Comparative analysis of risk ratings for the East European region

Mathematics and Computers in Simulation 68 (2005) 449–462 Comparative analysis of risk ratings for the East European region Suhejla Hoti School of Economics and Commerce, University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia Available online 3 March 2005 Abstract Following the aftermath of the 11 September 2001 events, the risks associated with engaging in international dealings have increased substantially, and become more difficult to analyse and predict for decision makers in the economic, financial and political sectors. The importance of country risk analysis is underscored by the existence of several prominent country risk rating agencies, which combine a wide range of qualitative and quantitative information regarding alternative measures of economic, financial and political risk into associated composite risk ratings. However, the accuracy of any rating agency with regard to any or all of these measures is open to question. For this reason, the paper provides a qualitative comparison of country risk rating systems used by seven leading rating agencies. The paper also provides a novel analysis of four risk ratings using univariate and multivariate volatility models for nine East European countries. These ratings are compiled by the International Country Risk Guide, which is the only risk rating agency to provide consistent monthly data for a large number of countries since 1984. The empirical results enable a comparative assessment of the conditional means and volatilities associated with county risk returns, defined as the rate of change in country risk ratings, across the nine East European countries. Moreover, the estimated constant conditional correlation coefficients provide useful information as to whether the countries are similar in terms of shocks to the four risk returns. © 2005 IMACS. Published by Elsevier B.V. All rights reserved. Keywords: Country risk; Rating agencies; Rating systems; ICRG; Economic risk; Financial risk; Political risk; Composite risk; Volatility; Conditional correlation; Shocks; GARCH 1. Introduction Country risk has become a topic of major concern for the international community in the last two decades. The debt crises of the early 1980s, political changes that occurred in the former Communist E-mail address: Suhejla.Hoti@uwa.edu.au. 0378-4754/$30.00 © 2005 IMACS. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.matcom.2005.02.014 450 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 Block countries in the late 1980s and early 1990s, the East Asian and East European financial crises that have occurred since 1997, the recent financial and banking crises in Argentina, and finally the tumultuous events flowing from 11 September 2001, clearly show that the risks associated with engaging in international relationships have increased substantially. Such events have also become more difficult to analyse and predict for decision makers in the economic, financial and political sectors. The increasing importance of country risk analysis by both official and private institutions is due to the fact that the globalisation of world trade and open capital markets are risky elements that can cause financial crises with rapid contagion effects, which threaten the stability of the international financial sector [6]. Furthermore, the increasing number of financial crises in developing countries, and their associated costs to the official institutions and private entities, are also major factors of risk that need to be considered. Given these developments, the activities of country risk rating agencies have increased substantially over the last two decades. Rating agencies compile country risk ratings as measures of the ability and willingness of countries to service their financial obligations. This is of particular importance for developing countries, for which there is limited information available. Country risk ratings help developing countries to gain access to capital markets and provide economic, financial and political operators with essential tools to assess and manage such risk. Consequently, the accuracy of risk rating agencies with regard to any or all country risk measures is crucial. The plan of the paper is as follows: Section 2 presents the nature of the country risk rating industry. Section 3 discusses the empirical findings of 50 published studies on country risk reviewed in [7]. The country risk rating systems of seven leading commercial analysts of country risk, namely Institutional Investor, Euromoney, Moody’s, Standard and Poor’s, International Country Risk Guide, and Political Risk Services, are assessed and compared in Section 4. Section 5 presents a novel analysis of country risk ratings, compiled by the ICRG for the East European region, using univariate and multivariate models of conditional volatility. Some concluding remarks are given in Section 6. 