Is the Effect of the Exchange Rate on Stock Prices Symmetric or Asymmetric? Evidence from Sudan

This study investigates asymmetry in the effect of the exchange rate on the Sudanese stock market prices. We applied the nonlinear autoregressive distributed lag (ARDL) model by Shin et al. (2014) to monthly data for the period from September 2003 to September 2019, using inflation, money supply, and Murabaha profit margin as control variables. No study found that test the nonlinearity effect of the exchange rate on stock prices in Sudan. This study proposed to fill this gap by examining the impact of the exchange rate of Sudanese Pound nonlinearity on the stock prices in the Khartoum Stock Exchange. The results show that the exchange rate has asymmetric effects on stock prices in both the short run and long run. The policy implication of this paper is that modeling the exchange rate and stock prices symmetrically may affect negatively the effectiveness of economic planning. Thus, nonlinear autoregressive distributed lag emerges as a more suitable model than the ARDL model for investigating such a relationship.


INTRODUCTION
Investigating the relationship between exchange rate and stock prices is of great importance to researchers and policymakers in the developed and developing economies. Research on this relation includes a large number of studies that investigate the link for different countries since the early 1970s. Researching this link has gained substantial consideration with the development of financial markets, the appearance of flexible currency market policies, and the new systems of foreign exchange.
There are two theoretical approaches to test the relationship between exchange rate and stock prices. That is, the flow-oriented model and the stock-oriented model. While the former assumes that the casualty runs from the exchange rate to stock prices, the latter assumes the opposite. The majority of studies investigated this relation used symmetric models, such as Ajaz et al. (2017), Luqman and Rehana (2018), Kwofie and Richard (2018), Bala and Hassan (2018). With the increasing recognition of nonlinearity in the exchange rate effect recently, there are some studies based on asymmetric models, such as Shin et al. (2014), Cheah et al. (2017), Ahmet et al. (2018), Oyinlola and Tirimisyu (2018), Benli et al. (2019), Yacouba and Halil (2019), Habibi and Chin (2019).
Sudan, during the last decades suffering from a massive deterioration in the value of the domestic currency, so it is vital to investigate the interrelation between the exchange rate and the other macroeconomic indicators to establish an effective economic policy in Sudan. However, unfortunately, a few studies have been empirically analyzed for the Sudanese economy to investigate such interrelation, and they are Abdalla (2013), Amin et al. (2014), Abdalla (2017), Dawai and Saeed (2017).
The objective of this paper is to determine whether changes in the nominal exchange rate of the Sudanese Pound have asymmetric or symmetric effects on the stock prices in the Khartoum Stock Exchange. For this purpose, we use monthly data from Sudan and employ the nonlinear autoregressive distributed lag (NARDL) This Journal is licensed under a Creative Commons Attribution 4.0 International License approach of Shin et al. (2014) beside the linear autoregressive distributed lag (ARDL) approach of Pesaran et al.'s (2001). The main contribution of this study is that the change in the nominal exchange rate is decomposed into the partial sum of positive changes and negative changes to determine whether the changes in exchange rates have symmetric or asymmetric effects on stock prices. To the best of my knowledge, similar studies are absent in the literature about Sudan.
We organize the remaining of our paper as follows; we review the related literature in section (2), outline our model and explain the methodology in Section (3). The results reported in Section (4), where we analyzed to reveal that exchange rate changes have asymmetric effects on stock prices. In section (5), we conclude the results and analysis for the sake of researchers and policymakers.

