Selection of Dutch non-listed real estate
investment funds using DEA
Stichting Gasunie Pensioenfonds
Selection of Dutch non-listed real estate
investment funds using DEA
Stichting Gasunie Pensioenfonds
H. Groen Damsterdiep 269-122 9713 EE Groningen Student number: 1666487 B.D. van ’t Veld Van Heemskerckstraat 3-16 9726 GB Groningen Student number: 1383728 Institution: Faculty of Economics University of Groningen Department:
Finance, Investment & Accounting Supervision N.V. Nederlandse Gasunie: Drs. J. de Boer CT
PREFACE
This thesis is end result of our internship at the treasury department of the N.V. Nederlandse Gasunie. The thesis completes our Master of Science education in Business Administration, specialization Finance. The goal of our five month internship is, to find out which Dutch non-listed real estate investment fund(s) is (are) most suited for investment. Scores on selection criteria for the funds are used and these scores are the input of a technique called data envelopment analysis. The investment recommendation is based on the results of data envelopment analysis in combination with the results of regression analysis.
We would like to thank drs. J. de Boer CT, dhr. P.C. Molenaar RBA, and drs. C. van Winsum for their support and nice cooperation.
Besides them we also would like to thank our university supervisor, dr. A. Plantinga for his critical remarks and instructions during our internship.
Groningen, August 2007 H. Groen & B.D. van ’t Veld.
MANAGEMENT SUMMERY
The pension fund of the N.V. Nederlandse Gasunie invests, like other pension funds, in different investment assets. Besides the well known investment assets as stocks and bonds, real estate is part of the investment portfolio of the pension fund. Real estate can be bought directly by acquiring, for example, office buildings. On the other hand, real estate can be acquired indirectly by means of buying shares in a real estate investment fund. The focus of this research is on non-listed real estate investment funds. More specifically, the funds have to be statutory founded in the Netherlands and got to have their investment portfolio in the Euro zone. The goal of this research is, to find out which non-listed real estate investment fund(s) is (are) most suited for investment by the pension fund of the N.V. Nederlandse Gasunie. Data envelopment analysis is used to accomplice this goal and regression analysis is used as a check of the results.
An important aspect of data envelopment analysis is the specification of inputs and outputs. The inputs are the fees, the leverage, and the risk. The outputs are the historic return, the transparency, and the tradability. In data envelopment analysis the inputs are minimized and the outputs are maximized. The most important criterion for the pension fund of the N.V. Nederlandse Gasunie is transparency. Transparency must be as high as possible. Therefore, the criterion transparency is given the highest weight range in data envelopment analysis.
The results of data envelopment analysis show that there are four funds with the maximum efficiency score of one. The funds are ING Dutch Residential Fund, ING Dutch Office Fund, Altera Vastgoed Winkels, and Altera Vastgoed Woningen. The best semi-efficient funds are Altera Vastgoed Bedrijfsruimten, Altera Vastgoed Kantoren, ING Dutch Retail Fund, and ING Real Estate Nordic Property Fund. It can be said that Altera Vastgoed and ING are most suited for investment by the pension fund of the N.V. Nederlandse Gasunie, according to data envelopment analysis.
The results of the regression analysis show that the funds of Altera Vastgoed and ING again rank highest. Among the five best funds, there are three funds of Altera Vastgoed and two funds of ING.
TABLE OF CONTENTS PREFACE ...iii MANAGEMENT SUMMERY... iv TABLE OF CONTENTS ... vi 1. INTRODUCTION... 1 2. LITERATURE ... 3
2.1. SELECTION CRITERIA OF REAL ESTATE FUNDS ... 3
2.2. FUND SELECTION USING DATA ENVELOPMENT ANALYSIS... 4
2.3. COMPARISON BETWEEN REGRESSION ANALYSIS AND DEA ... 5
3. DATA AND METHODOLOGY ... 7
3.1. METHODOLOGY... 7
3.1.1. DATA ENVELOPMENT ANALYSIS ... 7
3.1.2. REGRESSION ANALYSIS ... 11
3.1.3. CRITERIA ... 12
3.2. DATA... 14
3.2.1. RESTRICTIONS... 14
3.2.2. PROCESS OF INFORMATION GATHERING ... 15
3.2.3. SCORES... 16
3.2.4. CRITERIA ... 18
4. RESULTS... 20
4.1. EFFICIENCY SCORES... 20
4.2. WEIGHTS OF THE CRITERIA... 21
4.2.1. OUTPUT CRITERIA... 22
4.2.2. INPUT CRITERIA... 24
4.3. REGRESSION ANALYSIS RESULTS ... 26
5. RECOMMENDATIONS AND CONCLUSION... 29
5.1. RECOMMENDATIONS ... 29
5.2. CONCLUSION ... 31
REFERENCE LIST... 32
WEBSITES ... 33
1. INTRODUCTION
Stichting Gasunie Pensioenfonds (GUPF) manages the pensions of the employees of the N.V. Nederlandse Gasunie. The goal of the GUPF, like other pension funds, is to make sure that there are enough assets to match the current pension liabilities and the pension liabilities in the future. To attain that goal, the GUPF invests its assets in different asset classes, for example stocks, bonds, and real estate. This research explores the Dutch market for non-listed real estate investment funds.
The real estate market is a fast growing market (webmagazine CBS, 2005). Over the last years investors in real estate managed to realize good results on their real estate investments (Lymos notitie, November/December 2006). Nico Tates, CEO of Aberdeen Property Investors Europe, mentioned in an interview with NPN the recent value increase of listed real estate funds (NPN, juni/juli 2006). According to Tates the value of the funds increased above intrinsic values because of the high demand in the market. Changes in the prices of real estate funds were not fully representing the underlying value, but were partially the result of market sentiment (changes in the supply and demand in the market for the funds).
Institutional investors searched for alternative investment opportunities to get rid of the market sentiment but keep all the characteristics that real estate provides. The solution is non-listed real estate investment funds. Non-listed real estate funds are not listed on an exchange, and therefore less vulnerable to market sentiment. The result is that non-listed real estate funds are a good diversification tool in an investment portfolio.
