How Underlying Dimensions of Political Risk Affect Excess Return in Emerging and Developed Markets

Abstract

Political risk is expected to increase due to emerging markets’ increasing influence on the world economy. We identify legal, tension, conflict and policy as underlying dimensions through principal component analysis by using a disaggregated political risk index. Using a two-way error correction model, ethnic and religious tension is identified as a new and distinct dimension of political risk. Consequently, global investors are likely to benefit from understanding which dimension implies a reward. Investors in particular should direct their attention towards tension, which seems to command a risk premium regardless of both market and time.

JEL Classification: C33, F30, F50, G15

Keywords

Political risk emerging markets principal component analysis ethnic tensions religious tensions

1. Introduction

Foreign direct investments have grown over the past few decades (Jakobsen, 2012a). The rate of growth in emerging markets has proven to be even more pronounced, although the majority of foreign direct investments still flow to developed countries (Bilson, Brailsford, & Hooper, 2001). The world’s economic growth as such is increasingly driven by emerging markets (Bremmer & Keat, 2010), where political risk is considered to be of increased importance (Bilson, Brailsford, & Hooper, 2002; Bremmer & Keat, 2010; Diamonte, Liew, & Stevens, 1996). Political risk is further expected to increase due to emerging markets’ increasing influence on the world economy (Bremmer, 2012). One particular theme of interest to investors is the increasing magnitude of social upheaval and unrest that is also expected to increase going forward (International Labour Organization [ILO], 2017). Hence, increasing globalisation in conjunction with political risk is of great importance to global investors and a challenge worth studying in considerable detail.

Political risk is broadly defined by Jakobsen (2012b, p. 13) as ‘those events, actions, processes or characteristics of a socio-political nature that have the potential to—directly or indirectly—significantly and negatively affect the goals of foreign direct investors’. According to a standard portfolio theory, increased risk should lead to greater returns. Political risk is likely to violate this classical risk–return relation, as countries that are considered to be politically more at risk have been shown to receive lower returns than politically safer countries (Diamonte et al., 1996; Erb, Harvey, & Viskanta, 1996a). Recent studies confirm these findings. On the other hand, Erb et al. (1996a) and Bilson et al. (2002) suggested that political risk does have a positive effect on stock market returns. Lehkonen and Heimonen (2015) referred to this ongoing dispute as a political risk sign paradox. The empirical literature is also inconclusive with respect to the investigation of the impact of political risk on foreign direct investments. Jakobsen (2010, 2012a) suggested that this emanates from an under-prioritisation of the increasingly multidimensional phenomenon of political risk.

Despite its importance, the multidimensional nature of political risk is somewhat understudied. Two notable exceptions are Lehkonen and Heimonen (2015) and Dimic, Orlov, and Piljak (2015) who provided some insight into the underlying dimensions of political risk and their impact on stock market returns. However, as far as we know, no one has estimated the underlying dimensions of political risk using an econometric approach to investigate the effect on stock market returns. Being aware of the multidimensional nature of political risk, this study aims to identify whether underlying dimensions of political risk affect stock market excess returns differently and to contribute to an understanding of the political risk sign paradox. We suggest that an aggregated political risk measure sweeps away the potentially conflicting effects of underlying dimensions of political risk (Berggren, Bergh, & Bjørnskov, 2012; Harvey, 2004; Jakobsen, 2012a; Lam & Zhang, 2014) since it is unlikely that all political risk components are equally linked to stock market returns (Berggren, Bergh, & Bjørnskov, 2015). Hence, knowledge concerning compensation for bearing different types of political risk is of great importance for global investors.

The main research question is, How are stock market excess returns affected by underlying dimensions of political risk?

Our econometric analysis identifies legal, tension, conflict and policy as the underlying dimensions of political risk and demonstrate the increasing importance of ethnic and religious tension (tension) as a distinct risk dimension likely to influence financial markets going forward.

We contribute to existing research in two distinct ways. First, by means of principal component analysis, we disentangle the underlying dimensions of political risk in order to gain a more nuanced understanding of how they differ in their impact on stock market returns, thus identifying a novel and distinct dimension stemming from religious and ethnic tension. Second, we analyse the extent to which the political risk dimensions affect excess returns differently under different market types (e.g., developed vs. emerging markets) and over time, in order to determine which political risk is rewarded, with a particular emphasis on religious and ethnic tensions.

The article is organised as follows. Section 2 summarises previous research. Section 3 provides information regarding the data and methodology, while Section 4 presents our analysis and findings. Section 5 offers some concluding remarks.

2. Literature

Political instability has proven to be an important determinant of stock market performance. Previous research suggests that higher political risk is associated with higher expected returns (Bilson et al., 2002; Erb et al., 1996a; Erb, Harvey, & Viskanta, 1996b; Harvey, 2004). However, in some cases, political risk has proven to violate this classical risk–return relation (Diamonte et al., 1996; Dimic et al., 2015; Erb et al., 1996a; Lehkonen & Heimonen, 2015; Perotti & van Oijen, 2001). Lehkonen and Heimonen (2015) referred to this ongoing dispute as a political risk sign paradox.

The existing literature is inconclusive with respect to the direction of the effect of political risk on financial markets. On the one hand, empirical evidence suggests a negative political risk premium (RP), implying that investors are likely to accept reduced returns to hedge against political uncertainty (Brogaard & Detzel, 2015). On the other hand, others suggest a positive premium in support of the classical risk–return relation (Lam & Zhang, 2014). However, high political risk along one dimension could outperform low political risk in another and vice versa (Jakobsen, 2012a). Such complexity calls for a consideration of the underlying dimensions of political risk and their impact on stock market excess returns because previous research has primarily considered an aggregated and highly multidimensional political risk index (PRI). In other words, different dimensions of political risk could be seen as orthogonal to each other, hence representing vastly different effects on market return.