LITERATURE REVIEW
In the related literature, the interaction between stock market prices and exchange rate changes usually explained by two models. The flow-oriented model of exchange rate determination which assumes that the currency changes affect the real income and output of an economy through international competitiveness and the balance of trade position. Consequently, this will affects current and future cash flows of companies and their stock prices (Mundell, 1963(Mundell, , 1964Dornbusch andFisher, 1980, andGavin, 1989). The second model is the stock-oriented model, which assumes that causality will run from stock prices to exchange rate changes. If there is an increase in the stock price, it will lead to a rise in the wealth of the country, which, in turn, will increase demand for domestic currency, and currency will appreciate as a result (Branson, 1983and Frankel, 1983. Other theories used to explain the impact of the exchange rate on stock prices include the Efficient Market Hypothesis, Malkiel and Eugene (1970), and the Arbitrage Pricing Theory, Ross (1976).
Many studies investigated the relationship between stock prices and exchange rates around the world. One of the earliest studies that investigated such a relationship was by Aggarwal (1981). He used an aggregate index of stock prices and an effective exchange rate for monthly data for the period (1974)(1975)(1976)(1977)(1978) for the United States of America; he concluded that the two variables have a positive correlation. Soenen and Hennigar (1988) reached to findings opposite to the results by Aggarwal (1981). Bahmani-Oskooee and Sohrabian (1992) used monthly data for the period (1973)(1974)(1975)(1976)(1977)(1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988) and by using the Granger causality test and cointegration method. They found causality between stock prices and the effective rate of the United States of America in the short run and no relationship between the variables in the long run.
The majority of studies in the 1980s that investigate the relationship between exchange rate and stock prices focused on the determination of the direction of causality. While some investigate the causality between stock prices and exchange rates, some try to distinguish the long-run relationship from short-run effects by establishing cointegration between the two variables.
Going on the same lines of our analysis, we divide the literature broadly into two sub-groups. Firstly, studies assume an asymmetric relationship between exchange rate and stock prices. We review some of the studies that used Linear ARDL of Pesaran et al. (2001) in the Table 1. Table 1 have all assumed that the effects of exchange rate changes on stock prices are symmetric, that is to say, if depreciation has a negative impact on stock prices, appreciation has a positive effect. Secondly, studies which assume an asymmetric relationship between exchange rate and stock prices. In the Table 2, we review some recent studies from the literature that demonstrate how we can introduce asymmetry and run the test using nonlinear ARDL of Shin et al. (2014).

Studies reviewed and listed in
In the Table 3, we review some studies that carried out in Sudan.
No study was carried out to test the nonlinearity effect of the exchange rate on stock prices in Sudan. This study proposed to fill this gap by examining the impact of the exchange rate in Sudanese Pound nonlinearity on the stock prices in the Khartoum Stock Exchange.

METHODOLOGY
The study will employ monthly secondary data covering the period from September 2003 to September 2019, sourced from the Central Bank of Sudan, Khartoum stock exchange, and National Bureau of Statistics. To accomplish the study objective, we use Linear ARDL and NARDL models. We will start with linear error-correction: Where KSE denotes the stock prices in Khartoum financial market, EX is the official monthly nominal exchange rate per USD. By definition, a positive change in the exchange rate will mean depreciation. In contrast, a negative change will imply appreciation, PMP denotes Morabaha profit margin, INF is the level of inflation, and M2 is a measure of the nominal money supply. To account for the effects of other variables, we employ this multivariate model, which contains, besides the exchange rate, variables that affect stock prices, as mentioned by the literature on this topic. We intended to add a proxy for measuring economic activity to the model (following Bahmani-Oskooee and Saha (2015a). But, unfortunately, we could not add because of the unavailability of the monthly data in Sudan.
As found in the literature, the response of stock prices to changes in the exchange rate could be positive or negative, which depends,  as Bahmani-Oskooee and Saha (2015b) illustrated, on whether the firms in the country are export or import oriented. Since the firms in Sudan are import oriented, we expect that due to depreciation in the Sudanese Pound, the stock prices in Khartoum Financial Market will decline.
Regarding the relationship between stock prices and inflation, according to Anari and Kolari (2001), while in the short-run, there is a negative correlation between stock prices and inflation, this relation could be positive in the long run. Since usually, an increase in the money supply leads to a rise in inflation, we expect the same relationship between stock prices and money supply. To determine the response of stock prices to change in the Murabaha profit margin, we follow Amin et al. (2014). They found a negative correlation between stock prices and the Murabaha profit margin In Sudan. bounds testing approach to equation (3) so that we can judge short run symmetry or asymmetry as well as long run symmetry or asymmetry which is labeled non-linear ARDL model.

EMPIRICAL RESULTS AND ANALYSIS
This study tries to investigate the relationship between the nominal exchange rate and stock prices from the asymmetric perspective. We difficultly obtained monthly data from Sudan covering the period from September 2003 to September 2019 using linear ARDL and nonlinear ARDL.
Conducting unit root tests for the variables is the first task of doing this analysis. We tested stationary at the level and first difference using the Augmented Dickey-Fuller (ADF) test, putting in mind that the cointegration test procedure requires I (0) or I (1) variable only. We divided the results report and analysis into three divisions; firstly, we will outline unit root test results, then we will proceed to report linear ARDL estimates, and lastly, we will report nonlinear ARDL estimates.

