EFFECT OF SELECTED INSECTICIDE ON WHITEFLY (Bemisia tabaci) INFESTING BRINJAL PLANTS 113 THE FUTURE OF SAUDI ARABIA’S DATE EXPORTS USING A COINTEGRATION MODEL Mohamad Alnafissaa Mahmoud Alderinyb Yosef Alamric Jawad Alhashimd a,b,c,dDepartment of Agricultural Economics, College of Food and Agricultural Sciences King Saud University, Saudi Arabia.  malnafissa@ksu.edu.sa (Corresponding author) Article History ABSTRACT Received: 6 January 2021 Revised: 29 January 2021 Accepted: 15 February 2021 Published: 3 March 2021 Keywords Agricultural marketing Future structure market Export forecasting Dates market Comparative advantage ARDL model Cointegration model Saudi Arabia. Dates represent an important export crop for Saudi Arabia that attracts foreign exchange. There is an opportunity for Saudi Arabia to extend its date exports to the international market due to production exceeding consumption and comparative advantage in date production. This paper explores the future export market for Saudi dates by analyzing the factors that affect Saudi date exports and highlighting this crop’s comparative advantage. Also, the paper uses autoregressive and distributed lag methodology with data from 1980 to 2017 to predict the future for exported dates between 2020 and 2025. The relationship between the export of dates and other exogenous variables shows there are significant effects from domestic production, domestic consumption of dates, and the price of exported dates in both the long and short term. In contrast, the comparative advantage of Saudi dates shows a significant effect only in the short term, because there has been no active program to promote Saudi dates in global markets. The explanatory variables predict that 2019 Saudi date exports will increase by 3.8-fold in 2025, with the total export amount over 700 thousand tons. Thus, there is an opportunity to develop more programs that target the international market in support of date farms and exporters, and to support market research directed at satisfying the international market. Contribution/Originality: This study analyzed the future structure market of date export and highlights the comparative advantage of this crop for Saudi Arabia. It then predicts the future for date exports between 2020 and 2025. DOI: 10.18488/journal.ajard.2021.111.113.119 ISSN(P): 2304-1455/ ISSN(E): 2224-4433 How to cite: Mohamad Alnafissa --- Mahmoud Alderiny --- Yosef Alamri --- Jawad Alhashim (2021). The Future of Saudi Arabia’s Date Exports Using a Cointegration Model. Asian Journal of Agriculture and Rural Development, 11(1), 113-119. 10.18488/journal.ajard.2021.111.113.119 © 2021 Asian Economic and Social Society. All rights reserved. 1. INTRODUCTION Dates are one of Saudi Arabia’s major food crops and thus represent an important export crop attracting foreign exchange. Furthermore, Saudi Arabia has a comparative advantage in date production, which makes it worthwhile to study Saudi date exports to various importing markets. Saudi Arabia’s average date exports represented 5% of total agricultural exports during the period 1990–2016 (FAOSTAT, 2020). Asian Journal of Agriculture and Rural Development Volume 11, Issue 1 (2021): 113-119 http://www.aessweb.com/journals/5005 https://orcid.org/0000-0003-2234-531X https://orcid.org/0000-0002-5534-5052 https://orcid.org/0000-0001-5023-8998 https://orcid.org/0000-0002-1618-7119 mailto:malnafissa@ksu.edu.sa http://crossmark.crossref.org/dialog/?doi=10.18488/journal.ajard.2021.111.113.119 http://www.aessweb.com/journals/5005 Asian Journal of Agriculture and Rural Development, 11(3)2021: 113-119 114 Al-Abdulkader, Al-Kahtani, Ismaiel, Elhendi, and SaadIi (2016) showed that increasing the marketing efficiency of Saudi dates will lead to an increase of up to 30% of return for the agricultural sector. Therefore, the Saudi government intends to use dates as one of its strategic agricultural commodities in global markets due to its comparative advantage in date production. Because date production is increasing year by year, with an average annual rate of growth of 4.6 tons per year during the years 2000–2018, increase in production must be accompanied by a competitive ability to access external