Empirical Evidence from EU-28 Countries on Resilient Transport Infrastructure Systems and Sustainable Economic Growth

This paper examines the nexus between the main forms of transport, related investments, specific air pollutants, and sustainable economic growth. The research is important since transport may act as a facilitator of social, economic, and environmental development. Based on data retrieved from Eurostat, Organisation for Economic Co-operation and Development (OECD), and World Bank, the output of fixed-effects regressions for EU-28 countries over 1990–2016 reveals that road, inland waterways, maritime, and air transport infrastructure positively influence gross domestic product per capita (GDPC), though a negative link occurred in the case of railway transport. As concerning investments in transport infrastructure, the empirical results exhibit a positive impact on economic growth for every type of transport, except inland waterways. Besides, emissions of CO2 from all kind of transport, alongside other specific air pollutants, negatively influence GDPC. The fully modified and dynamic ordinary least squares panel estimation results reinforce the findings. Further, in the short-run, Granger causality based on panel vector error correction model pointed out a unidirectional causal link running from sustainable economic growth to inland waterways and maritime transport of goods, albeit a one-way causal link running from the volume of goods transported by air to GDPC. As well, the empirical results provide support one-way short-run links running from GDPC to investments in road and inland waterway transport infrastructure. In addition, a bidirectional short-run link occurred between carbon dioxide emissions from railway transport and GDPC, whereas unidirectional relations with economic growth were identified in the case of carbon dioxide emissions from road and domestic aviation. In the long-run, a bidirectional causal relation was noticed between the length of the railways lines, investments in railway transport infrastructure, and GDPC, as well as a two-way causal link between the gross weight of seaborne goods handled in ports and GDPC.


Introduction
Infrastructure ensures the provision of basic physical systems and structures that are indispensable for peoples and companies [1].Amongst the multifarious kinds of infrastructure, the policy makers regard transport infrastructure as one of the most essential, since related cost is critical in setting the location for firms, and hence the economic development of a region [2].Li and Li [3] reported that one dollar of road investment in China cut input inventory by about 2 cents each year over 1998-2007.As well, transport is one of the crucial sectors, inasmuch as an economy can benefit from transport services by quickening access to amenities, increasing the market mobility and direct employment [4], encouraging international tourism [5], saving time, and dropping business costs [6,7].Beyond, through attracting resources from other areas, transport infrastructure can act as a magnet of regional economic growth [8].According to Chandra and Thompson [9], highways increase the economic activity in the counties that they pass directly through, but draw activity away from nearby counties.Therefore, well-organized transportation networks engender employment and wealth, and lead to economic growth [10].Bottasso et al. [11] explored the main European ports of Organisation for Economic Co-operation and Development (henceforth "OECD") states and noticed that regional employment is positively associated to port throughput, although the number of passengers was not.Li et al. [12] found that road infrastructure influences the local economy by attracting foreign investments and stimulating real estate development.Contrariwise, a deficiency of infrastructure generates bottlenecks for sustainable growth and poverty reduction [13].
The classical location theory points out the significance of transport costs as a driver of the place of economic undertakings [14,15].Therewith, the view that the interaction amid transport costs and increasing returns to scale, offers corporations a reason to lay down near large markets, is a central part in the new economic geography literature [16].In addition, the theory of endogenous growth reinforces that public infrastructure, covering transport infrastructure, is a basis of economic growth via its role in technical change.Barro [17] predicted that only productive government expenditures will positively influence the long-run growth rate.However, well-developed transport infrastructure ensures returns through certain macroeconomic drivers of productivity, such as: "expansion of business activity, innovations, and investments, labor market, competition, domestic and international trade globally mobile activity, regional economic development, wellbeing of population, environment safety and health" [18].Aschauer [19] revealed that a higher level and enhanced quality of highway size increases transportation services and raises the marginal product of private capital.Thus, investments in transport infrastructure will increase the efficiency and lessen the prices of production inputs [20].Padeiro [21] examined by the means of logistic regressions whether small municipalities register employment growth based on their position in relation to transport infrastructures, and found that motorways influence very small regions, whilst airports influence bigger ones.
On the other hand, transport externalities such as carbon dioxide emissions (henceforth "CO 2 ") [22,23] and traffic mortalities [24] cause environmental degradation [25] and a lower quality of life [26].Likewise, Geist and Lambin [27] emphasized that infrastructure extension is one of the factors that prevail in affecting deforestation.Also, lubricant and substance spills from ships, either from operating activities or catastrophic mishaps, cause health hazards [28].Unfortunately, dropping emissions in transport is more expensive than in other sectors, inasmuch as transport still heavily relies on fossil fuels [29].There are environmental sustainability norms for biofuels and bioliquids, with a threshold of 35% savings of greenhouse gas emissions [30,31].The Paris Agreement [32] aims to restrict global warming to well below 2 • C above pre-industrial levels and confine the temperature increase to 1.5 • C, whilst White Paper [33] pursues a 60% greenhouse gas emission reduction by 2050 in relation to 1990.Hence, in order to accomplish the target of cutting transport emissions, the decarbonization of transport is required [29], alongside improvements in transport energy intensity, the substitution of road transport by rail transport, and the shift from oil products to electricity [34].As well, smart city solutions can unveil a vital role in alleviating transport emissions and fulfilling reduction targets [35].However, in 2015, transport (comprising aviation and shipping) contributed 25.8% of total greenhouse gas emissions in the EU-28 [36].Besides, the European Union (henceforth "EU") member states are required to fulfill the goal of 10% for the share of renewable energy in the transport sector by 2020 [30].As well, EU set a 30% drop of emissions from both cars and vans in 2030 compared to the 2021 targets [37].In this regard, investing in up-to-date infrastructure enables the use of more energy-efficient means and alternative technologies that positively affect the economy with minimizing negative externalities [38].Therefore, Kumari and Sharma [39] noticed a strong association between physical and social infrastructure, and economic development.
The current paper aims at empirically investigating the relationship between transport infrastructure systems and economic growth for EU-28 countries.Therefore, the paper's leading purposes are: (i) to examine the influence of the main forms of transport, namely rail, roads, water, and air on gross domestic product per capita; (ii) to explore the impact of transport infrastructure investments on economic growth; (iii) to analyze the effect of carbon dioxide emissions from transport and other specific air pollutants from transport on economic growth.Transport is a strategic segment of the EU economy that exerts a striking influence on social, economic, and environmental expansion.In the same vein, transport supports economic growth, employment, global competitiveness, and trade, allowing persons and goods to shift across Europe.Withal, transport is the crucial promoter of the four freedoms of displacement outlining the Single Market: citizens, merchandises, services, and funds [40].Previous papers on European countries studied the impact of port activities on local employment [11], the influence of port activities on local development [41], the relationship between air transportation and regional growth [42], as well as the impact of infrastructure investments on national productivity [43].To the best of our knowledge, this is the first study that analyzes the influence of transport infrastructure, related investments, and specific air pollutants on sustainable economic growth in EU-28.
The structure of the paper is as follows.Section 2 provides an overview of the previous findings towards the influence of transport infrastructure, associated investments, and related pollutant emissions on sustainable economic growth.Section 3 describes the sample, variables, and econometric methodology, whilst Section 4 discusses the empirical output.The last section summarizes the main conclusions and provides policy implications.

