The value of regional investment strategies: the region of Groningen case

(1)

The value of regional investment strategies:

the region of Groningen case

Mireille Bombeld

Student number: 1529714

Supervisors:

dr. A. (Auke) Plantinga

prof. dr. L.J.R. (Bert) Scholtens

Groningen, May 27, 2011

University of Groningen

Faculty of Economics and Business

Master Thesis, MSc Business Administration

(2)

2

The value of regional investment strategies:

the region of Groningen case

Abstract

Home biased investors currently have the opportunity to invest in index funds. Such funds do not always include companies that have their economic gravity center in the region. Since familiarity is one of the explanations for the home bias, I created portfolios based on companies that are situated in the Dutch region of Groningen. The performance of these portfolios are tested against appropriate benchmarks using Jensen‟s alpha, the Fama and French three factor model, and the Sharpe ratio. I find that the portfolios significantly outperform their benchmarks and conclude that home biased portfolios based on locally situated companies can be a sensible investment strategy.

JEL Codes: G11, D03

(3)

3

Currently, there is overwhelming evidence of the existence of the home bias and a lot of explanations are offered, all failing to provide the one overriding explanation. However, I argue that even though it is labeled as a bias, investing in a regional portfolio might be a sensible strategy. First, investors may have a strong preference for local investments. Second, there is also evidence that indeed individual investors and professional money managers are able to earn higher returns from their local investments (Coval and Moskowitz (1999)), (Ivković and Weisbenner (2005)). Currently, investors who wish to hold regional portfolios have the opportunity to buy index funds. However, these funds include locally quoted companies which does not necessarily mean that these are also the companies that have an economic gravity center in the region. This economic gravity center is what fosters familiarity and therefore such index funds do not completely satisfy the need of regional investors. Instead, regional investment products should ideally be based on locally quoted companies. This leads me to my research question:

How does a regional investment portfolio based on locally situated companies perform, compared to regional investment products based on locally quoted companies?

(6)

6

a lot of companies that have only marginal portfolio weights. Equal weighting is used as an alternative strategy. Finally, fundamental indexation is used determine the weights for a third home biased portfolio. Fundamental indexation extends the employee based weighting scheme by using multiple fundamentals to determine portfolio weights.

After the construction of these three home biased portfolios, I test their performance using Jensen‟s alpha to determine if the strategy is able to generate significant risk-adjusted excess returns over the benchmarks. Furthermore, I use the Fama and French three factor test to test if the results found using Jensen‟s alpha by including two additional risk factors. Hereby it is tested whether the results found with Jensen‟s alpha are due to risks that are not captured within that model. Finally, the Sharpe ratio is used to test for situations where diversifiable risk is present in portfolios. The benchmarks chosen are funds focused on the Netherlands that are combined into a portfolio of funds – since those are the substitute products for investors who wish to hold home biased portfolios – and the AEX index. The test is performed for the period January 1, 1996 until December 31, 2009.

Regressions using Jensen‟s alpha revealed both positive and significant outperformance of the home biased portfolios based on locally situated companies with respect to the benchmarks based on locally quoted companies. Also, in the situation where additional risk factors are introduced – as is the case with the Fama and French three factor model – positive and significant outperformance is found. Finally, this result also holds in the presence of unsystematic risk, given the positive results from the analysis of the Sharpe ratio. Thus, I find evidence to confirm that such a strategy might be valuable and rational.

(7)

7

(8)

8

I. Literature Review

A well documented and widely accepted theory in finance is the possible reduction of risk in a portfolio that stems from international diversification. International diversification can produce these significant reductions in the systematic risk of a portfolio when the returns in different equity markets are not perfectly correlated. Even though this is commonly recognized, French and Poterba (1991), amongst others, find that investors hold a disproportionally large share of their portfolio in domestic equities. This is referred to as the home equity bias. Especially when you look at the market size it appears that holdings in foreign equities are lower than would be expected. French and Poterba (1991) also show that this deviation from the market portfolio is quite costly. They use actual portfolio holdings to find implied return expectations and use these to show that investors expect the return on the domestic equities to be several hundred basis points above the returns in foreign markets. French and Poterba (1991) in addition show that the expected returns from domestic equities are systematically higher than the implied returns from a value-weighted strategy.

(9)

9

like to make about international diversification is based on Cai and Warnock (2005) who show that holding large domestic multinationals can lead to substantial international diversification, where the influence of foreign factors increases with the extent of the firm‟s foreign sales. This type of investment can thus be seen as home-made diversification when these type of companies are included in domestic portfolios.

(10)

10

relative risk aversion greater than one and hence, it is unlikely that the demand for hedging domestic inflation can explain the home equity bias. Michaelides (2002) takes a more macroeconomic approach to the home bias and uses an analytical framework for the optimal portfolio choice in the presence of both liquidity constraints and labor income risk that cannot be diversified away. He shows that either small trading costs for international investments or a slightly higher return on domestic stocks are sufficient to generate a home bias. This is because consumption fluctuations can be smoothed with a small amount of buffer stock, which limits the benefits of international diversification, and exchange rate risk makes international diversification less attractive to risk averse investors (Michaelides (2002)).

