Time Series Analysis among Tourism, Financial Development, FDI and Economic Growth in Jordan

This study investigates the relationships among tourism, financial development, FDI inflows and economic growth in Jordan for the (1985-2016) period. The current paper has used bounds testing approach to confirm the relationship among the study variables. Multivariate Granger causality test is used to determine the directions of causality between the study variables. The results confirmed that there is evidenced of relationships among tourism, financial development, FDI inflows and economic growth. Also, the multivariate Granger causality test confirmed deferent directions of causal among the study variables.


Background
Over the past decades, several studies argued the relationship between the economic growth and its determinants hypotheses to better understand the interaction among them. These studies examined the relationship between economic growths represented by gross domestic product (GDP) and other economic factors based on the work of Keynes (1936). In general, the economic hypotheses which are foreign direct investment (FDI), energy consumption (EC), tourism (T) and financial development led economic growth are discussed by many researchers (See, Khan et al., 2014;Shahbaz, 2012;Hamdi et al., 2014;He & Ahmed, 2012;Balaguer & Cantavella-Jorda, 2002).
give a clear picture for the policy makers about the effects of (T, FD and FDI) on economic growth the relationship and the directions of causality among the selected factors is analysed. The rest of the current study is structured as the following. Jordanian economy overview is presented in section 2. The previous studies are provided in section 3. Data collection and model specification are presented in section 4. Econometric framework is discussed in section 5. The results and concluding remarks are discussed in sections6 and 7 respectively.

Jordanian Economy Overview
Jordanian economy is considered as one of the smallest economy in Middle East countries with several economic obstacles such as (high level of poverty, high level of inflation rate, huge budget deficit and high level of unemployment). Also, Jordan has a few natural resources and depends on its energy requirements on external sources. Therefore, Jordanian economy faced several shocks in past decade, for example global financial crisis, Arab spring and Syrian crisis (Central Bank of Jordan, 2013;Bekhet & Matar, 2013;World Bank, 2014). Frome these facts, Jordanian policy maker trade to deal with these facts by address set of internal and external policies starting from 1997 Jordan has sign set of economic agreement whit other countries. New energy strategy has been developed in 2007 that aims to development indigenous and renewable energy resources (World Bank, 2011;Bekhet & Al-Smadi, 2012).
Nowadays, the tourism sector in Jordan is play a vital role in Jordanian economy and showed substantial growth in terms of revenues to became the second faster sector in Jordan (Jordan Inbound Tour Operators Association (JITOA, 2017). Also, Jordan became a member of the main international organizations in the world. As a result of economic policies and international agreements, the Jordanian economy became the most significant market in the Middle East (Bechtel & Al-Smadi, 2012;Ministry of Industry and Trade, 2012). However, to show the performance of Jordanian economy activities could be by testing the level of RGDP in Jordan at constant prices. Figure1 demonstrations that RGDP in Jordan at constant prices recorded an annual growth rate of 4.4% for the 1985-2016period.  Historically, Jordanian policy maker have made several steps to improve the level of Jordanian economy for example, (many economic policies and roles have been adopted to encourage the private sector, prepare towards a free market economy and new investment regulations was adopted to increase the level of business environment). Resulted of that, Jordanian economy is considered as one of the highs countries in the world in terms of attracting FDI inflows (Bekhet & Al-Smadi, 2015). Also, Jordan has witnessed structural reforms containing liberalization of the trade and investment administrative, introduction of modern regulations, and institutions, to become one of the most open economies in the Middle East Countries (Jordan Investment Board, 2012).  1985  1986  1987  1988  1989  1990  1991  1992  1993  1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016 RGDP(Billions,

JD)
ijef.ccsenet.org In term of money supply (M2) many studies confirmed that there is a positively relationship between M2 and economic growth, this is because the high level of M2 came as a result of the growth in GDP and M2 is consider as one of the most important factor that affective in economic growth . Moreover, the tourism sector has become an important sector that has an impact on the economic development. Also, for many countries it is consider as the most important source of welfare and the main benefits of the tourism sector are the income creation and generation of jobs (JITOA, 2017). For Jordan the tourism sector accounted around JD2.6billion in 2016. This is because, Jordan has witnessed several development projects in some of main tourist attractions, which cooperate in marketing Jordan as a tourist destination and raise its competitiveness within the region (JITOA, 2017).
Also, Jordan National Tourism Strategy vision (2010)(2011)(2012)(2013)(2014)(2015), is to create a special point in Jordan that encourages the foreign visitor to come to Jordan that will lead to increase the level of Jordanian economic. However, this strategy aims to improve the level of quality service, diversifying products and increase the number of tourist in Jordan (Jordan Tourism Board, 2017). As a result of that, Jordan ranked in position 53 out of 130 countries on the Travel and Tourism Competitive Index (Travel and Tourism Competitiveness Report, 2016). Figure 3 show that Jordanian tourism sector performance (Total Number of Arrivals, (TA)) noted an annual growth rate of 4.1% for the 1985-2016 periods.

