Do international banks' assessments of country risk follow a random walk? An emperical examination of the Middle East.

Alon, Ilan ; Qi, Min

ABSTRACT

The article empirically investigates the stochastic properties of a widely used indicator of country risk: Institutional Investor's creditworthiness ratings. It tests whether Institutional Investor's ratings of Middle Eastern countries follow a random walk by checking for unit root. It is important to test for unit root because estimated relationships between environmental variables and indicators of country risk may exhibit spurious relationships. Furthermore, if the variable contains a unit root the impact of changes in perceptions of creditworth may have a long-lasting effect rather than a temporary one. Our analysis reveals that country risk ratings for some countries in the Middle East follow a random walk, even after adjusting for structural changes.

INTRODUCTION

Country risk analysis has increased in importance in recent years. "The phenomenal growth of international capital flows is one of the most important developments in the world economy since the breakdown of the Bretton Woods system of fixed exchange rates in the early 1970s" (Council of Economic Advisers, 1999, p. 221). This increase in cross-border capital flows to developing countries created the need to understand the risks associated with these monies. In response to this growing need, which began to develop rapidly in the 1970s, a number of institutions have constructed methods to measure the country's credit worthiness (also referred to as country risk) to help investors and lenders evaluate their various exposures. Recent crises in the global economy serve to underline the importance of country-risk analysis.

While "country risk" refers to the ability and willingness of a country to service its foreign debt, private firms are also influenced by country risk because their ability to pay their foreign obligations can be seriously impaired by a sudden depreciation of the currency, exchange controls, or insufficient foreign currency in their respective central banks (Wells, 1997). Therefore, foreign companies have adjusted the level and type of investment, and the organizational form of entry into emerging countries based on their perceptions of country risk.

Because country-risk ratings are supposed to reflect the probability of default on foreign financial obligations, the extant literature on the topic shows that country risk impacts a variety of economic factors such as foreign direct investment (Gross & Trevino, 1996), equity ownership (Pan, 1996), stock market returns (Erb, Harvey & Viskanta, 1996), as well as bank loans, bond prices and bond yields (Scholtens, 1999).

Banks' loan practices in developed economies have been called into question when economic crises have unsettled emerging markets. In response to changes in a country's credit ratings, banking institutions have adjusted the volume and interest-rate spread for syndicated commercial loans to developing countries. Feder and Ross (1982) ascertained a systematic relationship between bankers' assessment of country risk and interest rate differential in the Euromarket. This paper is concerned with examining the stability of Institutional Investor's country-risk ratings over time for selected Middle Eastern countries using unit root analysis.

Measure of Country Risk

Despite the widespread use of Institutional Investor's country-risk ratings in academic research (e.g., Cosset & Roy, 1991; Lee, 1993; Somerville & Taffler, 1995; Grosse & Trevino, 1996; Pan, 1996; Haque et al., 1996), no known studies have examined the stationarity of this measure over time. The focus of this article is to find whether international banks' assessments of country risk follow a stationary path or a random walk or unit-root process. While several measures of country risk exist, we chose Institutional Investor's rating system because (1) it is the only measure that is based solely on the ratings of leading international bankers, (2) it is offered free to Institutional Investor's readers ensuring widespread dissemination, (3) it is a widely accepted measure by both industry and academia, and (4) it correlates closely with other leading measures of country risk.

As to the latter point, Cosset and Roy (1991) found a high correlation (r = 0.96) between the ratings of Euromoney and Institutional Investor. Brewer and Rivoli (1990) found that Institutional Investor and Euromoney's ratings react similarly to changes in relevant independent variables. To test for convergent validity, Dichtl and Koglmayr (1986) compared the German country-risk ratings of Manager Magazine--which is based on 225 experts from business, banks, Chambers of Commerce and other institutions--to Institutional Investor, finding a correlation coefficient of 83%.

Some country analysts have pointed to the general problem of country-risk measures' reliability and, therefore, their usefulness (Oetzel et al., 2001). They claim that country-risk models often fail to predict crises and discontinuous changes and are reactive rather than predictive. In addition, the political, social and economic context of a country or a region is so complex that it makes prediction virtually impossible. Thus, if the measure of country risk is not actually correlated with realized risk, what is the point of testing or using these measures?

We believe that country risk measures do matter, even if they fail to capture some of the discontinuous changes in the political economy. The main reason is a simple one: decision makers in banking, government agencies, and private companies use these measures and inevitably affect the participants' behaviors. While managers may not always be held accountable for overlooking the knowable, they may suffer repercussions if they ignore widely available information about country risk. Bankers' assessments are especially meaningful not only because they affect their own lending and pricing practices, but also because bankers are typically the international financial advisors of multinational companies and government agencies. While country-risk measures fail to predict discontinuous changes, they are relatively successful in explaining continuous and gradual changes, the kinds of changes that prevail the majority of the time.