2. Country risk ratings The importance of country risk analysis is underscored by the existence of several prominent country risk rating agencies, namely the Economist Intelligence Unit, Euromoney, Institutional Investor, International Country Risk Guide, Moody’s, Political Risk Services, and Standard and Poor’s. These risk rating agencies employ different methodologies and methods to determine country risk ratings, combining a wide range of qualitative and quantitative information regarding alternative measures of economic, financial and political risk into associated composite risk ratings. A primary function of country risk assessment is to anticipate the possibility of debt repudiation, default or delays in payment by sovereign borrowers [4]. However, the accuracy of any risk rating agency with regard to any or all country risk measures is open to question. For purposes of evaluating the importance and relevance of agency country risk ratings, it is necessary to analyse such agency rating systems according to established rating criteria. The primary purpose of each of these rating agencies is to measure the risk associated with investing in a foreign country. This paper provides a qualitative comparison of country risk rating systems used by Economist Intelligence Unit (EIU), Euromoney, Institutional Investor (II), International Country Risk Guide (ICRG), S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 451 Moody’s, the Political Risk Services (PRS), and Standard and Poor’s (S&P’s). A classification of the seven risk rating agencies is given according to the agency definition of country risk ratings, number of countries covered, frequency of the risk ratings, number and type of ratings compiled, number and type of risk component variables used, weights assigned to risk components, and the given range for the risk ratings. 3. Country risk literature For purposes of evaluating the accuracy of agency risk rating systems, it is also necessary to review the literature relating to empirical country risk models. A review of 50 empirical studies on country risk published over the last two decades was given in [7]. The studies were analysed according to established statistical and econometric criteria used in estimation, evaluation, and forecasting in order to evaluate the practicality and relevance of the economic, financial and political theories pertaining to country risk. Table 1 classifies the 50 studies according to the type of country risk variable used. Of the 50 studies, 27 examined debt re-scheduling on 36 occasions, 17 considered country risk ratings on 18 occasions, and 6 considered other types of dependent variables (see [7] for the definitions of these three types of variables). Although debt rescheduling is the most frequently used dependent variable in the country risk rating literature, this paper focuses on the agency country risk ratings, which is the second most frequently used variable in the literature and closely related with debt re-scheduling (for further details, see [8]). As country risk rating is primarily a measure of country creditworthiness, the lower is the creditworthiness of a country, the higher is the associated risk in investing in that country, and the higher is the probability that the country will re-schedule its future debt payments. Table 2 classifies the rating agencies used in the 17 country risk studies according to their frequency. Institutional Investor country risk ratings are the most frequently used ratings, and were used 13 times in total. Euromoney country risk ratings, which were used 6 times, are the second most frequently used ratings. Moody’s, S&P’s, and ICRG country risk ratings were each used twice, followed by EIU and PRS, each being used once. Table 3 reports four types of risk component variables used in the 17 country risk ratings studies, namely economic, financial, political, and composite. Composite risk variables are ratings or aggregates that comprise economic, financial and political risk component variables, and were used in all 17 studies. Of these studies, only two did not use economic variables and only one did not use financial variables. Political variables have been used less frequently, namely in 10 studies. Table 1 Type of dependent variable used Type Frequency Debt rescheduling Agency country risk ratings Others Total 27 17 6 50 452 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 Table 2 Agency ratings used Agency Frequency II Euromoney Moody’s S&P’s ICRG EIU PRS 13 6 2 2 2 1 1 Some studies used data from more than one agency. Table 3 Risk component variables used in country risk ratings Variables Frequency Economic Financial Political Composite Number of studies 15 16 10 17 17 Table 4 Frequency of risk component variables used in country risk ratings Risk components used Frequency 4 3 2 1 Total 10 4 3 0 17 Table 4 presents the number of country risk components used, as well as their frequency. All four country risk components have been used in 10 studies, 4 studies used variables representing three risk components, 3 studies used variables representing two risk components, and no study used variables representing only one risk component. 