Challenging CAPM: Including Imported Political Risk
into the Equation
Author: W.G. Tigranyan 1829378
Supervisor: J.O. Mierau
University of Groningen, Faculty of Economics and Business
Date: April 22th, 2013
ABSTRACT:
The recent developments in the financial world raised questions about the practicality and
reliability of the Capital Asset Pricing Model (CAPM). This paper investigates the reliability of
CAPM in context of 54 countries around the world. In addition, a variable is included to measure
Political Risk exposure of a country to measure its effect on CAPM. The results indicate that the
market risk premium is indeed one and the Imported Political Risk (IPR) has significant quadratic
effect on country index returns. Furthermore, market portfolio and IPR together do not explain all
Page | 1
The recent developments in the financial world, such as credit crunch in 2007, resulted in a boom
of academic articles concerning the causes and the effects of the credit crunch (Attinasi II,
Checherita-Westpal and Nickel, 2009; Goodhart, 2008; Hall, 2008). In addition, European
Central Bank implemented BASEL III in order to limit the risk taking behavior of the banks and
other financial institutions. To elaborate the intuition behind the regulation, one should keep in
mind that financial institutions work with valuation models, from which Capital Asset Pricing
Model (CAPM) is the backbone underlying all models. CAPM is developed gradually by Treynor
(1961) and Sharpe (1964) and extended by Lintner (1965a; 1965b), Mossin (1966), Fama (1968a;
1968b) and Long (1972). One of the assumptions of CAPM is the information symmetry which
results in correct valuation of securities. However, a crisis in any form indicates correction of the
market prices and if one believes in efficient financial markets with symmetric information
availability, than there should be no market corrections or crises in the long run, as the
information symmetry forces the security prices to correct gradually.
As the market mechanism indicates market inefficiencies, the question is raised why in the recent
financial credit crunch of 2007 large institutional investors would believe that for example
CDO’s offered higher risk adjusted returns than what market would indicate. This is against CAPM, which states that risk adjusted returns for all securities are constant.
Because the scholars dedicated much of their time to investigate the dynamics of market
inefficiencies, this paper will concentrate on more fundamental issue. This paper will investigate
the efficiency of CAPM in the context of information flow. The advances in telecommunication
resulted in better accessibility of relevant information by investors, resulting in faster market
reaction on news. When considering that the news regarding the political developments is
Page | 2
in the security prices in short amount of time. Changes in political risk should therefore have no
significant influence on security return in the long run. In order to test this, an index of PR is
designed such that it will capture the exposure of a country towards the political risk of its trade
partners. The Imported Political Risk (IPR) is assumed to be most publically known information,
which should be easily incorporated in the prices of securities1, therefore, when including IPR in
the research, it would not result in significant result. Many researches have investigated the
effect of political event on stock market volatility (Cutler, Poterba and Summers, 1989;
Bittlingmayer, 1988; Chan and Wei, 1996; Kim and Mei, 2001) and stock market performance
(Erb, Harvey and Viskanta, 1995, 1996a, 1996b; Cosset and Suret, 1995; Bekaert, 1995, Bekaert
and Harvey, 1997), however, those researches were based on single event and single country. For
example, Beaulieu, Cosset and Essaddam (2006) investigate the short run effect of 1995 Quebec
referendum on the stock return of Quebec companies. Therefore, to my knowledge, there are no
cross country security pricing studies investigating the information efficiencies of markets in
context of IPR. Accordingly, this research will investigate whether the information symmetry
prevails in context of IPR. In addition, test will be conducted to see whether the risk is rewarded
linearly.
This paper is structured as follows: Section A will describe the literature and define the
hypotheses, thereafter the methodology will be justified in Section B, which will be followed by
the data definition in Section C, the results of the research will be summarized in Section D and
then the conclusion and discussion in Section E.
Page | 3
A. Literature Review
A concept of equity valuation is developed in the 60’s. Initially, Treynor (1961) and Sharpe
(1964) introduced the basics of valuation model, which has been extended and improved futher
by Lintner (1965a; 1965b), Mossin (1966), Fama (1968a; 1968b) and Long (1972). The model,
nowadays known as Capital Asset Pricing Model (CAPM), states that the return on risky assets
depend only on equity specific risk; where the return is linearly related to market portfolio. In
order to be able of valuing equities CAPM adopted strong assumptions. CAPM assumes that (i)
there are no transaction costs or taxes, (ii) all important information concerning the equity is
publically available to all investors at no costs, (iii) the investors care only about the return and
the volatility of securities, and (iv) all investors can borrow and lend at risk free rate at all time
(Fama, 1970). The assumption that investors care only about the return and the risk of a security,
also known as homogenous beliefs assumption, is based on three pillars: (i) investors are assumed
to be rational and value securities rationally, (ii) when the investors are not acting rationally, their
investments are random and cancel each other out, and as last (iii) when the randomness does not
result in cancelation of mispricing, the professional arbitrageurs will cancel out the mispricing
(Fama, 1970).
The CAPM provides a simple and easily implementable framework for security pricing.
Elaborating more in detail, one can see that the first assumption is always violated as all investors
have transaction costs, and although some countries have no taxes on capital gains and dividends,
such as United Arab Emirates, considerable part of the countries have taxes. In addition, Shleifer
and Vishny (1997) prove the inability of arbitrageurs to bring the prices of securities back to their
fundamentals. In addition, Errunza and Losq (1985) and King (1993) prove the existence of
Page | 4
work of Blitz and van Vliet (2007) prove that securities with historically low volatility have
considerably higher risk adjusted return than high volatility securities in terms of Sharpe ratio and
CAPM alpha.
More importantly, the disagreements about the assumptions of information availability and
homogeneous beliefs of investors have created the field of behavioral finance where these
assumptions are discussed and tested intensively. In his book “Inefficient markets” Andrei
Shleifer (2000) provides an extensive summary of the literature concerning efficient market
theory and empirics.
CAPM predicts all efficient securities to lie on Security Market Line (SML), as described below.
( ) ( ) (1)
where ( ) is the expected return of security i, is the risk free rate, is the return on market portfolio and is the idiosyncratic risk of security i. Hence the expected return depends only on as the firm specific factor. The SML is illustrated in Figure 1 where market portfolio has beta equaling one. Any mispricing will trigger actions from arbitrageurs, thereby forcing the
prices to fall back to their fundamental values, balancing the correct pricing of securities. In the
short run, there might be inconsistencies with the CAPM, as the market needs time to recognize
the mispricing and react on it; however, when taking long perspective, all mispricing should have
been eliminated when the correction mechanism functions properly. Therefore, when the markets
are efficient, the individual securities represent together the market; hence have on average beta
of one. In order to test whether the securities are priced according to CAPM, the first hypothesis
is developed as:
Page | 5
Figure 1: The relation between expected return of a security against Beta of the security with market portfolio return of 15% and risk free return of 3%. The black dot represents the market portfolio.
Arguably, the most debated assumption of CAPM is the information availability to all investors.
It is difficult to belief that all investors have the same information always. The fact that almost all
countries imposed legislative restrictions for trade with insider information suggests information
asymmetry or at least different costs to gain the information. The availability of information
assumption is divided into three separate hypotheses: weak, semi-strong and strong forms (Fama,
1970). The weak form of efficiency states that the investors have only access to the information
of past performance and risk for certain security and attach the price based on that information.
The semi-strong information efficiency states that all publically known information is
incorporated in the price of the security; therefore no risk-adjusted excess returns can be obtained
in the long run. The strong form of information efficiency states that even insider information
will be incorporated in the security prices, as the information always leaks to the public.