Determinants of Country Risk

Researchers have attached a variety of economic and political explanations to the behavior of country-risk ratings (Feder & Uy, 1985; Dichtl & Koglmayr, 1986; Citron & Nickelsburg, 1987; Brewer & Rivoli, 1990; Cosset & Roy, 1991; Balkan, 1992; Oral et al., 1992; Lee, 1993; Haque et al., 1997; DeHaan et al., 1997). Although some dimensions of country risk have emerged, there is no agreement in the literature with respect to the exact specifications of an ideal country-risk model. Table 1 shows a list of explanatory variables used in the various studies along with their corresponding rational and expected influence on the perceived level of country risk. It should be noted that the purpose of the present study is neither to explain nor to predict country risk, but rather to examine the stability of the measure over time to ascertain whether it is even appropriate to include the country-risk measure in future empirical studies, regardless of its common-sense usage.

A number of researchers have focused on prediction instead of explanation to form a "best-fit" and a practical model, building on the premise that a good model is one that is able to accurately predict the correct level of country risk, even if it does not unveil the entire causal structure. Balkan, for example, stated: "the sign and the value of the coefficients, estimated from a structural model are not very meaningful and usually not robust with respect to alternative model specifications" (1992, p. 999). Because the empirical specification is often not amenable to devising robust structural models, Balkan suggested the replacement of the model-building approach with one that is forecast-based.

Others have criticized the empirical approach to measuring default risk, claiming that the predictive capabilities of the models are not high. The empirical literature on default risk suggests that country risk either follows a random walk or the market of international lending does not fully account for it (Citron & Nickelsburg, 1987). Our paper tests the first argument.

Importance of Unit Root for Country Risk

Understanding whether country-risk measures follow a stationary process has profound implications for international business theory and practice. If the unit root null hypothesis is rejected, or the measures of country risk are stationary over time, and the countries' risk measures will revert to their long run mean. If the unit root hypothesis is not rejected then one can say that the data follow a random walk (i.e., past observations do not provide information on the future). More to the point, the impact of a shock--whether positive or negative--on a random walk can last forever. Several important implications follow from this type of analysis. First, previously established relationships between selected environmental variables and country-risk ratings may exhibit spurious relationships (for problems of spurious relationships, see Granger and Newbold, 1974). Second, univariate forecasts of non-stationary country-risk data are not reliable because shocks are permanent and the mean does not revert to its long-run level. Since Institutional Investor's country-risk ratings have been rigorously used as proxies for country risk and/or bankers' perceptions of credit worthiness in developing and testing country-risk models, construction and interpretation, prediction and structural analysis can all be potentially improved from knowledge of the stationarity of the data.

The Middle East

We focused on the Middle East because it is an area that is of strategic importance to many developed countries. The region has a population of about 390 million people and is home to most of the world's oil reserves. Recognizing the strategic importance of the region, the United States spends as much as $60 billion a year to preserve its interests there (Abbas, 1999). Risk assumption is a given since uncertainty is a given. The Iranian revolution, the Iran-Iraq War, the Gulf War, and the Arab-Israeli conflict are just a few examples of the types of events that stem from the political instabilities that plague the region.

Country-risk ratings of one country in the Middle East may spill over to another because countries share external borders, which can lead to disputes over territories, resources (such as energy and water), and ideologies. Regional geopolitical influences both from outside--such as U.S. foreign policy--and inside--such as the Arab-Israeli conflict--may systematically affect all the countries in the region.

Many of the countries also share economic space. Members of the Organization of Petroleum Exporting Countries (OPEC), for example, share a common economic denominator: oil. The impact of oil prices on the balances of payments of oil-producing countries is exactly the opposite of their impact on developed oil-importing countries: the higher the oil prices, the greater the balance of payment surpluses. When world oil prices crashed in 1998, oil exporters experienced financial imbalances and shrinking GDPs. Iran, for example, did not meet all of its debt service obligations to Germany, Italy, and Japan as a result (Dun & Bradstreet, 1999).

A recent study shows that the U.S. and other industrialized nations might become increasingly dependent on oil from the Middle East in the next two decades and will need oil-rich countries in the Gulf to increase oil production by almost 80% to satisfy the world's growing demand (Fialka, 2001). OPEC countries produce about 40 percent of the world's oil and have more than 77 percent of its proven oil reserves. Seven of OPEC's eleven countries are in the Middle East. Saudi Arabia alone has 25% of the world's known reserves, while Iraq holds about 11% (Fialka, 2001). The next section (Section 2) contains a review of the methodology, including information about the data and the analytical technique used to test for unit root. The empirical results are explained in Section 3. Section 4 includes a discussion of the implications of our analysis. Conclusions and directions for future research are provided in the final section.

METHODOLOGY

Data

Institutional Investor provides a composite rating system that consists of a weighted average of leading international bankers' evaluations of various countries' creditworthiness. Perceptions of leading international banks regarding the risk environment impact the relative score given to each country. The ratings range from 0 (maximum risk of default) to 100 (minimum risk, most creditworthy). The ratings of international banks with the largest worldwide exposure receive more weight than those from smaller international banks. Our analysis divides OPEC from Non-OPEC countries. Figure 1 plots the ratings from 1979 to 1999 for all fourteen Middle Eastern countries considered in the present study, while Table 2 provides several descriptive statistics.