Non-listed real estate investment funds are the subject of this master thesis. A research is conducted to create an overview of relevant suppliers. In contrast to exchange traded real estate funds there is no complete overview of all the suppliers of non-listed real estate funds. The lack of an overview results in a time-consuming study to find the different suppliers. From the created overview the funds that do not suit the demands that are set by the GUPF are subtracted.
Second, if a good overview is realized a comparison between the different suppliers is more difficult to make than when dealing with listed funds. Non-listed funds are not required to give as much information as listed funds. Less information about non-listed funds is therefore available. A number of criteria is used for the comparison between the different funds. The number of criteria results in difficulties when using standard analysis tools. To cope with the difficulties a more advanced tool called data envelopment analysis is used.
What is (are) the best Dutch non-listed real estate investment fund(s) for the Stichting Gasunie Pensioenfonds to invest in?
To answer this main question, the next sub questions are taken into account: which relevant funds comprise the market for Dutch non-listed real estate? What are the criteria for selecting real estate investment funds? By using the data envelopment analysis method, we provide the GUPF with a well funded recommendation for the funds best suited for investment.
2. LITERATURE
In this section the relevant literature, with regard to the research question, will be discussed. First, a paper by Van Aert (2006) about selection criteria for real estate is discussed. The criteria can be of use in the research that is conducted in this thesis. After the criteria two papers about the use of data envelopment analysis for the selection of investment funds are discussed. These papers use the same methodology to select the best mutual fund on the one hand and the best hedge fund on the other hand. The same methodology is used in this thesis to select the best non-listed real estate fund. The last part of this section is a comparative study between data envelopment analysis and regression analysis.
2.1. SELECTION CRITERIA OF REAL ESTATE FUNDS
According to a survey (Van Aert, 2006), executed by the European association for investors in non-listed real estate vehicles (INREV), investors use seven criteria for the selection of real estate investment funds. These criteria are, in decreasing order of importance, the track record, the style, the sector, the location, the corporate governance, the structure, and the fees.
The track record of the fund manager is the most important criterion in the fund selection process (Van Aert, 2006). The track record consists, among others, of the past financial performance (of the manager’s funds) and experience of the manager, with regard to style and working area.
The style of a fund tells something about its risk and return profile. A distinction can be made between three styles; core, value-added and opportunistic. In short the expected return and leverage increase from core to opportunistic and the holding period decreases in the same direction. Core funds are composed of fully leased properties. The tenants usually have a contract for a longer period of time and the rents increase with at least the inflation rate. The value-added style creates added value for existing real estate properties by repositioning and redevelopment. The repositioning is done by buying properties with tenant contracts for short periods of time and under market value. These contracts are changed to at least market value contracts. The redevelopment is done by buying property and renovating it. Subsequently the new property is sold for more than the renovation and acquisition cost together and is thus creating added value. Opportunistic funds seek to generate return from the value increase in real estate, by using inefficiencies in the market. This can be done in different ways. One example is the ‘wholesale and retail’ transaction. A large portfolio of real estate is bought and next each property is sold separately to generate return.
portfolio of the investor. The current portfolio influences the investment decision. An absent sector should be added and therefore the sectors already present are not taken into account in the investment decision (Van Aert, 2006).
On the macro level investors look at the existing investments, the growth, and the return expectations. Together with the size of the market for real estate, the location where they are currently invested in, and the locations they are willing to invest in.
Two other selection criteria are the corporate governance and structure of the fund. The corporate governance is the system by which business corporations are directed and controlled. The non-listed real estate sector does not have generally accepted rules yet. Examples of legal structures are; limited partnerships, bonds, commanditaire vennootschappen and the Luxembourg’s funds common placement.
Fees, that an investor has to pay, are the last selection criterion that is mentioned. There are numerous ways to let an investor pay for the fund. Examples of the fees that can be charged is; set-up fees, placement fees, acquisition fees, base management fees, and performance fees.
2.2. FUND SELECTION USING DATA ENVELOPMENT ANALYSIS
A few papers are written about the performance measurement of investment funds and a few about the selection of investment funds using date envelopment analysis. Two of these papers will be discussed. One is written by McMullen and Strong (1998) and is about the selection of mutual funds. The second is written by Huyen (2006) and has the selection of hedge funds as topic.
In standard portfolio theory, investors only consider the risk and return profile of an investment for making an investment decision. Both authors add however additional variables to the model as a result of new insights, like behavioral finance etc.
Due to the lack of available information Huyen (2006) only uses statistical information like skewness and kurtosis to compare the different funds among each other. McMullen and Strong (1998) use more economic information, like change in sales and expense ratio, as variables in their model. All authors added additional variables to compare the funds among each other and pick the most suited fund.
The data envelopment analysis model is appealing because in both situations a large number of variables make up the model. This model can cope with multiple and different criteria and is able to add weights to the different variables. In this way the data and criteria correspond with the limitations and requirements set by the investor.
Both authors want to maximize the return - and skewness in the case of Huyen (2006) - and minimize the risk, kurtosis (Huyen, 2006) and expense ratio (McMullen and Strong, 1998). The minimization and maximization leads to a number of funds (McMullen 12 and Huyen 5) that are efficient. Efficient funds (and also the near efficient funds) have most of their variable-scores at a desirable level.
In the study of McMullen and Strong (1998) the larger stock funds do not meet the desirable scores on most of the variables. The conclusion McMullen and Strong (1998) therefore make is that large stock funds perform poorly for their fictitious individual investor.
Huyen (2006) only uses statistical variables in his analysis. He is therefore able to compare the outcome of data envelopment analysis with the Sharpe ratio to check the fund ranking. The fund rankings of the Sharpe ratio and data envelopment analysis are comparable and have a correlation of almost 1. This similarity is indirect evidence of the reliability of data envelopment analysis.
Huyen (2006) concluded his paper with a robustness test. With this test Huyen (2006) makes a distinction between the individual efficient funds that are generated by data envelopment analysis. The fund that is capable of weakening his inputs and outputs the most without losing efficiency is assumed to be the most desirable of the efficient funds.
Both papers illustrate that data envelopment analysis is a useful tool to select investment funds when investors face multi-dimensional problems. It is possible for the investor to add weights to certain criteria, but not necessary, to get the most suitable investment fund.
