Empirical analysis of the relationship between trade wars and sea—air transportation

Abstract The aim of this paper is to empirically analyse the relationship between the trade wars and modes of transport for selected countries. For this purpose the causality relationship between trade value and sea transport / air transportation for EU–G20 and US–G20 countries was examined. Panel causality analysis was used as a method in the study. The empirical findings of the study show the existence of a causality relationship between the trade value and modes of transport (sea transport and air transport) for country groups. This shows that the countries’ sea and air transport will be adversely affected by trade wars.


Introduction
Trade wars have been one of the most discussed issues both in the academic community and in politics recently. Trade wars manifest themselves largely as the use of additional customs duties and anti-dumping duties (Conybeare, 1987). If there is no agreement between countries in trade wars there is an increase in the costs of certain import products as a result of additional customs duties and trade barriers. Trade wars are also a form of overprotective economic conflict in which countries create tariffs or other trade barriers against each other. According to Conybeare (1987) there is a close relationship between the size of being affected by trade wars and the economic size of the country.
Accordingly large countries are largely unaffected by their own trade restrictions or restrictions with another country. However trade restrictions adversely affect small countries. One of the main reasons for this negative effect is the trade value asymmetry between the small and large countries. Therefore it is possible to mention the existence of a relationship between economic magnitudes and the level of influence in economic trade wars.
Trade wars have recently become one of the most frequently used methods to give countries a commercial advantage or punish another country. Therefore it is important to examine the negative effects of trade wars between countries. The negative impact of trade wars between countries is not limited to the trade value of countries. It is thought that trade wars may affect the goods or services related to the trade value and the sectors connected to it. The sectors covered in this study are sea transportation and air transportation. In this study it is assumed that trade wars between countries will affect the trade value of the country and therefore the sea and air transportation sectors used in the realization of trade activities will be affected.
This study examines the causality relationship between trade value and sea transport / air transportation for EU-G20 and US-G20 countries. This study is expected to contribute to the literature in several aspects. The first is that no studies have examined the relationship between trade wars and modes of transport in the literature. Thus it was aimed to fill this gap in the literature. Secondly, the impact of the trade wars on sea transport and air transport is to be examined in the context of the G20 countries. In this context it is aimed at determining the effect of trade wars 4 on G20 countries. The last is to examine the relationship between two modes of transport (sea transport and air transport) and commercial activities. This will enable the assessment of the impact of the trade wars on transport modes for the G20 countries.
The remainder of the paper is organized as follows. The first section presents the literature on the relationship between economy, trade and transportation. The second section contains detailed information about the variables and data used in the study. The third section describes the statistics of the sample and reports the main empirical results using firm-level data. The last section concludes the paper.

