Modelling the Country Risk of Zambia

Journal of Political Risk, Vol. 10, No. 3, March 2022

By Simon Muwando, Victor Gumbo, and Gelson Tembo

Abstract

The world has experienced a dramatic increase in the flow of transnational investments following increased internationalization and globalization of firms in the previous decade. Country risk exposure is a cause for concern for all the institutions that are engaged in multinational trade and finance. The main objective of this study is modelling Zambia’s country risk. A mixed method with concurrent research design was employed. Personal interviews were the main instrument for collection of primary data and snowball sampling was used to select the interviewees. Secondary data was collected from the Lusaka Stock Exchange (LSE), Ministry of Finance, Bank of Zambia and Central Statistical Office. An autoregressive distributed lag technique was employed on annual data for the 1994 to 2018 period. This approach was chosen as it works best for small samples. The findings of the study revealed that the short run drivers for country risk of Zambia are beta, current account balance, political risk, unemployment rate and weighted short term interest rates. Current account balance was found to positively affect country risk while beta, political stability, and weighted short term interest rates negatively influence it.  The study findings established that the long run determinants of country risk of Zambia are current account balance, betas, political risk, and unemployment rate. From the study findings, current account balance positively influences country risk of Zambia whereas beta, and political stability negatively influence country risk of Zambia. The study concluded that the major determinant of country risk of Zambia in the short run and long run is current account balance as it has significant positive influence. Effective policies need to be implemented by authorities to manage or reduce persistent current account deficits and political risk, in order to manage country risk.

1. Introduction

The computation and the trade-off between country risk and macroeconomic variables is a topic of concern within the multinational trade and finance circus. All institutions that are involved in managing international investment portfolios are vulnerable to country risk. According to Shapiro (1999), country risk is defined as “the general level of political and economic uncertainty in a country affecting the value of loans or investments in that country”. Political crises in the 1960s and 1970s, the oil crisis and collapse of the fixed exchange rate system in the 1970s, financial debt crises in the 1980s, financial crises in the 1990s and subprime lending of 2007, and the sovereign and private debt crisis in the Euro area and non-Euro member state countries, are instances of country risk (San-Martin-Albizuri and Rodriguez-Castellanos, 2018). All of these crises renewed the interest of different stakeholders in the concept of country risk and fuelled an on-going debate among rating agencies, policymakers (including public debt managers, bank regulators, fiscal authorities and central bankers) and academics on how to measure or estimate country risk (Blommestein and Turner, 2012). Modern contributors to this debate advocate for a set of indicators that they think capture country risk; these criteria range from macroeconomic to financial formulas through to credit ratings (Blommestein, Guzzo and Holland, 2010). Despite the pros and cons in each of the recommended measures, no single one has emerged entirely acceptable (Blommestein and Turner, 2012).

Empirical studies have used different proxies to measure country risk. Verma and Verma (2014) analysed the responsiveness of country risk in Asian markets to a group of domestic and global macroeconomic variables. The difference between the return from the equities and risk-free rate of return was used as the proxy for measuring country risk. The findings of the study established that, among the global factors, the exchange rate (US dollar price) had a significant positive effect on country risk except in the case of Malaysia. The findings on impact of exchange rate volatility on country betas concurs with Oetzel, Bettis and Richards (2000), Gangemi, Brooks and Faff (2000), Jeon (2001), Verma and Soydemir (2006) and Basu, Deepthi and Reddy (2011) who found that currency risk is a major driver of country risk. This is because we live in a global village where markets are highly integrated, thus global factors influence country risk the most. However, this contrasts with Muwando and Gumbo (2013) who found that political risk is the significant factor influencing Zimbabwe’s country risk.

Tourani-Rad, Choi and Wilson (2006) estimated the country risk of New Zealand from 1985 to 2000. Their macroeconomic variable set includes New Zealand’s commodity index, net trade balance, USD/NZD exchange rate, AUD/NZD exchange rate, M3 money supply, 90-day bill yield, 10-year government bond yield, food price index, monetary conditions index, and trade-weighted index. The study used the difference between local market index return and world market returns as a proxy for measuring country risk. A multivariate regression analysis technique was used to estimate country volatility. The findings of the study revealed that the USD/NZD exchange rate and monetary conditions index were the major drivers of country betas. Using the ARIMA technique, Andrade and Teles (2004) studied the country risk of Brazil from 1991 to 2002, then adopted the same proxy as Tourani-Rad, Choi and Wilson. This method is prone to large errors, putting the results into question. They found that manipulation of the nominal interest rate is essential to reducing country risk.

Goldberg and Veitch (2002) analysed the determinants of country risk for Argentina for the 1992 to 2008 period; the variables include consumer price index, exchange rate (Argentina), exchange rate (Brazil), exchange rate (Chile), exchange rate (Mexico), money supply, reserve money and international reserves. An ARIMA technique was employed to estimate the model. The findings of the study established that only Brazilian and Mexican exchange rate crises, not Argentinean macroeconomic fundamentals, were the major variables influencing Argentina’s country risk. This implies that the contagion effect was the main determinant of country risk changes. These findings are in line with Goldberg and Veitch (2010) who established that foreign exchange rates and gold prices were major drivers of South Africa’s risk prior to financial integration. However, they are in contrast with Gangemi, Brooks and Faff (2000) who, using the same approach, modelled Australia’s country risk for the 1974 to 1994 period. The trade-weighted index was found as the only variable with substantial positive impact on country risk and asset returns. The study also found that a gain in value of the local currency positively impacted Australian country risk and that external shocks are essential to the performance of the economy. Therefore, country risk is mainly influenced by economic, financial, and political variables (see also Erb, Harvey and Viskanta (1996); Groenewold and Fraser, 1997; Bracker and Koch, 1999; Oetzel, Bettis and Richards, 2011). They used the same approach and proxy, which again is prone to large errors, making it unreliable.

Below the paper is subdivided into five additional sections: Section 2 provides a brief overview of the Zambian economy. Section 3 presents the study’s conceptual framework. Section 4 outlines the econometric methodology that was adopted. Section 5 analyses, interprets and discusses the study’ findings. Finally, Section 6 presents the conclusions of the study.