2.3. COMPARISON BETWEEN REGRESSION ANALYSIS AND DEA
Both data envelopment analysis and regression analysis are methods that can be used to compare the performance of different producers (Thanassoulis, 1993). The differences in approach to solve an efficiency problem are discussed by Thanassoulis (1993).
Data envelopment analysis uses linear programming to define the efficiency of the producers - which are called decision making units (DMUs) - relative to an efficient producer. Regression analysis on the other hand uses linear programming to compare the performance of the producers with the average producer. Both methods can however cope with multiple inputs and multiple outputs to calculate the performance of the producers.
The major difference between the two methods is that data envelopment analysis estimates the efficient boundary, and therefore is a boundary method, and that regression analysis estimates the regression line that best fits the data, and therefore is a non-boundary method.
many separate sets of marginal costs. Regression analysis on the other hand only has one unique set of marginal costs. This difference is likely to give more accurate estimates when data envelopment analysis is used. The restriction is that data envelopment analysis is not allowed to give any zero marginal value to any producer; in that case the data is unreliable.
Target setting is the last comparison point made by Thanassoulis (1993). Target setting is the specification of input-output levels which the producers should achieve (Thanassoulis, 1993, p 1136). The result from Thanassoulis’ comparison is that the estimates made by data envelopment analysis are more accurate on the general level. Regression analysis produces more accurate estimates on an individual level, thus the efficiency of producer A relative to producer B. Added to this result is the fact that the estimates produced by data envelopment analysis are more volatile than those produced by regression analysis.
The result of the comparison is that data envelopment analysis is the preferable method over regression analysis on an individual maximum or minimum level.
On a last note Thanassoulis gives an overview of the advantages of the different methods over each other. The most important advantages of the different methods are mentioned here. The advantages of data envelopment analysis over regression analysis are that; it is a non-parametric method not requiring the user to hypothesize a mathematical form for the production function; it can cope more rapidly with multiple inputs and multiple outputs; it can identify sources of inefficiency in terms of excessive use of particular resources or low levels on certain outputs.
3. DATA AND METHODOLOGY
In this section the data and the methodology are described. First, data envelopment analysis is explained in detail and the general data envelopment analysis linear programming model is presented, following McMullen and Strong (1998). This is followed by an explanation of regression analysis. After that, the criteria that are used to describe and score the funds are explained. This section proceeds by giving the restrictions to the selection of the dataset. Also the process of information gathering is described, and the scores on the criteria and the description of the data are given.
3.1. METHODOLOGY
3.1.1. DATA ENVELOPMENT ANALYSIS
The data envelopment analysis method is developed by Charnes, Cooper and Rhodes and is presented in their 1978 article (Charnes et al, 1978). The method is used in order to find out how efficient producers are (Cooper et al, 2006). Data envelopment analysis is an extreme point method and compares each producer with an efficient producer. This efficient producer does not have to exist in the real world (it can be virtual), and is constructed from the existing producers. The efficient producer is constructed from all the information about the inputs and the outputs of all producers.
A single input and single output example clarifies this, see table 1 and figure 1 below (Cooper et al, 2006). Store E generates 4 units of output (sale) by using 5 units of input (employee). Then if the other stores operate efficiently they should be able to attain this same efficiency level of 0.8 (sale divided by employee). In addition, store B generates 3 units of output by using 3 units of input. The other stores should attain this same efficiency level, which is 1, for them to be efficient. When stores E, B and the rest of the stores are combined, the efficient composite store is formed. In this simple example, an existing store (B) is efficient. The result is an efficient frontier which shows all efficient stores and the associated combinations of inputs and outputs that are efficient.
Store A B C D E F G H
Employee 2 3 3 4 5 5 6 8
Sale 1 3 2 3 4 2 3 5
Sale/employee 0.5 1 0.667 0.75 0.8 0.4 0.5 0.625
Figure 1: Efficient frontier line (red) and regression line (blue)
The red line in figure 1 is the efficient frontier line and the blue line is the regression line. The efficient frontier line shows that only one store is on the efficient frontier, B, and therefore store B operates efficiently. All stores to the right of the efficient frontier line are inefficient. The slope of the line connecting each store to the origin measures the efficiency and it is the ratio of sales and employees. Store B has the highest slope and thus is the most efficient.
The difference between data envelopment analysis and regression analysis is explained by looking at the two lines in figure 1. The regression line goes through the middle of all data points and the points above the line are excellent and the points below the line are inferior. The degree of excellence or inferiority is measured by the magnitude of the deviation from the regression line. In contrast the efficiency of the stores when data envelopment analysis is used, is determined by deviations from the efficient frontier line.
The data envelopment analysis methodology is further explained by means of the general data envelopment analysis linear programming model, as presented in McMullen and Strong (1998). The three formulas in the frame below show this model. The explanation is not aimed at the producers in general, but at specific funds.
Max:
(
)
=+
=
s y yk yk yk kO
u
h
1µ
for fund k (1) Subject to:(
)
= = + r x xk xk xk v I 1 1ρ
for fund k (2) and0
1 1≤
+
−
= = xk r x xk xk s y yk ykv
I
u
O
µ
for fund k (3)This research looks at three output variables; transparency, tradability, and historic return of fund k, where k=1,2…n. The outputs are also called benefits, because they are beneficial for the investor. They are depicted by O1k, O2k, and O3k, and this means that s=3 outputs (equation 1,3). These outputs
have weights u1k,u2k, and u3k. In addition to three output variables three input variables are
investigated, that are the fees, the risk, and the leverage for fund k. The inputs are also called costs. They are depicted by I1k, I2k, and I3k and this means that r=3 inputs (equation 2,3). The weights of the
inputs are v1k, v2k, and v3k. The slack variables are µyk and yk. The values of these variables quantify
the inefficiency in fund k.
Equation (1) is the objective function and it shows that the objective is to maximize the efficiency of fund k, denoted by hk. The efficiency is maximized by the model by evaluating hk at
varying values of the weights uyk and vrk. The fund with the highest value of hk subject to the
constraints (2) and (3) is the most efficient. This feature of data envelopment analysis of presenting a single efficiency score makes it unique compared to other methods. The constraints (2) and (3) ensure that the efficiency of fund k does not exceed one.
