Next Article in Journal
Socioeconomic Sustainability for Low-Income Households: The Mediating Role of Financial Well-Being
Next Article in Special Issue
“I Am Here to Fly, but Better Get the Environment Right!” Passenger Response to Airport Servicescape
Previous Article in Journal
Identification and Classification of the Dissolved Substances from Sludge Biochar and Their Effects on the Activity of Acid Phosphomonoesterase
Previous Article in Special Issue
Borrowing Power from Potential Entrants and High-Speed Rail: Entry Pattern of China’s Low-Cost Carrier
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Does Air Cargo Matter in Chinese Regional Economic Development? An Empirical Granger Causality Test

1
Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao’an Road, Shanghai 201804, China
2
Antai College of Economics & Management, Shanghai Jiao Tong University, Shanghai 200240, China
*
Author to whom correspondence should be addressed.
Sustainability 2022, 14(15), 9753; https://doi.org/10.3390/su14159753
Submission received: 15 June 2022 / Revised: 23 July 2022 / Accepted: 2 August 2022 / Published: 8 August 2022
(This article belongs to the Special Issue Aviation Management and Air Transport Industry II)

Abstract

:
Using longitudinal data of 110 prefecture-level cities in China between 2006 and 2019, this paper conducts a Granger causality test between air cargo and regional economies through a multivariate VAR model, and attempts to explain the causal relationship in accordance with the endogenous and exogenous growth theory. Our results show that while the impacts of regional socioeconomic factors, including GDP and employment, on air cargo are larger in terms of magnitude than those of air cargo on the regional economy, the former effects are less robust when subsamples are included in the models. The benefits of air cargo growth for local economic growth are less dependent on the sizes, locations, and characteristics of cities. In accordance with the endogenous and exogenous growth theory, air cargo development policies need to be more favorable to regions with faster growth in secondary industry employment, as well as regions that are in urgent need of developing import and export trade.

1. Introduction

There has been an extensive collection of research regarding the impacts of air transportation on regional development. An academic consensus is generally achieved, in that the air transportation industry can exert a positive impact on the regional economy, though to various degrees. Debbage (1999) defined two ways in which the availability of air transportation affects the regional economy: first, the construction of an airport is a direct investment in the regional economy and generates on-site employment, and the effects of such a large investment multiplier can be significant in sectors such as wholesale goods and ground transportation; second, air transportation can change a region’s economic linkages with other regions, and create differences in regional competitiveness [1]. However, in previous studies, attention has been focused on air passenger transport, while research on air cargo is scarce. Such neglect may be partially due to the late emergence of air cargo as a major research topic, or simply because of the lack of data.
The main types of cargo typically transported by air include capital and transportation equipment, high-value consumer goods (gold and other valuables, electronics, apparel), intermediate products in manufacturing, and refrigerated agricultural products (fresh vegetables and seafood) [2]. In general, cargo transported by air has different characteristics than that transported by road, rail, or port, either because of its high unit quality or time sensitivity. Because of this, it has been argued that air cargo can provide new avenues in the process of product “self-discovery” by both expanding the set of products for which a region may be competitive, and by providing additional avenues for entrepreneurial experimentation by reducing ex ante investment costs [3]. The flower industry in Ethiopia is a good example [4].
Air transportation provides a competitive advantage, and further brings about agglomeration effects; agglomeration economies drive firms and employees to allocate geographically, but the nature of such a causal relationship is difficult to prove. Just as endogenous growth theory suggests that economic policies can positively affect the long-run growth rates of the regions in which they are implemented, while these faster-growing regions will attract more and more resources as a result of policy intervention, this difference will continue to widen as the economy grows [5]. Neoclassical economism, on the other hand, argues that inputs to the economy do lead to economic growth, but that the long-run growth rate is caused by technological change, and that economic policy can only temporarily raise the growth rate to the steady-state value to which it must eventually return [6,7]. This means that economic policies are exogenous growth factors for regional economies. The debate between these two theories makes it important for air cargo hub planning to effectively measure the relationship between air cargo and the regional economy.
In recent years, with the rapid development of e-commerce and express logistics in China, and the continuous relocation and upgrading of manufacturing industries, the demand for air cargo is increasing day by day. However, China has never had specialized air cargo or a complete air cargo airport system; thus, in the 13th Five-Year Plan in 2015, the Chinese government made it a priority to speed up the construction of cargo-oriented airports. In 2019, the National Development and Reform Commission (NDRC) and the Civil Aviation Administration of China (CAAC) also proposed the construction of an air cargo hub system consisting of comprehensive hub airports and specialized cargo hub airports [8], making the planning and construction of air cargo hubs a hot topic once again. As a large infrastructure system that is generally considered to promote economic development, the idea was put forward, and several provinces started planning and visioning to build regional or even national air cargo hubs; however, in China, the intrinsic relationship between air cargo and the economy is still unclear, and it is not clear how this large infrastructure system can best play a larger role. Disorderly planning and competition can lead to a great waste of resources and time. Therefore, this paper attempts to develop a panel Granger causality model between China’s air cargo indicators and regional macroeconomic indicators to explore some of the intrinsic linkages between China’s air cargo and regional economies, with a view to providing some insights into the planning of China’s air cargo hubs.

2. Literature Review

2.1. Infrastructure and Regional Economic Growth

There are countless studies on transportation infrastructure and regional economic development; for example, Solow (1957), a representative of neoclassical economic growth theory, analyzed the impact of transportation infrastructure investment on economic growth [8]. He believed that the impact of transportation infrastructure on economic growth is not sustainable due to the “law of diminishing marginal returns”, and although it can increase the capital stock of society, transportation infrastructure investment has only short-term effects on economic growth and cannot produce long-term effects, which means that transportation infrastructure investment is only an exogenous factor for regional economic growth. However, the endogenous economic growth theory, represented by Romer (1986, 1990) and Lucas (1988), argues that the externalities of infrastructure investment are a fundamental source of long-term economic growth [9,10,11]. As Bougheas (2000) argued, economic infrastructure not only enhances the marginal productivity of factors, but also facilitates the inter-regional mobility of labor and commodities, expanding the scope of markets, optimizing resource allocation, and deepening the division of labor, thus allowing the economy to gain endogenous growth momentum [12].
Many subsequent scholars have further investigated the positive impact of transportation infrastructure on regional economic development through empirical analysis. For example, Aschauer’s (1989) empirical paper found a strong positive correlation between physical infrastructure investment and macroeconomic performance, and yielded an output elasticity of 0.39 for infrastructure, although the idea was subsequently debated by multiple parties [13]. This conclusion is also supported by Munnell (1990a) using time series data from 1948–1987 for the U.S. [14]. Tatom (1991) argues that the variables in the time series data may be “pseudocorrelated”, because the model suffers from multicollinearity, and is unable to account for the heterogeneity of different cross-sectional data [15]. Since then, an increasing number of scholars have used panel data to study the impact of transportation infrastructure. For example, Nourzad and Vrieze (1995) analyzed panel data for seven OECD countries from 1960–1988, and found that the output elasticity of infrastructure was 0.05 [16]; Chandra and Thompson (2000) first studied the impact of the interstate highway system on total annual income through a single-digit SIC code and found that highway connections had a significant impact on financial, insurance, real estate, retail, and service revenues with a slight positive impact, but a slightly negative impact on manufacturing and agricultural revenues [17]; Duranton (2012) and (2014) found that, on average, a 10% increase in urban highway stock leads to an increase in employment of about 1.5%, while highway connections have a positive impact on intercity trade [18,19], etc. While the findings largely support the conclusion that transportation infrastructure has a catalytic effect on economic growth, the empirical evidence on whether the relationship is endogenous or exogenous to growth remains mixed, making it difficult to distinguish between a lasting or temporary impact between investment in infrastructure and regional economic growth.
As implied by Romer’s (1986) endogenous growth model, the growth of the investment rate determines the permanent growth of the long-run growth rate [9], i.e., the higher the investment rate in period t, the higher the economic growth rate from period t. This also suggests that the correlation between the past value of investment rate growth and the current growth rate should be positive once the past history of growth rates is controlled for. In Solow’s (1957) standard neoclassical growth model, long-run productivity growth, while driven by the rate of technological progress, changes in the share of investment can also lead to temporary changes in the growth rate [8]. Specifically, from the steady state, an increase in the investment rate in period t leads to a simultaneous increase in productivity, but its growth rate will gradually decline from period t until it approaches its steady state value. Thus, once the lagged value of the growth rate is controlled for, the correlation between the lagged value of the investment rate and the current growth rate is expected to be negative during the transition dynamics. Based on these definitions, Vanhoudt (1998) directly states that the process of testing the endogenous growth theory is very close to a simple Granger causality equation [20]. Since then, Hartwing (2009), in testing the endogenous relationship between physical investment, human capital investment, and GDP growth, gives us a very good framework for panel Granger testing, which is a dynamic panel regression model (GMM) to test multivariate Granger causality [21].

