University of Amsterdam
Department of Economics & Finance BSc Economics & Finance
The regional and world
integration of Dutch, French
and German firms in the 21
st
century: is the EMU CAPM
or world CAPM able to
outperform the domestic
CAPM?
Author: D.M. Pfundt
Student number: 10204431
Thesis supervisor: Jan Lemmen
PREFACE AND ACKNOWLEDGEMENTS
I experienced writing and working on this thesis as interesting. I was able to get a good description and feeling on academically research. Since I have been working on an empirical subject I got a good experience of the reality and not just theory. Furthermore, I was able to apply skills that I achieved during my study Economics and Finance, e.g. writing, theory and the use of statistical programs. I liked the whole idea of performing a study on a research question and eventually being able to answer that question. All by all, I can conclude that working on a research is not just time-consuming but on the contrary also satisfying.
I would like to specially thank Jan Lemmen of the department of Finance at the University of Amsterdam for refereeing my thesis as my supervisor. Thank you for the quick replies,
helpful comments, support, insights, as well as most of the material on this subject. It is partly due to you that this study turned out like this.
NON-PLAGIARISM STATEMENT
By submitting this thesis the author declares to have written this thesis completely by himself/herself, and not to have used sources or resources other than the ones mentioned. All sources used, quotes and citations that were literally taken from publications, or that were in close accordance with the meaning of those publications, are indicated as such.
COPYRIGHT STATEMENT
The author has copyright of this thesis, but also acknowledges the intellectual copyright of contributions made by the thesis supervisor, which may include important research ideas and data. Author and thesis supervisor will have made clear agreements about issues such as confidentiality.
Electronic versions of the thesis are in principle available for inclusion in any EUR thesis database and repository, such as the Master Thesis Repository of the Erasmus University Rotterdam
ABSTRACT
This study focuses on the appropriate choice of a benchmark market portfolio when using CAPM. From the regression results it follows that the Dutch, French and German markets are not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than their domestic CAPM. Furthermore, the European subprime
mortgage crisis had no impact on this conclusion. However, based on this sample, the European subprime mortgage crisis did have a positive effect on the overall performance of all the CAPMs in estimating of the cost of equity, which improved.
Keywords: Asset Pricing, CAPM, Beta, Stock return JEL classification: G120
TABLE OF CONTENTS
PREFACE AND ACKNOWLEDGEMENTS... ii
ABSTRACT... iii
TABLE OF CONTENTS ... iv
LIST OF TABLES ... v
CHAPTER 1 Introduction ... 1
CHAPTER 2 Literature and theory ... 3
2.1 Theoretical framework ... 3 2.2 Literature review ... 4 CHAPTER 3 Methodology ... 6 CHAPTER 4 Data ... 8 CHAPTER 5 Results... 9 5.1 Empirical results ... 9
5.2 Summary and discussion ... 19
CHAPTER 6 Conclusion ... 22
LIST OF TABLES
Table 5.1 Regression results France period 1 9 Table 5.2 Regression results France period 2 10 Table 5.3 Regression results France period 3 11 Table 5.4 Regression results Germany period 1 12 Table 5.5 Regression results Germany period 2 13 Table 5.6 Regression results Germany period 3 14 Table 5.7 Regression results the Netherlands period 1 15 Table 5.8 Regression results the Netherlands period 2 16 Table 5.9 Regression results the Netherlands period 3 17 Table 5.10 (a) Summary regression results France 19 Table 5.10 (b) Summary regression results Germany 19 Table 5.10 (c) Summary regression results the Netherlands 19
CHAPTER 1 Introduction
Harris et al. (2003) state that the estimatio n of the cost of equity is an important issue faced by financial managers, analysts and academics. According to Bartholdy and Peare (2003) an estimation of the cost of equity is used in many financial decisions within a firm such as capital-budgeting and performance evaluation. The estimates of the cost of equity are also important for the (semi) government who make decisions that affect the public, i.e. the firms and households. Since these decisions have to be made in advance of time the firms and the government are forced to accurately estimate the cost of equity today to minimize the cost of adaption.
From the survey findings of Bruner et al. (1998) it follows that the capital asset pricing model (CAPM) of Sharpe-Lintner is currently the preferred model when estimating the cost of equity among most firms. In line with this finding, Harris et al. (2003) state that from recent survey evidence it follows that the CAPM is the most widely used model among investors and US firms. Furthermore, Harris et al. (2003) enlighten the fact that one of the main decisions to be made when implementing the CAPM is the choice for the benchmark market portfolio. They emphasize the fact that – when using the CAPM – there is no evidence of a clear-cut answer to the question of which benchmark market portfolio is appropriate to use in the case of an open international financial market. They also state that in the case of a closed national financial market, theory suggests that a domestic benchmark market portfolio is more appropriate to use. Thus far empirical research failed to answer the question of which benchmark portfolio is appropriate to use in an open international financial market. Although the study of Harris et al. (2003) concludes that their data, on a sample of S&P500 firms, has a better overall fit with the domestic CAPM compared to the world CAPM. The choice of which benchmark market portfolio to use when estimating the cost of equity using the CAPM basically depends on the degree of integration of a particular market in relation to a regional or global market. If a market is integrated enough into a regional market or world market, those benchmark market portfolio might be more appropriate to use within the CAPM in an open international financial market.
Akdogan (1992) investigated regional integration of eight European countries into the European Community (EC) over time. He concluded that it seemed that all these countries were increasingly integrated into the EC index, based on his measure of integration. In other words, these eight countries were increasingly showing regional integration, into Europe, through time. He continued with supplementary studies on the subject of integration, e.g. Akdogan (1996), and his studies became the basis of the study of Barari (2004) on the integration of Latin American countries. These studies had similar findings on regional - and global integration of different economies as the first study of Akdogan, an increasingly regional integration and world integration of particular economies. The findings of these studies could suggest that economies are integrated enough nowadays for a regional
benchmark market portfolio or world benchmark market portfolio to be more appropriate to use within the CAPM in an open international financial market.
These studies are slightly out-dated, none of these studies linked the integration of an economy to the appropriate choice for a CAPM benchmark market portfolio and there is still no clear-cut answer to the issue of which benchmark market portfolio is appropriate to use in the case of an open international financial market. Through the 21st century financial markets
became more integrated due to removal of almost all trade barriers and the development of technology. Although Europe is too big for my research, these observations lead to my main research question: Are the Dutch, French and German economies integrated enough for the EMU CAPM or the world CAPM to provide better estimates of the cost of equity than their domestic CAPM?