markets. The economic importance of date production and the market environment for dates has recently resulted in an extensive literature that discusses different marketing aspects and issues. For instance, Dawood and Kinnucan (2018) estimated the world market demand for dates and price elasticity. Intezar, Abdallah, and Rao (2016) studied the opportunities and difficulties for farmers and traders in marketing dates, while Dhehibi, El Awar, Ben-Salah, and Aw- Hassan (2018) studied competitiveness in the date trade for Gulf Cooperation Countries (GCC). Also, value chain analysis and marketing opportunities for date palms in GCC countries were studied by Dhehibi., Salah, and Frija (2018) while the efficiency of date marketing was investigated by Ismaiel, Al-Abdulkader, Al-Kahtani, and Saad (2017), Alabdulkader, Elhendy, Al Kahtani, and Ismail (2017) and Al-Abdulkader et al. (2016). Al-Shreed et al. (2012) investigated export markets for Saudi dates and potential opportunities. The comparative competitive advantage for the export of dates was studied by Ali, Fahad, and El-Habbab (2014). Ali, Al-Mulhim, and El-Habbab (2014) revealed the demand for the export of Saudi dates. This research extends previous literature exploring the future export market for Saudi dates. The study analyzes factors that impact Saudi date exports and highlights the comparative advantage of this crop. It then predicts the future for date exports in the period 2020–2025. 2. METHODOLOGY Engle and Granger (1987), Johansen (1988) and Johansen and Juselius (1990) developed methods used to test cointegration, assuming the variables are stationery to the same degree. However, this assumption is not achieved in many applied aspects because the variables could integrate to different degrees. In this case, it is challenging to perform the analysis using these methods. Therefore, Pesaran, Shin, and Smith (2001) provided an appropriate model for treating this problem – the autoregressive and distributed lag model (ARDL) (Nkoro & Uko, 2016). The main concept of the methodology is to combine the autoregressive and distributed lag models to create a time series model. The dependent variable is a function of its lagged values and the values of the explained variables in the current and lag periods. Using ARDL methodology for the cointegration test has some advantages: First, determining the possibility of applying the model regardless of whether the variables under study are degree I(0), degree I(1), or both. They are used when the integration order is unknown or not uniform for all variables. The only condition for applying this test is that none of the variables can have an I(2) degree of integration. Second, it can be used if the sample size is small, which contrasts with most traditional cointegration tests that require the sample size to be large for efficient results. Third, the application of this methodology allows long- and short-run relationships to be estimated together using one equation. Assuming that Yt expresses the dependent variable in the period and that X1t, X2t, ..., Xkt represent k external explanatory variables in the period t, the relationship between the dependent variable and the set of explanatory variables is expressed in a multiple linear regression model: 𝑌𝑡 = 𝛽0 + 𝛽1𝑋1𝑡 + 𝛽2𝑋2𝑡 … + 𝛽𝑘𝑋𝑘𝑡 + 휀𝑡 (1) Equation 1 shows the estimation of a long-run relationship, while ARDL is applied to test the cointegration through Pesaran et al. (2001). According to ARDL methodology, the analysis is carried out in three steps: 1. A cointegration test using the unrestricted error correction model (UECM), this model is expressed as ∆𝑌𝑡 = 𝛼0 + 𝛾0𝑌𝑡−1 + 𝛾1𝑋1𝑡−1 + ⋯ + 𝛾𝑘𝑋𝑘𝑡−1 + ∑ 𝛼𝑖∆𝑌𝑡−𝑖 𝑝 𝑖=1 + ∑ ∑ 𝛽𝑗𝑙∆𝑋𝑗𝑡−𝑙 𝑞𝑗 𝑙=0 𝑘 𝑗=1 + 휀𝑡 (2) where 𝛾0, 𝛾1, … , 𝛾𝑘 represent long-run relationship coefficients, while (𝛼𝑖 , 𝛽𝑗𝑙), and 𝑙 = 0,1, … , 𝑞𝑗 are short-run relationship coefficients; p expresses the length of the lag period for the dependent variable Y, while 𝑞𝑗 expresses the length of the deceleration period for the interpreted variable 𝑋𝑗 , (𝑗 = 1, … , 𝑘), and the lagged times for the variables do not necessarily have the same number of lags (𝑝 ≠ 𝑞𝑗). The symbol ∆ also indicates the first difference of the variable; 