Dutch Mutual Funds Performance

(1)

1

Dutch Mutual Funds Performance

PHILIPP SPORKA*

ABSTRACT:

In this research I investigate 324 Dutch Mutual Funds from 2003 until 2009. I show that these funds underperform against a passive strategy of common factors from Carhart’s Four Factor model, and an additional and recently discovered endogenous benchmark factor. The new Group factor however, does not uncover its entire potential, due to weak group homogeneity. I include new and

alternative fund categories in my research and find evidence for differences in the performance of the individual groups of funds.

A lot has been written, and a lot of research has been carried out on the performance of Mutual Funds. Most of this research focused on the by far largest market in mutual funds, the United States.

Relatively little has been done about European funds, as several researches note (Bauer, Koedijk, Otten (2004), Otten and Bams (2002), Blake and Timmerman (1998)). If those decided to do so, they mainly focused on the largest mutual fmarkets in Europe, namely the United Kingdom, France, and Germany. The Netherlands is in terms of size a somewhat small country, in terms of stock market activity, and assets of mutual funds, it is one of the most important European nations. The Dutch market value of publicly traded shares amounted to $ 956,5 billion – ranked 17th in the world, and ranked 8th in Europe - by the end of the year 2008 (CIA world factbook). According to Otten and Bams (2002), the mutual fund industry in the Netherlands is ranked fifth in Europe, in terms of size. So far, only Ter Horst, Nijman, and de Roon (1998) monitored the Dutch mutual funds performance.

(2)

2

In this paper I examine the younger history of 324 mutual funds in the Netherlands, i.e. from July 2003 until October 2009. All kinds of different funds are represented in this sample, which is rather new to mutual fund research. The whole performance evaluation is based on Alpha (α), introduced and named after Jensen (1968). Jensen’s Alpha is part of the following equation that is based on the Capital Asset Pricing Model CAPM, and determines whether a fund is underperforming or outperforming relative to a common benchmark, typically an index tracking the market.

𝑅𝑖 – 𝑅𝑟𝑓 = 𝛼𝑖+ 𝛽𝑖(1) ∗ [𝑅𝑚 − 𝑅𝑟𝑓] + 𝜖 (1) where 𝑅𝑖 denotes the return of each individual mutual fund, 𝑅𝑟𝑓 the risk free rate, and 𝑅𝑚 the return of a market proxy. The term in the brackets is the market premium, 𝛽𝑖(1) is the Beta of the fund relative to the market, and 𝜖 is the error term with mean zero. That only leaves the intercept 𝛼𝑖, which, if positive, indicates outperformance of a fund relative to the benchmark, and if negative, the fund underperforms against the market.

I further extend the simple CAPM equation by two factors, namely SMB and HML, as it has been suggested by Fama and French (1993). SMB stands for “small minus big”, expressed in terms of market capitalization and takes the difference between small cap performance and large cap

performance into account. HML stands for “high minus low” and refers to the book-to-market ratio. It is the difference between the performance of value stocks with a high book to market ratio, and the performance of growth stocks with a low book to market ratio. In their paper Fama and French (1993) show how these two factors massively improve the explanatory power of their regression. They argue that “the intercept in the time-series regression of the managed portfolio’s excess return […] is the average abnormal return needed to judge whether a manager can beat the market”, and this is exactly what I intend to examine in this paper. The Fama and French Three Factor Model (from now on FF3) regression equation expands the previous CAPM equation by SMB and HML, and reads as below:

𝑅𝑖 – 𝑅𝑟𝑓 = 𝛼𝑖+ 𝛽𝑖(1) ∗ 𝑅𝑚 − 𝑅𝑟𝑓 + 𝛽𝑖 2 𝑆𝑀𝐵 + 𝛽𝑖 3 𝐻𝑀𝐿 + 𝜖 (2) 𝛽𝑖 2 and 𝛽𝑖 3 are the sensitivities of each fund to the factors SMB and HML, respectively.

Carhart (1997), proposes an extension of the FF3 model, and adds another relevant factor, which he calls Momentum (short MOM). This factor captures the phenomena of past winning stocks outperforming past losing stocks in the subsequent period to follow. In short it is the difference between recent winners and losers.

(3)

3

The factors SMB, HML and MOM are based on stock portfolios. This paper, however, is about funds, and Huij and Verbeek (2009) point this out as a weakness. They claim that “systematic biases” occur when applying stock portfolios on mutual funds. Instead they propose the use of mutual funds portfolios to reduce the bias. One could form fund portfolios of all available funds, hence exogenous benchmarks. Another way of dealing with this is to enlarge the Four Factor Model by an endogenously computed Group factor that captures some of the commonalities that only funds share. This group factor was established by Hunter et al. (2009). In contrast to the ordinary factors (SMB, HML, MOM), which mainly are exogenous by nature, the group factor is built “from the return on the group of funds to which a given fund belongs” (Hunter et al. (2009)). Further they note that “each fund manager chooses the group within which it intends to compete. This choice of a group by the manager indicates the set of strategies from which the manager will choose in order to compete”. And they add that this “approach accounts for the commonalities in fund strategies […]”. An investor usually desires a type of fund before he picks a specific fund. Hence, the investor compares a fund to its peer group, something that the Group factor greatly takes into account. In terms of only investigating Dutch funds, the peer group of a fund is not just the investment strategy or geographical focus of a fund, but also its manager(s). An investor from the Netherlands has a rather limited choice of funds within his category of interest. Access to Dutch funds is much easier and in many cases cheaper. Furthermore, access to foreign funds is difficult due to language barriers and these funds are not being promoted as much. Naturally the choice falls for one of the Dutch managed funds, operated by Dutch managers. Hence the peer group is reduced to the investor’s universe of choices. The investor evaluates the performance of a fund, let us say a fund that specializes in Asian equity, not to all Asian funds, but compares it to the funds that are available, in this case Dutch funds. This example illustrates the advantages of the Group factor over an Asian market index. The results of Hunter et al. (2009) clearly show improved regression outcomes in terms of less correlation of the residuals across funds within a group and across groups, enhancement in explanatory power, and thereby more precise Alpha estimation. The new factor GRP expands the previous equation to:

𝑅𝑖 – 𝑅𝑟𝑓 = 𝛼𝑖+ 𝛽𝑖(1) ∗ 𝑅𝑚 − 𝑅𝑟𝑓 + 𝛽𝑖 2 𝑆𝑀𝐵 + 𝛽𝑖 3 𝐻𝑀𝐿 + 𝛽𝑖 4 𝑀𝑂𝑀 +

𝛽𝑖 5 𝐺𝑅𝑃𝑖,𝑔 + 𝜖 (4)

GRP is a proxy of every fund’s peer group, the funds that it directly competes with. It is an endogenous factor that is easily obtained from the fund’s data. It can bundle some of the

(4)

4

All the models based on equations (1) to (4) should help to explain and estimate α. In fact the more we move from equation (1) to (4), the better the estimation of α should be. For this reason, I test whether each individual model positively reacts to the omitted variables test. This leads to the first set of hypotheses.

The first hypothesis is about the CAPM.

H1: The market factor is a relevant benchmark, and improves the regression.

The second hypothesis is about the move from the CAPM to the FF3 model.

H2: SMB and HML, both are relevant benchmarks, and jointly improve the regression.

The third hypothesis is about the move from the FF3 model to the CAR4 model.

H3: The MOM factor is a relevant benchmark, and improves the regression.

The fourth hypothesis is about the move from the CAR4 model to the 5F model.

H4: The GRP factor is a relevant benchmark, and improves the regression.

These hypotheses deal with the preciseness of the models. The main goal of this paper, however, is to estimate α. Different research so far has shown mixed results about whether α is positive or negative or even just about zero. All portfolio management firms that manage funds for investors exist on the basis of a positive α, at least theoretically. Early literature indicated that positive

α’s were present in the fund’s industry, however, Gruber (1996) and Malkiel (1995) in their US

studies showed that after accounting for funds that did not survive, results yield underperformance against the benchmark. When again some studies showed positive α, it was taken away by Carhart’s introduction of the MOM factor in 1997.