Our study builds on previous work emphasising the importance of different underlying dimensions of political risk. In several analyses, Lam and Zhang (2014) identified the political risk components ‘Bureaucracy Quality’ and ‘Government Stability’ as distinct dimensions of political risk. The former (latter) dimension commands an RP in the emerging (developed) markets. Harvey (2004) included several components of political risk, implicitly treating each as a distinct dimension. In contrast to the aggregated PRI, several components imply positive hedge returns in the emerging markets, especially in cases of ‘Government Stability’, ‘Investment Profile’ and ‘Internal Conflict’. When investigating subgroups of political risk, as defined by Bekaert, Harvey, and Lundblad (2005) and Bekaert, Harvey, Lundblad, and Siegel (2014), in addition to political risk components, both are suggested as being unique to specific markets (Dimic et al., 2015). However, tensions are associated with lower stock market returns in less developed markets. Hence, the subgroup ‘Tensions and Conflicts’ seems to violate the classical risk–return relation when the level of democracy is not taken into account (Lehkonen & Heimonen, 2015), supporting Pástor and Veronesi (2013), who suggested that the political RP is economic state dependent. Berggren et al. (2012) detected a negative relationship between the ‘Social Congruence’ dimension of political risk and growth in rich countries. Furthermore, improvements in ‘Legal’ and ‘Policy’ are suggested to be positively related to growth in rich countries, whereas instability hampers growth in poor countries. The positive relation in terms of ‘Policy’ further turns out to correlate positively with the instability of policies (Berggren et al., 2015). In summary, the social aspect of political risk seems to be an important underlying dimension that affects the stock market differently, depending on the country’s macroeconomic state, leading to the following hypothesis:

Hypothesis 1 (H1): Underlying dimensions of political risk affect stock market excess returns differently between markets.

In addition to the cross-sectional effect from the state-dependent nature of political risk, another branch of literature considers the time-variation in the same relationship. Arguably, the emerging market is becoming politically more secure, whereas the developed market is becoming politically riskier (Dimic et al., 2015; Erb et al., 1996a; Harvey, 2004). This phenomenon is referred to by Diamonte et al. (1996) as a global convergence. Consequently, the effect of political risk on financial markets is expected to vary over time. In association with a global convergence, and the fact that the classical risk–return relation in terms of political risk is considered a relatively recent phenomenon (Bilson et al., 2002; Lam & Zhang, 2014), we propose a further hypothesis:

Hypothesis 2 (H2): The underlying dimensions of political risk affect stock market excess return differently over time.

3. Data and Methods

Our particular research question calls for a two-step research process; first, we carve out four principal components from the political risk data from international country risk guide (ICRG). We then analyse these factors in a two-way error component model.

3.1 Data and Variable Description

The final sample size comprises 28 countries covering June 2001 to May 2015. Fifteen of the 28 countries are considered developed, and 13 are considered emerging (Table A1), according to well-established classification (Morgan Stanley Capital International [MSCI], 2016). To isolate the global financial crisis of 2008/2009, the sub-periods can be interpreted as pre-crisis (06/01–12/06), during-crisis (01/07–12/10) and post-crisis (01/11–05/15), including monthly observations.

3.1.1 Dependent Variable—Excess Return

Broad market indices are chosen to capture the movement of the country’s entire market. Returns are derived from monthly observations. The indices are measured in USD according to availability and comparability, reflecting the perspective of global investors. In line with the capital asset pricing model (CAPM), we use market-specific excess return as our dependent variable. This is constructed by subtracting the risk-free rate (proxied by the 10-year US treasury bond rate) from the return of the Morgan Stanley index for the focal market.¹

3.1.2 Independent Variable—Political Risk

This study makes use of the disaggregated PRI included in ICRG, published by the PRS Group. The guide provides a monthly rating covering 141 countries including three sub-categories of country-specific risk: political, financial and economic. The ratings are estimated based on subjective staff analyses and are thus considered to be a forward-looking measure making it predictive by nature (Bilson et al., 2002) and, further, suitable for stock market analysis. The components are aggregated into the PRI. The maximum rating on the PRI is 100. In each case, a lower (higher) risk rating reflects a higher (lower) risk.

3.1.3 Control Variables

Table A1 presents a detailed description of the sources of our control variables. The growth rate of Industrial Production (IP) is obtained by monthly observations, computing the first difference in the logarithm and included seasonally adjusted data (index 2010 = 100). The variable is lagged by 2 months as suggested by Chen, Roll, and Ross (1986) and Bilson et al. (2001).

Inflation (CPI) is estimated as the lagged first difference in the logarithm of the consumer price index and included seasonal adjustments (index 2010 = 100). The lagged variable represents short-term inflation expectations rather than the actual inflation.

RP is estimated so as to capture the effect of changes in risk aversion and is estimated by subtracting the 10-year US treasury bond rate from the MSCI all country world index (ACWI) as a proxy for the global market portfolio. To estimate the unanticipated movement, the moving 12-month average RP is subtracted each month.

To capture the influence of the shape of the Term Structure (TS), we subtract the previous month’s short-term interest rate from the current month’s long-term interest rate. The interest rates are converted from annual to monthly rates.

3.2 Approach Step One: Principal Component Analysis

We perform PCA on standardised data rather than mean-corrected data because we do not assume that a component’s variance indicates its importance when forming the PRI. The analysis is performed on the pooled sample because our prime interest is the overall grouping of components. We extract four principal components according to the ‘eigenvalue-greater-than-one’ rule and a logical interpretation, resulting in a total variance explanation of approximately 73 per cent. They will henceforth be referred to as ‘underlying dimensions of political risk’. When interpreting the principal components, we consider the rotated component matrix in Table 1, rotated by varimax² with Kaiser normalisation, in line with Toft (2008) and Berggren et al. (2012).

The 12 political risk components split nicely into four underlying dimensions of political risk. The components have a considerable amount in common with their respective dimensions as all of their uniqueness is below 0.4. Hence, the multidimensional phenomenon of political risk seems to prove applicable for the PRI of ICRG.