-60.00% -40.00% -20.00% 0.00% 20.00% 40.00% 60.00% -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 R et u rn o f se cu ri ti es w it h gi ve n b et a Beta of securities
Page | 6
Nowadays, the improvements in information and communication sector have led to fast
distribution of information worldwide. Even the political event in foreign countries is becoming
known by the public in very short amount of time. The ease of obtaining information regarding
political changes should have direct impact on security prices. Therefore, including the change in
the foreign political risk into CAPM framework should not influence the security prices in the
long run.
Companies having operations in foreign countries are faced with the foreign political risk2 on
everyday basis (Kobrin, 1979), which will eventually influence the firms performance when not
managed properly (Erb, Harvey and Viskanta, 1996a). A clear example of foreign political risk
influence on a company’s performance is the case of Royal Dutch Shell in Sakhalin, Russia in 2006, when, off the records, Russian government imposed a two billion affine to Shell for
violating environmental standard, which resulted in withdrawal of Shell from the region. In this
case, Shell did not anticipate the influence of the government in large commercial projects in
Russia which might give some strategic benefits to the political power of Russian Federation.
This major event is in line with Clark and Tunaru (2005) and Simon (1973), who find that
political events have significant negative economic and financial consequences.
Many researches have investigated the effect of political event on stock market volatility (Cutler,
Poterba and Summers, 1989; Bittlingmayer, 1988; Chan and Wei, 1996; Kim and Mei, 2001) and
stock market performance (Erb, Harvey and Viskanta, 1995, 1996a and 1996b; Cosset and Suret,
1995; Bekaert, 1995, Bekaert and Harvey, 1997), however, those researches were based on short
run, single event, domestic political risk or one country events. For example, Beaulieu, Cosset
Page | 7
and Essaddam (2006) investigate the short run effect of 1995 Quebec referendum on the stock
return of Quebec companies. To sum up, the political risk itself might not have as much impact
on a firm’s performance as the change in the foreign political risk or the exposure towards it. Related to the assumptions of CAPM regarding to the information availability, the political risk is
categorized in the semi-strong assumption of information. The political events are always
highlighted in the media, which might take less than half hour in order to recognize and broadcast
to the investors, hence in the long run, the change in the Imported Political Risk (IPR) should not
have significant explanatory power for the firm’s or country’s performance. Moreover, political
risk is country specific; therefore, it can be diversified when investing internationally; meaning
that the international investor will not price the IPR as idiosyncratic risk, whereas the national
investor will.
Although the political risk is converging over time (Diamonte, Liew and Stevens, 1996),
Pukthuanthong-Le and Visaltanachoti (2009) found significant pricing of idiosyncratic risk on
firm level, leading to significant pricing of political risk on country’s largest index. In the long
run IPR should have no effect on country index return. When IPR explains the variations of the
index returns, it violates the semi-strong assumption of information availability, thereby proving
the limited use of CAPM by practitioners. Formally, hypothesis 2 is formed as:
When considering different risk adjusted returns for securities (Errunza and Losq, 1985), one can
expect that the same nonlinear relation also holds also for IPR. To explain more in detail with
Page | 8
amount3, it is less likely that Shell would react double as fierce. Most probably Shell would leave
the country and stop all its operation in Russian Federation for long time. In accordance with this
logic, the third and last hypothesis is formed as:
Page | 9
B. Methodology
Within CAPM, the expected returns of any securities are described by the SML, formula (1). In
the setting of this research, the world is the market with individual securities on SML. Country
indices represent efficient portfolios consisting of securities listed in that country. When taking
into account the risk free rate requirement of SML, Formula (1), it becomes impossible to find
risk free rate applicable to every country (Solnik, 1974). Therefore, this research will employ the
methodology used by Jorion (1990), which does not require the specification of universal risk
free rate. The regression model to be estimated can be written as:
̃ ̃ (2)
where ̃ is the index return of country i at time t, is the regression intercept for country i, is the world equity market beta, which should equal one, ̃ is the return of market portfolio at time t, and is the disturbance term for country i at time t. In order to understand what measures, formula (1) is rearranged, as illustrated by formulae 3 – 6. ̃ ̃ (3)
̃ ( ) ̃ (4)
formula (2) – formula (4) => ̃ ̃ ( ) ̃ ̃ (5)
( ) (6) In context of CAPM, should equal one. Regardless of what the value of risk free rate is,
should always equal to zero when market beta equals one. In case that the intercept is anything
Page | 10
problem, disproving automatically the CAPM. For example, Ross (1976) proposed using
Arbitrage Pricing Theory (APT) as a substitute for CAPM. APT is a multifactor model where
also macroeconomic factors are included for asset pricing.
For investigation of IPR impact on the return of country index and the nonlinear relation of IPR,
the regression models become:
̃ ̃ (7)
̃ ̃ (8)
Page | 11
C. Data
The data of this research is limited to the availability of country bilateral trade weights and
political risk measures. While Bank of International Settlements (BIS) provides country level
bilateral trade weight for 61 countries4 from 1993 until 2010, the Political Risk Service Group
(PRS Group) provides data for political risk measures for 146 countries from 1984 until 2008 in
its International Country Risk Guide (ICRG). Therefore, when excluding Euro Area as a separate
country, the combined dataset includes 60 countries with data from 1993 until 2008. However,
not all countries have equity markets indices starting 1993; therefore, the countries which
introduced indices after 2000 are excluded from this research. The year 2000 is chosen in order to
have at least 50% of the observations included in the time series; hence the dataset includes at
least eight years observations. When deleting the countries with missing country index values, 54
countries remain in the dataset with a total of 847 observations. When accounting for one year
loss when calculating the return and maintaining the 50% criteria, 777 country – year
observations remain. The summary of countries and their indices with corresponding available
observations is illustrated in Appendix A.
C.1. Country Performance
Country performance is measured by a country’s dominant large cap index, as illustrated in
Appendix A. The large cap index prices are obtained on annual basis from Thomson Datastream
database. Ideally, one should include the total return indices. However, the total return indices are
introduced quite recently, resulting in many missing variables for the years 1993 – 2008. As a
result, only price indices are employed.
Page | 12
Using only price indices may lead to biased results, as it automatically assumes that the capital
gains are the most important measures for performance, thereby neglecting the dividend yield.
However, when comparing the price indices with the corresponding total return indices for
available countries, the correlations are very high5. More thorough investigation indicates that the
variation between the total return and price indices are also very similar. As Thomson Datastream
provides index prices instead of the returns, the index return for country i at time t is calculated
using formula (9).
(
) (9)
When investigating individual indices, some very interesting developments are observed. Figure
2 illustrates the developments in the two fastest growing equity indices; namely MSCI Brazil and
Istanbul SE National 100. When comparing with Figure 3, MSCI World had average
continuously compounded return of 7.48% while MSCI Brazil had average return of 45.11% and
Istanbul SE National 100 50.93%. The highest increase yielded in Brazil in 1993, accounting for
386.22%. The worst performing country in the dataset is Thailand which had on average negative
annual return of 3.23%. It is interesting to mention that after the financial crisis in 2007, without
exception all indices declined sharply. The minimum decline was the Venezuelan market by
7.06% and maximum decline was astonishing 194.19%6. In comparison, MSCI World had a
return of -67.87% and DJGL World -69.67%. This suggests the existence of outliers. In order to
prevent distortion in analysis results, the outliers are excluded from the analysis by using dummy
variable. Section C.4. will explain the construction of the dummy variable.