4. Comparison of country risk rating methodologies Country risk refers broadly to the ability and willingness of a country to repay its financial obligations to its foreign creditors. While the individual agencies use different definitions of country risk ratings, they all fall into this broad category. Institutional Investor and Euromoney define their country risk ratings as measures of the creditworthiness of a country as a whole. These ratings measure the economic, financial, and political performances of countries. Moody’s country risk rating is defined as a measure of the ability and willingness of a country’s S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 453 Table 5 Risk ratings debut year Agency Year Moody’s S&P’s II PRS Euromoney ICRG EIU 1914 1941 1979 1979 1983 1984 NA “NA” denotes not available. central bank to provide foreign currency to service the foreign debt held by the government and other borrowers residing in that country. This rating is not a direct evaluation of the creditworthiness of the government, but rather an assessment of the foreign liabilities of the country as a whole. Unlike Moody’s, S&P’s defines its country risk rating as a measure of a government’s ability and willingness to repay debt according to its terms. Standard and Poor’s ratings are sovereign ratings as they address the credit risk of the government and not of the other borrowers of a country [9]. The ICRG country risk rating is defined as the ability and willingness of a country to finance its official, commercial, and trade debt obligations. It measures the economic, financial and political structures of a country as a whole. Similarly, the EIU defines its country risk rating as a measure of the likelihood of a financial crisis in a country that would affect foreign investors in that country. The Economist Intelligence Unit ratings also provide a measure of the general risk associated with investing in a country. Finally, PRS defines its rating as a measure of the likely changes in the level of political turmoil and government intervention that affect the business climate. These ratings are known as forecast ratings. Table 5 classifies the seven rating agencies according to the year they started to compile country risk ratings. Clearly, Moody’s and S&P’s are the oldest agencies in the risk rating industry. Moody’s issued its first country risk ratings just before WWI. Standard and Poor’s was formed after Poor’s Publishing and Standard Statistics merged in 1941, and continued to compile risk ratings for several sovereign bond issues [2]. The remaining 5 agencies started publishing risk ratings almost four decades later, around the onset of the third World Debt Crisis. In 1979, II and PRS published their first risk ratings, followed by Euromoney in 1983 and ICRG in 1984. While EIU was founded in 1946, information about the debut year for the EIU risk ratings is not available. In Table 6, the seven rating agencies are classified according to the number of countries rated, as of April 2004, except for Moody’s and S&P’s, for which the information is available to July 2002. The number of rated countries ranges from 93 to 185. Of the seven rating agencies, Euromoney’s coverage is the largest, compiling ratings for 185 countries. The ICRG covers the second largest group of countries with 140, while II provides ratings for more than 135 countries [9]. As of July 2002, Moody’s and S&P’s have been providing ratings for 109 and 93 countries, respectively [2], with S&P’s having the smallest country coverage in the group. Virtually every one of the countries covered by Moody’s participates in the world’s capital markets. Both EIU and PRS provide ratings for 100 countries. The EIU covers key emerging and highly indebted countries that are monitored by the its Country Risk Service (CRS). 