The Causality Analysis of Airports and Regional Economy: Empirical Evidence from Jiangsu Province in China

Abstract

China is the second largest aviation country in the world. The Chinese aviation industry and economy both developed quickly in the last two decades. However, the interaction mechanisms of aviation and the regional economy were different in each province. Jiangsu was the most important province in the Yangtze River delta region. The GDP of Jiangsu ranked second in China, but air transportation didn’t have the same leading position in the last decade. Taking Jiangsu province as for analysis, this paper provided empirical evidence of the causality between airport development and regional economic growth. The results showed that: (1) From 2008 to 2018, the overall volume of airport passenger throughput and GDP in Jiangsu had a strong upward trend with strong seasonal fluctuations; (2) There was a bi-directional Granger causality relationship between the airport passenger throughput and GDP in Jiangsu based on the vector autoregression (VAR) model and the Granger causality test; (3) From the impulse response and variance decomposition, the inter-contribution of GDP and airport development was sustainable and increasing over time. However, the impact of economic growth on airports was more significant than the impact of airports on economic growth.

1. Introduction

The Chinese economy has been rapidly and steadily growing since the start of the economic reform and opening up to foreign trade in the mid-1990s. The volume of GDP increased from $4984.7 billion (2009) to $13,608 billion (2018), with an average annual growth of 8.01%. Meanwhile, the aviation industry in China was booming the rapidly economic growth. China has been the second-largest aviation market in the world since 2008. The volume of airport passenger throughput in China increased from 0.486 billion (2009) to 1.264 billion (2018) and the average annual growth was 12.8%. This indicated that aviation development speed outpaced economic growth in the last ten years. (Note: all of the data was downloaded from www.stats.gov.cn (accessed on 15 July 2019) and www.caac.gov.cn (accessed on 15 July 2019).

Air transportation and regional economic growth are highly interdependent. Communities that benefit from rapid economic development tend to invest more in infrastructure and the provision of air transportation services. In turn, the availability of reliable transportation services further strengthens regional sustainable development. Therefore, a deep understanding of the interaction mechanisms between aviation and the local economy will also facilitate the airports and regional economic development. However, Wu and Man (2018) found that the correlations of regional air transportation utilization and per capita GDP and urbanization rate in each province were very different in China [1]. An example is Jiangsu province, where the economic status and air transportation scale did not match. Jiangsu province is located in the east of China in theYangtze River delta region, which has a significant economic position. The overall GDP of Jiangsu ranked the second highest among the 31 Chinese administrative provincial units. Jiangsu experienced a relatively high level of economic development, greater opportunities for businesses, and high population growth in the last decade. Though air transportation had good prospects because of the strong economic position, the total airport passenger throughput in Jiangsu ranked eighth in 2019. Although the annual airport passenger throughput of Nanjing Lukou International Airport (NKG) was more than 30 million, it was far behind the neighboring airports, including Shanghai Pudong International Airport (PVG), Shanghai Hongqiao International Airport (SHA) and Hangzhou Xiaoshan International Airport (HGH). The ratio of air passenger volume to provincial GDP of Jiangsu was in the range of [0.0035, 0.0079], which was in the lowest category of 31 provincial units in China.

Jiangsu is located in the core area of the Yangtze River delta region, which is the most important economic region in China. The interaction mechanism between the aviation and regional economy was complicated because of the different development environments. It needed to demonstrate that the correlation between the regional economy and airports is focused on the particular area.

There were few studies evaluating the correlation between airports and the regional economy of the typical Chinese province, especially ones that used econometrical tests. In this paper, the Granger causality test, variance decomposition and impulse response functions were modeled to provide a holistic picture of the relationship between the Jiangsu airports and the regional economy. Since the direction of causality could vary across regions due to different spatial, economic and social characteristics, the research results would play a key background role in the policy-making of airport development funds and regional aviation route subsidies to meet the travel demands and Jiangsu economic development goals. It also offered a means of promoting Jiangsu aviation and economic competitiveness in the Yangtze River delta region.

2. Literature Review

2.1. Interactions between Aviation and Regional Economy

Airports have been associated with four main types of economic impact: direct impact, indirect impact, induced impact, and catalytic impact [2,3]. The impacts are mainly due to the employment and income generated by different sources.