Literature
In this study which examines the effects of trade wars between countries on maritime and air transport the literature will be discussed under the heading on the relationships between air and sea transportation and economic developments. In the literature there are many studies investigating the causality relationship between the transportation sector and economic growth (Gramlich, 1994). It is seen that the focus of these studies is the cause-effect relationship between economic growth and transportation sector. Tong and Yu (2018) analysed the cointegration and causal relationship between growth of economic and transportation in China for 2000-2015. The results found a granger causal relationship between transportation and the growth of economic.
There are also studies in the literature that examine the relationship between commercial activities and the transportation sector among countries. In this context Nguyen and Tongzon (2010) concluded that Australia's trade volume with China, Japan and the USA contributed to the development of Australia's transport sector for the period 2001-2004. Saidi and Hammami (2017) concluded that there is a two-way causality relationship between freight transport and economic growth in the 2000-2014 period for high, middle-and low--income countries. This finding is consistent with the results obtained by Pao, Yu and Yang (2011). In other studies in the literature the transportation and economic relationship in the United States (Alagic, 2017); the relationship between transportation and GDP for EU28 countries (Gardiner & Hajek, 2016) and the dynamic relationship between freight transport, energy consumption and GDP in the United States (Benali & Feki, 2018) were empirically examined. Donaldson (2018) analysed railroads for general equilibrium in the trade model and the findings are a decrease of trade costs, an increase of trade and GDP. Hummels's (2007) technological changes in sea transportation was the critical input to growing trade in the first era of globalization during the latter half of the nineteenth century. The technological change in air transportation and the declining cost of rapid transportation has been a critical input into a second era of globalization.
In the literature there are studies which examine the effect of transportation modes on the economy of the country. In this context Taghvaee, Omaraee and Taghvaee (2017) discussed the short-and long-term impact between sea transport and GDP. Park and Seo (2016) examined the impact of ports on regional economic growth. Konstantakis, Papageorgiou, Christopoulos, Dokas and Michaelides (2019) studied transport fluctuations in Greece for the period 1998-2015 by analysing granger causality so, the findings were that the maritime sector were not affected by local economy. Rashid Khan and others (2018) analysed panel econometric techniques accounting for cross-sectional dependence and heterogeneity for 24 upper middle and high-income countries in the period of 1990-2015. Container traffic at the port positively affected per capita income across countries. Martínez-Zarzoso and Nowak-Lehmann (2007) analysed the real distance is not a good proxy for transportation costs and identify the central variables influencing road and sea transportation costs. Road and sea transport costs are central explanatory factors of exports and they seem to deter trade to a greater extent than road or maritime transit time when endogeneity is considered.
On average, changes in transportation costs account for almost half of the changes in welfare. These findings suggest that the endogeneity of transportation costs is an important mechanism determining the welfare effects of such a policy change. Research suggests that trade costs decline when total bilateral trade, which includes all modes of transportation, increases (Asturias, 2020). Wessel (2019) analysed five different transportation infrastructure types with respect trade effects. The results are shown there is a relationship between air and rail trade in the corresponding infrastructure type. Transport infrastructure plays an evident role in the export performance of economic growth for a country.
There are many studies focusing on the economic impact of air transport in the literature. The impact of economic development in the US on air transport (Chi & Baek, 2013); the relationship between air transport and GDP for countries in the South Asia region (Hakim & Merkert, 2016); the long-term and short-term causality relationship between economic growth and domestic passenger traffic in China (Hu, Xiao, Deng, Xiao, & Wang, 2015); the cointegration and causality relationship between air transport demand and economic growth in Brazil (Marazzo, Scherre, & Fernandes, 2010); the symmetric and asymmetric causality between GDP and the demand for airline in Turkey (Kiraci, 2018); the relationship between air transport and macroeconomic variables in Turkey (Kiraci & Battal, 2018); the causality relationship between air transport demand and economic growth in Italy (Brida, Bukstein, & Zapata-Aguirre, 2016) and the long-run relationship between aviation demand and economic growth in India (Mehmood, Shahid, & Younas, 2013) were examined.
In other studies in the literature the impact of air cargo transportation on local economic development in the United States (Button & Yuan, 2013); the impact of air traffic on regional economic performance in Europe (Mukkala & Tervo, 2013); the impact of civil aviation activities on international trade in Europe (Brugnoli, Dal Bianco, Martini, & Scotti, 2018); the relationship between airline passenger traffic and economic growth for seven different geographical regions of the world (Profillidis & Botzoris, 2015); the short and long-term impact of regional air transport on regional economic growth in Australia (Baker, Merkert, & Kamruzzaman, 2015) were investigated. Costa, Caetano, Alves and Rossi (2019) studied relationship between air transport services and economic development by using the linear regression method. The results show ambigu-ous relationships between explicative and dependent variables. Accordingly it can be seen that empirical studies are rarely seen in which the effects of trade wars on transport modes (sea transport and air transport) as discussed in this study. Therefore this study is expected to fill this gap in the literature.

Data and method
In this study basically three different variables were used. The first is trade value data. In this context international trade in goods ($) data from both EU and USA to G20 countries were utilised. The second data used in the study is on sea transport. The data obtained here refers to the portion of the trade value from the EU and the US to the G20 countries, carried by sea. In other words the data related to the part of the total trade value made from the EU and the USA to the G20 countries using sea transportation was used. The last data used in the study is related to how much of the trade in the aforementioned countries takes place by air transportation. In other words data on the part of the total trade value from the EU and the US to the G20 countries using air transport was obtained. G20 countries realize approximately 75% of the international trade in the world. Therefore the countries that carry out trade wars and direct foreign trade were analysed. In addition countries that can be considered as related to trade wars are included.
In the study the total trade value from EU to G20 countries and transportation modes 5 (sea and air transportation) used in the trade between 2002 and 2016 were used. Since the data cannot be obtained for all countries the trade value, sea and air transport data from EU to fifteen countries were analysed in the mentioned period. Similarly data on the total trade value from the USA to the G20 countries in the period of 1999-2016 and the modes of transport (sea and air transport) used were used. In this study trade value, maritime and air transport data from the USA to sixteen countries were included in the analysis due to the lack of data. The data used in the study were obtained from the International Trade Administration (ITA) and Eurostat database.
Two different analyses were used to reveal the causality relationship between trade value and trade modes of transportation (maritime and air transport). The first of these is the bootstrap panel Granger causality analysis (based on the assumption of heterogeneity) developed by Kónya (2006). The second is the panel causality test used for heterogeneous mixed models developed by Emirmahmutoglu and Kose (2011). The reason for choosing these panel causality tests is that they are widely used in the literature. In addition these are the panel causality methods most suitable for the data of the study. These methods reveal the causal relationship between variables. Therefore it is appropriate to use panel causality methods in the study, Kónya (2006) and Emirmahmutoglu and Kose (2011).