휀𝑡 indicates the random error terms and assumes that its mean is 0, that the variance is constant from one time period to another, and the errors are not correlated, {𝐸(휀𝑡) = 0, 𝐸(휀𝑡 2) = 𝜎𝜀 2, 𝐸(휀𝑡휀𝑡`) = 0 𝑓𝑜𝑟 𝑡 ≠ 𝑡`}. In this model, the presence of a cointegration relationship test determines the extent of the statistical significance of the relationship between the variables in the long term by applying the bounds testing approach to cointegration, which depends on the Wald test. The null hypothesis is 𝐻0: 𝛾0 = 𝛾1 = ⋯ = 𝛾𝑘 = 0 and indicates that there is no cointegration relationship between the variables in Equation 1, whereas the alternative hypothesis is 𝐻1: 𝛾0 ≠ 𝛾1 ≠ ⋯ ≠ 𝛾𝑘 ≠ 0 and denotes the existence of a cointegration relationship between the variables in Equation 1. The F- test statistic depends on whether the variables included in the model integrate to degree I(0) or degree I(1), on the number of independent variables, on whether or not Equation 1 includes an intercept and trend, and on the sample size. Comparing the calculated F value with the tabular values of the lower and upper thresholds provided by Pesaran et al. (2001), a decision is taken regarding the null hypothesis where the lower bound was derived based on the assumption that the variables are integral to degree I(0). In contrast, the upper limits are derived based on the assumption that the variables are integral to degree I(1). If the calculated F value is less than the critical value of the lower limit, the null hypothesis is not rejected. This Asian Journal of Agriculture and Rural Development, 11(3)2021: 113-119 115 indicates that there is no cointegration. However, if the F value is higher than the critical value of the upper limit, the null hypothesis is rejected and cointegration exists between the variables in Equation 1. This means that the model allows us to estimate the long-run relationship among the variables. However, if the calculated F occurs between the lower and higher limits, the result is that the cointegration is not conclusive. 2. Estimating the long-run relationship between the dependent variable Y and the explanatory variables is expressed in the following equation: 𝑌 = ∅0 + ∅1𝑋1 + ⋯ + ∅𝑘𝑋𝑘 (3) where ∅0, ∅1, … , ∅𝑘 represent long-term coefficients, and their estimates derive from the estimates of the long- term transactions shown in Equation 2 as follows: {∅0 = −(𝛼0 𝛾0⁄ ), ∅1 = −(𝛾1 𝛾0⁄ ), … , ∅𝑘 = −(𝛾𝑘 𝛾0⁄ )} . The lagged rank in the ARDL model is chosen by the Akaike information criterion (AIC) or the Schwarz Bayesian Criterion (SBC) standard. However, Pesaran et al. (2001) recommend that the lagged period be a maximum of two periods for annual data. 3. The third step specifies and estimates the dynamic ARDL by constructing an error correction model (ECM). The ARDL model is popular in economic studies of different aspects such as economic growth (Okafor & Shaibu, 2016), inflation forecasting (Ülke, Sahin, & Subasi, 2018), tourism demand forecasting (Zhu, Lim, Xie, & Wu, 2018), and forecasting U.S. commercial property price indices (Van de Minne, Francke, & Geltner, 2018). It is expressed in the following equation: ∆𝑌𝑡 = 𝜋0 + ∑ 𝜋𝑖∆𝑌𝑡−𝑖 𝑝 𝑖=1 + ∑ ∑ 𝛿𝑗𝑙∆𝑋𝑗𝑡−𝑙 𝑞𝑗 𝑙=0 𝑘 𝑗=1 + 𝜓𝐸𝐶𝑇𝑡−1 + 𝑣𝑡 (4) The error correction term is represented by 𝐸𝐶𝑇𝑡−1. This variable represents the long-run relationship between the dependent variable and the independent variables, while its coefficient indicates the speed of adjustment from the short-run equilibrium. The value of the error correction coefficient falls between minus one (−1) and 0, where a large coefficient in absolute value indicates a higher speed of adjustment. If the coefficient has a 0 value, it means there is no long-run relationship. On the other hand, a significant coefficient means there are adjustments between the short run and long run. 3. MODEL SPECIFICATION The study assumes that the future quantity of Saudi date exports, 𝑌𝑡 , is determined by a set of economic factors such as the domestic production of dates, denoted by 𝑋1𝑡 , the amount of domestic consumption of dates, denoted by 𝑋2𝑡 , the average export price 𝑋3𝑡 , and the comparative advantage of date exports with respect other agricultural export