The few studies for European risk adjusted performance show results that are summarized in the table below

Table Ⅰ

Country Author / Year Results

UK Shukla and Imwegen 1995

-Blake and Timmerman 1998

-France Mc Donald 1973 +

Sweden Dahlquist, Engström, Söderlind 2000 mixed

(5)

5

As can be seen, there is no clear evidence in the available literature from which to draw definite conclusions. Shukla and Imwegen (1995) for example argue that the size of an economy and therefore the size of mutual funds are positively associated with performance. This would rather suggest negative performance relative to the benchmark for the Dutch fund industry. On the contrary Ter Horst, Nijman, and de Roon (1998) show that Dutch funds actually outperform the market.

The most extensive study about the European funds market was conducted by Otten and Bams (2002). Their up to date methodologies yield positive α results for the mutual fund industries of the following nations: UK, The Netherlands, France and Italy. Slightly negative results were obtained for Germany. Their research differs from mine, mainly because their paper was written, before my sample started, and also because they only restrict their research to funds that invest within Europe. Most of my examined funds however, invest everywhere in the world.

An investor based in the Netherlands, or a foreign investor is interested whether it is worth to invest in Dutch Mutual Funds. Furthermore, many people rely on funds, for their pensions. Moreover, managers of those funds are concerned about the overall performance of their whole peer group. Furthermore, efficient market theory suggests that abnormal returns are not possible, especially not for an entire industry. The previously mentioned studies however, show that abnormal returns occur. Given the ambiguity, about the sign of Alpha in the Dutch funds business, I test that Alpha is not significantly different from zero and translate this into a hypothesis.

H5: Dutch Mutual Funds do neither outperform nor underperform a passive strategy.

A passive strategy involves replicating a portfolio with exposure to all the proxies discussed (market, SMB, HML, MOM, GRP), whereas most traditional funds are actively managed. H5 is based on the results of the value and significance of α.

(6)

6

The CAR4 model further allows for direct comparison of the group’s averages and statistical evidence of whether there are significant deviations. Assuming that the GRP factor erased group specific factors, there should be evidence that in turn there are significant differences between the groups in the CAR4 model. This leads to a hypothesis version of the previous assumption.

H6: Some groups of funds do significantly outperform other groups of funds in the CAR4 model.

Although it is difficult to compute statistically significant outcomes for a similar econometric investigation of the 5F model, it is possible to infer some conclusions about the joint outcomes of hypothesis H6 and the answer to the assumption that the standard deviation of group α is larger in the CAR4 model. If both, cannot be rejected it is deducible that the GRP factor wiped out any differences in the group returns of funds. However, this would not be an empirical result, but it can point at the power of the GRP factor, and at how ineffective fund managers are in earning abnormal returns.

Another interesting question arises from how well Dutch fund managers compare

internationally. Since, this sample is restricted to Dutch funds; such a comparison is only possible based on existing literature. European comparison is summarized in table Ⅰ.

The next section deals with the data used, the Methodology applied and the models. The section that follows presents and explains the results, which are being discussed in the section after that. The paper ends with a short conclusion.

I. Data and Methods

A. Data

(7)

7

B. Capital Asset Pricing Model

In this paper I run regression analyses with different models. The first is the CAPM. The CAPM regression requires two variables, namely a market benchmark and a risk free rate. The

selection of a suitable risk free rate is easier than that of a market rate. Let me first define the “market” or “universe” before I determine the benchmarks.

(8)

8

Jones Euro Stoxx Total Market Index”, which covers 95% of free float market capitalization in Europe. The Index’ returns are available through the website of Stoxx.

Now that it is determined what the “investment universe” of a Dutch fund manager is, I choose an appropriate risk free rate. Risk free rates are estimated through the bond market’s activities. It is wise to choose a bond market that is liquid enough, and can is considered risk free. The safest bet on a risk free government bond is a German bond. German fiscal policy is well-known to be very reliable, and independent from politics. German bonds have historically been receiving the highest ratings possible (e.g. AAA from S&P), and liquidity of German bonds is very high. Kempf and Uhrig-Homburg (2000) prove that liquidity is a determinant in the pricing of bonds. The Deutsche Bundesbank daily computes the so-called “Umlaufrendite”, the risk free rate. Monthly returns are taken from averages of publicly issued bonds with a minimum maturity of four years. This rate has been computed since the year 1955 and is successful in its use in Germany. It is the closest and easiest way to obtain a European risk free rate.

C. The Three Factor Model

The next model (equation (2)) includes the factors SMB (Small minus Big) and HML (High minus Low) that were introduced by Fama and French (1993). The website of French offers ready to use monthly data on the SMB and HML factors, but only for the US. In this investment universe my model calls for Euro Zone SMB and HML data. I construct this data by taking the difference of two indices. Again, I make use of Dow Jones Euro Stoxx, also to remain consistent. In order to create the SMB benchmark for my model, I subtract the returns of the Dow Jones Euro Stoxx Total Market Index Large from the returns of the Dow Jones Euro Stoxx Total Market Index Small on a monthly basis. Similarly, I subtract the monthly returns of the Dow Jones Euro Stoxx Total Market Index Growth from the monthly returns of the Dow Jones Euro Stoxx Total Market Index Value to obtain the HML factor. Data on this is available throughout my sample period.

D. The Four Factor Model

(9)

9

E. The Group Factor Model

The Group Factor model from equation (4) requires the endogenous construction of different group factors - GRP in this research. In this model every fund belongs to one group only. The group return is based on the equally weighted average of all the group’s fund returns that are observable during a particular period of time. Hunter et al. (2009) suggest the use of a three year window during which the average is being calculated. The authors also show that the group return benchmark varies in time and therefore advocate for three years. My sample consists of 76 observations, i.e. roughly more than six years. This is why I split it in two periods of time, each comprising 38 months of data. Hence, the group return is based on the average of 38 data points in time for each period. Therefore, there are two different averages for every single group. The following sections discuss the formation of the groups.

In order to categorize all the funds into different groups I acquire the necessary information on the Dutch website of Morningstar. Morningstar analyzes funds, and usually also puts them in one of 180 predefined categories. Furthermore, if Morningstar does not give any information about a particular fund I check another source, namely Trustnetoffshore. Lastly, some fund’s official websites offer information about their products, however rarely, if these funds ceased to exist.

My research requires a couple of categories only. Unfortunately the previously described methodology of categorizing the funds with the help of morningstar.nl and trustnetoffshore.com results in a large amount of different and very specific categories. Therefore, I cluster similar categories into new upper level groups. My aim is to create groups that consist of at least five funds. The better a fund fits a category, the better the use of the group benchmark will be. This yields better and more precise outcomes for alpha. However, at the same time it is important to form groups that are large enough in order to be essential. Hence, although it might sound irrational, I put a fund that mainly invests in Japanese equity into the same category, namely “Asian equity funds”, as a fund that invests in Asian equity excluding Japan. At first, those two funds appear mutually exclusive, but they are undoubtedly related to some extent, being located in the same geographic region. I assume that Japanese equity is more correlated with Asian equity excluding Japan, than to the world’s equity. A separate group for Japanese funds only is not being created, since my data only contains two Japanese funds. A benchmark created from the average returns of the two funds would not be a good reference. With this logic in mind, I create all the other groups. Very important to mention here is that I do not have any groups in mind, before I look at the sample. I create the groups, based on available categories and not the other way around. This way guarantees that the groups are as homogeneous as possible.

(10)

10

asset classes, such as Real Estate, commodities, and funds of hedge funds become potential groups. Enough funds in this sample are claiming to be “Real Estate” funds; therefore it becomes a group in my research. In order to construct a significantly large group, I merge funds that invest in

commodities, and funds that invest in hedge funds, or claim to be using derivatives of any kind, into a single group. This group is simply named “hedge”. Most of the information that I gather on equity funds reveals the geographic area they are mainly invested in. Hence, I decide to form groups based on geographic regions. Those regions are: “Asia”, “Europe”, “Global”, “NL”, and lastly “USA”. As one can see, these regions are not mutually exclusive. Global with any other region, and Europe and NL are overlapping investment universes. Since this is a research about Dutch mutual funds, it should not surprise that The Netherlands is an independent group. One can assume that Dutch managers know their domestic market better than foreign markets. This is why it is also interesting to see how this group compares to other groups, such as Asia.

