THE EQUITY RISK PREMIUM AND ECONOMIC FLUCTUATIONS
EMPIRICAL EVIDENCE ON THE DIFFERENCES AND SIMILARITIES BETWEEN
NORTH WESTERN EUROPE AND THE P.I.G.S.
Tom Ruiter January, 2016
Abstract
This study aims to find significant differences in the ex post equity risk premium between North Western Europe and Portugal, Italy, Greece and Spain (PIGS) in the period 1999 - 2015. Additionally, the study focuses on the impact of the economic conditions, measured by the percentage change and volatility in industrial production, on the equity risk premium. No difference in the average equity risk premium is found between the North Western European region and the PIGS, which is in line with the integrated market hypothesis. The average equity risk premium in periods 1999 – 2007 and 2007 – 2015 displays a significant difference for the PIGS. This difference is not observed for the North Western European region, which can be due to the fact that the impact and aftermath of the crisis periods were less severe in this region than in the PIGS region. No clear significant impact of the industrial production on the equity risk premium is observed. However, when excluding the crisis periods from the sample, a significant impact of the economic conditions, in terms of volatility in industrial production, can be observed. Measuring the economic conditions in terms of the composite leading indicator leads to a significant impact on the equity risk premium for the entire sample.
Keywords: equity risk premium, economic fluctuations, North Western Europe, PIGS
University of Groningen Msc. Thesis
1. Introduction
The equity risk premium, the difference between the expected return on the market portfolio and the risk free interest rate, is of great interest to both investors and academics. If an investor would have invested 1000 US$ in 1927 in 3-month US treasury bills he would have received 19,737.70 US$ by the end of 2014. However, if the investor would have invested the same amount in a value weighted S&P index fund, he would have received 2,899,951.30 US$ by the end of the same term (www.nyu.edu). The return obtained on risky equity is almost 147 times higher than for the 3-month treasury bills. The difference between these returns, the so-called equity risk premium, can be attributed to the higher risk associated with investing in stock. Investors want to be compensated for their exposure to higher risk. The fact that the difference between the return on the relatively safe Treasury bills and risky equity is too large to be explained by economic models and the equilibrium theory is called the equity risk premium puzzle. This thesis will test the impact of the economic condition on the equity risk premium in North Western Europe and the PIGS. In addition, differences in the equity risk premium in both regions will be tested.
Merha and Prescott (1985) were the first to identify the equity risk premium puzzle phenomenon. Over the period 1889 – 1978, they observed an equity risk premium of 6.18%. For an equity risk premium of 6.18%, a level of relative risk aversion of 26 is needed (Merha and Prescott, 1985). According to the experimental theory, a level of relative risk aversion that is restricted to a maximum of 10 can be justified (Arrow, 1971; Altug, 1983; Friend and Blume, 1975; Hildreth and Knowles, 1982; Kydland and Prescott, 1982; Tobin and Dolde, 1971). Mehra and Prescott (1985) stated that this restriction is essential as, with larger values, any combination of equity risk premium and risk free rate can be obtained by making small changes in the process of consumption. This will not be the case when using a value of relative risk aversion of 10 or lower. To conclude, a value of relative risk aversion of 26 is highly unlikely.
Siegel (1992) extended the sample period to 1802 - 1990 and reviewed three different sub periods within this period, in order to question the sample period studied by Mehra and Prescott (1985). Siegel (1992) divided the sample period into the following sub-periods: 1802 – 1871, 1872 – 1925 and 1926 – 1990. The equity risk premiums he found for these periods are respectively 2.9%, 4.7% and 8.1%, and additionally, the study derives an equity risk premium of 5.3% for the overall sample period.
Shackman (2006) reached another conclusion, although using different measurements. Whereas Salomons and Grootveld (2003) used equity prices and the risk free rate denominated in US dollars, Shackman (2006) used local currencies and the local risk free rates. This resulted in lower equity risk premiums, which even turned out to be negative for Brazil. Both Shackman (2006) and Harvey and Bekaert (2002 and 2003) attributed the difference in the equity risk premiums between developed and emerging markets to the degree of market integration. Emerging markets are not fully integrated within the global economy, which results into higher equity risk premiums. This is due to the fact that investors are not able to fully diversify the higher risks which are perceived in emerging markets, as these markets are not perfectly integrated with the global market. This leads to higher equity risk premiums to compensate for the higher, non-fully diversifiable, risk. However, the extent of integration within the global market is changing over time. Hence, changes in the degree of market integration of emerging markets will partly influence the equity risk premium and the time variation of the premium. Shackman (2006) concluded that investors should rather take expectations on market integration into account instead of forecasted developments in economic growth.
Additionally, the impact of turbulent economic periods on the equity risk premium has been subject to a broad discussion. Harvey (1989) and more recently, academics as El Hedi Arouri and Jawadi (2010), found higher equity risk premiums during economic troughs than during peaks for US data. El Hedi Arouri and Jawadi (2010) stated that risk averse investors require higher returns if high risk is expected, and therefore require a higher equity risk premium in economic turbulent times. Their argument is supported by several examples of periods with high expected risks and returns (Oil crisis, 1973-1974; Gulf war, 1991-2003; etc.). Also the most recent financial crisis caused a rise in the price of risk implying a lack in the economy as whole.
As described, the variation in the equity risk premium is correlated with external factors as the sample period, economic conditions and geographic location. In addition, the methodology used can have great impact on the results derived. All these factors lead academics to remain their focus on the equity risk premium in order to adjust economic models and try to find an explanation to clarify the differences between the observed equity risk premiums.
time varying risk premiums can be accounted to the countercyclical changes in risk aversion (Campbell and Cochrane, 1999) or to the volatility of the consumption process during the business cycle (Bansal and Yaron, 2004).
The relation between time variation in risk aversion and the economic circumstances are suggested as an explanatory factor. The risk aversion increases during economic turbulent periods due to the higher uncertainty, which in return causes investors to require a higher equity risk premium. This relation between time varying aspect of the equity risk premium and risk aversion has been subject to extensive research in order to explain the equity risk premium puzzle. The changes in risk aversion are often linked to the concept of the habit formation model and the related habit persistence phenomenon (Pollak, 1970). In other words, risk aversion is dependent on the difference between consumption and habits formed by past consumption (Brandt and Wang, 2003). The fact that investors tend to be habit persistent results in more risk-averse investors whenever wealth and consumption decreases. Contrarily, lower risk aversion is observed when wealth and consumption are increasing (Salomons, 2008). Therefore, an economic or financial crisis causing a decrease in wealth and consumption results in an increase of investors’ risk aversion. As El Hedi Arouri and Jawadi (2010) argued, that it is very reasonable that, due to excessive fear during periods experiencing economic turbulence, higher risk aversion is observed. This higher risk aversion causes investors to require a higher equity risk premium to compensate for the higher risk perceived.