5
The correlation coefficient between the price indices and the corresponding total return indices are well above 0.9.
Page | 13
Figure 2: Equity market developments in Brazil for MSCI Brazil and Turkey for Istanbul SE National 100 in logarithmic scale. MSCI Brazil grew on average by 45.11% from 1993 until 2007 to fall by 51.25% in 2008. Istanbul SE National 100 grew on average by 50.93% from 1993 until 2007 to fall with 58.03% in 2008.
Figure 3: Equity market developments in MSCI World, Thailand for Bangkok S.E.T. and Philippine for Philippine SE I in logarithmic scale. MSCI World grew on average by 7.48% from 1993 until 2007 to fall with 50.18% in 2008. Bangkok S.E.T. decreased on average by 3.23% from 1993 until 2007 to fall with 51.77% in 2008 and Philippine SE I grew on average by 2.75% from 1993 until 2007 to fall by 49.47% in 2008.
1 10 100 1000 10000 100000 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Index developments for Brazil and Turkey
Brazil Turkey 1 10 100 1000 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Index developments for the world, Thailand and
Philippine
Page | 14
C.2. Imported Political Risk
Imported Political Risk (IPR) is the average of political risk exposure of a country towards its
trade partner, weighted with bilateral trade weights. The political risk measures are obtained from
ICRG provided by PRS Group, while the trade weights are obtained from BIS.
The PRS Group publishes ICRG on annual basis, which includes measures for political risk7
through 12 components. By employing the ICRG data, this research follows the steps of
Diamonte, Liew and Stevens (1996), who studied the political risk in emerging and developed
countries. In addition, Erb, Harvey and Viskanta (1996b) summarize various measures for
country risk with their reciprocal differences (Exhibit 7) between January 1984 and September
1995. The major advantage of ICRG political risk indicator is that it encompasses almost all
aspects of country risk; therefore, it is a proxy for country’s total risk index.
The ICRG country political risk includes 146 countries from 1984 until 2008. The PRS Group
attaches different values to different components, depending on the importance of the component.
The weight of risk components and the meaning of the indices are summarized in Table 1.
Realizing the maximum score of 12 points, the total Political Risk (PR) is calculated as the
average of 12 components of PR weighted against the maximum points the component can
receive. The calculation is summarized by formula (10).
(∑ ) (10)
where is the index score for risk component for country i at time t, Point is the maximum point the component can score, and is the total political risk of country i at time t.
Page | 15
Table 1: The weights used for calculation of total political risk index (left) and the relative meaning of the index (right)
Risk Components Weight Political Risk Assessment Government Stability 12 Very High <49.9%
Socioeconomic Conditions 12 High 50%-59.9%
Investment Profile 12 Moderate 60%-69.9%
Internal Conflict 12 Low 70%-79.9%
External Conflict 12 Very Low 80%-100%
Corruption 6
Military in Politics 6
Religious Tensions 6
Law and Order 6
Ethnic Tensions 6
Democratic Accountability 6
Bureaucracy Quality 4
Note 1: The highest weight in the components is 12, indicating the high importance of the component for the total political risk. Some components are constructed by summing other sub-components. For
example, Government Stability is the sum of three different sub-components: Government Unity, Legislative Strength and Popular Support. Every subcomponent receives at maximum a value of 4, resulting in a maximum of 12 points for Government Stability. Compared to Bureaucracy Quality, Government Stability is more important for political measure, because unstable governments result in regime change while Bureaucracy Quality results in inefficiencies in the government. Poor Bureaucracy Quality results rarely in political uncertainty.
Note 2: Political Risk Assessment illustrates the relative meaning of the total Political Risk. Lower index value corresponds with higher risk, while higher indicates lower risk.
In order to calculate IPR, data is needed to describe a country’s exposure towards other partner
countries. Jorion’s work (1990) is used as guideline in this context, where the revenues of the
multinationals from foreign countries were used as the weight of exchange rate exposure.
Although being a solid measure of foreign exposure, this methodology is only applicable at firm
level and in countries where the revenues from foreign operations are obliged to be reported. This
research uses country level data, which makes Jorion’s (1990) application impossible.
More detailed investigation suggests that countries import risk from their foreign partner relative
Page | 16
countries from 1993 until 2010. The weighting matrix is composed such that the weights between
the years 93-95, 96-98, 99-01, 02-04, 05-07, and 08-10 are constant. In order to be able of
constructing the total IPR, assumption is made that the trade is roughly constant over relatively
short period of time; therefore the values of bilateral trade can be used as yearly inputs. For
example, Figure 4 illustrates the development of trade with Euro Area for several important trade
partners8. One can observe that the trade with China is increasing rapidly, resulting in higher
exposure of Euro Area towards the developments in China. The increasing share of China seems
to be at the expense of other countries, as their relative share in Euro Area is declining.
Figure 4: Trade development of selected countries with Euro Area
The downside of BIS dataset is that the bilateral trade of Euro Area member countries with total
Euro Area is not provided; therefore, it is impossible to construct political risk for Euro Area
based only on BIS dataset. However, when realizing that the political risk is directly linked to the
8 Total share of the six countries to world GDP (current USD) fluctuates around 50%. 0.0 5.0 10.0 15.0 20.0 25.0 1993_1995 1996_1998 1999_2001 2002_2004 2005_2007 2008_2010
Trade with Euro Area
Page | 17
governing body of a country, it can be stated that the political risk of Euro Area is transferred
trough a common governing structure. The Euro Area countries are unified in European
Parliament as the main governing body where from the local political risk is being transferred to
the whole European Union. Therefore, the total political risk of Euro Area can be calculated by
taking the weighted average political risk of Euro Area countries with the weight being the share
of representatives of Euro Area countries in the European Parliament. Appendix B summarizes
the representatives of Euro Area countries in the European Parliament with their corresponding
weights. For example, Germany’s contribution to the total political risk of Euro Area can be calculated as follows: Germany has 20.63% of total representatives from the Euro Area and its
political risk in 1994 was 83.80, which results in 20.63%*83.80=17.28 as Germany’s
contribution of the political risk in the Euro Area. Hence the calculation can be summarized by
the formula:
∑(
), (11)
where is the total political risk of Euro Area at time t and
is the weight of risk exposure
based on political representation in the EU for country i at time t.
Once the political risk of the Euro Area is constructed, the Imported Political Risk (IPR) can be
calculated by using the formula:
∑( ), (12)
Page | 18
As IPR is publicly known information, a constant IPR will not influence the return. However,
when there is a change in IPR, market are expected to react on the change. The yearly change in
IPR is obtained by employing formula (9).
The calculation of IPR for USA will illustrate in detail the steps of IPR index construction.
Assume world consists of five countries: China, France, Germany, Greece and USA. The first
step is to calculate the political risk of the countries. The steps and results are summarized in
Table 2. The rows for score are the raw data received from ICRG. The weighted scores are
obtained by dividing the scored by the maximum points per component. After summing the
weighted scores, unscaled political risk is received; this is rescaled by multiplying with 100/12 in
order to receive the total political risk, as explained by formula (10).