products (𝑋4𝑡). The Saudi date export model can be written as 𝑌𝑡 = 𝑓(𝑋1𝑡 , 𝑋2𝑡 , 𝑋3𝑡 , 𝑋4𝑡) (5) 4. DATA This study uses aggregate data for dates, covering the period 1980–2019. The quantity of Saudi date exports and domestic consumption of dates is measured in tons, while the average export price is measured in U.S. dollars per ton. The study used secondary data collected from various sources: (1) The Ministry of Economy and Planning, General Authority for Statistics, and other official authorities in Saudi Arabia; and (2) the United States Department of Agriculture (USDA), U.N. Comtrade Database from United Nations Statistics Division and Food and Agriculture Organization (FAO) websites, as well as the World Development Indicators and the World Data Bank. 5. EMPIRICAL RESULTS 5.1. Order of Integration The study used the augmented Dickey–Fuller (ADF) test to exam the null hypothesis of a unit root. It used the SBC to determine the maximum length of lags and used the Eviews 11 program to obtain the results. Table 1 shows the ADF for the variables with standard variables and logarithmic form. Table-1. Integration results of the study variables when taking into consideration natural and logarithmic values. Variable Normal values Natural logarithm values Integration degree I Type Lags integration degree I Type Lags 𝑌 I(1) Trend 3 I(1) Intercept 2 𝑋1 I(1) Intercept 2 I(1) Intercept 4 𝑋2 I(0) Trend 2 I(1) Intercept 4 𝑋3 I(0) None 5 I(0) None 1 𝑋4 I(1) None 4 I(0) Trend 1 Table 1 shows that the forecast for Saudi date exports (Y), the domestic production of dates (𝑋1), and the comparative advantage of date exports with respect to other agricultural export products (𝑋4) are stationary at the first difference, then integrated at the level 𝑌, 𝑋1, 𝑋4~𝐼(1). On the other hand, the level of consumption of domestic dates (𝑋2) and the average export price (𝑋3) are stationary at 𝑋2, 𝑋3~𝐼(0). As for the variables in their logarithmic form, the result shows that Ln(Y), (Ln 𝑋1), and Ln(𝑋2) are stationary at first differences and I(1), while Ln(𝑋3) and Ln(𝑋4) are stationary at Ln(𝑋3), Ln(𝑋4) ~ I (0). Asian Journal of Agriculture and Rural Development, 11(3)2021: 113-119 116 5.2. Cointegration Test The test of stationarity gives the difference in the degree of integration of the study variables, as well as the absence of variables with the integration of I(2) or higher (Table 1). It is appropriate to follow the ARDL methodology to test cointegration between the variables. In this case, the UECM is used to test the cointegration among the variables. It is described in the following equation: ∆𝐿𝑛(𝑌𝑡) = 𝛼0 + 𝛾0𝐿𝑛(𝑌𝑡−1) + 𝛾1𝐿𝑛(𝑋1𝑡−1) + 𝛾2𝐿𝑛(𝑋2𝑡−1) + 𝛾3𝐿𝑛(𝑋3𝑡−1) + 𝛾4𝐿𝑛(𝑋4𝑡−1) + ∑ 𝛼𝑖∆𝐿𝑛(𝑌𝑡−𝑖) 𝑝 𝑖=1 + ∑ 𝛽1𝑗∆(𝑋1,𝑡−𝑗) 𝑞1 𝑗=0 + ∑ 𝛽2𝑗∆(𝑋2,𝑡−𝑗) 𝑞2 𝑗=0 + ∑ 𝛽3𝑗∆𝐿𝑛(𝑋3,𝑡−𝑗) 𝑞3 𝑗=0 + ∑ 𝛽4𝑗∆(𝑋4,𝑡−𝑗) 𝑞4 𝑗=0 + 휀𝑡 (6) The equation is denoted by 𝐴𝑅𝐷𝐿(𝑝, 𝑞1, 𝑞2, 𝑞3, 𝑞4), and is used to investigate the long-run relationships among variables. The long-run relationship exists if there is cointegration between the dependent variable and independent variables in Equation 6. The null hypothesis assumes cointegration does not exist (null hypothesis is 𝐻0: 𝛾0 = 𝛾1 = 𝛾2 = 𝛾3 = 𝛾4 = 0 ), while the alternative hypothesis ( 𝐻1: 𝛾0 ≠ 𝛾1 ≠ 𝛾2 ≠ 𝛾3 ≠ 𝛾4 ≠ 0 ) indicates there is a cointegrated relationship between the variables. The study used SBC to determine the number of lags by using Microfit 5.0 (a program developed by Pesaran and Pesaran (2009), as well as the results from estimating the UECM model. The result shows the appropriate model is 𝐴𝑅𝐷𝐿(1,0,1,0,1), as shown in Table 2. A lag of one time period has been specified for the dependent variable, logarithm of exports 𝐿𝑛(𝑌), and the independent variables 𝑋4, independent variables 𝑋1, 𝑋2, 𝑎𝑛𝑑 𝐿𝑛(𝑋3) have no lag. Table-2. Results of the estimation parameters of date exports using ARDL. Regressor Coefficient Standard error T-ratio [Prob.] 