[FIGURE 1 OMITTED]

Most countries except Jordan and Oman have positive skewness, indicating a distribution with an asymmetric tail extending more toward the right. Nine of these fourteen countries have negative excess kurtosis, which indicates a relatively thin-tailed distribution compared to the normal distribution. The first-order autocorrelation of the countries is fairly high, ranging between 0.50 (United Arab Emirates) and 0.98 (Iraq).

Unit Root Processes and Testing for Unit Root

Whether a time series is stationary or not has both economic and statistical implications. If a series contains a unit root, it is not predictable or mean reverting, the usefulness of point forecast diminishes as forecast horizon increases, the unconditional variance is unbounded, Detrending by regression is inappropriate, spurious regression might happen, and finally the effect of any shock will be permanent; however, for a stationary series, the effect of a shock will fade away over time. For instance, consider the model

where: [y.sub.t] = [[rho][y.sub.t-1] = [e.sub.t] [e.sub.t] ~ iid(O, [[sigma].sup.2.sub.e]

Suppose that in some time period, say, T, there is a jump C in [e.sub.t]. Then if D = 1 (indicating y is a unit root process), [y.sub.T],[y.sub.T+1],[y.sub.t+2] will all increase by C: thus, the effect of the shock C is permanent. ...,

On the other hand, if [rho] < 1 (indicating y is stationary), a jump by C starting with [y.sub.T] will increase the successive values of y, by [c.sub.D], c[D.sub.2], c[D.sub.3], ... thus, the effect of the shock fades away over time.

On the statistical side, there are two issues: The first is about the appropriateness of trend-removal methods (by regression or by differencing). Studies have shown that spurious autocorrelation will arise whenever a unit root process is de-trended by regression or differenced by a trend-stationary process. The second statistical problem is that the distribution of the least squares estimate of the autoregressive parameter has a nonstandard distribution (unlike the usual t or F distributions) when there is a unit root. This distribution has to be computed numerically on a case-by-case basis, depending on what other variables are included in the regression (constant term, trend, other lags, and so on). This in part accounts for the proliferation of unit root tests and their associated tables.

Testing for unit root was popular in the 1980s and 1990s among econometricians because of its importance in studying the impact of a policy change or an economic event on the economy's long run behavior. The Dickey-Fuller test (1979, 1981) is the unit root test used most widely by econometricians. The test considers three model specifications: a pure random walk, a random walk with a drift (or intercept), and a random walk with both a drift and a linear time trend. The methodology is precisely the same regardless of which of the three forms of the equation is estimated. However, the critical values of the t-statistics are dependent upon whether an intercept and/or time trend is included in the regression equation. The augmented Dickey-Fuller test is similar to the Dickey-Fuller test except that some lagged changes in the series are included in the regression. Dickey-Fuller tests require that errors are statistically independent and have a constant variance. Problems may arise if the error terms are correlated and have changing variance.

Phillips and Perron (1988) developed a generalization of the Dickey-Fuller procedure that allows the disturbances to be weakly dependent and heterogeneously distributed. We thus adopt the Phillips-Perron test in the present study. Our test proceeds as follows. Consider the following model specifications:

(1) [y.sub.t] = [alpha] + [beta] + [[rho].sub.t-1] + [e.sub.t]

(2) [y.sub.t] = [alpha] + [[rho].sub.t-1] + [e.sub.t]

(3) [y.sub.t] = [[[rho].sub.t-1] + [e.sub.t]

For each country, Model (1) is estimated first. The null hypotheses [alpha] = [beta], = 0 , and D = 1 are tested by the Phillips-Perron test as follows. First, the significance of " and $ are tested by their corresponding Phillips-Perron statistics. If both are significant, the parameter estimates and their Phillips-Perron test statistics are reported. If neither of the two coefficients is significant, then Model (2) is estimated, and the significance of is tested. If significant, results from Model (2) are reported; otherwise, the results of Model (3) are reported. Critical values are from Dickey and Fuller (1981), and Fuller (1976), and we use *, **, and *** to indicate significance at the 10%, 5%, and 1% significance level, respectively.