Another factor that might explain the equity risk premium puzzle is the survivorship bias (Brown et al., 1995; Jorion and Goetzmann, 1999). Rational investors take into account the small possibility of a catastrophe. The used dataset contains US data, which is considered as a survivor during the turbulent periods in the sample period. Hence, the equity risk premium obtained for US might underestimate the riskiness of the equity market, as the past data does not include the effect of this catastrophe. Therefore, the equity risk premium can be an exception (Rietz, 1988; Brown, Goetzmann and Ross, 1995; Goetzmann and Jorion 1999). Gielen (1994) and Hirose and Tso (1995) focused on equity risk premium in turbulent periods for both Germany and Japan. Their results showed that during periods of hyperinflation, equities regain their value whereas bonds experience a heavy impact. During the 20th century, the equity risk premium for these countries was actually higher than for the US. By studying more developing countries, the equity risk premium puzzle can still not be resolved, but on the contrary, it proves the robustness of the puzzle (Blanchard, 1993; Fase, 1997; Dimson, Marsh and Staunton, 2005).
The fact that there is still so much uncertainty about the equity risk premium is also of great importance to investors. The equity risk premium determines how much more an investor can earn if he allocates his assets to either the risk free government bonds or risky equity. The equity premium is, in this sense, of great importance for asset allocation, but in addition, also for estimating the cost of capital, wealth projections and many other implementations.
In literature, the terminology on the “ex-post” and “ex-ante” equity risk premium is used interchangeably. In this thesis, the equity risk premium will refer to the ex-post equity risk premium, meaning the realised excess return investors earn over the risk free rate. Using the ex-ante equity risk premium would be a better measurement as this is a predicting variable which is more relevant in terms of the equity risk premium. This is specifically due to the fact that equity risk premium is related to expected risk, and the returns investors require to compensate for this expected risk. However, deriving the ex-ante risk premium is complex, and in addition, there is much discussion on the methodology to derive the appropriate risk premium. Therefore, as there is no clear conformity on the derivation of the ex-ante equity risk premium, this thesis will use the ex-post equity risk premium to ensure the accuracy and comparability of the results.
market included in the sample will be reviewed and a comparison will be made between the North Western countries and the PIGS. The PIGS countries were considered as vulnerable economies during the European sovereign debt crisis, due the high national budget deficits compared to the gross domestic product (GDP). Therefore it is likely that the equity risk premium in the sample period might have differed. However, if the markets are fully integrated with the global market, the higher perceived risk in the PIGS countries would be diversifiable, leading to similar equity risk premiums. Therefore the following hypothesis will be tested:
H1: There is a significant difference between the equity risk premium in the North Western countries
and the PIGS.
In addition, this study will test whether the equity risk premium changed over time in North Western Europe and in the PIGS. It is likely that the equity risk premium has significantly changed after 2007, as the period 2007 – 2015 was dominated by the financial and sovereign debt crisis. The higher risks during this period will possibly have affected the equity risk premium, therefore the following hypothesis is set:
H2: The equity risk premium significantly differs between the periods 1999 – 2007 and 2007 – 2015.
As described, the equity risk premium is considered to be affected by the economic condition. Previous research showed that economic turbulent periods led to changes in the equity risk premium as it affects the risks perceived by investors. The industrial production is used as a proxy for the economic condition to test the impact on the equity risk premium. The relation between the economic conditions and the equity risk premium will be tested by using the industrial production in two different ways. Firstly, the relation will be tested by regressing the equity risk premium on the percentage change in industrial production. Secondly, the relation between the volatility in industrial production and the equity risk premium will be tested. In order to verify whether the economic condition will have a significant impact on the equity risk premium the following hypothesis is set:
H3: There is a significant relation between the economic condition and the equity risk premium.
This paper will be structured as follows; section two will describe the methodology used in this thesis, section three will briefly elaborate on the data used in this study, section four will briefly analyse the results of this study, section five will discuss the results and section six will provide the conclusion of this paper.
2. Methodology
This section will describe the methodology used to test the preconceived hypotheses as described in the previous section. Both the methods used to derive the equity risk premiums and the volatility in industrial production will be elucidated. In addition, the control variables, the regressions, significance and robustness tests used will be described.
2.1 Equity risk premium
The “ex post” equity risk premium is the excess return of equities in a particular market over the bond return in that specific market, used as a proxy of the risk free rate:
𝐸𝑅𝑃𝑖,𝑡 = 𝑅𝑖,𝑡𝐸 − 𝑅𝑖,𝑡𝐵
where RE
i,t is the monthly return on equity of a market i on time t and RBi,t is the return on a bond with one
-month maturity.
In order to ensure that the total return of the market is captured, the total return indices are used. These indices are adjusted for income from cash dividends or capital repayments, which are reinvested in the index. There are no further inferences made on the international capital asset pricing model (CAPM) (Lintner, 1965a and 1965b; Sharpe, 1964; Treynor, 1961 and 1962). The lognormal return of a particular market is derived from the various Morgan Stanley Capital International (MSCI) indices (www.msci.com). It is common practice to use the lognormal return when calculating stock or indices returns. The two key reasons to use lognormal returns are the facts that the returns are continuously compounded and time-additive (Brooks, 2008). Moreover, as stock prices cannot be negative, and are often assumed to be better described by the lognormal distribution than by the normal distribution. Therefore, in order to calculate the return of the various indices, the lognormal function is used:
𝑅𝑖,𝑡𝐸 = 𝐿𝑁 ( 𝐼𝑖,𝑡 𝐼𝑖,𝑡−1
)
In order be able to test the hypotheses, normality has to be assumed. Therefore, the returns will be tested for normality using the Kurtosis, Skewness and Jarque-Bera statistic (1980).
To get a more comprehensive view on the equity risk premium, the Sharpe and Sortino ratio are calculated. The Sharpe ratio (Sharpe, 1994) adjusts returns for their respective volatility and therefore gives a better insight in the relation between risk and return. As this ratio is taking into account the volatility, it is also called the reward to variability ratio. This ratio is assumed to be a better measurement than the return itself as it compensates for the risk taken by investors.
𝑆ℎ𝑎𝑟𝑝𝑒𝑖 =
𝐸(𝑅𝑖𝐸− 𝑅𝑖𝐵) 𝜎𝑖
where σi is the standard deviation of the equity risk premium of index i.
The Sharpe ratio treats both upside as downside volatility in the same way. The Sortino ratio (Sortino, 1991) is an equivalent measure to the Sharpe ratio; however, this ratio is adjusted for downside risk. It only takes into account the volatility of returns below a certain threshold, and rewards investors for returns obtained above this threshold. It therefore solely penalizes investors for negative volatility:
𝑆𝑜𝑟𝑡𝑖𝑛𝑜𝑖 = 𝐸(𝑅𝑖𝐸− 𝑅𝑖𝐵) − 𝑀𝐴𝑅 √1 𝑇∑ (𝑅𝑖,𝑡− 𝑀𝐴𝑅) 2 𝑇 𝑡=0, 𝑅𝑖,𝑡<𝑀𝐴𝑅
where MAR is the minimal acceptable excess return, which is zero in this case, rewarding investors for positive excess returns.
2.2 Volatility in industrial production
to adopt the common frequency approach. The growth in the industrial production index is calculated by the following formula: ∆%𝐼𝑃𝑖,𝑡 = ( 𝐼𝑃𝑖,𝑡 𝐼𝑃𝑖,𝑡−1 ) − 1
Where IPi,t is the level of industrial production of index i at time t.
