Improving the Forecast Accuracy of Oil-Exchange Rate Nexus in GCC Countries

This paper renders new evidence on the predictability of GCC dollar exchange rates using crude oil prices relying on the approach of Westerlund [1] [2] that accounts for salient features of the predictor. The results show the presence of significant in-sample predictability of exchange rates using crude oil prices (Brent and WTI prices) across the GCC countries. The results of forecast evaluation based on the root mean square error (RMSE), Campbell-Thompson (C-T) statistic and Diebold-Mariano (D-M) statistic are rather mixed. The superior forecast performance of the oil-based exchange rate model is highly sensitive to the choice of benchmark time-series models. We, however, conclude the overwhelming forecast performance of time-series models (namely, AR, ARMA, and ARFIMA) over our oil-based exchange rate model in predicting exchange rates across the GCC region.


Introduction
Many economies, developing and developed, depend on oil for a variety of needs especially in the production of many goods and services [3]. To further stress the importance of oil, fluctuations in oil prices have been linked to many economic challenges such as economic recessions, trade deficits, high inflation, low values for stocks and bonds and high uncertainty for investment [4] [5] [6]. In this light, a number of studies have highlighted how changes in oil price influence these macroeconomic variables [3]. Generally, empirical evidences that establish the relationship between oil price and exchange rate are less substantial and as well mixed. From the scanty literature, a shred of evidence suggests that movements in oil prices determine the value of a currency [3] [7]- [15]. Given the shred of evidence explaining the dynamics of oil price-exchange rate relationship, we explore this nexus for the GCC countries. There is a dearth of studies in this regard on the predictability of exchange rate with oil price for oil-exporting countries in general and the GCC in particular despite their high dependence on crude oil. This serves as a motivation for the present study.
Thus, the present study contributes to the literature in the following ways.
First, it evaluates both the in-sample and out-of-sample forecast performance of oil-based exchange rate model relative to time series models. Second, it accounts for some important features of oil price which may have implications on its forecast performance. Consequently, the approach of Westerlund [1] [2], which accommodates salient features such as endogeneity, persistence and conditional heteroscedasticity in the predictors of a series, is employed. The choice of GCC countries in this study is deliberate. Unlike other oil-exporting nations, these countries share similar characteristics given the increasing economic integration among them (an instance is the adoption of pegged exchange rates against the US dollars).
Following this section, the rest of the paper is structured as follows. The next section provides a brief review of the literature followed by Section 3, which presents the predictive model for estimation and the underlying forecasting procedures. Section 4 contains preliminary analyses of data features. Section 5 discusses the results while Section 6 concludes the paper.

Literature Review
The literature has widely taken care of the relationship between exchange rate and oil prices. It must be re-stated from the beginning of this review that most studies are impact analysis focus; very few consider exchange rate forecast through oil price, most especially in the GCC countries. The study of the relationship received much more attention for the United States dollar and currencies of industrial and developed economies, along with OPEC and other oil exporting countries. On the results discovered; the bulk of the studies attesting to the relationship between oil prices and exchange rates argue that movements in oil prices determine the value of a currency. For example, while studying real oil prices in the post Bretton Woods era and their relationship with 16 OECD countries' real exchange rates, Chaudhuri [7] among others found that, in 13 of the 16 economies studied, co-integration is discovered between the two series and oil price volatility cause movements in real exchange rate. In support of this discovery, Nikbakht [13] conducted panel analysis of 7 OPEC countries employing monthly data spanning the years 2000 to 2007. The research applies co-integration analysis also and discovered evidences that real oil prices drives fluctuations in the real exchange rates, confirming a long run link between the two for pooled series. VARs and ECMs have been adopted in the long run impacts study; many other approaches were also adopted such as that found in [7]. According to Amin [14], in the G7 countries, most results point out that oil  [15] confirm that the relationship between oil prices and nominal exchange rates assumes more significant pose after the financial crises of 2008 for some selected emerging countries. The mixed nature of the various results discovered is no more news, it is important to point out from here that while some impact analyses find positive relationship, some are negative. A group claims that a rise in oil prices leads to the appreciation of the currency under study. For instance, Amano and Norden [16] study Germany, Japan and the United States and confirm the major findings of Chaudhuri [7] and Nikbakht [13]. It is also contended that in the long run a rise in the price of oil will result in the real appreciation of the US dollar against the currencies of 15 other industrial countries post the Bretton Woods era. In agreement to this, the economies of the United States, Eurozone, OPEC and China are analysed byBénassy-Quéré, et al. [17]. Based on their analysis, in the long run, a 10 percent rise in the price of oil spurs about 4.3 percent appreciation of the US dollar, with a sluggish return of the US dollar exchange rate to its long run equilibrium value. For the GCC however, Alotaibi [18] concludes that positive oil price shocks dominate currency movements continuously in all GCC countries except in UAE and Qatar, where demand shocks are more persistent.
Coudert, et al. [10] also corroborated by deducing that an increase in oil prices promotes real appreciation of the oil exporter's exchange currency in the long run and that to a large extent, pegged currencies maintain the behavior of their anchors. Likewise, using four-variable structural VAR models, Korhonen, et al. [11] argue that a positive oil shock would cause currency appreciation against the USD.
For the other side of coin however, another group of researchers claims that a rise in oil prices would actually worsen the value of a currency leading to its depreciation against other currencies. For instance, as claimed by Akram [8] increasing oil prices, in a negative non-linear way, affect nominal exchange rates in the short run only, and claims that the strength of this effect depends on the level and trend of oil prices. Akram [8] finds the link to be insignificant in the long run. Comparatively, Trygubenko [9] employs a number of empirical models and concludes that rising oil prices significantly depreciate the USD in the short run. Like Akram [8] however, he finds no relation between the two variables in the long run. Concurring to this, Al-Mulali, et al. [19] emphasized a long run relationship between the UAE dirham's real exchange rate and oil prices but while recognizing the impact of positive oil price shocks on domestic prices, the authors conclude that a 1 percent increase in oil prices causes 0.16 percent depreciation in the real value of the dirham.
A major gap in the literature relates to the fact that virtually all the known studies involve impact (in-sample) analyses while the issue of predictability between oil price and exchange rate has received very little attention. This is the

Estimation Approach
We begin our methodology by specifying a bivariate single predictive model where crude oil price is hypothesized as a predictor of exchange rate: where t s is the log of dollar exchange rate for each of the six GCC countries' currencies involving Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirate; and t p is the log of crude oil prices where Brent price and WTI price are used separately in the estimation process. Thus, we have two predictive models for each of the oil price series across the six GCC countries considered (details about the data utilized are provided in the section that follows). The t ε is zero mean idiosyncratic error term on exchange rate and the coefficient λ measures the relative impact of crude oil prices on exchange rate and the underlying null hypothesis of no predictability is that 0 λ = .
In order to resolve any probable endogeneity bias resulting from the correlation between t p and t ε as well as any potential persistence effect, we follow the approach of Lewellen [20] and Westerlund [1] [2]. This bias is not unexpected since a single predictor (oil price) is accounted for in the predictive model for exchange rate as in Equation (1) whereas in reality there are several predictors that influence the latter which are not captured in the analyses. Excluding these variables will bias the regression estimates [20]. In addition, oil price is recently found to be endogenous as it responds to the interplay between supply shock (due to global oil supply) and demand shock (due to the real level of economic activity) [21] and therefore correcting for this inherent endogeneity becomes important in the estimation process. In fact, Lewellen [20] finds that ignoring such bias has implications on the forecast results. The underlying predictive model that accounts for these effects can be specified as: where the parameter , the (C-T hereafter) test [22] and the test (D-M hereafter) [23]. The C-T test statistic is computed as the mean square error (MSE) obtained from the restricted and the unrestricted models, respectively. In the present case, our oil-based exchange rate model in Equation (1) stands as the unrestricted model, whereas time-series models including AR(1), ARMA (1, 1) and ARFIMA (1, d, 1), where "d" is the order of integration which is neither zero nor unity. A positive value of the statistic implies that our oil-based exchange rate model is preferred to the time-series models in predicting exchange rates; otherwise, it does not. By implication, a positive C-T statistic obtains from the fact that RMSE associated with our predictive model is less than that associated with the time-series models; but the reverse is the case for a negative C-T statistic. The D-M test is also used as a complementary test and it tests whether the difference between the forecast errors of two competing predictive models is statistically significant (or different from zero).

While the D-M test is not suitable for small samples (which is not a concern
given the large samples used for analyses), the test is however valid when the forecast errors are found to be non-Gaussian, nonzero mean, serially correlated, and contemporaneously correlated. A negative value and statistical significance of the D-M statistic at the conventional levels of 1%, 5% and 10% imply that our oil-based exchange rate model significantly outperforms the time-series models; otherwise, it does not.

Data Description and Source
We focus attention on the foreign exchange markets of the six GCC countries,

Autocorrelation and Heteroscedasticity Test Results
Here, we conduct autocorrelation and heteroscedasticity tests using Ljung-Box test Q-statistics and Autoregressive conditional heteroscedasticity lagrangian multiplier (ARCH-LM) test F-statistics, respectively (see Table 1) over the full

The Persistence and Endogeneity Test Results
To further strengthen the choice of estimator, we test for persistence and endogeneity in the predictors, which comprise crude oil prices (Brent and WTI prices) in this case (see Table 2) over the full sample period. The persistence test has the null hypothesis of no persistence effect in the predictors. The coefficient of the AR(1) process was estimated for each predictor using OLS estimator and the results were found to be close or equal to one which is often the features of series with higher order of integration, thus, suggesting that the predictors (crude oil prices) contain persistent effects. We, however, observe that our predictors are

In-Sample Predictability Results: Do Oil Prices Matter in Exchange Rate Behaviour?
The predictability power of a potential economic predictor hinges on the statistical significance of the first-order autoregressive coefficient in the theoretical (predictive) model at the conventional levels of significance, namely, 1%, 5%, and 10%. It can be observed that irrespective of measures of oil price series (Brent and WTI prices), the null hypothesis of no predictability is rejected at 1% level of significance (see Table 3). We, therefore, conclude that crude oil prices play a significant role in predicting the behaviour of dollar exchange rates across the entire GCC countries. Our result affirms the previous findings of [12] 1 and [25]. We also establish a negative linkage between crude oil prices (Brent and WTI) and dollar exchange rates across the entire GCC region. This supports the findings of [11], [12], and [15] that higher oil price leads to appreciation of the net oil-exporting currencies against the US dollar. 1 Lizardo, et al. [12] Table 3. In-sample predictability of exchange rates using oil prices (Brent and WTI prices).
where δ denotes the coefficient on the predictor z, which in this case stands for crude oil prices. We employ both Brent and WTI prices as proxies for crude oil prices. ***implies the rejection of the null hypothesis of no predictability at 1% level of significance. The values in parentheses are the standard errors associated with the first-order autoregressive coefficients in our predictive model. Here, we consider 75% of the full sample data.