Table 1. The sources of four economic impacts.

Economic Impacts	Sources
Direct Impact	The direct construction and operation of airports.
Indirect Impact	The chain of suppliers of goods and services.
Induced Impact	The spending of income by employees created by the direct and indirect effects.
Catalytic Impact	The role of airports as a driver of productivity growth and then as an attractor of new firms.

shows the sources of the four types of impact.

Good airline services in a region contributed significantly to urban economic development [4]. An increase in flights was favorable to promoting foreign investment and improved the employment rate of a city. The empirical results showed that a 10% increase in passenger enplanements in a metro area led to 1% increase in employment rate in related service industries. Kasarda and Green (2005) examined the role of air cargo in economic development, and showed that air service liberalization, improving customs quality, and reducing corruption could enhance air cargo’s positive impact [5]. Blonigen (2012) exploited time variation in the long-term growth rate to estimate the effect of airline traffic on the local population, income, and employment growth by using the data of almost 300 Metropolitan Statistical Areas (MSAs) over the last two decades. The results showed that a 50% increase in an average city’s air traffic growth rate generated an additional stream of income over 20 years, which equaled 7.4% of real GDP [6].

Providing empirical support for the direction of causality between the development of air services and regional economic growth has important policy implications for governments in the development of air transport infrastructure along with its link to regional economic policies. There are four main types of causal relationships between airports and economic growth [7]. The arrows indicate the direction of the impacts, as shown in Figure 1.

Richard and Charlotta (2015) examined the role of airports in regional development and found that the impact of airports on regional development varies with their size and scale [8].

Focusing on different areas, Douglas Baker (2015) found that there was a significant bi-directional relationship between regional aviation and economic growth in Australia: airports had an impact on regional economic growth, and the economy directly impacted regional air transport [9]. Hakim and Merkert (2016) examined the causal relationship between air transport (total air passenger and air freight) and GDP in South Asian countries and found a long-run unidirectional Granger causality from GDP to air passenger and air freight volume [10]. Yingigba (2019) examined the causal relationship between economic and domestic air travel demand in Nigeria [11].

A review of studies on the economic impacts of airports included mainly three types of methods: (1) input-output models, (2) the collection of benefits, and (3) catalytic methods. In the catalytic methods, various econometric models were developed, including regression models, 2-stage least square regression [12]; dynamic panel data models [13]; short-run Granger causality analysis and long-run Granger causality analysis [14].

2.2. Airports Development and Regional Economy in China

Airports are the important elements of national and provincial air transportation infrastructures. They not only provide effective air transportation to meet travel demands but also promote the development of the local economy, including employment opportunities, industrial upgrading, and social welfare improvement. The goal of the 13th five-year aviation development plan from Civil Aviation Administration of China (CAAC) was to build a safe, convenient, efficient, and green modern civil aviation system to meet the demands of a moderately prosperous society (www.caac.gov.cn (accessed on 15 July 2019)). According to the National Transportation Airports Plan (www.caac.gov.cn (accessed on 15 July 2019)), there will be more than 260 civilian airports in 2022. The high quality and intensive sustainable development of airports and airline service are vital for the sustainable development of the transportation system and the national economy. It requires the regional economy and aviation industry to experience a virtuous circle.

Hong (2011) examined the linkage between transport infrastructure and regional economic growth by using data from a sample of 31 Chinese provinces from 1998 to 2007, and the result showed that the contribution of airway transport infrastructure is weak [15].