Table 2: Calculation of total Political Risk for China, France, Germany and Greece
Components 1 2 3 4 5 6 7 8 9 10 11 12 Max points 12 12 12 12 12 6 6 6 6 6 6 4 S cor e China 10.7 7.8 7.0 9.6 10.0 2.5 3.0 5.0 4.5 4.5 1.5 2.0 France 8.8 7.5 11.8 9.7 10.0 5.0 5.5 4.0 5.0 2.5 6.0 3.0 Germany 10.0 8.0 11.9 11.1 10.5 5.0 6.0 5.0 5.0 4.0 6.0 4.0 Greece 6.9 7.7 10.5 8.8 10.2 2.0 5.0 5.0 4.5 5.0 6.0 3.0 W eight ed S cor e China 0.89 0.65 0.58 0.80 0.83 0.42 0.50 0.83 0.75 0.75 0.25 0.50 France 0.73 0.63 0.98 0.81 0.83 0.83 0.92 0.67 0.83 0.42 1.00 0.75 Germany 0.83 0.67 0.99 0.93 0.88 0.83 1.00 0.83 0.83 0.67 1.00 1.00 Greece 0.58 0.64 0.88 0.73 0.85 0.33 0.83 0.83 0.75 0.83 1.00 0.75 Unscaled PR (sum weighted score) Total PR (x100/12) T otal P R China 7.76 64.65 France 9.40 78.33 Germany 10.46 87.15 Greece 9.01 75.07
Note 1: Max points refer to the maximum point a country can score per component.
Page | 19
At this stage, the total PR is known, however, according to BIS dataset9, USA trades for 40%
with Euro Area and 60% with China. As the PR of Euro Area is not known, the IPR cannot be
calculated. Assuming Euro Area consists only of Germany, France and Greece, who are
represented in the European Parliament by respectively 50%, 30% and 20% of the seats, the PR
can be approximated using formula (11).
In this case, USA imports 40% of 82.09 from Euro Area and 60% of 64.65 from China, resulting
in IPR of 77.88. The calculation is similar for all other countries in the dataset.
C.3. Market Performance
Market performance reflects the world equity market performance. This research employs two
measures for world performance: MSCI World and DJGL World. Both of them are denominated
in US dollar. Both of the indices will be used in simultaneously in order to cross check the
robustness of the results. The return on market portfolio is calculated by employing formula (9).
The proxies for world benchmarks are somewhat different regarding the exposure of index
towards several countries and industries, as illustrated by Appendix C. Although the benchmarks
are somewhat different, one cannot predict which one approximates the world market better,
however, it is not excluded that the regression results will yield slightly different results.
C.4. Outliers
As discussed before, this research will control for outliers by using dummy variable. The
criterion of outlier used in this research is:
Page | 20
̅ (13)
where, ̅ is the mean of country index returns and is the standard deviation of index returns. Three time positive and three time negative deviation from the mean will cover 99.7% of the
dataset, when assuming that the returns follow normal distribution. When the observations are
identified as outlier, the dummy variable receives value of one, otherwise zero. As a result, 43
outliers are identified from a total of 777 observations.
C.5. Domestic Political Risk
This research will also control for domestic Political Risk (PR). How the domestic PR is
influencing valuation practices is described by Agmon and Findlay (1982). The main argument
how domestic political risk will affect stock valuation is by increasing inflation. The mechanism
described by Agmon and Findlay (1982) is that weak governments, which are characterized by
high political risk, will grant subsidies to certain domestic companies, which will increase the
governmental spending. The government tries to close the gap in budget deficit, caused by large
subsidies. As the tax income is the main source of income for government, and the tax rates are
sticky in the short run, the government will use inflation as the tax mechanism of last resort. The
increase in inflation will force the discount rate for stock valuation to increase, resulting in
decrease of stock prices.
Table 2 illustrates the construction of total political risk, which is converted to domestic PR by
Page | 21
D. Results
Initially there were 60 countries involved in the research. However, because of lacking data 54
countries remained in the dataset with 15 yearly (1994 – 2008) observations, accounting for 777
country-year observations. Table 3 summarizes the correlation matrix and descriptive statistics of
the data employed in this research. Frankly, none of the variables seem to follow normal
distribution. Interesting is to mention the difference between the minimum and the maximum of
the returns between 1994 and 2008. MSCI Brazil faces an incredible return of 288% in 1994
whereas OMX Iceland all share face devastating decrease of 194%10. These large differences
indicate the existence of outliers which will influence the regression results. The influences of
these outliers are eliminated by using a dummy variable. As can be seen, the correlation between
the Return and IPR is very low at 0.14. Further, the correlation between world and country
indices are fairly high, indicating that the world indices explain a large part of the movements in
the country indices.
In order to analyze the relation and magnitude of IPR on country’s index, stepwise regressions is conducted using formulae (2), (7) and (8). Using two different world index proxies and three
models will result in six regressions. The regressions are summarized in Table 4.
10 Based on continuous compounding ( (
Page | 22
Table 3: Correlation matrix and descriptive statistics
Variable 1 2 3 4 5 6 1 Return 1.000 2 IPR 0.140 1.000 3 IPR2 0.043 -0.344 1.000 4 MSCI World 0.623 0.188 -0.084 1.000 5 DJGL World 0.640 0.176 -0.085 0.996 1.000 6 Domestic PR -0.061 0.245 -0.094 -0.001 -0.002 1.000 Mean 0.060 -0.001 -0.000 0.021 0.021 0.001 Standard Deviation 0.374 0.014 0.000 0.223 0.225 0.037 Minimum -1.942 -0.064 0.000 -0.678 -0.697 -0.367 Maximum 2.881 0.053 0.004 0.178 0.203 0.148 Skewness -0.402 -0.553 5.535 -2.050 -2.126 -1.435 Kurtosis 9.749 4.712 43.016 6.638 7.025 17.429 Note 1: N=777
Note 2: 9 countries provided index data later than 1993, which resulted in incomplete return data from 1993 until 2008. Missing countries are: Bulgaria until 2001, Croatia until 1998, Czech Republic until 1995, Estonia until 1997, Latvia until 1997,
Luxembourg until 2000, Malta until 1997, Romania until 1998, and Russian Federation until 1996.
Table 4: Regression model specification Model Specification 1 ̃ ̃ 2 ̃ ̃ 3 ̃ ̃ 4 ̃ ̃ 5 ̃ ̃ 6 ̃ ̃ Note 1: ̃ is the return on MSCI World in year t
Note 2: ̃ is the return on DJGL World in year t
Note 3: is the dummy variable of outliers with the value of one if ̃ exceeds the mean with more than three times the standard error and zero otherwise
Page | 23
The initial results of pooled OLS are summarized in Appendix D.1. The pooled OLS model show
very significant effects of IPR and IPR2 in the CAPM setting. Furthermore, the constant is as
predicted zero with slightly significant, negative coefficient for Domestic PR, indicating that a
country’s risk free rate is determined by the country risk characteristics. Even the test of coefficients proves the beta of the market is one. In order to see whether the results are BLUE,
diagnostic tests are conducted. The results of Breusch – Pegan / Cook – Weisberg test of
heteroscedasticity shows homoscedastic disturbance terms in Models 1 – 3 and 6. It is surprising
how the P-values of heteroscedasticity test change when including the variables for IPR and IPR2.
In Models 1, 2, 4 and 5 in Appendix D.1 the P-values of heteroscedasticity are very low until the
inclusion of political risk variables. Even when looking the differences between Models 5 and 6,
initially inclusion of IPR in Model 5 seems to have limited effect on the regression output;
however, when introducing IPR2 to correct for nonlinear relation between IPR and Return, the
P-value for heteroscedasticity increases dramatically from 0.085 to 0.999, almost perfect
homoscedastic. Further, the tests for autocorrelations indicate absence of autocorrelation. As last,
when investigating multicollinearity in Appendix D.2, the correlations between the disturbance
terms with its one year lagged values are very low.