454 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 Table 6 Classification by the number of countries covered Agency Number of countries Euromoney ICRG II Moody’s EIU PRS S&P’s 185 140 135 109 100 100 93 Table 7 Classification by frequency of ratings Agency Frequency of ratings ICRG EIU PRS II Euromoney Moody’s S&P’s Monthly Quarterlya Quarterly Semi-annual Semi-annual Annuala Annualb a b With monthly ratings updates. With weekly ratings updates. Published country risk ratings are made available on a monthly, quarterly, semi-annual, and/or annual basis. Table 7 classifies the seven rating agencies according to the frequency of their ratings. Of the seven rating agencies, ICRG is the only agency to provide consistent country risk ratings on a monthly basis. The Economist Intelligence Unit publishes quarterly risk ratings with monthly updates on these ratings. Political Risk Services provides quarterly ratings with no updates, while II and Euromoney publish their ratings semi-annually in the March and September issues of these monthly magazines. Moody’s and S&P’s provide annual credit reports with monthly and weekly ratings updates, respectively. Table 8 classifies the seven rating agencies according to the number of risk ratings or risk rating categories they compile. The number of agency compiled ratings ranges from 1 to 10. It should be emphasised Table 8 Classification by number of ratings compiled Agency Number of ratings Moody’s Euromoney S&P’s ICRG EIU PRS II Total 10 10 7 4 4 3 1 39 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 455 that while Moody’s and S&P’s compile ratings for both the issuer and specific debt instruments, the other five agencies compile ratings only for the issuer. Of the 39 risk ratings, more than half are compiled by Moody’s and Euromoney, with the remaining 19 ratings being compiled by S&P’s, ICRG, EIU, PRS, and II. For each nation, Moody’s publishes ratings in ten major areas, namely long-term (bonds and preferred stock), issuer, bank deposits, bank financial strength, national scale, managed fund, real estate fund, prime rating, and speculative grade liquidity. Moody’s country risk ratings act as sovereign ceilings or caps on ratings of foreign currency securities of any other borrowing entity. The ratings account for foreign currency transfer risk and systemic risk in the nation. Like Moody’s, Euromoney offers ten ratings for each country it covers. These ratings include one composite country risk rating and nine component risk ratings, namely political risk, economic performance, debt indicators, debt in default or re-scheduled, credit ratings, access to bank finance, access to short-term finance, access to capital markets, and discount on forfeiting. Standard and Poor’s ratings are provided for seven major areas, namely long-term debt, commercial paper, preferred stock, certificates of deposit, money market funds, mutual bond funds, and the claims-paying ability of insurance companies. Such ratings set the benchmark for the ratings assigned to other issuers in the country. International Country Risk Guide and EIU compile four types of risk ratings each. International Country Risk Guide ratings include one composite country risk rating and three component risk ratings, namely economic, financial and political. The Economist Intelligence Unit compiles one country risk rating and three specific investment ratings, namely currency risk (associated with accepting foreign exchange exposure against the US dollar), sovereign debt risk (associated with foreign currency loans to sovereign states), and banking sector risk (associated with foreign currency loans to banks). Political Risk Services offers three industry forecasts at the micro level, namely financial transfers (banking and lending), foreign direct investment (such as retail, manufacturing, and mining), and exports to the host country market. Finally, II offers only one risk rating, which is the country risk rating. Table 9 classifies the seven rating agencies according to the total number of risk component variables. The total number of risk component variables used in the rating systems of the seven agencies ranges from 9 (for II) to 76 (for EIU). Euromoney, ICRG and PRS use at least 20 component variables to compile their ratings, while Moody’s and S&P’s use at least 10 component variables. In terms of the individual risk component variables, the number of the economic risk variables used by each agency varies from 2 (for Euromoney) to 55 (for EIU). EIU uses the largest number of the economic risk variables, followed Table 9 Number of risk component variables used Agency ECO FIN POL Others Total EIU Euromoney ICRG PRS Moody’s S&P’s II 55 2 5 13 7 3 5 10 10 5 2 0 1 3 11 11 12 5 6 6 1 0 3 0 0 0 0 0 76 26 22 20 13 10 9 Economic, financial and political risk ratings are denoted as ECO, FIN, and POL, respectively. “Others” category refers to agency ratings. 