𝐿𝑛(𝑌𝑡−1) 0.3402 0.0985 3.456[.001] 𝑋1𝑡 0.0104 0.0019 5.580 [.000] 𝑋2𝑡 −0.0102 0.0020 −5.121[.000] 𝐿𝑛(𝑋3𝑡) −0.1348 0.0736 −1.832 [.073] 𝑋4𝑡 0.0020 0.0009 2.353[.023] 𝑋4𝑡−1 −0.0030 0.0006 −4.847 [.000] 𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 6.4489 0.9845 6.551[.000] R2 0.932 F-Stat. F(6,51) 115.81[.000] AIC −9.958 SBC −17.170 DW-statistic 2.114 Durbin’s h-statistic −659[.510] The results show that the R2 is 0.932, which indicates that the independent variables under study explain 93.2% of the variation in the dependent variable. The value of the test 𝐹 − 𝑆𝑡𝑎𝑡 = 115.81 is significant at the 0.0001 level, so the model explains much of the variation in Saudi date exports. Also, the model does not suffer from the problem of autocorrelation between errors, as the value of the modified Durbin’s h-statistic 𝐴𝑑. 𝐷𝑊 = −0.659. The cointegration test is shown in Table 3. The results show that the model rejects the null hypothesis of no cointegration, while it accepts the alternative hypothesis. The F-statistic is 6.406, which is higher than the critical upper limit (4.323) at the 5% level of significance. Thus, there is cointegration among the variables at the 5% level of significance. Table-3. Results of a cointegration test for model variables ARDL (1,0,1,0,1). Test statistics 95% Lower bound 95% Upper bound 90% Lower bound 90% Upper bound F-statistic = 6.406 3.044 4.323 2.578 3.706 W-statistic = 32.029 15.222 21.613 12.892 18.527 5.3. Estimating the Long-Run Relationship In the previous step, the statistical test demonstrated the existence of cointegration among the variables of the export model; thus, there is a long-run relationship among these variables. The long-run relationship mentioned in Equation 7 was estimated according to methodology of ARDL (1, 0, 1 ,0, 1) using Microfit 5.0. The results are as follows: 𝐿𝑛(𝑌𝑡) = 9.8 + 0.016𝑋1 − 0.015 𝑋2 − 0.20𝐿𝑛(𝑋3) − 0.001 𝑋4 (30.5)∗∗∗ (7.3)∗∗∗ (−6.3)∗∗∗ (−1.9) ∗ (−1.12) The domestic production of dates (𝑋1) has a significant positive effect while the level of domestic consumption of dates (𝑋2) has a negative effect on exports in the long term at the 1% level of significance. This means that a 1,000- ton increase in domestic production leads to a 1.6% increase in exports in the long run. On the other hand, increasing the domestic consumption of dates by 1,000 tons leads to a 1.5% decrease in exports. The price of date exports (𝑋3) has a significant negative impact on exports. A 10% increase in export prices leads to a 2.04% decrease in date exports. This relationship does not match with the economic logic that might explain why the government of Saudi Arabia has been buying date surpluses from farmers at promotional prices that exceed market prices and presenting them as donations to other countries. This is in the form of aid that has been provided by the government for more than thirty years to friendly countries when they suffer from problems like natural disasters and famine conditions. In Asian Journal of Agriculture and Rural Development, 11(3)2021: 113-119 117 addition, this result explains the inelasticity of the quantity of date exports with a change in price. The comparative advantage of date exports (𝑋4) does not show a significant effect on exports in the long run.Thus, the model results confirm the positive impact of domestic production and the negative impact of domestic consumption on Saudi date export levels and export values, which matches with economic logic. In contrast, the comparative advantage of Saudi dates in the international export market did not have any relationship with Saudi export quantity. Even though Saudi Arabian date exports have a comparative advantage, the analysis does not reflect that export advantage. The main reason could be because there has been no active program to promote Saudi dates in global markets throughout most time of the study period (1980–2017). The National Center for Palms and Dates was established in Saudi Arabia in 2011, and one of its goals is to take advantage of Saudi Arabia’s date production and to promote Saudi dates in global markets. 