Unit Root Test Results using ADF
The results presented in Tables 4 and 5 and the findings of the ADF test indicate that all variables in the model satisfied the required condition.
Using Akaike's information criterion (AIC) to select an optimum specification, we Impose a maximum of twelve lags on each first differenced variable. We estimated the model outlined in the previous section using ARDL and NARDL. The results reported in the following tables.

Estimates of the Linear Model
We will report here the results obtained by using the linear ARDL, which include both short and long-run results in addition to their diagnostic statistics.
The CUSUM test is plotted against the breakpoints. If the plot of CUSUM statistic stays within a 5% significance level, then estimated coefficients are said to be stable (Figure 1).  In the diagnostic statistics (reported in Tables 8), a significant negative coefficient obtained for ECM t−1 that supports the existence of cointegration in the long run. The size of the coefficient itself implies that 22% of adjustment takes place within one month. Some other diagnostics also reported. We found that the Lagrange multiplier (LM) insignificant at a 5% level of significance, which implies the absence of serial correlation problems. We test Ramsey's RESET statistic to judge misspecification. Given its t-value of 0.3004, the RESET statistic is insignificant, supporting the excellent specification of the model. Depending on the CUSUM test, we found stable estimates. We also calculate Adjusted R2 to judge the goodness of fit, which found to be 56%. To conclude this section, we note that the exchange rate has short-run but not long-run effects on stock prices in Sudan. We have to compare these results with that obtained using the non-linear ARDL to attain the objective of this study.

Estimates of the Nonlinear Model
We will report here the results obtained by using the nonlinear ARDL, which include both short and long-run results in addition to their diagnostic statistics.
The CUSUM test state that If the plot of CUSUM statistic stays within a 5% significance level, then estimated coefficients are stable ( Figure 2).
When we rely on nonlinear ARDL to investigate the relationship between LnEX and LnKSE in Sudan, the short-run results (reported in Tables 9), indicate that only changes in the LnINF have significant effects on the stock prices at significance level 5%, While LnEX-POS (currency appreciation) and LnMPM have significant effects on the stock prices through their lags. Nevertheless, LnEX-NEG (currency depreciation) and LnM2 have insignificant effects. This result implies that the exchange rate (LnEX) changes have an asymmetric effect on stock prices (LnKSE) in Sudan in the short run.
Regarding long run results (reported in Table 10), we must first establish cointegration using the bound test. Our calculated F statistic (8.093689) is higher than the upper bound critical value of 4.01 at all levels of significance, implying the existing of cointegration. However, in the long run, it clears that LnEX-NEG (currency depreciation) has significant effects on LnKSE while the other variables in the model, including LnEX-POS (currency appreciation), have insignificant effects on LnKSE.
To conclude, LnEX-POS (currency appreciation) is insignificant and has a negative sign. At the same time, LnEX-NEG (currency   *Indicates that the parameter is significant at 1%. **Indicates that the parameter is significant at 5% Figure 1: CUSUM test depreciation) is the significance with a positive sign, which again implies that exchange rate (LnEX) changes have an asymmetric effect on stock prices (LnKSE) in Sudan in the long run also.
In the diagnostic statistic (reported in Table 11), a significant negative coefficient obtained for ECM t−1 that supports the existence of cointegration in the long run. The size of the coefficient itself implies that 20% of adjustment takes place within one month. We also reported some other diagnostics. We found that the LM insignificant at a 5% level of significance, which implies the absence of serial correlation problems. We test Ramsey's RESET statistic to judge misspecification. Given its t-value of (2.17091), the RESET statistic is significant, supporting the misspecification of the model. Depending on the CUSUM test, we found stable estimates. We calculate the adjusted R 2 to judge the goodness of fit, which found to be 61%. To conclude this section, we note that the exchange rate has short-run but not long-run effects on stock prices in Sudan.

CONCLUSION
This study employs Linear ARDL and Nonlinear ARDL models to analyze both symmetric and asymmetric effects in the short run and long run of exchange rate on stock prices in the Khartoum  . These values usually used to judge the significance of ECM t−1 . * , ** and *** denote significance at the 1%, 5% and 10% levels, respectively  . We use these values to judge the significance of ECM t−1 . * , ** and *** denote significance at the 1%, 5% and 10% levels, respectively stock exchange. Our results show that the exchange rate has a asymmetric impact on stock prices in both the short run and long run. Hence, modeling the exchange rate and stock prices as a symmetric relation may lead to uneffective economic plans. Thus, NARDL emerges as a more suitable model than the ARDL model for investigating such a relationship.