(11)

11

based on familiarity, the portfolio returns for these local returns will on average not earn superior returns relative to their nonlocal investments, whereas they will earn superior returns in case of an informational advantage. The evidence points to the latter case: the performance of the individual investors‟ local investments were superior to their nonlocal investments. This robust result is particularly strong in case of less widely known stocks, i.e. non-S&P 500 companies (Ivković and Weisbenner (2005)).

The main criticism on the home bias found on fund level is that mutual funds have both the resources and skilled analysts available to analyze information, which makes the assumption of asymmetric information less appealing at this level. Dziuda and Mondria (2009) show that if the individual investors have to judge portfolio performance – and thus the performance of the fund manager – a rationale for a local bias rises again. Namely, these individual investors have more knowledge about the performance of local markets and businesses and thus are better able to evaluate fund managers operating in these markets. This makes it more attractive for fund managers to operate in local markets, as it creates less risk for their skill evaluations (Dziuda and Mondria (2009)). The second class of explanations which is concerned with investor behavior, is also recognized by other authors. Nofsinger (2008) explains that investors invest in companies with which they are familiar and that investors find more comfort from investing in a business that is visible to him. Thus, if investors invest in foreign stocks, they also choose foreign firms that are visible, meaning large firms that trade products that are recognizable to the investor. This preference holds for both individual investors and investment managers (Nofsinger (2008)). Another explanation that fits in this category is that “the statistical uncertainties associated with estimating

expected returns in equity markets makes it difficult for investors to learn that expected returns in domestic markets are not systematically higher than those abroad” (French and

Poterba (1991), p.225).

(12)

12

strategy that should be avoided. Hasan and Simaan (2000) show that under certain conditions investing in a home equity biased portfolio can be the rational choice and can be consistent with rational mean-variance portfolio choice. First, it has to be acknowledged that under full information, the optimal choice would obviously be international diversification. However, if the assumption of full information does not hold, there is a powerful incentive to deviate from international diversification. The costs of this divergence is dependent on the distance of the home market to the mean-variance efficient frontier. The benefit comes from a reduction in the estimation errors. In particular: “when the cross-market variability in the

estimation errors of international markets far exceeds the cross-market variability in the means themselves, domestic dedication dominates international diversification” (Hasan and

Simaan (2000), p.331). In addition, Hasan and Simaan (2000) show that, as the number of stocks in a portfolio increases, the marginal advantage of including an additional stock to the portfolio can be outweighed by the cost of estimation risk that is present in determining both the mean vector and the covariance matrix of asset returns.

(13)

13

(14)

14

II. Methodology

In this section, the difference between regional investment products based on locally quoted companies or based on locally situated companies is explained first. Second, Jensen‟s Alpha, the Fama and French three factor model and the Sharpe ratio are introduced as measures for portfolio performance. Finally, the portfolio weighting schemes and the rebalancing frequency are addressed.

A. Regional Investment Products

An individual investor that wishes to hold a regional portfolio currently has the opportunity to choose to invest in either the local stock market index or country portfolio‟s offered by banks. These funds, sometimes called index huggers, tend to replicate the benchmark portfolio, in this case the AEX index. The high value of R-squared for these funds suggests that the fund‟s performance is in line with the benchmark index. If this relationship holds, investors might prefer to invest in the index itself. This is because the index has a lower portfolio turnover and lower expenses1_{. An individual investor in the Netherlands, who}

wishes to hold a regional portfolio can either trade the AEX index or buy several Netherlands

funds, such as the ING AEX Shadow2_{. Table 1 gives an overview of AEX index related funds}

in the Netherlands. This table shows that an investor has a number of funds available if he or she wishes to hold a regional portfolio. The selection of these funds is done using the websites Fondsnieuws.nl and Morningstar.nl, two Dutch websites dedicated to mutual funds. Furthermore, the AEX index is included as well as a portfolio of funds. This portfolio of funds contains all the other funds described in the table and uses equal weighting to determine portfolio weights. This portfolio is thus the most interesting benchmark, since its performance can be seen as the performance of the average home biased investor who invests in locally quoted investment products.

1_{See: http://www.investopedia.com/terms/i/indexhugger.asp}

(15)

15

Table 1: Overview of index funds focused on The Netherlands This table lists the benchmarks that are used to test the value of the home biased portfolios. These funds are all focused on The Netherlands. The last column shows the period in which (return data for) these funds are/ were available. The last column shows the three year R2_{for each fund compared to the}

AEX index, as available from the Morningstar website.