456 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 Table 10 Type of risk component variables used Variables Frequency Economic Political Financial Others Total 90 52 31 3 176 “Others” category refers to agency risk ratings. by PRS (13 variables), Moody’s (7 variables), ICRG and II (5 variables, each), S&P’s (3 variables), and Euromoney (2). In the case of financial risk variables, the total number of variables used by each agency ranges from 0 (for Moody’s) to 10 (for EIU and Euromoney). Of the remaining four agencies, ICRG uses 5 financial variables in total, followed by II, PRS, and S&P’s, which use 3, 2, and 1 financial risk variables, respectively. The number of political risk variables used by each agency ranges from 1 (for II) to 12 (for ICRG). Of the remaining five agencies, EIU and Euromoney both use a total of 11 political variables, and are followed by Moody’s and S&P’s (with 6 variables each) and PRS (5 variables). Finally, regarding the “Others” category, which refers to agency risk ratings, Euromoney is the only agency which considers risk ratings compiled by 3 other agencies, namely Moody’s, S&P’s, and Fitch IBCA. The classification in Table 10 is given according to the type of risk component variables being used in the rating systems of the seven rating agencies. More than half of the risk component variables used by the seven risk rating agencies are predominantly economic in nature, with the remainder being political or financial in nature. Political variables are the second most frequently used risk components. The “Others” category refers to agency risk ratings, being used only in the case of the Euromoney risk rating system. In terms of the rating system used to compile composite country risk ratings, EIU, Euromoney, and ICRG differ from Moody’s, S&P’s, PRS and II, in that they calculate composite ratings using specific formulae, with predetermined weights assigned to each of the risk components. Table 11 reports the risk component weights in the rating systems of EIU, Euromoney, and ICRG. Referring to Table 11, the economic risk variables have the highest weight at 55% in the case of EIU, followed by Euromoney and ICRG, each assigning a weight of 25%. For financial risk component variables, Euromoney assigns the highest weight at 40%, followed by ICRG and EIU, assigning weights of 25 and 23%, respectively. For the political risk component variables, ICRG assigns the highest weight at 50%, followed by Euromoney and EIU with weights of 25% and 22%, respectively. Finally, in order to obtain the overall country risk score, Euromoney assigns a weight of 10% to agency ratings component. Table 11 Weights assigned to risk component variables (in percent) Agency ECO FIN POL Others Total EIU Euromoney ICRG 55 25 25 23 40 25 22 25 50 0 10 0 100 100 100 Economic, financial and political risk ratings are denoted as ECO, FIN, and POL, respectively. “Others” category refers to agency ratings. S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 457 Table 12 Types of risk rating grades Agency Grading range II Euromoney ICRG Moody’s S&P’s EIU PRS 0–100 1–100 0–100 AAA to C AAA to D A to E A+ to D− With respect to the rating systems of Moody’s, S&P’s, PRS, and II, composite risk ratings are determined on a subjective basis. For each country, Moody’s analysts weigh the risk component variables according to their assessment of the credit risk of the issuer. Similarly, in determining a risk rating, S&P’s analysts weight the risk component variables based on their perceptions of economic and fundamental business conditions for each country. PRS system compiles forecast ratings based on the component risk variables, and weighs them according to the assessed potential economic, financial and political risks to business investments and trade [9]. Institutional Investor differs from the other six agencies in that it uses no internal rating system. For each country, II asks 75–100 leading international banks to rate the risk components. The individual ratings are weighted using II’s formula, with greater weights assigned to responses based on the extent of a bank’s worldwide exposure and the degree of sophistication of a bank’s country risk model. The names of the participating banks are kept strictly confidential [9]. In the country risk literature, the II country risk assessment is known as the banker’s judgment. Finally, in Table 12 the seven rating agencies are classified according to the type and range of gradings they assign to country risk ratings. Institutional Investor, Euromoney, and ICRG provide quantitative country risk ratings, which range from 0 (lowest) to 100 (highest). On the other hand, Moody’s, S&P’s, EIU, and PRS publish qualitative letter ratings. The country risk ratings for Moody’s, S&P’s, EIU, and PRS range from AAA (highest) to C (lowest), AAA (highest) to C (lowest), A (highest) to E (lowest), and A+ (highest) to D− (lowest), respectively. In all cases, the lower (higher) is a given risk rating, the higher (lower) is the associated risk. 5. ICRG risk ratings analysis for nine East European countries 5.1. Univariate and multivariate models of conditional volatility for risk returns Monthly data can capture the time-varying volatility that is inherent in the underlying series. As risk ratings can be treated as indexes, their rates of change, or risk returns, can be analysed in the same manner as financial returns. The structure and properties of the Constant Conditional Correlation (CCC) Multivariate GARCH model of Bollerslev [3] will be discussed briefly in this section. 