CHAPTER 2 Literature and theory 2.1 Theoretical framework
In this study I will use the CAPM of Sharpe-Lintner, which, according to Fama & French (2004), is an extension of the mean-variance model of Markowitz. Berk & DeMarzo (2014, p.342) state that the CAPM is a model to estimate the expected return on equity based on a benchmark market portfolio volatility and relative exposure to the firm (Berk & DeMarzo, 2014, p. 381). According to Berk & DeMarzo the theoretical CAPM is known by the following formula:
𝐸(𝑅𝑖) = 𝑅𝑓 + 𝛽𝑖∗ (𝐸(𝑅𝑚) − 𝑅𝑓) (2.1)
where E(Ri) is the expected return on equity of firm i, Rf is the risk-free interest rate, βi is the
beta of firm i and E(Rm) is the expected return on the benchmark market portfolio (Berk &
DeMarzo, 2014, p.341-342). The beta, βi, represents the firm’s systematic risk exposure to a
benchmark market, i.e. the sensitivity of a firm’s return to the return of a particular benchmark market portfolio (Berk & DeMarzo, 2014, p.337). In other words, the beta measures the volatility of the stock caused by to the volatility of the market (p. 382). The benchmark market portfolio is assumed to be efficient, i.e. the portfolio with the least risk and highest return or in other words the portfolio with the highest Sharpe ratio (p. 377). The term E(Rm) – Rf, is known as the excess market return or the market risk premium (p. 340). The
multiplication of the excess market return by the firm’s specific beta is known as the risk premium of a specific firm (p. 381). The econometric CAPM formula, which is used in empirical research, is known as follows:
𝑅𝑖 − 𝑅𝑓 = 𝛼𝑖+ 𝛽𝑖∗ (𝑅𝑚− 𝑅𝑓) + 𝜀𝑖 (2.2)
where ε is the error term and αi is the intercept. The error term is the distance above/below the
best fitting line of a regression and equals zero on average, so that the formula can be shown as:
𝑅𝑖 − 𝑅𝑓 = 𝛽𝑖∗ (𝑅𝑚− 𝑅𝑓) + 𝛼𝑖 (2.3)
where the constant term, αi, is known as the stock’s alpha of firm i and represents the
risk-adjusted measure of the stock’s historical performance. According to the CAPM the αi should
not be significantly different from zero, where the βi should be significantly different from
zero (Berk & DeMarzo, 2014, p. 410). Furthermore, the CAPM has three underlying assumptions regarding the behavior of investors (Berk & DeMarzo, 2014, p. 379-380):
1. Investors can buy and sell all securities at competitive market prices (without
incurring taxes or transactions costs) and can borrow and lend at the risk-free interest rate.
2. Investors only hold efficient portfolios of traded securities, which are portfolios that yield the maximum expected return for a given level of volatility.
3. Investors have homogeneous expectations regarding the volatilities, correlations and expected returns of securities.
2.2 Literature review
The study of Akdogan (1992) on the integration of regional (EC) capital markets uses the CAPM with a European Community index, as a benchmark market portfolio, to test whether eight European markets are integrated into Europe. He picked Belgium, Denmark, France, Germany, Netherlands, Italy, Spain and the U.K. over a time period of 18 years, 1972 – 1990. He made eight data periods within his total time frame, based on data concerning legal
changes on trade barriers. As a measure of integration Akdogan used the following formula: 𝑝𝑖 = 𝛽𝑖2𝑣𝑎𝑟 (𝑅𝑤)
𝑣𝑎𝑟 (𝑅𝑖) (2.4)
His results show that in their last periods all eight countries seem to be increasingly integrated into the EC index, which is represented by an increasing value for pi. In other
words, Akdogan also shows that each country market’s proportion of systematic risk
increased over time due to removal of capital controls, trade barriers and the higher ability to trade in foreign markets.
A subsequent and more recent study of Akdogan (1996) on the integration versus
segmentation of countries states that an international fund manager should select the most segmented countries when investing in any opportunities. According to Akdogan (1996) it is a fact that most economies have become more interdependent and more integrated through time. He proposes the idea that this is due to the abolishing of capital controls and trade barriers. He acknowledges that evidence suggests that most markets could be classified as mildly or moderately integrated. According to Akdogan (1996) capital market integration implies the absence of risk premium differentials. He states that if risk is priced the same across world markets those markets are integrated. He also introduces another view on integration by comparing a countries’ market to a benchmark market. According to Akdogan (1996) the relevant measure for integration is the country’s systematic risk contribution to the benchmark market. He implies that a bigger systematic risk contribution implies a higher degree of integration. The relevant measure used in the study of Akdogan (1996) is:
𝑝𝑖 = 𝛽𝑖2𝑣𝑎𝑟 (𝑅𝑤)
𝑣𝑎𝑟 (𝑅𝑖) (2.4)
Akdogan states that a higher (or growing) fraction pi shows that the market is integrated
(more integrated). In his study he used a sample of 26 countries in a time frame from 1972 – 1989, which he cut into two periods (the decade of the 1970s and the decade of the 1980s). From his research he finds empirical evidence that some small to medium-sized European capital markets (Finland, Denmark, Spain & Italy) showed segmentation, while other countries (U.K., Japan, France, Australia and most of the emerging markets) showed integration through time.
The study of Barari (2004) on equity market integration in Latin America is an extension on the studies of Akdogan (1992, 1996). She attempts to investigate the regional integration and global integration of six Latin American countries in the time period 1988 – 2001. Barari extends the methodology of Akdogan (1996) by two dimensions. First of all, instead of
basing integration on one market portfolio she measures integration based on two market portfolios. In this way she can compare the status of regional integration and global
integration of a country by examining a country ratio of the regional integration score to the global integration score. Secondly, she uses a time-varying integration score instead of dividing the entire sample into periods. In this way she attempts to remove the sensitivity of the results to the cut-off dates. She states that such an exact choice of cut-off dates requires detailed data on the particular change. The measure of integration is more or less the same as that of Akdogan (1992, 1996). Barari concludes that during the late 1980s and the first half of the 1990s there was a move towards regional integration and away from global integration for most countries in her sample. From her results it becomes clear that from the mid-1990s the pace of global integration relative to regional integration accelerated and remained
accelerating. She states that the timing may suggest possible contagion effects, but that conclusion needs more detailed study.
The study of Harris et al. (2003) is about the choice between a global CAPM and a domestic CAPM. Their study focuses on a sample of S&P500 firms over the period 1983 – 1998. They state that the estimation of the cost of equity remains an important issue to financial
managers, analysts and academics. They highlight the fact that, when implementing CAPM, one of the main decisions is the choice on the appropriate benchmark market portfolio.
According to the theory a domestic market portfolio is the appropriate portfolio to use in case of a closed national market. However, thus far there is no clear-cut evidence on which market portfolio is appropriate to use in case of an open international market. In their study Harris et al. perform a time-varying OLS regression to estimate the annual betas for both CAPMs. They multiply these betas by the excess market return to achieve the CAPM estimates for the risk premiums. Further, they compare the CAPM risk premiums estimates with the ex ante risk premiums to conclude how close the estimates were. Finally, they conclude in how many years the domestic CAPM estimate was closer to the actual risk premium compared to the global CAPM. The study of Harris et al. (2003) leads to the conclusion that the domestic CAPM (S&P500 index) has a better overall fit with the actual risk premiums than the global CAPM (MSCI World Index). Although, the relatively small difference between the domestic CAPM and the global CAPM, both compared to the actual risk premiums, may suggest that the choice between these market portfolios is irrelevant for the S&P500 firms and investors. In other words, the American economy is not integrated enough for the world market
portfolio (global CAPM) to provide better estimates of the return on equity than the domestic market portfolio (domestic CAPM).
CHAPTER 3 Methodology
I will be using the CAPM with five different benchmark market portfolios (Dutch, French, German, EMU and world) to answer my research question - if the Dutch, French and German economies are integrated enough for the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the domestic CAPM. I will only use a sample of ten Dutch, French and German firms to represent the countries’ financial markets. Furthermore, I will cut my time frame into three time periods – overall (2000 – 2014), before the subprime mortgage crisis (2000 – 2006) and since the crisis (2007 – 2014). The latter two so that I will be able to analyse whether the European subprime mortgage crisis influenced the choice between the benchmark market portfolios, which might be due to a more integrated financial market since the start of the subprime mortgage crisis.
First of all, I will need to collect data on the stock market values of the firms as well as the stock market values of the indices. Subsequently, I will calculate the realized returns on the firms and on the indices from these stock market values.