The volatility in industrial production will be derived using an AR(12) GARCH(1,1) model. An autoregressive model with twelve lags is used in order to take into account the effects of a full year and therefore the model adjusts for seasonality effects. Using 24 lags increases the number of estimators, leading to an unnecessary extensive model. This did not improve the results, and additionally, the variation caused by the yearly seasonal fluctuations is already captured by using solely twelve lags.
The variance of the error terms is not likely to be constant over time as some periods experience more turbulence, resulting in higher risk than in other, more stable periods. As a result of this phenomenon, named volatility clustering, an alternative model needs to be used in order to adjust for the resulting heteroscedasticity. The autoregressive conditional heteroscedasticity model (ARCH) (Engle, 1982) allows to model volatility clustering, which is often the case for financial time series. Many models which are available are subject to wide criticism, but, as Robert Engle received the Nobel Prize in 2003 with the ARCH model, and in addition, as the model is widely used by academics it can be assumed that it is appropriate to adopt this model. Furthermore, the specifications of this model fits the purpose of this study, as the model adjusts for shortcomings of the standard ordinary least square method (OLS). The OLS method assumes constant and finite variance over time, namely homoscedasticity. But on the contrary, using the ARCH specification, the variance depends on the squared error term of the previous period, allowing heteroscedasticity. A model which does not take into account the presence of heteroscedasticity, will still lead unbiased coefficients, however, it results in standard errors and the confidence interval estimates which are too narrow (Engle, 2001).
The GARCH(1,1) is defined as:
∆%𝐼𝑃𝑖,𝑡 = 𝛼 + ∑ 𝜑𝑙∆%𝐼𝑃𝑡−1 12
𝑙=𝑖
+ 𝜀𝑖,𝑡
with variance equation of the error term εi,t, being
𝜎𝑖,𝑡2 = 𝛼0+ 𝛼1𝜀𝑡−12 + 𝛽𝜎𝑡−12
The volatility of industrial production will be defined as the conditional variance of this equation. The conditional variance is obtained as a result of the regression.
In order to verify whether the model is a good estimator, the residuals will be tested for the Gauss-Markov assumptions as described by Brooks (2008). If these assumptions are not satisfied, it will lead to a model with inappropriate standard errors and therefore can lead to misleading inferences. Firstly, as the model contains a constant, it can be assumed that the residuals have zero mean. In addition, the GARCH(1,1) specification is used, adjusting the model for heteroscedasticity. In order to test for autocorrelation of more than one lag the Breusch-Godfrey Serial correlation LM test and the method as described by Ljung and Box (1978) is used.
2.3 Control variables
Using only industrial production as independent variable might lead to an incorrectly defined regression, as important factors might not be included in the regression. As a consequence, a bias might arise as the model might compensate incorrectly for these omitted variables. As the equity risk premium is dependent on a broad set of factors, control variables are added to the regression. To assure that the model is adjusted correctly for the omitted variable bias, the control variables are added in line with previous research (Salomons and Grootveld, 2003; Shackman, 2006; Damodaran, 2012). Excluding these control variables might lead to incorrect estimates of the coefficients in the regression.
This reasoning is in line with findings of Damodaran (2012), who stated that the equity risk premium depends for a great extent on the state and predictability of the economy.
The variables described in this paragraph are included as control variables, in line with previous research as explained in the data section. Dependent on the variable, either the volatility or percentage change is used in this study. The volatility in the consumer price index (inflation) is obtained using the same methods as for the derivation of the volatility in the industrial production, using an autoregressive model with twelve lags including the GARCH(1,1) specification. The beta, defined as the co-movement of every market with the global market, is calculated by deriving the ratio between the returns of the local and global market. The short-term interest rate, the harmonized unemployment rate, the composite leading indicator, the consumer confidence index, the business confidence index and the USD/EUR exchange rate are included as the monthly percentage change.
Including control variables in the model might lead to multicollinearity between the various independent variables. Multicollinearity can bias the coefficient estimates of other independent variables in the regression. Therefore, the correlation between the independent variables is tested. A control variable will be excluded from the model if it is highly correlated with one or more other control variables.
2.4 Dummy variables
In addition to regressing the variables over the entire timespan, the regressions will also be performed excluding crisis periods. The Internet bubble in 2000 - 2001 and the subprime crisis during the period 2007 – 2009 will be excluded from the regression using a dummy variable. Furthermore, the period before and after 2007 will be regressed using dummy variables.
2.5 Relation between the equity risk premium and the economic condition
The regression described in this section will be used to test the effect of the volatility and the percentage change in industrial production on the equity risk premium. The control variables described in section 2.3 are added to the regression to assure that omitted variables will not bias the coefficients. White’s heteroscedasticity consistent standard errors are used in order to adjust for heteroscedasticity in the error terms (White, 1980), as incorrectly defined error terms can lead to misleading results. This led to the following regression:
𝐸𝑅𝑃𝑖,𝑡= 𝛼𝑖,𝑡+ 𝛽1𝐼𝑃𝜎,𝑖,𝑡 + 𝛽2𝐼𝑃∆%,𝑖,𝑡+ 𝛽3𝐶𝑃𝐼𝜎,𝑖,𝑡+𝛽4𝑆𝑇𝐼∆%,𝑖,𝑡+ 𝛽6𝐻𝑈𝑅∆%,𝑖,𝑡+ 𝛽7𝐶𝐿𝐼∆%,𝑖,𝑡+ 𝛽8𝐶𝐶𝐼∆%,𝑖,𝑡+ 𝛽9𝐸𝑅∆%,𝑖,𝑡+ 𝛽10𝐵𝐸𝑇𝐴𝑖,𝑡+ 𝜀𝑖,𝑡
IPσ: monthly volatility in industrial production
STI Δ%: monthly percentage change in the short-term interest rate
HUR Δ%: monthly percentage change in the harmonized unemployment rate CLI Δ%: monthly percentage change in the composite leading indicator CCI Δ%: monthly percentage change in the consumer confidence index ER Δ%: monthly percentage change in de USD/EUR exchange rate BETA: co-movement of local index with global market
The regression is derived for each of the seven countries, for North Western Europe and the PIGS for the entire sample period (1999 – 2015). In addition, regressions are performed for the periods 1999 – 2007, 2007 – 2015 and for the entire sample period excluding the crises periods (2000 – 2001 and 2007 – 2009). This leads to a total number of 36 regressions. For each of these regressions, the best fit is obtained by excluding control variables, which do not significantly impact the regression or bias the results of other coefficients. The Gauss-Markov assumptions for the residuals are tested in a similar manner as explained in the derivation of the volatility in industrial production.
This regression will be used to either confirm or reject hypothesis three, testing for a significant relation between the economic state and the equity risk premium. A significant coefficient for either the percentage change or volatility in industrial production proofs that the proxy for the economic condition has a significant influence on the equity risk premium. Additionally, the size and the sign of the coefficient shows the magnitude and the direction of the relation.