At first glance, the pooled OLS is BLUE, however, when investigating the autocorrelation for
separate countries, in Appendix D.3, and years, Appendix D.4, the regional and periodical
autocorrelations are obvious. For example, Thailand has correlation coefficient of 0.559. For the
total market, the correlation coefficient of the disturbance term between the years 2006 and 2007
is 0.486. The model seems to suffer from positive and negative autocorrelation, depending on the
country and year. In addition, when looking to the scatter plot of disturbance term of Model 3
Page | 24
probable heteroscedasticity in the long run. To correct for the autocorrelation and
heteroscedasticity, General Least Squared (GLS) regression analysis will be conducted taking
into account Durbin Watson d-statistic and ARCH(1) effect (Brooks, 2008). However, before
proceeding to the GLS regression, decision needs to be made concerning the use of fixed or
random effects by using regional and income dummy variables.
The definitions of regional dummy variables are summarized in Appendix E. The dummy
variable of North America is not included in the analysis in order to avoid the dummy trap. The
dummy variable of income level is based on current GNI per capita, using World Bank Atlas
method11. The dummy variable receives a value of one when the GNI per capital exceeds 12.476
USD, which is the level at which World Bank classifies the countries as high income country.
Hence, the dummy variable can take different values within the time span and country, making it
dynamic dummy variable, representing the state of a country for a specific year.
The results of robust OLS regression with the dummy variables are illustrated in Appendix F. The
results of the regressions indicate the presence of regional effects, as the dummy variable of Asia
is significantly positive and the joint hypothesis test of regional coefficients prove the existence
of regional effects on country returns. In addition, the income level also has significant effect on
a country’s return. Based on Appendix F, conclusion can be drawn that there are fixed effects in the dataset, therefore, the final GLS regression analyses are conducted taking into account fixed
effects.
Page | 25
Table 5: GLS panel data regression with control for ARCH effect, Durbin Watson d-statistic and Fixed Effects
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
IPR 0.590 2.829** 0.567 2.779** (0.791) (0.933) (0.775) (0.913) IPR2 126.5*** 127.1*** (29.89) (29.25) MSCI World 1.019*** 1.014*** 1.022*** (0.0442) (0.0450) (0.0447) DJGL World 1.026*** 1.021*** 1.029*** (0.0425) (0.0432) (0.0429) Outliers -0.540*** -0.534*** -0.507*** -0.517*** -0.512*** -0.487*** (0.0954) (0.0955) (0.0944) (0.0935) (0.0936) (0.0925) Domestic PR -0.270 -0.321 -0.286 -0.284 -0.334 -0.299 (0.286) (0.295) (0.293) (0.280) (0.289) (0.287) Intercept 0.0456*** 0.0473*** 0.0306** 0.0445*** 0.0462*** 0.0294** (0.0101) (0.0104) (0.0108) (0.00994) (0.0102) (0.0106) R2 Within 0.411 0.413 0.428 0.432 0.435 0.451 R2 Between 0.000 0.000 0.013 0.000 0.000 0.022 R2 Overall 0.392 0.394 0.410 0.413 0.416 0.433 Note 1: N=723
Note 2: Standard errors in parentheses Note 3: * p<0.05, ** p<0.01, *** p<0.001
Table 6: P-values of hypotheses tests
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Page | 26
Table 5 illustrates the final outcome of the analysis, whereas Table 6 summarizes the results of
hypotheses tests. For Models 1 and 4, the CAPM seems to hold for world market beta, as the
hypothesis tests fail to prove that the beta is other than one, the intercept is significant which
indicates that the risk free return is not dependent on domestic political risk. When introducing
IPR, the results do not change much in terms of significance. However, this is caused by
misspecification of functional form of IPR. When including IPR and IPR2 in Models 3 and 6, the
results change dramatically. The measure for IPR becomes very significant. When IPR increases,
the imported political environment becomes more risky, the return of country’s index also
increases, which is in line with higher risk higher reward perception. However, the risk adjusted
return is not constant, as IPR2 is significantly positive. Furthermore, proxies for market
performances do have a beta of one, as theory suggests. However, the intercept is significantly
positive and the domestic political risk is not. This indicates that the risk free return is not
explained only by domestic political risk. In addition, the R2 is always somewhat higher than
40%, which leaves almost 60% of the variation unexplained. When taking into account that the
dataset contains yearly observations, the noise or inefficiencies in the dataset are brought to
minimum. This means that the 60% unexplained variance is impossible to attribute to noise or
inefficiencies in the market. Therefore, it is safe to conclude that extending CAPM with IPR fails
to explain the market dynamics fully; hence there are other important factors which are not
included in this paper. The inclusion of IPR and IPR2 improves the explanatory power of the
model by only 2%. To sum up, the market has a beta of one; however, country index returns do
also depend on IPR which is not rewarded linearly.
In order to ensure the consistency of the results, this research employs two proxies for world
Page | 27
different method is also used to crosscheck the results. Appendix G.1 shows the final regression
results, based on the methodology of Fama and MacBeth (1978), whereas Appendix G.2 shows
the hypotheses tests of the models. The difference between the main methodology and Fama and
MacBeth (1978) is that they use excess returns of equities and markets, whereas Jorion (1990)
uses the returns directly as inputs for the model. The different approach yields in the same results
with somewhat better R2. Hence the results are robust across different methodologies and
Page | 28
E. Conclusion and Discussion
Current financial and political developments, such as credit crunch in 2007 and sovereign debt
crisis of Greece in 2010, indicate market inefficiencies with respect to information flow. CAPM
predicts correct market pricing under the assumption of information symmetry. Furthermore, the
excess return of a security should depend only on the risk of security. When considering the
correct security pricing under the assumption of symmetric information flow, the market should
correct the security prices gradually over time. When this is true, the mispricing should not
accumulate to create crises.
This paper has investigated the risk characteristics of country indices’ returns and the influences of political changes on index returns. Initially, the CAPM seemed to hold ground when the
measure of IPR was introduced, with exception of constant risk adjusted return of indices.
However, when including IPR2, the results changed significantly. The results of regression
analyses accept Hypothesis 1, that the market has indeed a beta of one, whereas Hypotheses 2
and 3 are also proven to be true, stating that the IPR is not initially priced and the pricing is not
constant in terms of risk adjusted return. In addition, there seems to be some regional influences.
For example, Asian indices yielded on average lower returns than elsewhere. Also the high
income countries had lower returns on their country indices.
When looking at researches investigating pricing theories with firm level data, price to book ratio
and price to earnings ratio seem to have significant influence on firm return (Danielson and
Dowdell, 2001). Even more fundamental, Fama and French (1992) show that the firm’s beta fails
to predict the returns in the year 1963 – 1990. In this setting, this research does not deviate from
prior research as different regions have different risk rewards (Errunza and Losq, 1985; Garcia
Page | 29
more appropriate to test security pricing theory, as the intercept is significantly positive, thereby
supporting Arbitrage Pricing Theory (Ross, 1976). In order to see what causes the significant
intercept, more research should be done using macroeconomic variables. One of the
macroeconomic factors could be the measures of international investments (Ferson and Harvey,
1993), which can be related to the semi-strong assumption of information flow and transaction
costs of CAPM.