Regarding my methodology, I have to make an important note on the risk-free interest rate. According to my methodology, an ordinary least squared (OLS) regression basically
estimates a coefficient and an intercept through some scatter points that minimizes the
deviation from these scatter points. In the case of CAPM, the scatter points are represented by the excess return on a stock on the y-axis, 𝑅𝑖 − 𝑅𝑓, and the excess return on the market on the x-axis, 𝑅𝑚− 𝑅𝑓, as shown by the following formulas:
𝑅𝑖 = 𝑅𝑓 + 𝛽𝑖∗ (𝑅𝑚− 𝑅𝑓) (3.1) 𝑅𝑖 − 𝑅𝑓 = 𝛽𝑖∗ (𝑅𝑚− 𝑅𝑓) (3.2)
The same risk-free interest rate is subtracted from both the return on the stock as well as the return on the market. In other words, the real values change but the relative relationship remains unchanged, so that it does not affect the value of the slope, i.e. the beta. Therefore, I do not collect the data on the risk-free interest rate.
Secondly, I will perform multiple OLS regressions of the return on the stocks, per country, on the return on the three relevant benchmark market portfolios. From these regressions I will achieve estimates of the beta (the slope), estimates of the alpha (constant), the coefficients of determination and the root-mean-square errors.
Thirdly, I will check whether the CAPM is significant by evaluating the t-test values using a significance level of 5%. According to the theory, as stretched by Berk and DeMarzo (2014), the constant should not be significantly different from zero, while the beta should be
significantly different from zero. If this holds the CAPM is valid. Furthermore, the central limit theorem states that the OLS standard error converges to the true standard error in the case of large amounts of data. This theorem implies that the heteroscedasticity and
autocorrelation in the data are no longer relevant issues. Since I have large amounts of data, heteroscedasticity and autocorrelation should not be an issue in my study. Therefore, I do not need to test my data on this.
Finally, I will be able to answer my research question by evaluating the performance of the three relevant CAPMs per country. To evaluate the performance of the CAPMs I will use the coefficients of determination, R2, and the root-mean-square error, RMSE. The coefficient of
determination reflects the percentage of the variation in the return on the stock that is explained by the variation in the return on the particular market portfolio. The root-mean-square error is a measure of the difference in the estimated CAPM returns and the actual realized returns. I will take the average of the measures of interest per country. I will compare the average of the coefficients of determination and the average of the root-mean-square errors to determine which benchmark market portfolio performed best, i.e. the highest value for R2 and the lowest value for RMSE or in other words which benchmark market portfolio
provided the best estimates of the cost of equity when used in the CAPM. This benchmark market portfolio outperformed the other benchmark market portfolios, and if this is not the domestic market portfolio we can conclude that for that particular country the firm is
integrated enough for either the EMU CAPM or the world CAPM to provide better estimates of the cost of equity than the domestic CAPM.
CHAPTER 4 Data
As stated before, I need the stock market values on thirty firms and five market portfolios from 2000 - 2014. Once I gained this data I will need to calculate the monthly returns, which I will do by using the following formula:
𝑅𝑖,𝑡 =𝑀𝑉𝑡 +1− 𝑀𝑉𝑡 𝑀𝑉𝑡
where MVt + 1 and MVt are the monthly stock market values of either the market portfolio or
a firm. Once all the monthly returns are calculated the data set is completed.
I used Datastream, a database provided by the University of Amsterdam, to collect the data on the stock market values. I collected the data on the stock market values of the firms and the indices. I chose a monthly basis, which is similar to the study of Akdogan (1992, 1996), Barari (2004) and Harris et al. (2003). I selected the thirty firms, ten per country, randomly. Furthermore, similar to the study of Harris et al. (2003) I used MSCI indices as the
benchmark market portfolios of the countries, region and world, instead of e.g. AEX, DAX and CAC. I used the MSCI indices because they have a similar structure and they cover for most of the stock market capitalization. The composition and summary of the MSCI indices are available on request with the author. Furthermore, as the European index I took the European Monetary Union index since all three countries are in the EMU and therefore it will be a better representation of the financial market of Europa for these countries.
CHAPTER 5 Results 5.1 Empirical results
I performed multiple OLS regressions of a total of thirty firms originating from France, Germany and the Netherlands on three benchmark market portfolios. I made a distinction between three different time periods – overall (1), before the crisis (2) and since the crisis (3). I will describe some of the results of my regressions per country and by relative time order. The complete STATA output and results are available on request with the author. In all the following tables the t-test values of respectively the alpha and the beta estimates are within brackets. The t-tests were based on the null-hypothesis that the alpha and the beta are equal to zero. I use a 5% significance level to determine whether the estimates were significantly different from zero. Furthermore recall that based on the central limit theorem I do not need to test on heteroscedasticity and autocorrelation.
France
Table 5.1 contains some of the regression results of the ten French firms in time period 1.
Firms Index Intercept, α Slope, β R2 RMSE
1. Air Liquide France 0,0073 (2,43) 0,5458 (7,47) 0,3441 0,0398 EMU 0,0073 (2,43) 0,5127 (7,30) 0,3268 0,0403 World 0,0064 (2,03) 0,5553 (7,98) 0,3183 0,0405 2. AXA France 0,0023 (0,45) 1,7257 (11,83) 0,6469 0,0673 EMU 0,0023 (0,48) 1,6901 (12,15) 0,6678 0,0653 World -0,0001 (-0,02) 1,6480 (9,54) 0,5272 0,0779 3. BNP Paribas France 0,0043 (0,95) 1,2727 (11,26) 0,5503 0,0607 EMU 0,0045 (0,96) 1,1952 (10,53) 0,5224 0,0626 World 0,0029 (0,55) 1,1349 (6,77) 0,3911 0,0706 4. Carrefour France -0,0038 (-0,84) 0,8055 (8,28) 0,3317 0,0603 EMU -0,0037 (-0,79) 0,7452 (7,49) 0,3056 0,0615 World -0,0046 (-0,92) 0,6908 (5,93) 0,2180 0,0653 5. Essilor France 0,0114 (3,05) 0,3699 (5,24) 0,1357 0,0493 EMU 0,0114 (3,06) 0,3513 (5,41) 0,1317 0,0494 World 0,0108 (2,84) 0,3523 (4,57) 0,1100 0,0500 6. Groupe Danone France 0,0063 (1,75) 0,4304 (4,99) 0,1873 0,0473 EMU 0,0064 (1,75) 0,3903 (4,70) 0,1658 0,0479 World 0,0057 (1,54) 0,4099 (4,74) 0,1519 0,0483 7. L'Oréal France 0,0045 (1,23) 0,6009 (7,31) 0,3032 0,0481 EMU 0,0046 (1,24) 0,5463 (6,59) 0,2696 0,0492 World 0,0039 (0,98) 0,5205 (5,93) 0,2032 0,0514 8. Orange France -0,0025 (-0,31) 0,9536 (4,28) 0,1779 0,1082 EMU -0,0026 (-0,32) 0,9525 (4,33) 0,1911 0,1073 World -0,0033 (-0,39) 0,7827 (4,11) 0,1071 0,1127 9. Renault France 0,0069 (1,05) 1,6101 (7,89) 0,4770 0,0890 EMU 0,0070 (1,08) 1,5798 (8,28) 0,4943 0,0875 World 0,0039 (0,60) 1,7169 (9,08) 0,4847 0,0883 10. Total France 0,0052 (1,67) 0,6677 (10,00) 0,4208 0,0413
EMU 0,0053 (1,64) 0,5987 (8,64) 0,3641 0,0433 World 0,0043 (1,27) 0,6164 (7,19) 0,3205 0,0448
AVG RMSE France 0,0611 AVG R2 France 0,3575
EMU 0,0614 EMU 0,3439 World 0,0649 World 0,2832 Table 5.1 Regression results France period 1
From table 5.1 it follows that, based on the t-test values, all the beta estimates are significantly different from zero. Also based on the t-test values it follows that all the alpha estimates are not significantly different from zero, except for all the Air Liquide and Essilor CAPM regressions. The finding that the alpha estimate is not significantly different from zero and the beta estimate is significantly different from zero implies that the CAPM is valid in these cases. In the exceptional cases of Air Liquide and Essilor all three CAPMs generate an alpha estimate that is significantly different from zero. Therefore, for all cases of Air Liquide and Essilor the CAPM is invalid. Furthermore, from table 5.1 it follows that the French CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the French market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the French CAPM.