2. Background of the possible financial, economic and political variables driving the country risk for Zambia

Despite its abundant natural resources, Zambia is perceived as one of the most impoverished countries in Africa with the majority (58%) earning less than the International Poverty Datum Line of $1.90 daily (as of 2015). Its per capita GDP rose from USD 611.87 in 1994 to USD 1352.16 in 2018 (World Bank, 2019). Zambia’s major investors include the US, South Africa, India, Japan, and several EU countries, namely the UK, Netherlands, and Sweden. Copper production is one of the key factors that drive its economic growth and development (Page and Velde, 2004). Its over-dependence on copper has exposed it to volatile commodity prices. The Zambian economy is perceived as a risk destination for FDI inflows due to increasing cases of political violence and corruption, as shown by low aggregate values of control of corruption index from 1980 to 2012 (Mbao, 2011; GAN Business Anti-corruption report, 2017; World Bank, 2013).  Electoral malpractices were rampant in Zambia during the 1980-2012 period (Yezi, 2013, p.17). Figure 1 above shows the international ranking of Zambia in terms of corruption. It indicates that Zambia is perceived as one of the most corrupt countries in the world with a higher ranking during of 2001 to 2018.

According to Maravi (2007) and Banda (2013), Zambia still needs to revise its investment policies to attract more FDI.  Most of the emerging economies are well ahead of the Zambian economy in terms of FDI performance. As a result of introducing new reforms in the 1990s, Zambia’s FDI inflows grew positively (Development Policy Research Unit, 2000; United Nations Conference on Trade and Development, 2006). The FDI and portfolio investment inflows decreased in 2018 (World, 2019).

Interest rates and inflation levels are still high in these countries. This is due to the continuous surge in cost of living; that is, high commodity prices for fuel and food. The Zambian authorities reduced the inflation rate to a single digit figure in 2006 (Central Statistical Office Zambia, 2012; World Bank, 2013). Inflation in Zambia has hovered around its medium term range of 6 – 8% (World Bank, 2019). Zambian exchange rates depreciate from 0.669 kwacha per USD to an average of 10.45 kwacha per USD in 2018 (Central Bank of Zambia, 2018; World Bank, 2019). This shows that the Zambian local currency fluctuates comparatively to other regional currencies.

Unemployment rates are extremely high in Zambia. The majority are languishing in poverty and remain unemployed (Yezi, 2013:18). In 2010, unemployment rates stood at 13%, respectively, vis-a-viz, the country’ constant economic growth rate of a minim 3% annually. This may imply that the country’s public resources are concentrated among very few individuals (Africa Development Bank, 2015; Organisation for Economic Co-operation Development, 2015 and United Nations Development Programme, 2015). The 2017 unemployment rate stood at 41.2% (Central Statistical Office Zambia, 2018).

The Zambian economy has performed relatively well within the Southern African Development Community (SADC) region despite a decline in GDP per capita and a lower economic growth rate caused by reduced production in the mining sector (Rasmussen, 2015; United Nations Development Programme, 2012). The SADC region faces an unusual heavy debt burden in comparison with other low-income countries. Zambia’s debt stood at 206.1% of GDP in 1991 and fell to 78.1% in 2018 (International Monetary Fund, 2012 and 2019). All these ratios are way above the recommended 3% of GDP indicating that this country may suffer from the debt trap in the long run.

Persistent current account deficits continue to haunt this country (World Bank, 2019). The Zambian government failed to meet the SADC target of a current account deficit of less than 9% in 2011 due to its over-dependence on imports (United Nations Development Programme, 2014 and 2015; SADC, 2014). Despite the risk associated with Zambia, the African Development Bank dedicated more than USD 1 billion on supporting the public sector infrastructure projects. Moreover, it also profited from debt relief under the Heavily Indebted Poor Country and Multilateral debt initiatives (African Economic Outlook, 2019).

According to the International Monetary Fund (IMF) report on Zambia (2019), the country’s medium-outlook is oblique as it is facing many economic challenges including severe debt exposures; drought and a subdued mining sector that are stunting the growth of the economy; widening current account deficits and inflationary pressures leading to exchange rate depreciation, increase in debt servicing costs and subsequently crowding out private social investment.

3. Conceptual framework for the current study

From the previous review of theoretical literature, the conceptual framework for the country risk as a function of economic, financial and political variables is in Figure 2 below.

4. Methodology

A mixed method with concurrent research design was employed in order to improve the accuracy of judgments by collecting different kinds of data surrounding the same phenomenon, namely, country risk. In terms of the quantitative approach, the study employed the Autoregressive Distributed Lagged (ARDL) Bounds test procedure on annual data collected from 1994 to 2018, in order to identify the determinants of Zambia’s country risk. This approach was adopted because all the variables were stationary at level and at the first difference; moreover, it accommodates for a small sample size. It also allows for the integration of the multiple independent variables in the model to estimate the dependent variable. In other words, an ARDL model was used to assess the explanatory power of macroeconomic factors on Zambia’s country risk. To complement the quantitative research design, an exploratory design through personal interviews was used. This was done to gain deeper comprehension of the main determinants of country risk, and to find out the respondents’ view on the impact of political, economic and financial risk on Zambia’s country risk.

The drivers of country risk are categorized into economic, financial and political variables (Hoti, 2005). Political variables are composed of legal factors, political instability; economic factors include per capita GDP, GDP deflator, current account balance, and unemployment rate; financial variables are composed of external debt balance, and short term interest rates.

4.1 Model Specification

According to Vij (2005), the country risk model can be expressed as follows:

(1)

The notation Xni indicates the values of the nth independent variable for the case i. The beta terms are unknown parameters and the εi terms are independent random variables that are normally distributed with mean zero and constant variance, δ2.

Erb, Harvey and Viskanta (1996a) express country risk as:

(2)

Where:     is the economic-related risk for country i in the period t;

 is the political-related risk for country i in the period t;

    is the financial-related risk for country i in the period t.

This means that country risk in equation (2) depends on economic related risk, political related risk and financial related risk.

In this study, the country beta model further takes the following form:

(3)

Where:            CRi is the country risk at time t;

                         α is the intercept or constant;

                        βi  to  βn are unknown parameters;

                       X1i  to Xni are country risk drivers;

                      εt terms are independent random variables that are normally distributed with mean zero and constant variance, σ2.

According to Choong et al. (2003) citing Pesaran, Shin and Smith (2001), the ARDL technique is applied by modelling the long-run equation [4] as a general vector autoregressive [VAR] model of order p in zt. This implies that:

(4)

Where  represents observation z at time t

             represents observation at time t-i

 represents [k + 1] – a vector of intercept [drift];

            α represents [k + 1] – a vector of trend coefficients;

            represents model coefficients.