5.4. Estimating the ECM Error Model The ECM describes the dynamic relationship between the dependent variable and the independent variables in the short run, where the model takes the following equation: ∆𝐿𝑛(𝑌𝑡) = 𝜋0 + ∑ 𝜋𝑖∆𝐿𝑛(𝑌𝑡−𝑖) 𝑝 𝑖=1 + ∑ 𝛿1𝑗∆(𝑋1,𝑡−𝑗) 𝑞1 𝑗=0 + ∑ 𝛿2𝑗∆(𝑋2,𝑡−𝑗) 𝑞2 𝑗=0 + ∑ 𝛿3𝑗∆𝐿𝑛(𝑋3,𝑡−𝑗) 𝑞3 𝑗=0 + ∑ 𝛿4𝑗∆(𝑋4,𝑡−𝑗) 𝑞4 𝑗=0 + 𝜓𝐸𝐶𝑇𝑡−1 + 𝑣𝑡 (8) The parameters of the short-term relationship model in Equation 8 were estimated as ARDL(1, 0, 0, 0, 1), and the results are summarized in Table 4. The short-run coefficient indicates the immediate impact of independent variable change on the dependent variable. For example, production ∆(𝑋1) has a positive and significant effect on log exports if production increases by 1,000 tons, and there is an immediate 1.04% increase in exports. Annual domestic consumption ∆(𝑋2) has a negative and significant effect on log exports if domestic consumption increases by 1,000 tons, and exports would decrease by 1.02%. If export price have a negative impact on date exports, a 1% increase in export prices leads to a 13.5% decrease in exports. The results also show that the change in comparative advantage variable has a significant positive effect on the increase of date exports, ∆𝐿𝑛 (𝑌), in the short term. Also, a one unit per year increase in the relative importance leads to a 0.2% increase in date exports. Table-4. Estimates for short-term coefficients for the EC model. Regressor Coefficient Standard error T-ratio [Prob.] ∆(𝑋1) 0.0104 0.0019 5.580[.000] ∆(𝑋2) −0.0102 0.0020 −5.121[.000] ∆𝐿𝑛(𝑋3) −0.1348 0.0736 −1.832 [.073] ∆(𝑋4) 0.0020 0.0009 2.353[.022] 𝐸𝐶𝑇𝑡−1 * −0.6598 0.0985 −6.701[.000] R2 0.598 F-Stat. F(5,25) 15.158[.000] AIC −9.958 SBC −17.170 DW-statistic 2.114 Note: *𝐸𝐶𝑇𝑡−1 = 𝐿𝑛(Y) − 0.0157(X1) + 0.0154(X2) + 0.2043 𝐿𝑛(X3) + 0.0014(X4) − 9.775 The error correction factor has a value of −0.66, which is statistically significant. The negative sign confirms the convergence of the equilibrium from the short- to long-term equilibrium. It reveals the speed of the return of the variable of date exports toward its long-term equilibrium value after date exports deviate from equilibrium in the previous period (t-1). Two-thirds (66%) of this disequilibrium is corrected each year. The adjustment factor is relatively large because the correction in exports needs more than one year to return to long-term equilibrium. 5.5. Results of Diagnostic Tests of the Model The appropriateness of the model was assessed through tests for serial correlation in the error term, model specification error, the normality of residuals, and the heteroscedasticity of residuals. Diagnostic tests were run on the data using Microfit 5.0, and the results are summarized in Table 5. Table-5. Results of diagnostic tests for ARDL(1, 0, 0, 0, 1). Test statistics LM version F version A: Serial correlation CHSQ(1) = .385 [.535] F(1,50)= .334 [.566] B: Functional form CHSQ(1) = 9.823 [.002] F(1,50)= 10.195 [.002] C: Normality CHSQ(2) = 36.806 [.000] Not applicable D: Heteroscedasticity CHSQ(1) = .491 [.483] F(1,56)= .478 [.492] A: LM, Lagrange multiplier test of residual serial correlation B: Ramsey’s RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values (1) The model tests the null hypothesis that there is serial correlation for the error term. The statistical result shows that the value of Chi-square was χ2 = 0.385, and its probabilistic value was pr(χ2 > 0.385 = 0.535). Asian Journal of Agriculture and Rural Development, 11(3)2021: 113-119 118 Thus, it is greater than the critical value and is significant at the 5% level. Therefore, the null hypothesis was accepted to reveal that there was no serial correlation between the error terms of the ARDL(1, 0, 0, 0, 1) model. (2) Model specification error was tested through the Ramsey RESET test. This test examines the null hypothesis that there are no omitted variables. The F-statistic was F(1,50) = 10.195, and its probability value was pr(F > 10.195 = 0.002), which was greater than 5%. Thus, the model does not suffer from omitted variables. (3) The normality of residuals was tested through the Jarque–Bera test. The null hypothesis is that the error term data follow a normal distribution. The result rejected the null hypothesis that the error terms follow a normal distribution because the probability value of this test, pr(χ2 > 36.806 = 0.000), was less than the level of statistical significance. Therefore, the model suffers from abnormal residual data problems but, because the sample size is large, the distribution of the residual data has an approximately normal distribution. (4) The study examined the residuals for heteroscedasticity. The result shows the model did not suffer from heteroscedasticity because the probability value, pr(F > 0.478 = 0.492), was greater than the 5% level of statistical significance. That means the model failed to reject the null hypothesis of homoscedasticity. 