2.2. Granger Causality Test between Air Transport and Regional Economic Growth

The literature on the Granger causality test between air transportation and regional economic development is mainly focused on the last 20 years, and Kulendran and Wilson (2000) is one of the first papers that used Granger causality analysis to study air transportation and economic development. Kulendran and Wilson (2000) applied cointegration and Granger causality tests to analyze the relationship between international trade (total, exports and imports) and international travel (total, business and holidays) in Australia between 1982 and 1997, and found that the causal relationship between air transport and international trade varies by trading partner due to the diversity of trade relationships [22]; however, he mainly analyzed the causal relationships across cross-sections, and did not test Granger causality for the entire panel.
Mukkala and Tervo (2013) pioneered the use of Hurlin and Venet’s (2001) three-step Granger causality analysis method (homogeneous non-causality test, causality nature test, regional subset causality test) to analyze the impact of air transport on regional development in 13 western European countries using panel data for the period 1991–2010 [23,24]. This approach allows the testing of the homogeneity and heterogeneity of the Granger causality between air transport and regional development. The results of the article show that regional development has a significant impact on air transport activities, but the impact of air transport activities on regional development seems to be more significant, especially in outlying regions; hence, the authors believed that “there are good reasons to protect local airlines because they are important for the development of remote areas”.
Later, Van de Vijver, Derudder, and Witlox (2016) similarly used Hurlin and Venet’s (2001) three-step Granger causality analysis approach to investigate whether there is significant Granger causality between employment and air passenger transport [25]. The authors revealed rare cases of mutual causality, and relatively common cases of no causality, by analyzing the total employment map in the framework of Mukkala and Tervo (2013). The results of the article show that air transport has a significant effect on employment in relatively peripheral European regions, especially in services employment, but no causal relationship with manufacturing employment, in most regions.
The study by Baker, Merkert, and Mamruzzaman (2015) goes deeper in terms of Granger causality tests [26]. The authors analyzed the Granger causality between passenger traffic growth and total real taxable revenue growth for 88 airports (including 51 regional and 37 remote airports) in less-developed regions of Australia, based on data for the period 1985 to 2011. Due to the long study period of this article, the authors set a maximum four-year time lag, and found that the positive impact of remote airline growth on economic growth remained significant even with a four-year lag.
These abovementioned studies focused on the relationship between air passenger transportation and regional economic development; in contrast, a relatively rare article by Button and Yuan (2012) explores the potential role of air transport in stimulating local and regional economic development by analyzing the Granger causality between air cargo and regional economic variables [27]. The authors used a binary finite vector autoregressive (VAR) model to investigate the Granger causality between air cargo volume and each regional economic variable (employment, personal income, and per capita income). The results again indicate that air cargo contributes positively to local economic development, while there is no significant Granger causality in terms of the effect of regional economic variables on air cargo, except for employment. Furthermore, the Granger cause of employment on air cargo is much weaker than the Granger cause of air cargo on employment, meaning that the evidence that air cargo is endogenous to the economic growth process at the metropolitan level is much weaker.
From these studies, it can be seen that the impact of air transport on regional economy is stronger than the impact of regional economy on air transport, especially in terms of service employment and development of remote areas. However, the above literature mainly focuses on air passenger transportation in developed countries, and all of the studies use bivariate Granger causality analysis. As described in Hartwig (2009), in bivariate Granger causality analysis, spurious causality may occur when both variables have “common causes” that are not present in the regression equation [21]. The spurious causality disappears only when all “common cause” variables are included in the regression. In determining whether there is a causal relationship, the conclusion that there is no causal relationship is valid only if no causal relationship between the two variables is tested in both multivariate and bivariate analyses.
In order to avoid spurious Granger causality, multivariate VAR models and bivariate VAR models between air cargo indicators and regional economic indicators are developed in this paper, and the Arellano–Bond two-step systematic GMM estimator in Stata is chosen to analyze the models. This paper constructs a multivariate VAR model to estimate the Granger causality between air cargo and regional economic development by carefully screening the existing data related to air cargo and regional economy, and tests the results using a bivariate VAR model, which makes the estimation results of Granger causality between air cargo and regional economic development more reliable. The results of this paper can provide new ideas to the study of the relationship between air transportation and regional economic development, and at the same time, fill some knowledge gaps in the study between air cargo and regional economic development in China. Finally, by combining the results of the panel Granger causality assessment with different schools of thought on economic development, this paper can provide some new insights into the sustainable development of air cargo, and can help local governments to better evaluate the role of air cargo hubs.
The rest of the paper is organized as follows: Section 3 details the data sources and composition, Section 4 presents the models, Section 5 analyzes the results of this paper in detail, Section 6 provides a robustness analysis and discussion of the results, and finally, summarizes the findings of this paper.

3. Data

The air cargo data in this paper were obtained from the annual reports published on the official website of the Civil Aviation Administration of China. Macroeconomic data were obtained from the 2006–2019 China City Statistical Yearbook. For the selection of cities, 110 prefecture-level cities were selected as samples based on a combination of data availability, mainly based on air cargo and mail throughput data, and related macroeconomic data, from 2006 to 2019. The total GDP of the selected prefecture-level cities was accounted for according to the 2019 data, accounting for about 70% of the total GDP of the country. Finally, the selected cities were divided according to the geographical regions of China, the details of which are shown in Table 1.
For the selection of variables, the air cargo and mail throughput data of each airport published by the Civil Aviation Administration of China were used as the air cargo index of the corresponding prefecture-level city. Alternatively, the total import and export trade; GDP; size of resident population; total employment; and employment in primary, secondary, and tertiary industries published by the statistical bureau of each prefecture-level city were used as the regional macroeconomic indicators. The relevant details are shown in Table 2.