To sum up, the CAPM is a good estimation of the security prices; however it is not the complete
picture. The R2 of the CAPM is nearly 40%, indicating that the largest amount of the security
price variation is not explained by CAPM. The assumption of CAPM for information symmetry
is also overruled, as the political risk has significant influence on security return. Moreover,
increasing IPR12 results in higher return, which is contra intuitive to what CAPM predicts. In
addition, the exposure to foreign country risk does not have linear influence on the country index
return. These results are robust when testing with two different measures of world market
performance and two different methodologies.
Page | 30
F. References
Agmon, Tamir, and Findlay, M. Chapman, 1982, Domestic political risk and stock valuation,
Financial Analysts Journal 38(6), 74-77.
Attinasi II, Maria Grazia, Checherita-Westpal, Cristina D., and Nickel, Christiane, 2009, What
explains the surge in Euro Area sovereign spread during the financial crisis of 2007-2009? ECB
Workin Paper No. 1131.
Beaulieu, Marie-Claude, Cosset, Jean-Claude, and Essaddam, Naceur, 2006, Political uncertainty
and stock market returns: evidence from the 1995 Quebec referendum, Conadian Journal of
Economics 39(2), 621-642.
Bekaert, Geert, 1995, Market integration and investment barriers in emerging equity markets,
World Bank Economic Review 9(1), 75-107.
Bekaert, Geert, and Harvey, Campbell R., 1997, Emerging equity market volatility, Journal of
Financial Economics 43, 29-77.
Bittlingmayer, George, 1998, Output, stock volatility, and political uncertainty in a natural
experiment: Germany, 1880-1940, The Journal of Finance 53(6), 2243-2257.
Blitz, David C., and van Vliet, Pim, 2007, The volatility effect: low risk without lower return,
The Journal of Portfolio Management Fall, 102-113.
Brewer, Thomas L., 1981, Political risk assessment for foreign direct investment decisions: better
Page | 31
Brooks, Chris, 2008, Introductory Econometrics for Finance, (Cambridge University Press, New
York, NY)
Cambel, John Y., and Taksler, Glen B., 2002, Equity volatility and corporate bond yields,
Working Paper No: 8961, Cambridge, MA.
Chan, Yue-cheong, and Wei, Hohn K.C., 1996, Political risk and stock price volatility: the case
of Hong Kong, Pacific-Basin Finance Journal 4(2-3), 259-275).
Clark, Ephraim, and Tunaru Radu, 2003, Quantification of political risk with multiple dependent
sources, Journal of Economics and Finance 27(2), 125-135.
Clark, Ephraim, and Tunaru Radu, 2005, The evolution of international political risk 1956-2001,
http://repec.org/mmfc05/paper37.pdf
Cosset, Jean-Claude, and Suret, Jean-Marc, 1995, Political risk and the benefits of international
portfolio diversification, Journal of International Business Studies 26(2), 301-318.
Culter, David M., Poterba, James M., and Summers, Lawrence H., 1989, What moves stock
prices? The Journal of Portfolio Management 15(3), 4-12.
Danielson, Morris G., and Dowdell, Thomas D., 2001, The return-stages valuation model and the
expectations within a firm’s P/B and P/E ratios, Financial Management 30(2), 93-124.
Diamonte, Robin L., Liew, John M., and Stevens, Ross L., 1996, Political risk in emerging and
developed markets, Financial Analysts Journal 52(3), 71-76.
Erb, Claude B., Harvey, Campbell R., and Viskanta, Tadas E., 1995, Country risk and global
Page | 32
Erb, Claude B., Harvey, Campbell R., and Viskanta, Tadas E., 1996a, Political risk, economic
risk and financial risk, Financial Analyst Journal 52(6), 29-46.
Erb, Claude B., Harvey, Campbell R., and Viskanta, Tadas E., 1996b, Expected returns and
volatility in 135 countries, The Journal of Portfolio Management Spring, 46-58.
Errunza, Vihang. and Losq, Etienne, 1985, International asset pricing under mild segmentation:
theory and test, The Journal of Finance 40(1), 105-124.
Fama, Eugene F., 1968a, Risk, return, and equilibrium, Chicago: Center for Mathematical Studies
in Business and Economics, No: 6831, University of Chicago.
Fama, Eugene F., 1968b, Risk, return, and equilibrium: some clarifying comments, Journal of
Finance 23, 29-40.
Fama, Eugene F., 1970, Efficient capital markets: a review of theory and empirical work, The
Journal of Finance 25(2), 383-417.
Fama, Eugene F., and French, Kenneth R., 1992, The cross-section of expected stock returns,
Journal of Finance 47, 427-465.
Fama, Eugene F., and MacBeth, James D., 1973, Risk, return and equilibrium: empirical tests,
The Journal of Political Economy 81(3), 607-636.
Ferson, Wayne E., and Harvey, Campbell R., 1993, The risk and predictability of international
equity returns, The Review of Financial Studies 6(3), 527-566.
Garcia, René, and Ghysels, Eric, 1998, Structural change and asset pricing in emerging markets,
Page | 33
Goodhart, Charles A.E., 2008, The background to the 2007 financial crisis, Internatioanl
Economics and Economic Policy 4, 331-346.
Hall, Maximilian J.B., 2008, The sub-prime crisis, the credit squeeze and Northern Rock: the
lessons to be learned, Journal of Financial Regulation and Compliance 16(1), 19-34.
Jorion, Philippe, 1990, The exchange-rate exposure of U.S. multinationals, The Journal of
Business 63(3), 331-345.
Kim, Harold Y., and Mei, Jianping P., 2001, What makes the stock market jump? An analysis of
political risk on Hong Kong stock returns, Journal of International Money and Finance 20(7),
1003-1016.
King, Alan J., 1993, Asymmetric risk measures and tracking models for portfolio optimization
under uncertainty, Annals of Operations Research 45(1), 165-177.
Kobrin, Stephen J., 1979, Political risk: a review and reconsideration, Journal of International
Business Studies 10(1), 67-80.
Linter, John, 1965a, Security prices, risk, and maximal gain from diversification, Journal of
Finance 20, 587-616.
Linter, John, 1965b, The valuation of risk assets and the selection of risky investments in stock
portfolios and capital budgets, Review of Economics and Statistics 47, 13-37.
Long, John B., Jr., 1972, Consumption-investment decision and equilibrium in security market, in
Michael C. Jensen, ed.: Studies in the Theory of Capital Markets (Preager Publishers).
Page | 34
Pukthuanthong-Le, Kuntara, and Visaltanachoti, Nuttawat, 2009, Idiosyncratic volatility and
stock returns: a cross country analysis, Applied Financial Economics 19, 1269-1281.
Ross, Stephen A., 1976, The arbitrage theory of capital asset pricing, Journal of Economic
Theory 13(3), 341-360.
Sharpe, William F., 1964, Capital asset prices: A theory of market equilibrium under conditions
of risk, Journal of Finance 19, 425-442.
Shleifer, Andrei, 2000, Inefficient markets: and introduction to behavioral finance (Oxford
University Press, New York).
Shleifer, Andrei, and Vishny, Robert W., 1997, The limits of arbitrage, The Journal of Finance
52(1), 35-55.
Simon, Jeffrey D., 1982, Political risk assessment: past trends and future prospects, Columbia
Journal of World Business Fall 17(3), 62-70.
Solnik, Bruno H., 1974, The international pricing of risk: an empirical investigation of the world
capital market structure, The Journal of Finance 29(2), 365-378.
