The influence of macroeconomic effects
on the Dutch stock market
BSc Finance and Organization Student: Coen van den Berg Studentnumber: 6068669 Field: Finance
Supervisor: Dr. J. Ligterink University of Amsterdam
Faculty of Economics and Business 3rdof February 2014
Abstract
This paper examines the macroeconomic effects on the Dutch stock market, i.e. the AEX index. A clear understanding of stock market determinants is crucial for investors, regulators and academics. The present paper analyzes relationships between a group of macroeconomic variables and the Dutch stock market index. The research reveals that some macroeconomic variables (U.S. GDP Growth, MSCI World Index and the Industrial Production) lead the Dutch stock market returns. This paper gives more insight in the way macroeconomic forces act on the stock market return during the monthly period from January 2002 till January 2012.
Table of Contents
INTRODUCTION ... 3
I.LITERATURE REVIEW ... 6
II.DATA METHODOLOGY ...10
A. Macroeconomic Factors…...10
B. Macroeconomic Factors including the MSCI World Index...12
III.RESEARCH RESULTS ...13
A. Macroeconomic Factors...13
B. Macroeconomic Factors including the MSCI World Index...14
IV.CONCLUSION AND DISCUSSION... 17
REFERENCES ... 19
TABLES... 20
As a stock analyst it is important to know to what extent macroeconomic factors influence stock prices of the twenty-five Dutch companies related to the Amsterdam Exchange Index the most. Since the 3rd of January 1983 the Amsterdam Exchange Index is composed of 25 of the most actively traded stocks. Each company operates globally and is affected by macroeconomic influences outside the Netherlands. This creates the starting point of examining the relationship between the Dutch stock market and macroeconomic factors. The rationale behind this is to find out which macroeconomic factors influence the AEX significantly. Thus, prompted forecasts can be made for the volatility in stocks and for stock markets. This research investigates the influence of macroeconomic factors on the monthly returns of the Dutch stock market and the underlying companies of it.
According to Kandir (2008), the first research in which specific macroeconomic variables were used as proxies for undefined variables was conducted by Chen, Roll and Ross (1986). They examined whether it is possible to express the equity returns as a function of macroeconomic variables. This was called the macroeconomic Factor Model. This model exists of the influence of economic forces such as interest rate on the expected dividends and discount rate. In turn, this influences the stock prices through the stock returns. The discount rate changes with the level of interest rates, term-structure and risk premium.
Basically not only interest rate influences the stock returns, but also macroeconomic factors such as the inflation rate, GDP growth, exchange rates, consumption rate, industrial production and oil price. Many researchers, such as Fama (1990) and Asprem(1989) among others, have tried to examine the relationship between the macroeconomic factors and the stock market of specific countries. Fama(1990) studied the US market from 1953 to 1987 and concluded that the stock market return is strongly correlated with the production growth rates. Other studies from Barro (1990) and Schwert (1990) have also found that there is a correlation between the U.S. stock market returns and its aggregate real economic activity.
Asprem (1989) discovered in his study about the European stock market returns that stock price changes are positively correlated to some measure of real economic activity. Gjerde and Seattem (1999) studied the Norwegian market and found that stock prices react negatively to changes in the interest rate.
Some empirical research of the correlation between macroeconomic factors and the stock market return of the Japanese stock market examined by Hamao (1988) supported the findings of Chen et al. (1986) who were the first to examine this relation. At the other hand the research of Poon and Taylor (1991), who studied the stock market returns in the United
Kingdom, did not provide enough support for the theory of Chen, Roll and Ross (1986). Based on that, empirically, the results of the studies that followed are ambiguous as suggested by Arnold and Vrugt (2006). Essentially this means that there is still need to investigate this relationship.
Because the Netherlands is a small country with an open economy surrounded by countries such as Great Britain and Germany, it is economically highly dependent of these countries. Additionally, the United States has an important role in the Dutch economy. To more extensively investigates these effects on the Amsterdam Stock Exchange, there is composed a comprehensive research which contains domestic and foreign variables which are relevant for the Dutch stock market. This research captures the United States as a benchmark by taking into consideration the economic growth of the US in terms of the USGDP growth rate. The choice of this U.S. macroeconomic variable is due to the fact that the United States trades mostly with the Netherlands. The foreign direct investment of Americans in the Netherlands is the largest destination, specifically 521.4 billion dollars. On the contrary the Netherlands is the 3rdlargest foreign direct investment partner in the United States. Moreover, this relationship has created 625000 jobs and has potential to grow further1. As a result the assumption is made that a United States macroeconomic variable has relatively the most effect on the Dutch stock return per month.
The research question is: what is the influence of macroeconomic forces on the Dutch stock market prices of the AEX weighted average index during the period 2002-2012? This empirical research formulates an answer to this question by focusing on the domestic macroeconomic factors of the Netherlands, an US macroeconomic variable and the MSCI World Index.