Pesaran, Shin & Smith (2001) further proposed the following vector error correction model [VECM] corresponding to [4]:

 (5)

Where ≡ 1- L is the difference operator,

In this study, Zt = (CA, CAPITA, DEFLATOR, ED, PSAV, UN, WSTIR). Γ is an n x n matrix (short run dynamics coefficients), = αβ′ where α is an n x 1 column vector (the matrix of loadings) denoting the speed of short run adjustment to disequilibrium and β′ is an 1 x n cointegrating row vector (the matrix of cointegrating vectors) representing the matrix of the coefficients of long run dynamics such that Yt converge in their long run equilibrium. Finally, εt is an n x 1 vector of white noise error term (Choong et al., 2003; Oteng-Abayie and Frimpong, 2006). In other words,  is the vector of variable  and       respectively; Yt  is an I(1) dependent variable denoted by CRt ;  (CA, CAPITA, DEFLATOR, ED, PSAV, UN, WSTIR) a vector matrix of I(0) and I(1).

The conditional Vector Error Correction Model (VECM) becomes:
(6)

The determinants of country risk used in this study were derived from previous empirical studies of country risk that dealt exclusively with emerging market equity returns and from the suggestion of theoretical research on sovereign and international borrowings (Basu, Deepthi and Reddy, 2011; Tourani-Rad, Choi and Wilson, 2006; Andrade and Teles, 2004; Gangemi, Brooks and Faff, 2000; Vij, 2005; Wdowinski, 2004; Goldberg and Veitch, 2002). Moreover, choice of variables was subject to data availability.  The set of macroeconomic factors (independent variables) chosen and assessed had a major domestic and international influence on the Zambian economy. These include political risk, GDP deflator, per capita GDP, external debt balance, current account balance, interest rate and unemployment rate.

The procedure for the selection of various independent variables used to estimate country risk was as follows:

Pesaran and Pesaran (2009) and Pesaran, Shin and Smith (2001) advocated an ARDL bound testing technique that was employed to test the impact on country risk, as measured by annual country betas; of economic, political and financial variables and also to establish the behaviour of country risk drivers in the short run and long run. The major merit of an ARDL method over other techniques is that it is used in time-series data notwithstanding their order of integration of variables, that is whether I(0), I(1) and/or fractionally integrated (Almahmoud, 2014 citing Pesaran and Pesaran, 2009). Furthermore, the technique can also test for cointegration by the bounds testing approach and then estimate the short run and long run dynamics (Almahmoud, 2014, p.89; Nkoro and Uko, 2016). It also captures the dynamic effects of both the lagged dependent variables that represent the autoregressive portion and lagged independent variables that constitute the distributed part of the model. Omission of variables and autocorrelation in the error term can be eradicated when the appropriate number of lags of regressor and regressand variables are factored into the model (Gujarat, 2012). The technique is also robust and efficient with samples of different sizes, especially those of smaller sizes for instance, the present study. From equation [6] above, the conditional VECM is expressed in the following form:

4.2 ARDL Bounds Testing Procedure

According to Kumar (2010), the ARDL Bounds test procedure fundamentally encompasses three steps. First, equation [7] is estimated using the Ordinary Least Squares (OLS) method in order to determine the presence of long run dynamics among the selected factors by performing a joint hypothesis F-test for the lagged variables (Oteng-Abayie and Frimpong, 2006; Saungweme and Odhiambo, 2019).

This implies that the following hypothesis is to be tested as follows:

The test which normalizes CRt is denoted by 

According to Kumar (2010) and Pesaran, Shin and Smith (2001, p.290), two asymptotic critical values bounds provide a test for cointegration when the explanatory variables are integrated at level d, that is I(d)  where 0 ≤ d ≤1.  The lower value of d assumes that the explanatory variables are stationary at level, I(0) and the upper value of d assumes that they are purely stationary at the first difference, I(1). Suppose the F-calculated is larger than the upper F-critical value, we reject the Ho and conclude that there is a long run relationship among the series despite the orders of integration for the time series. On the other hand, if the F-calculated is less than the lower critical value, we fail to reject the null hypothesis and conclude that there is no long run relationship among the series. Finally, if the F-calculated lies between the lower and the upper critical values, the result cannot be concluded (Nieh and Wang, 2005, Ben Jebli, 2016). The critical values used in this study were extracted from Pesaran, Shin and Smith (2001) table.

Second, if cointegration exists, the conditional ARDL(p,q1,q2,q3,q4,q5,q6,q7) long run model for CRtis estimated as follows:


The orders of the ARDL (p,q1,q2,q3,q4,q5,q6,q7) model in the seven variables is chosen using three criterions: Akaike Information Criterion (AIC), Schwarz Information Criterion (SIC) and Hannan-Quinn criterion (HQC) criterion (Pesaran and Smith, 1995).

Lastly, the Error Correction Model (ECM) is estimated to capture the short-run coefficients of the model. The ECM has the following specifications:

Betast is the outcome of the covariance between the local equity index return and World Market equity index return divided by the variance of the world market index return. The local equity index is the locally denominated stock indexes for Zambia (LSE). The proxy for the global market index is the MSCI emerging markets Index. MSCI emerging markets Index was chosen because it comprises stocks in emerging economies hence, it is the best benchmark for comparison with emerging economies of Zambia.

Stock Index Returns were computed using the formula given below:

Where           Rt represents Stock Index Returns at time t;

                       St represents Stock index at time t;

                        St-1 represents Stock index at time t lagged once.

The computed returns in Equation (10) are log-normalised in order to improve the normality of the Betastparameter and this confirms the significance of normality in all statistical analysis.

The economic, financial and political variables mentioned above serve as the explanatory variables that were used to compute the predictive power of the dependent variable Betast (Muwando and Gumbo, 2013). External debt and current account balance portrays the role of the fiscal authorities on the economy while interest rates reflect the monetary policy in Zambia. Political stability and absence of violence index was used as a proxy for political risk.

Rationality and consistency of the main assumptions made in the models were tested by performing the residual, stability and coefficient diagnostic tests.