2.6 Significance testing
To test whether the mean of the North Western European region significantly differs from the mean of the PIGS, the Wald t-test will be used (Wald, 1939). For testing the differences between the medians, the median chi squared test is used (Pearson, 1900). These tests solely test whether there can be significant differences observed between the time periods and regions, and will not take into account any effects of either the percentage change or the volatility in industrial production. By performing these test, both hypothesis one and two, as described in the previous section, can be confirmed or rejected.
3. Data
compared to GDP. Results of the North Western European region and the PIGS are based on an equal weighted index method of the respective countries, to assure that the results are not dominated by size effects of the included markets. The sample set used in this thesis includes data in the period January 1998 until August 2015. Using this timespan assures that both economic stable and turbulent times are incorporated in the dataset, which is in line with the main research question stated in this thesis.
3.1 The equity risk premium
To derive the excess return for the different markets the Morgan Stanley Capital International (MSCI) indices are used. As the MSCI indices incorporate large and mid capitalization companies, it will be an appropriate proxy for a particular market. In addition, this is also in line with earlier papers (Salomons and Grootveld, 2003; Shackman, 2006). The sample covers a timespan which starts at the introduction of the Euro and only contains countries being part of the Euro zone, therefore all data is measured in local currencies, namely the Euro. As a result of using a common currency dataset, currency fluctuations will not have influence on the outcomes and therefore, the results can be interpreted from the perspective of an international investor. The MSCI indices contain monthly returns and are adjusted for income from cash dividends or capital repayments, which are reinvested in the index.
As all the countries are part of the Euro zone, the EURIBOR, can be used as an appropriate proxy for the risk free rate (Brooks and Yan, 1999; Duffie and Stein, 2014). The EURIBOR is the average rate of interest for which banks lend unsecured funds in the Euro zone interbank market. This method is in line with similar research papers as the paper of Salomons and Grootveld (2003), who used the US monthly money market rate. As the returns on the MSCI indices are monthly, the one-month EURIBOR, obtained from the Bloomberg Terminal, will be used.
Table 1.
Sample size, mean, median, minimum, maximum, standard deviation, skewness, kurtosis, Jarque-Bera statistic and the corresponding probability of the equity risk premiums of the North Western European
countries on monthly basis.
France Germany Netherlands North Western Europe
No. 199 199 199 199 Mean 0.09% 0.11% 0.05% 0.08% Median 0.70% 0.69% 0.59% 1.00% Min -17.55% -28.80% -20.52% -22.29% Max 11.92% 17.84% 12.27% 12.74% St. dev 5.17% 6.33% 5.44% 5.43% Skewness -0.611 -0.861 -1.040 -0.882 Kurtosis 3.770 5.638 5.101 4.960 Jarque-Bera 17.299 82.277 72.437 57.678 Probability 0.000 0.000 0.000 0.000 Table 2.
Sample size, mean, median, minimum, maximum, standard deviation, skewness, kurtosis, Jarque-Bera statistic and the corresponding probability of the equity risk premiums of the Greece, Italy, Portugal and
Spain (PIGS) on monthly basis.
Greece Italy Portugal Spain PIGS
The equity risk premiums for the countries in the sample are tested for normality using the Jarque-Bera statistic (1980). For all the countries included in the sample the statistic is significant, meaning that the equity risk premiums are not normally distributed (Table 1 and 2). Graph 1 shows a visualisation of the distribution of the returns of both North Western Europe and the PIGS.
Graph 1.
Distribution of equity risk premium (monthly data) for North Western Europe and the PIGS. The bars indicate the number of observations per percentage return, the dotted line it the normal distribution based
on the mean and standard deviation of the sample.
North Western Europe PIGS
As also can be observed in Table 1 and 2 and Graph 1, all the equity risk premiums in the sample are negatively skewed and show a positive excess kurtosis. This is confirmed by the Jarque-Bera statistics, meaning that the equity risk premiums are not normally distributed. The same results for skewness and kurtosis are found in previous research on the equity risk premium (Bekaert, Erb, Harvey and Viskanta, 1998; Salomons and Grootveld, 2003).
are considered to be stronger economic countries within Europe, and therefore were able to sustain a positive growth in this period with severe economic conditions.
Table 3.
Descriptive statistics for the percentage change in industrial production for all the countries included in the sample. The monthly average, standard deviation, minimum and maximum are given.
Mean St. dev. Minimum Maximum
France -0.023% 1.297% -5.017% 4.491%
Germany 0.146% 1.582% -7.946% 4.632%
Netherlands 0.057% 2.242% -8.256% 5.757%
North Western Europe 0.060% 1.256% -5.115% 3.485%
Greece -0.032% 2.711% -7.925% 9.651% Italy -0.091% 1.362% -4.194% 3.642% Portugal -0.006% 2.388% -6.323% 7.487% Spain -0.046% 1.427% -4.125% 5.112% PIGS -0.043% 1.236% -3.562% 3.165% 3.2 Control variables
The control variables used in this study, are obtained from the OECD and Bloomberg. According to Salomons and Grootveld (2003) and Shackman (2006), the equity risk premium is to a great extent dependent on the degree of market integration. According to Bekaert and Harvey (1995), the degree of market integration can be derived from the risk-return ratio. Markets are fully integrated if the same amount of risk leads to the same expected return, independently of the geographical location. Contrarily, the expected returns of fully segmented markets will be not or little explainable by its covariance with the global market. Due to other sources of risk, the expected return for a certain amount of risk in segmented markets might differ. Factors as economical structure and dependence on a market style economy affect the degree of integration with the global market (Salomons and Grootveld, 2006). Bekaert and Harvey (1995) concluded that an increase in market integration will lower the cost of capital as it allows investors to diversify the unsystematic risks. A higher correlation with the global market should therefore result in a lower required return demanded by investors. The degree of market integration will be included in this thesis by creating a variable measuring the co-movement between a specific market and the global market.
and Wachter, 2007; Brandt and Wang, 2003). Hence, this thesis will include the volatility in the consumer price index, the percentage change in the short-term interest rate, the percentage change in USD/EUR exchange rate and both the absolute and the percentage change of the harmonised unemployment rate. Additionally, as explained in section 2.3, these variables partly help to explain the pattern observed in the error terms when regressing the equity risk premium solely on the percentage change and the volatility in industrial production.
Lastly, the composite leading indicator, the consumer confidence index and the business confidence index are included as control variables. Including these variables is in line with Salomons and Grootveld (2003), as they have certain strength in forecasting with regard to the state of the economy. Their research finds that these variables lead to a better fit with the equity risk premium than when using actual production. The composite leading indicator of the OECD is an index created to indicate early signs of peaks and troughs in the business cycle. The indicator is structured in such a way that it displays the movement of the economy around the long-term economic level. In addition, the indicator is based on qualitative factors to predict short-long-term fluctuations. The composite leading indicator showed over the past years, that it often correctly forecasts patterns in the business cycle. As the composite leading indicator is often considered to accurately forecast the business cycle, it will also be used as a proxy for the economic condition, in order to test for robustness of the outcomes of the industrial production variables. The consumer and business confidence index are both qualitative indices based on monthly surveys. The OECD derives the business confidence index by examining the assessment of the business sector on for example, the production, order and expectations. The consumer confidence index is based on households’, both current and near future, expectations on major purchases and on the economic state (www.oecd.org). A description and the corresponding source per control variable are given in Table 4.