Issuing party Official name of the fund Period (mm/yyyy) R2

AEX index 01/1996 – 12/2009

Portfolio of funds 01/1996 – 12/2009 Achmea Achmea Nederland Aandelenfonds 09/2000 – 06/2008 96.22 Add Value Add Value Fund N.V. 02/2007 – 12/2009 71.95 AEGON AEGON Equity Holland Fund 03/2004 – 12/2009 93.01 AEGON Index Plus Fonds 03/2004 – 12/2009 99.35 AXA AXA Aandelen Nederland - EUR 01/1996 – 11/2006 98.32 BNP Paribas BNP Paribas Netherlands Fund 02/2000 – 12/2009 94.77 BNP Paribas AEX Index Fund 01/1996 – 12/2009 95.87 BNP Paribas Small Companies NL 01/1996 – 12/2009 88.50 Delta Lloyd Delta Lloyd Nederland Fonds 05/2001 – 12/2009 73.98 Fortis Fortis ASR Fonds Nederlandfonds Acc 08/1997 – 04/2004 99.10 ING Investment Management ING AEX Shadow 08/1997 – 12/2009 95.77 ING Dutch Fund 01/1996 – 12/2009 96.07 Kempen Capital Management Kempen Orange Fund 01/1996 – 12/2009 78.43 Robeco Robeco Hollands Bezit 01/1996 – 12/2009 94.76 SNS Bank SNS Nederlands Aandelenfonds 01/1996 – 12/2009 95.73 Source: Fondsniews.nl & Morningstar.nl3

However, I argue that this is only one type of regional investing, namely investing in

locally quoted companies. The other type of regional investing is investing in locally situated companies. This point is best illustrated by table A1 in the appendix. This table shows the

comprehensive list of companies/ equities that are part of a portfolio based on the Dutch province of Groningen. This list is created by means of including all listed companies that have one or more business establishments in the province of Groningen. This also includes companies that are full subsidiaries of listed companies. A further elaboration on the

3_{See: http://www.morningstar.nl/nl/fundquickrank/default.aspx, and}

(16)

16

selection process is given in the data section. As can be seen, this list in Table A1 is more extensive than the AEX index. This list of all Groningen based listed companies amounts up to 121 companies over the entire period. In contrast, the AEX index only comprises the 25 most active securities in the Netherlands. The AEX index is believed to provide a fair representation of the Dutch economy4_{. However, as explained in the literature section, an}

investor that holds a portfolio consistent with the home bias does so because of - among other factors - the familiarity of the investor with the companies in such a portfolio. In addition, I argue that familiarity does not solely come from the company being amongst the most actively traded securities in a country, but also from direct links. Of course, these 25 actively traded securities belong to companies that are often quite large and well embedded in the Dutch economy and therefore make the news on a regular basis, increasing familiarity. Notwithstanding this fact, familiarity also comes from word of mouth and personal involvement. Therefore, the physical presence and the number of employees a company employs will also influence the perceived knowledge of the investor about a company. At this point, the following research question arises:

How does a regional investment portfolio based on locally situated companies perform, compared to regional investment products based on locally quoted companies?

B. Portfolio Performance

To answer the research question, the performance of the regional portfolio based on locally situated companies is compared to the performance of investment products based on regionally quoted companies. I use three measures of performance: Jensen‟s alpha, the Fama and French three factor model and the Sharpe ratio.

(17)

17

B.1 Jensen’s alpha

Jensen‟s alpha is calculated using an ordinary least squares time series regression which is analytically written as:

jt ft bt j j ft jt

R











(



)





(1)

Where

R

_jtis the return on the portfolio j at time t,

R

_ftis the return on a risk-free proxy, which is set to Euribor,

R

_bt is the return on benchmark indices (i.e. the funds listed in table 2),



_jtis the error term and the parameters



_jand



_jare the parameters to be estimated (Brooks (2008)). The significance of



_jdetermines if the fund earns significant abnormal returns in excess of the required return for the fund given its riskiness. Therefore, the null hypothesis is given by:

H

₀

:



_j



0

. Where a positive and significant result would mean that the portfolio has outperformed the benchmark.

B.2 Fama and French three factor model

In addition to Jensen‟s alpha, the Fama and French three factor model is used to test outperformance of the locally situated portfolios versus the locally quoted benchmarks. This model is developed to determine if the outperformance or underperformance found with Jensen‟s alpha is due to risks that are not captured by beta (Fama and French (1993)). This can be written analytically:

(18)

18

0 :

0 j



H



. Where a positive and significant result would mean that the portfolio has outperformed the benchmark.

B.3 Sharpe ratio

Both Jensen‟s alpha and the Fama and French three factor model are perfectly fine models if unsystematic risk can be diversified away. However, investors not always invest in multiple funds or investment products. This means that unsystematic risk might still be present. Therefore, the Sharpe ratio is used to determine the excess return over total risk incurred by the investor. The Sharpe ratio is calculated as:

x f x

E

r

E







(

)



(

)

(3)

Where

E

(

r

_x

)

is the expected return of the portfolio,

E

(

r

_f

)

is the risk free rate and



_x is the standard deviation of the portfolio (Benninga (2008)). The risk free rate is again set equal to Euribor. The Sharpe ratio measures the unit of excess return per unit of risk in the portfolio. Thus, a higher value for the Sharpe ratio is to be preferred over a lower value for the Sharpe ratio.

B.4 Transaction costs

This will be calculated twice: once gross of transaction costs, and once net of transaction costs. The transaction costs are calculated – based on Wermers (2000) – using the following approximation:

jt jt

jt

V

TO

C



0 .