Page | 35
Appendix
Appendix A: List of countries and indices
Country Index Code N
Argentina ARGENTINA MERVAL ARGMERV(PI) 16
Australia S&P/ASX 300 ASX300I(PI) 16
Austria ATX - AUSTRIAN TRADED INDEX ATXINDX(PI) 16
Belgium BEL 20 BGBEL20(PI) 16
Brazil MSCI BRAZIL MSBRAZL(PI) 16
Bulgaria BULGARIA SE SOFIX BSSOFIX(PI) 8
Canada S&P/TSX 60 INDEX TTOSP60(PI) 16
Chile CHILE SANTIAGO SE GENERAL (IGPA) IGPAGEN(PI) 16
China SHANGHAI SE A SHARE CHSASHR(PI) 16
Colombia MSCI COLOMBIA MSCOLML(PI) 15
Croatia CROATIA CROBEX CTCROBE(PI) 11
Czech PRAGUE SE PX CZPXIDX(PI) 14
Denmark OMX COPENHAGEN (OMXC20) DKKFXIN(PI) 16
Estonia OMX TALLINN (OMXT) ESTALSE(PI) 12
Finland OMX HELSINKI (OMXH) HEXINDX(PI) 16
France FRANCE CAC 40 FRCAC40(PI) 16
Germany DAX 30 PERFORMANCE DAXINDX(PI) 16
Greece ATHEX COMPOSITE GRAGENL(PI) 16
Hong Kong HANG SENG HNGKNGI(PI) 16
Hungary BUDAPEST (BUX) BUXINDX(PI) 16
Iceland OMX ICELAND ALL SHARE ICEXALL(PI) 15
India INDIA BSE (100) NATIONAL IBOMBSE(PI) 16
Indonesia MSCI INDONESIA MSINDFL(PI) 16
Ireland IRELAND SE OVERALL (ISEQ) ISEQUIT(PI) 16
Israel ISRAEL TA 100 ISTA100(PI) 16
Italy MSCI ITALY MSITALL(PI) 16
Japan TOPIX TOKYOSE(PI) 16
Korea KOREA SE KOSPI 200 KOR200I(PI) 16
Latvia NOMURA LATVIA NMLATVL(PI) 12
Luxembourg LUXEMBOURG SE GENERAL LUXGENI(PI) 9
Malaysia FTSE BURSA MALAYSIA KLCI FBMKLCI(PI) 16
Malta MALTA SE MSE MALTAIX(PI) 12
Mexico MEXICO IPC (BOLSA) MXIPC35(PI) 16
Page | 36
Appendix A: List of countries and indices (Continued)
Country Index Code N
New Zealand MSCI NEW ZEALAND MSNZEAL(PI) 16
Norway MSCI NORWAY MSNWAYL(PI) 16
Peru LIMA SE GENERAL(IGBL) PEGENRL(PI) 16
Philippine PHILIPPINE SE I(PSEi) PSECOMP(PI) 16
Poland MSCI POLAND MSPLNDL(PI) 15
Portugal PORTUGAL PSI-20 POPSI20(PI) 15
Romania ROMANIA BET (L) RMBETRL(PI) 11
Russia RUSSIA RTS INDEX RSRTSIN(PI) 13
Singapore MSCI SINGAPORE MSSINGL(PI) 16
Slovakia SLOVAKIA SAX 16 SXSAX16(PI) 15
South Africa MSCI SOUTH AFRICA MSSARFL(PI) 15
Spain IBEX 35 IBEX35I(PI) 16
Sweden OMX STOCKHOLM 30 (OMXS30) SWEDOMX(PI) 16
Switzerland SWISS MARKET (SMI) SWISSMI(PI) 16
Taiwan TAIWAN SE WEIGHED TAIEX TAIWGHT(PI) 16
Thailand BANGKOK S.E.T. BNGKSET(PI) 16
Turkey ISTANBUL SE NATIONAL 100 TRKISTB(PI) 16
UK FTSE 100 FTSE100(PI) 16
USA NASDAQ 100 NASA100(PI) 16
Venezuela VENEZUELA SE GENERAL VENGENL(PI) 15
World MSCI WORLD U$ MSWRLD$(PI) 16
World DJGL WORLD $ DJWRLD$(PI) 16
The table reports a summary of countries included in the research with their corresponding indices. Code of index refers to the code employed by Thomson Datastream database. As can be seen, one of the most important emerging markets, Russian Federation, has 13 years of observation for index return. RTS was introduced as early as in 1995. Likewise, most of formal Soviet countries or countries in formal Warsaw pact developed equity indices during late 90’ or beginning of the 21th
Page | 37
Appendix B: The number of representatives of Euro Area countries in the European Parliament and their share in political influence Euro Area Members Representatives Share of Representatives
Austria 19 3.96% Belgium 21 4.38% Cyprus 6 1.25% Estonia 6 1.25% Finland 13 2.71% France 74 15.42% Germany 99 20.63% Greece 22 4.58% Ireland 12 2.50% Italy 73 15.21% Luxembourg 6 1.25% Malta 6 1.25% Netherlands 26 5.42% Portugal 22 4.58% Slovakia 13 2.71% Slovenia 8 1.67% Spain 54 11.25% Total 480 100.00%
Appendix C: World Index composition
Sector Weights Regional Weights
MSCI DJGL MSCI DJGL
Financials 20.53% 22.06% USA 52.54% 44.48%
Information Technology 11.57% 9.41% UK 9.52% 7.99%
Consumer Discretionary 11.23% 10.23% Japan 8.43% 8.29%
Industrials 10.95% 12.80% Canada 4.73% 4.25%
Health Care 10.71% 8.73% France 4.11% 3.33%
Consumer Staples 10.64% 12.56% Other 20.67% 31.66%
Energy 10.41% 9.83%
Materials 6.79% 7.05%
Telecommunication Services 3.79% 3.79%
Utilities 3.38% 3.55%
Page | 38
Appendix D.1: Unrestricted pooled OLS
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
IPR 1.283 2.884*** 1.392 3.048*** (0.784) (0.848) (0.768) (0.829) IPR2 133.1*** 138.0*** (28.86) (28.30) MSCI World 1.023*** 1.009*** 1.014*** (0.0466) (0.0473) (0.0467) DJGL World 1.041*** 1.026*** 1.033*** (0.0452) (0.0459) (0.0452) Outliers -0.0795 -0.0813 -0.0790 -0.0580 -0.0600 -0.0573 (0.0936) (0.0935) (0.0923) (0.0920) (0.0919) (0.0905) Domestic PR -0.592* -0.716* -0.706* -0.588* -0.723* -0.712* (0.282) (0.292) (0.288) (0.277) (0.286) (0.282) Intercept 0.0446*** 0.0470*** 0.0226 0.0435*** 0.0460*** 0.0207 (0.0106) (0.0107) (0.0118) (0.0104) (0.0105) (0.0116) R2 0.392 0.394 0.410 0.413 0.416 0.433 Adj R2 0.389 0.391 0.406 0.411 0.413 0.430 Heteroscedasticity 0.165 0.170 0.705 0.070 0.085 0.999 Autocorrelation 0.086 0.096 0.117 0.085 0.094 0.113 Note 1: N=777
Note 2: Standard errors in parentheses Note 3: * p<0.05, ** p<0.01, *** p<0.001
Note 4: Heteroscedasticity is tested using Breusch-Pegan / Cook-Weisberg test. The values are the P-values of the test with null hypothesis being that the variance of the error term is constant.
Note 5: The value of autocorrelation is measured as correlation between the disturbance term and its one year lagged values.