Previous Studies
The relationship among GDP, TA, FD and FDI inflows is examined in previous empirical studies (See, Mishra, Rout, & Mohapatra, 2011) for India, Georgantopoulos (2013) for India, Kumar (2014) for Vietnam, Ngoasong and Kimbu (2016) for Cameroon, Ridderstaat and Croes (2017) for Canada, United Kingdom, and United States). All these studies have given conflicting results about the relationship among these variables. Table 1 summarises the results of these studies.  As discussed above the existing literatures, there are given conflicting results about the relationship amongGDP, TA, FD and FDI inflows.Subsequently, to achieve the objectives of the current paper, it could be formulating the following hypotheses:

Data Collection and Model Specification
Annual time series data was used and collected for the  period. However, it was obtained from different sources. TA was collected from the Ministry of Tourism and Antiquitiesdatabase (http://http://www.mota.gov.jo). The variables of (GDP, FDI and FD) were obtained from the World Bank (https://data.worldbank.org/country/jordan).Furthermore, all the variables transformations into natural logarithmic (L) to reduce the hetrosecedasticity problem and to obtain the growth rate of the variable Montgomery et al., 2008;Chen et al., 1986). Thus, followed the empirical literature (Table 1), it is plausible to form the long-run, short-run and causality as in Equation (1): LGDP t = α+ δ 1 LTA t +δ 2 LFD t +δ 3 LFDI t +ω t Where the intercept is (α), error term is (ω), the variables coefficients is δ i s (i= 1,…., 3) and the time period is (t).

Econometric Framework
Several studies confirmed that if the time series data are not stationary, the regression analysis would not be true or spurious regression (Bekhet, Yasmin, & Al-Smadi, 2017;Gujarati & Porter, 2009). However, to select the suitable time series models are depends on the results of stationarity and co-integration tests (Bekhet & Matar, 2013a;Pesaran et al., 2001). Therefore, in the present study the augmented Dickey-Fuller (ADF) [1979,1981] and Phillips-Perron (P-P) [1988] and Kwiatkowski, Phillips, Schmidt and Shin (KPSS) [1992] statistical tests are used to detect the level of stationarity either at I(0), I(1) or I(d) to selected the appropriate time series models.
To reach the objectives of the present study, the Autoregressive Distributive Lag (ARDL) bounds testing model is utilized. As discussed in many study the ARDL model developed by Pesaran et al. (2001) has several important advantage. First, allow for testing the relationship among the variables at different levels of stationary data either I(1), I(0) or both. Second, this model gives well results in case of small sample of data used. Third, this model can take the suitable lag order without losing any long run information. Finally, this model could be reducing the serial correlation problematic (Hamdi et al., 2014;Chandran & Munusamy, 2009;Pesaran, Shin, & Smith, 1999).
Generally, if the equilibrium relationship between the study variables is confirmed, this means that these variables are co-integrated (Bekhet, Yasmin, & Al-Smadi, 2017). Thus, to examine the long and short run relationship among (i.e., GDP, TA, FD, and FDI) ARDL model could be formulated as in Equation. (2). (2) Where, the first difference operator is (Δ), the intercepts is (α i s), the long run coefficients is (η ij s), while the short run coefficients is (β ij s), the error terms is (ε it s), the optimal lag length is (k), the lag order is (s), and i.j=1.,….,4.
Furthermore, several studies argued that the vector error correction Model (VECM) is a standard technique to observe the causality direction between the study variables (Hamdi et al., 2014;Shahbaz et al., 2014;Khan et al., 2014;Gujarati & Porter, 2009). This model is developed from VAR model established by Engle and Granger in (1987) to examine the long and short run causality between the study variables (Gujarati & Porter, 2009). However, if all the study variables are stationary at same level and co-integration then the VECM is used to observe the direction of causality between the study variables (Bekhet & Mugableh, 2012;Johansen & Juselius, 1990). (see the general form of VECM in Equation (3). ( Where, the first difference operator is (Δ), intercepts is (θ i s), the short run coefficients is (Π ij s), the error correction terms coefficients (ECT t-1 )is (λ i s) which use to examine the long run causality, and I,j=1.,…,.4

Quality Data, Stationarity and Co-integration Results
International Journal of Economics and Finance Vol. 10, No. 12; zero mean and constant variance (ε t s ˷ N(0, ζ 2 )). Moreover, Table 2 shows that all the variables are in acceptance range of correlation coefficients. Also, the results show that all the variables have positively relationship between each other, which means the effects of the multicollinearity is not existed (Menyah et al., 2014;Hamdi et al., 2014).  Table 3 confirm that the study variables are stationary at I (1), with constant and trend in ADF, P-P, and KPSS tests at significant levels of (1%, 5% and 10%). The results of ADF, P-P and KPSS tests are consistent with other many findings such as, Ohlan (2017)   Note.
(1) The significance statistical level at 1%, 5% and 10% are a, b and c.
(2) H0 for ADF and P-P tests are rejected if the variables have unit root. Source: E-Views 7.2 econometric software.
As shown in Table 3, that all study variables are stationary at I(1), this means that the bounds F-statistics test would be utilized to confirm if the selected variables are co-integrated. Thus, the results of the co-integration are determined based on F-statistic test and reported in Table 3.  Table 4 shows that the H 0 of no co-integration among the variables in the LGDP t , LFD t and LFDI t models are rejected at 5% significance level, while it rejected among the variables in LTA t model at 10% significance level. The above results are consistent with the findings of Bekhetand Al-Smadi, (2015) for Jordan; Bekhet, Yasmin and Al-Smadi (2017) for Malaysia; Ohlan (2017) for India.