Table 4.
Description and source of control variables
Variable Description
Consumer Price Index
The monthly growth rate of the consumer price index, measured by the OECD (www.oecd.org) as the change in the price of a basket containing specific, fixed set of goods and services, of constant quantity, which are commonly consumed by certain groups of households. This variable specifies the change in living standard of specific groups of households. The consumer price index is closely related with the economic activities and is therefore one of the main economic indicators of the OECD.
Short-term Interest Rate
The short-term interest rate are defined by the OECD (www.oecd.org) as either the short-term government rate or the rate at which financial institutions borrow to each other on the short-term (typically the three month money market rate or the three month treasury bill rate).
Harmonized Unemployment Rate
The OECD (www.oecd.org) defines the harmonized unemployment rate as the amount of unemployed persons, who are available for work, and are actively looking for a job by taking specific steps to find work, as a percentage of the total labour force (the total number of unemployed and civilian employed persons). In order to allow for international comparison, the rate is adjusted for differences in national definitions. In addition, the rate is adjusted for seasonal effects.
Composite Leading Indicator Δ%
The composite leading indicator of the OECD (www.oecd.org) is an index created to indicate early signals of peaks and troughs in the business cycle. The indicator is structured in such a way that is displays the movement of the economy around the long term economic level. In addition, the indicator is mainly based on qualitative factors to predict short-term fluctuations. In this thesis, the percentage change in the index rather than the absolute index amount is used, as the change in the index indicates a change in the economy. The change in this index has proven over past years to move closely together with the business cycle, and according to previous research it is an appropriate forecaster of the equity risk premium (Salomons and Grootveld, 2003).
Business Confidence Index Δ%
The business confidence index of the OECD (www.oecd.org) provides a perspective on the outlook of the business sector. The index is based on the qualitative assessment of the business sector on factors as production, stocks and orders. In addition, the index takes into account the assessment of the business sector on its current position and their expectation of the near future.
Consumer Confidence Index Δ%
The consumer confidence index is an index of the OECD (www.oecd.org) to provide a perspective on the consumer confidence based on qualitative information. The index is of monthly frequency and derived by the assessment of households on their, both current and expected, major purchases and perspective on the economic situation.
Beta The beta is calculated as the co-movement of the local MSCI index with the global MSCI index
(www.msci.com).
Exchange Rate (USD/EUR)Δ%
Table 5 shows the descriptive statistics of the control variables. The average harmonized unemployment rate shows a difference of approximately five percent between the North Western European region and the PIGS. Furthermore it is notable, that for almost every control variable, the standard deviation is higher for the PIGS region. The descriptive statistics per country are shown in Appendix A and B. Additionally, Graph 2 to 5 display the development of the control variables which had the most significant impact over the different regressions performed.
Table 5.
Descriptive statistics of the monthly control variables for North Western Europe and the PIGS. The monthly average, standard deviation, minimum and maximum are given.
North Western Europe PIGS
Mean Std. dev. Minimum Maximum Mean Std. dev. Minimum Maximum
Consumer Price Index 0.135% 0.058% -0.029% 0.277% 0.183% 0.112% -0.107% 0.353% Short-term Interest Rate 0.185% 0.129% -0.004% 0.416% 0.198% 0.142% -0.004% 0.454% Unemployment Rate 7.323% 0.659% 6.067% 8.667% 12.053% 4.297% 7.900% 20.800% Composite Leading Indicator Δ% 0.004% 0.215% -0.823% 0.621% -0.490% 7.056% -100.000% 0.455% Business Confidence Index Δ% 0.002% 0.158% -0.750% 0.308% -0.005% 0.168% -0.805% 0.336% Consumer Confidence Index Δ% -0.003% 0.150% -0.449% 0.322% -0.004% 0.148% -0.310% 0.479% Beta -1.528 32.469 -455.995 8.701 -7.339 103.564 -1456.519 35.218 Exchange Rate (USD/EUR)Δ% 0.074% 3.040% -9.645% 10.996% 0.074% 3.040% -9.645% 10.996%
Graph 2.
USD/EUR exchange rate over the period December 1998 – June 2015.
Graph 3.
Graph 4.
Composite leading indicator (OECD) for France, Germany, Greece, Italy, the Netherlands, Portugal and Spain over the period December 1998 – June 2015.
Graph 5.
4. Results
The results will be described in several steps. Firstly, the patterns in the indices will be presented. Secondly, the differences in the equity risk premiums between North Western Europe and the PIGS and between different periods will be tested. Thirdly, the regressions testing the effect of industrial production on the equity risk premium for the various countries, North Western Europe and the PIGS will be described. Lastly, the regressions will be performed for the various time periods.
As can be observed in Graph 6, the indices have a similar pattern until the beginning of 2009. The indices tend to move simultaneously during this period, whereas after 2009 different patterns for the sample can be observed. Furthermore, sharp drops in all the indices can be observed in the beginning of 2000 and mid-2007. This pattern is in line with the dot.com crash or Internet bubble (2000 – 2001) and the subprime crisis (2007 – 2009) described by Phillips and Yun (2011) and Adrian and Shin (2010). This sharp drop in the indices leads to include regressions in this thesis, which excludes the crisis periods. This will test whether the crisis periods have impact on the equity risk premiums and the regressions performed.
Graph 6.
4.1 Differences in the equity risk premium between the North Western European region and the PIGS
A difference between the equity market risk premium in North Western Europe and the PIGS can be observed. The North Western European region shows a slightly positive monthly equity risk premium, whereas the PIGS have a negative monthly return of -0.51%. However, the observed difference between the equity risk premium is not significant. Testing the difference between the medians of both equity risk premiums shows a significant result. Therefore, hypothesis 1, testing the difference between the equity returns in both regions, is rejected when testing in terms of the averages of North Western Europe and the PIGS. However, when testing the median of both regions, hypothesis one can be confirmed. The fact that there is a significant difference in terms of medians, but not in terms of means, can possibly be attributed to outliers as there are more negative outliers observed for the PIGS countries.
As can be observed in Table 6, the Sharpe ratio shows a positive value for North Western Europe, and contrarily, a negative value for the PIGS. Meaning, the amount of return received per unit of volatility is higher in North Western Europe. When taking into account the Sortino ratio, a lower value for the PIGS than for North Western Europe can be observed. The Sortino ratio is considered to be a better measurement as it takes into account solely downside risk. It purely looks at downside risk, resulting in rewarding investors for positive outliers and penalizing investors for negative returns, whereas the Sharpe ratio equally treats positive and negative outcomes.
Table 6.
Mean, median, Sharpe ratio and Sortino ratio for North Western Europe and the PIGS. In the last column the statistics to test the difference in mean and median between the respective groups are shown.
North Western Europe PIGS
Mean 0.08% -0.51% t1 = 1.056
Median 1.00% -0.47% χ2 2 = 2.905*
Sharpe3 0.053 -0.298
Sortino3 0.07 -0.39
* indicates that the statistic is significant at a 10% level, ** indicates that the statistic is significant at a 5% level and *** indicates that the statistic is significant at a 1% level.