002 



0 .

01 

(4)

(19)

19

Finally, note that returns used are simple returns. This is because they have the advantage that the simple return on a portfolio is the weighted average of the simple returns on the individual assets. The use of simple returns is possible since the frequency of the sampling of the data is already weekly, thus overcoming the general problem considered with simple returns: the fact that they are not time-additive (Brooks (2008)).

C. The Determination of Portfolio Weights

Three strategies are followed to determine the portfolio weights for the regional portfolio based on locally situated companies: weighting based on number of employees, weighting based on equal weights, and weighting using fundamental indexation. The rationale for these weighting schemes is explained below, followed by the analytical representation of the weighting schemes.

C.1 Weighting schemes rationale

(20)

20

Table 2: Portfolio weights for 2009 based on number of employees

This table shows the top 20 companies according to their size in the portfolio, when portfolio weights are calculated based on the number of employees. As can be seen, this top 20 accounts for 68.15% of the total portfolio. Companies that have an asterisk behind their names, are also included in the AEX index during 2009.

Company name Portfolio weight (%) Cumulative portfolio weight (%)

AHOLD * 11.31 11.31 TNT * _9.76 _21.07 Schuitema 4.95 26.02 KPN * 4.87 30.89 Atos _4.42 _35.31 Arriva 3.99 39.30 SDB 3.47 42.76 RANDSTAD * 3.16 45.93 Smurfit Kappa _2.64 _48.56 AKZO * 2.54 51.10 PPG _2.36 _53.47 McDonalds 2.10 55.56 Koninklijke BAM * 2.05 57.62 GRONTMIJ _1.87 _59.49 USG People 1.82 61.31

British American Tobacco _1.80 _63.11

G4S 1.71 64.82

ORDINA _1.67 _66.49

AEGON * 1.62 68.11

SCA 1.49 69.59

(21)

21

Moore (2005). They suggest that a portfolio weights should be determined using fundamental values of a firm as proxies for its size.

C.2 Portfolio weights based on number of employees

The portfolio weights are calculated by determining the number of employees a company has in the region of Groningen relative to the total number of employees that the entire group of selected companies employ in the region of Groningen. Analytically:



  _n j jt it it E E W 1 # # (5)

Where

#

E

is the number of employees in the region of Groningen, and

W

_it is the portfolio

weight for company

i

on time

t

.

C.3 Equally weighted portfolio weights

The portfolio weights are determined as the proportion of 1 divided by the total number of companies in the sample. This is thus equal for each company:

jt jt

N

W



1

(6)

Where

W

_jtis the portfolio weight for all selected companies at time

t

.

C.4 Portfolio weights based on fundamental indexation

(22)

22

The weights of the stocks in the fundamental indexes are calculated as follows:



  _n j jt it it FV FV W 1 (7)

The composite index is then calculated as the simple average of the weights for the individual fundamental indexes: 5 , , , , ,t SLt TAt POt NEt FC it W W W W W W      (8)

C.5 Rebalancing

(23)

23

III. Data and Descriptive Statistics

I obtained data on employment for the region of Groningen from the LISA database. This database contains information on employment for all business establishments in the Netherlands. Due to its ability to select data both on basis of different areas and on social-economic components, this database is used to obtain a list of all business establishments located in the region of Groningen. Since this database is complete and contains all companies, it is survivorship bias free. This list is then used to check for each company if it is listed on the stock exchange or if it is a full subsidiary of a listed corporation. If so, the company is included in the portfolio under its own name or the name of its respective parent company. In addition, the LexisNexis® database was used to check the exact dates of mergers, acquisitions, IPO‟s and privatizations. A company is then included in the portfolio for a given year if it listed at the beginning of the year (since rebalancing is done yearly). Total return data are retrieved from the Thomson Reuters Datastream database and cover the period from January 1996 – December 2009. In addition, data on the fundamentals used for the fundamental indexation approach is retrieved from the Thomson Reuters Worldscope database and also cover the period from January 1996 – December 2009. Finally, weekly data for the Fama and French factors are retrieved from the publicly available data library on the website of Kenneth R. French.

I use weekly returns because it overcomes the problem of non-normality and heteroscedasticity associated with daily data, while it does not necessarily result in loss of information. Monthly and yearly returns would destroy too much information, due to the limited number of observations that would be available when using this approach (Brooks (2008)).

(24)

24

used by Arnott, Hsu and Moore (2005): when for a certain fundamental fewer data was available than for all years needed, I used the average value of the fundamental from the years that is was available. I did not encounter cases where all but one fundamental variable was present, so I did not need to correct for such cases. Finally, for one company there were no fundamentals available at all and therefore this company was excluded from the portfolio based on fundamental indexation. However, since it is only a small proportion of the total portfolio I argue that this will not bias the results from the tests. Table 3 below shows the total number of companies included in the regional portfolio for each year.

Table 3: Number of companies included in the home biased portfolios

This table shows the total number of companies that are included in the regional portfolio for each year during the period 1996 – 2009.