Because multiple inputs and outputs are used, the efficient frontier is drawn with virtual inputs and outputs along the axes in a two dimensional graph. These virtual inputs and virtual outputs are the weighted sum of the inputs and the outputs. The formulas are as follows:
It is important to note that data envelopment analysis determines the optimal weights for the inputs and the outputs that result in the highest possible efficiency value. So the investor can state no preference for some inputs and/or outputs by setting specific weights for them. However, the investor can influence the outcome by setting constraints on the weights and thus force the weights in a specific direction. The Assurance Region Global (AGR) model can set upper and lower bounds for the ratio of the virtual weights of inputs and outputs against the total virtual weights. The upper and lower bounds that are used here are presented in the frames below. Equations (6a) to (6c) show upper and lower bounds for the inputs and equations (6d) to (6f) show upper and lower bounds for the outputs.
Fees: 0.2 0.4 3 3 2 2 1 1 1 1 ≤ + + ≤ x v x v x v x v (6a) Leverage: 0.2 0.4 3 3 2 2 1 1 2 2 ≤ + + ≤ x v x v x v x v (6b) Risk: 0.2 0.4 3 3 2 2 1 1 3 3 ≤ + + ≤ x v x v x v x v (6c) Historic return: 0.2 0.4 3 3 2 2 1 1 1 1 ≤ + + ≤ y u y u y u y u (6d) Tradability: 0.2 0.4 3 3 2 2 1 1 2 2 ≤ + + ≤ y u y u y u y u (6e) Transparency: 0.4 0.6 3 3 2 2 1 1 3 3 ≤ + + ≤ y u y u y u y u (6f)
It can be seen that the ratio of the virtual weight of each of the three inputs to the total virtual weight is the same for each input. The reason for this is that each input is equally important. The ratio of the virtual weight of each of the three outputs to the total virtual weight is not the same for each input. Transparency is more important than the other outputs and therefore transparency is assigned a greater ratio of virtual weight to total weight.
the number of efficient producers under the BCC model, it becomes clear that there are more efficient producers when the BCC model is used. The reason for this difference is that the BCC model measures the pure technical efficiency for a given scale of operation and the CCR model estimates the overall technical efficiency.
McMullen and Strong (1998) do not explicitly mention the DEA model they have chosen for the selection of mutual funds. But in the notes of the article they mention that the outcome of their analysis is an efficient frontier that has variable returns to scale as underlying assumption. From this it can be inferred that they use the BCC model. In contrast, Huyen (2006) uses the CCR model but he fails to give a reason for his choice. Huyen (2006) states the following (p. 16) “…when DMUs to be evaluated are not manufacturing ones, where the inputs and outputs are selection criteria and not physical elements, justifying for using CCR or BCC models is not easy.” Here the CCR variant of the AGR model is used, because constant returns to scale are assumed.
Data envelopment analysis has various strengths. It can handle multiple inputs and multiple outputs and it does not require an assumption of a functional form that relates the inputs to the outputs. Furthermore, producers are directly compared against each others and the inputs and the outputs can be measured in different units. Limitations of data envelopment analysis are for example that noise can cause problems because data envelopment analysis is an extreme point technique, and that data envelopment analysis is good at estimating the relative efficiency but poor at estimating the absolute efficiency. Other limitations are that statistical hypothesis tests are difficult to conduct because data envelopment analysis is a nonparametric method, and that large problems can be difficult to compute.
3.1.2. REGRESSION ANALYSIS
Regression analysis is used as an extra check to determine the performance of the various non-listed real estate investment funds.
In contrast to data envelopment analysis there is no efficient producer and therefore no efficient frontier. Instead regression analysis uses the average producer to determine the efficiency score of individual producers.
The regression line is defined as: u
x
y=
α
+β
+ (7)Where is defined as the horizontal interceptor and is the slope of the regression line. u is a random disturbance term explaining the difference between the individual points and the regression line.
In contrast to data envelopment analysis the weights assigned to the different criteria are not free. To calculate the input x, an equally weighted average of the three different criteria is taken. The weights assigned are therefore 0.33 for each of the three criteria.
The output y is calculated in a slightly different way. The output criterion transparency is more important than the other two output criteria and is given therefore a higher weight. The weights assigned to the different criteria are: tradability 0.3, return 0.3, and transparency 0.4.
3.1.3. CRITERIA
There are six different criteria by which the fund comparison will take place. According to the GUPF, these criteria are the most important for determining real estate fund allocation. The first criterion is the transparency of an investment fund and this criterion is very important for the GUPF. This is because the GUPF likes to know as much as possible about the funds they invest their money in or where they want to invest their money in. Due to this important topic for the pension fund, this is the most important criterion by which the different real estate funds are compared. Transparency, how it is defined by us, is the openness of the fund towards the investor. In order to measure the transparency, a checklist is used. This checklist is presented in appendix I. The scale by which transparency is measured, is a ratio. The ratio ranges in theory from 0/33 to 33/33 because the checklist contains 33 items. When for example only one of the checklist items is known, the transparency score is 1/33. This means that a fund is relatively closed and that an investor is not well informed about the development of the fund. A transparency score of 1 (33/33) means that a fund is distributing the relevant information to its investors so they are aware of the development of the fund.
the yearly performance. The scale to measure the fees is the percentage that is allocated by the firm as management costs and that has to be paid by the investor as compensation for running the fund.
The third criterion is the three year average historical performance of the fund subtracted by the three year average historical benchmark performance. This is an indicator of out- or underperformance. The benchmark that is used is the ROZ/IPD index. This is an index for the Dutch real estate market. Every real estate sector is covered by a sector index. These sector indices are the ROZ/IPD retail index, -office index, -industrial index, and -residential index. If a fund is invested in different sectors a weighted average of the sector indices returns is taken into account in order to compute the relative performance.
The risk of the fund is the fourth criterion and it is an indicator of non-recovery of initial investments or of not reaching the target return. The risk is usually measured as the volatility of past returns. Here the risk of the funds is measured by computing the standard deviation (volatility) of the difference between the historical fund returns and the historical returns on the corresponding ROZ/IPD index. The deviation from the normal method for volatility calculation has two reasons. The first reason is to eliminate the market volatility and only keep the true volatility of the fund. The second reason is to compare leveraged and unleveraged funds. If a fund is leveraged this usually means a higher expected return as well as a higher risk level when compared to the non-levered situation. Therefore, in the fund comparison also a volatility measure has to be taken into account.