Table 5.2 contains some of the regression results of the ten French firms in time period 2.
Firms Index Intercept, α Slope, β R2 RMSE
1. Air Liquide France 0,0075 (1,58) 0,3909 (2,91) 0,1811 0,0427 EMU 0,0075 (1,58) 0,3771 (3,07) 0,1826 0,0427 World 0,0072 (1,42) 0,4174 (2,68) 0,127 0,0441 2. AXA France 0,0029 (0,42) 1,6555 (8,09) 0,6293 0,0653 EMU 0,0028 (0,42) 1,6174 (8,78) 0,6507 0,0634 World 0,0017 (0,19) 1,7981 (5,97) 0,4568 0,0791 3. BNP Paribas France 0,0100 (1,90) 0,9345 (6,62) 0,5117 0,0469 EMU 0,0100 (1,85) 0,8731 (6,17) 0,4839 0,0482 World 0,0096 (1,47) 0,8752 (3,82) 0,2762 0,0571 4. Carrefour France -0,0057 (-0,88) 0,7767 (5,02) 0,3164 0,0587 EMU -0,0056 (-0,84) 0,6958 (4,48) 0,2751 0,0604 World -0,0060 (-0,82) 0,7108 (3,06) 0,163 0,0649 5. Essilor France 0,0134 (2,29) 0,2325 (1,96) 0,0496 0,0523 EMU 0,0134 (2,30) 0,2204 (2,08) 0,0483 0,0523 World 0,0133 (2,23) 0,2109 (1,27) 0,0251 0,0530 6. Groupe Danone France 0,0106 (1,85) 0,4323 (3,16) 0,1532 0,0522 EMU 0,0107 (1,83) 0,3787 (2,96) 0,1274 0,0530 World 0,0100 (1,74) 0,5508 (3,29) 0,153 0,0522 7. L'Oréal France 0,0011 (0,18) 0,5558 (3,87) 0,2082 0,0557 EMU 0,0012 (0,19) 0,4806 (3,30) 0,1686 0,0571 World 0,0010 (0,15) 0,4878 (2,59) 0,0987 0,0594 8. Orange France -0,0093 (-0,58) 1,5303 (3,01) 0,2362 0,1414 EMU -0,0095 (-0,61) 1,5625 (3,23) 0,2668 0,1386
World -0,0107 (-0,65) 1,7222 (3,23) 0,1841 0,1462 9. Renault France 0,0118 (1,39) 0,9433 (3,64) 0,2802 0,0777 EMU 0,0118 (1,39) 0,8919 (3,77) 0,2714 0,0782 World 0,0114 (1,22) 0,9248 (3,32) 0,1657 0,0837 10. Total France 0,0090 (1,94) 0,5645 (5,19) 0,3169 0,0426 EMU 0,0092 (1,87) 0,4768 (4,27) 0,2449 0,0448 World 0,0086 (1,72) 0,6231 (4,24) 0,2376 0,045
AVG RMSE France 0,0636 AVG R2 France 0,2883
EMU 0,0639 EMU 0,2720 World 0,0685 World 0,1887 Table 5.2 Regression results France period 2
From table 5.2 it follows that, based on the t-test values, all the beta estimates are
significantly different from zero, except for the Essilor world CAPM regression. Also based on the t-test values it follows that the alpha estimates are not significantly different from zero in all the cases, except for all Essilor regressions. These results imply that all CAPMs are valid apart from the exceptions. From the results in table 5.2 it also follows that the French CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the French market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the French CAPM.
Table 5.3 contains some of the regression results of the ten French firms in time period 3.
Firms Index Intercept, α Slope, β R2 RMSE
1. Air Liquide France 0,0070 (1,89) 0,6646 (8,49) 0,5005 0,0362 EMU 0,0072 (1,87) 0,6179 (7,26) 0,4633 0,0375 World 0,0056 (1,44) 0,6146 (8,22) 0,4733 0,0371 2. AXA France 0,0017 (0,23) 1,7797 (8,57) 0,6605 0,0695 EMU 0,0019 (0,26) 1,7466 (8,54) 0,6814 0,0673 World -0,0016 (-0,19) 1,5857 (7,43) 0,5799 0,0773 3. BNP Paribas France -0,0008 (-0,12) 1,5337 (9,70) 0,6058 0,0674 EMU -0,0005 (-0,07) 1,4464 (8,59) 0,5771 0,0698 World -0,0030 (-0,39) 1,2507 (5,66) 0,4456 0,0799 4. Carrefour France -0,0022 (-0,34) 0,8270 (6,53) 0,3435 0,0623 EMU -0,0020 (-0,31) 0,7830 (5,88) 0,3298 0,0629 World -0,0034 (-0,49) 0,6812 (5,11) 0,2578 0,0662 5. Essilor France 0,0095 (2,02) 0,4759 (6,10) 0,2395 0,0462 EMU 0,0096 (2,03) 0,4533 (6,11) 0,2328 0,0464 World 0,0087 (1,77) 0,4143 (4,97) 0,2008 0,0473 6. Groupe Danone France 0,0025 (0,56) 0,4303 (3,88) 0,2307 0,0428 EMU 0,0026 (0,58) 0,4003 (3,66) 0,214 0,0433 World 0,0019 (0,40) 0,3536 (3,59) 0,1724 0,0444 7. L'Oréal France 0,0074 (1,75) 0,6345 (6,48) 0,4199 0,0406 EMU 0,0076 (1,75) 0,5964 (6,20) 0,3973 0,0414 World 0,0064 (1,37) 0,5320 (5,41) 0,3265 0,0438
8. Orange France 0,0036 (0,62) 0,5092 (4,93) 0,1847 0,0583 EMU 0,0038 (0,64) 0,4776 (4,44) 0,174 0,0587 World 0,0031 (0,51) 0,3777 (4,16) 0,1124 0,0608 9. Renault France 0,0023 (0,26) 2,1229 (9,05) 0,6289 0,0888 EMU 0,0024 (0,29) 2,1146 (10,07) 0,6683 0,0840 World -0,0026 (-0,30) 2,0593 (10,02) 0,6545 0,0857 10. Total France 0,0017 (0,42) 0,7480 (8,50) 0,5111 0,0399 EMU 0,0019 (0,46) 0,6941 (7,43) 0,4713 0,0414 World 0,0006 (0,13) 0,6170 (5,75) 0,3845 0,0447
AVG RMSE France 0,0552 AVG R2 France 0,4325
EMU 0,0553 EMU 0,4209 World 0,0587 World 0,3608 Table 5.3 Regression results France period 3
From table 5.3 and based on the t-test values it follows that all the beta estimates are significantly different from zero and all the alpha estimates are not significantly different from zero, except for Essilor French CAPM and world CAPM regressions. This implies that all the CAPMs are valid apart from the latter exceptions. From the results in table 5.3 it also follows that the French CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the French market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the French CAPM.
Germany
Table 5.4 contains some of the regression results of the ten German firms in time period 1.