Year

Number of companies

1996 58 1997 63 1998 64 1999 72 2000 76 2001 79 2002 83 2003 83 2004 84 2005 88 2006 91 2007 93 2008 108 2009 107

(25)

25

the portfolios yield a mean return of 15.6%. However, there are considerable differences between the different home bias portfolios. For starters, when looking at the median return, the equally weighted portfolio ranks first providing with a median return of 0.6% on a weekly basis, whereas the portfolio based on fundamental indexation provides a 0.4% median weekly return. Also, the standard deviation in the returns of the portfolios differ. The equally weighted portfolio yields the most stable returns, with a standard deviation of 0.022 on a weekly basis. Using the rule that the standard deviation tends to increase with the square root of time, this implies a yearly standard deviation of 15.86%. For both the number of employees based portfolio and the fundamental indexation portfolio, the yearly standard deviation is 18.75%. The differences between the minimum and the maximum return, are consistent with the results from the standard deviation: the spread in returns is smallest in the equally weighted portfolio, second smallest for the number of employees portfolio and largest for the fundamental indexation portfolio. This leads me to conclude that even though the portfolios yield the same mean return, the equally weighted portfolio seems to do so at a lower risk than the other home bias portfolios.

(26)

26

Panel A: Definition of variables

Variable name Definition

Number of employees Weekly returns for the home bias portfolio based on number of employees as determinant for portfolio weights

Equally weighted Weekly returns for the home bias portfolio based on an equal weights as determinant for portfolio weights

Fundamental

indexation Weekly returns for the home bias portfolio based on fundamental indexation as determinant for portfolio weights

Risk free rate Euribor: 1 year

Jensen's alpha (α) Measure of outperformance relative to a benchmark Market (β1) Measure of sensitivity of returns relative to a benchmark

HML (β2)

SMB (β3)

Measure of sensitivity of returns to the book to market factor Measure of sensitivity of returns to the size factor

Panel B: Descriptive statistics based on weekly data

number of employees weighted equally fundamental indexation

Mean 0.003 0.003 0.003 Median 0.005 0.006 0.004 Standard deviation 0.026 0.022 0.026 Minimum -0.100 -0.082 -0.154 Maximum 0.127 0.132 0.211 Skewness -0.319 -0.315 0.252 Kurtosis 5.71 6.20 10.62 Jarque-Bera 235.32 324.66 1777.38 Number of observations 731 731 732

Table 4: Variable definition and descriptive statistics of the returns for the home biased portfolios

(27)

27

IV. Results

Table 5 shows the results of the Jensen‟s alpha regression, using equation (1) and a classical linear regression. The ARCH-LM test is performed to test for heteroscedasticity and since no significant ARCH effects are found, the classical linear regression model is used. The table shows the values for Jensen‟s Alpha (α) and the value for the sensitivity of portfolio returns to the returns of the benchmark (β), and the value for the R-squared measure. This regression is performed both against the AEX index and against the portfolio of funds, which is considered to be the return for an average home biased investor. It is easy to see that all values for Jensen‟s alpha are both positive and significant, at either the 95% or 99% level. This means that the home biased portfolios were able to outperform the benchmarks on a risk adjusted basis (note: I refer to the locally situated based portfolios as „the home biased portfolios‟). Furthermore, the three different weighting schemes all lead to such positive and significant results, thereby confirming the robustness of this result.

(28)

28

Table 5: performance evaluation of the home biased portfolios - Jensen's alpha

This table shows the results of the regression given by equation (1). The sample spans the period of January 1, 1996 until December 31, 2009. Panel A1 and A2 give the results of the regression using the AEX index and the portfolio of funds as benchmarks. Panel B1 and B2 give the same information, including transaction costs as described by equation (4). Coefficients denoted with *** are significant at a 99% level, coefficients with ** at a 95% level and coefficients with a * at a 90% level. Standard errors are denoted between parentheses. Returns are on a weekly basis.

Panel A1: AEX - no trading costs

number of employees equally weighted fundamental indexation

Jensen's alpha 0.002 *** 0.002 *** 0.002 ***

(0.00) (0.00) (0.00)

Market (β1) 0.641 *** 0.558 *** 0.612 ***

(0.01) (0.01) (0.02)

R-squared 78.3% 77.2% 66.4%

Panel A2: Portfolio of funds - no trading costs

Jensen's alpha 0.001 ** 0.002 *** 0.001 **

(0.00) (0.00) (0.00)

Market (β1) 0.953 *** 0.856 *** 0.927 ***

(0.02) (0.02) (0.03)

R-squared 70.6% 74.3% 62.1%

Panel B1: AEX - including trading costs

Jensen's alpha 0.001 *** 0.002 *** 0.001 **

(0.00) (0.00) (0.00)

Market (β1) 0.646 *** 0.561 *** 0.618 ***

(0.01) (0.01) (0.02)

R-squared 78.3% 77.1% 66.7%

Panel B2: Portfolio of funds - including trading costs

(29)