Many funds claim that they are “mixed funds”, hence hard to classify. However, they do give information as to what their strategy is like. Some mixed funds are cautious, others are balanced, and some are aggressive. With a reasonable amount of funds for each of these three strategies in the sample, I create the groups: “Mixed Cautious”, “Mixed Balanced”, and “Mixed Aggressive”.

Other funds claim to be multimanager or multimarket funds, i.e. being managed by more managers, each managing running his or her own area of knowledge. Each area may be an asset class, a regional market, a sector, or anything else. By nature those funds can be very different from each other. However, they have one thing in common. They are being managed by more professionals, who divide their tasks. Since there is a common denominator, and enough funds available, “MM”

(Multimanager/Multimarket) is another group in my research.

A further group is called “Guarantee”. These funds usually promise a certain percentage payout at some predefined time. The respective fund managers plan carefully ahead.

Some funds exclusively invest in a particular sector, such as the energy or the consumer goods sector. Unfortunately, the sample does not give enough funds to form groups for every single sector. As a matter of fact, only one fund per sector is available in this sample. Although they invest in different sectors, each with different properties, they have a few things in common. Good performance is difficult to achieve, if the sector as a whole is not doing well. Moreover, the correlation with market indices is rather low. Lastly, the investment universe of sector funds is small and managers know the respective industries very well, giving them some skills to possibly perform well. Therefore, all sector funds are put in a group called “sectors”.

(11)

11

Table Ⅱ summarizes the different groups and the amount of funds belonging to each group. Additionally, table A in the Appendix contains a detailed list of all the funds included in this research.

The Groups, Size and Performance

Group Funds Total Return Average Monthly Return Sectors 7 24,7% 0,3% NL 13 17,1% 0,2% Asia 12 15,5% 0,2% Europe 24 14,1% 0,2% RE 11 12,1% 0,2% Hedge 10 11,9% 0,2% Bonds 49 10,8% 0,1% Global 31 9,0% 0,1% Mix Balanced 23 8,4% 0,1% Mix Aggressive 11 6,3% 0,1% Guarantee 75 6,2% 0,1% Mix Cautious 11 6,2% 0,1% Others 33 -8,2% -0,1% USA 6 -14,3% -0,2% MM 8 -52,1% -0,7% All Funds 324 6,2% 0,1% Equity 275 5,4% 0,1%

Table Ⅱ: This table shows the amount of funds belonging to each group, as well as the overall average return of a group for the whole period (76 months). The last column displays the average monthly return.

(12)

12

Not only do I examine the individual groups, but I also run a regression with all 324 funds in the sample. Furthermore, I run another regression model without the bond funds. Fama and French (1993) suggest a Two Factor model for the bond funds, and Hunter et al. (2009) even propose a Six Factor model and six categories of different fixed income funds. My group of bond funds however, is not large enough to cover six groups of different types of bonds. Moreover, the access to

categorization that I have is not sufficient to allow for a separation into different groups. One can therefore assume that the above specified models, largely applicable to equity funds, do not explain much of the bond fund’s regression. This is why I introduce a new group category, excluding all bond funds. For simplicity I call this group “Equity”. Not all remaining funds in the “Equity” group are purely invested in true equity, but most of them are to at least some extent. The “All Funds” and “Equity” groups are also placed in table Ⅱ.

F. Methodology

I use least squares panel analyses techniques for all regressions, in order to obtain common values for Alpha, i.e. values that are universally valid for either group or the whole sample. Equations (1) to (4) are all based on the assumptions of a time series, now these equations need to be transformed in order to be valid for a panel analysis:

𝑅𝑖,𝑡 – 𝑅𝑟𝑓 ,𝑡 = 𝛼𝑔 + 𝛽𝑔(1) ∗ [𝑅𝑚 ,𝑡− 𝑅𝑟𝑓 ,𝑡] + 𝜖 (5) 𝑅𝑖,𝑡 – 𝑅𝑟𝑓 ,𝑡 = 𝛼𝑔 + 𝛽𝑔(1) ∗ 𝑅𝑚 ,𝑡− 𝑅𝑟𝑓 ,𝑡 + 𝛽𝑔 2 𝑆𝑀𝐵 + 𝛽𝑔 3 𝐻𝑀𝐿 + 𝜖 (6) 𝑅𝑖,𝑡 – 𝑅𝑟𝑓 ,𝑡 = 𝛼𝑔 + 𝛽𝑔(1) ∗ 𝑅𝑚 ,𝑡− 𝑅𝑟𝑓 ,𝑡 + 𝛽𝑔 2 𝑆𝑀𝐵 + 𝛽𝑔 3 𝐻𝑀𝐿 + 𝛽𝑔 4 𝑀𝑂𝑀 + 𝜖 (7) 𝑅𝑖,𝑡 – 𝑅𝑟𝑓 ,𝑡 = 𝛼𝑔 + 𝛽𝑔(1) ∗ 𝑅𝑚 ,𝑡− 𝑅𝑟𝑓 ,𝑡 + 𝛽𝑔 2 𝑆𝑀𝐵 + 𝛽𝑔 3 𝐻𝑀𝐿 + 𝛽𝑔 4 𝑀𝑂𝑀 + 𝛽𝑔 5 𝐺𝑅𝑃𝑖,𝑔 + 𝜖 (8)

The subscripts to the Beta values change on the right hand side of the equations to “g”. Instead of showing the sensitivity of a fund to those factors, equations (5) to (8) measure group sensitivities.

The panel estimation methods require the use of either a random or a fixed effects model. This is why at first I perform the Hausman test to all regressions in order to determine which kinds of effects exist.

(13)

13

improve the outcomes, and hence one of my hypotheses. Additionally, for every model I carry out an omitted variables test to see whether the variables of the more advanced models do need to be included in the regression.

In order to observe differences across groups, I compute the average return for every group. This is a dynamic average return series with 76 different observations for every group, as opposed to the GRP factor, which only has two different static observations. The following equation (9) shows how these averages are used in order to obtain differences across groups:

𝐴𝑉𝑔− 𝐴𝑉𝑘 = 𝛼𝑑𝑖𝑓𝑓 + 𝛽𝑔−𝑘(1) ∗ 𝑅𝑚 − 𝑅𝑟𝑓 + 𝛽𝑔−𝑘 2 𝑆𝑀𝐵 + 𝛽𝑔−𝑘 3 𝐻𝑀𝐿 +

𝛽𝑔−𝑘 4 𝑀𝑂𝑀 + 𝜖 (9)

The subscripts g and k denote one of the fifteen groups, and the equation is restricted to g≠k. Hence, I run 15 times 14 different equation regression. The left hand side of equation (9) subtracts two group averages from each other. The right hand side is similar to that of equation (3), namely the CAR4 model. Of all the models available in this paper this is the most advanced to test for group differences. Adding the GRP factor does not make sense, since it is directly correlated with the group average return, and further it would not be clear which GRP one should use. So far all the models are estimated with panel analyses. The model based on equation (9), on the contrary is a time series model, for which I use ordinary least squares estimation techniques. The intercept 𝛼𝑑𝑖𝑓𝑓 is the difference between two group’s averages, and thus signals whether a specific group of fund managers is outperforming or underperforming against another group’s fund managers. The common factors, market, SMB, HML, and MOM again control for effects outside a manger’s abilities.

This paper researches five main areas of interest. Firstly, how well the models describe the regressions for all 15 groups, and the other two sample covering groups. Secondly, what the individual GRP coefficient values are and whether these are statistically significant. Thirdly, and that is the most important field in this research, what the group and sample 𝛼 coefficients are, and how significant these outcomes are. Fourthly, what the differences across group 𝛼 are, and whether there appear improved results from adding the GRP factor. Lastly, I test the statistical significance of the

(14)

14

II. Results

A. The Hausman Test

Table B in the Appendix illustrates the use of the Hausman test to determine whether random effects or fixed effects occur. For every individual group and in turn for every different model within a group this table shows the results of the test. Most of the tests for the individual groups suggest the use of the random effects model. The Hausman tests yield insignificant chi-square statistics, rejecting the null that there are differences between the random and fixed effects model, in which case the random effects model is to be used. For the groups Global, Guarantee, Mixed Balanced, and Multi Manager the Hausman test recommends the use of a fixed effects model. Moreover, applying the Hausman test to the sample groups All Funds and Equity also results in the use of the fixed effects model. The third column of Table A displays the important chi-squared statistics, and column 4 shows which kind of effects are to be used from now on as a result of the test.