Transport Infrastructure and Economic Growth
Transport facilities contribute to economic growth by enhancing the whole productivity of production units, by stimulating technological spillovers across economies, as well as by increasing the profitability of transport-connected businesses [44].Using provincial Chinese panel data from 1993 to 2009, Xuelian [45] indicated that the total output elasticity of transport infrastructure for regional economic growth fluctuates between 0.05 and 0.07.However, there was exhibited that means of transport perform in different ways, respectively modal change of freight transport from road to rail is positive with regard to its impact in total output and the total wages of the economy, whilst freight transport by road has a superior influence in employment [46].Ikumo [47] noticed that the positive economic effect of high-speed railway service will occur in the regions along the paths, but this will engender a negative influence with other aspects.Nevertheless, large investments in high-speed train infrastructure are not reasonable, as long as its economic development benefits are not assured [48].Ding [49] revealed that a one percent increase in urban road density will increase the manufacturing gross domestic product share of the city-proper by 0.095 per cent, whereas a one percent growth in major regional roads is related to a 0.144 per cent increase in the manufacturing GDP share in the city-proper.Consequently, road infrastructure can boost economic growth [50].
Air cargo empowers states, regardless of location, to efficiently link to remote markets and global supply chains in a quick, trustworthy mode.Hence, in the current fast-cycle logistics period, nations with suitable air cargo connectivity show competitive commerce and production benefits over those without such capabilities [51].Fernandes and Pacheco [52] revealed for Brazil, from 1966 to 2006, a unidirectional Granger causal link from economic growth to domestic air transport demand.Yao and Yang [53] depicted for 31 Chinese provinces and municipalities, over 1995-2006, that a positive association ensued between air transport and economic growth, industrial structure, population density, as well as openness.By means of the European-level annual data from eighty-six regions, Mukkala and Tervo [42] noticed a causal relationship from air traffic to regional growth in peripheral regions, although this connection was less evident in central areas.Baltaci et al. [54] examined 26 sub-regions at NUTS 2 level in Turkey, over the period of 2004-2011 and provided support via fixed-effects and two stage least squares methods with instrumental variable approaches that airline transport positively influence economic growth.For Italy, during 1971-2012, Brida et al. [55] found a unidirectional causality running from aircraft movements to real gross domestic product.
Likewise, ports are essential for the support of economic activities in the surrounding areas, as they act as a critical association amid sea and land transport [56].Besides, according to Golebiowski [57] water carriage is the most energy effective way of transportation.For Iran, during 1978-2012, Taghvaee et al. [58] observed via a dynamic log-linear approach, both in the short-run and long-run, the occurrence of a positive link between maritime transportation and GDP.By exploring a panel of 40 heterogeneous countries, Khan et al. [59] revealed a positive link between container port traffic and per capita income.
Table 1 points out a brief review of previous papers that examined the relationship between transport infrastructure and economic growth.