empirical findings
In this study international trade in goods by mode of transport were analyzed. The main purpose of the study is to reveal the effect of the spread of trade wars on the modes of transportation in countries. Descriptive statistics of the variables included are presented in Table 1.

Cross-sectional dependency
Panel causality analyses were performed and firstly whether there is a crosssectional dependency in the series (Table 2) was examined. The cross-sectional dependence relates to whether the shock occurring in any of the series is affected by all units (countries) included in the panel data. Breusch and Pagan (1980), Pesaran (2004) and Pesaran, Ullah and Yamagata (2008) cross-sectional dependence tests were used. Table 2 shows the cross-sectional dependence test results. It shows that the H 0 hypothesis was rejected in both country groups included in the analysis. This shows that there is a horizontal cross-section dependence in the series. Given the developments in globalization and free movement of capital, trade relations between countries are expected to be versatile and affect each other.
Therefore the results of horizontal cross-sectional dependence are in line with the expectations.

Kónya (2006) panel causality test
The panel causality test developed by Kónya (2006) uses the seemingly unrelated regressions estimator instead of least squares. In the bootstrap panel causality analysis proposed by Kónya (2006) bootstrap test statistics are used instead of asymptotic critical test statistics in the Wald test. In this way cross-sectional dependence and heterogeneity are taken into consideration. In addition pre-tests such as stationarity and cointegration are not required for the series. In this method the direction of causality is analysed based on country-specific bootstrap critical values in the Wald test and does not require a common hypothesis for all members of the panel (Altıntas & Mercan, 2015, p. 328).
Kónya's panel causality approach describes a system that includes two sets of equations. The bootstrap based panel causality method can be expressed by the following equation series. and In this equation y represents trade (TRADE) between countries and x represents sea or air transport (SEA-AIR). In addition N represents the number of units (countries) in the panel (j = 1, …, N), t represents the time period (t = 1, …, T), and l indicates the number of delays. ly 1 and lx 1 represent the maximum delay lengths of the variables in the first set of equations, and ly 2 and lx 2 represent the maximum delay lengths of the variables in the second equation system. As a result of the application for a unit in the panel (i), if the all coefficients δ 1, i are not equal to zero and the all coefficients β 2, i β are equal to zero, therefore there is a one-way causality relationship from x to y. Similarly, if all of the coefficients β 2, i β are not equal to zero and all of the coefficients δ 1, i δ are equal to zero, there is a one-way causality relationship from y to x. In addition if the coefficients δ 1, i and β 2, i δ are not all equal to zero, then there is a bi-directional causality relationship between the variables. Finally, if the coefficients δ 1, i and β 2, i are equal to zero it is concluded that there is no causality between the variables. The bootstrap panel causality test results obtained from the analysis are presented in Table 3. Table 3 presents the results of the causality analysis of trade and sea transport from the European Union (EU) countries to the G20 countries. According to the analysis there is a causality relationship from trade to sea transport from EU to Brazil, Canada, Turkey and the United States. This situation indicates that trade value between EU and mentioned countries will be affected by sea transport depending on the growth opportunities. In contrast none of the countries included in the analysis have a causality relationship from sea transport to trade value. Table 4 presents the results of the causality analysis for trade value and air transport from the European Union (EU) countries to the G20 countries. Accordingly, none of the countries included in the analysis have a causality relationship from trade value to air transport or from air transport to trade value. Table 5 presents the results of the causality analysis of trade value and sea transport from the US to the G20 countries. None of the countries included in the analysis have a causality relationship. Table 6 presents the results of the causality analysis of trade value and sea transport from the United States to the G20 countries. None of the countries included in the analysis have a causality relationship.
[31]  Note: TRADE → SEA: It means that trade is the cause of sea transport. SEA → TRADE: It means that sea transport is the cause of trade. The values of *, ** and *** indicate that the test statistic is significant at 1%, 5% and 10% significance levels, respectively. The optimal lag length was determined according to the Akaike information criterion. Bootstrap number is 1000. The maximum lag length is 3. Source: Own study based on the International Trade Administration (ITA) and Eurostat database. Note: TRADE → SEA: It means that trade is the cause of sea transport. SEA → TRADE: It means that sea transport is the cause of trade. The values of *, ** and *** indicate that the test statistic is significant at 1%, 5% and 10% significance levels, respectively. The optimal lag length was determined according to the Akaike information criterion. Bootstrap number is 1000. The maximum lag length is 3. Source: Own study based on the International Trade Administration (ITA) and Eurostat database. Note: TRADE → SEA: It means that trade is the cause of sea transport. SEA → TRADE: It means that sea transport is the cause of trade. The values of *, ** and *** indicate that the test statistic is significant at 1%, 5% and 10% significance levels, respectively. The optimal lag length was determined according to the Akaike information criterion. Bootstrap number is 1000. The maximum lag length is 3. Source: Own study based on the International Trade Administration (ITA) and Eurostat database.