4. Model

The GMM method has evolved from differential GMM to systematic GMM. The difference between the two is that differential GMM only estimates the difference equation, and uses the lag term of the level value as the instrumental variable of the difference equation, while systematic GMM associates the level and difference equations, and uses the lag term of the level value for the instrumental variable of the difference equation, and also uses the lag term of the difference variable as the instrumental variable of the level equation. The panel data in this paper are large in cross-section and short in time series; when the horizontal lag term is often a weak instrumental variable for the endogenous variables in the difference equation [28], and the use of differential GMM may lead to biased estimates, systematic GMM can effectively avoid this problem. In the estimation process, a multivariate VAR model consisting of air cargo throughput and macroregional economic variables is first estimated by GMM, and a preliminary conclusion of causality is drawn based on the multivariate estimation results. Subsequently, a bivariate VAR model consisting of air cargo throughput and macroregional economic variables is estimated by GMM to test whether the non-causal relationship of the multivariate VAR estimation holds. Finally, according to the estimation results, the Wald test is conducted by applying the coefficients of the lagged explanatory variables. The null hypothesis of the Wald test is that the lagged coefficients of the explanatory variables are jointly equal to zero. If the p-value of the Wald test is significant, the null hypothesis is rejected, at which point, the significant Granger causality in the estimated results is considered to be strong; otherwise, even if there is significant Granger causality, the causality is considered to be weak.
The basic model structure is shown in Equation (1):
X i t = α 0 + s = 1 2 β s X i t s + s = 0 2 γ s Y i t s + s = 0 2 η s Z i t s + s = 0 2 κ s W i t s + s = 0 2 ω s E i t s + μ t + ε i t
The model is a multivariate VAR model, in which air cargo and mail throughput, total import and export trade, GDP, size of resident population, and total employment are taken as a set of variables; each variable in the set is taken in turn as the dependent variable, and the remaining variables other than the dependent variables are taken as independent variables, and are estimated separately by GMM. As an example, when air cargo and mail throughput is the dependent variable, Xit, Yit, Zit, Wit, and Eit represent air cargo and mail throughput, total import and export trade, GDP, size of resident population, and total employment, respectively, at that time, whereas when total import and export trade is the dependent variable, Xit, Yit, Zit, Wit, and Eit represent total import and export trade, air cargo and mail throughput, GDP, size of resident population, and total employment, respectively, at that time, etc. The μt represents the year fixed effects to control for the effects of macroeconomic or political shocks, and εit is the disturbance term, which is assumed to be independently distributed across cities with zero mean. Heteroskedasticity-robust standard errors are also used in the model analysis to control for the cross-sectional heteroskedasticity problem.
In order to further study the impact of air cargo on different sectors in the region, this paper continues to replace the total number of employment in the basic model with three indicators: employment in the primary industry, employment in the secondary industry, and employment in the tertiary industry. In China, the primary sector refers to agriculture, forestry, animal husbandry, and fishery (excluding agriculture, forestry, animal husbandry, and fishery services); the secondary sector refers to mining (excluding mining auxiliary activities), manufacturing (excluding metal products, machinery, and equipment repair), electricity, heat, gas, and water production and supply, and construction; the tertiary sector, i.e., services, refers to industries other than the primary and secondary sectors. At this point, the set of VAR model variables in the basic model is expanded from the original (air cargo, total import and export trade, GDP, size of resident population, total employment) to air cargo, total import and export trade, GDP, size of resident population, employment in primary industry, employment in secondary industry, employment in tertiary industry. For the convenience of description, in this paper, the basic model is referred to as model 1, and the extended model is referred to as model 2; GMM estimation is performed for these two models separately.
Prior to Granger analysis, it is necessary to check the data series for the presence of time trends that may lead to unreliable results of the Granger test. In order to eliminate possible unit roots and to achieve temporal smoothness, all variables in this paper were logged and differenced; this means that the analysis in this paper was performed in terms of growth rates, ensuring that the endogenous and exogenous growth factors of the indicators of interest could also be analyzed by comparing the Granger causality between growth rates. In addition, due to the short period of the study (2006–2019), all variables were set with a time lag of two years.
According to the definition of Granger causality, a smooth time series, Y, is said to “cause” another smooth time series, X, if, under the assumption that all other information is uncorrelated, the inclusion of lagged values of Y significantly reduces the variance of the prediction error of X. We combine the definition of Granger causality with the definition of endogenous and exogenous growth models of Romer (1986) and Solow (1957), for the purpose of this paper. This means that, if the growth rate of the air cargo indicator is taken as the dependent variable and the growth rates of regional economic indicators as the independent variables, after controlling for the lagged value of the growth rate of air cargo and the growth rates of regional economic indicators in the same period, the Granger causality derived from the lag of the growth rates of regional economic indicators to the growth rate of air cargo is found to be significantly positive, then regional economic growth is considered to be an endogenous growth factor of air cargo. The opposite is true for exogenous growth factors. The growth rate of regional economic indicators was used as the dependent variable, and the growth rate of air transport as the independent variable.

5. Empirical Results

5.1. The Results of Model 1

Before analyzing the data, the relationship between the variables can be understood through a bivariate scatter plot, as shown in Figure 1, Figure 2, Figure 3 and Figure 4. The gateway cities in the figure are the 35 large- and medium-sized cities consisting of Chinese municipalities, sub-provincial cities, and other provincial capitals. If the city in the sample is among these 35 large- and medium-sized cities, the city is ranked as a gateway city; otherwise it is defined as a non-gateway city. For the convenience of description, the total import and export trade in the figures and tables of this paper will be expressed as Trade, the size of the resident population as Population, and the total employment as Employment. The results presented in Figure 1, Figure 2, Figure 3 and Figure 4 show that China’s air cargo is mainly distributed in gateway cities, and air cargo seems to show a more obvious contemporaneous correlation with total import and export trade and total employment.
Table 3 and Table 4 show the results of the multivariate VAR and bivariate VAR estimation of model 1. The bivariate estimation results were used to test for the absence of causality in the multivariate estimation results. The key information related to air cargo is mainly presented here. The small sample correction proposed by Windmeijer (2005) [29] was implemented during the analysis. Since heteroskedasticity-robust standard errors were used in the data analysis, the Hansen test volume was chosen to test the validity of the instrumental variables. As can be seen from the results in Table 3 and Table 4, the null hypothesis that the instrumental variables are valid is accepted, except for the Hansen test volume in the bivariate estimation “air cargo to Population” direction, which rejects the null hypothesis at the 10% level. Additionally, from the results of the Arellano–Bond test (AB test) without second-order autocorrelation in the first differential equation perturbation term, it can be found that the test fully accepts the null hypothesis.
According to the results of the multivariate estimation of model 1, it can be found that in the direction of air cargo to the regional economy, there is no significant Granger causality, apart from in the direction of air cargo to the size of the resident population; there is also no significant Granger causality in this direction in the bivariate estimation. Air cargo shows a strong Granger causality to import and export trade, and a weak causality to GDP and total employment. In the direction of regional economy to air cargo, both multivariate and bivariate estimation results have Granger causality only in the direction of GDP to air cargo, and both Wald coefficients reject the null hypothesis that the coefficients are jointly equal to zero at the 1% level, which means that this direction has strong Granger causality.
Table 3. Model 1 multivariate VAR estimation results.
Table 3. Model 1 multivariate VAR estimation results.
VariablesAir Cargo–TradeAir Cargo–GDPAir Cargo–PopulationAir Cargo–Employment
Air Cargo to Regional Economy
Lag10.036 **0.003 *−0.001−0.003
(2.38)(1.72)(−0.85)(−0.77)
Lag2−0.022−0.002−0.001−0.006 *
(−1.62)(−0.86)(−0.55)(−1.82)
Wald test (p-level)0.0460.2210.5680.182
Hansen test (p-level)0.6150.4200.1890.525
AB test (p-level)0.9180.0860.4690.624
Regional Economy to Air Cargo
Lag10.0100.423 **0.018−0.222
(0.26)(2.21)(0.05)(−1.31)
Lag20.0430.509 ***−0.7080.126
(1.04)(2.60)(−1.45)(0.70)
Wald test (p-level)0.5850.0010.1720.394
Hansen test (p-level)0.3190.3190.3190.319
AB test (p-level)0.2190.2190.2190.219
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Table 4. Bivariate VAR estimation results for model 1.
Table 4. Bivariate VAR estimation results for model 1.
VariablesAir Cargo–TradeAir Cargo–GDPAir Cargo–PopulationAir Cargo–Employment
Air Cargo to Regional Economy
Lag10.034 **0.002−0.000−0.003
(2.04)(1.11)(−0.53)(−0.94)
Lag2−0.026−0.002−0.001−0.006 **
(−1.61)(−1.15)(−0.50)(−2.22)
Wald test (p-level)0.0790.3950.7560.067
Hansen test (p-level)0.2050.2940.0910.611
AB test (p-level)0.8640.1410.8750.303
Regional Economy to Air Cargo
Lag10.0250.319 ***0.457−0.163
(0.72)(2.63)(1.01)(−0.78)
Lag20.0250.221−0.5140.153
(0.69)(1.37)(−1.06)(0.80)
Wald test (p-level)0.5800.0090.5350.674
Hansen test (p-level)0.2830.4750.3180.394
AB test (p-level)0.3470.1250.2290.230
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