B. The Model’s Regression Explanatory Power

Table Ⅲ summarizes the regression outcomes for all the individual groups. The focus here is on how well the model exhibits explanatory power, and whether additional variables should be

employed or omitted. The models, based on equations (5) to (8) are listed in the first column, while the variables that expand the previous model to the actual are listed in the second column. The adjusted R

squared findings are presented in the third column. The fourth column displays the F statistic that

determines whether one can reject the null hypothesis that all variables are zero. The fifth column and the sixth column present the Omitted Variables Test results. In the random effects model only an F

statistic is displayed, while in the fixed effects model an additional Log Likelihood ratio is shown. The

last row of each group’s statistics sheet illustrates the outcome for jointly omitting all variables, while all the preceding tests only check the results for omitting the variable that was added to the previous model. First of all it is worth mentioning that all group’s regressions are significant in terms of the F

statistic (column 4). One can therefore - reject at the level of 1% - the null that all variables are jointly

zero, in each and every model. For every group it is easily observable that with moving from the simple CAPM to the more advanced FF3 and CAR4 models adjusted R squared values rise

(15)

15

respond to the omitted variables test on the GRP factor are: Asia, Global, Guarantee, RE, Equity, and All Funds. For these six groups, H4 cannot be rejected.

Regression Summary

Omitted Var. Test

Model Added

Variables Adjusted R^2 F statistic F statistic Log Likelihood Ratio ASIA

CAPM Market 4,6% 33,5 ** 32,9 **

FF3 SMB, HML 17,8% 49,4 ** 54,4 **

CAR4 MOM 39,5% 110,1 ** 239,1 **

5F GRP 40,0% 90,4 ** 7,3 **

All Jointly Omitted: 89,6 **

EUROPE

CAPM Market 5,9% 91,4 ** 91,4 **

FF3 SMB, HML 26,6% 174,1 ** 202,6 **

CAR4 MOM 50,6% 367,4 ** 693,7 **

5F GRP 50,7% 294,6 ** 2,2

GLOBAL

CAPM Market 3,6% 70,5 ** 70,5 **

FF3 SMB, HML 21,9% 3,8 ** 221,2 ** 436,6 **

CAR4 MOM 43,7% 8,6 ** 646,2 ** 608,0 **

5F GRP 43,9% 8,7 ** 7,0 ** 7,8 **

All Jointly Omitted: 279,0 ** 1128,7 **

GUARANTEE

CAPM Market 4,1% 155,5 ** 155,5 **

FF3 SMB, HML 8,9% 119,2 ** 96,9 **

CAR4 MOM 20,5% 5,1 ** 545,7 ** 539,7 **

5F GRP 20,8% 5,2 ** 14,6 ** 15,6 **

HEDGE

CAPM Market 5,9% 43,5 ** 43,5 **

FF3 SMB, HML 22,8% 67,3 ** 74,5 **

CAR4 MOM 46,1% 145,1 ** 291,0 **

5F GRP 46,1% 116,1 ** 0,6

MIX AGG

CAPM Market 7,7% 54,1 ** 54,0 **

FF3 SMB, HML 35,8% 118,3 ** 138,4 **

CAR4 MOM 56,6% 206,8 ** 302,5 **

5F GRP 56,6% 165,9 ** 1,4

MIX BAL

CAPM Market 7,7% 114,6 ** 114,6 **

FF3 SMB, HML 34,5% 240,0 ** 279,3 **

CAR4 MOM 51,5% 7,3 ** 503,5 ** 500,8 **

5F GRP 51,4% 7,3 ** 0,0 0,0

MIX CAUT

CAPM Market 9,0% 71,1 ** 71,1 **

FF3 SMB, HML 24,8% 79,0 ** 75,6 **

CAR4 MOM 34,3% 93,7 ** 103,4 **

5F GRP 34,3% 75,0 ** 0,4

MM

CAPM Market 3,6% 3,0 ** 1,5 1,5

FF3 SMB, HML 6,6% 4,0 ** 7,6 ** 15,3 **

CAR4 MOM 9,2% 4,9 ** 12,7 ** 12,9 **

5F GRP 9,5% 4,7 ** 2,4 2,5

(16)

16 NL CAPM Market 4,9% 38,1 ** 38,1 ** FF3 SMB, HML 33,1% 119,3 ** 151,9 ** CAR4 MOM 63,5% 313,0 ** 595,7 ** 5F GRP 63,5% 250,1 ** 0,2

OTHERS

CAPM Market 3,6% 72,7 ** 72,7 **

FF3 SMB, HML 11,2% 81,9 ** 83,5 **

CAR4 MOM 21,3% 131,3 ** 248,0 **

5F GRP 20,4% 6,1 ** 0,0 0,0

RE

CAPM Market 8,7% 60,4 ** 60,4 **

FF3 SMB, HML 26,0% 74,1 ** 73,9 **

CAR4 MOM 35,2% 85,6 ** 88,9 **

5F GRP 35,8% 70,7 ** 7,4 **

SECTORS

CAPM Market 6,7% 39,2 ** 39,2 **

FF3 SMB, HML 21,3% 48,7 ** 49,8 **

CAR4 MOM 29,7% 57,0 ** 64,5 **

5F GRP 29,8% 46,0 ** 1,5

USA

CAPM Market 1,5% 6,4 6,4

FF3 SMB, HML 20,5% 30,6 ** 42,0 **

CAR4 MOM 26,0% 31,3 ** 26,4 **

5F GRP 25,9% 25,1 ** 0,5

BONDS

CAPM Market 1,4% 43,6 ** 43,6 **

FF3 SMB, HML 1,8% 20,2 ** 8,4 **

CAR4 MOM 1,8% 15,3 ** 0,5

5F GRP 1,9% 12,8 ** 2,9

ALL FUNDS

CAPM Market 3,5% 684,7 ** 684,7 **

FF3 SMB, HML 13,7% 985,3 ** 1095,3 **

CAR4 MOM 24,8% 1533,6 ** 2742,5 **

5F GRP 24,4% 7,5 ** 86,0 ** 90,3 **

EQUITY

CAPM Market 4,0% 640,8 ** 640,8 **

FF3 SMB, HML 15,1% 4,1 ** 1102,6 ** 2177,2 **

CAR4 MOM 29,3% 8,3 ** 2953,7 ** 2852,2 **

5F GRP 29,5% 8,4 ** 47,9 ** 50,7 **

Table Ⅲ

(17)

17

The first four hypotheses tested are summarized in the table below.

Groups and Hypotheses

Group H1 H2 H3 H4 Asia     Global     Guarantee     RE     Equity     All Funds     Bonds     Europe     Hedge     Mix Agg     Mix Bal     Mix Caut     NL     Others     Sectors     MM     USA     Table Ⅳ

: Hypothesis cannot be rejected; : Hypothesis rejected. Both at the confidence interval of 99%. Results are based the omitted variables test (see table Ⅲ).