It should be noted that in performing unit root tests, special care must be taken if it is suspected that structural change has occurred. When structural breaks are present, the various Dickey-Fuller and Phillips-Perron test statistics are biased toward the non-rejection of a unit root (Perron, 1989). Perron (1997) proposed a test for unit root that does not require an a priori fixed date of possible structural change, but treats it as an unknown. This test is particularly useful in the present study because the Middle East has been a region affected by numerous political and economic events, both within and outside the region, that span the entire period of investigation. Perron (1997) studied the various methods used to select the break points and the asymptotic and finite sample distributions of the corresponding statistics. To avoid the possible bias of the Phillips-Perron (1988) test because of the ignorance of the potential structural change, we also conduct the Perron (1997) test. The model we considered is

(4) [y.sub.t] = [alpha] + [theta]D[U.sub.t]+([delta]D([T.sub.b]).sub.t] + [beta] + [gamma]D[U.sub.t] + [[rho]y.sub.t-1] + [e.sub.t]

(5) [y.sub.t] = [alpha] + [theta]D[U.sub.t]+([delta]D([T.sub.b]).sub.t] + [beta] ++ [[rho]y.sub.t-1] + [e.sub.t]

where Tb denotes the time at which the change in the intercept occurs, D[U.sub.t] = 1(t>[T.sub.b]) and D[([T.sub.b]).sub.t] = 1(t=[T.sub.b] + 1) where (*) is the indicator function. For each country, model (4) is estimated first. + = = b t

If Y is significant, the results of model (4) are reported, otherwise, estimate model (5). If [theta] is significant, the results of model (5) are reported, otherwise the results from model (1) are reported. For models (4) and (5), the null hypothesis [rho] = 1 is tested using the critical values are from Perron (1997). Again, we use *, **, and *** to indicate significance at 10%, 5%, and 1%, respectively.

It is obvious that the range of the country risk measure considered in this study is between 0 and 100. Since not a single country hit these boundaries during the entire sample period, unit root tests are still applicable. This is why unit root tests have been widely applied to unemployment rate which is also bounded (e.g., Nelson & Plosser, 1982b).

EMPIRICAL RESULTS

The results for the Phillips-Perron test are reported in Table 3. The null of unit root is strongly rejected (1% significance level) for seven countries: Iran, Iraq, Saudi Arabia, Qatar, Israel, Syria, and Cyprus. The null of unit root is rejected at 5% significance level for United Arab Emirates. The null of unit root is not rejected at 5% significance level for six countries: Kuwait, Jordan, Egypt, Lebanon, Oman, and Bahrain. One interesting observation is that while the ratings for most of the OPEC countries are stationary, the case for unit root is stronger for Non-OPEC countries.

While most (11 out of 14) of the Middle East countries have a significant drift term, few (5 out of 14) present a significant time trend. The five countries that show a positive significant trend in the ratings are Iran, Israel, Syria, Oman, and Cyprus, largely agreeing with the plots in Figure 1.

The results from the Perron test are reported in Table 4. The unit root test results from the Phillips-Perron test in Table 3 largely hold in Table 4 despite the fact that the Perron test detected structural breaks for all the countries but Syria and Cyprus. The null of unit root was not rejected in Table 3 for Kuwait and Bahrain without structural break, but is rejected when the structural change is considered in Table 4. The Perron test reinforces the proposition that the country-risk ratings for most OPEC countries are stationary; the case for random walk is much stronger for most of the Non-OPEC countries.

In Table 4 a casual inspection of [T.sub.b], the year at which the change in the intercept occurs, and the change in the intercept in the following year, shows the impact of some historical events on country-risk ratings in the Middle East. For example, Tb for both Iraq and Kuwait detected the Iraqi invasion that led to the Gulf War. The single largest year drop of 19 points in the rating for Kuwait from 1990 to 1991 (the so-called innovational outlier) was successfully captured by the unusually large negative of -19.59. In the next section, we will examine this event in more detail.

DISCUSSIONS AND CONCLUSIONS

The empirical results of this study imply the following. First, there is a presence of unit root in Institutional Investor's ratings of country risk, suggesting a potential for spurious regression in econometric studies. Second, OPEC risk ratings are on average higher and more stable compared with non-OPEC Middle Eastern countries. Third, the Gulf War marks a change in the structure of country-risk ratings of six out of fourteen countries in the Middle East, but, ironically, Saudi Arabia is unaffected. Fourth, sub-regional categorization of Middle Eastern countries can help explain differences in the country-risk ratings of international bankers.

Unit Root in the Middle East The presence of unit root in the data of some developing countries suggests that the use of country-risk ratings in regression modeling is problematic both in longitudinal and in cross-sectional research designs using the measure. This is a significant finding given the many studies that use Institutional Investor to explain and predict country risk. Such models could potentially report only a spurious relationship. The instability of country-risk ratings for some countries in the Middle East is also indicative of these countries' political and economic structures. Countries with non-stationary country-risk ratings are more likely to experience disturbances with lasting effects.

OPEC vs. Non-OPEC

The results show that non-OPEC countries of the Middle East are more likely to exhibit unstable country-risk ratings, in addition to being more likely to have less- favorable country-risk ratings, compared with their OPEC-member neighbors. This challenges the previous proposition that country risk is likely to be adversely affected by a country's dependence on fuel exports (Haque et al., 1996). Furthermore, in the context of the Middle East, Kassicieh and Nassar (1982) claimed that the economic dependency of OPEC countries on oil has destabilized their economic and political structures because of fluctuating oil prices leading to debt repayment problems. Building on this argument, Alon et al. (1998) suggested that fluctuating revenues have made it difficult for the governments of OPEC countries in the Middle East to maintain their welfare states, resulting in a disgruntled citizenry. In contrast to these arguments, the present analysis of the Middle East suggests that countries that have oil will actually benefit from fewer fluctuations in bankers' evaluations of their country-risk structure. Evidenced in Figure 1 and Table 2, of the six OPEC countries listed, four (Saudi Arabia, United Arab Emirates, Kuwait, and Qatar) have been receiving consistently higher ratings than the included non-OPEC countries.