Appendix D.2: Summary Variance Inflation Factor per regression model
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
IPR 1.11 1.33 1.10 1.33 IPR2 1.23 1.23 MSCI World 1.01 1.05 1.05 DJGL World 1.01 1.04 1.05 Outliers 1.01 1.01 1.01 1.01 1.01 1.01 Domestic PR 1.00 1.07 1.07 1.00 1.07 1.07 Mean 1.01 1.06 1.14 1.01 1.06 1.14
Note: A VIF value above 10 indicates presence of multicollinearity; however the outcome of
Page | 39
Appendix D.3: Correlation coefficients of regression disturbance with theirs one year lagged value per country for Model 3
Country Correlation N Country Correlation N
Argentina 0.109 14 Korea -0.087 14 Australia 0.455 14 Latvia 0.080 11 Austria 0.584 14 Luxembourg 0.209 8 Belgium -0.058 14 Malaysia 0.199 14 Brazil -0.550 14 Malta -0.045 11 Bulgaria 0.034 7 Mexico -0.176 14 Canada -0.110 14 Netherlands 0.326 14
Chile 0.240 14 New Zealand 0.489 14
China 0.072 14 Norway 0.014 14 Colombia 0.099 14 Peru 0.289 14 Croatia 0.130 10 Philippine 0.145 14 Czech 0.599 13 Poland 0.187 14 Denmark -0.196 14 Portugal 0.134 14 Estonia 0.090 11 Romania 0.250 10 Finland 0.325 14 Russia -0.094 12 France 0.470 14 Singapore 0.069 14 Germany 0.220 14 Slovakia 0.265 14
Greece 0.085 14 South Africa 0.411 14
Hong Kong 0.019 14 Spain 0.202 14
Hungary -0.022 14 Sweden 0.073 14 Iceland 0.276 14 Switzerland -0.408 14 India 0.183 14 Taiwan -0.415 14 Indonesia 0.238 14 Thailand 0.559 14 Ireland 0.175 14 Turkey -0.441 14 Israel -0.261 14 UK -0.249 14 Italy -0.341 14 USA 0.389 14 Japan -0.025 14 Venezuela -0.159 14
Note: Although the total correlation coefficient signifies absence of autocorrelation with the
Page | 40
Appendix D.4: Correlation coefficients of regression disturbance with theirs one year lagged value per year for Model 3
Year Correlation N Year Correlation N
1995 -0.206 45 2002 0.349 54 1996 0.400 46 2003 0.434 54 1997 0.409 47 2004 0.375 54 1998 -0.067 50 2005 0.459 54 1999 0.108 52 2006 -0.338 54 2000 -0.140 52 2007 0.486 54 2001 0.029 53 2008 -0.218 54
Note: The correlation of the indices seem to follow a specific pattern which is not described by
IPR or world equity benchmark. Typically, in the booming periods, the indices move together. During the dot com crisis, the markets did not behave as much similar as during the crisis of 2006 - 2007.
Appendix D.5: Scatterplot of disturbance term of Model 3 across time
Page | 41
Appendix E: Definitions of the regional dummy variables
Region Countries in region
Asia China, Hong Kong, India, Indonesia, Japan, Korea, Malaysia, Philippine, Singapore, Taiwan, Thailand Europe Austria, Belgium, Bulgaria, Croatia, Czech Republic,
Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Latvia,
Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Romania, Russia, Slovakia, Spain, Sweden, Switzerland, Turkey, UK
Latin America Argentina, Brazil, Chile, Colombia, Peru North America Canada, Mexico, USA
Oceania Australia, New Zealand
Page | 42
Appendix F: Robust OLS regression with regional and income dummy variables
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
IPR 1.203 2.689* 1.325 2.868* (1.095) (1.320) (1.078) (1.301) IPR2 122.3*** 127.3*** (36.07) (35.92) MSCI World 1.028*** 1.014*** 1.019*** (0.0518) (0.0514) (0.0512) DJGL World 1.043*** 1.029*** 1.036*** (0.0499) (0.0490) (0.0492) Outliers -0.134 -0.135 -0.133 -0.111 -0.113 -0.110 (0.445) (0.440) (0.432) (0.440) (0.435) (0.426) Domestic PR -0.579 -0.694* -0.688* -0.575* -0.702* -0.695* (0.301) (0.347) (0.332) (0.291) (0.337) (0.322) Asia -0.0902** -0.0895** -0.0747* -0.0900** -0.0892** -0.0738* (0.0290) (0.0295) (0.0310) (0.0283) (0.0288) (0.0304) Europe -0.00508 -0.00750 0.0123 -0.00696 -0.00960 0.0110 (0.0259) (0.0267) (0.0271) (0.0257) (0.0265) (0.0270) Latin America 0.0350 0.0356 0.0392 0.0355 0.0362 0.0399 (0.0527) (0.0528) (0.0518) (0.0518) (0.0518) (0.0509) Oceania -0.0482 -0.0475 -0.0312 -0.0492 -0.0484 -0.0314 (0.0326) (0.0325) (0.0352) (0.0320) (0.0320) (0.0347) High Income -0.0695** -0.0694** -0.0661** -0.0676** -0.0675** -0.0639** (0.0241) (0.0241) (0.0242) (0.0237) (0.0237) (0.0237) Intercept 0.104*** 0.107*** 0.0682* 0.103*** 0.106*** 0.0655* (0.0244) (0.0244) (0.0279) (0.0240) (0.0241) (0.0275) R2 0.421 0.423 0.436 0.441 0.443 0.458 Adj R2 0.415 0.416 0.428 0.435 0.437 0.450 Regional Test 0.001 0.002 0.003 0.001 0.002 0.002 Income Test 0.004 0.004 0.006 0.004 0.005 0.007 Note 1: N=765
Note 2: Standard errors in parentheses Note 3: * p<0.05, ** p<0.01, *** p<0.001
Note 4: Regional test is the joint coefficient test of Asia, Europe, Latin America and Oceania equaling zero.
Page | 43
Appendix G.1: GLS panel data regression with control for ARCH effect, Durbin Watson d-statistic and Fixed Effects
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
IPR 0.585 2.828** 0.567 2.781**
(0.789) (0.932) (0.773) (0.911)
IPR2 126.5*** 127.0***
(29.89) (29.25)
Risk Premium MSCI 1.022*** 1.017*** 1.025*** (0.0453) (0.0460) (0.0457) Risk Premium DJGL 1.028*** 1.023*** 1.030*** (0.0435) (0.0441) (0.0438) Outliers -0.540*** -0.534*** -0.506*** -0.517*** -0.512*** -0.487*** (0.0954) (0.0955) (0.0945) (0.0936) (0.0937) (0.0925) Domestic PR -0.271 -0.321 -0.286 -0.284 -0.334 -0.300 (0.286) (0.295) (0.293) (0.280) (0.289) (0.286) Intercept 0.0465*** 0.0480*** 0.0316** 0.0457*** 0.0472*** 0.0307** (0.0102) (0.0104) (0.0108) (0.0100) (0.0102) (0.0106) R2 Within 0.466 0.465 0.477 0.487 0.487 0.499 R2 Between 0.197 0.196 0.111 0.205 0.205 0.110 R2 Overall 0.440 0.440 0.449 0.463 0.463 0.473 Note 1: N=723
Note 2: Standard errors in parentheses Note 3: * p<0.05, ** p<0.01, *** p<0.001
Appendix G.2: P-values of hypotheses tests
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
IPR 0.459 0.002 0.463 0.002
IPR2 0.000 0.000
Risk Premium MSCI 0.627 0.709 0.583