4.2 Time varying aspect of the equity risk premium
The sample period is split in order to test whether the subprime crisis, which started in 2007, caused a structural change. The results are shown in Table 7. Both the median and mean show a significant difference for the PIGS between the periods 1999 – 2007 and 2007 - 2015. The average equity risk premium decreased from an average
1 Wald t-test
2 Median Chi-square test
of 0.14% to -1.23% in the period 2007 – 2015. This leads us to confirm hypothesis two for the PIGS countries, stating that a significant difference can be observed between the periods 1999 – 2007 and 2007 - 2015. However, conducting the same tests for North Western Europe leads to contradictory results, namely, hypothesis two is rejected in both terms of mean and median.
Table 7.
Mean, median for North Western Europe and the PIGS in the periods 1999 – 2007 and 2007 - 2015. The last column shows the test statistic for the difference in mean and median between the two groups per period.
The bottom row shows the test statistic for the difference between the two periods per group.
North Western Europe PIGS
1999 - 2007 Mean 0.12% 0.14% t4 = -0.033 Median 1.22% 0.95% χ25= 0.077 2007 - 2015 Mean 0.04% -1.23% t = 1.434 Median 0.74% -1.22% χ2 = 6.084** t = 0.108 t = 1.687* χ2 = 0.128 χ2 = 6.911***
* indicates that the statistic is significant at a 10% level, ** indicates that the statistic is significant at a 5% level and *** indicates that the statistic is significant at a 1% level.
Comparing North Western Europe and the PIGS results in the median equity risk premium showing a significant difference in the period 2007 – 2015, whereas both the average and median average risk premium are not showing a significant difference in the period 1999 – 2007.
4.3 Impact of the economic conditions on the equity risk premium
The impact of the percentage change and volatility in industrial production on the equity risk premium is tested by reviewing the correlation between both variables and by performing various regressions. As can be observed in Graph 7 and 8 on the next page, showing the monthly change in industrial production and the equity risk premium, correlation between the variables is evident in the overall period. Correlation above 0.6 is observed for specific time periods within the sample. However, the observed correlation is in general higher for North Western Europe than for the PIGS. A peak in the equity risk premium can be observed after the subprime crisis in 2007 – 2009. According to Ameur, Gnégné and Jawadi (2013) this indicates the increased uncertainty and risk aversion after the crisis.
4 Wald t-test
Graph 7.
Monthly change in industrial production and the equity risk premium North Western Europe in the sample period January 1999 – July 2015.
Graph 8.
The first regressions are based on the entire sample period, 1999 – 2015. The equity risk premium is regressed against the volatility and percentage change in industrial production together with the control variables. Before including the control variables, the same regressions were performed only using the percentage change and volatility in industrial production. These regressions led to poor results and extremely low goodness-of-fit statistics. In addition, as explained in section 2.3, the similar pattern observed in the residuals of the different regressions indicates that the residuals could be partly explained by a missing variable. As these results therefore lack added value, this thesis will only display results of the regression including the control variables.
As can be observed in Table 8 and 9, the volatility in industrial production has only significant impact on the equity risk premium in Germany. The effect of the percentage change in industrial production is positive for the countries in North Western Europe; however, these results do not display any significance. The short-term interest rate, the harmonized unemployment rate and the exchange rate did not have significant impact for each of the countries in North Western Europe, and are therefore excluded from the regressions.
Table 8.
Regression of the volatility of industrial production, percentage change in industrial production and control variables on the equity risk premium for France, Germany, the Netherlands and North Western Europe for
the period 1999 – 2015 using White consistent standard errors. The table shows the coefficient, the
corresponding p-value and the adjusted goodness-of-fit statistic (R2).
France Germany Netherlands North Western Europe
% change industrial prod. Δ% 0.017 0.251 0.028 0.205
0.954 0.458 0.868 0.541
Volatility industrial prod. σ -8.975 -12.944* -7.007 25.416
0.599 0.079 0.733 0.131
Consumer price index σ**** 0.351* -0.908** -0.014***
0.056 0.045 0.000
Composite leading indicator Δ% 12.905*** 6.950*** 6.449** 6.465**
0.000 0.000 0.011 0.026
Consumer confidence index Δ% 5.536* 7.3997*
0.075 0.075
Beta**** -0.041***
0.000
Adjusted R2 0.147 0.112 0.161 0.166
# observations 199 199 199 199
* indicates that the statistic is significant at a 10% level, ** indicates that the statistic is significant at a 5% level and *** indicates that the statistic is significant at a 1% level. **** coefficient multiplied with the factor 1000
co-movement of the local market with the global market (beta) has a significant impact, however, the coefficient is rather small. In addition, in contrast to the countries in North Western Europe, the USD/EUR exchange rate is significant for each of the PIGS countries, having a negative impact on the equity risk premium. An increase in the USD/EUR exchange rate means a depreciation of the euro in relation to the Dollar. Appreciation or depreciation of a currency is directly related to the demand and supply of the currency. Changes in the exchange rate are therefore often linked to the state of the economy.
Table 9.
Regression of the volatility of industrial production, percentage change in industrial production and control variables on the equity risk premium for Portugal, Italy, Greece, Spain and the PIGS for the period 1999 - 2015 using White consistent standard errors. The table shows the coefficient, the corresponding p-value
and the adjusted goodness-of-fit statistic (R2).
Portugal Italy Greece Spain PIGS
% change industrial prod. Δ% -0.082 0.201 -0.199 -0.664** -0.229
0.593 0.559 0.4449 0.018 0.402
Volatility industrial prod. σ 23.138 -19.695 0.093 -7.670 17.725
0.496 0.529 0.997 0.795 0.749
Consumer price index σ**** 0.066*** -4.481**
0.000 0.013
Short-term interest rate Δ% -0.0148* -0.010*
0.068 0.060
Unemployment rate Δ% -0.541*
0.0769
Composite leading indicator Δ% 6.188*** 9.953*** 9.349*** 11.355** 9.907***
0.000 0.000 0.002 0.010 0.002
Consumer confidence index Δ% 7.546
0.039
Exchange rate (USD/EUR) Δ% -0.378*** -0.293** -0.788*** -0.390** -0.414***
0.001 0.024 0.004 0.012 0.002
Beta**** -0.048*** -0.059*** -0.062*** -0.056***
0.000 0.000 0.000 0.000
Adjusted R2 0.165 0.146 0.108 0.156 0.226
# observations 199 199 199 199 199
* indicates that the statistic is significant at a 10% level, ** indicates that the statistic is significant at a 5% level and *** indicates that the statistic is significant at a 1% level. **** coefficient multiplied with the factor 1000
cycle. The remaining control variables show significance or attributed to an improved regression dependent on the country.
The results of the regressions described can be biased as this sample period includes two severe crisis periods. Therefore, another regression is performed excluding the crises periods, in order to test whether the results would differ if the sample period would solely include a more stable economic period.
Table 10.
Regression of the volatility of industrial production, percentage change in industrial production and control variables on the equity risk premium for France, Germany, the Netherlands and North Western Europe for
the period 1999 – 2015, excluding the crisis periods, using White consistent standard errors. The table
shows the coefficient, the corresponding p-value and the adjusted goodness-of-fit statistic (R2).