Forecast Evaluation: Can Oil-Based Exchange Rate Model Beat Time Series Models?
We further compare the in-sample and out-of-sample forecast performance of our oil-based exchange rate model with three time-series models including AR, ARMA, and ARFIMA using the RMSE, the C-T and the D-M statistics (see Tables 4-9). Our results are however mixed. For Kuwait and Saudi Arabia, we find that our oil-based predictive exchange rate model significantly outperforms the time series models both in-sample and out-of-sample, irrespective of the choice of oil price series and the choice of benchmark time-series models 2 . This conclusion is reached on the basis of positive C-T statistics (see Table 6 and Table 7), negative and significant D-M statistics (see Table 8 and Table 9), with the RMSE associated with our predictive model being smaller than that of the time-series models (compare Table 4 and Table 5).
We however establish an opposing conclusion of the superior forecast performance of the time-series models (AR, ARMA and ARFIMA) over our predictive exchange rate model in the cases of Bahrain, Qatar, and the United Arab Emirate both in-sample and out-of-sample, irrespective of the choice of oil price series, and the choice of benchmark time-series models. This conclusion is reached on the basis of negative C-T statistics (see Table 6 and Table 7), positive and significant D-M statistics (see Table 8 and Table 9), with the RMSE associated with our predictive model being greater than that of the time-series models (compare Table 4 and Table 5). In addition, our results show that Oman's dollar exchange rate behaviour is highly sensitive to the choice of benchmark time-series models: while AR and ARMA models predict the dollar exchange rate better than our predictive model, the reverse is the case for the ARFIMA model using both in-sample and out-of-sample periods. 2 This result parallels the findings of [8]. Note: Capturing 75% of the full sample, we evaluate the in-sample and out-of-sample forecast performance (using 30and 60 days as the forecast horizons) of our predictive model, which in this case is the oil-based exchange rate model (using Brent and WTI prices) with the aid of root mean square error (RMSE). The smaller the root mean square error (RMSE), the greater the predictive power of a model and vice versa.  On the whole, we conclude the overwhelming predictive power of time-series models over our oil-based exchange rate model across the GCC region. That time-series models (AR, ARMA, and ARFIMA), in the majority of cases, predict exchange rates better than our oil-based exchange rate model can do is reflected in the predictability graphs associated with both models (Compare Figure 1 and     Note: The Campbell-Thompson (C-T) test statistics as used here compares the unrestricted model, which in this case is the oil-based exchange rate model (using WTI price) with the time-series models (AR, ARMA, and ARFIMA), which constitute the class of restricted models. Positive C-T stat implies that the oil-based exchange rate model (using WTI price) is preferred to AR, ARMA, and ARFIMA models in predicting exchange rates using the in-sample data covering 75% of the full sample and the out-of-sample forecast horizons of 30 and 60 days. On the other hand, negative C-T stat implies that AR, ARMA, and ARFIMA models are preferred to the oil-based exchange rate model (using WTI price) in predicting exchange rates using the in-sample data covering 75% of the full sample the out-of-sample forecast horizons of 30 and 60 days.   and the restricted model comprising the time-series models (AR, ARMA, and ARFIMA). The negative and statistical significance at 1% (***), 5% (**) and 10% (*) implies that the oil-based exchange rate model (using Brent price) significantly outperforms the AR, ARMA, and ARFIMA models using in-sample data covering 75% of the full sample and out-of-sample forecast horizons of 30 and 60 days. However, the positive and statistical significance at 1% (***), 5% (**) and 10% (*) implies that the AR, ARMA, and ARFIMA models significantly outperform the oil-based exchange rate model (using Brent price) using in-sample data covering 75% of the full sample and out-of-sample forecast horizons of 30 and 60 days. in the region, and as noted by Amin [14], the inability of the GCC nominal exchange rates, which are pegged completely or partially to the US dollar, to adjust to the oil price shocks through appreciation or depreciation means that the impact would be transferred to the GCC economies in the form of domestic inflation and higher prices relative to the foreign prices, with a large resultant effect on the GCC real exchange rates.

Conclusions
We offer new evidence on the predictability of exchange rates using crude oil prices, namely Brent and WTI prices, across the six GCC countries comprising Bahrain, Kuwait, Oman, Qatar, Saudi Arabia and the United Arab Emirate. Driven by the need to account for some salient features usually present in high frequency time-series data, we employ the estimator proposed by Lewellen [20] and Westerlund, et al. [1] [2] in order to account for possible persistence, endogeneity, serial correlation, and conditional heteroscedasticity effects in our predictors investigate if accounting for structural breaks and asymmetries in oil-exchange rate nexus would improve the predictability of exchange rates using crude oil prices. A considerable attention can also be drawn towards the use of real exchange rate, which is a measure of a country's international competitiveness in the foreign market.

Conflicts of Interest
The author declares no conflicts of interest regarding the publication of this paper.