5.2. The Results of Model 2

The estimated results of model 2 are shown in Table 5 and Table 6. The estimated results between air cargo and total import and export trade, GDP, and size of resident population in the bivariate VAR estimation are consistent with the bivariate estimation results of model 1; thus, only the estimated results between air cargo and primary employment, secondary employment, and tertiary employment are presented. Consistent with model 1, the bivariate estimation results were only used to test the direction of no causality in the multivariate estimation results.
In model 2, in the direction of air cargo to regional economy, air cargo has a significant Granger causality to import and export trade and secondary employment only, and, as seen in the Wald test, air cargo has strong Granger causality to these two variables. Compared to model 1, the weak Granger causality in the direction of air cargo to GDP disappears in model 2. In the direction of regional economy to air cargo, consistent with model 1, only GDP has a significant Granger causality to air cargo, but the coefficient and significance of this result are decreased compared to model 1.
In a nutshell, in terms of scope, air cargo has a broader impact on the regional economy, but in terms of coefficients, the regional economy has a stronger influence on air cargo. Finally, based on the positivity and negativity of the correlation coefficients, it can be found that GDP growth is an endogenous growth factor for air cargo, and air cargo growth is an exogenous growth factor for total employment and secondary employment. The one-year lagged coefficient of air cargo on total import and export trade in model 2 is significantly positive, and the two-year lagged coefficient is significantly negative, making it unclear whether air cargo growth is an endogenous or exogenous growth factor for total import and export trade; however, in terms of coefficient values and significance, it seems to be more supportive of the endogenous growth theory.
Table 5. Model 2 multivariate VAR estimation results.
Table 5. Model 2 multivariate VAR estimation results.
VariablesAir Cargo–
Trade
Air Cargo–
GDP
Air Cargo–
Population
Air Cargo–Primary
Employment
Air Cargo–Secondary
Employment
Air Cargo–Tertiary
Employment
Air Cargo to Regional Economy
Lag10.037 **0.003−0.0000.0210.006−0.004
(2.51)(1.46)(−0.48)(0.91)(1.34)(−1.10)
Lag2−0.020 *−0.001−0.001−0.005−0.013 ***0.001
(−1.77)(−0.56)(−0.58)(−0.27)(−3.10)(0.44)
Wald test (p-level)0.0230.2940.7580.6570.0040.541
Hansen test (p-level)0.6060.4400.4220.3890.3270.230
AB test (p-level)0.8200.1830.4620.2330.6370.682
Regional Economy to Air Cargo
Lag10.0270.316 **0.055−0.051−0.007−0.026
(0.70)(2.02)(0.13)(−1.24)(−0.06)(−0.13)
Lag20.0410.334 *−0.7420.0590.0620.283
(0.84)(1.78)(−1.40)(1.13)(0.39)(1.55)
Wald test (p-level)0.5810.0150.2510.3290.9160.294
Hansen test (p-level)0.4850.4850.4850.4850.4850.485
AB test (p-level)0.2350.2350.2350.2350.2350.235
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Table 6. Bivariate VAR estimation results for model 2.
Table 6. Bivariate VAR estimation results for model 2.
VariablesAir Cargo–Primary EmploymentAir Cargo–Secondary EmploymentAir Cargo–Tertiary Employment
Air Cargo to Regional Economy
Lag10.0120.005−0.003
(0.56)(0.98)(−0.80)
Lag2−0.009−0.013 ***−0.001
(−0.52)(−3.19)(−0.43)
Wald test (p-level)0.7830.0030.262
Hansen test (p-level)0.3370.3680.795
AB test (p-level)0.1240.4830.365
Regional Economy to Air Cargo
Lag1−0.051−0.026−0.034
(−1.57)(−0.21)(−0.29)
Lag20.0680.1170.152
(1.35)(0.76)(1.10)
Wald test (p-level)0.2130.5130.540
Hansen test (p-level)0.6050.3160.440
AB test (p-level)0.2760.1460.084
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

6. Robustness Analysis and Discussion

In order to perform a robustness analysis on the estimation results presented in the previous section to exclude the effects of outliers, we chose to remove each of the seven geographic regions in turn, and then re-run the multivariate VAR estimation and bivariate VAR estimation for model 1 and model 2 using the Arellano–Bond two-step system GMM estimator. The results are shown in Table 7, Table 8, Table 9, Table 10, Table 11 and Table 12, where Table 7, Table 8 and Table 9 show the robustness analysis results of air cargo to regional economy direction, and Table 10, Table 11 and Table 12 are the robustness analysis results of regional economy to air cargo direction. Once again, in order to save space, only the key information related to air cargo is listed here, and the results which are all insignificant when excluding any of the regions are also omitted.