The Gulf War

On August 2, 1990, Iraq invaded Kuwait, triggering an immediate international backlash including UN economic sanctions, the entry of U.S. forces into Saudi Arabia, a U.S. blockade, an Arab League vote to commit troops to Saudi Arabia, and pledges by Saudis, Kuwaitis and the Japanese to contribute billions of dollars to opposition forces (Congressional Quarterly, 2000). The years surrounding Iraq's invasion of Kuwait and the Gulf War have been marked by a structural change in the country-risk ratings of eight of the fourteen Middle Eastern countries in our study. Four of six OPEC countries--Iraq, Qatar, Kuwait, and United Arab Emirates--as well as four of eight non-OPEC members--Israel, Egypt, Oman and Bahrain--are impacted by these events. The Gulf War sent country-risk shockwaves that spread beyond the borders of Iraq and Kuwait, perhaps starting a new era of regional political economy in the region. The Gulf War was a turning point in Middle Eastern global politics: it is the first time in recent history that a regional Middle Eastern war had powerful repercussions on the global economy, culminating in the direct involvement of a superpower.

Ironically, the overall country-risk ratings of Saudi Arabia were not affected by the Gulf War, despite the fact that so much Allied military activity was launched from within its borders. The military and financial support provided for Saudi Arabia has shielded its country-risk ratings from deteriorating in the eyes of international bankers. Since Saudi Arabia is by far the largest producer of oil in OPEC, the international community had a large stake in preserving the stability of Saudi Arabia's borders.

Based on the Perron test (1997), the risk ratings for most countries in the Middle East experienced a downward break in the intercept terms. The multi-country effect displayed in the country-risk ratings of the Middle East supports Alon et al.'s (1998) findings that political risk spillovers are present in the Middle East and Nigh's (1986) call to examine political risks within a regional framework. The data suggest that international bankers make global evaluations of the region in addition to the country-specific variables affecting country-risk climate.

Sub-categorization of the Middle East

OPEC and non-OPEC countries in the Middle East can be further grouped into smaller regions that can help explain the magnitude and stability of the ratings. From the standpoint of a country-risk environment, Iran and Iraq stand apart from the rest of the OPEC countries because of the long war they waged in the 1980s. Among all fourteen countries, the country-risk rating for Iraq is the most volatile with the largest standard deviation (16.29); further, it has been deteriorating over time. Both countries have been receiving relatively low credit ratings. Iraq emerged from its war with Iran only to enter into a war with Kuwait and, later, with combined Allied forces.

Although most countries in the region initially declared neutrality, by 1982 the Gulf Cooperation Council had openly expressed financial and logistical support for Iraq. Jordan, Saudi Arabia, and Kuwait declared Iran an enemy of Arabism and Islam (Kassicieh & Nessar, 1986). Using Institutional Investor's country-risk correlation analysis, Alon et al. (1998) proposed that a strong positive spillover effect of country risk exists between Iraq and the Arab nations that supported it. This article does not find the onset of the Iran-Iraq War to have been a significant event leading to a structural change in their country-risk ratings.

Those countries involved in the Arab-Israeli conflict constitute an additional subset of the non-OPEC countries. Egypt, Jordan, Syria, and Lebanon all share a border with Israel and have been directly involved in wars and land disputes (often stemming from religious disagreements) with Israel. These conflicts have destabilized the region and have made it a less desirable location for nesting international capital. Therefore, the country-risk ratings of these countries are, on average, comparatively low. Alon et al. (1998) proposed that the improvements seen in the 1990s and the strong correlation in the country-risk ratings of Israel, Jordan and Egypt are attributable to the peace treaties these countries signed, suggesting that a resolution to the conflict will lower the country-risk environment of these countries, and perhaps the entire region, and will spur regional economic growth.

RESEARCH LIMITATIONS AND FUTURE RESEARCH

This research is not without limitations. First, we have concentrated on one region, i.e., the Middle East. Results of unit root analysis may vary significantly by region. Second, we only examined one country risk variable, Institutional Investor's ratings. Future research can empirically examine other country risk assessments to see if the same conclusions can be reached.

A number of additional implications for future research can be drawn from this study. First, country-risk ratings should be examined for unit root prior to their inclusion in explanatory and predictive models. Researchers should examine country risk within a regional framework. The causes of country risk and the weight given to each cause can vary widely between regions and, sometimes, even within regions. Examining the countries by geographic region controls for common outside influences and inter-regional cooperation and conflict, providing context for the event and a framework for comparative research. Given the importance of oil and oil prices to the Balances of Payments of both industrialized nations and oil-producing nations, we suggest that future studies control for oil-related variables such as oil price, oil reserves, OPEC membership, etc.