Firms Index Intercept, α Slope, β R2 RMSE
1. Allianz Germany -0,0015 (-0,32) 1,3379 (11,81) 0,6445 0,0627 EMU -0,0001 (-0,02) 1,4883 (11,26) 0,5996 0,0666 World -0,0022 (-0,38) 1,4344 (9,45) 0,4624 0,0772 2. Bayer Germany 0,0077 (1,82) 1,0370 (11,64) 0,5632 0,0577 EMU 0,0089 (1,93) 1,1133 (10,04) 0,4879 0,0624 World 0,0075 (1,43) 1,0306 (8,02) 0,3472 0,0705 3. BMW Germany 0,0079 (1,68) 0,9438 (11,58) 0,4798 0,0621 EMU 0,0089 (1,80) 1,0090 (10,00) 0,4123 0,0660 World 0,0072 (1,39) 1,0571 (8,50) 0,3757 0,0680 4. Deutsche Bank Germany -0,0026 (-0,46) 1,2569 (10,70) 0,5228 0,0758 EMU -0,0016 (-0,30) 1,5027 (10,87) 0,5618 0,0727 World -0,0042 (-0,75) 1,5738 (8,65) 0,5117 0,0767 5. Deutsche Telekom Germany -0,0048 (-0,91) 0,7143 (6,38) 0,3037 0,0683 EMU -0,0040 (-0,74) 0,7653 (5,45) 0,2621 0,0703 World -0,0046 (-0,80) 0,6416 (4,52) 0,153 0,0753 6. E.ON Germany 0,0032 (0,72) 0,7155 (8,79) 0,378 0,0579 EMU 0,0038 (0,88) 0,8436 (9,47) 0,395 0,0571 World 0,0026 (0,56) 0,8099 (7,83) 0,3023 0,0614
7. Salzgitter Germany 0,0096 (1,46) 0,8939 (7,48) 0,2888 0,0886 EMU 0,0103 (1,59) 1,0700 (7,93) 0,311 0,0872 World 0,0087 (1,29) 1,0499 (6,49) 0,2486 0,0910 8. SAP Germany 0,0047 (0,67) 1,1732 (7,85) 0,3763 0,0954 EMU 0,0061 (0,83) 1,2200 (6,55) 0,306 0,1006 World 0,0047 (0,59) 1,1142 (5,59) 0,2119 0,1072 9. Siemens Germany 0,0027 (0,61) 1,2708 (15,83) 0,636 0,0607 EMU 0,0040 (0,84) 1,4098 (13,74) 0,5885 0,0645 World 0,0026 (0,44) 1,2284 (9,62) 0,3709 0,0798 10. Volkswagen Germany 0,0127 (1,54) 0,6874 (3,92) 0,142 0,1067 EMU 0,0138 (1,63) 0,6458 (2,80) 0,0942 0,1096 World 0,0130 (1,45) 0,5824 (1,86) 0,0636 0,1115
AVG RMSE Germany 0,0736 AVG R2 Germany 0,4335
EMU 0,0757 EMU 0,4018 World 0,0818 World 0,3047 Table 5.4 Regression results Germany period 1
From table 5.4 it follows that based on the t-test values all the alpha estimates are not significantly different from zero, while all the beta estimates are significantly different from zero except the Volkswagen world CAPM regression. Therefore, all the CAPMs are valid apart from the latter exception. From table 5.4 it also follows that the German CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the German market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the German CAPM.
Table 5.5 contains some of the regression results of the ten German firms in time period 2.
Firms Index Intercept, α Slope, β R2 RMSE
1. Allianz Germany -0,0019 (-0,25) 1,4264 (8,28) 0,6357 0,0718 EMU -0,0039 (-0,45) 1,6593 (7,42) 0,5564 0,0793 World -0,0048 (-0,47) 1,7548 (5,03) 0,3535 0,0957 2. Bayer Germany 0,0052 (0,71) 1,1973 (8,97) 0,5857 0,0670 EMU 0,0035 (0,45) 1,4314 (8,03) 0,5414 0,0705 World 0,0023 (0,26) 1,6234 (5,88) 0,3956 0,0809 3. BMW Germany 0,0077 (1,12) 0,7985 (9,11) 0,4217 0,0622 EMU 0,0066 (0,93) 0,9514 (7,91) 0,3873 0,0640 World 0,0060 (0,76) 1,0229 (4,89) 0,2543 0,0706 4. Deutsche Bank Germany 0,0066 (1,04) 0,9530 (9,58) 0,5454 0,0579 EMU 0,0051 (0,82) 1,1869 (8,86) 0,5472 0,0578 World 0,0041 (0,58) 1,3956 (6,19) 0,4297 0,0648 5. Deutsche Telekom Germany -0,0143 (-1,60) 0,9301 (4,63) 0,3727 0,0803 EMU -0,0156 (-1,69) 1,1313 (3,86) 0,3566 0,0813 World -0,0164 (-1,66) 1,2432 (3,70) 0,2446 0,0881 6. E.ON Germany 0,0126 (2,06) 0,3972 (3,81) 0,1853 0,0554
EMU 0,0119 (1,97) 0,5134 (3,81) 0,2002 0,0549 World 0,0118 (1,82) 0,4776 (2,49) 0,0984 0,0583 7. Salzgitter Germany 0,0351 (3,69) 0,7380 (4,34) 0,2407 0,0872 EMU 0,0340 (3,58) 0,9172 (4,57) 0,2404 0,0872 World 0,0335 (3,27) 0,9469 (3,51) 0,1456 0,0925 8. SAP Germany 0,0095 (0,74) 1,7188 (7,34) 0,4849 0,1178 EMU 0,0072 (0,53) 1,9835 (5,89) 0,4177 0,1253 World 0,0056 (0,38) 2,2674 (4,48) 0,31 0,1364 9. Siemens Germany 0,0047 (0,59) 1,4316 (10,64) 0,6319 0,0727 EMU 0,0025 (0,32) 1,7937 (10,82) 0,6415 0,0717 World 0,0021 (0,20) 1,6332 (4,92) 0,3021 0,1001 10. Volkswagen Germany 0,0106 (1,23) 0,8666 (6,24) 0,3464 0,0792 EMU 0,0094 (1,06) 1,0218 (5,21) 0,3114 0,0813 World 0,0087 (0,91) 1,1160 (4,96) 0,211 0,087
AVG RMSE Germany 0,0751 AVG R2 Germany 0,4450
EMU 0,0773 EMU 0,4200 World 0,0874 World 0,2745 Table 5.5 Regression results Germany period 2
From table 5.5 it follows that based on the t-test values all the beta estimates are significantly different from zero. Furthermore, all the alpha estimates are not significantly different from zero except all Salzgitter CAPM, E.ON German CAPM and E.ON EMU CAPM regressions. Therefore, all the CAPMs are valid apart from the latter exceptions. From table 5.5 it also follows that the German CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the German market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the German CAPM.
Table 5.6 contains some of the regression results of the ten German firms in time period 3.