The population of study encompasses bank executives, monetary authorities, fiscal authorities, potential and existing investors, especially one that appreciates the concept of country risk and the impact of economic, financial and political fundamentals on country risk. Personal interviews were mainly used as the instruments for collecting primary data, while secondary data was collected from secondary sources (Central Bank, CSO, Ministry of Finance, IMF, World Bank, newspapers and journal articles) for the chosen financial, economic and political variables, and then recorded systematically. The researcher carefully selected interviewees through snowball sampling and scheduled appointments with the key informants on the relevant study. Snowball sampling assumes relevant respondents are connected so that those connections can be used to construct a sample from a small initial sample. In other words, it involves building a sample through referrals, as each respondent recommends others (Bacon-Shone, 2013). Snowball sampling was used to explore the interviewees’ perception of the main determinants of country risk, their impact on country risk and ways of managing country risk.

5 Interpretation, Analysis and Discussion of the results

5.1 Personal interviews response rate

The personal interviews response rate is shown below.

Table 1: Interview response rate

Scheduled Interviews 15
Actual Interviews conducted 8
Response rate 53%

Source: Researcher’s own analysis using Ms Excel

Out of fifteen interviews planned to be conducted in each country, eight interviews were conducted in Zambia. This implies a 53% response rate.

5.2 The estimated annual country betas

The annual country betas (Betast) were computed by dividing the covariance of the local index returns and world market returns by the variance of world market returns. This gives the numerical value of country risk which is objective and reflective of the risk inherent in a particular country. The results of the estimated annual betas are shown Figure 3 below:

The results in Figure 3 indicate that Zambia is a moderately risky destination for investments because most of its estimated annual betas are slightly bigger. This concurs with the interviewees’ response, as they perceived Zambia’s economy to be less stable. Generally, the annual country beta values are smaller and this converges with empirical literature that emerging markets have lower beta than developed markets (Wdowinski, 2004 citing Harvey, 1995 and Erb, Harvey and Viskanta, 1996). The sharp increase in forecasted beta in Zambia may be attributed to deteriorating economic, financial and political factors.

5.3 Multicollinearity test

Two variables, per capita GDP and GDP deflator, were highly correlated. In designing the model, GDP deflator was excluded because it had a high probability value. The outcomes of multicollinearity tests are shown below:

Table 2: Correlation matrix  

CA CAPITA PSAV ED UN WSTIR
CA  1.000000  0.656600  0.626400 -0.665896 -0.205368  0.276803
CAPITA  0.656600  1.000000  0.725999 -0.909195 -0.752508  0.553528
PSAV  0.626400  0.725999  1.000000 -0.749934 -0.513872  0.153651
ED -0.665896 -0.909195 -0.749934  1.000000  0.798350 -0.336502
UN -0.205368 -0.752508 -0.513872  0.798350  1.000000 -0.354907
WSTIR  0.276803  0.553528  0.153651 -0.336502 -0.354907  1.000000

Source: Researcher’s own analysis using EViews 10

From the table above, there is no multicollinearity problem because all the correlation coefficients are less than 0.8

5.4 Normality distribution tests

The results of the Jarque-Bera tests are presented in Table 3 below.

Table 3: Normality Test  

BETAS CA CAPITA DEFLATOR PSAV ED UN WSTIR
 Mean  0.165632 -3.429200  943.9525  76.63896  0.224800  0.953671  11.00000  2.353600
 Median  0.120000 -3.300000  1030.282  66.51300  0.200000  0.534000  10.61000  1.650000
 Maximum  1.712100  7.500000  1839.537  201.3040  0.660000  2.081900  18.50000  6.260000
 Minimum -0.735000 -16.50000  330.2830  6.346000 -0.280000  0.186500  7.750000  0.660000
 Std. Dev.  0.571677  6.708557  547.4558  59.29485  0.238574  0.747204  3.049057  1.789408
 Skewness  0.966783 -0.311951  0.193542  0.539574 -0.230161  0.320238  0.780679  0.998420
 Kurtosis  3.936484  2.376537  1.425575  2.133782  2.611433  1.347535  2.970018  2.795727
 Jarque-Bera  4.807998  0.810374  2.738176  1.994679  0.378001  3.271716  2.540352  4.196975
 Probability  0.090356  0.666852  0.254339  0.368859  0.827786  0.194785  0.280782  0.122642
 Sum  4.140800 -85.73000  23598.81  1915.974  5.620000  23.84177  275.0000  58.84000
 Sum Sq. Dev.  7.843551  1080.114  7192989.  84381.11  1.366024  13.39954  223.1220  76.84758
 Observations  25  25  25  25  25  25  25  25

Source: Researcher’s own analysis using EViews 10

Since Jargue-Bera p-values in Table 3 are more than 0.05, we fail to reject Ho and conclude that all the residuals are normally distributed. Hence, the statistical data for Zambia follows a normal distribution.

 

5.5 Stationarity tests

The results of unity root tests for the model type ‘Intercept without trend’ are shown in Table 4 below.

Table 4: Augmented Dickey-Fuller (ADF) Unit roots test results for stationarity

Variable Significance Level Test statistic[I(0)] Critical values[I(0)] P-value

I(0)]

Test statistic

[I(I)]

Critical values [I(I)] P-value

[I(I)]

Annual Country Betas
1% level
5% level
10% level
 

-4.946

-3.738
-2.992
-2.636
 

0.0006

Current account as % of GDP
1% level
5% level
10% level
-1.889 -3.738

-2.993

-2.636

0.3315 -5.871 -3.753

-2.998

-2.639

0.000
Per Capita GDP
1% level
5% level
10% level
 

-0.735

-3.738
-2.992
-2.636
 

0.8193

-4.183 -3.753

-2.998

-2.639

0.004
Political Stability and Absence of Violence Index

 

 

1% level
5% level
10% level
 

 

-1.943

 

-3.738

-2.992
-2.636
 

0.3083

-5.370
-3.7536
-2.9984
-2.6392
0.000
External debt as % of GDP 1% level

5% level

10% level

-1.326 -3.753

-2.999

-2.639

0.5996 -4.643 -3.753

-2.999

-2.639

0.001
Unemployment Rate 1% level

5% level

10% level

-2.012 -3.770

-3.005

-2.642

0.2798 -4.910 -3.770

-3.005

-2.642

0.000
Weighted Short Term Interest rates 1% level

5% level

10% level

-0.462 -3.753

-2.998

-2.639

0.8819 -8.767 -3.753

-2.998

-2.639

0.000

Source: Researcher’s own compilation from EViews 10

In the Table 4 above only annual country betas are stationary at level [I(0)] while the other variables, such as current account balance, per capita GDP, political stability and absence of violence index, external debt and weighted short term interest rates were differenced once[I(1)] for them to be stationarity.