29

The values for the R-squared factor range from 62.1% to 78.3% using Jensen‟s alpha and from 62.7% to 78.5% using the Fama and French three factor model. This thus means that both models are rather similar in terms of their explanatory power with respect to the portfolio performance. Alternatively, the R-squared measure can be interpreted as the similarity between the different investment strategies. A low value for the R-squared measure means that the portfolio and the benchmark are rather different strategies. It can be seen that even though the home biased portfolios are not highly similar to the benchmarks, there is still more co-movement than would be expected from the portfolio composition. As shown in table 2, the AEX companies form only a small portion of the home biased portfolios. From the companies with largest portfolio weights accumulating to almost 70%, only 38% is accumulated by companies which are also present in the AEX index. Furthermore, in total for the year 2009, AEX companies accounted for only 37% in the home biased portfolio based on number of employees. In that same year, only 13 of 25 AEX companies were included in the home biased portfolios. Finally, table A1 in the appendix shows a comprehensive list of the funds included in the home biased portfolios. Indeed, the home biased portfolios do include both rather different equities and a larger number of equities. Therefore they do not necessarily yield very similar return patterns when compared to the AEX or the portfolio of funds based on this index.

(30)

30

Table 6: performance evaluation of the home biased portfolios - Fama and French three factor model

Panel A1: AEX - no trading costs

Alpha 0.001 *** 0.002 *** 0.002 *** (0.00) (0.00) (0.00) Market (β1) 0.640 *** 0.555 *** 0.616 *** (0.01) (0.01) (0.02) HML (β2) 0.000 ** 0.001 ** 0.000 (0.00) (0.00) (0.00) SMB (β3) 0.000 0.001 ** 0.000 (0.00) (0.00) (0.00) R-squared 78.5% 77.4% 66.5%

Panel A2: Portfolio of funds - no trading costs

(31)

31

Table 6: continued

Panel B1: AEX - including trading costs

Alpha 0.001 ** 0.001 *** 0.001 ** (0.00) (0.00) (0.00) Market (β1) 0.644 *** 0.558 *** 0.621 *** (0.01) (0.01) (0.02) HML (β2) 0.001 *** 0.001 ** -0.000 (0.00) (0.00) (0.00) SMB (β3) 0.000 0.001 ** -0.001 (0.00) (0.00) (0.00) R-squared 78.5% 77.4% 66.8%

Panel B2: Portfolio of funds - including trading costs

(32)

32

Jensen‟s alpha and the Fama and French model fail to recognize that investors do not always diversify their portfolios in such a way that it eliminates unsystematic risk. Therefore the significant outperformance of the home biased portfolios is tested using the Sharpe model as well. Table 7 below shows the results for the Sharpe ratio, calculated using equation (3) and subsequently ranked descending from the highest value for the Sharpe ratio to the lowest value of the Sharpe ratio. Since a higher Sharpe ratio is to be preferred over a lower Sharpe ratio, it is clear that the home biased portfolios based on locally situated firms are to be preferred over the benchmarks, even is unsystematic risk is present.

Table 7: Sharpe ratios

This table gives the Sharpe ratios - as calculated using equation (3) – for the home biased portfolios (with weighting schemes based on number of employees, equally weighted and fundamental indexation) and the benchmarks. The Sharpe ratios are given both without and including trading costs, as calculated using equation (4). The sample spans the period of January 1, 1996 until December 31, 2009. Data used is on a weekly basis. The portfolios and benchmarks are ranked from highest to lowest value for the Sharpe ratio.

Portfolio / Benchmark No trading costs

Home Bias - equal weights 0.0977

Home Bias - fundamental indexation 0.0768

Home Bias - number of employees 0.0744

Portfolio of funds focused on the Netherlands 0.0298

AEX 0.0147

Portfolio / Benchmark Including trading costs

AEX 0.0325

(33)

33

in the data. In addition, the data is split in two part according to the late-2000s financial crisis. November 1, 2007 is used as the date that marks the start of this financial crisis. The results of both Jensen‟s alpha and the Fama and French three factor model can be found in tables A2A and A2B in appendix 2.

(34)

34

V. Conclusions

I constructed home biased portfolios based on locally situated companies for the province of Groningen, using three different weighting strategies: with weights based on the number of employees, with weights based on equal weights, and with weights based on fundamental indexation. The home biased portfolios were constructed for the period January 1, 1996 until December 31, 2009. A period that is merely chosen due to the availability of employment data for the region Groningen. Finally, the AEX index and a portfolio of index funds that represents the average Dutch home biased investor were used as benchmarks.

Regressions using Jensen‟s alpha revealed both positive and significant outperformance of the home biased portfolios based on locally situated companies with respect to the benchmarks based on locally quoted companies. Also, in the situation where additional risk factors are introduced – as is the case with the Fama and French three factor model – positive and significant outperformance is found. Finally, this result also holds in the presence of unsystematic risk, given the positive results from the analysis of the Sharpe ratio. Taking all these results together, it is possible to answer the research question: home biased portfolios based on locally situated companies significantly outperform the benchmark which is made up by a portfolio of locally quoted companies based funds. Thus, apart from the

rationale for regional investing products based on locally situated companies (i.e. to provide

an answer to demand for such products and the fact that different types of investors are already able to exploit informational advantages and earn higher returns on local investments), there now is evidence that such products can also be valuable since the test yield significant outperformance.