Firms Index Intercept, α Slope, β R2 RMSE
1. Allianz Germany -0,0006 (-0,10) 1,2459 (8,88) 0,6629 0,0539 EMU 0,0032 (0,58) 1,3547 (8,97) 0,6765 0,0528 World 0,0001 (0,02) 1,2962 (8,27) 0,6395 0,0557 2. Bayer Germany 0,0109 (2,31) 0,8680 (8,97) 0,5596 0,0467 EMU 0,0138 (2,68) 0,8652 (7,63) 0,4798 0,0508 World 0,0121 (2,18) 0,7745 (6,75) 0,3971 0,0547 3. BMW Germany 0,0072 (1,13) 1,0940 (8,20) 0,54 0,0613 EMU 0,0110 (1,57) 1,0531 (6,84) 0,4319 0,0681 World 0,0082 (1,18) 1,0707 (6,92) 0,461 0,0663 4. Deutsche Bank Germany -0,0123 (-1,45) 1,5827 (7,91) 0,566 0,0841 EMU -0,0076 (-0,92) 1,7493 (8,13) 0,5969 0,0810 World -0,0115 (-1,32) 1,6562 (6,81) 0,5525 0,0854 5. Deutsche Telekom Germany 0,0047 (0,89) 0,4791 (4,61) 0,2379 0,0520 EMU 0,0063 (1,16) 0,4785 (4,17) 0,2048 0,0532 World 0,0057 (0,98) 0,3766 (3,47) 0,131 0,0556
6. E.ON Germany -0,0068 (-1,29) 1,0565 (12,96) 0,6007 0,0523 EMU -0,0035 (-0,62) 1,1017 (10,46) 0,5639 0,0546 World -0,0054 (-0,85) 0,9584 (7,64) 0,4407 0,0618 7. Salzgitter Germany -0,0135 (-1,57) 1,0870 (5,94) 0,3853 0,0833 EMU -0,0103 (-1,23) 1,1944 (6,32) 0,4016 0,0822 World -0,0128 (-1,52) 1,1136 (5,64) 0,3604 0,0850 8. SAP Germany 0,0035 (0,63) 0,6148 (5,49) 0,3301 0,0531 EMU 0,0055 (0,99) 0,6280 (5,14) 0,2973 0,0544 World 0,0040 (0,70) 0,6250 (4,84) 0,3041 0,0542 9. Siemens Germany 0,0019 (0,39) 1,1069 (12,89) 0,6741 0,0467 EMU 0,0056 (1,03) 1,1117 (10,65) 0,5869 0,0526 World 0,0031 (0,54) 1,0561 (10,00) 0,5469 0,0551 10. Volkswagen Germany 0,0156 (1,11) 0,4995 (1,51) 0,055 0,1256 EMU 0,1768 (1,30) 0,3530 (0,95) 0,0237 0,1277 World 0,0168 (1,17) 0,3523 (0,83) 0,0244 0,1276
AVG RMSE Germany 0,0659 AVG R2 Germany 0,4612
EMU 0,0677 EMU 0,4263 World 0,0701 World 0,3858 Table 5.6 Regression results Germany period 3
From table 5.6, based on the t-test values, it follows that all the alpha estimates are not significantly different from zero except for all Bayer CAPM regressions. It also follows that all the beta estimates are significantly different from zero except for all Volkswagen CAPM regressions. These results imply that all CAPMs are valid apart from the exceptions. From table 5.6 it also follows that the German CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the German market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the German CAPM.
The Netherlands
Table 5.7 contains some of the regression results of the ten Dutch firms in time period 1.
Firms Index Intercept, α Slope, β R2 RMSE
1. Aegon Dutch -0.0066 (-1.15) 1.7059 (12.82) 0,6141 0,0766 EMU -0.0054 (-0.91) 1.7439 (11.59) 0,5998 0,0780 World -0.0080 (-1.18) 1.7290 (9.36) 0,4895 0,0881 2. Ahold Dutch 0.0016 (0.24) 0.8998 (6.07) 0,2406 0,0905 EMU 0.0026 (0.37) 0.7876 (5.31) 0,1723 0,0945 World 0.0020 (0.27) 0.6511 (3.77) 0,0977 0,0987 3. Akzo Nobel Dutch 0.0029 (0.70) 1.1379 (13.21) 0,5818 0,0546 EMU 0.0040 (0.87) 1.0493 (11.20) 0,4623 0,0620 World 0.0023 (0.46) 1.0652 (9.64) 0,3956 0,0657 4. ASML Dutch 0.0080 (1.06) 1.6050 (8.74) 0,4529 0,0999 EMU 0.0091 (1.22) 1.6878 (9.42) 0,4681 0,0985
World 0.0075 (0.86) 1.4333 (6.07) 0,2803 0,1146 5. Heineken Dutch 0.0052 (1.35) 0.5230 (6.65) 0,2544 0,0507 EMU 0.0057 (1.45) 0.4988 (6.15) 0,2163 0,0520 World 0.0050 (1.21) 0.4773 (5.10) 0,1644 0,0537 6. ING Group Dutch 0,0004 (0.07) 1,7617 (11.26) 0,623 0,0776 EMU 0.0015 (0.27) 1.8847 (12.01) 0,6664 0,073 World -0.0014 (-0.21) 1.8593 (10.41) 0,5385 0,0859 7. Philips Dutch 0.0000 (0.01) 1.3897 (14.10) 0,6163 0,0621 EMU 0.0011 (0.24) 1.3859 (13.51) 0,5728 0,0655 World -0.0007 (-0.01) 1.1667 (8.60) 0,337 0,0816 8. Reed Elsevier Dutch 0.0051 (1.45) 0.6081 (9.46) 0,3427 0,0477 EMU 0.0057 (1.53) 0.5494 (7.75) 0,2614 0,0506 World 0.0051 (1.28) 0.5090 (5.55) 0,1863 0,0531 9. Royal Dutch Shell Dutch 0.0026 (0.74) 0.6231 (7.48) 0,3611 0,0469 EMU 0.0032 (0.87) 0.5823 (6.47) 0,2946 0,0493 World 0.0022 (0.56) 0.6132 (5.63) 0,2713 0,0501 10. Unilever Dutch 0.0061 (1.60) 0.5308 (6.65) 0,2639 0,0502 EMU 0.0068 (1.67) 0.4323 (4.91) 0,1636 0,0535 World 0.0063 (1.49) 0.3896 (4.08) 0,1103 0,0552
AVG RMSE Dutch 0,0657 AVG R2 Dutch 0,4351
EMU 0,0677 EMU 0,3878 World 0,0747 World 0,2871 Table 5.7 Regression results the Netherlands period 1
From table 5.7 it follows that all the alpha estimates are not significantly different from zero and all the beta estimates are significantly different from zero. This implies that all the CAPMs are valid. From table 5.7 it follows that the Dutch CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the Dutch market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the Dutch CAPM.
Table 5.8 contains some of the regression results of the ten Dutch firms in time period 2.
Firms Index Intercept, α Slope, β R2 RMSE
1. Aegon Dutch -0,0061 (-0,71) 1,7575 (10,73) 0,6172 0,0793 EMU -0,0079 (-0,85) 1,7894 (8,24) 0,5572 0,0853 World -0,0091 (-0,82) 1,9500 (5,84) 0,3758 0,1013 2. Ahold Dutch -0,0034 (-0,26) 1,3561 (5,31) 0,2998 0,1188 EMU -0,0047 (-0,34) 1,3149 (4,94) 0,2454 0,1233 World -0,0059 (-0,42) 1,5627 (3,68) 0,1969 0,1272 3. Akzo Nobel Dutch 0,0029 (0,48) 1,0136 (9,17) 0,5244 0,0553 EMU 0,0021 (0,31) 0,9236 (6,03) 0,3792 0,0632 World 0,0013 (0,17) 1,0827 (5,77) 0,296 0,0673 4. ASML Dutch 0,0035 (0,26) 2,1497 (6,86) 0,4996 0,1233 EMU 0,0004 (0,04) 2,5630 (10,43) 0,6184 0,1077
World -0,0010 (-0,07) 2,6849 (6,11) 0,3855 0,1367 5. Heineken Dutch 0,0042 (0,71) 0,2350 (2,46) 0,0606 0,0531 EMU 0,0040 (0,68) 0,2019 (2,15) 0,0389 0,0537 World 0,0043 (0,71) 0,0462 (0,30) 0,0012 0,0547 6. ING Group Dutch 0,0070 (1,12) 1,3998 (9,35) 0,6643 0,0571 EMU 0,0055 (0,83) 1,4336 (7,89) 0,6067 0,0618 World 0,0048 (0,56) 1,4452 (4,98) 0,3502 0,0794 7. Philips Dutch 0,0028 (0,35) 1,5836 (8,29) 0,5964 0,0747 EMU 0,0007 (0,10) 1,8529 (12,54) 0,711 0,0632 World 0,0002 (0,02) 1,7442 (6,05) 0,3579 0,0942 8. Reed Elsevier Dutch 0,0039 (0,68) 0,6024 (7,42) 0,2969 0,0531 EMU 0,0036 (0,57) 0,5138 (4,76) 0,1881 0,0571 World 0,0029 (0,46) 0,6894 (4,74) 0,1923 0,0570 9. Royal Dutch Shell Dutch 0,0027 (0,53) 0,6840 (6,68) 0,426 0,0455 EMU 0,0022 (0,39) 0,5841 (4,05) 0,2705 0,0513 World 0,0015 (0,26) 0,7527 (4,25) 0,2551 0,0519 10. Unilever Dutch 0,0053 (0,83) 0,4912 (4,29) 0,1925 0,0577 EMU 0,0050 (0,74) 0,3789 (2,67) 0,0998 0,0609 World 0,0050 (0,70) 0,3332 (1,61) 0,0438 0,0628
AVG RMSE Dutch 0,0718 AVG R2 Dutch 0,4178
EMU 0,0728 EMU 0,3715 World 0,0832 World 0,2455 Table 5.8 Regression results the Netherlands period 2
From table 5.8 it follows that all the alpha estimates are not significantly different from zero. However, the beta estimates are significantly different from zero except for the world CAPM regressions of Heineken and Unilever. This implies that all the CAPMs are valid apart from the latter exceptions. From table 5.8 it follows that the Dutch CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that, based on this time period, the Dutch market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the Dutch CAPM.