(18)

18

Regressions and Coefficients

Group 5F Model CAR4 model

adj. R^2

Panel unbal. observations

GRP Factor

Coeff. t stat α Coeff. t stat α Coeff. t stat NL 63,5% 718 0,061 0,1 (-)0,008 ** -6,6 (-)0,008 ** -6,5 Mix Agg 56,6% 633 0,336 1,2 (-)0,006 ** -5,3 (-)0,005 ** -5,3 Mix Bal 51,4% 1362 0,027 0,2 (-)0,004 ** -6,9 (-)0,004 ** -7,3 Europe 50,7% 1431 0,228 1,5 (-)0,006 ** -6,2 (-)0,006 ** -6,0 Hedge 46,1% 674 0,142 0,8 (-)0,005 ** -4,1 (-)0,004 ** -4,0 Global 43,9% 1851 0,460 ** 2,6 (-)0,006 ** -7,6 (-)0,006 ** -7,3 Asia 40,0% 671 0,400 ** 2,7 (-)0,005 ** -3,1 (-)0,004 ** -2,7 RE 35,8% 625 0,387 ** 2,7 (-)0,006 ** -3,4 (-)0,006 ** -3,2 Mix Caut 34,3% 711 0,135 0,6 (-)0,003 ** -6,3 (-)0,003 ** -6,7 Sectors 29,8% 531 0,282 1,2 (-)0,006 ** -2,9 (-)0,004 ** -2,6 USA 25,9% 346 0,276 0,7 (-)0,009 ** -3,9 (-)0,009 ** -4,3 Guarantee 20,8% 3620 0,697 ** 3,8 (-)0,004 ** -6,2 (-)0,003 ** -5,3 Others 20,4% 1929 0,027 0,1 (-)0,006 ** -5,4 (-)0,006 ** -5,1 MM 9,5% 422 0,656 1,6 (-)0,008 -1,6 (-)0,015 ** -5,0 Bonds 1,9% 3067 2,773 1,7 (-)0,007 -2,3 (-)0,002 ** -4,3 Equity 29,5% 15524 0,369 ** 6,9 (-)0,005 -16,7 (-)0,005 ** -16,0 All Funds 24,4% 18591 0,476 ** 9,3 (-)0,005 -17,6 (-)0,005 ** -16,3 Average (-)0,006 (-)0,006 St. Dev 0,002 0,003 Table Ⅴ

** = significant values (confidence interval of 99%); Average is an equally weighted value of the 15 individual groups (excluding Equity and All Funds). St. Dev is also based on these 15 groups.

The group NL ranks first, based on adjusted R squared possibly indicating that the variables chosen represent the Dutch investment universe very well. Further it is expected that with European variables (market, SMB, HML, MOM) the group Europe should be well explained which it

comparably is. Surprising is the fact that the groups Hedge and RE are well explained by a model that is based on equity. Very poorly explained is the Bonds group, which was anticipated in this paper, and supports the decision to exclude it in the Equity Group. As can be seen at the bottom of the table, the Equity group is considerably better explained by the model than the All Funds group. Furthermore, the rather different group Multi Manager, mainly dissimilar in terms of organization and management, is poorly described by all models.

(19)

19

Funds group, as well as the Equity group. There is no simple and clear explanation as to why exactly these groups gain from the GRP factor. Those groups share medium adjusted R squared values in this sample. They are neither a good nor a poor fit. The best ranked five groups in terms of adjusted R

squared (table Ⅴ) do not benefit from the addition of the GRP factor. Perhaps the available variables

already capture all or most of the commonalities of these funds that would otherwise be stored in the GRP factor. The groups that benefit from GRP, on the other hand have in common that they are only partly (still significant in all six cases) correlated with the European market, SMB, HML, and MOM factors. Their additional correlation with for example the Asian or Global market or MOM factor may be captured by GRP. Although many individual groups do not benefit from the inclusion of the GRP factor, the two most important and representative ones, All Funds and Equity do.

Now that the explanatory power of the models and the usefulness of the GRP factor are discussed, I turn towards the coefficient of the GRP factor of the statistically significant results. Table Ⅴ below shows the estimated GRP coefficient of all the groups in column 4, while the respective t

statistic is shown in column 5. All estimated coefficients are positive and lie between 0,027 and 0,698.

Significant values are available for the groups Asia, Global, Guarantee, RE , All Funds, and Equity. Those are the same groups for which the omitted variable test signaled the need to include the GRP variable. The significant coefficients lie in the range of 0,369 to 0,698, and show that the funds in these groups strongly react on movements of the GRP factor.

C. The Alpha Coefficients

The main goal of this research is to determine whether Dutch fund managers earn abnormal returns compared to what an investor could earn with a passive strategy mimicking a portfolio of the market, SMB, HML, MOM, and some exposure to the group factor. Table Ⅴ presents the intercept, i.e. the Alpha coefficient in column 6. The respective t statistic is in column 7. From that information, one can reject the null in two cases, namely for the Bonds group and the MM group. The other group’s t

statistics imply significant and negative Alpha. Especially the two groups covering most and all of the

(20)

20

D. Differences in Alpha Across the Groups

A comparison of the group’s α in column 6 of table Ⅴ shows that there are differences across groups. Some group’s managers do perform better than others. The disparity between the best group (Mixed Cautious Funds) and the worst (USA funds) is large. On a monthly basis, the Mixed Cautious group loses about 0,3% against the passive strategy, while the USA group loses about 0,9%. Clearly, this opposes my assumption of similar α across the groups for the 5F model. The last row shows the

standard deviation of how the group α differ. This statistic needs to be compared to the standard deviation of group α in the CAR4 model. The CAR4 model α coefficients and respective t statistics

are displayed in the last two columns of table Ⅴ. The standard deviation of α is larger in the CAR4 model than in the 5F model, supporting my assumptions.

Now I turn to the groups, which benefit from the inclusion of the GRP factor (based on the omitted variables test). The α coefficients with respective t statistics, and an average α are presented in table below.

Differences in Alpha: Comparison of 5F and CAR4

5F model CAR4 model

Group α Coefficient

t

statistic Group α Coefficient

t statistic Guarantee (-)0,0037 ** -6,2 Guarantee (-)0,0029 ** -5,3 Asia (-)0,0053 ** -3,1 Asia (-)0,0044 ** -2,7 RE (-)0,0059 ** -3,4 RE (-)0,0056 ** -3,2 Global (-)0,0063 ** -7,6 Global (-)0,0060 ** -7,3 equally weighted average α: equally weighted average α:

-0,0053 -0,0047

All Funds (-)0,0049 ** -17,6 All Funds (-)0,0045 ** -16,3 Equity (-)0,0053 ** -16,7 Equity (-)0,0050 ** -16,0

Table Ⅵ

(21)

21

E. Time Series Regression and Significant Group Alpha Differentials

Next I analyze the time series regression OLS for equation (5). 14 times 15 different regressions yield 210 results for 𝛼𝑑𝑖𝑓𝑓. However, only half of the results are needed, since the other half is symmetrical and with the opposite sign. When subtracting Hedge from Europe, or the other way around, outcomes for 𝛼𝑑𝑖𝑓𝑓 only differ by the sign. The following grid summarizes the results.

Figure 1: The entire grid reads from left to right. The group in the first row is subtracted from the group in the first column. The lower left triangle of the grid is symmetrical to the upper right part. It only highlights the significant (at both the 1%, and 5% level) Alpha differences, instead of copying the opposite values of the upper right part.

One glance at figure 1 suffices to see that underperformance and outperformance occur across these groups. Especially the lower left part of the grid helps to easily understand the outcomes. The Bonds group significantly outperforms most of the other groups, while the groups MM and Others significantly underperform against all others. Reading from left to right shows that e.g. for the group Others many minuses are seen, meaning that it underperforms significantly in these cases. The horizontal concentration of many minuses for a specific group (MM, Others) in a row, as well as vertical concentration of minuses in a column (Bonds, Guarantee, Mixed Cautious, Mixed Balances) signals that differences across groups are observable. These results are in support of hypothesis 6, which cannot be rejected.

The support of the assumption of a higher standard deviation in the CAR4 model (see table Ⅴ), and the support of hypothesis 6, carefully signals that the GRP factor is capable of erasing some of the variation in α across groups.

Asia Bonds Europe Global Gua'tee Hedge Mix Agg Mix Bal Mix Caut MM NL Others RE Sectors USA

(22)

-22

III. Discussion

First of all, I would like to turn to the first four hypotheses. It is expected that the conventional models CAPM, FF3, and CAR4 add value. Too many researchers have already proven that. This is why it might at first appear surprising that for some groups H1 is rejected. However, the groups MM, and USA, which both do not appreciate the CAPM, as the omitted variables test shows, may not correlate with the chosen market proxy, because of geographical reasons. The USA group would probably be better explained by one of the many American market benchmarks. The MM group is not as easy to be judged. The categories Multi Market and Multi Manager suggest investments in

different, also geographically different markets. Apparently they do not invest a lot in the Euro Zone. Furthermore, this could explain their very bad performance based on α. One explanation might be that this group also invested in regions that did not do very well during the last few years (financial crisis 2008, beginning 2009), and this was not captured by the Euro Zone proxy, signaling a potential weakness of the choice of an appropriate market benchmark. However, this exposure should be picked up by the GRP factor, which however also does not add much explanatory power to the regression of the two groups. Lastly, it is important to mention that both groups are fairly small in terms of cross sections, which is why the regression outcomes may be so poor.