ACKNOWLEDGEMENTS

The authors wish to thank Raj Aggarwal, James Baker, John Dunning, Richard Lee, Cornelis Los, Jennifer Oetzel, and the participants of EIBA for their useful comments.

REFERENCES

Abbas, A. J. (1999). Middle East competitiveness in the 21st century's global market. The Academy of Management Executives, 13(1), 102-107.

Alon, I., M. Martin & D. McKee. (1998). Political risk spillovers in the Middle East. In A. Best (Ed.), Annual Research Volume: Global Competitiveness. American Society for Competitiveness: 19-25.

Balkan, E. M. (1992). Political instability, country risk and probability of default. Applied Economics, 24, 999-1008.

Bargas, S. (2000). Direct Investment Positions for 1999. From the July 2000 Survey of Current Business. Bureau of Economic Analysis (retrieved 9/11/2000). www.bea.doc.gov/bea/ai/0700dip/maintext.htm.

Brewer, T. L. & P. Rivoli. (1990). Politics and perceived country creditworthiness in international banking. Journal of Money, Credit, and Banking, 22(3), 357-369.

Citron, J.T. & G. Nickelsburg. (1987). Country risk and political instability. Journal of Development Economics, 25, 385-392.

Congressional Quarterly. (2000). The Middle East. Washington D.C.: CQ Press.

Cosset, J.C. & J. Roy. (1991). The determinants of country risk ratings. Journal of International Business Studies, (First Quarter): 135-142.

DeHaan, J., C. L. J. Siermann & E. Van Lubek. (1997). Political instability and country risk: New evidence. Applied Economics Letters, 4(11), 703-707.

Dichtl, E. & H. G. Koglmayr. (1986). Country risk rating. Management International Review, 26(4), 4-11.

Dickey, D.A. & W.A. Fuller. (1979). Distribution of the estimates for autoregressive time series with a unit root. Journal of the American Statistical Association, 74, 427-31.

Dickey, D.A. & W.A. Fuller. (1981). Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica, 49, 1057-72.

Dun & Bradstreet. (1999). Country Risk Analysis: Middle East. Economic Analysis Group, Dun & Bradstreet Limited: 19-22.

Erb, C. B., C. R. Harvey & T. E. Viskanta. (1996). Political risk, economic risk and financial risk. Financial Analyst Journal, 52(6), 29-46.

Feder, G. & K. Ross. (1982). Risk assessments and risk premiums in the euro-dollar market. Journal of Finance, 37(3), 679-691.

Feder, G. & L. Uy. (1985). The determinants of international creditworthiness and their policy implications. Journal of Policy Modeling, 7, 327-344.

Fuller, W.A. (1976). Introduction to Statistical Time Series. New York: John Wiley & Sons.

Fialka, J. J. (2001). Study predicts U.S. need for oil from Middle East will increase. Wall Street Journal, (February 15): A4.

Granger, C. & P. Newbold. (1974). Spurious regressions in econometrics. Journal of Econometrics 2, 111-120.

Grosse, R. & L. J. Trevino. (1996). Foreign direct investment in the united states: An analysis by country of origin. Journal of International Business Studies, 27(1), 139-155.

Haque, N., M.S. Kumar & D.J. Mathieson. (1996). The economic content of indicators of developing country creditworthiness. International Monetary Fund Staff Papers, 43(4), 688-724.

Haque, N., D. Mathieson & N. Mark. (1997). Rating the raters of country creditworthiness. Finance & Development, 34(1), 10-13.

Institutional Investor. (1990-1997). Country Risk Ratings. Yearly Editions.

Kassicieh, S. K. & J. R. Nassar. (1982). Political risk and the multinational corporation: A study of the impact of the Iranian revolution on Saudi Arabia, Kuwait and the United Arab Emirates. Management International Review, 22(3), 22-32.

Kassicieh, S. K. & J. R. Nassar. (1986). Political risk in the Gulf: The impact of the Iran-Iraq war on governments and multinational corporations, California Management Review, 28, 69-86.

Lee, S. H. (1993). Relative importance of political instability and economic variables on perceived country credtiworthiness. Journal of International Business Studies, (4th Quarter): 801-812.

Nelson, C. R. & C. I. Plosser. (1982). Trends and random walks in macroeconomic time series. Journal of Monetary Economics, 10, 139-162.

Nigh, D. (1986). Political events and the foreign direct investment decision: An empirical examination. Managerial and Decision Economics, 7, 99-106.

Oetzel, J. M., R.A. Bettis & M. Zenner. (2001). Country risk measures: How risky are they? Journal of World Business, 36(2), 128-145.

Oral, M., O. Kettani, J.C. Cosset & M. Daouas. (1992). An estimation model for country risk rating. International Journal of Forecasting, 8(4), 583-593.