The dimensions are supported by both theoretical and empirical research. Moreover, the dimensions are comparable with the dimensions obtained by Berggren et al. (2012, 2015), implicitly suggesting that the correlations between the components are relatively constant over time because our sample includes more recent data. However, unlike this earlier work (2015), we include a fourth dimension with heavy loadings from ethnic and religious tensions. When we interpret the principal components restricted to three dimensions, the additional dimension, normally observed as separate from the others and independent of the rotation technique and sample specifications, is lost. Hence, four dimensions are preferable because the additional dimension is distinct and highly robust. This implies that ethnic and religious tensions have become more important in recent years, which may be a result of increased globalisation, leading to the increased effects of political risk evolving from tensions. Furthermore, terror-related incidents have increased in the twenty-first century, driven by ethnic and, especially by religious tensions. Hence, adding a fourth component is a contribution to the understanding of political risk.

Table 1.
Principal Component Analysis

Political Risk Components Principal Components

Legal Tension Conflict Policy Uniqueness

Bureaucracy quality 0.859 0.133 0.094 0.218 0.188

Corruption 0.829 0.133 0.186 0.200 0.221

Democratic accountability 0.723 0.063 0.040 −0.305 0.378

Ethnic tensions 0.113 0.805 −0.104 0.009 0.328

External conflict 0.005 −0.162 0.860 0.015 0.233

Government stability −0.037 −0.042 0.004 0.870 0.239

Internal conflict 0.328 0.384 0.606 0.414 0.206

Investment profile 0.404 0.303 0.237 0.539 0.399

Law and order 0.789 0.322 0.205 0.096 0.223

Military in politics 0.426 0.353 0.665 0.006 0.252

Religious tensions 0.249 0.785 0.203 0.111 0.268

Socioeconomic conditions 0.568 0.326 0.062 0.471 0.345

Political Risk Components	Principal Components
Bureaucracy quality	0.859	0.133	0.094	0.218	0.188
Corruption	0.829	0.133	0.186	0.200	0.221
Democratic accountability	0.723	0.063	0.040	−0.305	0.378
Ethnic tensions	0.113	0.805	−0.104	0.009	0.328
External conflict	0.005	−0.162	0.860	0.015	0.233
Government stability	−0.037	−0.042	0.004	0.870	0.239
Internal conflict	0.328	0.384	0.606	0.414	0.206
Investment profile	0.404	0.303	0.237	0.539	0.399
Law and order	0.789	0.322	0.205	0.096	0.223
Military in politics	0.426	0.353	0.665	0.006	0.252
Religious tensions	0.249	0.785	0.203	0.111	0.268
Socioeconomic conditions	0.568	0.326	0.062	0.471	0.345

Source: The authors.

Notes: Loadings and uniqueness. The rotated component matrix, using varimax with Kaiser Normalisation, is estimated out of the disaggregated political risk index (International Country Risk Guide). Loadings in boldface are referred to as ‘heavy loadings’ in the text, using 0.5 as cut-off.

The political risk components with heavy loadings on the first dimension are bureaucracy quality, corruption, democratic accountability, law and order and socioeconomic conditions, all of which measure the quality of a country’s legal system and its consequences (Howell, 2011). This dimension is highly correlated with rule of law from other (The World Justice Project, 2016) and bears similarities to the ‘Political Institutions’ dimension included in Jakobsen (2012a) and both ‘Quality and Institutions’ and ‘Democratic Tendencies’, suggested by Bekaert et al. (2005, 2014). Moreover, the dimension is similar to the ‘Legal’ dimension of Berggren et al. (2012) and the ‘Legal-administrative’ dimension of Berggren et al. (2015). Hence, the dimension is interpreted as legal.

The middle dimensions obtain heavy loadings from ethnic and religious tensions in addition to military in politics, external conflict, and internal conflict, all of which reflect unrest. They are comparable to the ‘Social Congruence’ dimension suggested by Berggren et al. (2012), the ‘Social Harmony’ dimension by Berggren et al. (2015) and the subgroup of ‘Conflicts’ suggested by Bekaert et al. (2005, 2014). The second dimension is interpreted as tension, and the third is interpreted as conflict. Tension is observed to correlate relatively highly with the fractionalisation data in terms of ‘Ethnic’ and ‘Language’ (Alesina, Devleeschauwer, Easterly, Kurlat, & Wacziarg, 2003). Hence, tension could be interpreted as an important determinant of political economy, reflecting instability, social unrest and political violence (Alesina & Perotti, 1996). Conflict is seen to correlate weakly with the data. The finding as such provides support whereby the middle dimensions indeed reflect two distinct dimensions, unlike previous research which highlights the social aspect of political risk in general. Both tension and conflict are considered sources of risk because they are indicators of ‘future trouble’ (Jakobsen, 2012a), indicating a lagged effect with respect to the influence on stock market returns. However, Jakobsen (2012b) recommended looking into both contemporary and recent history. Hence, both tension and conflict are lagged by only 1 month.

The fourth dimension obtains considerable loadings from government stability and investment profile, similar to the subgroup ‘Government Actions’ of Bekaert et al. (2005, 2014). The former component reflects the government’s ability to declare its programme and to stay in office (Howell, 2011). The latter component is a measure of contract viability, profit repatriation and payment delays (Howell, 2011). The component influences the ‘Economic Governance’ dimension used by Jakobsen (2012a). The dimension correlates relatively highly with the categories ‘Rule of Law’, ‘Regulatory Efficiency’ and ‘Open Markets’ from Index of Economic Freedom (The Heritage Foundation, 2016). This dimension is further identical with the ‘Policy’ dimension in Berggren et al. (2012) and is thus interpreted as policy.

Summary statistics from the pooled sample, including the 12 political risk components, are presented in Table A2.