Leverage is the fifth criterion that is used. Real estate is usually financed using debt and equity. Leverage is the amount of debt divided by the total value of the portfolio and it will be measured as such. If a fund is highly leveraged then the risk is also higher due to the larger amount of interest- and redemption costs that need to be paid.
3.2. DATA
In this section the data used to perform the comparison between the different non-listed real estate funds is presented. First the restrictions to the data set are mentioned. Afterwards the process of information gathering is described. At last the scores on the different criteria are mentioned and discussed.
3.2.1. RESTRICTIONS
The research focuses on non-listed real estate investment funds that are statutory founded in the Netherlands (Beleggingsplan, 2007) and have their investment portfolio in the Euro zone. The limitation of investing only in the Euro zone has been made to eliminate the exchange rate effect of different currencies on for example the return. The task of comparing the funds among each other becomes more easily due to the absence of these exchange rate changes.
Table 2 gives an overview of the different sources used and the number of funds per source. It can be seen that 6 different sources are used. These sources are screened by using the above mentioned limitations in order to obtain the right set of funds.
Source Number of funds
Vereniging Vastgoed Participanten (VVP) 100
Vereniging van Institutionele Beleggers in Vastgoed, Nederland (IVBN) 32 PropertyNL 29 INREV 24 www.workfresh.org 19 www.vastgoedsites.nl 45
Table 2: Overview of sources
3.2.2. PROCESS OF INFORMATION GATHERING
The data is gathered using three different sources. The first source is the annual report of the fund. In this report there is a large amount of information summarized. This information is used to fill in the checklist which is used to measure the transparency of a fund. Also information about the selection criteria that can be derived from the annual report is used to rate the corresponding criteria.
The second source is the prospectus of the fund. The prospectus is used as an information source for the same reasons as the annual report. By using the prospectus and the annual report the checklist is filled in, and the other criteria are scored.
The remaining source of information is an interview or a telephone call with a representative of the fund. Interviews are held with representatives of organizations that offer the most funds, namely Achmea, Altera Vastgoed and ING. They all have four funds that are included in this research. With this additional information source the missing information will be gathered to complete the scores on the criteria.
3.2.3. SCORES
All scores on the criteria in table 3 are the result of the study using the sources described in section 3.2.2. During the research it became clear that not every fund had such thing as a prospectus. The absence of a prospectus is the result of the Dutch law about non-listed funds. Because some funds are not under control of the Netherlands Authority for the Financial Markets (AFM), they are not obliged to have a prospectus.
The absence of a prospectus is counterbalanced by the presence of a similar document. In some cases this document is called an information memorandum and in other cases it is called a business profile. These documents roughly have the same content as a prospectus and are therefore justified to replace the prospectus.
In table 3 an overview is given of the different funds. This overview is given by the scores on the different criteria that are mentioned in section 3.1.3. All scores are presented up to four decimals to ensure that data envelopment analysis is able to make a distinction between the funds. However, a supplier can have more than one fund and therefore some scores can give the same result (fees, transparency and tradability).
Fees (%) Leverage (%) Historic Return (%) Risk (std dev) Transparency (score) Tradability (%)
Achmea Dutch Retail Fund 0.4400 7.0000 -0.2000 3.6651 0.5152 0 Achmea Dutch Industrial Fund 0.4400 0.0000 1.0000 1.6263 0.5455 0 Achmea Dutch Office Fund 0.4400 0.0000 5.3667 4.0451 0.5455 0 Achmea Dutch Residential
Fund
0.3800 0.0000 1.8333 2.6406 0.5152 0
Altera Vastgoed Bedrijfsruimten
0.3110 1.0100 -0.6700 1.8501 1.0000 20 Altera Vastgoed Kantoren 0.3110 4.1100 0.1667 2.7508 1.0000 20 Altera Vastgoed Woningen 0.3110 0.3900 0.3667 2.4079 1.0000 20 Altera Vastgoed Winkels 0.3110 0.6600 -1.1000 1.4056 1.0000 20 Vastgoed Fundament Fonds 2.7500 39.5600 16.0000 3.0470 0.4545 0 AZL Vastgoed Kantoren 0.6000 55.7500 -5.7667 1.1150 0.8788 10 AZL Vastgoed Winkels 0.5500 49.1100 -1.6333 2.7301 0.8788 10 AZL Vastgoed Woningen 0.5500 11.3500 0.7000 1.9079 0.8788 10
Eurindustrial 0.6000 40.9500 -1.9567 2.0757 0.6364 10
Shopping Parks 0.5099 73.4500 -4.1000 4.1000 0.3636 100
ING Dutch Office Fund 0.6000 25.4000 1.2667 0.7301 0.8182 100 ING Dutch Residential Fund 0.6000 9.1600 -0.4000 0.9338 0.7879 100 ING Dutch Retail Fund 0.6000 15.8700 2.7667 2.2109 0.8182 100 ING Nordic Property Fund 0.7500 50.0000 2.3000 2.3000 0.6667 100 Kantoren Fonds Nederland 6.8000 45.5100 -1.1667 2.1510 0.5455 0 Synvest Real Estate Fund 0.7000 60.4900 0.5605 8.1112 0.6061 100 Wilgenhaege Stedekroon 0.6000 56.9100 -10.9219 10.9334 0.5152 10 Average 0.9121 26.0324 0.2101 2.9875 0.7128 34.7619 Standard deviation 1.4409 25.0714 4.9134 2.3982 0.2061 42.8508 Median 0.5500 15.8700 0.1667 2.3000 0.6667 10 Minimum 0.3110 0.0000 -10.9219 0.7301 0.3636 0 Maximum 6.8000 73.4500 16.0000 10.9334 1.0000 100 Table 3: Scores on the criteria1
3.2.4. CRITERIA
The fees that are charged by the funds are on average 0.91% of the Gross Asset Value. Two funds (Vastgoed Fundament Fonds and Kantoren Fonds Nederland) have fees larger than 1%. The remaining funds all have fees smaller than or equal to 0.75%. The two high fee percentages are the total expense ratios of the funds. The reason that these ratios are taken here, and not the management fees, is that detailed information about the management fees is absent for these funds. The only expense figure that is known, is the total expense ratio. The total expense ratio also covers other costs made by the fund and therefore is higher than the management fees alone. It can be said that the lack of transparency in the construction of fees leads by two funds to higher fees and they are therefore penalized by a higher score on fees. When the two high fee funds are not taken into account there is still a big difference in management fees between the funds. The remaining fund with the highest fees is the ING Nordic Property fund with fees of 0.75% and the funds with the lowest fees are the Altera Vastgoed funds with fees of 0.33%.