emirmahmutoglu and Kose (2011) panel causality test
The panel causality test developed by Emirmahmutoglu and Kose (2011) is a method based on meta-analysis in mixed heterogeneous panels. In the meta--analysis developed by Fisher (1932), tests are performed for N units and the significance levels (probability values) of this test statistic are used. The superior side of this test, which is the panel data version of the causality test developed by Toda and Yamamota (1995), is that it reduces information loss by modelling the series with level values, allows the delay length to be differentiated for each series and take into account the horizontal cross-section dependence (Zeren & Ergün, 2013, p. 233;Buberkoku, 2016, p. 189).
In the panel causality test developed by Emirmahmutoglu and Kose (2011) a standard panel VAR estimate is made at the first stage and the appropriate delay length (p) is determined. In the next step, the integration level (d max ) of the variable with the highest degree of integration is added to the appropriate delay length. Finally a panel VAR model is estimated using the level values of the variables for the delay level (p + d max ) (Emirmahmutoglu & Kose, 2011, pp. 871-872;Topallı, 2016, p. 89). In Emirmahmutoglu and Kose (2011) test, panel VAR model is estimated for each horizontal section as follows.
In the analysis the test is performed with the corrected Wald (modified Wald) test for the estimated k i lag length. The hypothesis H 0 is established as there is no causality relationship from the variable y to the variable x. Table 7 presents the results of the causality analysis of trade value and sea transport from the European Union (EU) countries to the G20 countries. There is a causality relationship from trade value to sea transport from EU to Australia, Indonesia and Mexico. In addition there is a causality relationship between sea transport to trade value from the EU to Australia, Indonesia, South Korea, Mexico and Russia. In this context there is a two-way causality relationship from trade value to sea transport from EU to Australia, Indonesia and Mexico. The results of Fisher's test statistics, which generally evaluate the findings for all countries in the table, show that there is a two-way causality relationship from trade value to sea transport and from sea transport to trade value. Table 8 presents the results of the causality analysis of trade value and air transport from the European Union (EU) countries to the G20 countries. There is a causality relationship from trade value to air transport from EU to Japan and USA. In addition, there is a causality relationship from the EU and the US air transport to trade value. In this respect there is a two-way causality relationship from trade value from EU to USA to air transportation and from air transportation to trade value. Table 9 presents the results of the causality analysis of trade value and sea transport from the US to the G20 countries. There is a causal relationship from trade value sea transportation from USA to Brazil. In addition, there is a causal relationship from sea transportation to trade value from the USA to France, Italy and South Korea. The Fisher test statistics, in which the findings are gen- Note: TRADE → SEA: It means that trade is the cause of sea transport. SEA → TRADE: It means that sea transport is the cause of trade. The values of *, ** and *** indicate that the test statistic is significant at 1%, 5% and 10% significance levels, respectively, k i shows the optimal lag length. The optimal lag length was determined according to the Akaike information criterion. Bootstrap number is 1000. The maximum lag length is 3.
erally evaluated for all countries in the table, show that there is a two-way causality relationship from sea transport to trade value. Table 10 presents the results of the causality analysis of trade value and air transport from the US to the G20 countries. There is a causal relationship from trade value to air transportation from USA to France, Italy and Saudi Arabia. In addition there is a causality relationship from the US to France, Germany, Italy and South Africa between air transport to trade value. In this context there is a bi-directional causality relationship between air transport and trade value from the USA to France and Italy. The Fisher test statistics, in which the findings are generally evaluated for all countries in the table, show that there is Note: TRADE → SEA: It means that trade is the cause of sea transport. SEA → TRADE: It means that sea transport is the cause of trade. The values of *, ** and *** indicate that the test statistic is significant at 1%, 5% and 10% significance levels, respectively, k i shows the optimal lag length. The optimal lag length was determined according to the Akaike information criterion. Bootstrap number is 1000. The maximum lag length is 3.
a two-way causality relationship between trade value to air transport and from air transport to trade value.