5.6 Optimum lag length

To perform a cointegration test among the variables in the ARDL bound testing, it is a prerequisite to establish the optimal lag to avoid the hypothesis of serially correlated residuals in the cointegrated equation. The researcher limits the estimation to two lags since the possibility of serially uncorrelated residuals will occur when the number of lags is increased. However, it has to be done parsimoniously to avoid an over-parameterization problem (Pesaran, Shin and Smith, 2001). The results of optimum lag selection are shown in Table 5 below.

Table 5: Optimum lag selection

Lags AIC SIC HQC
1 2.08004 2.47532 2.17970
2 0.94054** 1.68444** 1.11578**

Source: Researcher’s own compilation from EViews 10

NB: ** denotes optimal lag chosen.

In Table 5 above, lag 2 was chosen as the optimum lag for an ARDL model of Zambia as it has the lowest value for the entire three criterions.

5.7 Co-integration Testing using ARDL Bound Test

The results of the ARDL Bound test for cointegration are shown in Table 6 below.

Table 6: ARDL Bound test for Cointegration

Unrestricted intercept and no trend

Dependent variable F-statistic Upper Bound Lower Bound Remark What is next??
Betast Fbetas = 20.18

 

2.45 3.61 Cointegration exist Estimate ECM(Error Correction Model)

Source: Researcher’s own compilation from EViews 10

From the table above, the F-Statistic (20.18) is greater than I(1) the critical values (3.61) and so we reject the null hypothesis at the 5% level and conclude that there is cointegration among the variables; there is a long run relationship between country risk and a set of selected economic, political and financial variables (current account balance, per capita GDP, external debt, political stability and absence of violence index, unemployment rate and weighted short term interest rates). This is affirmed by the interviewees as they perceived current account, per capita GDP, external debt, political stability and absence of violence index, unemployment rate and weighted short term interest rates to be major drivers of Zambia’s country risk. Thus, a long run ARDL can be estimated with 2 lags for both countries.

 

5.8 Long-run dynamics results

The results of the long run ARDL model coefficients are shown in the Table 7 below.

Table 7: Estimated long run ARDL model coefficients

C Coefficient Std. Error t-Statistic P-value
C 4.58858 1.09541 4.18892 0.0030
Betas(-1) -1.15884 0.21500 -5.38988 0.0007   *
Betas(-2) -0.02909 0.15192 -0.19150 0.8529
CA(-1) 0.10487 0.01920 5.46355 0.0006   *
CA(-2) 0.08479 0.024482 3.41651 0.0091    *
Capita(-1) 0.000083 0.00068 0.12092 0.9067
Capita(-2) -0.00111 0.00060 -1.86808 0.0987
ED(-1) -0.37506 0.50720 -0.73948 0.4807
ED(-2) 0.56065 0.46501 1.20567 0.2624
PSAV(-1) -1.96895 0.51397 -3.83086 0.0050   *
PSAV(-2) -1.63500 0.56232 -2.90761 0.0197   *
WSTIR(-1) -0.15719 0.08390 -1.87378 0.0978
WSTIR(-2) 0.10437 0.06442 1.62027 0.1438
UN(-1) 0.07300 0.06649 1.09794 0.3042
UN(-2) -0.25581 0.08427 -3.03556 0.0162   *

Source: Research estimation results from EViews 10

NB   * denotes significance at 0.05

From Table 7, it can be observed that Betas in one year lag [Betas(-1)] has a significant long run relationship with country risk. One-year lagged beta is statistically significant at 5% level of significance because its p-value is less than 5%. With a coefficient of -1.15884, country risk decreases by 1.16% when annual beta increases by 1%, ceteris paribus. The long run p-values suggest that current account balance in one-year lag [CA(-1)] and two-year lag [CA(-2)] have a significant long run relationship with country risk because their p-values are less than 5%. If current account lagged once increases by 1%, country risk increases by 10% (0.10487), ceteris paribus. Furthermore, country risk increases by 8% (0.0849) when current account lagged twice increases by 1%, ceteris paribus. These findings concur with Ferreira (2010) who found that current account significantly influences the country risk of Brazil. It can also be observed that political stability and absence of violence index in one-year lag [PSAV(-1)] and two-year lag [PSAV(-2)] have a significant negative long run relationship with country risk. Political stability and absence of violence index in one-year lag and two-year lag are statistically significant at 5% level of significance since their p-values are less than 0.05. In conclusion, country risk decreases by 1.97% (-1.96785) when political stability and absence of violence index in one-year lag increases by 1%, ceteris paribus. In addition, when one-year lagged political stability and absence of violence index increases by 1% country risk decreases by 1.64% (-1.635), ceteris paribus. This is in line with Vij (2005), Basu, Deepthi and Reddy (2011) and Muwando and Gumbo (2013) who established that political risk and absence of violence is the main driver of country risk. The long run p-values indicate that unemployment rate lagged twice [UN(-2)]  has a significant  influence on country risk. Unemployment rate lagged twice is statistically significant at 5% level of significance since its p-value is less than 5%. In conclusion, country risk decreases by 25% (-0.25581) when unemployment rate increases by 1%, ceteris paribus. This contradicts the apriori conditions that an increase in unemployment increases country risk. Therefore, the long run determinants of country risk of Zambia are current account balance, betas, political risk and unemployment rate.

5.9 Error Correction Model (ECM)

The results of the error correction model are presented in the Tables 8 below.