Pan, Y. (1996). Influences on foreign equity ownership level in joint ventures in China. Journal of International Business Studies, 27(1), 1-26.

Perron, P. (1989). The great crash, the oil price shock and the unit root hypothesis. Econometrica, 57, 1361-1401.

Perron, P. (1997). Further evidence on breaking trend functions in macroeconomic variables. Journal of Econometrics, 80, 355-385.

Phillips, P. & P. Perron. (1988). Testing for a unit root in time series regression. Biometrika, 75, 335-46.

Scholtens, B. (1999). On the co-movement of bond yield spreads and country risk ratings. Journal of Fixed Income, 8(4), 99-103.

Somerville, R.A. & R.J. Taffler. (1995). Banker judgment versus formal forecasting models--The case of country risk assessment. Journal of Banking and Finance, 19(2), 281-297.

Wells, R. K. (1997). Country risk management: A practical guide. Business Credit, 99(10), 20-24.

Ilan Alon, Rollins College

Min Qi, Kent State University Table 1. Factors of Country Risk Factor Variables Rational Wealth GNP per Capita * Wealthier countries can lower Growth of PCGDP consumption to implement an austerity program. Growth Propensity to Invest Countries with higher growth Change in GDP have a high opportunity cost of defaulting and are more likely to finance their debt. Financial Current Account/GNP, These variables are sources of Ability to Export Growth Rate, hard currency needed to Generate Hard Export Volume, BOP maintain debt service. It Currency Surplus, Capital should be noted that some of Inflw/Debt the variables are expressed as stocks and some as flows. Domestic Increase Money Supply High domestic prices will lead Economic Rise in Prices to capital flight. Strong Structure Consumption/GDP domestic economy can deal better with financial shocks. Economic Export Variability High export variability can Sensitivity Import Volumes sensitize the country to Export Structure currency crisis. However, the Current Acct/GDP default risk can be smaller because they depend on frequent borrowing to smooth out consumption (Cosset & Roy, 1991). Political Changes in Gov't Political instability reduces Instability Changes in Leadership the country's willingness to Political Legitimacy pay because the costs of Armed Conflict obtaining a larger share of GNP through taxation increases the possibility of governmental collapse (Citron & Nickelsburg, 1987). Assumes new government is weaker. International Credits from IMF While no explanation was Institutional Debt Rescheduling offered, Ditchtl and Koglmayr Affiliation or Service (1986) found that credits from Difficulties the IMF worsen country risk, Claims to IMF while credits from BIS improves Credits from BIS it. Leverage Net Debt/Exports Highly leveraged countries are Gov't Debt/GDP more likely to experience Foreign Debt/GDP disturbances in debt payments Debt Payment/Export during hard times. Total Int. Debt/GNP Liquidity Gold Stocks While liquidity cannot solve Int. Reserves/Imports persistent BOP problems, it can Int. Reserves help a country deal with Currency Reserves short-term fluctuations. Factor Expected Sign Wealth + Growth + Financial + Ability to Generate Hard Currency Domestic - Economic Structure Economic +/- Sensitivity Political - Instability International - Institutional + Affiliation Leverage - Liquidity + Table 2. Descriptive Statistics Skew- Kurto- Correla- Country Mean Std. Min Max ness sis tion (1) OPEC Countries Iran 23.09 6.53 12.8 36.2 0.22 -0.78 0.67 Iraq 19.87 16.29 7.1 60.4 1.61 1.69 0.98 Saudi Arabia 63.41 8.91 53.7 85.4 0.91 -0.02 0.96 Qatar 54.82 3.31 49.7 65.8 1.84 5.51 0.71 United Arab 59.77 2.52 54.2 66.2 0.03 1.79 0.50 Emirates Kuwait 59.43 8.99 41.8 79.3 0.23 0.08 0.84 Non-OPEC Countries Israel 39.99 8.64 28.3 54.3 0.52 -1.23 0.90 Syria 22.77 5.76 16.8 39.3 1.73 3.04 0.90 Jordan 32.87 7.29 20.7 44.7 -0.52 -0.81 0.93 Egypt 31.59 6.29 22.4 44.4 0.13 -0.69 0.89 Lebanon 17.43 8.71 7.3 32.5 0.55 -1.15 0.92 Oman 50.73 2.45 45.4 53.5 -0.84 -0.56 0.62 Cyprus 44.73 8.18 33.1 57.3 0.23 -1.28 0.97 Bahrain 54.11 3.80 48.1 62.9 0.41 -0.24 0.82 Table 3. Phillips-Perron (1988) Test for Unit Root in Country Credit Ratings Country [sup.