IN SEARCH OF AN EMPIRICALLY DETERMINED NONLINEAR MULTIVARIATE THEORETICAL MODEL FOR FORECASTING...

ABSTRACT

A number of models have been proposed for forecasting foreign exchange currency rates (XRs) variations. A present study, utilizing linear and nonlinear regression models, finds inconsistencies in the claim that the predictive variables, (GDP, inflation, unemployment rate,

and interest rates) can each singularly provide predictive measures of XR variations. Although inflation rates with seemingly strong co-linearity with interest rates do appear to explain the XR variations for several countries, these relationships are both positive and inverse, making the reliance on inflation rates as an XR predictor inconclusive. The predicative variables impact the XR forecasting relations differently, in part due to the problem of size, and, thus, there is an absence of a universally applicable model for XR forecasting. This is tantamount to each country developing its own XR forecasting model.

INTRODUCTION

The operations management literature characterizes a winning competitiveness in the heroic championship of quality, price, and flexibility. Behavioral and human resource experts believe that competitive attainability is possible through effective cross-cultural communication and political risk abatement. Functional strategies are believed to provide a coordinated co-production of strategic objectives. Likewise, economics and finance are also bestowed important roles. As with modern art and music, each discipline performs its own function, not with the intention to supplant the other, but harmoniously. Foreign exchange (FEX) fluctuations can pose economic, transaction, and translation risks to a firm's finances and, thus, to its competitiveness. FEX fluctuations can also impact a country's economic macroeconomic variables, and its sociopolitical and economic objectives. Thus, the importance of forecasting on many of the factors that impact strategy planning and decision-making should not be de-emphasized.

Factors such as interest rates, GNP growth, inflation and unemployment are considered as explanatory variables. However, we have attempted to examine the simultaneous impact of these factors on the exchange rates (XRs). As such, our approach is more relevant to many real-world phenomenon, judging that effect is induced by the simultaneous impact of several factors. This approach could provide a formidable basis for business firms and speculators for investment decisions, and avert financial and foreign exchange exposures. Approximately two trillion dollars are transacted daily in the Futures Markets and Foreign Currency Markets for speculative, precautionary, and transaction purposes. While the extent of governments' currency interventions is unknown, it has nevertheless contributed to this ever-increasing volume.

The GATT's Paraguay Round has swayed trade restrictions to voluntary export restraint (VER) by countries with consistent positive trade balance, or BOP. However, similar to quotas, the host country is not insulated from the negative economic consequences of VER, (Kreinen, 1998). In fact, quotas could be more detrimental to the economy. Negative BOP leads to capital inflow, i.e., a country must borrow from foreigners to pay for excess imports. As with other policy options, imports may respond to a country's anti-inflationary objectives. Virtually, all countries are dependent on imports to keep their industries in operation and to satisfy their consumers. Ipso facto, other than the national security objectives, health and welfare, and infant industry arguments, trade restrictions cannot be defended, since such restrictions would hamper economic growth or social welfare. In some cases, however, trade restrictions have paradoxically led to industrialization, especially due to the desire of MNCs seeking international monopolies. Trade restrictions may also aggravate economic objectives, vis-a-vis business cycles, and economic development.

THE STATE OF THE LITERATURE

Extensive research into currency exchange literature uniformly concludes that the foreign exchange currency (FEX) affects the domestic economy and international trade, and, thus, the competitiveness of nations. Likewise, foreign XR forecasting has a significant impact on the firm's cash flows, and, thus, on its financial and foreign exchange exposures. Such exposures may impact the firm's competitiveness and marketing and resource strategies. In general XRs are affected by many factors, such as the country's BOP, political stability, productivity and income growths, as well as, interest rates, inflation, the government monetary and fiscal policies, and macroeconomic reforms. Major international crises, such as that which occurred in the fall of 1997 in Southeast Asia, can also have an impact on foreign exchange market.

Ceteris paribus, an increase in demand for currency (i.e., a forward demand shift) will cause foreign XRs to move upward, or appreciate, and vice versa. The principles also state that, ceteris paribus, an increase in currency supply will cause FEX currency to depreciate, and vice versa. Therefore, the causes of FEX currency appreciation are an increase in demand, a decrease in supply, ceteris paribus, and vice versa, i.e. the market forces. Aware of this economic phenomenon, governments resort to currency interventions by buying and selling currencies to induce the desired results, when faced with the BOP problems. This mechanism is similar to the central bank's open market activities, i.e., bond transactions to adjust the money supply and interest rates. Conversely, an inappropriate decision to revalue or devalue currency will respectively create surplus or shortage of currency in the world market.