6.1. Air Cargo to Regional Economy Direction

In the multivariate VAR estimation results of air cargo to regional economy direction of model 1 (Table 7), it can be found that the one-year lagged coefficient of air cargo to total import and export trade direction increases in significance after excluding Northeast China, decreases in significance after excluding East China, and Granger causality becomes no longer significant after excluding Northwest China. This phenomenon seems to be more evident in the estimation results of model 2 (Table 8). The Granger causality of this direction in model 2 is no longer significant after excluding either East China or Northwest China, while the significance and coefficients increase after excluding Northeast China. It seems that the Northeast China data in the sample reduce the significance of the full sample, while the East China and Northwest China data increase the significance of the full sample. However, the Granger causality in this direction remains significant in the bivariate VAR estimation results, even after excluding Northwest China (Table 9), and the Wald test rejects the null hypothesis at the 10% level; therefore, the estimation results for this direction in the full sample cannot be rejected.
The analysis of the scatter plots of total import and export trade and air cargo volume in Northeast, East, and Northwest China (Figure 5, Figure 6 and Figure 7) shows that the import and export trade is higher in port cities, and has a significant correlation with air cargo volume, but the proportion of port cities in Northeast China is low. It also reveals that air cargo volume is also dispersed by non-major import/export trade cities at the same time, thus reducing the full sample significance in this direction. The northwest region, being deeply inland, relies to a greater extent on air transport for import and export trade, which in turn enhances the significance in this direction for the full sample. From this perspective, air cargo hubs seem to be more appropriate in regions with more port cities and that are deep inland, where air transport services are often a key element of the transport system in remote areas, despite the relatively low share of air cargo in deep inland areas [30]. As Appold and Kasarda (2013) argued, air cargo is not only a facilitator of existing trade, but should also be considered a trade generator, as it allows markets to be connected that would not otherwise be so [31].
Table 7. Robustness test of model 1—multivariate VAR estimation of air cargo to regional-economy-excluded regions.
Table 7. Robustness test of model 1—multivariate VAR estimation of air cargo to regional-economy-excluded regions.
RegionNortheast ChinaEast ChinaCentral ChinaNorth ChinaSouth ChinaNorthwest ChinaSouthwest China
VariablesTradeTradeTradeTradeTradeTradeTrade
Lag1 Air Cargo0.047 ***0.037 *0.041 **0.037 **0.037 **0.0230.028 **
(2.71)(1.97)(2.62)(2.13)(2.36)(1.60)(2.24)
Lag2 Air Cargo−0.020−0.029−0.025 **−0.021−0.023 *−0.021−0.014
(−1.28)(−1.51)(−2.01)(−1.51)(−1.74)(−1.44)(−1.54)
Wald test (p-level)0.0290.0980.0170.0880.0400.1900.021
Hansen test (p-level)0.2290.8690.7720.7880.7750.7540.606
AB test (p-level)0.8490.8510.8350.7900.9290.9490.863
VariablesGDPGDPGDPGDPGDPGDPGDP
Lag1 Air Cargo0.0010.0060.004 *0.004 *0.003*0.0050.004
(0.33)(1.44)(1.85)(1.85)(1.77)(1.57)(1.56)
Lag2 Air Cargo−0.002−0.003−0.001−0.002−0.0020.001−0.003
(−0.68)(−0.80)(−0.47)(−1.23)(−1.02)(0.36)(−1.00)
Wald test (p-level)0.7550.2890.1810.1310.1940.1670.268
Hansen test (p-level)0.3290.6520.3510.5470.6400.2580.327
AB test (p-level)0.1990.0740.0850.1090.0800.0870.103
VariablesEmploymentEmploymentEmploymentEmploymentEmploymentEmploymentEmployment
Lag1 Air Cargo−0.003−0.006−0.001−0.003−0.004−0.003−0.000
(−0.97)(−1.21)(−0.28)(−0.75)(−1.01)(−0.72)(−0.15)
Lag2 Air Cargo−0.007 **−0.007−0.006 *−0.006 *−0.007 *−0.006−0.006 *
(−2.28)(−1.59)(−1.94)(−1.72)(−1.96)(−1.44)(−1.89)
Wald test (p-level)0.0680.2190.1570.2140.1170.3240.160
Hansen test (p-level)0.1880.7300.5340.3310.5350.4620.505
AB test (p-level)0.5690.3840.4590.4840.4070.5000.649
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Table 8. Robustness test of model 2—multivariate VAR estimation of air cargo to regional-economy-excluded regions.
Table 8. Robustness test of model 2—multivariate VAR estimation of air cargo to regional-economy-excluded regions.
RegionNortheast ChinaEast ChinaCentral ChinaNorth ChinaSouth ChinaNorthwest
China
Southwest China
VariablesTradeTradeTradeTradeTradeTradeTrade
Lag1 Air Cargo0.045 ***0.0320.039 **0.033 *0.029 *0.0210.024 *
(2.91)(1.35)(2.40)(1.86)(1.84)(1.54)(1.91)
Lag2 Air Cargo−0.017−0.034 *−0.023 *−0.019−0.021 *−0.018−0.020 *
(−1.38)(−1.69)(−1.88)(−1.58)(−1.84)(−1.45)(−1.83)
Wald test (p-level)0.0160.1280.0210.0900.0630.1780.020
Hansen test (p-level)0.3800.2490.3630.2510.3700.4320.174
AB test (p-level)0.6800.7290.8110.8700.7550.6810.692
VariablesGDPGDPGDPGDPGDPGDPGDP
Lag1 Air Cargo0.0010.0040.0040.005 *0.0030.0040.004
(0.51)(1.18)(1.53)(1.94)(1.21)(1.59)(1.64)
Lag2 Air Cargo−0.002−0.002−0.002−0.002−0.0020.002−0.003
(−0.72)(−0.56)(−0.68)(−0.65)(−0.65)(0.99)(−1.16)
Wald test (p-level)0.6850.4000.2710.1420.4440.1280.218
Hansen test (p-level)0.3190.3120.4710.2210.2510.2640.276
AB test (p-level)0.2420.1240.1720.1290.1890.1910.193
VariablesSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary Employment
Lag1 Air Cargo0.0020.0070.0080.0060.0060.011 *0.007
(0.57)(1.09)(1.54)(1.12)(1.22)(1.74)(1.33)
Lag2 Air Cargo−0.011 ***−0.015 **−0.014 ***−0.013 ***−0.012 ***−0.016 ***−0.014 ***
(−2.96)(−2.55)(−3.09)(−2.86)(−2.85)(−3.01)(−2.93)
Wald test (p-level)0.0090.0160.0020.0140.0070.0040.011
Hansen test (p-level)0.3400.2980.3760.4640.2470.3860.307
AB test (p-level)0.7510.7730.6300.8270.5930.6720.574
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Table 9. Robustness test of bivariate VAR estimation of air cargo to regional-economy-excluded regions.
Table 9. Robustness test of bivariate VAR estimation of air cargo to regional-economy-excluded regions.
RegionNortheast ChinaEast ChinaCentral ChinaNorth ChinaSouth ChinaNorthwest ChinaSouthwest China
VariablesTradeTradeTradeTradeTradeTradeTrade
Lag1 Air Cargo0.044 ***0.039 *0.038 **0.034 **0.037 **0.032 **0.026 *
(2.70)(1.70)(2.02)(2.01)(2.24)(2.11)(1.60)
Lag2 Air Cargo−0.029 *−0.037−0.029 *−0.024−0.027 *−0.019−0.023 **
(−1.72)(−1.64)(−1.85)(−1.43)(−1.79)(−1.17)(−2.22)
Wald test (p-level)0.0240.1320.0760.1080.0440.0960.020
Hansen test (p-level)0.2310.3790.2060.3560.2680.1670.237
AB test (p-level)0.8010.8670.9170.9340.9720.8840.820
VariablesGDPGDPGDPGDPGDPGDPGDP
Lag1 Air Cargo−0.0000.0030.0030.004*0.0020.0030.004
(−0.12)(0.84)(1.23)(2.13)(1.10)(0.97)(1.17)
Lag2 Air Cargo−0.001−0.003−0.002−0.002−0.0020.000−0.004
(−0.63)(−1.17)(−0.97)(−1.21)(−1.22)(0.25)(−1.60)
Wald test (p-level)0.8080.3670.3810.1820.3600.5630.253
Hansen test (p-level)0.2900.3360.3050.3980.2560.2230.171
AB test (p-level)0.3090.1250.1400.1570.1810.2850.161
VariablesEmploymentEmploymentEmploymentEmploymentEmploymentEmploymentEmployment
Lag1 Air Cargo−0.005−0.007−0.003−0.003−0.005−0.002−0.000
(−1.20)(−1.60)(−0.64)(−0.80)(−1.19)(−0.46)(−0.13)
Lag2 Air Cargo−0.005 *−0.008 **−0.006 **−0.008 **−0.007 **−0.006 *−0.007 *
(−1.83)(−2.29)(−2.05)(−2.57)(−2.24)(−1.81)(−1.98)
Wald test (p-level)0.0890.0530.1180.0290.0440.1900.142
Hansen test (p-level)0.6140.6890.4590.4840.6890.3310.545
AB test (p-level)0.4630.2490.3060.2450.3760.2390.325
VariablesSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary EmploymentSecondary Employment
Lag1 Air Cargo0.0030.0040.0080.0030.0040.0070.004
(0.52)(0.42)(1.50)(0.48)(0.77)(1.31)(0.76)
Lag2 Air Cargo−0.010 ***−0.013−0.013 ***−0.012 ***−0.012 **−0.013 ***−0.013 ***
(−2.72)(−1.59)(−2.93)(−2.96)(−2.63)(−3.04)(−2.64)
Wald test (p-level)0.0200.1100.0050.0100.0180.0070.016
Hansen test (p-level)0.1340.2300.2900.1450.1460.1410.283
AB test (p-level)0.7520.4660.4080.6990.5500.4930.242
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Figure 5. Scatter plot of air cargo and Trade in Northeast China.
Figure 5. Scatter plot of air cargo and Trade in Northeast China.
Sustainability 14 09753 g005
Figure 6. Scatter plot of air cargo and Trade in East China.
Figure 6. Scatter plot of air cargo and Trade in East China.
Sustainability 14 09753 g006
Figure 7. Scatter plot of air cargo and Trade in Northwest China.
Figure 7. Scatter plot of air cargo and Trade in Northwest China.
Sustainability 14 09753 g007
It can also be seen in Table 7, Table 8 and Table 9 that the two-year lagged coefficients of the direction of air cargo to import and export trade are not significant after excluding the northwest region; therefore, we can reject the hypothesis that there is Granger causality of the two-year lagged growth coefficient of air cargo to the import and export trade after excluding the northwest region.
In the direction of air cargo to GDP, the one-year lagged coefficients are significant in model 1 only after excluding Central, North, and South China, and in model 2 after excluding North China, and the Wald test results support the null hypothesis. The bivariate estimates also have significant Granger causality in this direction, after excluding North China. This means that the weak Granger causality in this direction in the full sample results holds only after excluding North China, but it seems that the full sample results should not be rejected due to the individual region. The reason for such anomalies in North China seems to be due to the out-migration of industries from Beijing since 2008 [32].
In the multivariate VAR estimation of air cargo to total employment direction, the one-year lagged coefficient is no longer significant after excluding East China and Northwest China, and the Wald test results indicate that the null hypothesis can be rejected only after excluding Northeast China, while the one-year lagged coefficients in the bivariate VAR estimation are all significant. Therefore, the weak Granger causality in this direction cannot be rejected in the full sample. The reason for the inter-regional differences observed here is considered to be similar to that for the direction of air cargo to import and export trade; i.e., where air cargo volume has a more significant effect on total employment in regions with more ports or that are deep inland, while air cargo volume has less effect on total employment in regions with fewer port cities. The results of the Wald test, presented in Table 9, show that after excluding Central China, Northwest China, and Southwest China, respectively, the Granger causality in this direction is weaker. Furthermore, the percentage stacking of regional city categories in Figure 8 shows that cities in Central China, Northwest China, and Southwest China are almost all inland, which further confirms that inland regions are more dependent on air transport.
The estimates of the directions of air cargo to size of resident population, primary industry employment, secondary industry employment, and tertiary industry employment are all consistent with the full sample results. There is no significant Granger causality in the directions of air cargo to size of resident population, primary industry employment, or tertiary industry employment, while there is very strong Granger causality in the direction of air cargo to secondary industry employment, after excluding any region with negative coefficients, and it can be said that the full sample results for this direction are very robust.
Overall, of the significant Granger causality in the full sample results of the air cargo to regional economy direction, there are varying degrees of inter-regional variability in all results, except for a strong causality effect on secondary employment, which is very robust. However, none of the full sample results are rejected except for the two-year lagged causality of air cargo growth to import and export trade growth. The results of the analysis of the regions show that the regional economy of the inland areas and port cities are more dependent on air cargo.