The fact that hypotheses 2 and 3are not rejected shows how well - even under difficult and complicated settings – the factors SMB, HML, and MOM work. Especially the MOM factor turns out to be very effective in explanatory power of all groups. This speaks for the good construction of the factor.

The result for hypothesis 4 may be disappointing, given the great success of the GRP factor in the paper of Hunter et al. (2009). The representative Equity group signals an explanatory power of almost 30% under the 5F model. This value is very low compared to other researcher’s results. Direct comparison with the Five Factor model of Hunter et al. (2009) further deepens this discrepancy. Their

R squared values do not fall below 80%. The reasons for these differences are, however logically

explainable. My sample covers six years and 4 months, while their research covers the time period from 1984 until 2007, hence 23 years. Furthermore, their groups are clearly defined based on strategy and do not include more exotic funds, such as hedge funds, commodity funds, real estate funds and others. This is also why their group factor captures so much of the explanatory power, as opposed to my model, where the group selection is not based on strategy, but largely on other features, mainly geographical. For future research, I therefore recommend to base group selection on strategy, instead of totally ignoring the GRP factor. Some groups were well chosen others less, as the omitted variables tests for the GRP factor show.

(23)

23

until 2009. This is in contrast to most of the results so far obtained, where usually Dutch funds performed well in international comparison. A yearly underperformance of 6% before fees does not speak for investments with Dutch funds, and supports the view of many researchers.

The differences between group α decrease, when moving from the CAR4 model to the 5F model. This is shown in table Ⅴ, and supports the assumption that variation in α across groups is larger in the CAR4 model than it is in the 5F model. Hence, after accounting for group specific common exogenous powers, α differentials between groups decline.

Figure 1 also shows that especially the groups of funds that are heavily invested in fixed income securities do perform well. Bond funds are by nature restricted to fixed income securities only, Mixed Cautious, and Mixed Balanced funds need to maintain low risk, and Guarantee funds also require large investments in fixed income securities and low risk products to ascertain their returns. These four group’s managers outperform the other group’s managers. The question remains whether an exogenous fixed income proxy would take away these differences in α. Alternatively, one could argue that these groups of funds are being managed more passively, further supporting the view that active fund management does not yield high returns. Without further empirical results this cannot be answered. But it remains an interesting field of research.

(24)

24

IV. Conclusion

(25)

25

Appendix

Table A

Group Fund Group Fund Group Fund

ASIA AEGON EQ.PACIFIC FD.CAP. EUROPE ACHMEA ER.AANDELENFONDS CAP DEAD GUARANTEE AA BEST OF WORLD GARANTIE CAP.

BLACKR.ML INV.MGRS.LX. IIF WLD.EN.C2 EUR AEGON EUROPA INDEX CAP ABN AMRO PARAPLU FONDS 5 X 5 GARANTIE FONDS DS. CAPITAL INTERTIOL FD.CIF JAPAN EQ.C EUR ALLIANZ FD.EUROPEES FONDS CAP. ABN AMRO PARAPLU FONDS RENTE GIGANT GARANTIE FALCON JAPE.AANDELEN FON.CAP ASN BELEGGINGSFONDSEN NV ASN SML.& MIDCAP FON.DS. ALLIANZ FD.GARANTIE FONDS 1992 CAP. FALCON VERRE OOSTEN AANDELEN FON.CAP DELTA LLOYD INSTL. EUROPA FONDS CAP ALLIANZ FD.GARANTIE FONDS 1993 CAP. FORTIS APHI FD19 OPF DELTA LLOYD INSTL.BLUE RET.FD.CAP. CINTINU GARANTIEFONDS CAP. INSINGER ASTMGMT.IDB MULTI MAGER JAPAN DELTA LLOYD SLT.DIVIDEND FONDS NV DS. CORDARES GARANTIEFONDS CAP. NN VERRE OOSTEN FONDS CAP FALCONEUROPESE AANDELEN FONDS CAP DE GOUDSE VERZEKERD RENDEMENT TR2 CAP. PICTET GESTION FUNDS JAP.EQ.SELECTION HP CAP HAUSSMANN HDG.HAUSSMANN C EURO DE GOUDSE VERZEKERD RENDEMENT TR3 CAP. REAAL PP AANDELENFONDS VERRE OOSTEN CAP. IDX.UMBR.POOL EUR.EQ. IDX.POOL DS. DE GOUDSE VERZEKERD RENDEMENT TR4 CAP. ROBECO JAPAN JUL 00/07 CAP ING PREMIUM DIVIDEND FD. DRESDNER HOLLAND CONTINU KLIK GARANTIEWRD STEWART ASIAN HOLDINGS CAP. INSINGER ASST.MAN.IDB MM EMRG.COS.ER. FORTIS GLOBAL CLICK CAP.

BONDS ACHMEA ER.OBLFD.CAP DEAD MEERWAARDE DEPOT CAP FORTIS PARAPLU FONDS 5X5 GARANTIE II DS. ALLIANZ FD.AMERIKAANSE DOLLAR FONDS CAP. MEI ROM EN BUL FD FORTIS PARAPLU FONDS RENTE GIGANT GNT.II DS. ALLIANZ FD.AUSTRALISCHE DOLLAR FONDS CAP. NN EUROPA FONDS CAP GARANTIEFONDS CAP.