[alpha]]Z [sup.[beta]]Z ([t.sub.a]) ([t.sub.[beta]] OPEC Countries Iran 15.72 4.81 *** 0.68 4.05 *** Iraq Saudi Arabia 10.57 3.25 ** Qatar 28.48 5.38 *** United Arab Emirates 34.58 3.41 ** Kuwait 15.24 2.34 * Non-OPEC Countries Israel 11.07 6.40 *** 0.61 10.25 *** Syria 5.43 3.99 ** 0.19 3.74 *** Jordan 3.83 2.69 ** Egypt Lebanon Oman 33.09 3.38 ** 0.21 2.57 * Cyprus 21.33 4.71 *** 0.69 5.18 *** Bahrain 15.21 2.45 * Country [sup.[rho]]Z ([t.sub.[rho]] OPEC Countries Iran 0.29 -5.07 *** Iraq 0.85 -6.05 *** Saudi Arabia 0.81 -3.8 *** Qatar 0.47 -5.54 *** United Arab Emirates 0.42 -3.43 ** Kuwait 0.73 -2.57 Non-OPEC Countries Israel 0.72 -6.49 *** Syria 0.73 -4.79 *** Jordan 0.87 -1.81 Egypt 1.02 0.40 Lebanon 1.00 -0.34 Oman 0.35 -3.38 * Cyprus 0.54 -4.49 *** Bahrain 0.71 -2.58 Note: (1.) For each country, Model (1): [y.sub.t]=[alpha]+[beta]+[rho] [y.sub.t-1]+[e.sub.t] is estimated first. The of and are tested their corresponding Phillips-Perron statistics. If both are significant, the parameter estimates and their Phillips-Perron test statistics are reported in Table 2. If either of the two coefficients is not significant, then Model (2), [y.sub.t]=[alpha]+[rho][y.sub.t-1]+[e.sub.t] is estimated, and the significance of is tested. If significant, results from Model (2) are reported, otherwise, report the results of Model (3): [y.sub.t] = [rho] [y.sub.t-1]+[e.sub.t] (2.) The null hypotheses [alpha] = 0, [beta] = 0, [rho] = 1 are tested by the Phillips-Perron (1988) test. *, **, and *** indicate significant at 10%, 5%, and 1% significance level, respectively. Critical values are from Dickey and Fuller (1981), and Fuller (1976). Table 4. Perron (1997) Test for Unit Root with Endogenous Time Break Country [T.sub.b] [alpha] [theta] [delta] [beta] OPEC Countries Iran 1993 4.41 -8.63 3.51 1.20 (2.80) (-5.87) (1.77) (10.19) Iraq 1990 4.95 -2.83 -3.45 0.00 (1.74) (-1.69) (-2.01) (0.02) Saudi 1985 65.55 -7.90 2.56 -0.51 Arabia (6.63) (-5.04) (1.53) (-3.60) Qatar 1990 37.10 -0.66 -4.50 -0.07 (6.20) (-0.66) (-3.85) (-0.75) United Arab 1989 71.42 -19.71 4.95 -0.52 Emirates (5.81) (-4.61) (2.98) (-2.62) Kuwait 1989 89.54 -57.63 -19.59 -1.91 (12.26) (-12.09) (9.67) (-7.88) Non-OPEC Countries Israel 1991 7.83 3.38 -4.91 -0.43 (4.61) (2.85) (-3.68) (4.94) Syria NA 5.43 0.19 (3.99) (3.74) Jordan 1988 14.58 -17.59 3.98 -0.21 (4.34) (-4.29) (2.18) (-1.17) Egypt 1990 47.47 -58.22 1.94 -1.94 (4.45) (-4.26) (0.83) (-4.49) Lebanon 1988 17.31 -40.73 2.65 -1.22 (2.60) (-2.80) (0.95) (-2.23) Oman 1990 36.37 -3.61 -2.06 0.52 (4.48) (-2.44) (-1.30) (3.65) Cyprus NA 21.33 0.69 (4.71) (5.18) Bahrain 1989 64.99 -12.05 5.52 -0.54 (5.42) (-3.48) (3.11) (-2.61) [t.sub.- Country Y [rho] [theta]] OPEC Countries Iran 0.29 -9.89 *** Iraq 0.71 -6.48 *** Saudi 0.13 -7.02 *** Arabia Qatar 0.34 -6.65 *** United Arab 1.48 -0.14 -6.06 ** Emirates (4.41) Kuwait 3.64 -0.19 -13.14 *** (10.76) Non-OPEC Countries Israel 0.65 -8.80 *** Syria 0.73 -4.79 *** Jordan 1.01 (3.70) 0.64 -4.82 Egypt 4.61 -0.09 -4.46 (4.63) Lebanon 3.68 0.21 -3.45 (2.95) Oman 0.20 -4.58 Cyprus 0.54 -4.49 *** Bahrain 0.67 -0.09 -5.67 ** (2.67) Note: 1. Model: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] denotes the time at which the change in the intercept occurs. [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] where 1(*) is the indicator function. The numbers in parenthesis are the t statistics. 2. The null hypothesis [rho] = 1 is tested by the Perron (1997) test. *, **, and *** indicate significant at 10%, 5%, and 1% significance level, respectively. Critical values are from Perron (1997)."