6.2. Regional Economy to Air Cargo Direction

In the full sample results in the direction of regional economy to air cargo, only GDP growth has a strong Granger causality effect on air cargo growth. However, the results of the robustness test (Table 10, Table 11 and Table 12) show that the Granger causality is no longer significant after the exclusion of the northwest region, while the results of the bivariate estimation showed that the Wald test support the null hypothesis that the coefficient is jointly equal to zero after the exclusion of the northwest region in this direction, although the one-year lagged coefficient is significant at the 10% level. That is, with the exclusion of the northwest region, the growth of GDP has only a weak Granger causality effect on the growth of air cargo; thus, we reject the hypothesis of strong Granger causality in this direction in the full sample results. In Figure 9, it can also be observed that although the air cargo volume is lower in the northwest, it is overly concentrated in gateway cities compared to other regions.
Table 10. Robustness test of model 1—multivariate VAR estimation of regional economy to air-cargo-excluded regions.
Table 10. Robustness test of model 1—multivariate VAR estimation of regional economy to air-cargo-excluded regions.
RegionNortheast ChinaEast ChinaCentral ChinaNorth ChinaSouth ChinaNorthwest ChinaSouthwest China
VariablesAir CargoAir CargoAir CargoAir CargoAir CargoAir CargoAir Cargo
Lag1 GDP0.523 ***0.299 **0.473 ***0.391 **0.302 **0.3070.410 ***
(2.68)(2.23)(3.41)(2.52)(2.46)(1.53)(2.69)
Lag2 GDP0.3170.461 ***0.375 **0.340 *0.393 **0.1720.384 *
(1.12)(3.05)(2.10)(1.89)(2.31)(0.80)(1.90)
Lag1 Employment−0.196−0.104−0.182−0.166−0.228−0.154−0.488 **
(−1.06)(−0.54)(−0.93)(−0.81)(−1.11)(−0.81)(−2.31)
Lag2 Employment0.169−0.0140.1620.0760.1230.1470.270
(0.72)(−0.08)(0.80)(0.37)(0.59)(0.67)(0.77)
Wald test (p-level)
GDP
0.0070.0010.0000.0040.0070.0490.007
Wald test (p-level)
Employment
0.5050.8630.5700.7150.5270.5770.061
Hansen test (p-level)0.2920.8310.3940.4540.3200.2350.122
AB test (p-level)0.2940.8910.1420.1110.2530.3960.256
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Table 11. Robustness test of model 2—multivariate VAR estimation of regional economy to air-cargo-excluded regions.
Table 11. Robustness test of model 2—multivariate VAR estimation of regional economy to air-cargo-excluded regions.
RegionNortheast ChinaEast ChinaCentral ChinaNorth ChinaSouth ChinaNorthwest ChinaSouthwest China
VariablesAir CargoAir CargoAir CargoAir CargoAir CargoAir CargoAir Cargo
Lag1 GDP0.475 *0.1820.405 ***0.306 *0.2440.2190.370 **
(1.95)(1.24)(2.77)(1.87)(1.54)(1.42)(2.06)
Lag2 GDP0.3230.407 **0.345 *0.347 *0.391 **0.1460.314 *
(1.18)(2.37)(1.83)(1.80)(2.08)(0.91)(1.68)
Lag1 Tertiary Employment0.0440.067−0.0710.106−0.032−0.050−0.313 *
(0.18)(0.31)(−0.37)(0.48)(−0.17)(−0.27)(−1.76)
Lag2 Tertiary Employment0.2640.2150.3320.2320.294 *0.2680.280
(1.44)(1.26)(1.65)(1.15)(1.72)(1.37)(0.98)
Observations104585811001089106710781023
Number of id957810099979893
Year FEYESYESYESYESYESYESYES
Wald test (p-level)
GDP
0.0410.0310.0070.0370.0320.2270.038
Wald test (p-level)
Tertiary Employment
0.2740.3520.2560.2790.2290.3820.213
Hansen test (p-level)0.2010.3510.2670.2820.4560.2790.206
AB test (p-level)0.3430.2120.1900.1410.2920.3500.226
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Table 12. Robustness test of Bivariate VAR estimation of regional economy to air-cargo-excluded regions.
Table 12. Robustness test of Bivariate VAR estimation of regional economy to air-cargo-excluded regions.
RegionNortheast ChinaEast ChinaCentral ChinaNorth ChinaSouth ChinaNorthwest ChinaSouthwest China
VariablesAir CargoAir CargoAir CargoAir CargoAir CargoAir CargoAir Cargo
Lag1 GDP0.458 *0.2180.366 ***0.326 **0.239 **0.236 *0.303 **
(1.76)(1.64)(2.92)(2.50)(2.23)(1.84)(2.04)
Lag2 GDP0.2610.318 **0.2160.294 *0.2170.0590.362 **
(0.77)(2.39)(1.31)(1.68)(1.33)(0.38)(2.02)
Wald test (p-level)0.0490.0190.0050.0150.0090.1490.034
Hansen test (p-level)0.1720.7050.3510.3640.5990.6030.226
AB test (p-level)0.1830.6850.1340.0970.1230.2250.293
Lag1 Employment−0.139−0.109−0.140−0.147−0.180−0.064−0.416 **
(−0.50)(−0.62)(−0.56)(−0.62)(−0.80)(−0.29)(−2.41)
Lag2 Employment0.1600.1220.1700.0760.1430.0890.301 *
(0.78)(0.67)(0.81)(0.35)(0.65)(0.43)(1.67)
Wald test (p-level)0.7350.7250.7100.8210.7060.9030.060
Hansen test (p-level)0.6580.7160.4000.1730.3370.3470.330
AB test (p-level)0.3210.9620.1750.1480.1950.2760.205
Lag1 Secondary Employment0.0090.0480.0000.091−0.0600.037−0.038
(0.06)(0.42)(0.00)(0.65)(−0.48)(0.32)(−0.37)
Lag2 Secondary Employment−0.0180.0170.144−0.0350.1510.0610.224 **
(−0.11)(0.14)(0.83)(−0.21)(0.99)(0.38)(2.14)
Wald test (p-level)0.9930.7610.3440.6350.5270.5590.062
Hansen test (p-level)0.4870.5150.1760.1480.3740.2050.282
AB test (p-level)0.2510.8190.1270.0820.1420.2300.115
Lag1 Tertiary Employment−0.096−0.001−0.0740.099−0.046−0.057−0.241 *
(−0.70)(−0.01)(−0.55)(0.68)(−0.36)(−0.53)(−1.75)
Lag2 Tertiary Employment0.1490.1700.2430.0900.1920.2050.127
(0.89)(1.12)(1.43)(0.63)(1.37)(1.61)(0.76)
Wald test (p-level)0.6460.5110.3560.4370.3600.2760.221
Hansen test (p-level)0.5630.6610.1650.1480.5330.3940.347
AB test (p-level)0.1380.5430.0620.1400.0890.1370.158
Note: t-statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.
Figure 9. Box Diagram of air cargo:(a) Northwest China; (b) excluding Northwest China.
Figure 9. Box Diagram of air cargo:(a) Northwest China; (b) excluding Northwest China.
Sustainability 14 09753 g009
Surprisingly, in the results presented in Table 10, the Granger causality in the direction of total employment to air cargo is significantly negative after excluding the southwest region, and the Wald test results lead us to reject the null hypothesis at the 10% level. In the results of Table 11, the same Granger causality is observed in the direction of tertiary employment to air cargo, but the Wald test shows a weak causality at this point. Meanwhile, in the results of Table 12, the Granger causality in the direction of secondary employment to air cargo is significantly positive, and the Wald test results also cause us to reject the null hypothesis at the 10% level. That is, after excluding the southwest, total employment growth, secondary employment growth, and tertiary employment growth are all Granger causes of air cargo growth, rejecting the results for all three directions in the full sample, and coefficient values show that the effect of employment on air cargo is greater than the effect of air cargo on employment.
It can also be seen from the log scatter plots in Figure 10 and Figure 11 that the log scatter plots between air cargo and total employment and secondary employment in the southwest region both show greater volatility, and less volatility after excluding the southwest region, especially with respect to secondary employment.
In brief, the strong Granger causality of the full sample results for the direction of GDP to air cargo is rejected due to the overdependence of air cargo on GDP in the northwest increasing the significance of the full sample. In contrast, total employment, secondary employment, and tertiary employment in the full sample without significant Granger causality in the direction to air cargo are Granger causal after excluding the southwest region. The greater volatility between employment growth and air cargo growth in the southwest reduces the significance of the full sample.