Transport Infrastructure Investments and Economic Growth
Transport infrastructure investments suppose supplementary transport volume, augmented trustworthiness, and an improved quality of transport services.In turn, this determines lower transport costs, as well as shorter transit times.As well, enhanced transport infrastructure is the essential part for business expansion [70].Pereira and Andraz [71] estimated for Portugal, in the long-term, that one euro in public investment rises output by 9.5 euros, namely a rate of return of 15.9%.Hu and Liu [72] explored 28 provinces in China from 1985 to 2006 via a spatial econometric model and found that each 1% increase of the transportation investment in China leads to 0.28% increase on GDP.Yu et al. [73] unveiled a one-way Granger causal link between transport investment and economic growth at Chinese national level.Based on a sample of 563 estimates obtained from 33 studies, Melo et al. [74] reported that a growth of 10% in public investment in transport infrastructure is related with a rise in output of almost 0.5%.Holmgren and Merkel [75] conducted a meta-analysis of 776 estimates with regards to elasticity of production concerning infrastructure and noticed that the estimated outcome of investing in infrastructure fluctuates from −0.06 to 0.52.However, Deng et al. [76] revealed that transport infrastructure investment can stimulate economic growth, but the extent of outcome variations hinge on the scale of the existing transport network.Hence, there was noticed an insignificant positive association between transport infrastructure accumulation and economic growth when the highway network density is lower than 0.17 km/km 2 .Besides, a significant positive effect on economic growth occurred when the highway network density is among 0.17 and 0.38 km/km 2 , or higher than 0.38 km/km 2 [76].As well, it was established that private sector investment in the transport sector and disposal of rail infrastructure lessen transport CO 2 emissions [77].For Pakistan, Mohmand et al. [78] revealed no causal link in the short run between economic growth and transportation infrastructure, but a unidirectional causality from economic development to infrastructure investment in the long run.Yang et al. [79] established that urban transportation investment increases emissions in the short run because it might determine road jams that would heighten the emissions of the low-speed traffic, but in the long run, the urban transportation investment exhibits a positive influence on decreasing emissions since it could extend the roads and make the traffic system more accessible.
A summary review of papers that explored the link between transport infrastructure investment and economic growth is shown in Table 2.

Pollutant Emissions from Transport and Economic Growth
Roads carry noteworthy economic benefits and are crucial for development, specifically in densely inhabited rural regions that are unrelated to markets and economic activity, but are frequently the forerunners to deforestation and biodiversity harm [88].However, the transport sector, which comprises the aviation, road, navigation, and railway subsectors, is presently the second highest emitter of greenhouse gases, after power and heat production [89], and it accounts for about a quarter of CO 2 emissions globally [90].Environmental issues such as air pollution are the consequences of the rise and concentration of greenhouse gases in the atmosphere, which comprises CO 2 , methane, and nitrogen oxides [91].The climate influence of aviation is determined by long-term effects from CO 2 emissions, and shorter-term impacts from non-CO 2 emissions and effects, which comprise the emissions of water vapor, particles, and nitrogen oxides [92].Ben Abdallah et al. [93] noticed that a reciprocal causal relationship between transport value added, road transport-related energy consumption, road infrastructure, and CO 2 emissions occurs in the Tunisian transport sector, in the long-run.Sousa, Roseta-Palma and Martins [89] found a monotonically increasing relationship between economic growth and CO 2 emissions from transport in Portugal.However, Atte-Oudeyi et al. [94] revealed for a sample of six emerging countries an inverted U-shaped Environmental Kuznets Curve between CO 2 emissions per capita due to road transport and the level of GDP per capita.On the contrary, Alshehry and Belloumi [95] did not confirm the inverse-U association between transport CO 2 emissions and economic growth in Saudi Arabia, but noticed in the long run a unidirectional causality running from economic growth to transport CO 2 emissions.Yang and He [96] revealed a long-term equilibrium link between road transport subsectors and Air Pollution Index in Beijing, whereas both the freight and passenger subsectors are unidirectional Granger causing Air Pollution Index.
Table 3 reveals a brief review of past studies that examined the relationship between pollutant emissions from transport and economic growth.

Quantitative Framework
In order to investigate the impact of transport infrastructure status, related investments, carbon dioxide emissions from transport, and other specific air pollutants on sustainable economic growth, we primary employed a panel data multivariate linear regression model [19,34,49,51,54,66,68,79,86,88,94] with the basic form being as follows: where Y stands for the dependent variable, respectively gross domestic product per capita, X denotes a vector of explanatory variables capturing the status of the main forms of transport (railway, road, inland waterways, maritime, and air), W signifies a vector of explanatory variables regarding investments in the leading types of transport, Z describes a vector of explanatory variables regarding pollutant emissions from transport (carbon dioxide emissions, sulphur oxides, nitrogen oxides, non-methane volatile organic compounds, and ammonia), and Controls is a vector of country-level control variables catching additional features that may impact economic growth.The constant is depicted by α, β 1 -β 4 are the coefficients to be estimated, ε is the residual term, i is the subscript of EU-28 countries, and t is the subscript of the time span.The fixed-effects approach was selected since the study was limited to a particular group of states [34].
Onward, the nexus between transport infrastructure, associated investments, emissions of carbon dioxide, and economic growth was explored through a panel vector autoregressive model [2,8,39,44,50,55,59,62,64,65,69,71,82,83,87,93,101,105].Firstly, the unit root of each variable was examined, then the long-run cointegration link among the variables is investigated, and afterwards the panel vector error correction model (henceforth "PVECM") was estimated to infer the Granger causal associations [44].A nonstationary panel process occurs if at least one of its moments (mean, variance, or covariance) is time-independent [93], whilst a stationary process occurs whether its statistical features persists constantly [62].In case of the occurrence of nonstationarity, a cointegration and vector error correction model were put forward to explore the link amid such variables [8].Provided that a non-stationary series is differenced d times to be converted stationary, then it is assumed to be integrated of order d: I(d) [64].However, as long as none of the panel unit root tests were free from statistical weaknesses with respect to size and power properties, we applied several unit root tests to find out the order of integration of the variables [44], as follows: Levin, Lin and Chu (henceforth "LLC"), Im, Pesaran, and Shin (henceforth "IPS"), Augmented Dickey-Fuller (henceforth "ADF"), Phillips-Perron (henceforth "PP"), and Breitung.The autoregressive specification for panel data is formulated as below [62]: where X denotes the exogenous variables in the model, covering any fixed effect or individual trends, α i states the autoregressive coefficients, and ε it signifies the error terms which are supposed to be jointly independent.Y i is weakly stationary if α i is less than one, whilst Y i has a unit root when α i is equal to one [62].The ADF test considers merely the existence of autocorrelation in the series, but the PP test furthermore takes into account the hypothesis of the occurrence of a heteroskedasticity extent in the series [93].The LLC test examines the heterogeneity of intercepts across members of the panel, whereas the IPS test investigates the heterogeneity within the intercepts as well as in the slope coefficients [44].The null hypothesis posits that all series are non-stationary, whilst the alternative one presumes the stationarity of all series [100].

Descriptive Statistics, Correlation Examination, and Unit Root Testing
Table 5 gives the descriptive statistics of the variables.With reference to goods transportation, road transport showed the utmost mean value (69,372.64 million tone-km), whereas air transport shows the lowest mean value (0.60 million tone).On the contrary, in terms of passenger transportation, air transport showed the highest mean value (41.02 million passengers).Even if the investments in road transport infrastructure highlighted the paramount mean value, in terms of network length, EU-28 railways lines registered the highest mean value.By 2030, the length of the existing European high-speed rail network should be tripled, and by 2050 the majority of medium-distance passenger transport should go by rail [33].Source: Authors' computations.Notes: For the definition of variables, please see Table 4.
Figure 1 shows the mean volume of transported goods in EU-28, over 1990-2016.Alike Table 4, road transport is the main mode of freight transport, followed by rail freight.Germany registered the highest mean values in relation to all types of transport, except for the gross weight of seaborne goods handled in ports (United Kingdom reveals the highest mean value), whereas the lowest mean values are reported in Ireland (rail transport), Cyprus (road transport), the Czech Republic (inland waterways transport), Malta (gross weight of seaborne goods handled in ports), and Croatia (air transport).
Figure 2 exposes the mean investments in transport infrastructure in EU-28, over 1990-2016.As in Table 4, investments in road transport infrastructure showed the utmost mean value, succeeded by investments in railway transport infrastructure.Again, Germany revealed the highest mean value of investments in roads and inland waterways transport, along with Italy (investments in railway transport), Spain (investments in maritime port infrastructure), and the United Kingdom (investments in airport infrastructure).In contrast, the lowest mean value of investments in roads and maritime port infrastructure was exhibited in Malta, as well as Estonia (investments in railway transport), Luxembourg (investments in inland waterways transport), and Slovenia (investments in airport infrastructure).Source: Authors' computations.Notes: For the definition of variables, please see Table 4. Notes: For the definition of variables, please see Table 4.

NH3_non
Figure 2 exposes the mean investments in transport infrastructure in EU-28, over 1990-2016.As in Table 4, investments in road transport infrastructure showed the utmost mean value, succeeded by investments in railway transport infrastructure.Again, Germany revealed the highest mean value of investments in roads and inland waterways transport, along with Italy (investments in railway transport), Spain (investments in maritime port infrastructure), and the United Kingdom (investments in airport infrastructure).In contrast, the lowest mean value of investments in roads and maritime port infrastructure was exhibited in Malta, as well as Estonia (investments in railway transport), Luxembourg (investments in inland waterways transport), and Slovenia (investments in airport infrastructure).4.
With regard to air pollutants, both in terms of the share of CO2 emissions from road transport in total carbon dioxide emissions from transport and other specific air pollutants, road transport was the most damaging environmental stressor in EU-28 countries.Besides, air transportation revealed the lowest share of CO2 emissions from domestic aviation in total carbon dioxide emissions from transport, harmonized with the White Paper [33].
The correlations between selected variables are exposed in Table 6.We acknowledged strong linear associations between several variables, with the value associated to the correlation coefficient exceeding 0.7.However, the issue of multicollinearity was removed by considering the aforesaid variables in distinct regression models.
Table 7 summarizes the outcome of the panel unit root tests.When first differences were considered, the null hypothesis of non-stationarity was rejected for all variables.As such, a part of the variables was not stationary in level, but stationary in the first difference.Therefore, the series were integrated at I(1) [44].4,000,000,000 6,000,000,000 8,000,000,000 10,000,000,000 12,000,000,000 14,000,000,000 16,000,000,000 18,000,000,000 20,000,000,000  4.

Panel Data Regression Models Output
With regard to air pollutants, both in terms of the share of CO 2 emissions from road transport in total carbon dioxide emissions from transport and other specific air pollutants, road transport was the most damaging environmental stressor in EU-28 countries.Besides, air transportation revealed the lowest share of CO 2 emissions from domestic aviation in total carbon dioxide emissions from transport, harmonized with the White Paper [33].
The correlations between selected variables are exposed in Table 6.We acknowledged strong linear associations between several variables, with the value associated to the correlation coefficient exceeding 0.7.However, the issue of multicollinearity was removed by considering the aforesaid variables in distinct regression models.
Table 7 summarizes the outcome of the panel unit root tests.When first differences were considered, the null hypothesis of non-stationarity was rejected for all variables.As such, a part of the variables was not stationary in level, but stationary in the first difference.Therefore, the series were integrated at I(1) [44].

Panel Data Regression Models Output
Table 8 shows the econometric outcomes concerning the effect of railway transport infrastructure, related investments, and air pollution on gross domestic product per capita.We noticed a negative impact of the length of railways lines (Equations ( 4), ( 7) and ( 12)) and rail passenger transport (Equation ( 2)) on economic growth.The negative outcome may have occurred due to regions being short of high-speed rail lines, and respectively, the absence of ripple effects across such areas.Consequently, there may be an influx of companies and dwellings to the extents owing high-speed infrastructure, whereas the rest of zones will exhibit a quite lesser desirability, and scarcer human resources.Accordingly, the progress of a particular area may not be sufficient to develop the entire state [47].Nevertheless, rail goods transport (Equation ( 6)) and investments in railway transport infrastructure positively influenced economic growth (all the estimated models).
In line with Chen, Xue, Rose and Haynes [84], investment in rail infrastructure may stimulate economic growth through job creation, enhancement of regional openness, and a drop in transport price.Therewith, along with the expansion of rail infrastructure, several matters such as road traffic congestion or air contamination from cars and air travel will be settled [84].With regard to air pollution, the share of CO 2 from railway transport in total carbon dioxide emissions from transport negatively influences GDPC (Equations ( 4)-( 6)); this relationship was also proven in case of emissions of sulphur oxides (Equations ( 7) and ( 8)), and emissions of ammonia (Equation ( 12)).Hence, even if railway investment entails more benefits [87], the greater use of rail leads to a rise in electricity demand, causing higher CO 2 emissions [99].Also, local air pollution, climate change, noise, and land use [48] are the consequences of rail networks, especially high-speed rail systems.
The estimations concerning the impact of road transport infrastructure, associated investments, and air pollution on sustainable economic growth are revealed in Table 9.The results showed that there was a positive influence of the length of motorways (Equations (3), ( 9) and ( 13)), the number of passenger cars per 1000 inhabitants (all the estimated models), as well as goods transported by road on GDPC (all the estimated models).Likewise, a strong positive association was noticed between investments in road transport infrastructure and economic growth (Equations ( 4), ( 7) and ( 10)).Investments in roads drives the creation of direct employment [4], attracts many investors, which bring physical and socio-economic development to surrounding areas [24], and improves the productivity of corporations [3].Source: Authors' computations.Notes: Superscripts ***, **, *, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively.Figures in brackets depict t-statistic.For the definition of variables, please see Table 4.
Table 9. Fixed-effects regressions results on the effect of road transport infrastructure, related investments, and air pollution on sustainable economic growth.

Equations
(1) Therewith, all the estimated equations point out that both the share of CO 2 emissions from road transport in total carbon dioxide emissions from transport and other specific air pollutants, except NH 3 , negatively influence GDPC, due to their adverse effects on the atmosphere, on health, and on climate change [25].
Table 10 shows the empirical outcomes regarding the effect of inland waterways transport infrastructure, related investments, and air pollution on sustainable economic growth.The econometric results provide support for a positive influence of inland waterways goods transports on economic growth (Equations (1), ( 3) and ( 8)), except Equation ( 4), due to its features such as reduced energy consumption, reduced external costs, and a reduced number of accidents [57].
Table 10.Fixed-effects regressions results on the effect of inland waterways transport infrastructure, related investments, and air pollution on sustainable economic growth.

Variables Equations
(1) However, investments in inland waterways transport negatively influenced GDPC, but the relationship was statistically weakly significant merely in Equation (5).In line with Hong, Chu and Wang [68], investment in water transport infrastructure will positively influence economic growth only after the investment scale surpasses a threshold level.Besides, we noticed a negative association between all other specific air pollutants and GDPC (all the estimated models), but the share of CO 2 emissions from international maritime bunkers in total carbon dioxide emissions positively influenced economic growth (Equation ( 5)).However, inland water networks are usually overfilled in CO 2 , being documented as a prominent side in the global carbon cycle [22].
The results, with reference to the influence of maritime transport infrastructure, associated investments, and air pollution on sustainable economic growth, are shown in Table 11.Like inland waterway goods transport (see Table 10), the gross weight of seaborne goods handled in ports exerted a strong positive effect on the GDPC (Equations (1), (3), ( 5)), in line with previous studies [58,59].Similar to Song and van Geenhuizen [86], we reinforced that investments in maritime port infrastructure positively influence economic growth (Equations ( 6) and ( 8)), apart from Equation (4).In case of air pollutants, it was revealed that emissions of sulphur oxides (Equation ( 7)) and emissions of ammonia (Equation ( 9)) negatively influenced the GDPC.
Table 11.Fixed-effects regressions results on the effect of maritime transport infrastructure, related investments, and air pollution on sustainable economic growth.

Variables Equations
(1) The estimation results with respect to the impact of air transport infrastructure, associated investments, and air pollution on sustainable economic growth, are presented in Table 12.As in previous studies [53,54,65,67], the total number of passengers (Equations (1), ( 3)), alongside the volume of goods transported in Europe (Equation ( 4)) positively influenced GDPC.Likewise, investments in airport infrastructure positively influenced economic growth (Equations ( 4)-( 8)).With regard to transport pollution, we noticed a positive connection between the share of CO 2 emissions from domestic aviation in total carbon dioxide emissions from transport and GDPC (Equation ( 4)), but other specific air pollutants, except NH 3 , negatively influence GDPC (all the estimated models).In relation to country-level control variables, trade openness and domestic credit to private sector positively influence economic growth since the high volume of freight boosts economic prosperity [12].In the case of urbanization, there are noticed mixed relationships, whereas energy consumption of transport relative to GDP negatively influences the gross domestic product per capita.

Cointegration and Causality Examination
Considering that all of the selected variables have a unique order of integration (see Table 7), the cointegration examination was further employed.Table 13 shows the results of the Pedroni test [109,110].There are two sets covering tests for homogeneous and heterogeneous panels: the tests of the first set average the results of single state test statistics, whereas tests of the second set pool the statistics along the within-dimension [38].Thereby, several statistics provide support for the cointegration connection among the variables from the five models.Onward, as in [62,65,103], the long-run equilibrium between variables was checked via the Kao [111] test presented in Table 14, as well as the Johansen [112] test shown in Table 15.Hence, the null hypothesis of no cointegration between variables was rejected, thus reinforcing the existence of cointegration.Therefore, transport infrastructure, related investments, emissions of carbon dioxide, and economic growth has a long-run relationship in the EU-28 countries.For robust statistical estimates, we employed a non-parametric approach, namely a panel with a fully modified ordinary least squares (henceforth "FMOLS") estimator as in [5,59,65,77,82,107], but also an alternative parametric method, respectively a dynamic ordinary least squares estimator (henceforth "DOLS"), as in [5].The FMOLS is the desirable econometric way to assess the long-run parameter estimates that handle both the endogeneity and serial correlation concerns [59], whilst DOLS covers the lagged first difference [77].Table 16 displays the parameter estimates by FMOLS and DOLS.As in Table 8, we noticed a negative impact of the length of railways lines on GDPC (Equation (1)).Contrariwise, consistent with [82] and analogous to Table 9, the length of motorways positively influenced economic growth (Equation ( 2)).With regard to the carriage of goods by the main forms of transport, a positive and statistically significant impact was observed (Equations ( 3)-( 5)), similarly [59].Likewise, investments in transport infrastructure positively influenced economic growth (Equations (1)-( 3)), apart from investments in airport infrastructure that negatively influenced GDPC (Equation ( 5)).In the case of investments in maritime port infrastructure, a statistically non-significant impact was observed (Equation ( 4)), contrary to [82].In relation to air pollutants, the share of CO 2 emissions from railway transport in total carbon dioxide emissions from transport negatively influenced GDPC (Equation (1)), as in Table 8.Nevertheless, the share of CO 2 emissions from international maritime bunkers in total carbon dioxide emissions had a positive influence on GDPC (Equation ( 3)), consistent with [107].Since the variables were cointegrated, we could estimate the panel vector error correction model to assess the direction of the causality [62].Therefore, Table 17 reveals the PVECM Granger causalities where the ECT denotes the long-run dynamics, whilst the differenced variables shows the short-run dynamics between the variables [44].As such, when ∆GDPC served as the dependent variable, the error correction term was statistically significant, except for model 5.This denotes that economic growth tends to converge to its long-run equilibrium path in response to changes in its regressors [44].With respect to the first model, regarding railway transport, in short-run, we noticed a bidirectional causal link between the share of CO 2 emissions from railway transport in total carbon dioxide emissions from transport and GDPC, similar to [107].Besides, the empirical results provided support for a one-way causality link running from investments in railway transport infrastructure to CO 2 _rail.Thereby, investments in rail infrastructure entail higher amounts of energy, mostly in the steel industry, and per se, greater CO 2 emissions [99].Besides, we noticed long-run causality running from Length_rail, Inv_rail, and CO 2 _rail to GDPC, from GDPC, Inv_rail, and CO 2 _rail to Length_rail, as well as from GDPC, Length_rail, and CO 2 _rail to Inv_rail.
In the case of the second model, concerning road transport, in short-run, we acknowledged a bidirectional causal association between investments in road transport infrastructure and the length of motorways.Further, we noticed a unidirectional causal relation running from GDPC to investments in road transport infrastructure, as well as a unidirectional causal relation running from GDPC to CO 2 _road, consistent with [95].In the long-run, causal links ran from Length_motorways, Inv_roads, and CO 2 _road to GDPC, from GDPC to Inv_roads, similar to [78], and from Length_motorways, CO 2 _road to Inv_roads.
The third model towards waterways transport revealed in the short-run a couple of one-way causal links running from GDPC to inland waterways goods transports, and from GDPC to investments in inland waterways transport, similar to [103].As well, further short-run causal relationships ran from investments in this transportation type to Goods_water.In the long-run, we noticed a causal link running from Goods_water, Inv_water, and CO 2 _maritime to GDPC.
The model regarding maritime transport revealed a unidirectional causal link running from economic growth to the gross weight of seaborne goods handled in ports, consistent with [2,59], alongside a unidirectional causal relation running from Goods_sea to investments in maritime port infrastructure.The following causal links were observed in the long-run: Goods_sea, Inv_ports, and CO 2 _maritime to GDPC, along with GDPC, Inv_ports, and CO 2 _maritime to Goods_sea.
The fifth model with regard to air transport revealed in the short-run a unidirectional causal association running from the volume of goods transported by air to GDPC, like [42,64], and from the share of CO 2 emissions from domestic aviation in total carbon dioxide emissions from transport to GDPC.Moreover, we remarked on long-run causalities running from GDPC to Goods_air alike [62], also from Inv_airports and CO 2 _aviation to Goods_air, and from GDPC, Goods_air, Inv_airports to CO 2 _aviation.

Concluding Remarks and Policy Implications
The transport sector reveals itself as a fundamental factor towards economic growth, ensuring the efficient distribution of resources and mobility for people [107].This paper examined the relationship between the main types of transport, related investments, specific air pollutants, and sustainable economic growth for EU-28 countries over 1990-2016.By the means of fixed-effects regressions, the empirical results provide support for a positive impact of road, inland waterways, maritime, and air transport infrastructure on economic growth.Likewise, investments in transport infrastructure positively influence gross domestic product per capita for every form of transport, apart from inland waterways, whilst CO 2 emissions from transport and other specific air pollutants exhibit a negative impact on economic growth.Overall, the output of panel fully modified OLS and dynamic OLS regressions confirm the findings.
According to Granger causality based on a panel vector error correction model, we noticed a short-run one-way link running from the volume of goods transported by air to GDPC, as well as from inland waterways goods transports and the gross weight of seaborne goods handled in ports to GDPC.In the case of the length of railways lines and the length of motorways, no short-run causal relationship with economic growth was noticed.With reference to transport investments, we found a short-run causal connection running from investments in road transport infrastructure and investments in inland waterways transport to GDP.Concerning air pollutants, a bidirectional link occurred between CO 2 emissions from railway transport and GDPC, while unidirectional relations appeared from economic growth to CO 2 from road, and from CO 2 related to domestic aviation to economic growth.In the long-run, we notice a two-way causal link between the length of railways lines, investments in railway transport infrastructure, and economic growth, as well as a bidirectional causal connection between the gross weight of seaborne goods handled in ports and gross domestic product per capita.
With reference to policy decision-making processes, the EU-28 nations should continue the enlargement of the railway system in order to follow the targets established in the White Paper [33].As well, in order to achieve the EU goals concerning emissions of air pollutants from transport [30,37], manufacturers should aim to decrease the usage of fossil fuels [99], whilst designing more fuel efficient-vehicles.With respect to maritime transport, even if CO 2 discharges per 1 tonne/kilometer are up to five times less in inland waterway carriage than road transport [57], replacing current fleet with bigger vessels may be considered in terms of the decrease in greenhouse gases [23].Therewith, investments within airport and maritime port infrastructure should be speeded up.Likewise, second generation biofuels should be promoted with regard to the challenge of transportation sector decarbonization.
A limitation of the current study ensues from the fact that high-speed rail network was not considered distinctly from traditional rail, whilst the density of transport networks in relation to the total area and the number of citizens was not taken into account.For future research, this study can be extended by segregating the investments in transport infrastructure into public, private, and public-private partnerships, along with urbanized kilometers of road.

Figure 1 .
Figure 1.The weight of transported goods (mean values) by type of transport.Source: Authors' work.Notes: For the definition of variables, please see Table4.

Figure 1 .
Figure 1.The weight of transported goods (mean values) by type of transport.Source: Authors' work.Notes: For the definition of variables, please see Table4.

Figure 2 .
Figure 2. The investments in transport infrastructure (mean values) by type of transport.Source: Authors' work.Notes: Data for Cyprus not available.For the definition of variables, please see Table4.

Figure 2 .
Figure 2. The investments in transport infrastructure (mean values) by type of transport.Source: Authors' work.Notes: Data for Cyprus not available.For the definition of variables, please see Table4.

Table 1 .
Summary review of prior studies towards transport infrastructure and economic growth.

Table 2 .
Brief review of previous studies regarding the impact of transport infrastructure investments on economic growth.

Table 3 .
Summary of previous studies on environmental impact of transport infrastructure.
Source: Authors' work based on literature review.

Table 4 .
Description of the variables.
(26) NOx_non-road Emissions of nitrogen oxides (NOx) from non-road transport (tonnes) (logarithmic values) Eurostat (tsdpc270) '90-'15 (27) NMVOC_road Emissions of non-methane volatile organic compounds (NMVOC) from road transport (tonnes) (logarithmic values) Eurostat (tsdpc280) '90-'15 (28) NMVOC_non-road Emissions of non-methane volatile organic compounds (NMVOC) from non-road transport (tonnes) (logarithmic values) Panel E: Country-level control variables (31) Energy_transportEnergy consumption of transport relative to GDP computed as the ratio between the energy consumption of transport and GDP (chain-linked volumes, at 2010 exchange rates).The energy consumed by all types of transport (road, rail, inland navigation and aviation) is covered, including commercial, individual, and public transport, with the exception of maritime and pipeline transport (Index, 2010 = 100)Source: Authors' work.The definitions of the variables are retrieved from Eurostat, OECD, and World Bank.

Table 7 .
Panel unit root tests output.

Table 8 .
Fixed-effects regressions results on the effect of railway transport infrastructure, related investments, and air pollution on sustainable economic growth.
Figures in brackets depict t-statistic.For the definition of variables, please see Table 4.
Figures in brackets depict t-statistic.For the definition of variables, please see Table 4.

Table 12 .
Fixed-effects regressions results on the effect of air transport infrastructure, related investments, and air pollution on sustainable economic growth.Figures in brackets depict t-statistic.For the definition of variables, please see Table4.

Table 13 .
The output of the Pedroni (Engle Granger-based) panel cointegration test.
length.For the definition of variables, please see Table4.

Table 14 .
The output of the Kao (Engle Granger-based) panel cointegration test.Heteroskedasticity and autocorrelation consistent.Schwarz Info Criterion was selected for lag length.For the definition of variables, please see Table4.

Table 15 .
The output of the Fisher (combined Johansen) panel cointegration test.

Table 16 .
The results of panel fully modified OLS (FMOLS) and dynamic OLS (DOLS) regressions.Panel method: Pooled estimation.Heterogeneous variances.The Schwarz lag and lead method are used in the case of DOLS estimation.Figures in brackets depict t-statistic.For the definition of variables, please see

Table 17 .
Panel vector error correction model (PVECM) Granger causalities.Source: Authors' computations.Notes: Superscripts ***, **, *, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively.ECT reveals the coefficient of the error correction term.The number of appropriate lag is two according to Schwarz information criterion.For the definition of variables, please see Table4.