Conclusions
In this study the causal relationship between trade value and transportation modes (sea transportation or air transportation) is examined empirically. In the study, trade value data for 15 countries from the EU and 16 countries from Note: TRADE → SEA: It means that trade is the cause of sea transport. SEA → TRADE: It means that sea transport is the cause of trade. The values of *, ** and *** indicate that the test statistic is significant at 1%, 5% and 10% significance levels, respectively, k i shows the optimal lag length. The optimal lag length was determined according to the Akaike information criterion. Bootstrap number is 1000. The maximum lag length is 3.
Source: Own study based on the International Trade Administration (ITA) and Eurostat database.
the USA and data on the modes of transport used to provide this trade value are included in the analysis. Within the scope of the study, panel Granger causality developed by Kónya (2006) and panel causality analyses developed by Emirmahmutoglu and Kose (2011) were used to reveal the causality relationship between these variables.
According to Kónya (2006) the panel Granger causality results showed that there is a causality relationship from trade value to sea transport from EU to Brazil, Canada, Turkey and the United States. This situation shows that a positive or negative situation in the trade value in these countries will affect the sea The results of the causality analysis for air transport from the EU to the G20 countries show that there is a Granger causality relationship from the trade value to air transport from the EU to Japan and the US. Moreover, there is a Granger causality relationship air transport to trade value the EU and US. In this context developments that may affect the trade value from the EU to Japan and the USA are expected to affect air transportation.
According to the results of the analysis of trade value and sea transport from the US to the G20 countries there is a Granger causality relationship from trade value to sea transport from the US to Brazil. Therefore the findings indicate that developments that may affect trade value between the USA and Brazil may also affect sea transport. Furthermore, there is a Granger causality relationship between sea transport to trade value from the US to France, Italy and South Korea. The results show that a positive or negative situation in sea transport from the USA to France, Italy and South Korea may affect trade value. The results of Fisher's test statistics, which generally evaluated the findings for all countries included in the analysis, indicate the presence of a two-way causality relationship between sea transport to trade value.
According to the results of the analysis of trade value and air transport from the US to the G20 countries, there is a Granger causality relationship between trade value to air transport between the US to France, Italy and Saudi Arabia. Hence, developments that may affect the trade value between these countries and the USA are expected to affect air transportation. Besides, the findings suggest that there is a Granger causality relationship between air transport to trade value from the US to France, Germany, Italy and S.Africa. The Fisher test statistics, which generally evaluated the findings for all countries included in the analysis, show that there is a bi-directional Granger causality relationship between trade value to air transport and between air transport to trade value. Different findings were found after the Kónya (2006) and Emirmahmutoglu and Köse (2011) panel causality analysis. This is because the econometric models behind these tests are different. There is no information in the literature about which test is superior but the aim here is to reveal the causal relationship through different panel causality tests. In addition it is seen that the causal results of sea and air transport are different. In other words while there is a causal relationship between trade value and sea transport in one country, there is no causal relationship with air transport in the same country. The main reason for this is that the products carried by sea and air transport have different characteristics. For example, heavy but relatively inexpensive products are carried by sea. In air transport light and expensive products are carried. Therefore the causal relationship may differ depending on the mode of transport. In international trade the weight and price of the products transported is the reason why the causal relationship differs.
When the findings obtained within the scope of the study are evaluated in general terms it is expected that trade wars, currency wars and the protective policies of the countries will affect the trade value. It has been empirically demonstrated that trade contraction between countries may also affect sea and air transport. Due to the close relationship between trade value and transport modes developments that may occur between countries and affect trade value are also expected to affect transport modes. Therefore the stakeholders of the sea and air transport sectors should take into account the trade value between the countries and the expected developments in the trade value. In future studies low income countries can feature and the effect of international trade in goods on transportation modes can be analysed. In addition examining the long-term relationship in studies may contribute to the literature.