Long Run and Short Run Results
Several studies confirmed that, if the co-integration relationship among the variables in modelis warranted, then the long run and short run relationship between the study variables can be utilized (Bekhet & Al-Smadi, 2015;Khan et al., 2014;Uddin et al., 2013). However, in this study the lag order is selected based on the lowest value of Hannan-Quinn information criterion (HQ), Schwarz information criterion (SC), Akaike information criterion (AIC) tests Granger, 1981). However, the results confirm that the optimal lag length (k) is one lag. Table 4 shows that the long and short run relationship between the study variables is confirmed.  Table 5 confirm the relationship between LGDP t model and (LTA t and LFD t ) variables in the long run at 1% significance level, which means that an increase of the numbers of tourist arrival and the size of financial development will lead to increase the level of the economic growth. Also, all the coefficients results have a correct sign as discussed by several empirical studies see Ohlan (2017) for India; Bassil, Hamadeh, and Samara (2015) for Lebanese ;Lee, (2012) for Singapore. Furthermore, the result confirmed the relationship between FDI inflows and economic growth in the long run at 10% significance level. This result is similar to many studies and confirmed by the endogenous growth theory which recommended that FDI help economic growth in a capital scarce economy by increasing the volume of money supply as well as efficiency of the physical investment (Bekhet, Yasmin, & Al-Smadi, 2017;Bekhet & Al-Smadi, 2015;Romer, 1986;De Mello, 1999). Table 5 also presents the short-run dynamics equilibrium relationship results between the LGDP t and the study variables. At 1% significance levels, the financial development is positively associated with economic growth. However, the ΔLFDI t is positively associated with ΔLGDP t model at 5% significance levels. In addition, the coefficients of ECT t-1 are significant with appropriate signs in absolute value with 45%. This implies that this model ΔLGDP t is corrected from the short-run towards the long-run equilibrium by45%, in other word the long-run would be shortly corrected back by 1.8 year.
Therefore, the stability of co-integration is examined by conducting the CUSUM and CUSUMQ tests. The results of these tests are displayed in Figure 4.The CUSUM and CUSUMQ tests results confirmed that the co-integration estimates are reliable and consistent because both diagrams are within critical bounds at (5%) of significance level .

The Results of Multivariate Granger Causality Tests
The multivariate granger causality tests are utilized to find out the long-run and short-run directions of causality among the study variables. The empirical results are based on applying VEC model in Equation (3) and given in Table 6.  Table 6 confirms that in this study there is long run Granger causality (bidirectional) running among the study variables. These results were detected using t-statistics test at 1% and 5% significant levels. The above results are consistent with the findings of Ohlan (2017) for India; Seghir, Mostefa, Abbes, and Zakarya (2015) for 49 countries; Georgantopoulos, (2013) for India. However, the results of the short run causality are summarized in Figure 5. (1) represent the short run unidirectional results; (2) represent the short run bidirectional results. Source: Table 5.   Figure 5 shows that bidirectional causality running from economic growth to FDI inflows, from economic growth to financial development and from tourism to financial development is determined. Unidirectional causality running from economic growth to tourism, from financial development to FDI inflows, from tourism to FDI inflows is existed. However, these results are consistent with the findings of Ohlan (2017) for India; Seghir, Mostefa, Abbes, and Zakarya (2015) for 49 countries; Başarir and Ç akir (2015) for Turkey, France, Spain, Italy and Greece; Chulaphan, and Barahona (2017) for Thailand.

Concluding Remarks and Recommendations
This study is identified the long and short run linkage and causality directions between economic growth, tourism sector, financial development and FDI inflows for the 1985-2016 period. Stationarity tests, ARDL Model and Multivariate Granger Causality test are used. The results show that the increase of the numbers of tourist arrival and the financial development are absolutely lead to increase the level of the economic growth. Also, the result confirmed that there is long run relationship between FDI inflows and economic growth in Jordan. The multivariate Granger causality results show that there is bidirectional Granger causality running among the study variables in the long run. Also, there is bidirectional causality running from economic growth to FDI inflows, from economic growth to financial development and from tourism to financial development in the short run.
In the policy context, the finding of this study offer justification for Jordanian Government to give more careful consideration toward encouraging inbound tourism. Jordanian policy makers should give more attention for the current regulations and continue implementing the economic plans that ultimately lead to Increase the number tourist arrive in Jordan and create more productivity power in the Jordanian economy. These results are important for academics, corporations and foreign investors since they are interested in the relationship between economic growth, financial development, tourism and FDI inflows. Finally, the results of this study it appears to have no evidence that the financial development and FDI inflows are played a role in increase the level of tourism sector performance in Jordan.