6. FORECASTING EXPORT OF DATES FROM 2020 TO 2025 One of the main objectives of this study was to predict the export of dates from Saudi Arabia for the period 2020– 2025. The values of the independent variables were predicted using ARIMA(p,d,q) models, which are the domestic production of dates (𝑋1), the amount of domestic consumption of dates (𝑋2), the average export price (𝑋3), and the comparative advantage of date exports with respect to other agricultural export products (𝑋4). The Eviews 11 program was used with the ARDL(1, 0, 0, 0, 1) model to predict 𝐿𝑛(𝑌)̂ , and then to predict export values. Table 6 presents the predictions of the explanatory variables and date exports according to the ARDL(1, 0, 0, 0, 1) model, which was calculated using Microfit 5.0. Table-6. Prediction results using the ARDL model (1,0,0,0,1). Year (X1) 1000 ton (X2) 1000 ton (X3) (X4) ∆𝑳𝒏(𝑿𝟏) ∆𝑳𝒏(𝑿𝟐) ∆(𝑿𝟑) ∆𝑳𝒏(𝑿𝟒) Ln(y) Y ARMA(3,0) ARMA(3,0) ARMA(2,0) ARMA(1,1) Predict. Predict./ton 2019 1285* 1129* 1.261* 41.846* 12.114* 182317* 2020 1338 1182 1.115 34.446 12.338 228091 2021 1405 1234 1.039 30.499 12.605 297896 2022 1472 1284 0.891 28.078 12.897 399233 2023 1507 1315 0.785 26.395 13.071 474872 2024 1558 1359 0.644 25.092 13.232 557769 2025 1622 1411 0.526 23.996 13.459 700185 Note: *Actual values. Table 6 shows the model expectations from 2020 to 2025. The results show an increase in the quantity of export by 700,000 tons between 2020 and 2025. Thus, the prediction shows that exports would increase by 3.8-fold in 2025 compared to 2019. The average annual export level of dates in Saudi Arabia from 2020 to 2025 is expected to be about 443,008 tons, with a standard deviation of 172,910 tons. These expected export levels are restricted by the effects of the four factors that built this prediction model. Saudi Arabia has the opportunity to multiply its date exports at a high level of production, stability of consumption, and to increase returns by targeting the appropriate world market. 7. CONCLUSION Saudi Arabia has an opportunity to extend date exports to the international market because its production exceeds consumption. The government of Saudi Arabia is interested in increasing the competitiveness of dates in global markets as a strategic agricultural commodity, because the country has so many farms that produce the crop. This research analyzed the future market structure for Saudi date exports, highlighting this crop’s comparative advantage by using a cointegration model for the period 2020–2025. The study used the ARDL model with data covering the period 1980–2017. The estimated results of the relationship between Saudi date exports and other exogenous variables showed a significant effect from domestic production, domestic consumption of dates, and the price of exported dates in both the long and short term. In contrast, the comparative advantage variable for Saudi dates showed a significant effect only in the short term. The explanatory variables predict an expected increase in Saudi date exports in 2025 by more than 3.8-fold that of 2019 exports. Thus, dates have an export comparative advantage over other agricultural products in Saudi Arabia. Accordingly, there is an opportunity to develop further programs targeting the international market in support of date farms and exporters, and to support market research directed at satisfying the international market. Asian Journal of Agriculture and Rural Development, 11(3)2021: 113-119 119 Funding: This study received no specific financial support. Competing Interests: The authors declare that they have no competing interests. Contributors/Acknowledgement: All authors contributed equally to the conception and design of the study. Views and opinions expressed in this study are those of the authors views; the Asian Journal of Agriculture and Rural Development shall not be responsible or answerable for any loss, damage, or liability, etc. caused in relation to/arising out of the use of the