3.3 Approach Step Two: Two-way Error Component Model

We specify a linear model, assuming both factors (entity in terms of market or country and time) and covariates (the independent variables) are linearly related to the dependent variable. Additionally, we cannot ignore the presence of entity-specific or time-specific fixed effects. Hence, we operate with a two-way error component model that allows both entity-fixed and time-fixed effects within the same model using maximum likelihood (Brooks, 2008). Including interaction terms between factors and covariates allows us to investigate whether the linear relation between an underlying dimension of political risk and the dependent variable varies by entity or over time. The market type (e.g., developed market) fixed effect is used to test the first hypothesis. Hence, the absence of fixed effects within countries included in each market is implicitly assumed, which could lead to a country bias. We therefore also test for country-fixed effects.³

To investigate how underlying dimensions of political risk affect stock market excess returns, we expand the multifactor model of Chen et al. (1986) with principal components extracted through a PCA. The factors included in the model of Chen et al. (1986) are sources of systematic asset risk, defined as economic state variables, consistent with the arbitrage pricing theory introduced by Ross (1976). We suggest underlying dimensions of political risk as economic state variables, systematically affecting stock market returns. Only the factor-specific loadings are considered and not the factor prices following Fama and MacBeth (1973), as in Chen et al. (1986). We estimate an equation of the form

E R_{i t} = α + \sum_{k} β_{k} X_{k i t} + \sum_{j} γ_{j} P_{j i t} + δ_{j i t} + ρ_{j i t} + μ_{i} + λ_{t} + ν_{i t}

where α is the overall constant term, and β_k are factor-specific loadings on the state variables of Chen et al. (1986); X_kit, and γ_j are underlying dimensions of political risk; P_jit. k denotes the presence of certain macroeconomic variables (e.g., CPI), while j refers to the four principal components (e.g., tension). δ_ijt (ρ_ijt) denotes an interaction term creating entity-specific (time-specific) betas, μ_i (λ_t) denotes a dummy for entity (time) and v_it denotes the remainder error term after subtracting both entity- and time-fixed effects; i and t denote country and months, respectively, and are operationalised by introducing a dummy for each country and year less one. Regarding the interaction terms, M and P denote markets (i.e., developed or emerging) and sub-periods (i.e., pre-crisis, during-crisis and post-crisis), respectively, whereas j again refers to the four dimensions of political risk (e.g., tension). The interaction terms are constructed using dummies for developed market and the during-crisis and post-crisis periods (M and P), multiplied with each of the political risk factors. The estimated coefficients of the market-specific interaction term (δ_ijt) capture the difference in the effect of the political risk factors on excess returns in different market types (i.e., developed or emerging). In other words, these show whether the political risk—excess return relationship—is moderated by the market type to which the country belongs.

Figure 1.

Source: The authors.

\begin{matrix} δ_{j i t, M} = L e g a l_{i t} \cdot D_{D M} + T e n s i o n_{i t} \cdot D_{D M} + C o n f l i c t_{i t} \cdot D_{D M} \\ + P o l i c y_{i t} \cdot D_{D M} ​ \end{matrix}

Similarly, the time-specific interaction term (ρ_ijt) captures the moderating effect of the different sub-periods (i.e., pre-crisis, during-crisis and post-crisis).

\begin{matrix} ρ_{j i t, p} = L e g a l_{i t} \cdot D_{D C} + T e n s i o n_{i t} \cdot D_{D C} + C o n f l i c t_{i t} \cdot D_{D C} \\ + P o l i c y_{i t} \cdot D_{D C} + L e g a l_{i t} \cdot D_{P C} + T e n s i o n_{i t} \cdot D_{P C} \\ + C o n f l i c t_{i t} \cdot D_{P C} + P o l i c y_{i t} \cdot D_{P C} \end{matrix}

Note that these parameters are essential in our investigation into the differing effects of political risk between markets and over time. Figure 1 sums up the empirical model for this article.

4. Analysis and Findings

4.1 Analysis Step One: Analysis of the Components

We perform two distinct preliminary analyses in line with Harvey (2004).⁴ First, we perform an ex ante analysis of the relation between the average return and the average disaggregated PRI. Two portfolios are constructed based on the risk level, low- and high-risk portfolios, using the cut-off points suggested by the PRS group; low-risk (high rating): 80 per cent–100 per cent and high-risk (low rating): 0 per cent–70 per cent (Howell, 2011). The hedge portfolio reflects a long position in the high-risk portfolio and a short position in the low-risk portfolio. Hence, a positive return in the hedge portfolio indicates a reward of political risk.

Our findings suggest that different components of political risk command varying political RPs, and that hedge returns in the emerging markets are higher; both findings are in line with Harvey (2004). Government stability produces a negative hedge return in the all-country sample, whereas socioeconomic conditions produce the highest positive hedge return in the all-country sample. Ethnic tensions contribute with a positive hedge return of 5.606 per cent p.a., of which 6.734 per cent p.a. is related to the emerging markets. A similar trend is detected for religious tensions.

Moreover, the cross-sectional relations (Table A3) between the country’s total equity RPs,⁵ estimated by Damodaran (2016), and underlying dimensions of political risk are examined. In the case of the emerging markets,⁶ there does not seem to be any relation at all except with tension, which shows a negative correlation. This suggests that increased political risk is associated with an increased RP, in line with Harvey (2004). In the case of the developed markets, both legal and policy correlate negatively with the country’s total equity RP. The opposite trend is observed for tension, further in line with Harvey (2004).

These preliminary analyses confirm our assumption that the aggregated PRI is too coarse because there seems to be differences between both components and dimensions. Tension (ethnic and religious) is noted as an important and distinct dimension, which is the political risk aspect that is rewarded. This conflicts with previous research which highlights the social aspect of political risk in general.

4.2 Analysis Step Two: Two-way Error Component Model

We extend the multifactor model of Chen et al. (1986) in various ways. First, we identify the incremental effect of each underlying dimension of political risk resulting in five different model specifications. Second, the hypotheses are addressed by introducing interaction terms between the political risk dimensions and both market and time separately, resulting in two different model specifications (nos. 6 and 7). Finally, the main model (no. 8) seeks to address the main question by including the interaction terms of both market and time simultaneously. Significantly negative coefficients for the underlying dimensions of political risk reflect a positive risk–return relation according to the construction of the rating system where a low (high) rating indicates high (low) risk. Significantly positive coefficients will violate the classical risk–return relation.

We present the results of introducing one underlying dimension of political risk at a time in Table 2. Legal is significantly positive in models 2–4. In model 5, legal is insignificant, whereas policy is significantly negative. Two of 12 political risk components load moderately on the dimensions and are illustrated in Section 3.2. Both dimensions are considered positively related to economic growth in Berggren et al. (2012). As such, legal and policy are likely to reflect partially similar dimensions of political risk. However, the markets seem more differentiated with respect to legal, as the significance level of the market-fixed effect⁷ deteriorates in model 5. The time-fixed effect is highly significant during all model specifications.