Long and Shi (2013) collected data from 1978 to 2010 to study the relationship between civil aviation and the economy of China [16], which showed that a long-term stable relationship had not been established. Yang (2014) pointed out that the increased volume of air passengers had promoted the development of the urban economy by using 15 years of data of 35 large and medium-sized cities in China [17]. Li (2015) found that economic growth benefited from the development of civil aviation using a co-integration test, a Granger causality test, an impulse-response and a variance decomposition test using a data set of the national turnover volume of air transportation and GDP from 2002 to 2012 [18]. Shen and Zou (2016) showed that the Granger causality between air transportation and the economy in China was bidirectional. In particular, economic growth in China promoted the growth of air transportation more than the opposite direction of causality [19]. Hu et al. (2015) applied a bi-variate panel VectorError Correction Model (VECM) to analyse both short and long-run equilibrium and Granger causality relationships between economic growth and domestic air passenger traffic based on quarterly panel data for 29 provinces in China for the period 2006Q1–2012Q3. The authors found a long-run cointegration and bi-directional Granger causal relationship between the two series [20]. Lu and Li (2016) reported that there was a long-term stable equilibrium relationship between aviation mileage and the national income level in China. The increase in national income levels had promoted the growth of aviation mileage [21]. Li and Li (2016) found that air transportation significantly improved as a result of local economic development by using the Grainger causality test with the data of airport passenger throughput, cargo throughput, and GDP from 1990–2014 in China [22].

Ryerson and Ge (2014) showed that turboprops could serve the second tier and emerging cities with lower fuel costs and reduced environmental impact to balance the increase in travel time for the passenger by undertaking a spatial analysis of the Chinese short-haul aviation network [23].

The published literature was in good agreement that aviation development was influenced by the economic growth. However, the economic growth in different regions had different characteristics [24]. This led to the complex development and evolution of aviation and also affected future policy settings. Therefore, the causality relationship test results were highly influenced by the data of specific time-series and regions.

2.3. Aims, Objectives and Contributions

This paper aims to study the causality between regional economy and airports development in typical Chinese provincial units by taking Jiangsu province as an empirical study. Jiangsu province is an economically prosperous region in China, but the progress of the aviation industry is lagging behind the economy, as discussed previously. The government of Jiangsu has the goal to increase investment in aviation to improve air passenger throughput in the near future based on the fast pace of economic development. Hence, the main purpose is hereby to address the methodology and indicate the heterogeneous relationships that can occur between aviation and economic development in China. More specifically, the objectives of this paper are:

• Test the stationarity and cointegration of the economy time series data and airport passenger throughput time series data to determine whether there is a long-term causality relationship in Jiangsu province.
• Test the type of causality between airports development and the regional economy to determine whether there is a correlation over the same period of the study.
• To detect the dynamic reactions between the air passenger volume and regional economy by impulse response and variance decomposition testing.

The contributions of this paper are:

• It establishes the empirical evidence for determining causal relationships between regional aviation/airports and economic growth in the typical provincial unit in China.
• It also provides evidence and references to the local government for deciding what policies can ensure airports’ viability and improve airports subsidy benefits. Especially in the polycentric urban region, better aviation development policies could realize multiple poles of economic growth.
• Jiangsu is the most important province in the Yangtze River delta region. The research results of Jiangsu will give suggestions for economic development strategies to the other provinces for reference.

3. Research Data and Processing

3.1. Research Data and Sources

The nine airports of Jiangsu were selected as research objects, and included Nanjing, Lianyungang, Yancheng, Xuzhou, Nantong, Changzhou, Wuxi, Huaian and Yangzhou. Nanjing, the capital city of Jiangsu, is the metropolitan region, with a population of 8.4 million compared with the total population of 80.3 million for Jiangsu province. Wuxi, Changzhou, and Nantong are the southern city clusters, which are very close to the most developed city: Shanghai. Xuzhou, Lianyungang, Huaian and Yancheng are the northern city clusters, which are less developed compared to Nanjing and the southern city clusters. Yangzhou City is located in the center of Jiangsu province, which is a well-known tourist destination. The layout of Jiangsu airports was shown in Figure 2.

An empirical analysis was conducted by using quarterly data of airport passenger throughput (PAX), and the local Gross Domestic Product (GDP) of the Jiangsu province. To test the causal relationship between airports and regional economic development, PAX was employed as an index to measure the development of airports, and GDP was used as a measurement of the regional economic development. Quarterly data of the total PAX of the nine selected airports and GDP of Jiangsu province from 2008 to 2018 was collected from the Statistical Yearbook by the Civil Aviation Administration of China and the China Statistical Yearbook by the National Bureau of Statistics. (all of the data were downloaded from www.caac.gov.cn (accessed on 15 July 2019) and www.stats.gov.cn (accessed on 15 July 2019)).

3.2. Data Processing

Data preprocessing tasks were conducted to prepare the dataset for modelling. First, as the GDP data obtained from the statistical yearbook was the current value of GDP, the influence of price factors should be removed to obtain the constant price of GDP. The constant value of the GDP was calculated using 2008 as the annual base period. Second, seasonal adjustments were required to remove any seasonality effects that may be present in the dataset. Figure 3a,b showed the raw data of GDP and PAX compared with the processed data GDPSA(constant price of GDP with seasonal adjustments) and PAXSA(PAX with seasonal adjustments). Therefore, the processed data was without the effect of price and seasonality. From Figure 3a,b, it can be seen that the overall development of airports and the economy in the Jiangsu region was fast. Moreover, there was a strong upward trend over time with strong seasonal fluctuations.

3.3. Unit Root Test

Granger (1988) suggested that for two series to be integrated, they must be integrated in the same order (more than zero), or that both series should contain a deterministic trend. The data clearly showed a time trend and seasonality. Therefore, it was necessary to conduct a Phillips-Perron test (PP test) for Unit Root with and without a time trend for the series of PAXSA and GDPSA. Both series were modeled as follows in Equation (1). Note that the time trend was tested using a quadratic term in the PP regression.

$${\mathrm{Y}}_{\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\mathrm{time}+{\mathsf{\beta }}_2{\mathrm{time}}^2+{\mathsf{\beta }}_3{\mathrm{D}}_{\mathrm{Q}1}+{\mathsf{\beta }}_4{\mathrm{D}}_{\mathrm{Q}2}+{\mathsf{\beta }}_5{\mathrm{D}}_{\mathrm{Q}3}+{\mathsf{\epsilon }}_{\mathrm{t}}$$

(1)

where Y_t was GDPSA or PAXSA; time was a time trend variable; D_Qx was a dummy variable that was equal to 1 in quarter $\mathrm{x}$, and zero otherwise; and $\mathsf{\epsilon }$_t was the error term.

The results of the PP test were shown in

Table 2. Phillips-Perron test Unit Root Test.

Variable	Test Form	PP Test Statistic	5% Critical Value	Prob.	Conclusions
GDPSA	Level (No Trend)	−0.3740	−2.933	0.9045	Accept Hypothesis, non-stationary
GDPSA	First difference (No Trend)	−61.9329	−2.933	0.0001	Refuse Hypothesis, stationary
PAXSA	Level (No Trend)	1.6182	−2.931	0.9994	Accept Hypothesis, non-stationary
PAXSA	First difference (No Trend)	−10.7602	−2.931	0.0000	Refuse Hypothesis, stationary

. The two series were non-stationary in their original form. However, PAXSA and GDPSA were stationary when the time component was removed. The test for stationarity confirmed that the detrended and de-seasonalised PAXSA and GDPSA were stationary.

4. Model Development

4.1. Vector Autoregression Function (VAR)

From the above analysis, we could establish the Vector Autoregression (VAR) function of the GDPSA and PAXSA variables. The VAR functions could be specified as follows:

$${\mathrm{GDPSA}}_{\mathrm{t}}={\displaystyle {\sum }_{\mathrm{k}=1}^{\mathrm{p}}{\mathsf{\alpha }}_{1\mathrm{k}}{\mathrm{GDPSA}}_{\mathrm{t}-\mathrm{k}}}+{\displaystyle {\sum }_{\mathrm{k}=1}^{\mathrm{p}}{\mathsf{\beta }}_{1\mathrm{k}}{\mathrm{PAXSA}}_{\mathrm{t}-\mathrm{k}}}+{\mathsf{\epsilon }}_{1\mathrm{t}}$$

(2)

$${\mathrm{PAXSA}}_{\mathrm{t}}={\displaystyle {\sum }_{\mathrm{k}=1}^{\mathrm{p}}{\mathsf{\alpha }}_{2\mathrm{k}}{\mathrm{GDPSA}}_{\mathrm{t}-\mathrm{k}}}+{\displaystyle {\sum }_{\mathrm{k}=1}^{\mathrm{p}}{\mathsf{\beta }}_{2\mathrm{k}}{\mathrm{PAXSA}}_{\mathrm{t}-\mathrm{k}}}+{\mathsf{\epsilon }}_{2\mathrm{t}}$$

(3)

In Equations (2) and (3), t was the lag order, GDPSA and PAXSA were the detrended and de-seasonalised GDP and PAX. Using the unrestricted VAR type, lag order 4 was selected from the VAR lag order selection criteria.

Table 3. VAR lag order selection criteria.

Lag	LogL	LR	FPE	AIC	SC	HQ
0	−618.7136	NA	1.03 × 1011	31.03568	31.12012	31.06621
1	−535.4174	154.0980	1.96 × 109	27.07087	27.32420	27.16247
2	−519.5074	27.84250	1.08 × 109	26.47537	26.89759 *	26.62803
3	−515.1696	7.157436	1.07 × 109	26.45848	27.04959	26.67220
4	−506.9142	12.79584 *	8.71 × 108 *	26.24571 *	27.00570	26.52050 *

*: Indicates lag order selected by the criterion. LR: sequential modified LR test statistic (each test at 5% level); FPE: Final prediction error; AIC: Akaike information criterion; SC: Schwarz information criterion; HQ: Hannan-Quinn information criterion.

showed the results. So the lag intervals for endogenous were 1 to 4, the estimated results were shown in

Table 4. Vector Autoregression Estimates.

	PAXSA	GDPSA		PAXSA	GDPSA
GDPSA (-1)	2.152570 * [1.28129]	0.105620 [0.67245]	PAXSA (-1)	0.557930 *** [3.14050]	0.026976 * [1.62412]
GDPSA (-2)	−1.294377 [−0.69824]	0.357623 ** [2.06344]	PAXSA (-2)	0.145074 [0.73929]	−0.011172 [−0.60894]
GDPSA (-3)	2.932682 * [1.51387]	−0.039202 [−0.21645]	PAXSA (-3)	0.329673 * [1.75797]	−0.037794 [−2.15561]
GDPSA (-4)	−2.560298 [−1.25295]	0.513495 ** [2.68784]	PAXSA (-4)	−0.123664 [−0.63563]	0.028991 * [1.59384]
c	108.0575 [1.8660]			−704.7399 [−1.1378]
R-squared	0.98077			0.9773
Adj. R-squared	0.97581			0.97152
F-statistic	197.67			167.29
Akaike AIC	10.799			10.934

t-statistics in []. *, **, and *** indicate the significance levels of 10%, 5% and 1%.

. This indicated that air passenger throughput (PAX) was influenced by PAX of lag order 1, 3 and GDP of lag order 1, 3; GDP was influenced by PAX of lag order 1, 4 and GDP of lag order 2, 4. The occurrence was usually a two-way relationship: the economy affected the airports, and the airports also affected the economy. Expectations regarding the future were the primary aspects of this interplay.

4.2. Granger Causality Test

It further estimated the Granger causal relationship of the two series based on the VAR specification in Equations (2) and (3). The Granger causality test revealed the direct or indirect effects between GDPSA and PAXSA.

Table 5. Results of Granger Causality Test.

Null Hypothesis	Lag Order	F-Statistic	Prob.	Conclusions
PAXSA does not Granger Cause GDPSA	1	11.0191	0.0019	Rejection Null Hypothesis
GDPSA does not Granger Cause PAXSA		8.96375	0.0047	Rejection Null Hypothesis
PAXSA does not Granger Cause GDPSA	2	2.14138	0.1318	Accept Null Hypothesis
GDPSA does not Granger Cause PAXSA		2.44693	0.1004	Accept Null Hypothesis
PAXSA does not Granger Cause GDPSA	3	2.39876	0.0850	Rejection Null Hypothesis
GDPSA does not Granger Cause PAXSA		2.54990	0.0720	Rejection Null Hypothesis
PAXSA does not Granger Cause GDPSA	4	2.17916	0.0946	Rejection Null Hypothesis
GDPSA does not Granger Cause PAXSA		2.43724	0.0679	Rejection Null Hypothesis
PAXSA does not Granger Cause GDPSA	5	1.80603	0.1441	Accept Null Hypothesis
GDPSA does not Granger Cause PAXSA		2.23480	0.0786	Refuse Null Hypothesis

contained the Granger causality test output.

The test results showed that there was bi-directional short-run Granger causality in lag order 1, 3 and 4 at the 10% significance level. These results indicated that there was a bi-directional Granger causality relationship between GDPSA and PAXSA in Jiangsu province. It also concluded that the airports and regional economy in Jiangsu province were mutually promoting growth.

4.3. Impulse Response and Variance Decomposition

• Impulse response

Impulse responses are the most commonly applied tools for describing these dynamic reactions in vector autoregressive analysis. The impulse response function is used to reflect the response of an endogenous variable on the impact of innovation (shock) in an exogenous viable. Specifically, if a standard deviation impact is given to the random error term, it will affect the current and future values of the endogenous variables.

Figure 4 showed the impulse response of PAXSA and GDPSA to Cholesky one S.D. innovation. The blue line was the impulse response estimate for the horizon period, and the two red lines were the one-standard error confidence bands. From Figure 4a, when there was an innovation on GDPSA for horizon periods H = 30, the response of PAXSA fluctuated greatly, which reached a peak in the fourth period and stabilized after the 20th period. Therefore, the impact of GDPSA on PAXSA was persistent and stably increasing in the long-term development. From Figure 4b, when there was an innovation on PAXSA, the response of GDPSA also fluctuated significantly in the first six periods, which reached the peak in the second period, then dropped quickly to the lowest in the fourth, and gradually stabilized. But after the 20th period, the response was unstable. These results indicated that the impact of PAXSA on GDPSA existed for a long period and was uncertain.

The mutual impacts of passenger throughput and economic growth were positive in Jiangsu province, and there was a more substantial impact in the short period. The impact of economic growth on passenger throughput was more significant than the impact of passenger throughput to economic growth. This result showed that airport growth was more sensitive to the change of economy in Jiangsu province.

2.

Variance decomposition

Impulse response functions trace the effects of a shock on one endogenous variable on the other variables in the VAR. Variance decomposition separates the variation in an endogenous variable into the component shocks to the VAR. Thus, the variance decomposition provides information about the relative importance of each random innovation in affecting the variables in the VAR.

From Figure 5a,b, it could find that the interaction between airport passenger throughput and the local economy was lagging. Figure 5a showed the percentage of GDPSA variance explained by self-change that was decreasing gradually, however, it became stable at about 60% after the 35th period. The proportion of variance explained by PAXSA increased gradually, and stabilized at 40% after 35 periods. Taken together, the proportion of passenger throughput contribution to the change of GDP was 40%. Figure 5b showed the change variance of PAXSA, which was mostly caused by its fluctuation in the beginning. The proportion of self-disturbance decreased from 92% to 50% in the 50th period. On the contrary, the proportion of variance explained by the change of GDPSA increased from 0% to about 30%. In the long term, the proportion of GDP contribution to passenger throughput was about 45%. Therefore, the inter-contribution of GDP and aviation development was sustainable and increasing over time.

5. Conclusions

This paper aimed to establish empirical evidence for determining the causal relationships between aviation/airports and economic growth in a typical Chinese provincial area—Jiangsu province. Using a VAR model and a Granger causality test, it found that economic growth and the development of airports were interdependent and mutually promoting. The results of Impulse Response and Variance Decomposition showed that the impact of GDP on airport passenger throughput was persistent and stably increasing in the long-term development; the impact of airport passenger throughput on GDP also existed for a long period but was unstable.

Jiangsu is located in the core of the Yangtze River delta region, which is the polycentric urban region. Jiangsu faces regional economic competition from Shanghai and the Zhejiang province. Jiangsu airports are less competitive than the Shanghai and Zhejiang airports. From the research, the GDP had a long-term impact on airport development, but the impact of airports on the GDP was less. It is therefore important to leverage airport development to improve the regional economy. The contribution of aviation is not only about the airport itself but also about the induced and catalytic impacts, such as the funding of infrastructure and the attraction of new firms. Governments could support some kind of long-term strategic spatial framework to inform key investment decisions of airport-based sub-centers in regional polycentric growth [25,26] and promote closer collaboration in the Jiangsu provincial airports, including the airline network, flight schedules and air service quality.