458 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 Consider the following specification: yt = E(yt |Ft−1 ) + εt , εt = Dt ηt , (1) where yt = (y1t ,. . .,ymt )′ , ηt = (η1t ,. . .,ηmt )′ is a sequence of independently and identically distributed (iid) 1/2 1/2 random vectors, Ft the past information available to time t, Dt = diag(h1t , ..., hm1t ), m (=4) is the number of country risk returns, and t = 1,. . .,221 monthly observations for the period 1984(1) to 2002(5). The CCC model assumes that the conditional variance for a risk return i, hit , i = 1,. . .,m, follows a univariate GARCH process, that is, hit = ωi + r
αij ε2i,t−j + j=1 s
βij hi,t−j (2) j=1 where αij represents the ARCH effects, or the short-run persistence of shocks to risk return i, and βij r  represents the GARCH effects, or the contribution of shocks to long-run persistence, namely αij + j=1 s  βij . j=1 It is important to note that the conditional correlation for the CCC model is assumed to be constant. As Γ = E(ηt η′t |Ft−1 ) = E(ηt η′t ), the (constant) conditional correlation matrix of the unconditional shocks, εt , is equivalent to the (constant) conditional correlation matrix of the standardized shocks, ηt , where Γ = {ρij } for i, j = 1,. . .,m. When the number of risk returns is set to m = 1, such that a univariate model is specified rather than the multivariate model, Eqs. (1) and (2) become:  εt = ηt ht , ht = ω + r
j=1 αj ε2t−j + s
βj ht−j (3) j=1 and ω > 0, αj ≥ 0 for j = 1,. . .,r and βj ≥ 0 for j = 1,. . .,s are sufficient conditions to ensure that the conditional variance ht > 0. Using results from [11,12,15] and [13], the necessary and sufficient condition for the existence of the second moment of εt , that is E(ε2t ) < ∞, for the case r = s = 1 is α1 + β1 < 1. The parameters in Eqs. (1) and (3) are typically obtained by maximum likelihood estimation (MLE) using a joint normal density for the standardized shocks. When ηt does not follow a joint multivariate normal distribution, the parameters are estimated by Quasi-MLE (QMLE), which is less efficient than MLE. Ling and McAleer [14] showed that the QMLE for GARCH(r,s) is consistent if the second moment is finite, that is E(ε2t ) < ∞. Jeantheau [10] showed that the log-moment condition given by E(log(α1 η2t + β1 )) < 0 is sufficient for the QMLE to be consistent for GARCH(1,1), while [5] showed that the QMLE is asymptotically normal for GARCH(1,1) under the same condition. It should be noted that the logmoment condition is weaker than the second moment condition. However, the log-moment condition is more difficult to compute in practice as it is the expected value of a function of an unknown random variable and unknown parameters. S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 459 Table 13 Summary of AR(1)-GARCH(1,1) estimates Conditions Outcome α>0 β>0 Log-moment Second moment Number of equations 25 31 19 29 36 5.2. Empirical results The univariate AR(1)-GARCH(1,1) are used to provide estimates of the volatilities associated with the four risk returns for the nine East European countries. Ratings for Yugoslavia are available from January 1984 to May 2002, for Hungary and Romania from August 1984 to May 2002, for Bulgaria and Poland from December 1984 to May 2002, for Czech Republic and Slovak Republic from January 1993 to May 2002, for Albania from October 1985 to May 2002, and for Russia from April 1992 to May 2002. The univariate results enable a validation of the regularity conditions underlying the model, highlight the importance of economic, financial and political risk ratings as components of a composite risk rating, and evaluate the usefulness of the ICRG risk ratings. Constant conditional correlations between pairs of country risk returns are also estimated. This gives an indication of the relationship between shocks to the economic, financial, political and composite risk returns, as well as the direction of any causality in the four risk ratings across the nine east European countries. All the estimates are obtained using EViews 4 and the Berndt, Hall, Hall and Hausman (BHHH) [1] algorithm. Using the RATS 6 econometric software package gave virtually identical results. The AR(1)-GARCH(1,1) estimates for East Europe are summarised in Table 13, with the α and β estimates being positive fractions in 25 and 31 cases, respectively. The second moment condition is satisfied in 29 of 36 cases for the 9 countries and 4 risk returns, while the log-moment condition is satisfied 19 