follow a moderate contrarian strategy. On the other hand, the positive abnormal performance could be the result of overweighting certain sectors compared to the aggregate market portfolio. Some sectors might be more positively affected by Ramadan than others. If a fund manager assigns a higher portfolio weight to these more promising sectors than they actually possess in the market, the sector allocation will contribute positively to the mutual fund's alpha. Note that such an investment decision is also attributed to the fund manager's selection ability. Foreign equity funds investing in the Turkish stock market also deliver risk-adjusted returns during Ramadan that are more than three times larger than during the other months. This result suggests that euro-based investors can benefit from the anomaly by buying fund shares in advance of the Muslim holy month, or alternatively, by delaying already planned sales until after the Feast of Ramadan. Surprisingly, domestic index funds constitute the only fund group that yields negative risk-adjusted returns during the Muslim holy month. The reason for this seemingly inferior performance lies primarily in increased money inflows. Apparently, individual investors try to jump on the bandwagon through purchasing passive index funds, which exhibit the highest equity exposure and are thus more likely to benefit from the Ramadan anomaly. These speculative attempts are not entirely in vain since index funds earn the largest average raw returns of all domestic mutual fund groups during Ramadan. However, local index fund managers seem to have difficulty in converting the new cash inflows into profitable investments in a timely manner. Acknowledgments Tomasz Wisniewski would like to acknowledge the support of the University of Leicester's sabbatical scheme. We would also like to thank Ron Balvers, Bruce Hearn, Nikolaos Tessaromatis and William N. Trumbull as well as participants of the 19th Annual Conference of the Multinational Finance Society in Krakow, the 10th INFINITI Conference on International Finance in Dublin, the 9th ACDD Conference in Strasbourg and the economics seminar at West Virginia University for valuable comments and suggestions. We gratefully acknowledge the provision of mutual fund data by Thomson Reuters Lipper and Takasbank, the ISE Settlement and Custody Bank Inc. References Abadir, K.M., Spierdijk, L., 2005. Liquidity Constraints and the Demand for Assets: An Application to the Festivity Effect. Working Paper. Imperial College, London (October 2005). Akdeniz, L., Altay-Salih, A., Aydogan, K., 2000. A cross-section of expected stock returns on the Istanbul Stock Exchange. Russian and East European Finance and Trade 36 (5), 6–26. Al-Hajieh, H., Redhead, K., Rodgers, T., 2011. Investor sentiment and calendar anomaly effects: a case study of the impact of Ramadan on Islamic Middle Eastern markets. Research in International Business and Finance 25 (3), 345–356. Ayçiçegi-Dinn, A., Caldwell-Harris, C.L., 2011. Individualism–collectivism among Americans, Turks and Turkish immigrants to the US. International Journal of Intercultural Relations 35 (1), 9–16. Baele, L., Farooq, M., Ongena, S., 2012. Of Religion and Redemption: Evidence from Default on Islamic Loans. Working Paper. Tilburg University (June 2012). Bekaert, G., Harvey, C.R., 2000. Foreign speculators and emerging equity markets. Journal of Finance 55 (2), 565–613.