Table 2.
Two-way Error Component Model

Estimates of Fixed Effects Including Type III Tests of Fixed Effects

Model
1
2
3
4
5

Variables
Estimates

Estimates of intercept

Intercept 0.0024
(0.0018) 0.0043
(0.0020) 0.0039
(0.0020) 0.0038
(0.0020) 0.0012
(0.0023)

μ −0.0033
(0.0016) −0.0072*
(0.0024) −0.0064*
(0.0024) −0.0063
(0.0025) −0.0043*
(0.0026)

λ₁ 0.0104*
(0.0023) 0.0052*
(0.0019) 0.0061*
(0.0019) 0.0062*
(0.0019) 0.0090***
(0.0022)

λ₂ −0.0040*
(0.0022) −0.0064*
(0.0021) −0.0064*
(0.0021) −0.0063
(0.0021) −0.0049
(0.0021)

Estimates of independent variables

RP 1.0400*
(0.0149) 1.0404*
(0.0150) 1.0560*
(0.0150) 1.0560*
(0.0150) 1.0556*
(0.0150)

TS −0.0007*
(0.0002) −0.0007*
(0.0002) −0.0007*
(0.0002) −0.0007*
(0.0002) −0.0007*
(0.0002)

CPI 0.5683
(0.2478) 0.5424
(0.2480) 0.5282
(0.2490) 0.5268
(0.2491) 0.5166
(0.2490)

IP 0.1692*
(0.0293) 0.1680*
(0.0293) 0.1675*
(0.0291) 0.1675*
(0.0292) 0.1688*
(0.0290)

Legal 0.0027
(0.0012) 0.0022
(0.0012) 0.0021*
(0.0012) 0.0015
(0.0012)

Tension −0.0011
(0.0008) −0.0011
(0.0008) −0.0013
(0.0008)

Conflict −0.0001
(0.0008) −0.0003
(0.0008)

Policy −0.0023
(0.0010)

Significance level of fixed effects

Market 0.041 0.002* 0.000* 0.009*** 0.096*

Time 0.000* 0.000* 0.000* 0.000* 0.000***

Information criteria

AIC −14,076 −14,079 −14,059 −14,057 −14,061

BIC −14,018 −14,015 −13,988 −13,980 −13,977

N 4,732 4,732 4,704 4,704 4,704

Estimates of Fixed Effects Including Type III Tests of Fixed Effects
Estimates of intercept
Intercept	0.0024 (0.0018)	0.0043 (0.0020)	0.0039 (0.0020)	0.0038 (0.0020)	0.0012 (0.0023)
μ	−0.0033** (0.0016)	−0.0072*** (0.0024)	−0.0064*** (0.0024)	−0.0063** (0.0025)	−0.0043* (0.0026)
λ₁	0.0104*** (0.0023)	0.0052*** (0.0019)	0.0061*** (0.0019)	0.0062*** (0.0019)	0.0090*** (0.0022)
λ₂	−0.0040* (0.0022)	−0.0064*** (0.0021)	−0.0064*** (0.0021)	−0.0063** (0.0021)	−0.0049** (0.0021)
Estimates of independent variables
RP	1.0400*** (0.0149)	1.0404*** (0.0150)	1.0560*** (0.0150)	1.0560*** (0.0150)	1.0556*** (0.0150)
TS	−0.0007*** (0.0002)	−0.0007*** (0.0002)	−0.0007*** (0.0002)	−0.0007*** (0.0002)	−0.0007*** (0.0002)
CPI	0.5683** (0.2478)	0.5424** (0.2480)	0.5282** (0.2490)	0.5268** (0.2491)	0.5166** (0.2490)
IP	0.1692*** (0.0293)	0.1680*** (0.0293)	0.1675*** (0.0291)	0.1675*** (0.0292)	0.1688*** (0.0290)
Legal		0.0027** (0.0012)	0.0022** (0.0012)	0.0021* (0.0012)	0.0015 (0.0012)
Tension			−0.0011 (0.0008)	−0.0011 (0.0008)	−0.0013 (0.0008)
Conflict				−0.0001 (0.0008)	−0.0003 (0.0008)
Policy					−0.0023** (0.0010)
Significance level of fixed effects
Market	0.041**	0.002***	0.000***	0.009***	0.096*
Time	0.000***	0.000***	0.000***	0.000***	0.000***
Information criteria
AIC	−14,076	−14,079	−14,059	−14,057	−14,061
BIC	−14,018	−14,015	−13,988	−13,980	−13,977
N	4,732	4,732	4,704	4,704	4,704

Source: The authors.

Notes: Includes the extended multifactor model of Chen et al. (1986), introducing one underlying dimension of political risk at a time, using maximum likelihood. The standard error is reported beneath each estimate. The emerging market and the latter time period are set as redundant variables. ***, ** and * denote statistical significance at 1%, 5% and 10%, respectively. μ (λ) denotes dummy for market (time). Dependent variable: Excess return. The sample period spans from June 2001 to May 2015. AIC is the Akaiki Information Criterion and BIC is the Bayesian Information Criterion.

We investigate the impact of the market in interaction with underlying dimensions of political risk to identify whether the dimensions affect excess returns differently between markets. The results are presented as model 6 in Table 3. The market-fixed effect is significant, whereas the time-fixed effect is highly significant. Policy seems to be derived from the market effect, given the highly significant interaction term, as legal turns significant. The effect of policy is positive (negative) in the emerging (developed) markets. Hence, policy affects excess return differently across markets and provides empirical support for the first hypothesis.

We investigate the impact of time in interaction with underlying dimensions of political risk to identify whether the dimensions affect excess returns differently over time. The results are presented as model 7 in Table 3. In cases where the interaction term with respect to time becomes significant in Type III tests of fixed effects, it is not necessarily possible to reveal which period is significant compared to another. This applies especially in cases where the significance level is close to 0.05. The highly significant time-fixed effect overrules the market-fixed effect, possibly as a result of severe changes in the country’s politics during and after the global financial crisis (Hoshi, 2011). Furthermore, the global convergence finally seems present, indicating a correlation across markets with respect to the underlying dimensions of political risk. Only the main effect of tension is (weakly) significant. However, the effects are significant in interaction with time in the case of policy, legal and conflict (0.01, 0.05 and 0.10, respectively), that is, they affect excess returns differently over time, ultimately leading to negative coefficients in the first sub-periods. This provides strong empirical support for the second hypothesis.