Table 5.9 contains some of the regression results of the ten Dutch firms in time period 3.
Firms Index Intercept, α Slope, β R2 RMSE
1. Aegon Dutch -0,0069 (-0,86) 1,6609 (7,91) 0,6115 0,0749 EMU -0,0032 (-0,42) 1,7080 (8,14) 0,6424 0,0718 World -0,0071 (-0,85) 1,6343 (7,28) 0,6074 0,0753 2. Ahold Dutch 0,0077 (1,66) 0,4906 (4,50) 0,2671 0,0460 EMU 0,0092 (1,84) 0,3768 (3,68) 0,1565 0,0493 World 0,0090 (1,69) 0,2574 (2,56) 0,0754 0,0516 3. Akzo Nobel Dutch 0,0023 (0,42) 1,2480 (9,46) 0,632 0,0539 EMU 0,0055 (0,90) 1,1463 (9,89) 0,5297 0,0609 World 0,0032 (0,47) 1,0570 (7,80) 0,4651 0,0649 4. ASML Dutch 0,0140 (2,02) 1,1181 (6,05) 0,4963 0,0637
EMU 0,0170 (2,33) 1,0068 (6,08) 0,3998 0,0695 World 0,0151 (1,89) 0,8948 (4,52) 0,326 0,0737 5. Heineken Dutch 0,0051 (1,07) 0,7764 (7,68) 0,4946 0,0444 EMU 0,0070 (1,43) 0,7287 (6,71) 0,4329 0,0470 World 0,0056 (1,04) 0,6598 (5,83) 0,3665 0,0497 6. ING Group Dutch -0,0066 (-0,76) 2,0892 (9,54) 0,6426 0,0881 EMU -0,0022 (-0,29) 2,2357 (10,94) 0,731 0,0764 World -0,0068 (-0,75) 2,0402 (9,74) 0,6287 0,0898 7. Philips Dutch -0,0017 (-0,36) 1,2216 (14,57) 0,6803 0,0474 EMU 0,0018 (0,29) 1,0235 (9,01) 0,4744 0,0607 World -0,0002 (-0,03) 0,9216 (6,91) 0,3972 0,0650 8. Reed Elsevier Dutch 0,0061 (1,38) 0,6118 (6,15) 0,3933 0,0430 EMU 0,0076 (1,67) 0,5765 (6,12) 0,3469 0,0446 World 0,0070 (1,38) 0,4306 (3,89) 0,1999 0,0494 9. Royal Dutch Shell Dutch 0,0028 (0,55) 0,5694 (4,63) 0,3066 0,0484 EMU 0,0041 (0,82) 0,5806 (5,06) 0,3168 0,0481 World 0,0028 (0,54) 0,5535 (4,24) 0,2973 0,0487 10. Unilever Dutch 0,0067 (1,46) 0,5649 (4,97) 0,353 0,0433 EMU 0,0083 (1,73) 0,4733 (4,18) 0,2461 0,0467 World 0,0074 (1,49) 0,4125 (3,93) 0,1931 0,0483
AVG RMSE Dutch 0,0553 AVG R2 Dutch 0,4877
EMU 0,0575 EMU 0,4277 World 0,0616 World 0,3557 Table 5.9 Regression results the Netherlands period 3
From table 5.9 it follows that all the alpha estimates are not significantly different from zero except for the ASML Dutch CAPM and EMU CAPM regressions. However, all the beta estimates are significantly different from zero. This implies that all the CAPMs are valid apart from the exceptions. From table 5.9 it follows that the Dutch CAPM has the lowest average root-mean-square error and the highest average coefficient of determination.
Therefore, I can conclude that, based on this time period, the Dutch market is not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the Dutch CAPM.
5.2 Summary & Discussion
The following tables contain summaries of the measures of interest, i.e. the average coefficient of determination and the average root-mean-square error.
(a) (b)
(c)
Tables 5.10 (a, b, c) Summary regression results France, Germany & the Netherlands From these tables it follows that among all three countries, and across every time period, the domestic CAPM has the lowest average root-mean-square error and the highest average
Germany
AVG RMSE Germany EMU World
2000 - 2014 0,0736 0,0757 0,0818 2000 - 2006 0,0751 0,0773 0,0874 2007 - 2014 0,0659 0,0677 0,0701
AVG R2 Germany EMU World
2000 - 2014 0,4335 0,4018 0,3047 2000 - 2006 0,4450 0,4200 0,2745 2007 - 2014 0,4612 0,4263 0,3858
France
AVG RMSE France EMU World
2000 - 2014 0,0611 0,0614 0,0650 2000 - 2006 0,0636 0,0639 0,0685 2007 - 2014 0,0552 0,0553 0,0587
AVG R2 France EMU World
2000 - 2014 0,3575 0,3439 0,2832 2000 - 2006 0,2883 0,2720 0,1887 2007 - 2014 0,4325 0,4210 0,3608
The Netherlands
AVG RMSE Dutch EMU World
2000 - 2014 0,0657 0,0677 0,0747 2000 - 2006 0,0718 0,0728 0,0832 2007 - 2014 0,0553 0,0575 0,0616
AVG R2 Dutch EMU World
2000 - 2014 0,4351 0,3878 0,2871 2000 - 2006 0,4178 0,3715 0,2455 2007 - 2014 0,4877 0,4276 0,3557
coefficient of determination. Therefore, I can conclude that the Dutch, French and German market are not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than their domestic CAPM. I can also conclude that the European subprime mortgage crisis had no impact on this outcome. Since the crisis neither the EMU CAPM nor the world CAPM was able to provide better estimates of the cost of equity than the domestic CAPM in any of the three countries.
However, from these tables it does follow that there is a significant difference in performance between time period 2 (2000-2006) and time period 3 (2007 – 2014), measured by the
measures of interest. The distinction between the time periods 2 and 3 was based on the commencement of the European subprime mortgage crisis, which basically kicked of in 2007. From table 5.10(a) it becomes clear that all three CAPMs performed better during time period 3 (2007 – 2014) when compared to time period 2 (2000 – 2006), based on the measures of interest. From table 5.10(a) it follows that the average coefficient of determination increased while the average root-mean-square error decreased in time period 3 (2007 – 2014) relative to time period 2 (2000 – 2006). This does not only hold for France, as follows from tables 5.10(b)(c). From these tables it follows that for Germany as well as for the Netherlands all three CAPMs performed better in time period 3 (2007 – 2014) relative to time period 2 (2000 – 2006), i.e. the average coefficient of determination increased while the average root-mean-square error decreased in time period 3 (2007 – 2014) relative to time period 2 (2000 – 2006). Among all three countries the average coefficient of determination increased while the
average root-mean-square error decreased since the commencement of the European subprime mortgage crisis. Therefore, I can conclude that the European subprime mortgage crisis had a positive effect on the performance of all CAPMs in estimating of the cost of equity, which improved.
Although these are remarkable conclusions one has to be careful with interpreting the results. There are some problems with this study, e.g. not all the return on a stock is explained by the return on the market, and the results of this study. The problems can be divided into two categories: the CAPM and the data.
One issue that could drive the results might be the CAPM itself. First of all, the CAPM is a single factor model that is heavily debated among scientist, e.g. Jagannathan & McGratten (1995) and Bartholdy & Peare (2004). One of the problems with a single factor model, like CAPM, is that it possibly exhibits omitted variable bias. The return on the benchmark market portfolio is not the only variable affecting the return on a stock, as implied by CAPM. Therefore, the results could be influenced by omitted variable bias. Secondly, underlying the CAPM are three assumptions, as stated in the theoretical framework. These simplifying assumptions could also drive the results. In reality it is unlikely that investors do not face taxes or transaction costs. The other assumptions - investors can borrow and lend at the risk-free interest rate, investors only hold efficient portfolios and investors have homogeneous expectations – are also unlikely to hold in reality. Therefore, the results could be influenced by the assumptions.
Another issue that could drive the results might be the data. There could be multiple problems with this issue. First of all, the time period (2000 – 2014) might be inappropriate to test the CAPM and influence the results. This could be due to the crisis that emerged at which time the CAPM does not hold. It also could be due to the length of the time period, e.g. in the study of Jagannathan & McGratten (1995) a longer time period, 66 years, is used to evaluate the CAPM. Furthermore, I used monthly data as suggested by Akdogan (1992, 1996), Barari (2004) and Harris et al. (2003). However, a daily or weekly basis for the data could have been more appropriate to use. The results could be influenced by these choices on different aspects of the time. Secondly, the results could be influenced due to a sample selection error, i.e. the assignment of the thirty firms for which by chance CAPM is invalid. Finally, the results could be influenced due to the sample size error, i.e. a sample size of thirty firms is not considered to be a size to perform tests that achieve accurate results.
In my study the domestic CAPM provides better estimates of the cost of equity than the EMU CAPM and world CAPM throughout all three of my time periods (2000 – 2014), which is in line with the study of Harris et al. (2003). From the study of Harris et al. (2003) it follows that for a sample of S&P500 firms the ex ante expected return estimates show a better overall fit with the domestic CAPM than with the global CAPM through their time period (1983 – 1998). Furthermore, although the Dutch, French and German markets are not integrated enough for the EMU CAPM or world CAPM to provide better estimates of the cost of equity than the domestic CAPM, one could say that the markets exhibit increasing regional as well as global integration throughout the 21st century (2000 – 2014). This is based on the
improved performance of the EMU and world CAPMs, i.e. the coefficient of determination increased while the root-mean-square error decreased, over time. Since the commencement of the crisis the EMU CAPM and world CAPM were more accurate, which could be due to domestic markets that became more integrated into the regional and world market. This result would be in line with Akdogan (1992, 1996) and Barari (2004), but needs further
investigation, since the improved performance of the EMU CAPM and the world CAPM could be due to other reasons than integration.
CHAPTER 6 Conclusion
This study focused on the appropriate choice of a benchmark market portfolio when using CAPM by comparing three benchmark market portfolios. If markets are integrated enough either the regional CAPM or world CAPM should be able to produce better estimates of the cost of equity than the domestic CAPM. From the regressions results it follows that among all three countries, and across every time period, the domestic CAPM has the lowest average root-mean-square error and the highest average coefficient of determination. Therefore, I can conclude that the Dutch, French and German markets are not integrated enough for either the EMU CAPM or world CAPM to provide better estimates of the cost of equity than their domestic CAPM. I can also conclude that the European subprime mortgage crisis had no impact on this outcome, as before and since the commencement of the crisis neither the EMU CAPM nor the world CAPM was able to provide better estimates than the domestic CAPM in any of the three countries. However, since the commencement of the European subprime mortgage crisis, among all three countries, the average coefficient of determination increased while the average root-mean-square error decreased. Therefore, I can conclude that the European subprime mortgage crisis did have a positive effect on the overall performance of all the CAPM, measured by these measures. One has to be careful with interpreting the results from this study, due to the CAPM (omitted variable bias, simplifying assumptions), the data (crisis, time span, frequency data, sample selection error, sample size error) and the methodology (risk-free interest rate, performance measures) that could influence the results and conclusion of this study.
REFERENCES
Akdogan, H. (1992). Behavior of systematic risk in a regionally integrated model for stock prices. Elsevier, 39 (2), 213-216.
Akdogan, H. (1996). A Suggested Approach to Country Selection in International Portfolio Diversification. Journal of Portfolio Management, 23 (1), 33-39.
Barari, M. (2004). Equity market integration in Latin America: A time-varying integration score analysis. International Review of Financial Analyses, 13 (5), 649-668.
Bartholdy, J., Peare, P. (2003). Unbiased estimation of expected returns using CAPM.
International Review of Financial Analyses, 12 (1), 69-81.
Berk, J., & DeMarzo, P. (2014). Capital markets and the pricing of risk. Optimal portfolio choice and the capital asset pricing model. Estimating the cost of capital. In J. Berk & P. DeMarzo, Corporate Finance (3rd edition, pp. 312-436). Essex, Harlow, England: Pearson.
Bruner, R.F., Eades, K.M., Harris, R.S., Higgins, R.C. (1998). Best Practices in Estimating the Cost of Capital: Survey & Synthesis. Financial Practice & Education, 8, 13-28 Fama, E.F., French, K.R. (2004). The Capital Asset Pricing Model: Theory and Evindence.
Journal of Economic Perspectives, 18 (3), 25-46.
Harris, R.S., Marston, F.C., Mishra, D.R., O’Brien, T.J. (2003). Ex Ante Cost of Equity Estimates of S&P500 Firms: The Choice Between Global and Domestic CAPM. Financial
Management, 32 (3), 51-66.
Jagannathan, R., & McGrattan, E.R. (1995). The CAPM debate. Federal Reserve Bank of
Minneapolis Quarterly Review, 19 (4), 2-17.
Morgan Stanley Capital International. (2015). MSCI Netherlands IMI. Consulted on 16 may 2015, at
https://www.msci.com/resources/factsheets/index_ fact_sheet/msci-netherlands- imi.pdf Morgan Stanley Capital International. (2015). MSCI Germany Index. Consulted on 16 may 2015, at
https://www.msci.com/resources/factsheets/index_fact_sheet/msci-germany- index- net.pdf Morgan Stanley Capital International. (2015). MSCI France Index. Consulted on 16 may 2015, at
https://www.msci.com/resources/factsheets/index_fact_sheet/msci- france- index-eur-net.pdf Morgan Stanley Capital International. (2015). MSCI EMU Index. Consulted on 16 may 2015, at
Morgan Stanley Capital International. (2015). MSCI World Index. Consulted on 16 may 2015, at