The relationship between the interest rate and real XR is shown by:

(1) EP/Pw = g + vr

where, EP is the foreign price of US goods, r is the interest rate and g and v are constants.

Because of the importance of the relationship between interest rate, XR and inflation, we can rewrite the macroeconomics model in the following terms:

(2) Y=C+I+G+(X-M)=a+b(I-t)Y+e-dr+ G+ g-mY-n(EP/Pm)

(3) Thus, r=(a+b+ e)/d-({1-b(1-t)+m}/d)Y - n/d(EP/Pm)+1/dG

where, Y is National Income or GDP, C is personal consumption, I is investment expenditure, X is export, M is import, a, d, and e are constants and b, m, and n are coefficients to be estimated.

This equation relates the interest rate, r to output Y, and real XR EP/Pm and government expenditure, G. It shows that the interest rate is negatively related to the XR. The dollar appreciation shifts the IS curve downward, and depreciation of the dollar raises the curve. To get the IS curve, we can eliminate the XR through the above equations.

(4) r=(a+ e+ g-ng)/(d+nv) - {1-b(1-t)+m}/(d+nv).Y+{1/(d+nv)}G

[Relations 2-4 are attributed to Hall and Taylor (1996)].

All the above terms are divided by d+nv term, and therefore Relationship (3) can be further simplified:

(5) r=[(a+ b+ g- ng)-{1-b(1-t)+m}.Y+G]/d+nv

The real XR is a measure of XR adjusted for difference in price levels between two countries. Therefore,

(6) XR=Px/Pm= (E)Px/Pm and net export, (X-M)=g-mY-n (E)Px/Pm

where, x and m represent export and import respectively, g is a constant, m and n are coefficients.

Trade Deficits

Trade deficits have a definite impact on a country's FEX currency XR. Usually, negative trade deficit has a negative impact. Kooros (2000) has attributed the U. S. trade deficit to the Britton Wood Agreement, where Allies chose the dollar as the convertible currency, requiring other nations to maintain reserves in the dollar. This led to increased demand for the dollar and thus appreciation in its value." A strong dollar led to the U. S. trade deficit. Otherwise, in a floating exchange rates system, the BOP equilibrium is believed to be in constant check by the currency depreciation or appreciation, which would make foreign goods cheaper and thus in more demand, or expensive, and, thus, in less demand. Theoretically, currency fluctuations are the work of the market, whereas currency devaluation/evaluation are of the government. It is doubtful whether foreign currency speculative demand can be considered a natural market force, since such a demand is predicated on purely profit motive, rather than transaction and precautionary satisfaction.

Currency Fluctuations

Currency fluctuations impact international trade and the domestic economy, (Brool and Eckwer, 1999). De Grauwe (1992) classified 12 major industrial countries into two groups; those within the European Monetary System (EMS) with relatively stable XRs, and others with their XRs fluctuating considerably. He concluded that during the 1980s, "the growth rates of output and exports have, on average, been significantly lower in the EMS countries than in the non-EMS countries." This was due to the ability of the EMS to effectuate policy on XR stability through only one central bank, whereas other currencies abided by their independent central bank policies.

Although currency fluctuations impact economic development and the international trade, however, according to studies by Broll and Eckwer, (1999), strategic global marketing decisions do not solely evolve around XRs. The McDonald Company is a case in point. McDonald's involvement in Japan entailed importation of almost all the hamburger ingredients into Japan. With the tremendous decline in Japanese Yen, and the recent recession and financial crisis in South East Asia, McDonald's long term insight into the futures market provided the needed insulation, (Ball and McCullogh, 1999).

Chen's investigation (1999) into the context of a nonlinear discrete-time dynamic system within an open macroeconomics framework, postulated a domestic price flexibility that real XR can exhibit with complex dynamic behavior. In Chen's view, this is in the absence of the Marshall-Lerner condition and "the dynamic complexity is due to the insignificant capital mobility." In terms of international macroeconomics, perhaps the most striking features shown by the data are the recurrent, large cycles in nominal and real XRs. The large and persistent fluctuations of the rates have stimulated the development of various versions of XR, overshooting theory pioneered by Dornbusch (1976). Dornbusch's model shows that a monetary expansion induces an immediate domestic money depreciation in excess of its long-run equilibrium value, i.e., an overshooting of the XR due to an exogenous shock. Subsequent studies attributed such large fluctuations to exogenous shocks. In this class of models, a unique equilibrium path toward the steady state is ensured. No persistent XR fluctuations can occur in a stationary economy which does not experience any exogenous shock to its economic fundamentals (Chert 1999).

Meese and Rogoff (1983) first challenged the explanatory power of existing models of XR determination by showing their inferior out-of-sample forecasting performance relative to a simple random walk model. Many empirical studies since then noted that real XRs are also volatile and that "measurable fundamentals do not explain the XR well, even contemporaneously." However, a simple deterministic classical trade model, incorporating capital mobility, can generate complex dynamics of the real XR, totally independent of movements in underlying economic fundamentals.

The importance of currency fluctuations among countries is demonstrated by the emergence of the eurodollar in January 1999, unifying eleven European national currencies. This was followed by other trade unions such as MERCOSUR, the negotiation of a single currency between Russia and Belarus, Argentina's attempt to dollarize the peso, as well as the efforts of other countries towards a unification of their currencies. This regional currency unification can undermine or strengthen the SDR, or gold standards. But, it lead to the conclusion that the current system is inadequate to respond to many countries' needs (Hanke, 1999). The General Motors' sponsored foreign exchange benchmarking study in 1988, led to a summary report of the best foreign exchange risk management practices, including seven general best practices that were being followed by a large majority of the companies in the group. Those companies have written FX policies, defining which kinds of exposures should be hedged. Companies with centralized FX management and sufficient systems enable them to manage and price the derivatives and to underlie the business exposures being hedged. This study, which included 31 MNCs with average sales of $ 50 billion, including the world's largest automobile and oil companies, and five of Fortune 10 companies, concluded that by pooling their knowledge, they can collectively establish the best FX practices (Wallace, 1998).

The International Monetary Fund believes that XR's volatility is present when it is allowed to float freely (Bayoumi and MacDonald, 1999). Such volatility has led economists "to advocate moving away from flexible XR international monetary regime and towards one based on greater XR fixity" (McKinnon, 1988; Mundell, 1992; and Williamson, 1987). This argument is advanced by the proponents of a greater monetary integration in Europe. Because prices in the goods markets are generally regarded as being sticky, nominal XRs volatility is transferred into comparable real XRs. To reduce the FX risk for countries affected by market volatility, "a new exchange-rate mechanism which integrates the best of the `fixed and market floating system' is required" (Wachtel, 1995). U. S. historical data indicates that with low real XR, export is high and vice versa; thus, the presence of an inverse relationship between strong XR and net export.

Current Theories

Currently, there are two prevailing theories that attempt to explain variations in XRs. First, is the PPP. As mentioned earlier, this theory states that under a floating XR system, price increases in one country will be accompanied by a decrease in that country's currency (currency depreciation) in the international market, therefore restoring the BOP equilibrium. The floating XR system, to which many countries have reverted, has become a natural force behind the BOP equilibrium. This also means that a strong currency would lead to excessive imports, and thus negative BOP, and a weak XR would produce the opposite effect. Because PPP is consumer-based, expenditures eliminate differences in relative prices: "marketers use the data to analyze the comparison of consumption changes with the level of development" (Ball and McCulloch, 1999, p. 208). Table 1 shows the PPP based GNP.

TABLE 1
Selected Countries' GNPs vs. PPP-based GNPs

Country        GNP/Capita     GNP/Capita
               Based on XR    Based on PPP
                              Values in 1993 US$

US             $24,740        $24,740
Switzerland     35,760         23,660
Germany         23,560         16,850
Canada          19,970         20,230
Denmark         26,730         19,560
Kenya           270            1,290
India           300            1,220
Zambia          380            1,040
Tanzania        90             580

Source: World Development Indicators, 1997,pp. 182-183

This can be shown by the country's per capita GNP based on XRs vs. per capita GNP based on PPP, (column 3). As can be seen, the use of PPP provides a more realistic GNP value, especially for the developing countries, since among other things, GNP can be exaggerated by the industrialized countries' pervasive higher relative prices, the extent of disproducts and unnecessary expenditures. The second theory is the random walk theory or efficient market hypothesis, (EMH) which assumes that perfect information exists in the market place and that variations in XRs depend on such information, and that the new information will cause a movement in the XR. The nature of the information determines the direction of the XR variation (Pass & Lowes, 1993).

In the early Eighties, there was substantial support for PPP as a valid approach for forecasting XRs. A study by Frenkel in 1981 claimed that PPP was indeed a statistically sound method of forecasting XRs (Hoque, 1995). Empirical evidence in the last fifteen years, however, suggests that PPP has only a marginal relationship, at best, with XR variations (Norrbin & Conover, 1998). This raises the question of the conditions that have caused unfavorable research on PPP over this period. Hoque (1995) suggests that early statistical tests with regard to PPP and XR variations are invalid, due to the false assumption that spot XRs and price levels are stationary. This, in turn, deems all critical values associated with these models inappropriate for statistical testing, therefore, making it impossible to claim that PPP has a significant relationship with XR fluctuations.

The empirical evidence suggesting the PPP's somewhat limited ability to forecast XR variations does not inhibit its usefulness. A study of variations in the Canadian dollar versus the U.S. dollar suggests that PPP out-performs all other techniques in the long run, especially for periods beyond a three-year horizon. When a country's economies are subject to tariffs, quota systems, and domestic market distortions, which prevent the market from operating freely, there is some indication that its PPP has some worth in XR forecasting and in GNP determination (Kooros, 2000).

Beyond the PPP approach to FX forecasting, the "Big Mac" claimed to reflect the differential XRs among countries. While PPP has exhibited some validity in the long run, especially for comparing GNP performance, the Big Mac has been received with skepticism and mixed results, because of its particularism and the absence of significant local or indigenous content. As with the CPI, dealing with a single good price change to measure inflation is flawed and invalid. A single good, especially when relatively exogenous to an economy is not a valid measure of XRs, regardless of who is promoting the concept, as a marketing ploy. A bunch of grapes produced indigenously in a country may be a far better measure of XR differential than the Big Mac. On the other hand, the PPP approach to XR determination is far more reliable than a single good, especially in the long run. Some empirical studies indicate the failure of PPP to hold continuously (Foot and Rogoff, 1995, and MacDonald, 1995). However, there is now growing evidence to suggest that PPP holds as a long-run phenomenon (Edison, 1987; Frankel, 1988; and Diebold, Husted, and Rush, 1991). When the predominant force upsetting the PPP relationship is nominal, it will have a transitory effect on deviations from PPP, which is essentially the Dornbusch's model. Conversely, if, as suggested by Stockman, (1987) the source of PPP disturbances are truly "real" in nature, this will have a permanent effect on the real XR (Bayoumi and MacDonald, 1999).

This is where many researchers turn to the EMH to help explain and forecast XR variations. The random walk theory assumes that since perfect information exists in the market place, therefore, the current level of XR is the best predictor of future values. An analysis of the Canadian market shows that EMH appears to be true in the short run, i.e. all available information is already incorporated in the XR. In the long run, however, it becomes harder to use the EMH to estimate XRs as information about expected economic performance two years down the road is bound to be somewhat ambiguous (Spiro, 1995).

Some similarities exist between the stock market and the foreign exchange markets. The real problem associated with using the EMH as an XR forecasting tool is its underlying assumptions, i.e. perfect information, zero transactions costs, zero information costs, and the existence of mere normal profit earnings. These assumptions being unrealistic and theoretical, any empirical testing carried out to ascertain that the random walk theory holds is fundamentally flawed (Ziets, 1995). This is not to say, however, that the random walk theory is useless, as in the short run it makes sense that XR provides an accurate representation of the information currently available in the market. However, to ignore the underlying market conditions seems unrealistic.

METHODOLOGY

In this study, an attempt was made to develop a universal predicative model to forecast XRs. The idea was that observations from many countries on their predicative variables, such as inflation, interest rates, per capita GDP growth, and unemployment could have provided sufficient number of observations to be employed in the development of a universal model. This approach could have satisfied the necessary assumptions for the use of multiple regressions. Actual observations from ten countries proved the absence of a uniform data in these countries. Furthermore, the problem of the size of the XR, and the manner by which inflation and other data had been reported, would have made the development of any universal model untenable. It was, therefore, decided to develop predicative models for XRs for each country, by

* First, applying simple linear and nonlinear relationships of each individual predictive variables with XR, and

* Second, applying multivariate regressions models, both linear and nonlinear. The results of these approaches appear in sequence. The nonlinear multivariate approach has attempted to identify the variables that appear to explain the XR movements and the nature of these movements in each country under study. These movements have been described both mathematically and in detail.

Our analysis starts with the study of the Pearson's correlation coefficient of XR with three selected variables INF, INT, and GDP.

Correlation Analysis

Table 2, gives the correlation coefficient between XR with inflation rate, interest rate, and GDP for each country. These correlation coefficients indicate that:

* The Canadian XR is significantly linearly related to the INF, INT, and GDP individually, the strongest being with INT.

* France's XR is not significantly correlated to any of the three variables.

* Germany's XR is significantly linearly related to the INT.

* UK's XR is significantly linearly related to the INF, INT.

TABLE 2
Correlation Coefficients

Country       INF         INT          GDP

            -0.47736    -0.59044     0.46717
Canada       (0.05)      (0.01)      (0.05)

            0.11779      0.26543     0.11419
France      (0.6416)    (0.2871)    (0.6519)

Germany     0.37341     -0.52441    -0.18613
            (0.1398)    (0.0255)    (0.4596)

UK          0.72378     0.64924     -0.41257
            (0.0007)    (0.0036)    (0.0889)

Numbers in the parentheses indicate significance levels (P-values).

If one is to single out one variable to be linearly related to XR for these countries collectively, that variable is the interest rate. Except for France, the variable INT has a statistically significant linear relationship with XR. There is no significant linear model representing XR in terms of INF, INT, and GDP to account for a large portion of variation in XR. Therefore, we explored more sophisticated models, including second degree polynomial with variables

INF, INT, GDP, and INFSQ=INF*INF, INTSQ=INT*INT, GDPSQ=GDP*GDP, the squares and INTGDP=INT*GDP, INFINT=INF*INT, the interaction terms.

Our selection is based on the contribution of each variable to the model and R-square (the proportion of variation in XR accounted by the model.)

Non-linear Model for Canada

Our analysis of the data (see Appendix for the raw data) for Canada leads to the following nonlinear model which includes the interaction term INT*GDP.

(1) XR=1.2231 +0.0704 (GDP) - 0.0056 (INT*GDP)

This model, which is highly significant, captures about 61% of the total variation in XR (R-square = 0.6143). A partial SAS output and graphical presentation of the raw data, the model, and residuals are given below.

Analysis of Variance

             DF   Sum of Squares    Mean Square      F      Prob>F

Regression    2     0.08363511       0.04181756    11.94    0.0008
Error        15     0.05251321       0.00350088
Total        17     0.13614832

            Parameter    Standard
Variable    Estimate     Error        Sum of Squares   F         Prob>F

INTERCEP    1.22312800   0.02030812   12.70350905      3628.66   0.0001
GDP         0.07036640   0.01476397    0.07952466        22.72   0.0002
INTGDP     -0.00563207   0.00143508    0.05392094        15.40   0.0014

This graph indicates the existence of an interaction between GDP and INT. High values of GDP with low values of INT produce high XR.

[GRAPH OMITTED]

The graph of this nonlinear model shows its resemblance to the previous graph.

[GRAPHS OMITTED]

The following three residual graphs (residuals against variables in the model) do not exhibit any serious abnormality and lack of appropriateness of the model proposed.

Non-linear Model for France

Our analysis of the data (see Appendix for the raw data) for France leads to the following nonlinear model which includes the interaction term INF*INT.

(2) XR=5.1976 - 0.1045 ([INF.sup.2]) - 0.0307 ([INT.sup.2]) + 0.1661 (INF*INT).

This model, which is highly significant, captures about 76% of the total variation in XR (R-square = 0.7590). A partial SAS output and graphical presentation of the raw data, the model, and residuals are given below.

Analysis of Variance

             DF   Sum of Squares   Mean Square     F     Prob>F

Regression    3   20.17193901      6.72397967    14.69   0.0001
Error        14    6.40649215      0.45760658
Total        17   26.57843116

           Parameter     Standard
Variable   Estimate      Error        Sum of Squares      F     Prob>F

INTERCEP    5.19762717   0.48776232   51.96201852      113.55   0.0001
INFSQ      -0.10450923   0.01658289   18.17528577       39.72   0.0001
INTSQ      -0.03065569   0.00953579    4.72934310       10.33   0.0062
INFINT      0.16600835   0.02667655   17.73400144       38.75   0.0001

The above graph (raw data) indicates the existence of an interaction between INF and INT. High values of INF with high values of INT produce high XR.

[GRAPH OMITTED]

The graph of this nonlinear model shows its resemblance to the previous graph.

[GRAPH OMITTED]

The three residual graphs (residuals against variables in the model) do not exhibit any serious abnormality and lack of appropriateness of the model proposed.

[GRAPH OMITTED]

Non-linear Model for Germany

Our analysis of the data (see Appendix for the raw data) for Germany leads to the following nonlinear model which includes the interaction term INF*INT.

(3) XR=1.9194 + 0.4462 (INF) - 0.0211([GDP.sup.2]) - 0.0555 (INF*INT).

This model, which is highly significant, captures about 62% of the total variation in XR (R-square = 0.6164). A partial SAS output and graphical presentation of the raw data, the model, and residuals are given below.

Analysis of Variance

             DF   Sum of Squares   Mean Square     F    Prob>F

Regression    3     2.38563159      0.79521053   6.96   0.0049
Error        13     1.48446293      0.11418946
Total        16     3.87009452

            Parameter     Standard
Variable    Estimate       Error      Sum of Squares      F     Prob>F

INTERCEP    1.91944624   0.17607574    13.56996992     118.84   0.0001
INF         0.44623118   0.10081221     2.23727593      19.59   0.0007
GDPSQ      -0.02109043   0.01135266     0.39409538       3.45   0.0860
INFINT     -0.05549551   0.01408361     1.77302035      15.53   0.0017

The above graph (raw data) indicates the existence of an interaction between INF and INT. Moderate values of INF and INT produce high XR.

[GRAPH OMITTED]

The graph of this nonlinear model shows its resemblance to the previous graph.

[GRAPH OMITTED]

The residual graphs (residuals against variables in the model) do not exhibit any serious abnormality (except for one relatively large residual) and lack of appropriateness of the model proposed.

[GRAPH OMITTED]

Non-linear Model for UK

Our analysis of the data (see Appendix for the raw data) for United Kingdom leads to the following nonlinear model which includes the interaction term INF*INT.

(4) XR=1.5422-0.1266(GDP)- 0.0285([GDP.sup.2]) +0.0017(INF*INT).

This model, which is highly significant, captures about 64% of the total variation in XR [R-square = 0.7376). A partial SAS output and graphical presentation of the raw data, the model, and residuals are given below.

Analysis of Variance

             DF   Sum of Squares   Mean Square     F     Prob>F

Regression    3     0.87633056      0.29211019   13.12   0.0002
Error        14     0.31172190      0.02226585
Total        17     1.8805246

             Parameter     Standard
Variable      Estimate      Error      Sum of Squares      F     Prob>F

INTERCEP     1.54217344   0.06818683    11.38950254     511.52   0.0001
GDPSQ       -0.12658513   0.04493785     0.17667704       7.93   0.0137
GDPSQ        0.02853170   0.01058293     0.16183882       7.27   0.0174
INFINT       0.00171409   0.00070218     0.13267997       5.96   0.0285

The above graph (raw data) indicates the existence of an interaction between INF and INT. High values of INF and INT produce high XR.

[GRAPH OMITTED]

The residual graphs (residuals against variables in the model) do not exhibit any serious abnormality and lack of appropriateness of the model proposed.

[GRAPHS OMITTED]

In summary, the following models have been developed for various countries:

Country                                  Model              [R.sup.2]

Canada   XR=1.2231+0.0704 (GDP) - 0.0056 (INT*GDP)            0.61
France   XR=5.1976-0.10451 ([INF.sup.2])-0.0307
           ([INF.sup.2])+0.1661 (INF*INT)                     0.76
Germany  XR=1.9194 + 0.4462 (INF) - 0.0211([GDP.sup.2]) -
           0.0555 (INF*INT)                                   0.62
UK       XR=1.5422-0.1266(GDP)- 0.0285([GDP.sup.2]) +
           0.0017(INF*INT)                                    0.74

CONCLUSION

The results of this study have important implications for scholars, governments, and business leaders. Business is concerned with minimizing the financial and foreign currency exposures; governments are concerned about the BOP positive stability, economic prosperity, and international economic policy. Our analysis has shown the inconsistencies between any of the four variables, i.e. GDP, inflation rate, interest rate, unemployment, and the XR, initially perceived to be formidably predictive of XRs. However, there is some evidence that the inflation rate has a significant relationship with XR, but this relationship is neither positive nor inverse, depending on the country. Similar behavior has also been observed between interest rates and XRs, and between interest rates and inflation rates.

These results agree with previous research studies that indicate a difficulty in building a universal model, which can help forecast XRs. The underlying reason is the apparent existence of many other variables, both indigenous and exogenous, that are not (could have been) included in the statistical models presented in this paper. Furthermore, XRs are not free from government interventions or from speculative behaviors, and, therefore, they do not freely interact with these variables through the market system. For example, government fiscal and monetary policies may distort the PPP theory, which was conceded in our analysis, indicating negative correlation between inflation rate and the XR in India. Also, protectionist measures, which prevent a free market system from operating efficiently, i.e. EMH, make it difficult to predict XR volatility when economic indicators change. Until the day that markets can operate under complete freedom with perfect information, forecasting XR variations will remain an arduous and ambiguous task.

Furthermore, the plausibility of the presence of a strong co-linearity between interest rates and inflation is detected, which can further restrict the results. Of course, there are other factors, which impact XRs. These include a country's economic and political and even macroeconomic structural adjustments and privatization, which need to be quantified to fit other more complex models.

REFERENCES

Amrhein, Denise G. (1998, Fall). XRs: Their importance, purpose and role in translation. Multinational Business Review.

Ball, Don, and Wendell McCulloch.(1999). International business. Boston: McGraw-Hill.

Bayoumi, Tamim, and Ronald MacDonald. (1999, March). Deviations of XRs from purchasing power parity: A story featuring two monetary unions, International Monetary Fund.

Brigham, Eugene F., Louis C. Gapenski, and Michael C. Ehrhardt. (1999) Financial management: Theory and practice. (4th ed.). New York: Dryden Press.

Broll, Udo, and Bernhard Eckwert. (1999, July). Exchange rate volatility and international trade. Southern Economic Journal.

Burdekin, Richard C. and Pierre Siklos. (1999, May) XR regimes and shifts in inflation persistence: Does nothing else matter? Journal of Money, Credit, and Banking.

Business cycles, international linkages, and XRs. Available http://www.imf.org/external/pubs/ft/weo/weo0598/index.htm

Chen, Shikuan. (1999, Summer). Complex dynamics of the real XR in an open macroeconomic model. Journal of Macroeconomics.

Cornwall, John and Wendy Cornwall. (1997, Summer). The unemployment problem and the legacy of Keynes. Journal of Post Keynesian Economics.

Epstein, Gene. (1999, May). Economic beat: The Calvin Coolidge theory of unemployment doesn't compute in the world of labor statistics. Barron's.

Eiteman, David K., Arthur I. Stonehill, Michael H. Moffet. Multinational Business Finance. (8th ed.). New York: Addison-Wesley.

Hanke, Steve H. (1999, Winter). "Reflection on XR regime." Cato Journal, 18.

Harvey, John, T., and Stephen F. Quinn. (1997, June). Expectations and rational expectations in the foreign exchange market, Journal of Economic Issues.

Hoque, Asural. (1995, March) "A test of the purchasing power parity hypothesis." Applied Economics, 27 (3) 311(5).

Hu Michael Y., Guoqiang Zhang, Christine X Jiang and B Eddy Patuwo. (1999, Winter). A cross-validation analysis of neural network out-of sample performance in XR forecasting. Decision Sciences.

Kaminsky, Graciela; Lizondo, Saul and Reinhart, Carmen. (1998, March) "Leading indicators of currency crisis." International Monetary Fund Staff Papers, 45 (1) 1 (48).

Kreinin, Mordechai . E. (1998) International economics: A policy approach, (8th ed.). Florida: Dryden.

Kooros, Syrous K. (2000, November). "An investigation into the plausibility of developing a general model for forecasting foreign exchange currency rates. "Journal of Current Research In Global Business.

Miller, Kent D., and Jeffrey J. Reuer. (1998) Firm strategy and economic exposure to foreign XR movements. Journal of International Business Studies.

Murray, John. (1997, Summer). XRs and Monetary Policy: Conference summary. Bank of Canada Review.

Achane, D.M. (1997, November). "Purchasing power parity: An analysis based on the evolutionary spectrum." Applied Economics, 29 (11), 1515(10).

Norrbin, Stefan and Conover, Mitchell. (1998, Spring). "How much is purchasing power parity worth?" Quarterly Journal of Business and Economics, 47 (2) 49(14).

Parru, Robert T. (1996, January). Monetary policy in a dynamic, global environment. Business Economics, 31 (1), 18(3).

Spiro, Peter. (1991, October). "Forecasts of the canada U.S. XR: Efficient markets versus purchasing power parity". Business Economics, 26 (4) 56(6).

Wallace, Jeffrey. (1998, Nov/Dec). Best practices in foreign exchange risk management. TMA Journal.

Ziets, Joachim. "A Note on Tests of Efficient Market Hypothesis: The Case of the Forward XR". Atlantic Economic Journal, December 199, v23, n4 p310(8).

APPENDIX: THE DATA

TABLE 1
Raw Data for Canada

Year    Xr     Int     Gdp     Inf

1980   1.19   17.26    1.50   10.20
1981   1.19   14.66    3.70   12.50
1982   1.23   10.26   -3.20   10.80
1983   1.24   10.04    3.20    5.80
1984   1.32   10.16    6.30    4.30
1985   1.40    9.49    4.80    4.00
1986   1.38    8.49    3.30    4.20
1987   1.30    8.66    4.30    4.40
1988   1.19   11.17    4.90    4.00
1989   1.16   12.47    2.40    5.00
1990   1.16   11.78   -0.20    4.80
1991   1.15    7.67   -1.70    5.60
1992   1.21    5.69    0.71    1.50
1993   1.29    4.90    3.20    3.30
1994   1.37    5.54    3.90    5.40
1995   1.37    6.89    2.30    2.10
1996   1.36    4.53    1.20    1.40
1997   1.38    3.52    3.80    1.60
TABLE 2.
Raw Data for France

Year    Xr     Int     Gdp     Inf

1980   4.23    9.50    1.60   13.30
1981   5.43    9.50    1.20   13.40
1982   6.57    9.50    2.50   11.80
1983   7.62    9.50    0.70    9.60
1984   8.74    9.50    2.80    7.40
1985   8.99    9.50    2.30    5.80
1986   6.93    9.50    2.30    2.50
1987   6.01    9.50    1.40    3.30
1988   5.96    9.50    3.60    2.70
1989   6.38    9.50    3.70    3.50
1990   5.45    9.50    5.70    3.40
1991   5.64   10.36    1.70    3.10
1992   5.29   10.00    1.20    1.90
1993   5.66    8.15   -1.00    2.10
1994   5.55    7.70    1.80    1.70
1995   4.99    6.66    2.20    1.80
1996   5.12    3.89    1.10    2.00
1997   5.89    3.37    2.30    1.20
TABLE 3
Raw Data for Germany

Year    Xr     Int     Gdp     Inf

1980   1.96    7.50    1.00    5.40
1981   2.25    7.50    0.10    6.30
1982   2.37    5.00   -1.10    5.30
1983   2.72    4.00    1.70    5.30
1984   3.15    4.50    2.80    2.40
1985   2.46    4.00    2.30    2.20
1986   1.94    3.50    2.30   -0.10
1987   1.58    2.50    1.40    0.20
1988   1.78    3.50    3.60    1.30
1989   1.70    6.00    3.70    2.80
1990   1.49    6.00    5.70    2.70
1991   1.66   11.31    4.50
1992   1.56   12.65    1.60    2.50
1993   1.65   10.16   -1.10    2.10
1994   1.62    9.43    2.50    1.20
1995   1.43   10.94    1.90    1.80
1996   1.50   10.02    1.30    1.50
1997   1.73    9.13    2.20    1.80
TABLE 4
Raw Data for UK

Year      Xr     Int     Gdp     Inf

  1980   2.39   15.62   -2.20   18.00
  1981   1.91   13.12   -1.30   11.90
  1982   1.61   11.36    1.80    8.60
  1983   1.45    9.09    3.70    4.60
  1984   1.16    7.62    2.40    5.00
  1985   1.44   10.78    3.80    6.10
  1986   1.47   10.68    4.20    3.40
  1987   1.87    9.66    4.40    4.10
  1988   1.81   10.31    5.20    4.90
  1989   1.61   13.88    2.10    7.80
  1990   1.93   14.68    0.60    9.50
  1991   1.76   11.50   -0.02    6.50
  1992   1.76    9.50   -0.01    4.60
  1993   1.50    5.90    0.02    3.20
  1994   1.53    5.50    0.04    1.60
  1995   1.58    6.70    0.03    2.40
  1996   1.56    6.00    0.02    3.10
  1997   1.64    7.60    0.03    2.60

Syrous K. Kooros is a Professor of Economics at Nicholls State University. Badiollah R. Asrabadi is a Distinguished Service Professor at Nicholls State University.