The average leverage of all funds in the data set is about 26% with a standard deviation of 25. This means that there are large differences between the leverage of the funds. Three funds of Achmea have a leverage of zero and all funds of Altera Vastgoed have leverage close to zero. These funds have (almost) no debt to finance their real estate. On the other side of the scale is Shopping Parks that has almost 74% of its investments financed by debt. A distinction can be made between low leveraged funds (<15%) and high leveraged funds (>25%). Between the two groups there are no other funds present. This leads to the suggestion that a fund has a distinct no leverage policy or a high leverage policy.
The risk and return are on average 2.99% and 0.21%. The risk is larger than the average return that leads to a number of funds that underperformed during the last three years. The high average risk is largely due to the Wilgenhaege Stedekroon fund. The Wilgenhaege Stedekroon fund is a relatively new fund and has therefore only one return. In this situation the absolute value of the return is taken as the risk. Also the Synvest Real Estate Fund has a high average risk. These two high risk funds are responsible for the high average risk. When these two funds are not taken into account, the risk ranges from 0.73% to 4.05%. Return figures differ greatly between the funds. The Vastgoed Fundament Fonds has the greatest return (16.00%) and the Wilgenhaege Stedekroon fund has the worst performance (-10.92%).
4. RESULTS
In the next section the results from data envelopment analysis and regression analysis are presented. The results are discussed and interpreted but no recommendation on the basis of the results is given yet.
4.1. EFFICIENCY SCORES
In table 4 the efficiency scores of the producers are given. It can be seen that there are two suppliers with funds that are efficient or close to efficient. The two suppliers are ING and Altera Vastgoed. Both suppliers give the investor the possibility to invest in one of their four funds for Dutch real estate.
Rank Producers Score
1 ING Dutch Residential Fund 1
1 ING Dutch Office Fund 1
1 Altera Vastgoed Winkels 1
1 Altera Vastgoed Woningen 1
5 Altera Vastgoed Bedrijfsruimten 0.8488
6 Altera Vastgoed Kantoren 0.6670
7 ING Dutch Retail Fund 0.6597
8 ING Real Estate Nordic Property Fund 0.4984
9 AZL Vastgoed Woningen 0.3152
10 Synvest Real Estate Fund 0.2735
11 Shopping Parks 0.2608
12 AZL Vastgoed Kantoren 0.2484
13 AZL Vastgoed Winkels 0.2208
14 Eurindustrial 0.2107
15 Achmea Dutch Industrial Fund 0.0390 16 Achmea Dutch Residential Fund 0.0390
17 Achmea Dutch Office Fund 0.0390
18 Wilgenhaege Stedekroon 0.0001
19 Achmea Dutch Retail Fund 0.0001
20 Kantoren Fonds Nederland 0.0000
21 Vastgoed Fundament Fonds 0.0000
It can be seen that two of the four efficient funds are residential funds. Together with table 3 it can be derived that the efficiency of the residential funds is largely due to the low leverage of the residential funds. Scores on other criteria of the residential funds do not deviate much from funds in other sectors. The efficient funds are distributed over three of the four sectors. Next to the two residential funds there is an office fund and a retail fund present in the group of efficient funds. All four efficient funds are real estate funds that are invested in the Netherlands.
4.2.1. OUTPUT CRITERIA
In table 5 an overview is given of the weights that are assigned to the different output criteria.
Producer Score Transparency Tradability Return
Achmea Dutch Retail Fund 0.0001 0.4 0.2 0.4
Achmea Dutch Industrial Fund 0.0390 0.4 0.2 0.4
Achmea Dutch Office Fund 0.0390 0.4 0.2 0.4
Achmea Dutch Residential Fund 0.0390 0.4 0.2 0.4
Altera Vastgoed Bedrijfsruimten 0.8488 0.4 0.2 0.4
Altera Vastgoed Kantoren 0.6670 0.4 0.2 0.4
Altera Vastgoed Woningen 1 0.6 0.2 0.2
Altera Vastgoed Winkels 1 0.6 0.2 0.2
Vastgoed Fundament Fonds 0.0001 0.4 0.2 0.4
AZL Vastgoed Kantoren 0.2484 0.6 0.2 0.2
AZL Vastgoed Winkels 0.2208 0.6 0.2 0.2
AZL Vastgoed Woningen 0.3152 0.6 0.2 0.2
Eurindustrial 0.2107 0.4 0.2 0.4
Shopping Parks 0.2608 0.4 0.4 0.2
ING Dutch Office Fund 1 0.6 0.2 0.2
ING Dutch Residential Fund 1 0.4462 0.3537 0.2
ING Dutch Retail Fund 0.6597 0.4 0.4 0.2
ING Real Estate Nordic Property Fund 0.4984 0.4 0.4 0.2
Kantoren Fonds Nederland 0.0000 0.4 0.2 0.4
Synvest Real Estate Fund 0.2735 0.4 0.4 0.2
Wilgenhaege Stedekroon 0.0001 0.6 0.2 0.2
Table 5: Weights assigned to output criteria
As can be seen in table 5 all weights are between 0.2 and 0.6. The weights given to all criteria are restricted within certain boundaries. For the criterion transparency this restriction is a weight between 0.4 and 0.6. For the other two criteria the restriction is a weight between 0.2 and 0.4.
4.2.2. INPUT CRITERIA
The next section is interesting for the readers who would like to know the more technical information behind data envelopment analysis. The readers who are not interested in the technical story can skip this section and proceed to the recommendations in section 5.
In table 6 the weights assigned to the input criteria are shown together with the efficiency scores of the corresponding funds.