Table 3.

Two-way Error Component

Estimates of Fixed Effects Including Type III Tests of Fixed Effects
Model	6	7	8
Variables	Estimates
Estimates of intercept
Intercept	0.0008 (0.0027)	0.0023 (0.0023)	0.0015 (0.0028)
μ	−0.0058** (0.0027)	−0.0017 (0.0027)	−0,0022 (0.0029)
λ₁	0.0104*** (0.0023)	0.0098*** (0.0023)	0.0109*** (0.0024)
λ₂	−0.0040* (0.0022)	−0.0068*** (0.0023)	−0.0061*** (0.0023)
Estimates of independent variables
RP	1.0552*** (0.0150)	1.0571*** (0.0150)	1.0568*** (0.0149)
TS	−0.0007*** (0.0002)	−0.0007*** (0.0002)	−0.0007*** (0.0002)
CPI	0.5271** (0.2494)	0.4685* (0.2484)	0.4824* (0.2489)
IP	0.1665*** (0.0291)	0.1674*** (0.0290)	0.1657*** (0.0290)
Legal	0.0011** (0.0019)	0.0035 (0.0019)	0.0011 (0.0024)
Tension	−0.0013 (0.0010)	−0.0013* (0.0014)	−0.0016* (0.0016)
Conflict	0.0007 (0.0011)	0,0000 (0.0015)	0.0008* (0.0016)
Policy	0.0006** (0.0015)	0,0016 (0.0016)	0.0042 (0.0020)
Estimates of market-specific interaction terms
δ_DM,Legal	0.0031 (0.0026)		0.0035 (0.0026)
δ_DM,Tension	0.0003 (0.0018)		0.0001 (0.0018)
δ_DM,Conflict	−0.0025 (0.0020)		−0.0030* (0.0017)
δ_DM,Policy	−0.0062*** (0.0018)		−0.0052*** (0.0019)
Significance level of time-specific interaction terms
ρ_T,Legal		0.0250**	0.1690
ρ_T,Tension		0.7590	0.5930
ρ_T,Conflict		0.0560*	0.0520*
ρ_T,Policy		0.0000***	0.0000***
Estimates of time-specific interaction terms
ρ_T1,Legal		−0.0047** (0.0019)	−0.0023 (0.0021)
ρ_T2,Legal		−0.0052** (0.0023)	−0.0043* (0.0023)
ρ_T1,Tension		0.0005 (0.0019)	0.0011 (0.0020)
ρ_T2,Tension		−0.0009 (0.0020)	−0.0008 (0.0020)
ρ_T1,Conflict		0.0005 (0.0019)	0.0012 (0.0020)
ρ_T2,Conflict		−0.0041* (0.0022)	−0.0035 (0.0022)
ρ_T1,Policy		−0.0105*** (0.0022)	−0.0107*** (0.0022)
ρ_T2,Policy		0.0015 (0.0026)	0.0008 (0.0026)
Significance level of fixed effects
Market	0.035**	0.539	0.438
Time	0.000***	0.000***	0.000***
Information Criteria
AIC	−14073	−14091	−14098
BIC	−13963	−13955	−13937
N	4704	4704	4732

Source: The authors.

Notes: This model includes the extended multifactor model of Chen et al. (1986), using maximum likelihood. Model 6 reports the model including interaction terms between the underlying dimensions of political risk and market. Model 7 reports the model including interaction terms between the underlying dimensions of political risk and time. Model 8 combines both interaction terms. The standard error is reported beneath each estimate. The emerging market and the latter time period are set as redundant variables. ***, ** and * denote statistical significance at 1%, 5% and 10%, respectively. δ (ρ) denotes market-specific (time-specific) interaction terms. μ (λ) denotes dummy for market (time). The sample period spans from June 2001 to May 2015.

The main model is specified by Equation (1), and the results are presented in model 8 in Table 3. The underlying dimensions of political risk, except from legal,⁸ are significant in one way or another. The control variables are highly significant, except for CPI, which is weakly significant. The time-fixed effect is highly significant, whereas the market-fixed effect is not. The main effects of tension and conflict are weakly significant in addition to both interaction terms with respect to conflict. Policy is highly significant in interaction with both time and market, indicating that the interaction terms overrule the main effect.

4.3 Discussion

Tension is the only variable that becomes significantly negative across markets and over time, indicating that this is an economic state variable that systematically contributes with a positive correlation to stock market excess returns. Excess return is expected to decrease 0.156 per cent in month t, given a unit increase in month t-1 (i.e., decreased risk), ceteris paribus.

Conflict correlates negatively with stock market excess returns in the emerging market. Excess returns are expected to decrease 0.22 per cent (note that the interaction term must be included) in month t, given a unit increase in month t–1 (i.e., decreased risk), ceteris paribus, in the case of the developed market in the latter sub-period. The developed market differs significantly from the emerging market and contributes with a 0.30 per cent decrease in excess returns. The results are in line with Harvey (2004), who suggested that ‘External Conflict’ and ‘Internal Conflict’ are drivers of positive hedge returns in the developed market, whereas the impacts in the emerging market are conflicting. The social aspect of political risk (in the meaning of both tension and conflict) contributes consistently to increased excess returns in developed markets. In emerging markets, tension has a larger impact on excess returns than conflict. The two coefficients further contribute with opposite signs, ultimately suggesting distinct underlying dimensions of political risk. This is in contrast to the studies of Toft (2008), Berggren et al. (2012, 2015) and Lehkonen and Heimonen (2015). The social aspect of political risk is suggested to be negatively related to both growth and return (Berggren et al., 2012; Lehkonen & Heimonen, 2015). Hence, the effect of conflict could possibly outperform the effect of tension and thus result in an overall negative relation, contributing to an understanding of the political risk sign paradox. Finally, our findings support the social aspect of political risk as an important dimension. However, the impact of tension and conflict differs, making the dimensions highly important to consider separately.