Table 8: Estimated Error Correction Results

Variable Coefficient Std. Error t-Statistic P-value
C -0.01292 0.06718 -0.19236 0.8538
D(Betas(-1)) -1.08437 0.15893 -6.82301 0.0005 *
D(Betas(-2)) -0.06602 0.13338 -0.49501 0.6382
D(CA(-1)) 0.10076 0.01691 5.95905 0.0010 *
D(CA(-2)) 0.08088 0.01806 4.47877 0.0042 *
D(Deflator(-1)) 0.00031 0.00051 0.61540 0.5609
D(Deflator(-2)) 0.00085 0.00047 -1.80920 0.1204
D(ED(-1)) -0.17030 0.45398 -0.37512 0.7205
D(ED(-2)) 0.48840 0.40122 1.21729 0.2692
D(PSAV(-1)) -2.15318 0.42085 -5.11630 0.0022 *
D(PSAV(-2)) -1.82465 0.40413 -4.51504 0.0040 *
D(UN(-1)) 0.11060 0.05573 1.98460 0.0944
D(UN(-2)) -0.30197 0.06812 -4.43275 0.0044 *
D(WSTIR(-1)) -0.18847 0.06601 -2.85524 0.0290 *
D(WSTIR(-2)) 0.07199 0.06846 1.05163 0.3335
ECT(-1) -1.40825 0.48287 -2.91643 0.0268 *

Source: Research estimation results from EViews 10

NB   * denotes significance at 0.05 level

In Table 8, ECT(-1) = -1.4082 is statistically significant at the 5% significance level, implying that the speed of adjustment towards long run equilibrium is 140.82%. If there is shock in any of the short term variables, the whole system gets back to long run equilibrium at a speed of 140.82%. Since the model is correctly specified, a high coefficient of ECT(-1) (above 1 with negative sign and significant) may imply that the system is convergent, yet, has an oscillatory adjustment process; the error correction process fluctuates around the long run value in a dampening manner. However, once this process is complete, convergence to the equilibrium path is rapid.

Table 8 also indicates that differenced one-year lagged beta [D(Betas(-1))] has a significant short run relationship with country risk. Beta in one-year lag is statistically significant at 5% level of significance since its p-value is less than 5%. In conclusion, country risk has a 1.08% negative change when beta increases by 1%, ceteris paribus. The short run p-values also suggest that differenced one-year lagged current account balance D(CA(-1))] and differenced two-year lagged current account balance [D(CA(-2)) have a significant relationship with country risk. Current account balance in one-year lag and in two-year lag are statistically significant at the 5% significance level because their p-value is lower than the 5%. In conclusion, country risk increases by 10% when one-year lagged current account balance increases by 100%, ceteris paribus. Furthermore, when two-year lagged current account balance increases by 100%, country risk increases by 8%. This finding is in line with apriori conditions and Cline (1984) who argues that current account surplus is inversely related to the default risk whilst current account deficit is positively related to country risk and mostly equates to the amount of new financing required by a country.

It can also be observed that differenced one-year lagged political stability and absence of violence index [D(PSAV(-1))]and two-year lagged political stability and absence of violence index [D(PSAV(-2))] have a significant negative short run relationship with country risk. Political stability and absence of violence index in one-year lag and two-year lag are statistically significant at 5% level of significance since their p-values are less than 0.05. In conclusion, country risk decreases by 2.15% (-2.15318) when political stability and absence of violence index in one-year lag increases by 1%, ceteris paribus. In addition, when political stability and absence of violence index in two-year lag increases by 1% country risk decreases by 1.82% (-1.82465), ceteris paribus. The short run p-values indicate that unemployment rate in two-year lag [D(UN(-2))]  has a significant  influence on country risk. Two-year lagged unemployment rate is statistically significant at 5% level of significance since its p-value is less than 5%. In conclusion, country risk decreases by 30% (-0.30197) when unemployment rate increases by 100%, ceteris paribus. This finding contrasts the apriori conditions.  It can also be observed that weighted short term interest rates in one-year lag [D(WSTIR(-1))] have a significant short run relationship with country risk. Weighted short term interest rates in one-year lag are statistically significant since its p-value is less than 5%. In conclusion, country risk decreases by 18.84% (0.18847) when weighted short term interest rates rise by 100%, ceteris paribus.  This is in line with Andrade and Teles (2004) who argue that short term interest rates are inversely related to country risk.

5.10 Residual diagnostic Tests of the Error Correction Model

The Error Correction Model (ECM) was tested for serial autocorrelation and heteroscedasticity by conducting the Breusch-Godfrey Serial correlation LM test and Breusch-Pagan-Godfrey test, respectively. The results are shown in Table 9 below.

Table 9: Summary of Serial Correlation and Heteroscedasticity test

Residual diagnostics Type of test F-statistic P-value
Serial Autocorrelation Breusch-Godfrey Serial correlation LM test 0.4778 0.6514
Heteroscedasticity Breusch-Pagan-Godfrey test 0.2055 0.9940

Source: Researcher’s own compilation from EViews 10

Since p-value is greater than 0.05 for the serial autocorrelation tests, we fail to reject the null hypothesis and conclude that the model does not have serial correlation. For the heteroscedasticity test, p-value is greater than 0.05, we fail to reject the null hypothesis and conclude that the model is homoscedastic.

5.11 Stability diagnostic Tests

The results of CUSUM and CUSUM square test are shown in Figure 4 and Figure 5 below.

In Figures above, CUSUM and CUSUM squares lie within the 5% boundary, implying that the error correction model is stable and reliable to determine country risk for Zambia.

5.12 Model Specification Test

The Ramsey RESET test to check specification errors was done. A correctly specified model will generate an adequate picture of the relationship between country risk and its drivers. The Ramsey test results are shown in Table 10 below:

5.13 Interviewees’ views on determinants of country risk

The interviewees perceived that GDP per capita, GDP deflator, external debt balance, current account balance, weighted short term interest rates, unemployment rate and political risk influence country risk. All of the interviewees point that external debt followed by current account balance, and then political risk are the major divers of country risk in Zambia. They further point to the heavy dependence on external loans, signalling an increase in default risk and country risk increase. This is in line with Ofstad and Tjønneland (2019) who argue that Zambia has borrowed heavily from China since 2012 and is now at risk falling into the debt trap. The authors further state that Zambia once benefited from the Heavily Indebted Poor Country debt relief in 2005.They also point to the persistent current deficits caused by overreliance on copper production. Fluctuation in the copper price on the international market worsens the terms of trade by negatively affecting the export receipts. This leads to an increase in Zambia’s country risk. They also mentioned that unemployment rate is extremely high leading to an increase in cases of social unrest and political risk.

5.14 Interviewees’ views on expected relationship country risk and its drivers

The results of the views of interviewees on the expected relationship between country risk and independent economic, financial and political variables indicates that the interviewees’ responses were in line with a priori conditions. They perceived that all the mentioned explanatory factors [per capita GDP(-); GDP deflator(+); external debt(+); current account balance(+/-); weighted short term interest rates(+/-); political risk(+); unemployment(+)] had the expected sign.