France Germany Netherlands North Western Europe
% change industrial prod. Δ% -0.146 0.422 -0.041 0.071
0.630 0.244 0.820 0.850
Volatility industrial prod. σ 57.119** 11.748 34.935* 83.382*
0.030 0.573 0.053 0.068
Consumer price index σ**** 0.416** -0.015
0.0142 0.0005
Composite leading indicator Δ% 13.279*** 5.275** 7.399***
0.001 0.036 0.003
Consumer confidence index Δ% 12.191***
0.000
Beta**** -0.040*** -0.066***
0.000 0.003
Adjusted R2 0.086 0.036 0.082 0.077
# observations 157 157 157 157
* indicates that the statistic is significant at a 10% level, ** indicates that the statistic is significant at a 5% level and *** indicates that the statistic is significant at a 1% level. **** coefficient multiplied with the factor 1000
As can be observed in Table 10, no significant pattern can be observed in the percentage change in industrial production in the North Western European countries. Excluding the crisis periods from the regression does not lead to a more significant coefficient of the percentage change in industrial production, neither to a higher adjusted goodness-of-fit statistic. The control variables show in general similar results to the regression performed including the crises periods. The most notable difference due to excluding the crisis periods is the positive significant impact of the volatility in industrial production for France, Netherlands and the North Western European region.
is in line with the effects found in North Western Europe. Furthermore, in contrast to previous regressions, the percentage change in industrial production shows a significant negative impact for Greece, Spain and the PIGS. Similar to previous regressions, including the crises periods, the USD/EUR exchange rate and the composite leading indicator significantly impact the equity risk premium with approximately the same magnitude. In addition, the adjusted goodness-of-fit statistic has similar values as for the previous regressions.
Table 11.
Regression of the volatility of industrial production, percentage change in industrial production and control variables on the equity risk premium for Portugal, Italy, Greece, Spain and the PIGS for the period 1999 – 2015, excluding the crisis periods, using White consistent standard errors. The table shows the coefficient,
the corresponding p-value and the adjusted goodness-of-fit statistic (R2).
Portugal Italy Greece Spain PIGS
% change industrial prod. Δ% 0.051 0.313 -0.5293* -0.867** -0.543*
0.756 0.429 0.073 0.039 0.083
Volatility industrial prod. σ 51.817*** 97.116** 41.936* 96.776** 170.333***
0.008 0.017 0.097 0.017 0.001
Consumer price index σ**** 0.067*** -4.758***
0.000 0.006
Short-term interest rate Δ% -0.019***
0.009
Unemployment rate Δ% -0.419** -1.059*** -0.684*
0.021 0.001 0.056
Composite leading indicator Δ% 5.431*** 10.052*** 9.714* 12.336*
0.000 0.001 0.097 0.000
Consumer confidence index Δ% 6.028***
0.008
Exchange rate (USD/EUR) Δ% -0.446*** -0.362** -0.837*** -0.406***
0.001 0.028 0.007 0.001
Beta**** -0.053*** -0.059*** -0.076*** -0.063***
0.000 0.000 0.000 0.000
Adjusted R2 0.141 0.068 0.106 0.129 0.170
# observations 157 157 157 157 157
* indicates that the statistic is significant at a 10% level, ** indicates that the statistic is significant at a 5% level and *** indicates that the statistic is significant at a 1% level. **** coefficient multiplied with the factor 1000
These findings leads us to reject hypothesis 3, as this study does not find a significant impact of both the volatility or the percentage change in industrial production on the equity risk premium. However, significant evidence is found for the impact of the volatility in industrial production on the equity risk premium when excluding the crisis periods.
impact for the PIGS countries in the period 2007 – 2015, whereas the effect is less obvious in the period 1999 – 2007. However, the model shows a higher adjusted goodness-of-fit statistic for all the countries included in the sample, when the regressions are performed only in the period 2007 – 2015. The statistic ranges between 0.23 and 0.33 for the countries included in the sample, meaning that it seems that the regressions are better explaining the relation in the period 2007 -– 2015.
The regressions containing all variables, not excluding insignificant variables, are included in Appendix G to J in order to test the robustness of the outcomes. The significance, sign and magnitude of most variables is in general in line with the regressions where the insignificant variables are excluded. This finding proofs the robustness of the outcomes and the conclusions derived from these results.
5. Discussion of results
5.1 The equity risk premium, geographical differences and the time varying aspect
This thesis tests the differences between North Western Europe and the PIGS, and between the periods 1999 – 2007 and 2007 – 2015. No significant difference in average equity risk premiums is found between North Western Europe and the PIGS. This finding cannot directly be compared to other studies as the time period used differs from other studies. But, the fact that no difference is observed in the average equity risk premium between North Western Europe and the PIGS is in line with the integrated market hypothesis (Harvey and Bekaert, 2002 and 2003; Shackman, 2006). It states that, independently of the geographical location, in fully integrated markets, the same amount of risk will lead to the same amount of expected return. A positive Sharpe ratio is observed in North Western Europe, whereas this ratio is negative for the PIGS. Observing the Sortino ratio, solely taking into account downside risk, shows a lower ratio for the PIGS than for North Western Europe. Also this finding can be explained by the fact that both North Western Europe and the PIGS can be considered as fully integrated with the global market. Meaning, all unsystematic risk can be diversified away, and therefore, the higher downside risk in the PIGS is diversifiable. This results in a lower return to downside risk ratio for the PIGS countries. Whereas, in not fully integrated markets as the emerging markets, a higher Sortino ratio can be observed, as the higher risk is not fully diversifiable (Salomons and Grootveld, 2006).
5.2 Impact of the economic conditions on the equity risk premium
The relation between the industrial production and the equity risk premium seems evident when reviewing the correlation between the variables in the overall period. Correlation above 0.6 is observed for specific time periods within the sample. But when regressing the equity risk premium against the state of the economy, measured as the volatility and percentage change in the industrial production, the relation is not significantly observed for both the North Western European countries and the PIGS countries. However, excluding the crisis periods from the sample period results in a significant impact of the volatility in industrial production on the equity risk premium for North Western Europe (with exemption from Germany) and the PIGS countries. The fact that the Internet bubble in 2000 – 2001 and the financial crisis in 2007 – 2009 had a high overall impact on the countries included in the sample, might explain why including these periods bias the results. The uncertainty in the economy, measured in terms of volatility in industrial production, positively impacts the equity risk premium. More volatility in the economy leads to more insecurity for investors, and as a result higher perceived risk. The fact that an increase in the volatility in industrial production leads to an increase in the equity risk premium might therefore be due to investors requiring to be compensated for the higher perceived risk. According to Damodaran (2012), this result is not surprising, as not the level of growth in industrial production is of importance, but rather the uncertainty about the growth in industrial production. Viewing the results in the context of previous studies leads to contrary findings. Salomons and Grootveld (2003) found a relation between the industrial production and the equity risk premiums of the G7 countries without excluding crisis periods. This difference might be due to the fact that they used a different sample period, namely 1975 – 2001, whereas the sample period used in this study is dominated by two severe crisis periods.
In addition, using the OECD composite leading indicator led to a better fit according to Salomons and Grootveld (2003). This finding is in line with the findings of this study, which shows a significant impact of the composite leading indicator on the equity risk premium for each country included in the sample. This result proofs the robustness of the test, as the composite leading indicator is also often considered as an appropriate proxy for the economic condition.
related to the demand and supply of the currency and therefore often linked to the state of the economy. The effect might be more obvious for the PIGS countries as the macroeconomic effects of the financial and sovereign debt crises on these countries were more severe.
6. Conclusion
The main contribution of this study to the current literature is twofold. Firstly, this study aims to find evidence for differences between the equity risk premiums in North Western Europe and the PIGS. Differences between the average and median equity risk premiums in North Western Europe and the PIGS are tested. In addition, this study tests whether the equity risk premium significantly changed in North Western Europe and the PIGS after the subprime crisis of 2007.
Secondly, the thesis aims to find evidence for the effect of the economic state on the equity risk premium. In order to do so, several regressions are performed to test the relation between the equity risk premium and the industrial production, a proxy for the state of the economy. This study tests whether changes in the percentage change and volatility in industrial production, which defines a more turbulent economic state, results in a changes in the equity risk premium. The total sample includes 1791 index observations obtained through the MSCI index. Various control variables are added to the regressions in line with previous research, to avoid the omitted variables bias. Therefore, this study will include the volatility in the consumer price index, the percentage change in the short-term interest rate, the percentage change in the harmonized employment rate, the composite leading indicator of the OECD, the consumer confidence index, the USD/EUR exchange rate and lastly, the co-movement of the local index with the global economy. To test for robustness the Gauss Markov assumptions for the error terms are tested.
No significant difference between the average equity risk premium in North Western Europe and the PIGS is found, which is in line with the integrated market hypothesis (Harvey and Bekaert, 2002 and 2003; Shackman, 2006). If markets are fully integrated, the perceived unsystematic risk is diversifiable. A significant difference for the PIGS is found between the periods 1999-2007 and 2007-2015, which can be due to the severe impact of the financial and sovereign debt crises on this region.
premium. It is probable that an increase in volatility results in higher uncertainty and risk for investors, for which investors would like to be compensated leading to an increase equity risk premium.
References
Adrian, T. and Song Shin, H., 2010. The changing nature of financial intermediation and the financial crisis of 2007-09. Federal Reserve Bank of New York Staff Report 439.
Altug, S.J., 1983. Gestation lags and the business cycle, an empirical analysis. Carnegie-Mellon working paper, Stanford University, Pittsburgh.
Ameur, H.B., Gnégné, Y. and Jawadi, F., 2013. Measuring the time-varying equity risk premium in the context of financial crisis: do developed and emerging markets differ?. Applied Economics Letters 20, 1673 - 1677.
Arrow, K., 1971. Essays in the Theory of Risk-Bearing. Amsterdam, North-Holland.
Bansal, R. and Yaron, A., 2004. Risks for the long run, a potential explanation of asset pricing puzzles. Journal of Finance 59, 1481-1509.
Barberis, N., Huang, M. and Santos, T., 2001. Prospect theory and asset prices. Quarterly Journal of Economics 116, 1–53.
Benartzi, S. and Thaler, R., 1995. Myopic loss aversion and the equity risk premium puzzle. Quarterly Journal of Economics 110, 73–92.
Bekaert, G., Erb, C.B., Harvey, C.R. and Viskanta, T.E., 1998. Distributional characteristics of emerging market returns and asset allocation, analysing returns that cannot be summarized by a normal distribution. Journal of Portfolio Management 24, 102-116.
Blanchard, O., 1993. Movements in the equity risk premium. Brookings Papers on Economic Activity 24, 75–118.
Breusch, T.S., 1978. Testing for autocorrelation in dynamic linear models. Australian Economic Papers 17, 334-355.
Bollerslev, T., 1986. Generalized autoregressive conditional heteroscedasticity. Journal of Econometrics 31, 307-327.
Brandt, M.W. and Wang, K.Q., 2003. Time-varying risk aversion and unexpected inflation. Journal of Monetary Economics 50, 1457-1498.
Brooks, C., 2008. Introductory, econometrics for finance, second edition. Cambridge University Press, New York.
Brooks, R. and Yan, D.Y., 1999. London inter-bank offer rate (LIBOR) versus treasury rate evidence from the parsimonious term structure model. The Journal of Fixed Income 9, 71-83.
Brown, S., Goetzmann, W. and Ross, S., 1995. Survival. Journal of Finance 50, 853–873.
Campbell, J. and Cochrane, J., 1999. By force of habit, A consumption-based explanation of aggregate stock market behaviour. Journal of Political Economy 107, 205-251.
Dimson, E., Marsh, P. and Staunton, M., 2001. 101 years of investment returns. Millennium Book. London, London Business School and ABN Amro.
Dimson, E., Marsh, P., Staunton, M., 2005. Global evidence on the equity risk premium. Journal of Applied Corporate Finance 15, 27-38.
Duffie, D. and Stein, J.C., 2014. Reforming LIBOR and other financial-market benchmarks. Journal of Economic Perspectives 29, 191-212.
El Hedi Arouri, M. and Jawadi, F., 2010. On the impacts of crisis on the risk premium, evidence from the US stock market using a conditional CAPM. Unpublished paper.
Engle, R.F., 1982. Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica 50, 987-1007.
Engle, R. F., 2001. GARCH(1,1): the use of ARCH/GARCH models in applied econometrics. Journal of Economic Perspectives 15, 157-168.
Fama, E., 1970. Efficient capital markets, a review of theory and empirical work. Journal of Finance 25, 383-417.
Fase, M., 1997. The risk premium on stocks in the European Union. Dutch Central Bank Research Memorandum 487.
Friend, I. and Blume, M., 1975. The demand for risky assets. American Economic Review 65, 900–922. Gibson, R. and Mougeot, N., 2004. The pricing of systematic liquidity risk, empirical evidence from the US
stock market. Journal of Banking and Finance, 28, 157–78.
Gielen, G., 1994. Konen Aktienkurse noch Steigen?. Wiesbaden, Ger., Gabler
Ghysels, E., Sinko, A. and Valkanov, R., 2007. MIDAS Regressions, further results and new directions. Econometric Reviews 26, 53-90.
Godfrey, L.G., 1978. Testing against general autoregressive and moving average error models when the regressors include lagged dependent variables. Econometrica 46, 1293-1301.
Goetzmann, W. and Jorion, P.J., 1999. Global stock markets in the twentieth century. Journal of Finance, 54, 953-980.
Gordon, M., 1962. The investment, financing and valuation of the corporation. Homewood, Ill., Richard Irwin. Gourio, F., 2012. Macroeconomic implications of time-varying risk premia. ECB Working Paper Series, 1463. Harvey, C. and Bekaert, G., 2002. Research in emerging markets finance: looking to the future. Emerging
Markets Review 3, 429–448.
Harvey, C. and Bekaert, G., 2003. Emerging markets finance. Journal of Empirical Finance 10, 3–56. Harvey, C.R., 1989. Time-varying conditional covariances in tests of asset pricing models. Journal of