(35)

35

portfolios used in this thesis were created for the province of Groningen. It is of course possible to construct such portfolios for other regions and countries as well.

(36)

36

Appendices

Appendix 1:

Table A1: Overview of funds included in the home biased portfolios

This table presents a comprehensive overview of the companies/ equities that are – at some time during the period of January 1, 1996 until December 31, 2007 – included in the home biased portfolios.

Accor Exxon Mobil PPG

Adecco FMC RANDSTAD

AEGON Foot Locker Rexel

Ahold Ford Motor RWE

AKZO Fortis Saint Gobain Group

AON FREE RECORD SHOP SCA

Arcelor Mittal FUGRO NV Scania

Arriva G4S Schuitema

AS Watson GDF Suez SDB

Atos GEA Group SGSN

Ballast Nedam Gerry Weber SHELL

BAT GRONTMIJ Sligro Food Groop

Bayer AG H&M Smurfit

BIDVEST Henkel SNS Reaal

BILFINGER Hogg Robinson Group SODEXO

Björn Borg Holmen Solar a/s group

BP Hornbach Holding STORK

Brunel Hunter Douglas NV Suez Environment

Burger King ICT TataChemeq

Cameron Imtech Teijin Limited

Canon ING The Carphone Warehouse

Cap Gemini Intersport THOMAS COOK INTL

Cargotec JC Decaux Tieto

Carpetright KBB TMG

Charles Vogele Kentucky Fried Chicken TNT

Chevron Kone Cranes Tommy Hilfiger

COATS Koninklijke BAM TOTAL

Cofely KPN Trelleborg

Compass Group Laurus UK Mail Group

CRH Logica Unicharm

CSM l'oreal USG People

Deutsche Telekom LVMH UTX

Douglas Holding Macintosh Retail Group Van Lanschot

Dow Chemical Manpower Vendex

DPZ McDonalds Vendex KBB

Draka Holding MEDIQ Vodafone

DSM Metro

Econosto Michelin

Elior NBTY

ERIKS nv Olympia Flexgroup

(37)

37

Appendix 2:

Table A2A: Robustness check - Pre November 1, 2007 - Jensen's alpha & Fama and French three factor model

This table shows the results of the regressions given by equation (1) and equation (2). The sample spans the period of January 1, 1996 until September 31, 2007. Panel A contains the results for Jensen‟s Alpha, panel B contains the results for the Fama and French three factor model. Coefficients denoted with *** are significant at a 99% level, coefficients with ** at a 95% level and coefficients with a * at a 90% level. Standard errors are denoted between parentheses. Returns are on a weekly basis.

Panel A1: Jensens Alpha - AEX

Jensen's alpha 0.001 *** 0.002 *** 0.002 ***

(0.00) (0.00) (0.02)

Market (β1) 0.604 *** 0.505 *** 0.534 ***

(0.01) (0.01) (0.02)

R-squared 74.9% 74.2% 64.8%

Panel A2: Jensen's Alpha - Portfolio of funds

Jensen's alpha 0.002 ** 0.002 *** 0.002 ***

(0.00) (0.00) (0.00)

Market (β1) 0.916 *** 0.796 *** 0.839 ***

(0.02) (0.02) (0.03)

R-squared 69.2% 74.0% 64.2%%

Panel B1: Fama and French three factor model - AEX

Alpha 0.001 *** 0.002 *** 0.002 *** (0.00) (0.00) (0.00) Market (β1) 0.602 *** 0.500 *** 0.534 *** (0.01) (0.01) (0.02) HML (β2) 0.001 0.000 ** -0.000 (0.00) (0.00) (0.00) SMB (β3) 0.000 0.001 ** -0.000 (0.00) (0.00) (0.00) R-squared 75.3% 74.5% 64.8%

Panel B2: Fama and French three factor model - Portfolio of funds

(38)

38

Appendix 2, continued:

Table A2B: Robustness check - Post November 1, 2007 - Jensen's alpha & Fama and French three factor model

This table shows the results of the regressions given by equation (1) and equation (2). The sample spans the period of November 1, 2007 until December 31, 2009. Panel A contains the results for Jensen‟s Alpha, panel B contains the results for the Fama and French three factor model. Coefficients denoted with *** are significant at a 99% level, coefficients with ** at a 95% level and coefficients with a * at a 90% level. Standard errors are denoted between parentheses. Returns are on a weekly basis.