ALLIANZ FD.BRITSE PONDEN FONDS CAP. ORANGE SENSE FUND DS. GENERALI GARANIEFONDS CAP. ALLIANZ FD.DEPOSITO FONDS CAP. PSP AANDELENFONDS EUROPA CAP. GENERALI GARANTIEFONDS II CAP. ALLIANZ FD.MANHIGH CAP. REAAL PP AANDELENFONDS EUROPA CAP. GLOBAL FIX CERTIFICATEN 100 % 0104 ALLIANZ FD.OBLIGATIE FONDS CAP. REAAL PP OPTIMAALFONDS EUROPA CAP. GLOBAL FIX CERTIFICATEN 100 % 0105 CAP. AXA VGF ABN AMRO OBLFD. CAP. ROBECO EUROSTOXX50 GARANT DEC09 CAP GLOBAL FIX CERTIFICATEN 100 % 0204 AXA VGF DEPOSITOFONDS CAP ROBECO EURST 50 APR 01/08 CAP GLOBAL FIX CERTIFICATEN 100 % 0205 CAP. AXA VGF DOLLARS VASTRENTEND CAP ROBECO EUS OKT 01/08 CAP GLOBAL FIX CERTIFICATEN 100 % 0304 AXA VGF EMS PLUS RENTEFONDS CAP RVS EUROPA FONDS CAP GLOBAL FIX CERTIFICATEN 100 % 0305 CAP. AXA VGF OBLIGATIE CAP SNS EURO AANDELEN PLUSFONDS CAP. GLOBAL FIX CERTIFICATEN 100 % 0404 AXA VGF VASTRENTENDE WAARDEN CAP GLOBAL AA GTD.GLB.FD.CAP DEAD - DEAD GLOBAL FIX CERTIFICATEN 100 % 0504 AXA VGF VREEMDE VALUTA CAP AEGON WERELD INDEX CAP GLOBAL FIX CERTIFICATEN 100 % 0505 CERTIFICAAT TRIODOS BANK DS. AEGON WORLD EQ.FD.CAP. GLOBAL FIX CERTIFICATEN 100 % 0604 CORDARES ZEKERHEIDSFONDS CAP. ALLIANZ FD.AANDELEN FONDS CAP. GLOBAL FIX CERTIFICATEN 100 % 0605 CAP. DELT.LLOYD INSTITUTIONEL OBLFD.XLT CAP. ALLIANZ FD. INTERTIOAL FONDS CAP GLOBAL FIX CERTIFICATEN 100 % 0704 DELTA LLOYD INSTL. CREDIT FONDS HR CAP AXA PARAPLU FONDS AXA MODEL FD.V CAP. GLOBAL FIX CERTIFICATEN 100 % 0705 CAP. DELTA LLOYD INSTL. CREDIT FONDS LR CAP AXA VGF AANDELEN INTERTIOAL CAP GLOBAL FIX CERTIFICATEN 100 % 0805 CAP. DELTA LLOYD INSTL. OBLIGATIEFONDS LT CAP AXA VGF GLOBAL FUND CAP GLOBAL FIX CERTIFICATEN 100 % 0905 CAP. DELTA LLOYD INSTL. OBLIGATIEFONDS MT CAP AXA VGF GLOBAL SELECT CAP GLOBAL FIX CERTIFICATEN 100 % 1004 CAP. EIGERFONDS CAP. CORDARES AANDELENFONDS CAP. GLOBAL FIX CERTIFICATEN 100 % 1104 CAP. FALCON BLUE STAR CAP DELTA LLOYD INSTL. WERELD FONDS CAP GLOBAL FIX CERTIFICATEN 100 % 1203 CAP. FALCON ER.DEPOSITO FON. CAP DELTA LLOYD INSTL.GLB. DEAD 28/09/06 GLOBAL FIX CERTIFICATEN 100 % 1204 CAP. FALCONER.OBLIGATIE FON. CAP DELTA LLOYD INSTL.GLB. TRENDS FONDS CAP GLOBAL FIX CERTIFICATEN 100 % 904 CAPITAL FLEXICLICK DS. DEAD 01/01/08 FALCON MILIEU AANDELEN FON.CAP GLOBAL FIX CERTIFICATEN 100% 0405 CAP. GENERALI DEPOSITOFONDS CAP. FALCON ORANGE STAR CAP GLOBAL FIX CERTIFICATEN 75 % 0105 CAP. GENERALI OBLIGATIEFONDS CAP. FORTIS GHIE FD18 OF GLOBAL FIX CERTIFICATEN 75 % 0205 CAP. HOLLAND STALLINGSREKENING CAP. FORTIS STRATEGIE FONDSEN DUURZAAM DOTIE FON.DS. GLOBAL FIX CERTIFICATEN 75 % 0305 CAP. ING FIRST CLASS OBLIGATIE FONDS GLOBAL FUTURE FD. CAP. GLOBAL FIX CERTIFICATEN 75 % 0505 CAP. ING HOOG DIVIDEND OBLIGATIE FONDS HOLLAND INTERTIOAL AANDELENFONDS CAP. GLOBAL FIX CERTIFICATEN 75 % 0605 CAP. INSINGER ASST.MAN.IDB DOM GLB.CV.ER. HOLLAND WERELD GLOBAL FIX CERTIFICATEN 75 % 0705 CAP. IVM PARALUFONDS IVM RENTE FONDS CAP ING DYMIC MIX V CAP. GLOBAL FIX CERTIFICATEN 75 % 0805 CAP. KEMPEN EURO CREDIT INSINGER ASST.MAN.IDB MM GLB.EQ.ER. GLOBAL FIX CERTIFICATEN 75 % 0904 CAP. MEESMAN IDX.UMBR.FD. MEESMAN EURO BD.IDX.FD. MEESMAN IDX.UMBR.FD. MEESMAN GLB.STK.IDX.FD. GLOBAL FIX CERTIFICATEN 75 % 1004 CAP. NN EUROPA RENTE FONDS CAP NN AANDELEN FONDS CAP GLOBAL FIX CERTIFICATEN 75 % 1104 CAP. NN GELDMARKET FONDS CAP PIONEER INV.MAN.FUND.TOP GLB.PLAYERS A USD DS. GLOBAL FIX CERTIFICATEN 75 % 1203 CAP. NN INTER RENTE FONDS QUANTUM GROUP ENDOWMENT A2 USD NAV GLOBAL FIX CERTIFICATEN 75 % 1204 CAP. NN RENTE FONDS CAP QUANTUM GROUP ENDOWMENT E2 EUR NAV GLOBAL FIX CERTIFICATEN 75 % 0905 CAP. PSP DEPOSITOFONDS NEDERLAND CAP. REAAL PP AANDELENFONDS INTERNATIONAAL CAP. GLOBAL FIX CERTIFICATEN 75% 0104 PSP VASTRENTEVDFONDS DEAD 03/03/08 REAAL PP OPTIMAALFONDS INTERNATIONAAL CAP. GLOBAL FIX CERTIFICATEN 75% 0204 REAAL OBLIGATIEFONDS INTERNATIONAAL CAP. SNS GROEIMARKT PLUSFONDS CAP. GLOBAL FIX CERTIFICATEN 75% 0304 REAAL PP DEPOSTIOFONDS NEDERLAND CAP. WINTERTHUR AANDELENFONDS CAP. GLOBAL FIX CERTIFICATEN 75% 0404 REAAL PP VASTRENTENDFOND NEDERLAND CAP. GLOBAL FIX CERTIFICATEN 75% 0405 CAP. RVS OBLIGATIEFONDS GLOBAL FIX CERTIFICATEN 75% 0504 SNS OBLIGATIE PLUSFONDS CAP. GLOBAL FIX CERTIFICATEN 75% 0604 TRIODOS FAIRSHARE FUND CAP GLOBAL FIX CERTIFICATEN 75% 0704 WINTERTHUR OBLIGATIEFOND NEDERLAND CAP. HOLLAND GARANTIE CERTIFICAAT CAP. WINTERTHUR OBLIGATIEFONDS CAP. HOLLAND GARANTIE FONDS CAP.

(26)

26

Table B: This table exhibits all the names of the 324 funds in the sample, sorted by their membership to any of the 15 groups.

Group Fund Group Fund Group Fund

MIX AGG ALLIANZ FD.COMBITIE FONDS CAP. NL AEGON NEDERLAND INDEX CAP HEDGE ATTICA LONG/SHORT SELECTOR DS. AXA PARAPLU FONDS AXA MODEL FD.IV CAP. FALCON NEDSE.AANDELEN FON.CAP GAM OVERLAY CHF R

AXA VGF MIX FUND CAP GENERALI AANDELENFONDS CAP. IDB DOM ALCHEMY ER

FALCON RED STAR CAP HOLLAND FUND INSINGER ASST.MAN.IDB ABS.RTN.DIRCTL MGRS.A HOLLAND OFFENSIEF MIX FONDS CAP. ING AEX SHADOW INSINGER ASST.MAN.IDB ABS.RTN.DIRECT MGRS B ING DYMIC MIX IV CAP. NN NEDERLAND FONDS CAP INSINGER ASST.MAN.IDB ABST.RTN.MKT.NTL.A PSP LOMBARD ODIER MIXFUND CAP. REAAL PP AANDELENFONDS NEDERLAND CAP. INSINGER ASST.MAN.IDB ABST.RTN.MKT.NTL.B SNS COMBINATIE PLUSFONDS CAP. REAAL PP OPTIMAALFONDS NEDERLAND CAP. LEVERAG CAPHOLD

SNS OPTIMAAL ROOD DS. ROBECO AEX MRT 02/07 CAP DEAD 02/04/07 ING BASIC MATS.FD.DS. VAN LANSCHOT MS OFFENSIEF CAP. ROBECO AEX OKT 02/07 CAP DEAD 29/10/07 RO COMMODITY FUND WINTERTHUR MIXFONDS NEDERLAND CAP. RVS AANDELEN FONDS CAP MM ATTICA NEUTRAL SELECTOR DS.