times. The consistency and asymptotic normality of the QMLE are not guaranteed for financial risk returns for Yugoslavia, as neither the second moment condition nor the log-moment condition is satisfied. Apart from this case, when the second moment condition is not satisfied, the log-moment condition ensures that the QMLE are consistent and asymptotically normal in the presence of infinite second moments. Similarly, with the exception of financial risk returns for Yugoslavia, the second moment condition is satisfied in all cases when the log-moment condition is either not satisfied or could not be computed. Based on the monthly standardized residuals of the univariate AR(1)-GARCH(1,1) model, the corresponding constant conditional correlations have been calculated for the economic, financial, political and composite risk return volatilities for the nine East European countries. The constant conditional correlation coefficients for the nine countries by four risk returns are reported in Table 14. For economic risk returns, 24 of the 36 conditional correlations are less than 0.200. Of the remaining one-third, 7 range from (0.200, 0.299), 3 from (0.300, 0.399), and 2 from (0.400, 0.499). The largest conditional correlation is for (Poland, Czech Republic), followed by (Poland, Hungary), (Russia, Czech Republic) and (Slovakia, Romania). Overall, while the majority of conditional correlations are close to 0, all 9 East European countries have correlations that exceed 0.200 with one or more countries. 460 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 Table 14 Constant conditional correlation coefficients for nine countries by risk returns Country Albania Economic risk return Albania 1.000 Bulgaria Czech Republic Hungary Poland Romania Russia Slovakia Yugoslavia Financial risk return Albania 1.000 Bulgaria Czech Republic Hungary Poland Romania Russia Slovakia Yugoslavia Political risk return Albania 1.000 Bulgaria Czech Republic Hungary Poland Romania Russia Slovakia Yugoslavia Composite risk return Albania 1.000 Bulgaria Czech Republic Hungary Poland Romania Russia Slovakia Yugoslavia Bulgaria Czech Republic Hungary Poland Romania Russia Slovakia Yugoslavia 0.312 1.000 0.146 0.011 1.000 0.212 0.031 0.141 1.000 0.221 0.093 0.475 0.406 1.000 0.081 0.126 0.212 −0.068 0.038 1.000 −0.021 −0.076 0.392 0.027 0.135 0.299 1.000 0.207 0.048 0.160 0.084 0.182 0.313 0.212 1.000 0.119 0.166 −0.195 −0.121 −0.218 0.196 −0.041 0.076 1.000 0.129 1.000 0.028 0.394 1.000 −0.059 0.414 0.737 1.000 0.153 0.254 0.623 0.519 1.000 0.101 0.484 0.529 0.425 0.448 1.000 −0.125 −0.314 −0.148 −0.048 −0.111 −0.391 1.000 −0.153 0.133 0.443 0.620 0.389 0.156 0.183 1.000 −0.037 0.129 0.307 0.177 0.073 0.209 −0.085 0.085 1.000 0.316 1.000 0.050 0.179 1.000 0.255 0.229 0.230 1.000 0.156 0.264 0.454 0.328 1.000 0.080 0.239 0.347 0.223 0.269 1.000 0.086 −0.004 0.117 0.092 0.042 −0.009 1.000 −0.106 0.050 0.250 0.137 0.360 0.075 −0.016 1.000 0.243 0.170 −0.070 0.230 0.061 0.093 0.155 −0.099 1.000 0.223 1.000 0.123 0.229 1.000 0.175 0.211 0.305 1.000 0.224 0.253 0.453 0.511 1.000 0.044 0.298 0.137 0.095 0.234 1.000 −0.069 −0.077 0.158 0.077 0.025 0.004 1.000 0.013 0.117 0.145 0.191 0.265 0.304 0.094 1.000 0.153 −0.027 −0.014 0.092 −0.011 −0.059 0.185 −0.058 1.000 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 461 The constant conditional correlations for financial risk returns can be very high. More than half of the 36 conditional correlations are less than 0.200. Of the remaining 17 conditional correlations, 2 range from (0.200, 0.299), 5 from (0.300, 0.399), 5 from (0.400, 0.499), 2 from (0.500, 0.599), 2 from (0.600, 0.699), and 1 from (0.700, 0.799). The highest conditional correlation is for (Hungary, Czech Republic), with the next two highest correlations being (Poland, Czech Republic) and (Slovakia, Hungary). Overall, 8 of the 9 countries have correlations that exceed 0.200 with two or more countries. Albania is the only country in the region with independent effects. For political risk returns, almost two thirds of the conditional correlations are less than 0.200. Of the remaining 15 conditional correlations, 10 range from (0.200, 0.299), 4 from (0.300, 0.399), and 1 from (0.400, 0.499). Thus, the conditional correlations are generally low. The highest correlation holds for (Poland, Czech Republic), followed by (Slovakia, Poland) and (Romania, Hungary). Moreover, Russia is the only independent country in the region. As for political risk returns, the conditional correlations for composite risk returns are generally low. Two-thirds of the 36 conditional correlations are less than 0.200. Of the remaining 12 conditional correlations, 8 range from (0.200, 0.299), 2 from (0.300, 0.399), 1 from (0.400, 0.499), and 1 from (0.500, 0.599). The two highest correlations are for (Poland, Hungary) and (Poland, Czech Republic). Moreover, Russia and Yugoslavia seem to be independent, as their conditional correlations with the remaining countries within the region are less than 0.200. Overall, the strongest conditional correlations and the largest range of variation are for financial risk returns, followed by composite risk returns. There is a higher range for the conditional correlations for economic risk returns than for their political returns counterparts. However, the number of conditional correlations that exceeds 0.200 is higher for political risk returns than for economic risk returns. Independent effects for various countries are observed for all risk returns, particularly for political and composite risk returns. Based on the estimated constant conditional correlations, Russia seems to be independent in shocks to political and composite risk returns, Yugoslavia for composite risk returns, and Albania for financial risk returns. 6. Concluding remarks The paper provided a qualitative comparison of the country risk rating systems of seven leading commercial analysts of country risk, namely Institutional Investor, Euromoney, Moody’s, Standard and Poor’s, International Country Risk Guide, and Political Risk Services. Such an evaluation permitted a critical assessment of the importance and relevance of agency rating systems. The paper also provided a novel analysis of four risk ratings using univariate and multivariate volatility models for nine East European countries. These ratings were compiled by the International Country Risk Guide, which is the only risk rating agency to provide consistent monthly data for a large number of countries since 1984. The empirical results enabled a comparative assessment of the conditional means and volatilities associated with county risk returns, defined as the rate of change in country risk ratings, across the nine East European countries. Moreover, the estimated constant conditional correlation coefficients provided useful information as to whether the countries are similar in terms of standardised shocks to the four risk returns. Overall, independent effects for various countries are observed for all risk returns, particularly for political and composite returns. Russia seems to be independent in the standardised shocks to political and composite risk returns, Yugoslavia for composite risk returns, and Albania for financial risk returns. 462 S. Hoti / Mathematics and Computers in Simulation 68 (2005) 449–462 Acknowledgements The author wishes to thank Michael McAleer and Felix Chan for helpful discussions and suggestions, and to acknowledge an Australian Research Council Research Fellowship. References [1] E.K. Berndt, B.H. Hall, R.E. Hall, J.A. Hausman, Estimation and inference in nonlinear structural models, Ann. Econ. Social Meas. 3 (1974) 653–665. [2] A.V. Bhatia, Sovereign credit ratings methodology: an evaluation, International Monetary Fund Working Paper, 02/170, 2002. [3] T. Bollerslev, Modelling the coherence in short-run nominal exchange rates: a multivariate generalized ARCH approach, Rev. Econ. Stat. 72 (1990) 498–505. [4] F.N. Burton, H. Inoue, An appraisal of the early-warning indicators of sovereign loan default in country risk evaluation systems, Manage. Int. Rev. 25 (1) (1985) 45–56. [5] F. Boussama, Asymptotic normality for the quasi-maximum likelihood estimator of a GARCH model, Comptes Rendus de l’Academie des Sciences 331 (Serie I) (2000) 81–84 (in French). [6] N.L. Hayes, Country risk revisited, J. Lending Credit Risk Manage. 80 (5) (1998) 61. [7] S. Hoti, Snapshot images of country risk ratings: an international comparison, in: A. Rizzoli, A.J. Jakeman (Eds.), Proceedings of the International Conference on Environmental Modelling and Software, vol. 2, Lugano, Switzerland, 2002, pp. 532–537. [8] S. Hoti, M. McAleer, An empirical assessment of country risk ratings and associated models, J. Econ. Surveys 18 (4) (2004) 539–588. [9] L.D. Howell, The Handbook of Country and Political Risk Analysis, 3rd ed., The PRS Group, New York, 2001. [10] T. Jeantheau, Strong consistency of estimators for multivariate ARCH models, Econ. Theory 14 (1998) 70–86. [11] S. Ling, W.K. Li, On fractionally integrated autoregressive moving-average models with conditional heteroskedasticity, J. Am. Stat. Assoc. 92 (1997) 1184–1194. [12] S. Ling, M. McAleer, Stationarity and the existence of moments of a family of GARCH processes, J. Econ. 106 (2002) 109–117. [13] S. Ling, M. McAleer, Necessary and sufficient moment conditions for the GARCH(r,s) and asymmetric power GARCH(r,s) models, Econ. Theory 18 (2002) 722–729. [14] S. Ling, M. McAleer, Asymptotic theory for a vector ARMA-GARCH model, Econ. Theory 19 (2003) 278–308. [15] D.B. Nelson, Stationarity and persistence in the GARCH(1,1) model, Econ. Theory 6 (1990) 318–334.