Producer Score Fees Leverage Risk
Achmea Dutch Retail Fund 0.0001 71439.15 71439.15 35719.57 Achmea Dutch Industrial Fund 0.0390 5.1259 10.2519 10.2519 Achmea Dutch Office Fund 0.0390 10.2540 10.2540 5.1270 Achmea Dutch Residential Fund 0.0390 5.1266 10.2532 10.2532 Altera Vastgoed Bedrijfsruimten 0,8488 0.4712 0.2356 0.4712 Altera Vastgoed Kantoren 0,6670 0.5997 0.2999 0.5997
Altera Vastgoed Woningen 1 0.4 0.4 0.2
Altera Vastgoed Winkels 1 0.3286 0.4 0.2714
Vastgoed Fundament Fonds 0.0001 154535.4 309070.8 309070.8 AZL Vastgoed Kantoren 0.2484 1.6103 0.8052 1.6103
AZL Vastgoed Winkels 0.2208 1.8113 0.9057 1.8113
AZL Vastgoed Woningen 0.3152 1.2690 1.2690 0.6345
Eurindustrial 0.2107 1.8984 0.9492 1.8984
Shopping Parks 0.2608 1.5335 0.7667 1.5335
ING Dutch Office Fund 1 0.4 0.2254 0.3746
ING Dutch Residential Fund 1 0.4 0.2 0.4
ING Dutch Retail Fund 0.6597 0.6063 0.3032 0.6063 ING Real Estate Nordic Property Fund 0.4984 0.8026 0.4013 0.8026 Kantoren Fonds Nederland 0.0000 147165.5 294330.9 294330.9 Synvest Real Estate Fund 0.2735 1.4623 0.7312 1.4623 Wilgenhaege Stedekroon 0.0001 30430.14 15215.07 30430.14 Table 6: Weights assigned to input criteria
The boundaries are the same for all three input criteria and are set between 0.2 and 0.4. It should therefore not be possible for an input criterion to have a weight that is more than two times that of the other criteria.
If a closer look is taken at table 6 it becomes clear that most of the weights are not within the boundaries set before. One fund even has weights assigned to two of its criteria larger than 300000. This asks for an explanation.
Because the inputs are used to generate the desired outputs the different funds are compared among each other. If fund A can generate the same output as fund B but with lower inputs than B a higher weight is assigned to the input of fund B to generate the same efficiency level as fund A. This is exactly what happened in the analysis of the funds. Due to high input scores or low output scores most funds are not performing efficiently.
If criteria have weights larger than 0.4 or the sum of the weights is larger than 1.0 this will result in inefficiency. It can be seen that the funds that are efficient have their input weights in line with the boundaries that are set and have a total weight of the input criteria of 1.0.
4.3. REGRESSION ANALYSIS RESULTS
The results of regression analysis are presented in table 7, figure 2, and table 8. Table 7 shows the weighted input, weighted output and regression output for each producer. As mentioned in the methodology section, the weighted input and weighted output are the weighted average of the
individual inputs and the individual outputs respectively. By the regression output is meant the output as calculated by the formula for the regression line.
Producer Weighted Input (x) Weighted Output (y) Regression Output (f) Achmea Dutch Retail Fund 3.7017 3.4227 3.5462 Achmea Dutch Industrial Fund 0.6888 3.7948 0.6599 Achmea Dutch Office Fund 1.4951 5.1048 1.4323 Achmea Dutch Residential Fund 1.0069 4.0327 0.9646 Altera Vastgoed Bedrijfsruimten 1.0570 9.4756 1.0126 Altera Vastgoed Kantoren 2.3906 9.7266 2.2902 Altera Vastgoed Woningen 1.0363 9.7866 0.9928
Altera Vastgoed Winkels 0.7922 9.3466 0.7589
Vastgoed Fundament Fonds 15.1190 8.2584 14.4840
AZL Vastgoed Kantoren 19.1550 4.8981 18.3505
AZL Vastgoed Winkels 17.4634 6.1381 16.7299
AZL Vastgoed Woningen 4.6026 6.8381 4.4093
Eurindustrial 14.5419 5.9441 13.9312
Shopping Parks 26.0200 32.1921 24.9271
ING Dutch Office Fund 8.9100 33.9839 8.5358
ING Dutch Residential Fund 3.5646 33.4718 3.4149
ING Dutch Retail Fund 6.2270 34.4339 5.9654
Figure 2 shows the data points of each producer and the regression line. The formula for the regression line is:
x
y=0.958 (8)
Where y is the weighted output and x is the weighted input.
y = 0,958x 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 Weighted input Weighted output
The efficiency of the funds according to regression analysis is calculated by dividing the weighted output of the funds by the weighted output as calculated by the regression formula.
The scores on regression efficiency are presented in table 8 and they show that there are four funds with a regression efficiency of at least 9. These funds are; Altera Vastgoed Winkels, Altera Vastgoed Woningen, ING Dutch Residential Fund, and Altera Vastgoed Bedrijfsruimten (in decreasing order of efficiency).
Regression efficiency
Altera Vastgoed Winkels 12,3153
Altera Vastgoed Woningen 9,8578
ING Dutch Residential Fund 9,8017
Altera Vastgoed Bedrijfsruimten 9,3572
ING Dutch Retail Fund 5,7722
Achmea Dutch Industrial Property Fund 5,7507
Altera Vastgoed Kantoren 4,2471
Achmea Dutch Residential Fund 4,1805
ING Dutch Office Fund 3,9813
Achmea Dutch Office Fund 3,5641
ING Real Estate Nordic Property Fund 2,0208
AZL Vastgoed Woningen 1,5508
Synvest Real Estate Fund 1,5222
Shopping Parks 1,2914
Achmea Dutch Retail Property Fund 0,9652
Vastgoed Fundament Fonds 0,5702
Eurindustrial 0,4267
AZL Vastgoed Winkels 0,3669
AZL Vastgoed Kantoren 0,2669
Kantoren Fonds Nederland 0,1808
Wilgenhaege Stedekroon 0,1467
5. RECOMMENDATIONS AND CONCLUSION
From a diversification point of view it is sensible that the GUPF invests in listed real estate as well as in non-listed real estate. This allocation leads to a good exposure to the different sorts of risk and opportunities.
5.1. RECOMMENDATIONS
To fully profit from the different opportunities that are present the GUPF should invest in all the different real estate sectors. On the hand of the criteria that are used during data envelopment analysis the GUPF should allocate the investment in non-listed real estate to ING or Altera Vastgoed.