Concerning policy, we see that excess returns are expected to decrease 1.178 per cent in month t, given a unit increase in month t (i.e., decreased risk), ceteris paribus, in the case of the developed market in the first sub-period. The positive correlation contradicts the findings of Dimic et al. (2015). Furthermore, the pre-crisis period differs significantly from the post-crisis period, leading to a significant decrease in the impact of policy. This is consistent with the shift in the declining trend of policy. This trend is somewhat more prominent in the developed market, possibly explaining why the market differs significantly from the emerging market, contributing with a 0.519 per cent decrease in excess returns. The negative correlation in the emerging market during the latter periods is in line with Berggren et al. (2012), who suggest a negative relation between ‘Policy’ and economic growth in poor countries.

Finally, legal turns insignificant, in line with the findings of Dimic et al. (2015). Nevertheless, legal varies significantly during the global financial crisis as compared to the other periods, implying that a certain risk level is necessary to obtain a RP. Legal is also significantly positive during several model specifications as illustrated in Table 2. Hence, legal is apparently an important political risk factor.

In sum, our analysis indicates that underlying dimensions of political risk are economic state variables such as tension, conflict and policy, systematically affecting stock market excess returns. Tension affects excess returns positively both across markets and over time, in line with the classical risk–return relation. Both conflict and policy vary across markets and over time, leading to a positive (negative) risk–return relation with excess return in the case of developed (emerging) markets. Hence, underlying dimensions of political risk affect stock market excess returns differently, which is in line with the expectations in Hypothesis 1. Furthermore, we observe that political risk has additionally increased, in general, during the twenty-first century, something which could be explained by emerging markets’ increasing influence on the world economy. This trend is especially prominent in terms of policy in the developed market, where investors seem to be rewarded in terms of both policy and conflict. These dimensions vary both across markets and over time, confirming both hypotheses. However, when assessing the interaction terms separately, only policy varies between markets. All dimensions, except tension, vary over time. This provides partial support for Hypothesis 2.

4.4 Limitations

A word of caution is, however, warranted. First, the results may be biased resulting from the omission of countries with extreme values of political risk due to the availability of the macroeconomic factors. Choosing control variables from the criterion of availability could have led to an even larger sample size, as in Lehkonen and Heimonen (2015). A larger sample size would further allow computing factor prices in addition to factor-specific loadings, following the two-step regression of Fama and MacBeth (1973). Second, a linear model may not be appropriate because of non-symmetric data and weak relations between the underlying dimensions of political risk and excess return. Using quantile regression would offer a richer insight into the entire distribution of the dependent variable. Information lags are furthermore likely to vary across both countries and markets (Bilson et al., 2001). Specifying a model for each market could thus be more appropriate. However, if we had done so, we would have lost the ability to test for significant differences between markets. Third, because it may be challenging to distinguish political risk from country-specific risk (Jakobsen, 2012b), including both the economic and financial aspects from ICRG, this could have led to an even richer insight into the dimensions of political risk.

Finally, we acknowledge that using the US treasury bond as a risk-free rate results in some problems as excess equity returns in each particular country entail currency risk which the US treasury bond yield does not have. Foreign exchange (FX) risk could affect stock market returns, especially in cases of large and sudden exchange rate changes. Our model is not able to capture the FX exposure for equity premium.

5. Concluding Remarks

Our study acknowledges the multidimensional nature of political risk and contributes to an understanding of the political risk sign paradox. We managed to operationalise political risk into four distinct dimensions, which we show to be a contribution in itself. Moreover, we identify tension, representing ethnic and religious tensions, which improves the understanding of political risk reward compared to previous research, highlighting the social aspect of political risk in general. We clearly show how aggregated political risk measures are too coarse for both analytical and investment purposes.

Political risk has increased in general throughout the twenty-first century, and recent developments in the international political arena have actualised this topic more than ever. However, stock market excess returns are highly context-sensitive as underlying dimensions of political risk affect stock market excess returns differently, thereby answering the main question of the study and posing great challenges for global investors. Although the total risk is what matters, an implication of this study is that internationally oriented investors are likely to benefit from knowledge regarding which of the dimensions implies a reward. Investors, in particular, should direct their attention towards tension, which seems to command a RP regardless of both market and time.

Future research should pursue the findings of this study by constructing portfolios that implement this knowledge in order to be more normative in explaining how to achieve excess returns.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: Frode Kjærland acknowledges funding from the regional research fund of Mid-Norway (sub project 68095104).

Footnotes

Acknowledgements

We express our gratitude to associate professors Randi Hammervold and Jo Jakobsen, the International Monetary Fund, and the Political Risk Service Group for kindly meeting our requests regarding methodical, theoretical and data support.

Appendix

Table A3.

Correlation Matrix of the Dependent Variable (ER) and Independent Variables

	RP	TS	CPI	IP	Legal	Tension	Conflict	Policy	ER
RP	1	0.006	−0.034	−0.002	−0.013	−0.001	0.007	0.021	0.709
TS	0.006	1	−0.060	−0.039	−0.004	0.007	0.007	0.016	−0.043
CPI	−0.034	−0.060	1	0.030	−0.141	−0.150	−0.049	−0.017	0.009
IP	−0.002	−0.039	0.030	1	−0.024	−0.014	−0.003	0.013	0.063
Legal	−0.013	−0.004	−0.141	−0.024	1	0.000	0.000	0.000	−0.019
Tension	−0.001	0.007	−0.150	−0.014	0.000	1	0.000	0.000	−020
Conflict	0.007	0.007	−0.049	−0.003	0.000	0.000	1	0.000	0.000
Policy	0.021	0.016	−0.017	0.012	0.000	0.000	0.000	1	0.001
ER	0.709	−0.043	0.009	0.063	−0.019	−0.020	0.000	0.001	1

Notes: Excess return (ER), risk premia (RP), term structure (TS), inflation (CPI) and industrial production (IP). The sample period spans from May 2001 to May 2015.