5.15 Interviewees’ views on whether country risk is diversifiable or not?
All of the interviewees perceived that country risk form a systematic risk (non-diversifiable) and is beyond the control of private investors. They suggest that, even if they diversify their portfolios in different countries, the contagion effect associated with this risk may adversely affect the return on their investments. This coincides with empirical literature that country risk is systematic in nature and cannot be diversified in the country’s financial portfolio (Erb, Harvey and Viskanta, 1997; Naumoski, 2011; Gangemi, Brooks and Faff, 2000; Damodaran, 2003; Esch, Keiffer and Lopez, 2005). Interviewees also point out that fiscal and monetary policy makers should implement effective and efficient policies to manage financial, economic and political variables that drive country risk.

6. Conclusion

The study concluded that the short run drivers for the country risk of Zambia are beta, unemployment rate, political risk, weighted short term interest rates, and current account balance, even though current account is not one-to-one responsive to country risk. These findings converge with Ferreira (2010) who found that current account as a percentage of GDP largely influences country risk. The study concluded that the long run determinants for the country risk of Zambia are annual betas, political stability and absence of violence, unemployment and current account balance. The finding that interest rates are not statistically significant concurs with Gangemi, Brooks and Faff (2000) and Verma and Soydemir (2006) who established that interest rates had a trivial influence on Australian and Latin American country risk.

The study concluded that if there is a shock in the short term variables, the whole economy of Zambia adjusts with a speed of 140.83% to reach its equilibrium in the long run. Its error correction process fluctuates around the long-run value in a dampening manner (oscillatory adjustment process). However, once this process is complete, convergence to the equilibrium path is rapid.

These results are critical to different stakeholders in managing country risk. The key to country risk management is to critically assess its drivers and then the government must implement the policies necessary to manage these determinants. Based on the conclusion above, the Zambian authorities need to implement policies necessary to reduce persistent current account deficits, political risk, unemployment rate and external debt. Potential and existing investors have to engage the services of both private and public political risk insurers like International Finance Corporation (IFC) and Multilateral Investment Guarantee Agency (MIGA), which provide them with cover against expropriation, currency blockage, breach of contract, sequestration, and confiscation. Creditors (especially exporters) should rely on export cover and insurance guarantees; for example, most Organisation for Economic Co-operation and Development (OECD) countries have established official export credit agencies (ECAs) to enhance exports and foreign investment while managing country risk.

 


 

Simon Muwando ([email protected]) is  both a PhD candidate with University of Lusaka in Zambia and Lecturer at the National University of Science and Technology in Zimbabwe. His PhD supervisors, Professor Victor Gumbo ([email protected]) and Professor Gelson Tembo ([email protected]) are senior lecturers at the University of Botswana and the University of Zambia, respectively.  

 


 

References

Africa Development Bank (AfDB), Organisation for Economic Co-operation Development (OECD) and United Nations Development Programme (UNDP), 2015. Africa economic outlook: Zambia

African Economic Outlook (AEO) (2019). Zambia Economic Outlook, African Development Bank Group.

Almahmoud, A.I. (2014). Country Risk Ratings and Stock Market Movements: Evidence from emerging economy.International Journal of Economics and Finance, 6(10), 88-96.

Andrade, J. P and Teles, V. K. (2004). An Empirical Model of the Brazilian county risk – An extension of the Beta Country Risk Model. Applied Economics, 38(11), 1271-1278.

Banda, A. (2013). The effect of foreign direct investment on the host country economic growth: Zambia’s economic outlook after trade liberalization, Analytical paper submitted to TRAPCA in partial fulfilment of the requirement for the award of certificate in international trade and development.

Bank of Zambia (2013) retrieved from www.boz.co.zm.

Bacon-Shone, J. H. (2013). Introduction to Quantitative Research Methods. Hong Kong: Graduate School, The University of Hong Kong.

Ben Jebli, M. (2016). On the causal links between health indicator, output, combustible renewables and waste consumption, rail transport, and CO2 emissions: the case of Tunisia. Environmental Science and Pollution Research, 23(16), 16699-16715.

Basu, S., Deepthi, D. and Reddy, J. (2011). Country Risk Analysis in Emerging Markets: The Indian Example, Working Paper No. 326, Indian Institute of Management Bangalore.

Blommestein, H. J., Guzzo,V and Holland, A. (2010). Debt Markets in the Post-Crisis Landscape, OECD Journal: Financial Market Trends, 1, 1 -27.

Blommestein, H. J. and Turner, P. (2012). The Debate over sovereign risk, safe assets, and the risk-free rate: What are the implications for sovereign issuers? Ekonomi-tek Volume / Cilt, 1(3), 55-70.

Bracker, K. and Koch, P. D. (1999). Economic determinants of the correlation structure across international equity markets. Journal of Economics and Business, 51, 443– 471.

Central Statistical Office (CSO) Zambia (2012). Statistical Data

Central Statistical Office (CSO) Zambia (2018). Statistical Data

Choong, C., Yusop, Z., Law, S., and Liew, V. K. (2005). Financial development and economic growth in Malaysia: The perspective of stock market. Journal of Investment Management and Financial Innovations, 4, 105-115.

Cline, W. (1984). International Debt: Systemic Risk and Policy Response. Institute of International Economics: Washington, DC.

Damodaran, A. (2003). Country risk and company exposure: theory and practice.  Journal of Applied Finance, 64-76.

Development Policy Research Unit (2000). What are the major trends and determinants of foreign direct investment in SADC countries? DPRU Policy Brief No. 00/P2 July 2000. Industrial Project Strategy. Regional Integration Economic Cooperation in Southern Africa, University of Cape Town.

Erb, C. B., Harvey, C and Viskanta, T. (1996a). Political risk, economic risk and financial risk. Financial Analysts Journal, 52, 28-46.

Erb, C. B., Harvey, C and Viskanta, T. E. (1996b). Expected returns and volatility in 135 countries. Journal of Portfolio Management Spring, 46-58.

Esch, L. Keiffer, Rand Lopez, T. (2005). Asset and Risk Management: Risk Oriented Finance. John Wiley & Sons Ltd, United Kingdom.

Ferreira, A. L.  (2010). The determinants of default risk in Brazil. Applied Economics Letters 17(17), 1703-1708.