7. Conclusions

This paper examines the Granger causality between air cargo growth and regional economic development for 110 prefecture-level cities in mainland China over the period 2006–2019 using the Arellano–Bond two-step system GMM estimator, and attempts to account for the endogenous relationship between the two. Unlike previous Granger causality tests between air transport and regional economic development, this paper employs a multivariate VAR framework to reduce the impact of possible spurious causality from bivariate tests, and to test for the absence of causality in the multivariate estimates based on the bivariate VAR estimation results. Although the robustness analysis performed by excluding different geographical regions in turn reveals that some of the Granger causality is not reliable for the full sample results due to strong inter-regional variability, we can still draw some conclusions, as follows.
From the results of the full sample, in the direction of the regional economy to air cargo, only the growth of GDP has strong Granger causality to air cargo growth; however, this result is rejected in the robustness analysis because the overdependence of air cargo on GDP in the northwest increases the significance of the full sample. Instead, total employment and secondary employment growth have strong Granger causality to air cargo growth, and tertiary employment growth has weak Granger causality to air cargo growth, albeit after excluding the southwest region.
The full sample results of the direction of air cargo to the regional economy show strong Granger causality of air cargo growth to the growth of total import and export trade, and the growth of secondary employment, and weak Granger causality to the growth of GDP and total employment. None of the results for the full sample are rejected in the robustness analysis in this direction, except for the negative impact of air cargo growth to import and export trade growth with a two-year lag. Overall, although it seems that the regional economy has a stronger impact on air cargo than air cargo has on the regional economy in terms of coefficient values, the results of the robustness analysis show that air cargo has a more reliable impact on the regional economy, especially on the development of the secondary sector.
From the perspective of the distribution between air cargo and regional economy, China’s air cargo is still mainly distributed in 35 large- and medium-sized cities, while the economic development of inland cities and port cities is more dependent on air cargo.
Finally, in terms of the positivity and negativity of the coefficients in the direction with strong Granger causality, the lagged coefficient of air cargo is significantly positive for the growth of total import and export trade, and the lagged coefficient of secondary employment is significantly positive for the growth of air cargo, both of which are more supportive of the endogenous growth model. The lagged coefficient of total employment on air cargo growth, and the lagged coefficient of air cargo on secondary industry employment growth, are significantly negative, which seems to be more supportive of the exogenous growth theory. From such results, it seems that policies regarding air cargo development need to be more favorable to regions with faster secondary industry employment growth, as well as regions that are in urgent need of import and export trade development.
China’s air cargo industry is currently lagging significantly behind passenger air transport, as well as other modes of freight transportation; nevertheless, the development of the air cargo industry can certainly support China’s industrial upgrading as it continues to transform its industrial chain. On a national and regional level, the planning of air cargo hubs needs to address how to attract this type of transportation and what it can bring. The conclusions of this paper are not conclusive regarding the causal relationship between air cargo and regional economic development, but rather show that air cargo does have a positive causal relationship with regional economic development, particularly secondary employment and import and export trade. We hope that our study will motivate national and local governments to be more prudent in their air cargo hub planning. Of course, it is not enough to measure the relationship between air cargo industry and regional economic development through panel Granger causality assessment alone; this paper is just the beginning. Later, we will use various statistical and econometric methods to analyze the relationship between air cargo and regional economic development from the supply-side and demand-side perspectives, and to analyze the development direction of China’s air cargo industry from a more intuitive perspective. We hope that the research in this paper will help China’s air cargo to develop in a more sustainable direction.

Author Contributions

Conceptualization, J.Z.; Data curation, J.Z. and L.L.; Formal analysis, J.Z.; Investigation, J.Z.; Methodology, J.Z.; Resources, J.Z. and L.L.; Software, J.Z. and L.L.; Supervision, Q.Y. and X.S.; Validation, J.Z.; Visualization, J.Z.; Writing—original draft, J.Z.; Writing—review & editing, J.Z., Q.Y. and X.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

Publicly available datasets were analyzed in this study. This data can be downloaded from the official website of the Civil Aviation Administration of China and the official website of the China Statistics Bureau, here: (http://www.caac.gov.cn/index.html, http://www.stats.gov.cn/) (accessed on 1 June 2022). Part of the data were obtained from the local statistical office.

Conflicts of Interest

The authors declare that they have no conflict of interest regarding the publication of this paper.