The macroeconomic factor of the US that will be discussed in this paper is, as mentioned earlier, the United States GDP growth. The GDP, in comparison with the GNP, is chosen primarily because it only takes in account the products and services that are produced within the territorial boundaries of the United States. The Dutch macroeconomic factors are the Inflation Rate, Industrial Production and the Term Spread. Also the effect of the Oil Price and the Exchange Rate (€/$) is examined. Unlike the Oil Price does not affect the United States or the Netherlands individually, it serves as a global factor because it is estimated as a significant factor. Moreover, the MSCI world index is added to the research because it includes 1610 constituents of 23 Developed Market countries and is therefore a
1 http://www.minbuza.nl/en/news/2011/11/economic-ties-between-the-usa-and-the-netherlands-a-partnership-that-works.html
miscellaneous estimator capturing worldwide volatility. Finally, the OLS-regression method determines which variables will have a significant influence on the monthly stock returns of the AEX. All data is retrieved from Thomson Reuters DataStream.
To answer the research question stated above, the paper is structured as follows. The next section focuses on the literature review that provides the theoretical basis for the research question. Also this section discusses the previously published findings on this field. Besides, this section outlines why specific variables have been selected in the papers and highlight their significance. In order to gain an understanding of how the research is constructed section II provides how the data is retrieved and which research method is going to be used. Section IIIexamines the descriptive statistics of the data and the analysis of the imposed regressions illustrated by tables and graphs. Finally, section
IV functions as the summary and conclusion to this thesis and revisits the results of the regression analysis, aiming to answer the research question. Besides, this section seeks to facilitate future research and suggestions on this topic.
I. Literature Review
This section serves as a literature review and starts with background information. Subsequently, this section discusses the relevant papers regarding the macroeconomic factors in this research. Third, this section describes and discusses the significance of global factors at specific markets to determine the stock returns. Finally, the last part examines papers of several authors who apply macroeconomic forces to cross-border research.
Asset pricing endures when new information enters the financial stock markets according to Fama (1970) in case of market efficiency. Stock prices react sensitive to information and this involves the stock market. As a result asset prices adjust with changes in the market. To determine the return of an equity after adjustments in time there have to be applied a financial theory or a fundamental framework. Macroeconomic factors belong to a substance of systematic risk which involves the whole market. In the literature a lot of authors have been researching to explain and correctively estimate the impact of macroeconomic forces on the returns of securities. Initially, the CAPM theory was established in 1974 to estimate the expected return of an individual asset by using the beta as measurement for volatility. Two years later, in 1976, Stephen A. Ross built a new model, called the Arbitrage Pricing Theory (APT). Papers which examined the influence of macroeconomic factors conducted the principle of the APT in the research to derive the stock returns. The issue contains to examine macroeconomic factors that are most important to explain the returns of the stock markets. This, in turn, depends on the research method conducted and on the particular stock market(s) investigated.
Chen, Roll and Ross (1986) examined the U.S. stock market and used seven macroeconomic state variables which measured the relationship between asset prices and real economic activities. By constructing the pricing formula of an individual asset as showed beneath. ‘P’ stands for the asset price, ‘E(c)’ stands for the expected dividend stream which is considered constant and ‘k’ corresponds to the discount rate. The factor k implicates whether there is more demand to bonds than to stocks, since the discount rate is explained by the interest rate. The assumption contains that the stock price is mature and it is a zero-growth stock.
(1)
Thus, the asset price can be determined by dividing the expected cash flow stream through the discount rate which is actually the measurement of present risk in the market.
The real economic activities, which are used as proxies of the macroeconomic factors in this research, contain interest rate, inflation, production, productivity, yield spread, industrial production and so forth. By conducting a time series regression of all these variables the validity of the returns can be derived and this is estimated according the Arbitrage Pricing Theory. The aim is to look for economic variables such as industrial production, real interest rate and consumption which are correlated with stock returns. U.S. data are used from the period 1958 to 1984. The research attempts to address the economic interpretation of factors. This is left unsatisfied in the macroeconomic factors analytic approach, because of the not considered correlation. By testing the validity of the Arbitrage Pricing Theory there is concluded that macroeconomic variables are causally related to share returns. Since Chen et al. (1986) many papers in this field have relied upon this evidence in accordance with the APT.
Hamao (1988) examined the Japanese stock market with an empirical investigation regarding the Arbitrage Pricing Theory using Japanese macroeconomic factors. Factors examined include industrial production, inflation, investor confidence, interest rate, foreign exchange and oil price. All latter factors seem to have strong evidence, except for the Japanese monthly production. In addition, Hamao (1988) tests the validity of the CAPM beta. The result shows that the CAPM beta does not capture any extra risk that may have been missed by the macroeconomic factors.
Otherwise, the paper of Rubio (1989) used Spanish data and found no evidence for a pricing relationship between market returns and macroeconomic variables. Also the empirical paper of Poon and Taylor (1991) contradicts with the results of Chen et al. (1986). The study examined the influence of the macroeconomic variables on the United Kingdom stock market returns during the period 1958 to 1984. There is used the same method as the study of Chen, Roll and Ross (1986) but is in turn applied to the UK stock market. The results show that macroeconomic variables do not appear to affect share returns in the United Kingdom as derived in the U.S. according to Chen, Roll and Ross (1986). The authors use a time series ARIMA model to test the data and use the residuals from the model as innovations. Poon and Taylor (1991) suggest that either different macroeconomic factors have an influence on share returns in the United Kingdom or the methodology employed by Chen, Roll and Ross (1986) is inefficient and indicates a spurious regression. This could be because Chen, Roll and Ross (1986) make use of the expected and the unexpected component of share returns and macroeconomic variables. Whereas Poon and Taylor (1991) argues that only the unexpected component should be taken into account.
Gjelte and Saettem (1999) examined Norway which is a small open economy and used the findings from previous research in major economies as a base for comparison. The U.S. and Japanese findings about the interest rate have hold in this study, such that the influence of the inflation and the oil price as well accurately responds to the stock returns. They utilized the VAR approach to Norwegian data. The multivariate vector autoregression modeling technique is a useful alternative to the conventional structural modeling procedure. VAR-analysis works with unrestricted reduced forms, treating all variables as potentially endogenous. This study demonstrates that significant results from major economies can be applied to smaller, open economies and are therefore valid. The suggestion from the research is that the VAR-method works for other countries as well. The multivariate vector autoregression modeling technique is a useful alternative to the conventional structural modeling procedure. VAR-analysis works with unrestricted reduced forms, treating all variables as potentially endogenous.
Kwon and Shin (1999) examined whether real economic activities determine Korean stock market returns by the response of stock prices to macroeconomic fluctuations. Stock prices were cointegrated with a set of macroeconomic variables. That was composed of the foreign exchange rates, trade balance, production level and money supply. Hamilton and Ling (1996) carried out a unique research between the stock market volatility and the business cycle. They concluded that 60% of the variance of the stock market returns is clarified by economic recessions.
During the 1990s Kaneko and Lee (1995) have re-examined the U.S. market and the Japanese market individually. In contrast to Hamao (1988), they found that global factors have become relevant. That meant changes in oil prices and exchange rate were significant in Japanese stock returns. For the U.S. market economic news about risk premiums, term premiums, and the growth rate in industrial production was most significant.
The global factors serve to gain a better understanding of the stock returns in a specific country. The significance of global factors or factors from other countries is evidenced in the research of Cheung and Ng (1998). The research covered the quarterly stock indices and macroeconomic data of five countries. By performing a cointegration relationship analysis of each country’s stock index and aggregate economic variables as proxies for expected returns it followed the same research methodology as the paper of Fama (1990). To improve the results of the cointegration analysis an Error Correction Model (ECM) was applied and this resulted in more explanatory power of stock returns. Cheung and Ng (1998)
suggested to examine whether foreign macroeconomic factors do influence the domestic stock returns. This suggestion has been the cause to include the macroeconomic factors, domestic and foreign, in this research.
The research of macroeconomic factors extended over time to cross-border research. Asprem (1989) researched the effect of stock price changes according to macroeconomic variables among ten European countries including the Netherlands. The result was that the relationship between stock prices and macroeconomic variables was largest in Germany, the Netherlands, Great Britain and Switzerland. Besides, there was concluded in the paper that American macroeconomic variables such as the yield spread influence European stock markets and there was a high correlation between S&P 400’s returns and the European indices.
Rapach, Wohar and Rangvind (2005) described the explanatory power of macroeconomic factors in predicting stock returns. Their sample was composed of twelve industrialized countries. Interest rate had the strongest predictive ability among all factors according to their paper. Although inflation found to be significant in the Netherlands and the U.S. particularly. The unemployment rate and the industrial production had nearly no predictive ability, whereas the term spread had little predictive ability. They considered both in-sample and out-of-sample tests.
Most of the studies in this field have focused on the United States as benchmark for developing the macroeconomic variables, but abovementioned papers highlight either other markets than the United States. The papers create the theoretical benchmark for researching the Dutch stock market. There has not been a similar research particularly for the Dutch stock market. In addition, some papers contain an opposite effect concerning the application of macroeconomic variables. This indicates a contradiction among results. To get a clear overview of which macroeconomic variables to pick for the Dutch stock market, this is further discussed in the next section. This research addresses the previous papers to examine whether the resulting factors contribute to the Dutch stock market. Since the Netherlands is a small, open economy the expectation is that global factors affect the Dutch stock market. According to the research of Asprem (1989) an U.S. macroeconomic variable is decisive in the analysis of the expected return on stock markets.
II. Data Methodology
This research investigates the monthly stock returns of the AEX index on the Dutch stock market. This section starts describing the methodology of the macroeconomic factors that determine the monthly stock return. Secondly, the methodology regarding the inclusion of the MSCI World Index is mentioned. Finally, a short conclusion of the two methodologies is given.
Previous studies contained only time series (stationarity) in the way of a Vector Autoregressive Regression (VAR) or an ARIMA model because of the that is the normal empirical procedure to test the APT. Nonetheless, the APT testing procedure has many problems according to Virtanen and Yli-Olli (1992). There are different interpretations in the procedure of time series, so there is in fact no consistency regarding the time series methodology which results in different outcomes.
However, the effect of an OLS-regression has never been exploit on stock returns. Hence, the effect is unknown of what OLS does according to the influence of macroeconomic factors. Macroeconomic variables are used as a proxy for the macroeconomic factors to determine the monthly stock returns of the AEX index. The following macroeconomic variables are considered in this research: Oil Price, Exchange Rate, Industrial Production, Inflation Rate, Term Spread, USGDP Growth and the MSCI World Index. Thus, the research induces a multiple regression model.
Appendix A shows all the variables and abbreviations used in this research. There is a constant added in the model since having a starting point is necessary to execute the regression and making a descriptive statistical analysis. The regression contains either k=6 and k=7 explanatory variables, dependent whether the MSCI World Index is included. The epsilon stands for the noise term in the regression which contains the residuals of the model. A. Macroeconomic factors
In order to create an answer to the research question the following regression equation is used:
AEXi= 0+ 1∗ i+ 2∗ i+ 3∗ i+ 4∗ i+ 5∗ i+ 6∗ i
RTAEXiis the dependent variable and yields the return on the AEX per month i. To determine
this dependent variable DataStream is used to obtain the data with the values for the AEX. There are different ways to calculate the AEX returns. The AEX returns have been calculated through taking the difference of the index numbers at the end of each month divided by the initial index. The independent variables are described as follows.
EX represents the exchange rate in EUR/USD (€/$) per month. The exchange rate is an important factor to take into account for countries which have a lot of import and export. Especially for the Netherlands this is therefore a considerable factor to include. The exchange rate is derived from The WM Company based on data provided by Reuters for the closing spot rate.
OP stands for the change in oil price per month. The oil price has an important role on the supply side of the economy and is part of the producer price index. Because the Netherlands is a small open economy the expectation is that the oil price has a substantial effect on the returns of the AEX. The Producer Price Index of Crude Petroleum is used as a proxy for the oil price. This indicator measures the average changes in prices received by producers of the crude-oil for their output. The data is retrieved from DataStream originating from the Bureau of Labor Statistics in the United States and is generated on a monthly basis.
IF represents the change in the inflation rate per month. The consumer price index gauges the purchasing power every month. This macroeconomic variable measures the consumption in a country and is therefore a good indicator to the stock returns on the AEX. The data is obtained from the Central Agency of Statistics Netherlands (CBS) through DataStream.
IP stands for the change in the industrial production per month. This macroeconomic variable gives insight on the industrial production in the Netherlands per month. The data is obtained from DataStream and is given on a monthly basis. The data originates from the Central Agency of Statistics Netherlands (CBS).
TS yields the term spread per month. A long-term government bond in the Netherlands takes ten years to expire. In case of the short-term government bond there is used the short-term Euribor rate on a 3-month term as a proxy. The Euribor 3-month rate is used since the Netherlands belong to the European Monetary Union and the Netherlands are aligned on this rate due to the euro. Every month the rates adjust so the difference is derived every month to obtain the term spread of the Netherlands. The Euribor short-term rate is obtained from the OECD, whereas the long-term government bond rates are derived from the Dutch Central Bank (DNB).
USGDP relies on the index figure of the U.S. Gross Domestic Product Growth per month. This variable serves as a proxy for the influence the U.S. economic activity has on the Dutch stock market. According to the research of Kaneko and Lee (1995) on the Japanese market, the expectation is that foreign macroeconomic variables do influence small open economies, such as the Netherlands. The data from the USGDP is obtained from the Bureau of Labor Statistics in the United States.
B. Macroeconomic factors including the MSCI World Index
The expectation is that the regression model could be more reliable if there is added a seventh variable. In this case the MSCI World Index is chosen. This means that this is an important variable to extend the model because it captures worldwide effects. This variable lowers the probability of omitted variable bias and serves as a robustness check. The MSCI World Index increases the explanatory power of the regression model, which is indicated by a higher R-squared. The regression equation with the MSCI World Index included is:
AEXi= 0+ 1∗ i+ 2∗ i + 3∗ i+ 4∗ i+ 5∗ i+ 6∗ i +
7 ∗ i+ i (3)
MSCI stands for the Morgan Stanley Capital International world equity market and represents 23 developed markets around the world with 1610 constituents. It is a widely used benchmark for equity returns and performance management of hedge funds and mutual funds. The returns are calculated by taking the sequential months per period and divide this number by the initial index. The monthly data is obtained from DataStream.
Concluding, there is one regression model which is employed to serve the macroeconomic factors as mentioned above. By including the MSCI World Index to the model the results are expected to obtain more explanatory power such that this derives better estimates. So it reveals which of the two outputs of the regressions is the best indicator to predict the Dutch stock market returns the best. The next section describes the resulting outputs.
III. Research Results
This section describes and analyses the empirical findings from the OLS-regression method. First, this section examines the regression results regarding equation (2). The second part discusses the regression results with the MSCI World Index included.
A. Macroeconomic factors
This paragraph analyzes the results of the regression equation (2) and examines the underlying determinants of the regression.
In TableI the summary statistics are shown. It represents the mean, median, standard deviation, minimum value and maximum value of the variables. Most striking in this table is the minimum value of the term spread which has a negative value. This means that the short-term rate exceeded the long-term rate in a period. This negative spread was in October 2008 and can be explained by the financial crisis.
The OLS-regression results are presented in Table II. The coefficients are shown for each variable together with the t-value between brackets. To derive whether the variables have a positive or negative impact the sign of the coefficients should be considered. The positive sign on the term spread means that there is a positive relation with the returns on the AEX. The positive effect of the term spread supports the theory of Asprem (1989). The positive sign of the coefficient of the oil price shows that it has a positive relationship with the returns on the AEX. This means that the stock prices of the AEX do benefit from an increase in the oil price. This is in line with the results in the United Kingdom from Poon and Taylor (1991) and Norway from Gjerde and Saettem (1999). The sign of the exchange rate is positive which supports theory. Depreciation of the euro would lead to a higher demand of Dutch products from theory, so the outcome must have shown a positive coefficient.
It can be concluded from Table IIthat USGDP Growth has solely significant effect on the monthly stock return, even at the 1% level, because the p-value is equal to 0.000. In general, by analyzing the p-value it is possible to conclude whether a variable is significant for a given significance level. In this case there are three stars behind the coefficient which indicates significance at the 1% level. From this can be mentioned that the US macroeconomic variable indeed has validity and affects the monthly periods 2002-2012 on the Dutch stock market. The USGDP Growth implicates a positive coefficient. This means that if the USGDP growth is positive, then the monthly returns will go up. So in an economic downturn negative growth will result in lower monthly stock returns.
When analyzing the results of the output there is derived the R² as shown in Table II. The R² measures the explained variation in the model and is also called the goodness-of-fit of the model. Therefore, the R² is an important indicator whether the model has explanatory power. The value always lies between 0 and 1. This regression has on average a low R² of 19.43%, which is also shown in Table II.
The F-statistic is also shown in Table II. The F-value is equal to 4.52. This is significant. This significant F-value indicates that the observed R-squared is reliable, and is not a spurious result of oddities in the data set. So the relationship between the dependent variable and the set of predictors is statistically reliable, and is useful.
Appendix B(2) illustrates the Pearson correlation matrix which shows the correlation of the independent variables. As shown from the matrix there are no high correlations presented among the independent variables. This means that there is no threat to the model due to multicollinearity or any other suspicious high or low correlations which the model can suffer from.
The regression resulted in normal standard errors on first instance when compared to the coefficients. Robust standard errors make sure that there is no chance on heteroskedasticity in the residuals. Heteroskedasticity could obviously lead to biased parameter estimates. The Breusch-Pagan / Cook-Weisberg tests the null hypothesis such that the error variances are all equal versus the alternative hypothesis that the error variances are a multiplicative function of one or more variables. A large chi-square indicates that heteroskedasticity is present. Derived from Appendix C table (2), the test results in a large chi-square which indicates that it suffers from heteroskedasticity. The solution for this problem is to regress with robust standard errors such that the standard errors are more trustworthy and heteroskedasticity cannot affect the standard errors anymore. Therefore, robust standard errors are applied to the regression model.
The next paragraph derives whether the addition of the MSCI World Index could result in a higher goodness-of-fit of the model and therefore more consistent outcomes. B. Macroeconomic factors including the MSCI World Index
This paragraph discusses and describes the results of including the MSCI World Index to the regression.
Table IIIshows the summary statistics of all the variables involved. The table reports the mean, median, standard deviation, the minimum and maximum value per variable. The MSCI values of the returns lie between -18.8% and 14.5% which is close to the values of the
AEX, so it has nearly the same range.
In Table IVthe coefficients of the explanatory variables are mentioned. The t-values are between parentheses. Also the R² and the F-statistic are listed. The macroeconomic factors show a different view of the relationship with the monthly return than when the MSCI is omitted. The impact of the MSCI World Index is substantial.
Table IV shows that the MSCI World Index and the Industrial Production are highly significant variables for all levels of significance on a range of 1%, 5% and 10%. Both variables have a p-value for the MSCI of respectively p = 0.000 and for the IP of p = 0.006. This shows that both explanatory variables have a substantial effect on the monthly returns of the 25 companies examined in the Amsterdam Exchange Index.
Examining Table IV the value of the R² contains 0.7042 which indicates 70.42% of the variance is explained in the model. This means that the explanatory power of the model has increased a lot. It suggests that the addition of the MSCI World Index has a very positive effect on the validity of the model. Moreover, the MSCI World Index is either a highly significant variable of the AEX monthly return and should therefore not be omitted from the regression.
The F-statistic is also shown in Table IV. The F-value is equal to 38.42. This is highly significant. This shows clearly that the observed R-squared is very reliable. So the relationship between the dependent variable and the set of predictors is statistically reliable, and is useful.
The test for heteroskedasticity of the standard errors is done through the Breusch-Pagan / Cook-Weisberg test. This test is executed to derive if there is heteroskedasticity in the standard errors. In Appendix C(3) the test with the results is tabulated. This indicates a lower chi-squared than the critical value and also the p-value of 0.3408 indicates insignificance regarding the test. This means that there is no heteroskedasticity present in the model. So robust standard errors do not have to be used according to this test.
Appendix B(3) gives an overview of the Pearson correlations among the independent variables. As noticeable from the matrix one number is bolded. This is the correlation between the RTMSCI and the RTAEX. This indicates that there is a high positive correlation between these two variables of 0.8240 which is equal to 82.40%. The high correlation is common, because stock markets move rather in the same direction, but there have to be checked for multicollinearity. So because of this high correlation there needs to be accounted for this threat in the model. There are several ways to see whether there is multicollinearity apparent in the model. There is executed a Variance Inflation Factor (VIF) analysis which is
represented by Table 1 in Appendix D. Also high standard errors indicate that there could be multicollinearity. Therefore, the standard errors are represented in Table 2 in Appendix D. The VIF analysis has two ways to prove multicollinearity. This is on one hand through the VIF value itself which as a rule of thumb needs to be at least 10. On the other hand the 1/VIF value also called the tolerance value has to be lower than 0.1 as a rule of thumb. Both values look fine when analyzing Table 1 which indicates no suffering from multicollinearity. Table 2 shows the robust standard errors and this errors seem to be fine either. No high values relative to the coefficients are recognized, so the decision can be made that multicollinearity doesn’t appear in the model.
V. Conclusion and Discussion
This paper has examined whether a set of six macroeconomic variables or a set of the same six macroeconomic variables with the MSCI World Index included do in fact influence the stock market returns. From the Dividend Discount Model, as illustrated by Chen et al. (1986), and the insights from previous empirical research in this field follows that an OLS-regression analysis could be executed. The variables that have been justified by theory are the inflation rate, term spread and the industrial production and the GDP of the United States. Besides, based on the results of Asprem (1989) and Kaneko and Lee (1995) who rely on the addition of international factors such as exchange rate and oil price, there is chosen to include the oil prices and the EUR/USD exchange rate into the regression. The obtained data were monthly data from 1 January 2002 till 1 January 2012.
The OLS-regression has shown that there needs to be a selection of variables based on theoretical outcomes before applying the OLS-regression. It is relatively impossible to obtain all factors that influence asset prices. However, there can be made a reliable OLS model by choosing the most important factors. Thus, this research shows that a regression with OLS is plausible and that there can be obtained significant results from the regression output. The empirical findings underpin the theoretical evidence of the impact of macroeconomic variables on the monthly returns of the Dutch stock market.
The findings of this experiment carry out that the oil price does not support the research of Poon and Taylor (1991) and Gjerde and Saettem (1999), because it is not significant to the stock returns. The same yields for the exchange rate. In contradiction with the research of Kaneko and Lee (1995) who derived that international factors such as the exchange rate and the oil price do have a significant effect. On the other hand the United States macroeconomic variable of the GDP growth has shown significance when the MSCI is omitted and has a positive relationship with the returns of the AEX, as supported by the theory of Asprem (1989). Consequently, it is confirmed that the GDP growth in the United States has relevance in the stock returns of the AEX, and the Netherlands is dependent of the United States for the stock return determination.
Additionally, the MSCI World Index is included in the model. This results that the MSCI World Index and also the Industrial Production have an important effect on the monthly stock returns of the AEX. So the significant effect of the U.S. GDP growth is no longer present.
insights for future research based on the OLS regression analysis. For example the MSCI World Index can be extended in further research as explanatory variable.
The findings of this research can be used as an argument to obtain further information regarding macroeconomic effects on stock markets of small, open economies. This aligns to the research of Gjerde and Saettrom (1999) which basically involves a small open economy. This explores the market characteristics of these economies regarding macroeconomic variables to estimate the effects on stock returns.
Contrarily, further research should also take the limitations into account of the OLS regression. The sample size is relatively small in the regression because of the time period which is only ten years. An increase in sample size decreases the standard errors which will result in a more sustainable outcome. A larger time period for the regression is therefore recommended. Another suggestion is to perform a similar research with European macroeconomic variables from the United Kingdom and Germany instead of a United States macroeconomic variable. In this case the examination of the effect of Europe on the Dutch stock returns could be estimated.
Reference list
Arnold, Ivo J.M. and Vrugt, Evert B., (2006). Stock Market Volatility and Macroeconomic Uncertainty: Evidence from Survey Data, The Nyenrode Research Group (NRG), Working Paper no. 06-08, March 2006.
Asprem, M., (1989). Stock Prices, Asset Portfolios and Macroeconomic Variables in Ten European Countries, Journal of Banking and Finance. Vol. 13 (1989), pp. 589-612.
Barro, R.J. (1990). The Stock Market and Investment, Review of Financial Studies, Vol. 3, pp. 115-131.
Chen N., Roll R. and Ross S. A., (1986). Economic Forces and the Stock Market, The Journal of Business, Vol. 59, No.3, pp. 383-403.
Gjerde O. and Saettem F., (1999), “Causal Relations among Stock Returns and Macroeconomic Variables in a Small, Open Economy”, Journal of International Financial Markets, Institutions and Money, Vol. 9, No. 1, pp. 61-74.
Hamao, Y. (1988), “An Empirical Investigation of the Arbitrage Pricing Theory”, Japan and the World Economy, Vol. 1, pp. 45-61.
Hamilton, J.D. and Lin, G. (1996), Stock Market Volatility and the Business Cycle, Journal of Applied Econometrics, Vol. 11, pp. 573-593.
Kwon, C.S. and Shin, T.S. (1999). Cointegration and Causality Between Macroeconomic Variables and Stock Market Returns, Global Finance Journal, Vol. 10, pp. 71-81.
Poon, S. and Taylor S.J. (1991). Macroeconomic factors and the UK Stock Market, Journal of Finance and Accounting, Vol. 18, pp. 619-636.
Schwert, G.W. (1989). Stock Returns and Real Activity: A Century of Evidence, Journal of Finance, Vol. 45, pp. 1237-1257.
Rapach, D.A., Wohar, M.E. and Rangvid, J. (2005). Macro Variables and International Stock Return Predictability, International Journal of Forecasting, Vol. 21, pp.137-166.
Ross, Stephen A. (1976). The Arbitrage Theory of Capital Asset Pricing, Journal of Economic Theory, Vol. 13, pp. 341-360.
Stock, J.H. and Watson, M.W. (2012). Introduction to Econometrics, 3rd edition, Pearson Education.
Virtanen, I. and Yli-Olli, P. (1992). Some Empirical Tests of the Arbitrage Pricing Theory Using Transformation Analysis. Empirical Economics, Vol. 7, pp. 507-522.
Table ISummary Statistics
This table reports the characteristics of all the variables that are represented in the regression equation (2) between 1 January 2002 – 1 January 2012. Column one up to column five report the mean, median, standard deviation, minimum and maximum value of the variables.
Variable Obs Mean Median Std. Dev. Min Max
(1) (2) (3) (4) (5) RTAEX 121 -.0016469 0.0040766 .0655262 -.2207413 .1190165 EX(€/$) 121 1.276.172 1.29095 .1552876 .8733 159.715 OP 121 .0202862 0.0289242 .1019453 -.3074884 .278415 IP 121 .5187471 -0.0408163 9.040.519 -28 91.8 IF 121 .0014035 0.0012005 .0045609 -.0106735 .0118282 TS 121 1.305.282 1.4359 .9635041 -.8831 28.079 USGDPGrowth 121 .0043907 0.0055342 .0069679 -.0215171 .016895
TableIICoefficients and t-ratios (2)
This table represents the OLS-regression equation (2) results of the monthly returns on the underlying determinants for the period 2002-2012. All variables are defined in Appendix A. T-values are presented between the brackets. *, ** and *** indicate, respectively, statistical significance at the 10%, 5% and 1% levels.
Variables (1) Coefficients (2) OP .092294 (1.54) EX .0627324 (1.70) IP .0002787 (0.44) IF -.1748761 (-0.13) TS 0.00878 (1.39) USGDP 3.125*** F( 6; 114) = 4.52 R² = 0.1943 (3.52)
Table IIISummary Statistics
This table reports the characteristics of all the variables that are represented in regression equation (3) between 1 January 2002 – 1 January 2012. Column one up to column five report the mean, median, standard deviation, minimum and maximum value of the variables.
Variable N Mean Median Std. Dev. Min Max
(1) (2) (3) (4) (5) RTAEX 121 -.0016469 0.0040766 .0655262 -.2207413 .1190165 RTMSCI World 121 .0027765 0.0072699 .0523889 -.1881257 .145277 EX(€/$) 121 1.276172 1.29095 .1552876 .8733 159.715 OP 121 .0202862 0.0289242 .1019453 -.3074884 .278415 IP 121 .5187471 -0.0408163 9.040519 -28 91.8 IF 121 .0014035 0.0012005 .0045609 -.0106735 .0118282 TS 121 1.305282 1.4359 .9635041 -.8831 28.079 USGDPGrowth 121 .0043907 0.0055342 .0069679 -.0215171 .016895
TableIV Coefficients and t-ratios (3)
This table represents the OLS-regression (3) results of the monthly returns on the explanatory variables for the period 2002-2012. All variables are defined in Appendix A. T-values are presented between the brackets. *, ** and *** indicate, respectively, statistical significance at the 10%, 5% and 1% levels.
RTAEX Coefficients RTMSCI 1.0258*** (13.96) OP -.00698 (0.05) EX 0.0078 (0.57) IP 0.00096*** (2.78) IF 0.4363 (0.55) TS -.00091 (-0.23) USGDP .90621 F( 7; 113) = 38.42 R² = 0.7042 (1.61)
Appendix A List of Variables
This table gives an overview and description of the variables that are used in this research. The variables are listed in column one, described in column two, and their source is presented in column three.
Variable(1) Description (2) Sources (3)
RTiAEX Return of the AEX index per month i
Thomson Reuters Datastream
RTiMSCI Monthly return of the MSCI World Index
Thomson Reuters Datastream
OPi Represents the change in the PPI Crude Petroleum per month as a proxy to the oil price Thomson Reuters Datastream
EXi Represents the Exchange Rate in EUR/USD on monthly basis Thomson Reuters Datastream
IPi IP defines the change in industrial production of the Netherlands per month Thomson Reuters Datastream
IFi IF defines the change in inflation rate of the Netherlands per month Thomson Reuters Datastream
TSi Difference between the long-term government bond rate in the Netherlands and the short-term 3-month Euribor rate used as a proxy for the short-term government bond rate Thomson Reuters Datastream
USGDPi US macroeconomic variable that measures the GDP growth per month Thomson Reuters Datastream
β0 Constant variable
Appendix B Correlation Matrices
This table (2) and (3) represents the Pearson correlation matrix of the pairwise correlation. The Pearson correlation coefficient measures the degree of linear relationship between two independent variables. The bold number indicate that there is a high level of correlation which may indicate multicollinearity.
(3) RTAEX RTMSCI EX OP IP IF TS USGDP
RTAEX 1.00 RTMSCI 0.8240 1.00 EX 0.0597 0.0485 1.00 OP 0.2353 0.2649 -0.0104 1.00 IP -0.0263 -0.1907 0.0136 0.0051 1.00 IF 0.0079 -0.0285 -0.0181 0.3331 0.0882 1.00 TS 0.2253 0.2555 -0.2575 0.1925 0.0601 -0.0027 1.00 USGDP 0.3720 0.3941 -0.1654 0.2174 -0.2233 -0.0847 0.3146 1.00 (2) RTAEX EX OP IP IF TS USGDP RTAEX 1.00 EX 0.0597 1.00 OP 0.2353 -0.0104 1.00 IP -0.0263 0.0136 0.0051 1.00 IF 0.0079 -0.0181 0.3331 0.0882 1.00 TS 0.2253 -0.2575 0.1925 0.0601 -0.0027 1.00 USGDP 0.3720 -0.1654 0.2174 -0.2233 -0.0847 0.3146 1.00
Appendix C Tests for Heteroskedasticity
In these two tables the results to test for heteroskedasticity of the error terms are shown. The results are derived from the default Breusch-Pagan / Cook-Weisburg test. The results are represented for equation (2) and (3).
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity (2) Ho: Constant variance
chi2(1) = 7.07 Prob > chi2 = 0.0079
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity (3) Ho: Constant variance
chi2(1) = 0.91 Prob > chi2 = 0.3408
Appendix D Tests for Multicollinearity
Table 1 represents the variance inflation factor analysis. This is a measure to detect multicollinearity. As a rule of thumb the VIF needs to be at least 10 to account for multicollinearity. The tolerance rate is represented by 1/VIF. Also this measure has a rule of thumb that states that a tolerance value lower than 0.1 accounts for multicollinearity. Table 2 shows the standard errors of the variables.
2
Variables (1) Standard errors (2)
RTMSCI 0.0735006 EX(€/$) 0.0227882 OP 0.0372498 IP 0.0003895 IF 0.7951337 TS 0.0038963 USGDPGrowth 0.5631121 1
Variable VIF 1/VIF
RTMSCI 1.32 0.758064 OP 1.28 0.779442 EX(€/$) 1.11 0.897580 IP 1.10 0.906407 IF 1.17 0.854636 TS 1.25 0.797548 USGDPGrowth 1.37 0.730082 Mean VIF 1.23