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

231

Białkowski, J., Etebari, A., Wisniewski, T.P., 2009. Piety and Profits: Stock Market Anomaly during the Muslim Holy Month. Working Paper. University of Canterbury (September 2009). Białkowski, J., Etebari, A., Wisniewski, T.P., 2012. Fast profits: investor sentiment and stock returns during Ramadan. Journal of Banking & Finance 36 (3), 835–845. Bildik, R., Gülay, G., 2007. Profitability of contrarian strategies: evidence from the Istanbul Stock Exchange. International Review of Finance 7 (1–2), 61–87. Bollen, N.P.B., 2007. Mutual fund attributes and investor behavior. Journal of Financial and Quantitative Analysis 42 (3), 683–708. Bollen, N.P.B., Busse, J.A., 2001. On the timing ability of mutual fund managers. Journal of Finance 56 (3), 1075–1094. Bollen, N.P.B., Busse, J.A., 2005. Short-term persistence in mutual fund performance. Review of Financial Studies 18 (2), 569–597. Bollerslev, T., Wooldridge, J.M., 1992. Quasi-maximum likelihood estimation and inference in dynamic models with time-varying covariances. Econometric Reviews 11 (2), 143–172. Borensztein, E., Gelos, R.G., 2003. A panic-prone pack? The behavior of emerging market mutual funds. IMF Staff Papers 50 (1), 43–63. Capital Markets Board of Turkey, 2011. 2010 CMB Annual Report. Carhart, M.M., 1997. On persistence in mutual fund performance. Journal of Finance 52 (1), 57–82. Central Intelligence Agency, 2011. The World Factbook 2011. Cesari, R., Panetta, F., 2002. The performance of Italian equity funds. Journal of Banking & Finance 26 (1), 99–126. Chong, R., Hudson, R., Keasey, K., Littler, K., 2005. Pre-holiday effects: international evidence on the decline and reversal of a stock market anomaly. Journal of International Money and Finance 24 (8), 1226–1236. Christopherson, J.A., Ferson, W.E., Glassman, D.A., 1998. Conditioning manager alphas on economic information: another look at the persistence of performance. Review of Financial Studies 11 (1), 111–142. Chui, A.C.W., Titman, S., Wei, K.C.J., 2010. Individualism and momentum around the world. Journal of Finance 65 (1), 361–392. Comer, G., Larrymore, N., Rodriguez, J., 2009. Controlling for fixed-income exposure in portfolio evaluation: evidence from hybrid mutual funds. Review of Financial Studies 22 (2), 481–507. Edelen, R.M., 1999. Investor flows and the assessed performance of open-end mutual funds. Journal of Financial Economics 53 (3), 439–466. Edison, H.J., Warnock, F.E., 2003. A simple measure of the intensity of capital controls. Journal of Empirical Finance 10 (1–2), 81–103. Elton, E.J., Gruber, M.J., Blake, C.R., 2012. An examination of mutual fund timing ability using monthly holdings data. Review of Finance 16 (3), 619–645. Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33 (1), 3–56. Federation of Euro-Asian Stock Exchanges, 2010. FEAS Historical Statistics 2010. Ferson, W.E., Schadt, R.W., 1996. Measuring fund strategy and performance in changing economic conditions. Journal of Finance 51 (2), 425–461. Ferson, W.E., Warther, V.A., 1996. Evaluating fund performance in a dynamic market. Financial Analysts Journal 52 (6), 20–28. Ferson, W.E., Sarkissian, S., Simin, T., 2003. Is stock return predictability spurious? Journal of Investment Management 1 (3), 1–10. Glosten, L.R., Jagannathan, R., Runkle, D.E., 1993. On the relation between the expected value and the volatility of the nominal excess return on stocks. Journal of Finance 48 (5), 1779–1801. Gonenc, H., Karan, M.B., 2003. Do value stocks earn higher returns than growth stocks in an emerging market? Evidence from the Istanbul Stock Exchange. Journal of International Financial Management & Accounting 14 (1), 1–25. Griffin, J.M., Ji, X., Martin, J.S., 2003. Momentum investing and business cycle risk: evidence from pole to pole. Journal of Finance 58 (6), 2515–2547. Grinblatt, M., Titman, S., 1994. A study of monthly mutual fund returns and performance evaluation techniques. Journal of Financial and Quantitative Analysis 29 (3), 419–444. Haugen, R.A., Baker, N.L., 1996. Commonality in the determinants of expected stock returns. Journal of Financial Economics 41 (3), 401–439. Henriksson, R.D., Merton, R.C., 1981. On market timing and investment performance. II. Statistical procedures for evaluating forecasting skills. Journal of Business 54 (4), 513–533. Huij, J., Derwall, J., 2008. “Hot hands” in bond funds. Journal of Banking & Finance 32 (4), 559–572. Huij, J., Post, T., 2011. On the performance of emerging market equity mutual funds. Emerging Markets Review 12 (3), 238–249. İmişiker, S., Özlale, Ü., 2008. Assessing selectivity and market timing performance of mutual funds for an emerging market: the case of Turkey. Emerging Markets Finance and Trade 44 (2), 87–99. Jagannathan, R., Korajczyk, R.A., 1986. Assessing the market timing performance of managed portfolios. Journal of Business 59 (2), 217–235. Jiang, G.J., Yao, T., Yu, T., 2007. Do mutual funds time the market? Evidence from portfolio holdings. Journal of Financial Economics 86 (3), 724–758. Ling, S., McAleer, M., 2002. Stationarity and the existence of moments of a family of GARCH processes. Journal of Econometrics 106 (1), 109–117. MacKinnon, J.G., 2002. Bootstrap inference in econometrics. Canadian Journal of Economics 35 (4), 615–645. Matallín-Sáez, J.C., 2003. Asymmetric relation in omitted benchmarks and market timing in mutual funds. Applied Economics Letters 10 (12), 775–778. Menkhoff, L., Schmidt, U., 2005. The use of trading strategies by fund managers: some first survey evidence. Applied Economics 37 (15), 1719–1730. Odabaşi, Y., Argan, M., 2009. Aspects of underlying Ramadan consumption patterns in Turkey. Journal of International Consumer Marketing 21 (3), 203–218. Oğuzsoy, C.B., Güven, S., 2004. Holy days effect on Istanbul Stock Exchange. Journal of Emerging Market Finance 3 (1), 63–75. Otten, R., Bams, D., 2007. The performance of local versus foreign mutual fund managers. European Financial Management 13 (4), 702–720. Sandikci, Ö., Omeraki, S., 2007. Globalization and rituals: does Ramadan turn into Christmas? Advances in Consumer Research 34, 610–615. Schmidt, P.S., von Arx, U., Schrimpf, A., Wagner, A.F., Ziegler, A., 2011. On the Construction of Common Size, Value and Momentum Factors in International Stock Markets: A Guide with Applications. Working Paper. University of Zurich (January 2011). Schwert, G.W., 2003. Anomalies and market efficiency. In: Constantinides, G.M., Harris, M., Stulz, R.M. (Eds.), Handbook of the Economics of Finance, pp. 939–974.

232

J. Białkowski et al. / Emerging Markets Review 15 (2013) 211–232

Seyyed, F.J., Abraham, A., Al-Hajji, M., 2005. Seasonality in stock returns and volatility: the Ramadan effect. Research in International Business and Finance 19 (3), 374–383. Sirri, E.R., Tufano, P., 1998. Costly search and mutual fund flows. Journal of Finance 53 (5), 1589–1622. Treynor, J.L., Mazuy, K.K., 1966. Can mutual funds outguess the market? Harvard Business Review 44 (4), 131–136. Van Dijk, M.A., 2011. Is size dead? A review of the size effect in equity returns. Journal of Banking & Finance 35 (12), 3263–3274. World Federation of Exchanges, 2010. WFE 2010 Annual Report & Statistics. Wright, C., Banerjee, P., Boney, V.R., 2008. Behavioral finance: are the disciples profiting from the doctrine? Journal of Investing 17 (4), 82–90.