content. REFERENCES Al-Abdulkader, A. M., Al-Kahtani, S. H., Ismaiel, S. M., Elhendi, A. M., & SaadIi, A. (2016). Enhancing marketing efficiency of the Saudi dates at the national and international markets. Alamri Yosef, al-Dakhil AI. International Journal of Economics and Finance, 8(8), 53-70. Al-Shreed, F., Al-Jamal, M., Al-Abbad, A., Al-Elaiw, Z., Abdallah, A. B., & Belaifa, H. (2012). A study on the export of Saudi Arabian dates in the global markets. Journal of Development and Agricultural Economics, 4(9), 268-274. Alabdulkader, A. M., Elhendy, A. M., Al Kahtani, S. H., & Ismail, S. M. (2017). Date marketing efficiency estimation in Saudi Arabia: A two-stage data envelopment analysis approach. Pakistan Journal of Agricultural Sciences, 54(2), 475–485. Available at: 10.21162/PAKJAS/17.3109. Ali, A., Al-Mulhim, F., & El-Habbab, S. (2014). The competitiveness of Saudi dates in selected European Union countries. Agricultural Science Research Journal, 5, 68-76. Ali, A. M., Fahad, A.-M., & El-Habbab, M. S. (2014). Saudi date exports demand in selected markets. International Journal of Agriculture and Crop Sciences, 7(11), 827. Dawood, A., & Kinnucan, H. (2018). Source–specific demand elasticities for dates in international trade (No. 2015-2018-108). The 2018 Southern Agricultural Economics Association Annual Meeting, February 2-6, 2018, Jacksonville, Florida - USA. Dhehibi, B., El Awar, F., Ben-Salah, M., & Aw-Hassan, A. (2018). Dates trade in GCC countries competitivity and complementarity. Paper presented at the The Sixth International Date Palm Conference (SIDPC) Abu Dhabi -UAE: 19 – 21 March1). Dhehibi, B., Salah, M. B., & Frija, A. (2018). Date palm value chain analysis and marketing opportunities for the Gulf Cooperation Council (GCC) countries. Chapter 2. Agricultural Economics Book. IntechOpen. London, UK. DOI: 10.5772/intechopen.82450. Engle, R. F., & Granger, C. W. J. (1987). Co-integration and error correction: Representation, estimation, and testing. Econometrica, 55(2), 251-276. Available at: 10.2307/1913236. FAOSTAT. (2020). FAOSTAT database. Retrieved from http://www.fao.org/faostat/en/. Intezar, M., Abdallah, A., & Rao, V. (2016). Opportunities and challenges of dates industry in Saudi Arabia: A study of Alkharj region. Journal of Applied Economic Sciences, 11(3), 428-434. Ismaiel, S. M., Al-Abdulkader, A. M., Al-Kahtani, S. H., & Saad, A. I. (2017). Scoring marketing obstacles and recommended solutions: The case of the date marketing system in Saudi Arabia. International Journal of Agricultural Marketing, 4(1), 129-141. Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control, 12(2–3), 231–254. Available at: 10.1016/0165-1889(88)90041-3. Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration with applications to demand for money. Oxford Bulletin of Economics and Statistics, 52(2), 169-210. Available at: 10.1111/j.1468-0084.1990.mp52002003.x. Nkoro, E., & Uko, A. K. (2016). Autoregressive distributed lag (ARDL) cointegration technique: Application and interpretation. Journal of Statistical and Econometric methods, 5(4), 63-91. Okafor, C., & Shaibu, I. (2016). Modelling economic growth function in Nigeria: An ARDL approach. Asian Journal of Economics and Empirical Research, 3(1), 84-93. Available at: 10.20448/journal.501/2016.3.1/501.1.84.93. Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289-326. Available at: 10.1002/jae.616. Pesaran, M., & Pesaran, B. (2009). Time series econometrics: Using microfit 5.0 (Window Version). Oxford: Oxford University Press. Ülke, V., Sahin, A., & Subasi, A. (2018). A comparison of time series and machine learning models for inflation forecasting: Empirical evidence from the USA. Neural Computing and Applications, 30(5), 1519-1527. Available at: 10.1007/s00521- 016-2766-x. Van de Minne, A., Francke, M., & Geltner, D. (2018). Forecasting US commercial property price indexes using dynamic factor models. Available at SSRN: https://ssrn.com/abstract=3148680. Zhu, L., Lim, C., Xie, W., & Wu, Y. (2018). Modelling tourist flow association for tourism demand forecasting. Current Issues in Tourism, 21(8), 902-916. Available at: 10.1080/13683500.2016.1218827. http://www.fao.org/faostat/en/ http://www.fao.org/faostat/en/