GAN Business Anti-corruption report for Zambia, 2017.

Gangemi, M.A.M., Brooks, R.D & Faff, R.W.  (2000), Modeling Australia’s country risk: A country beta approach.  Journal of Economics and Business, 52, 259-276.

Goldberg, C.S. and  Veitch, J.M. (2002). Country Risk and the Country Contagion Effect. Journal of Management Research, 2(1), 13-21.

Goldberg, C.S. and  Veitch, J.M. (2010). Country risk and financial integration – A case study of South Africa. Research in International Business and Finance, Elsevier, 24(2), 138-145.

Groenewold, N. and Fraser, P. (1997). Share prices and macroeconomic factors. Journal of Business Finance and Accounting, 24, 1367-1383.

Gujarat,D. (2012). Econometrics by Example. Palgrave Macmillan, UK

Harvey, C.R. (1995a). The risk exposure of emerging equity markets. The World Bank Economic Review, 9, 19-50.

Harvey, C.R. (1995b). Predictable risk and return in emerging markets. Review of Financial Studies, 8, 773-816.

Hoti,S.(2005). Modelling country spillover effects in country risk ratings. Elsevier Emerging Markets Review, 6, 324-345.

International Monetary Fund. (2012), Staff  Report for the 2012 Article IV Consultation-Debt Sustainability  Analysis Zambia, Supplement of IMF Country Report No. 12/200.

International Monetary Fund. (2019), Article IV Consultation-Press release; Staff report; and Statement by the Executive Director for Zambia, IMF Country Report No. 19/263.

Jeon, B.M. (2001). Macroeconomic influences on country risk. Working paper, Rice University.

Kumar, S. (2010). Determinants of Real Exchange Rate in India: An ARDL Approach. Reserve Bank of India Occasional Papers, 30(1), 33-64.

Maravi (2007). Political stability and economic growth in Zambia accessed on 13/08/15 on http://maravi.blogspot.com/2007/03/political-stability-and-economic-growth.html.

Mbao, M. L. M. (2011). Prevention and combating of corruption in Zambia. The Comparative and International Law Journal of Southern Africa, 44(2), 255-274.

Muwando. S. and Gumbo,V. (2013). A Country Beta Model for Zimbabwe: The Pioneer Journal of Theoretical and Applied Statistics, 6 (1), 25-41.

Naumoski, A. (2012). Estimating country risk premium in emerging markets: The case of Republic of Macedonia.Financial Theory and Practice, 36 (4), 413-434.

Nieh, C.C.  and  Wang, Y.S. (2005). ARDL Approach to the Exchange Rate Overshooting in Taiwan. Review of Quantitative Finance and Accounting, 25(1), 55–71.

Nkoro, E. and Uko, A.K. (2016). Autoregressive Distributed Lag (ARDL) cointegration technique: application and interpretation, Journal of Statistical and Econometric Methods, 5(4), 63-91.

Ofstad, A and Tjønneland, E. (2019). Zambia’s looming debt crisis- is China to blame? Chr. Michelsen Institute Insight number 1.

Oetzel. J, Bettis,M and  Richards A. (2001). Country risk measures: How risky are they? Journal of World Business, 36, 128-145.

Oteng-Abayie E. F and Frimpong J.M. (2006). Bounds Testing Approach to Cointegration: An Examination of Foreign Direct Investment Trade and Growth Relationships. American Journal of Applied Sciences, 3 (11): 2079-2085.

Page,S and Velde, D.W. (2004). Foreign Direct Investment by African Countries, Draft,  Papers prepared for InWent / UNCTAD meeting on FDI in Africa 22-24 November 2004, UNECA, Addis Ababa.

Pesaran, M. H., and Pesaran, B. (2009). Time Series Econometrics Using Microfit 5.0. New York: Oxford University Press.

Pesaran, M. H., Shin, Y. and  Smith, R. J. (2001). Bounds Testing Approaches to the Analysis of Level Relationships. Journal of Applied Econometrics, 16(3), 289–326.

Pesaran, M. H., and Smith, R. J. (1995). Estimating long run relationship from dynamic heterogeneous panels. Journal of Econometrics, 68, 79-113.

San-Martín-Albizuri, N. and Rodríguez-Castellanos, A. (2018). Measuring Country Risk: A Topic of interest, Chapter 5, 81-98.

Saungweme,T and Odhiambo, N.M. (2019). Government debt, government debt service and economic growth nexus in Zambia: a multivariate analysis. Cogent Economics  and  Finance , (7)1, 1-17.

Shapiro, A. (1999). Multinational Financial Management, 6th edition, Prentice Hall: London, UK.

Tourani-Rad, A., Choi, D.F.S and Wilson, B.J. (2006). A time-varying beta approach to measuring New Zealand’s country risk. Journal of managerial finance, 32 (3), 257-269.

United Nations Development Programme (UNDP). (2013). The Rise of the South: Human Progress in a Diverse World. Explanatory note on 2013, Human Development Report composite indices Zambia.

United Nations Development Programme (UNDP) (2015b). Human development Report 2015: Technical note.

United Nations Development Programme (UNDP). (2014). Human development Report for Zambia, 2015.

Verma. R and .Soydemir, G. (2006). Modelling country risk in Latin America: A country beta approach. Global Finance Journal, 17, 192-213.

Verma, R and  Verma P. (2014). Country Risk and Macroeconomic Factors: Evidence from Asian Markets. Journal of Applied Business and Economics, 16(5), 51-62.

Vij, M. (2005). The determinants of country risk analysis: An empirical approach. University of Delhi, 5 (1), 20-31.

Wdowinski, P. (2004). Determinants of Country Beta Risk in Poland, University of Lodz: CESifo (Center for Economic Studies and Ifo Institute for Economic Research). CESifo Working Paper Series No. 1120.

World Bank (2019). The World Bank in Zambia. Last Updated: Oct 13, 2019.

World Bank (2019). World Development indicators. Washington, DC – USA: World Bank Databank

World Bank (2013). Zambia at a Glance.

World Bank (2011). Zambian FDI inflows from 2001 to 2009.

World Organization against Torture (OCMT), Geneva (2001). Zambia: Violations of Economic, Social and Cultural Rights, Violence and the Protection against Torture. Report on the Implementation of the Convention Against Torture by Zambia.

Yezi consulting and associates (2013). Political governance study in Zambia. A study commissioned by Diakonia, Zambia.