Panel A1: Jensens Alpha - AEX

Jensen's alpha 0.001 *** 0.002 *** 0.002 ***

(0.00) (0.00) (0.02)

Market (β1) 0.604 *** 0.505 *** 0.534 ***

(0.01) (0.01) (0.02)

R-squared 88.2% 70.8% 75.5%

Panel A2: Jensen's Alpha - Portfolio of funds

Jensen's alpha 0.002 ** 0.002 *** 0.003 **

(0.00) (0.00) (0.00)

Market (β1) 1.058 *** 1.020 *** 1.164 ***

(0.06) (0.05) (0.09)

R-squared 74.7% 77.4% 62.3%

Panel B1: Fama and French three factor model - AEX

Alpha 0.001 ** 0.003 ** 0.003 (0.00) (0.00) (0.00) Market (β1) 0.746 *** 0.706 *** 0.846 *** (0.03) (0.03) (0.04) HML (β2) 0.001 * 0.001 ** -0.000 (0.00) (0.00) (0.00) SMB (β3) -0.001 -0.001 -0.004 ** (0.00) (0.00) (0.00) R-squared 88.5% 88.3% 77.2%

Panel B2: Fama and French three factor model - Portfolio of funds

(39)

39

Appendix 3:

Table A3: Robustness check - Sharpe ratios

This table gives the Sharpe ratios - as calculated using equation (3) – for the home biased portfolios (with weighting schemes based on number of employees, equally weighted and fundamental indexation) and the benchmarks. The Sharpe ratios are given both without and including trading costs, as calculated using equation (4). The sample spans the period of January 1, 1996 until December 31, 2009. Data used is on a weekly basis. The portfolios and benchmarks are ranked from highest to lowest value for the Sharpe ratio.

Portfolio / Benchmark Pre November 1, 2007

AEX 0.0422

Portfolio / Benchmark Post November 1, 2007

Home Bias – number of employees -0.0102

Home Bias – equal weights -0.0150

Home Bias – fundamental indexation -0.0222

AEX -0.0926

(40)

40

References

Arnott, Robert D., Hsu, Jason, and Philip Moore, 2005, Fundamental Indexation, Financial

Analysts Journal 61-2, 83-99.

Benninga, Simon, 2008, Financial Modeling, Third Edition (The MIT Press, Cambridge, MA).

Brooks, Chris, 2008, Introductory Econometrics for Finance, Second Edition (Cambridge University Press, Cambridge).

Cai, Fang, and Francis E. Warnock, 2005, International diversification at home and abroad, NBER Working Paper No. W12220, National Bureau of Economic Research.

Coval, Joshua D., and Tobias J. Moskowitz, 1999, Home bias at home: Local equity preference in domestic portfolios, The Journal of Finance 54-6, 2045-2073.

Dziuda, Wioletta, and Jordi Mondria, 2009, Asymmetric information, portfolio managers and home bias, AFA 2010 Atlanta Meetings Paper, Kellogg School of Management.

Errunza, Vihang, Hogan, Ked, and Mao-Wei Hung, 1999, Can the gains from international diversification be achieved with trading abroad?, The Journal of Finance 54-6, 2075-2107. Fama, Eugene F., and Kenneth R. French, 1993, Common risk factors in the returns on stocks and bonds, Journal of Financial Economics 33-1, 3-56.

French, Kenneth R., and James M. Poterba, 1991, International diversification and international equity markets, The American Economic Review 18-2, 222-226. French, Kenneth R., The Kenneth R. French Data Library, 1995 – 2009,

http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/index.html, Retrieved on Saturday April 23, 2011, 10:24 AM.

Hasan, Iftekhar, and Yusif Simaan, 2000, A rational explanation for home country bias,

Journal of International Money and Finance 19-3, 331-361.

(41)

41

LISA – Het werkgelegenheidsregister van Nederland, 1995-2009 (Stichting LISA,

Enschede)

Michaelides, Alexander, 2002, International portfolio choice, liquidity constraints and the home equity bias puzzle, CEPR Discussion paper no. 3066, London School of Economics. Nofsinger, John R., 2008, The Psychology of Investing, 3rd_{edition (Pearson Education Inc.,} Upper Saddle River, NJ).

Thomson Reuters Datastream, 1995-2009 (Thomson Reuters, New York, NY) Thomson Reuters Worldscope, 1995-2009 (Thomson Reuters, New York, NY)

Treynor, Jack, 2005, Why market-valuation-indifferent indexing works, Financial Analysts

Journal 61-5, 65-69.

Uppal, Raman, 1992, The economic determinants of the home country bias in investors‟ portfolios: a survey, Journal of International Financial Management and Accounting 4-3, 171-189.

Wermers, Russ, 2000, Mutual fund performance.: an empirical decomposition into stock-pricing talent, style, transactions costs, and expenses, The Journal of Finance 55-4, 1655-1695.

You, Leyuan, and Robert T. Daigler, 2010, Is international diversification really beneficial?,

The value of regional investment strategies: the region of Groningen case

The value of regional investment strategies:

the region of Groningen case

Mireille Bombeld

Student number: 1529714

Supervisors:

dr. A. (Auke) Plantinga

prof. dr. L.J.R. (Bert) Scholtens

Groningen, May 27, 2011

University of Groningen

Faculty of Economics and Business

Master Thesis, MSc Business Administration

The value of regional investment strategies:

the region of Groningen case

Abstract

Table of contents

I.

Literature Review

II.

Methodology

A.

Regional Investment Products

B.

Portfolio Performance

R

R

R

R











(



)





R

R

R









H

:





0

0

:



H



E

r

r

E







(

)



(

)

E

(

r

)

E

(

r

)



V

TO

C