MIX BAL ACHMEA ER.MIX 50/50 CAP DEAD SNS NEDERLANDS AANDELEN PLUSFONDS CAP. FALCINVEST MASTER FUND GLB.MARKETS FUND B CAP. ACHMEA ER.MIX80/20 CAP DEAD WINTERTHUR AANDELENFONDS NEDERLAND CAP. FALCINVEST MASTER FUND OPPS.FUND B CAP. ALLIANZ FD.MANLOW CAP. OTHERS ARNHOLD & S BLEICHROEDER FIRST EAGLE B FINLES EUR.SELECTOR FON. CAP

ARNHOLD & S BLEICHROEDER SOFIRE LTD ARNHOLD & S BLEICHROEDER FIRST EAGLE E PERMAL MLTMGR.HAUSMANN ASIAN HOLDINGS A $ AXA PARAPLU FONDS AXA MODEL FD.III CAP. ARNHOLD &.S BLEICHROEDER FIRST EAGLE C USD PERMAL MM JAPAN HLDG. B/C USD

AXA VGF ACTIEF BEHEER CAP ASN NOVIB FONDS CAP VERMEER LOW VOLATILITY FUND CAP. AXA VGF COMBITIE DS. AXA UNIVERSAL LIFE AXA MODEL DEPOT I CAP. VERMEER MID VOLATILITY FUND CAP. CORDARES COMBITIEFONDS CAP. AXA UNIVERSAL LIFE AXA MODEL DEPOT II CAP. RE AGEON WERELDW VASTGOED DELTA LLOYD PROFIELFONDSEN 1 CAP. AXA UNIVERSAL LIFE AXA MODEL DEPOT III CAP. AXA VGF VASTGOED CAP DELTA LLOYD PROFIELFONDSEN 2 CAP. AXA UNIVERSAL LIFE AXA MODEL DEPOT IV CAP. AZL VASTGOED WONINGEN DS. DELTA LLOYD PROFIELFONDSEN 3 CAP. AXA UNIVERSAL LIFE AXA MODEL DEPOT V CAP. DELTA LLOYD INSTL.REAL ESTATE FD.CAP. DELTA LLOYD PROFIELFONDSEN 4 CAP. AXA VGF COLLECTIEF BEHEER CAP FORTIS ASR VASTGOEDFONDS CAP. DELTA LLOYD PROFIELFONDSEN 5 CAP. AXA VGF GARANDMENT CAP MIDDLE EUROPE REAL ESTATES NV FALCON PURPLE STAR CAP AXA VGF HOLLANDHAVEN CAP NN VASTGOED FONDS CAP

FORTIS ALLINC OF AXA VGF ROBECO CAP REAAL PP VASTGOEDFONDS INTERNATIONAAL CAP. GENERALI COMBITIEFONDS CAP. DELTA LLOYD BK.FD.ER. ROBECO HOOG DIVIDEND ONROEREND GOED DS. HOLLAND GROEN RENTEFONDS DS. DELTA LLOYD BK.ROLINCO ER. SNS VASTGOED PLUSFONDS CAP.

ING DYMIC MIX III CAP. DELTA LLOYD INSTL.BLUE SLT.FD.CAP. VASTGOED FUNDAMENT FONDS NV C NN MIX FONDS CAP DELTA LLYOD INSTITUTIONEEL SECTORS AXA VGF FINANCIELE SECTOR CAP SNS OPTI ORANJE 18 FINLES STAR SELECTOR FUND CAP. ING DLY.CSM.GOODS FUND VAN LANSCHOT MS GROEIGERICHT CAP. GESTION ACTIEF BEHEER FONDS CAP. ING ENERGY FUND VAN LANSCHOT MS NEUTRAAL CAP. HOMBURG INVEST CAP. ING INDLS.FD. VPV VALUE FUND CAP. INSINGER ASST.MAN.IDB DOM ER.MIX ER. ING LUX.CSM.GOODS FD. MIX CAUT AXA PARAPLU FONDS AXA MODEL FD.I CAP. INTEREFFEKT CHI WARRANT CAP. ING TELC.SVS.FD.

AXA PARAPLU FONDS AXA MODEL FD.II CAP. JUNGFRAUFONDS CAP. ING UTILS.FD.

AXA VGF ABN AMRO ALL IN FUND CAP. OIKOCREDIT NEDERLAND FONDS CAP USA AEGON EQ.NORTH AMERICA FUND CAP. AXA VGF ALL IN FUND CAP PANEUROLIFE HIGH POWER FUND DS. ALLIANZ HOLLAND AMERIKA

FALCON GREEN STAR CAP PRISMA PLUS FONDS CAP FALCON NOORD AMERIKAANSE AANDELEN FON.CAP HOLLAND DEFENSIEF MIX FONDS CAP. RABO.LEDENCERTIFICAAT 3 DS. IDX.UMBR.POOL NTH.AMER. EQ.IDX.POOL DS. ING DYMIC MIX I CAP. RABOBANK LENDERCERTIFICAAT 1 CAP. OPTIMIX AMERICA FD

ING DYMIC MIX II CAP. RABOBANK LENDERCERTIFICAAT 2 CAP. REAAL PP AANDELENFONDS AMERIKA CAP. RVS MIX FONDS CAP ROBECO WLDTOP50GARANT ME I12 DS

SNS OPTIMAAL GEEL DS. SNS SPAARTEWUSTFONDS CAP. VAN LANSCHOT MS DEFENSIEF CAP. TRIODOS CULTUURFONDS DS.

(27)

27

Table B

The Hausman test is based on the Chi-square Distribition. The CAPM model has 1 degree of freedom, FF3 has 3 degrees of freedom, CAR4 has 4 degrees of freedom, and lastly the 5 Factor model has 5 degrees of freedom. The Chi-square statistics are bold, if the null is rejected at a confidence interval of 99%.

Hausman Test Hausman Test

Group Model Chi-2 stat Effects? Group Model Chi-2 stat Effects?

ASIA CAPM 2,1828 random NL CAPM 0,4299 random

FF3 3,5538 random FF3 3,8596 random

CAR4 4,4720 random CAR4 5,0463 random

5F 3,6931 random 5F 8,6239 random

EUROPE CAPM 6,6324 random OTHERS CAPM 0,0000 random

5F 5,3431 random 5F 16,6461 fixed

GLOBAL CAPM 2,7262 random RE CAPM 0,7746 random

FF3 18,6276 fixed FF3 2,0689 random

CAR4 36,3999 fixed CAR4 1,1726 random

5F 41,4105 fixed 5F 1,0645 random

GUARANTEECAPM 0,9276 random SECTORS CAPM 0,5101 random

CAR4 16,4408 fixed CAR4 0,0000 random

5F 18,0279 fixed 5F 0,0000 random

HEDGE CAPM 0,0002 random USA CAPM 0,1830 random

MIX AGG CAPM 0,0080 random BONDS CAPM 0,0832 random

MIX BAL CAPM 0,0390 random

FF3 4,7705 random

CAR4 13,5670 fixed

5F 16,7400 fixed

MIX CAUT CAPM 2,6589 random ALL FUNDSCAPM 5,1737 random

5F 2,7750 random 5F 19,7247 fixed

MM CAPM 7,3933 fixed EQUITY CAPM 2,1536 random

FF3 19,4982 fixed FF3 14,7072 fixed

CAR4 22,0276 fixed CAR4 25,6036 fixed

(28)

28

References

Bauer, R., Koedijk, K., Otten, R., 2004, International evidence on ethical mutual fund performance and investment style, Journal of Banking & Finance, Vol. 29, 1751-1767

Blake, D., and Timmerman, A., 1998, Mutual fund performance: evidence for the UK, Eurpoean

Finance Review, Vol. 2, 57 - 77

Carhart, M., 1997, On persistence in mutual fund performance, Journal of Finance, Vol. 52, 57 - 82 Dahlquist, M., Engström, S., Söderlind, P., 2000, Performance and characteristics of Swedish mutual funds, Journal of Financial and Quantitative Analysis, Vol. 35, 409 - 423

Fama, E. and French, K.R., 1993, Common risk factors in the returns on stocks and bonds, Journal of

Financial Economics 33, 3-56

Gruber, M., 1996, Another puzzle: the growth in actively managed mutual funds, Journal of

Finance, Vol. 51, 783 - 806

Huij, J., and Verbeek, M., 2009, On the Use of Multifactor Models to Evaluate Mutual Fund Performance, Financial Management, Spring 2009, 75-102

Hunter, D., Kandel, E., Kandel, S., and Wermers, R., 2009, Endogenous Benchmarks, Working Paper Jensen, M., 1968, The performance of mutual funds in the period 1945 – 1964, Journal of

Finance, Vol. 23, 389 - 416

Kempf, A., Uhrig-Homburg, M., 2000, Liquidity and its Impact on Bond Prices, Schmalenbach

Business Review, Vol.52, 26-44

Malkiel, B., 1995, Returns from investing in equity mutual funds 1971 – 1991, Journal of

Finance, Vol. 50, 549 - 573

McDonald, J., 1973, French mutual fund performance: evaluation of internationally-diversified portfolios, Journal of Finance, Vol. 28, 1161 - 1180

Otten, R., and Bams, D., 2002, European Mutual Fund Performance, European Financial

Management, Vol. 8, 75 - 101

(29)

29

Databases and WebPages