Notes

References

Alesina

, & Perotti

(1996). Income distribution, political instability, and investment. Journal of the European Economic Association, 40(6), 1203–1228.

Alesina

, Devleeschauwer

, Easterly

, Kurlat

, & Wacziarg

(2003). Fractionalization. Journal of Economic Growth, 8(2), 155–194.

Bekaert

, Harvey

C. R.

, & Lundblad

(2005). Does financial liberalization spur growth? Journal of Financial Economics, 77(1), 3–55.

Bekaert

, Harvey

C. R.

, Lundblad

C. T.

, & Siegel

(2014). Political risk spreads. Journal of International Business Studies, 45(4), 471–493.

Berggren

, Bergh

, & Bjørnskov

(2012). The growth effects of institutional instability. Journal of Institutional Economics, 8(2), 187–224.

Berggren

, Bergh

, & Bjørnskov

(2015). What matters for growth in Europe? Institutions versus policies, quality versus instability. Journal of Economic Policy Reform, 18(1), 69–88.

Bilson

C. M.

, Brailsford

T. J.

, & Hooper

V. J.

(2001). Selecting macroeconomic variables as explanatory factors of emerging stock market returns. Pacific-Basin Finance Journal, 9(4), 401–426.

Bilson

C. M.

, Brailsford

T. J.

, & Hooper

V. J.

(2002). The explanatory power of political risk in emerging markets. International Review of Financial Analysis, 11(1), 1–27.

Bremmer

(2012). The return of emerging market political risk. Harvard Business Review November.

10.

Bremmer

, & Keat

(2010). The fat tail: The power of political knowledge in an uncertain world. Oxford: Oxford University Press.

11.

Brogaard

, & Detzel

(2015). The asset-pricing implications of government economic policy uncertainty. Management Science, 61(1), 3–18.

12.

Brooks

(2008). Introductory econometrics for finance (2nd ed.). Cambridge: Cambridge University Press.

13.

Chen

, Roll

, & Ross

S. A.

(1986). Economic forces and the stock market. The Journal of Business, 59(3), 383–403.

14.

Damodaran (2016). Data: Archives. Retrieved from http://pages.stern.nyu.edu/~adamodar/

15.

Diamonte

R. L.

, Liew

J. M.

, & Stevens

R. L.

(1996). Political risk in emerging and developed markets. Financial Analysts Journal, 52(3), 71–76.

16.

Dimic

, Orlov

, & Piljak

(2015). The political risk factor in emerging, frontier, and developed stock markets. Finance Research Letters, 15(C), 239–245.

17.

Erb

C. B.

, Harvey

C. R.

, & Viskanta

T. E.

(1996a). Political risk, economic risk, and financial risk. Financial Analysts Journal, 52(6), 29–46.

18.

Erb

C. B.

, Harvey

C. R.

, & Viskanta

T. E.

(1996b). Expected return and volatility in 135 countries. The Journal of Portfolio Management, 22(3), 46–58.

19.

Fama

E. F.

, & MacBeth

J. D.

(1973). Risk, return and equilibrium: Empirical tests. Journal of Political Economy, 81(3), 607–636.

20.

Harvey

C. R.

(2004). Country risk components, the cost of capital, and returns in emerging markets (Working Paper). Retrieved from http://ssrn.com/abstract=620710

21.

Hoshi

(2011). Financial regulation: Lessons from the recent financial crises. Journal of Economic Literature, 49(1), 120–128.

22.

Howell

L. D.

(2011). International country risk guide methodology. Retrieved from http://www.prsgroup.com/wp-content/uploads/2012/11/icrgmethodology.pdf

23.

International Labour Organization (ILO) (2017). World employment and social outlook: Trends 2017. Retrieved from http://www.ilo.org/global/research/global-reports/weso/2017/WCMS_541211/lang--en/index.htm

24.

Jakobsen

(2010). Old problems remain, new ones crop up: Political risk in the 21st century. Business Horizons, 53(5), 481–490.

25.

Jakobsen

(2012a). Hands off my investment! A time series cross-section analysis of the relationship between a new measure of political risk and foreign direct investment. In DeSare

& Caprioglio

(Eds.), Foreign investment: Types, methods, and impacts (Ch. 5). New York, NY: Nova Science Publishers.

26.

Jakobsen

(2012b). Political risk and the multinational company: Concepts, theories and evidence. Trondheim: Tapir Academic Press.

27.

Lam

S. S.

, & Zhang

(2014). Does policy uncertainty matter for international equity markets? (Working Paper). Retrieved from http://ssrn.com/abstract=2297133

28.

Lehkonen

, & Heimonen

(2015). Democracy, political risks and stock market performance. Journal of International Money and Finance, 59(C), 77–99.

29.

Morgan Stanley Capital International (MSCI) (2015). Market classification framework. Retrieved from https://www.msci.com/documents/10199/0ebf2d14-67a9-4b5f-96b0-d2a265a7c25d

30.

Morgan Stanley Capital International (MSCI) (2016). MSCI global investable market indexes methodology. Retrieved from https://www.msci.com/eqb/methodology/meth_docs/MSCI_May2016_GIMIMethodology.pdf

31.

Pástor

, & Veronesi

(2013). Political uncertainty and risk premia. Journal of Financial Economics, 110(3), 520–545.

32.

Perotti

E. C.

, & van Oijen

(2001). Privatization, political risk and stock market development in emerging economies. Journal of International Money and Finance, 20(1), 43–69.

33.

Ross

S. A.

(1976). The arbitrage theory of capital asset pricing. Journal of Economic Theory, 13(3), 341–360.

34.

S&P Dow Jones Indices (2013). Index consultation. Retrieved from https://www.djindexes.com/mdsidx/downloads/index_consultation/S&P_Dow_Jones_Country_Classification_Consultation_08142013.pdf

35.

The Heritage Foundation (2016). About the index. Retrieved from http://www.heritage.org/index/about

36.

The World Justice Project (2016). Methodology. Retrieved from http://worldjusticeproject.org/methodology

37.

Toft

(2008). Project finance and political risk (Master’s thesis). Sweden: Aarhus School of Business.