One of the options is to invest the full allocation in non-listed real estate into one supplier. In the case the GUPF prefers one supplier over multiple Altera Vastgoed is the one that should be chosen. In that case two of the efficient funds are present in the portfolio of the GUPF and two semi-efficient funds are present. The semi-semi-efficient funds have the highest efficiency score of the funds that are not efficient and are therefore preferred over the ING funds that are not efficient.
A second possibility for the GUPF is to select the sector funds that are the most efficient and add them together in a portfolio. In case the most efficient funds are chosen a choice has to be made in the sector residential, because in that sector two funds are efficient. Both funds can be added in for example equal weights or one of the two can be chosen on the basis of one of the criteria. Because the criterion transparency is the most important criterion for the GUPF the choice for Altera Vastgoed Woningen should be made over ING Dutch Residential.
Rank Producers Score
1 ING Dutch Residential Fund 1
1 ING Dutch Office Fund 1
1 Altera Vastgoed Winkels 1
1 Altera Vastgoed Woningen 1
5 Altera Vastgoed Bedrijfsruimten 0.8488
6 Altera Vastgoed Kantoren 0.6670
7 ING Dutch Retail Fund 0.6597
8 ING Real Estate Nordic Property Fund 0.4984 Table 9: Top 8 efficient funds
If the outcome of data envelopment analysis is compared with the outcome of regression analysis it becomes clear that the efficiency rating of regression analysis is different from the rating of data envelopment analysis.
Although the ranking is not entirely the same most of the efficient funds of data envelopment analysis are present in the top of regression analysis. The top three efficient funds of regression analysis are also efficient funds in data envelopment analysis.
If the GUPF wants to invest in all different sectors the best performer of each sector is picked from the list. This results in a recommendation of investment in all funds of Altera Vastgoed, according to regression analysis. The funds of Altera Vastgoed are the best performers in all sectors and should therefore be preferred over the other suppliers of non-listed real estate.
The options presented above lead to different portfolios with different amounts of efficient funds. The last option of data envelopment analysis, represented in table 9, is the option with the most efficient funds present in the portfolio and should therefore be preferred over the other options.
5.2. CONCLUSION
If the recommendations are compared with the present non-listed real estate portfolio of the GUPF it can be concluded that some of the efficient funds are already present in the portfolio. The funds already present are those of Altera Vastgoed.
To optimize the portfolio the remaining sector, industrial, should be added to realize a better allocation over the different sectors. The fund that should be added is Altera Vastgoed Bedrijfsruimten.
REFERENCE LIST
Brooks C. (2005), Introductory econometrics for finance, Cambridge, New York.
Centraal Bureau voor de Statistiek (2005), Institutionele beleggers doen meer in vastgoed, Webmagzine CBS, 28 februari 2005.
Charnes A., Cooper W. and Rhodes E. (1978), Measuring the efficiency of decision-making units, European Journal of Operational Research, Vol 2, No. 6, November 1978, pp. 429-444.
Cooper W., Seiford L. and Tone K. (2006), Introduction to Data Envelopment Analysis and Its Uses, Springer, New York.
Huyen N.T.T. (2006), On the use of data envelopment analysis in hedge fund selection, working paper, University of New Orleans.
Lymos (2006), Implementatie van een vastgoedbeleggingsstrategie in niet-beursgenoteerde vastgoedfondsen, notitie, November/December 2006.
McMullen P. R. and Strong, R. A. (1998), Selection of Mutual funds using data envelopment analysis, Journal of Business and Economic Studies, Vol. 4, No. 1, Spring 1998 .
Nederlands Pensioen- en Belegginsnieuws (2006), Beursgenoteerd vastgoed in the slop, NPN Magazine, June/July 2006.
Raitio V. (2006), Fee structures in non-listed real estate funds for institutional investors, VBA Journaal, No.4, Winter 2006, pp. 40-44.
Stichting Gasunie Pensioefonds, Memorandum beleidsplan beleggingen, 2007, p11.
Thanassoulis E. (1993), A Comparison of Regression Analysis and Data Envelopment Analysis as Alternative Methods for Performance Assessments, The Journal of Operational Research Society, Vol. 44, No. 11, November 1993, pp. 1129-1144.
WEBSITES
Vereniging Vastgoed Participanten,
http://www.vvp.nu/Overzicht%20aanbieders/Profielen%20aanbieders.html, 16-04-2007. A DEA homepage, http://www.etm.pdx.edu/dea/homedea.html, 16-04-2007.
Vereniging Vastgoed Participanten, http://www.vvp.nu/home, 16-04-2007. Vereniging van Institutionele Beleggers in Nederland,
http://www.ivbn.nl/webgen.aspx?WebsiteID=1&PageID=0&FramesetID=1, 18-04-2007. Vastgoedsites.nl, http://www.vastgoedsites.nl/, 18-04-2007.
PropertyNL, http://www.propertynl.com/propertynl/propertynl/, 18-04-2007. European Association for Investors in Non-Listed Real Estate Vehicles,
http://www.inrev.org/default_frames.aspx?myurl=http://www.inrev.org/content/content.aspx?ctype=ur l&id=8&cname=ucPressReleasedetail&newsid=14, 23-04-2007.
Workfresh, http://www.workfresh.org/index.php?option=content&task=view&id=103&Itemid=76, 23-04-2007.
Nederlands Pensioen en Belegginsnieuws,
APPENDIX I
Checklist Transparency
o Name fund manager
o Name of staff behind fund manager o Track record fund manager
o Working experience of fund manager in real estate o Allocation over sectors
o Availability of exact location details o Dispositions
o Previous returns of fund
o Direct return (in comparison with benchmark) o Indirect return (in comparison with benchmark) o Total return
o Risk of the fund
o Leverage of the fund (in percentage or true value) o Number of shareholders
o Total value of each shareholder invested o Total value of assets
o Number of reports per year
o Change in position by participants (share transactions) o Availability of board of directors
o Participants in board of directors o Times the portfolio is revaluated o Information about construction of fees
o Does fund have to buy position of participant if he wants to get out o Locations under consideration
o Degree of occupation o Profit and loss account o Balance sheet
o Cash flow statement o Dividend payments o Organization costs
o Relevant macro economic environment o Tenants