References

  1. Debbage, K.G. Air transportation and urban-economic restructuring: Competitive advantage in the US Carolinas. J. Air Transp. Manag. 1999, 5, 211–221. [Google Scholar] [CrossRef] [Green Version]
  2. Matha, S.G.; Goldstein, P.; Lu, J. Air transportation and regional economic development: A case study for the new airport in south albania. CID Work. Pap. 2020, 56. [Google Scholar]
  3. Hausmann, R.; Rodrik, D. Economic development as self-discovery. J. Dev. Econ. 2003, 72, 603–633. [Google Scholar] [CrossRef] [Green Version]
  4. Gebreeyesus, M.; Iizuka, M. Discovery of the flower industry in Ethiopia: Experimentation and coordination. J. Glob. Dev. 2012, 2, 42. [Google Scholar] [CrossRef] [Green Version]
  5. Button, K.J. The economist’s perspective on regional endogenous development. In Endogenous Regional Development; Edward Elgar Publishing: Cheltenham, UK, 2011. [Google Scholar] [CrossRef]
  6. Solow, R.M. A contribution to the theory of economic growth. Q. J. Econ. 1956, 70, 65–94. [Google Scholar] [CrossRef]
  7. Solow, R.M. Technical change and the aggregate production function. Rev. Econ. Stat. 1957, 39, 312–320. [Google Scholar] [CrossRef] [Green Version]
  8. National Development and Reform Commission; Civil Aviation Administration. Opinions of the National Development and Reform Commission Civil Aviation Administration on Promoting the Development of Air Cargo Facilities. 2020. Available online: http://www.gov.cn/zhengce/zhengceku/2020-09/04/content_5540595.htm (accessed on 1 June 2022).
  9. Romer, P.M. Increasing returns and long-run growth. J. Political Econ. 1986, 94, 1002–1037. [Google Scholar] [CrossRef] [Green Version]
  10. Romer, P.M. Endogenous technological change. J. Political Econ. 1990, 98, S71–S102. [Google Scholar] [CrossRef] [Green Version]
  11. Lucas, R.E. On the mechanics of economic development. J. Monet. Econ. 1988, 22, 3–42. [Google Scholar] [CrossRef]
  12. Bougheas, S.; Demetriades, P.O.; Mamuneas, T.P. Infrastructure, specialization, and economic growth. Can. J. Econ./Rev. Can. D’économique 2000, 33, 506–522. [Google Scholar] [CrossRef]
  13. Aschauer, D.A. Is public expenditure productive? J. Monet. Econ. 1989, 23, 177–200. [Google Scholar] [CrossRef]
  14. Munnell, A.H. Why has productivity growth declined? Productivity and public investment. N. Engl. Econ. Rev. 1990, 30, 3–22. [Google Scholar]
  15. Tatom, J.A. Public capital and private sector performance. Fed. Reserve Bank St Louis Rev. 1991, 73, 3–15. [Google Scholar] [CrossRef] [Green Version]
  16. Nourzad, F.; Vrieze, M.D. Public capital formation and productivity growth: Some international evidence. J. Product. Anal. 1995, 6, 283–295. [Google Scholar] [CrossRef]
  17. Chandra, A.; Thompson, E. Does public infrastructure affect economic activity? Evidence from the rural interstate highway system—ScienceDirect. Reg. Sci. Urban Econ. 2000, 30, 457–490. [Google Scholar] [CrossRef]
  18. Duranton, G.; Turner, M.A. Urban growth and transportation. Rev. Econ. Stud. 2012, 79, 1407–1440. [Google Scholar] [CrossRef]
  19. Duranton, G.; Morrow, P.M.; Turner, M.A. Roads and Trade: Evidence from the US. Rev. Econ. Stud. 2014, 81, 681–724. [Google Scholar] [CrossRef]
  20. Vanhoudt, P. A fallacy in causality research on growth and capital accumulation. Econ. Lett. 1998, 60, 77–81. [Google Scholar] [CrossRef]
  21. Hartwig, J. A Panel Granger-Causality Test of Endogenous vs. Exogenous Growth. SSRN Electron. J. 2009, 34. [Google Scholar] [CrossRef] [Green Version]
  22. Kulendran, N.; Wilson, K. Is there a relationship between international trade and international travel? Appl. Econ. 2000, 32, 1001–1009. [Google Scholar] [CrossRef]
  23. Mukkala, K.; Tervo, H. Air Transportation and Regional Growth: Which Way Does the Causality Run? Environ. Plan. A Econ. Space 2013, 45, 1508–1520. [Google Scholar] [CrossRef]
  24. Venet, B.; Hurlin, C. Granger Causality Tests in Panel Data Models with Fixed Coecients. Doc. De Rech. Leo 2001, 33. [Google Scholar]
  25. Van de Vijver, E.; Derudder, B.; Witlox, F. Air Passenger Transport and Regional Development: Cause and Effect in Europe. PROMET–Traffic Transp. 2016, 28, 143–154. [Google Scholar] [CrossRef] [Green Version]
  26. Baker, D.; Merkert, R.; Md, K. Regional aviation and economic growth: Cointegration and causality analysis in Australia. J. Transp. Geogr. 2015, 43, 140–150. [Google Scholar] [CrossRef]
  27. Button, K.; Yuan, J. Airfreight Transport and Economic Development: An Examination of Causality. Urban Stud. 2013, 50, 329–340. [Google Scholar] [CrossRef]
  28. Blundell, R.; Bond, S. Initial conditions and moment restrictions in dynamic panel data models. J. Econom. 1998, 87, 115–143. [Google Scholar] [CrossRef] [Green Version]
  29. Windmeijer, F. A finite sample correction for the variance of linear two-step GMM estimators. J. Econom. 2005, 126, 25–51. [Google Scholar] [CrossRef]
  30. Bråthen, S.; Halpern, N. Air transport service provision and management strategies to improve the economic benefits for remote regions. Res. Transp. Bus. Manag. 2012, 4, 3–12. [Google Scholar] [CrossRef]
  31. Appold, S.J.; Kasarda, J.D. The airport city phenomenon: Evidence from large US airports. Urban Stud. 2013, 50, 1239–1259. [Google Scholar] [CrossRef]
  32. Boyang, G.; Weidong, L.; Dunford, M. State land policy, land markets and geographies of manufacturing: The case of Beijing, China. Land Use Policy 2014, 36, 1–12. [Google Scholar] [CrossRef]
Figure 1. Scatter plot of air cargo and Trade.
Figure 1. Scatter plot of air cargo and Trade.
Sustainability 14 09753 g001
Figure 2. Scatter plot of air cargo and GDP.
Figure 2. Scatter plot of air cargo and GDP.
Sustainability 14 09753 g002
Figure 3. Scatter plot of air cargo and Population.
Figure 3. Scatter plot of air cargo and Population.
Sustainability 14 09753 g003
Figure 4. Scatter plot of air cargo and Employment.
Figure 4. Scatter plot of air cargo and Employment.
Sustainability 14 09753 g004
Figure 8. Percentage stacking chart of city categories by region.
Figure 8. Percentage stacking chart of city categories by region.
Sustainability 14 09753 g008
Figure 10. Scatter plot of ln(air cargo) and ln(Employment): (a) Southwest China; (b) excluding Southwest China.
Figure 10. Scatter plot of ln(air cargo) and ln(Employment): (a) Southwest China; (b) excluding Southwest China.
Sustainability 14 09753 g010
Figure 11. Scatter plot of ln(air cargo) and ln(secondary employment): (a) Southwest China; (b) excluding Southwest China.
Figure 11. Scatter plot of ln(air cargo) and ln(secondary employment): (a) Southwest China; (b) excluding Southwest China.
Sustainability 14 09753 g011
Table 1. Names and locations of the selected cities.
Table 1. Names and locations of the selected cities.
RegionsNumberCity Names
Northeast China15Chifeng, Dalian, Daqing, Dandong, Harbin, Heihe, Jixi, Jiamusi, Jinzhou, Mudanjiang, Qiqihar, Qinhuangdao, Shenyang, Tongliao, Changchun
East China32Anqing, Changzhou, Dongying, Fuzhou, Ganzhou, Hangzhou, Hefei, Huangshan, Jinan, Jingdezhen, Lianyungang, Linyi, Nanchang, Nanjing, Nantong, Ningbo, Qingdao, Quzhou, Quanzhou, Xiamen, Shanghai, Weihai, Weifang, Wenzhou, Wuxi, Xuzhou, Yantai, Yancheng, Huai’an, Jining, Taizhou, Zhoushan
Central China10Changde, Luoyang, Nanyang, Wuhan, Yichang, Zhangjiajie, Changsha, Zhengzhou, Huaihua, Yongzhou
North China11Baotou, Beijing, Datong, Hohhot, Shijiazhuang, Taiyuan, Tianjin, Yuncheng, Changzhi, Erdos, Handan
South China13Beihai, Guangzhou, Guilin, Liuzhou, Nanning, Shenzhen, Zhanjiang, Zhuhai, Foshan, Haikou, Jieyang, Sanya, Wuzhou
Southwest China17Baoshan, Chengdu, Guiyang, Kunming, Lijiang, Luzhou, Mianyang, Nanchong, Panzhihua, Yibin, Zhaotong, Chongqing, Dazhou, Guangyuan, Lhasa, Lincang, Pu’er
Northwest China12Jiayuguan, Lanzhou, Wuhai, Xi’an, Yan’an, Yinchuan, Yulin, Hanzhong, Karamay, Tianshui, Urumqi, Xining
Table 2. Descriptive statistics.
Table 2. Descriptive statistics.
VariableUnitObs.MeanStd. Dev.MinMax
Air cargo and mail throughputThousand1540112.713406.24004231.740
Total import and export tradeBillion1540153.203428.7390.24434,120
GDPBillion1540324.387435.6557.0173798.755
Resident populationThousand15405728.6594557.419176.131,878.4
Total employmentThousand15401616.2042673.97742.116,681.6
Primary industry employmentThousand1540193.939607.2150.0316643.5
Secondary industry employmentThousand1540623.626902.1896.55147.4
Tertiary industry employmentThousand1540788.9211467.6549.210,580.8
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Zhou, J.; Leng, L.; Yuan, Q.; Shi, X. Does Air Cargo Matter in Chinese Regional Economic Development? An Empirical Granger Causality Test. Sustainability 2022, 14, 9753. https://doi.org/10.3390/su14159753

AMA Style

Zhou J, Leng L, Yuan Q, Shi X. Does Air Cargo Matter in Chinese Regional Economic Development? An Empirical Granger Causality Test. Sustainability. 2022; 14(15):9753. https://doi.org/10.3390/su14159753

Chicago/Turabian Style

Zhou, Jun, Liang Leng, Quan Yuan, and Xiaofa Shi. 2022. "Does Air Cargo Matter in Chinese Regional Economic Development? An Empirical Granger Causality Test" Sustainability 14, no. 15: 9753. https://doi.org/10.3390/su14159753

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop