A Cross-Country European Efficiency Measurement of Maritime Transport: A Data Envelopment Analysis Approach

Abstract

Maritime transport, which includes shipping and port operations, is the fundamental basis of international trade and globalization. In transportation management, efficiency is critical for verifying performance and proposing the best countermeasure to meet predetermined goals. Various efforts in this field have been made to solve this problem satisfactorily. However, the significant proportion of conventional approaches are based on long-term observations and professional expertise, with only a few exceptions based on practice-based historical data. Data Envelopment Analysis (DEA) is a non-parametric technique for analyzing various output and input variables parallelly. The efficiency of maritime transport in European countries is explored using a two-stage DEA approach based on Malmquist and Epsilon-Based Measure (EBM). First, the Malmquist model analyses countries’ total productivity growth rates and their breakdown into technical efficiency (catch-up) and technology change (frontier-shift). Second, the EBM model is used to determine the efficiency and inefficiency of the maritime transportation systems in each European country. Apart from identifying the best-performing countries in specific areas over the study period (2016–2019), the results highlight that the gap in applying the EBM method to maritime transport has been successfully closed and that the emerging paradigm, when combined with the Malmquist model, can be a sustainable and appropriate evaluation model for other research areas.

1. Introduction

Maritime transportation has long been regarded as the backbone of global commerce, accounting for more than 80% of global trade volume [1]. Due to maritime transport’s essential role in facilitating international trade is critical to the global economy and globalization. Maritime transport is a critical element of an economic structure, encompassing everything from shipbuilding to port development to maritime logistics services [2]. The pandemic of coronavirus disease (COVID-19) has highlighted nations’ global interdependence [3] and sparked new trends that will reshape the maritime transport landscape with a 4.1% decline in the volume of international maritime trade in 2020. The world economy experienced severe implications as a result of supply chain disruptions, demand contractions, and global economic uncertainty brought about by the COVID-19 pandemic [4].

Moreover, sea shipping is a derived demand that is heavily influenced by global economic and trade trends. Global economic growth slowed in 2019 as trade tensions remained high and policy uncertainty remained high. Worldwide economic implications were severe as a result of supply chain disruptions, demand contractions, and global economic uncertainty induced by the COVID-19 pandemic. The continuation of trade conflicts between China and the United States, as well as the global economy’s general downturn, hurt both developed and emerging economies [4]. Despite its environmental credentials and the congested state of Europe’s roadways, maritime transport remains the second most popular means of short-distance transit in Europe, trailing only road transport. Because marine freight has various advantages over other means of transport, particularly road freight, such as reduced prices and a lesser environmental effect [5]. For external and internal trade, the European Union (EU) significantly rely on maritime transit. In total, 77% of EU imports and exports, as well as 35% of intra-Union commerce, transit through European seaports. Despite a projected drop in maritime activity in 2020 as a result of the COVID-19 pandemic, this sector is likely to expand dramatically over the next few decades [6].

Due to the critical importance of maritime transport in global trade and economic development in general, particularly in European countries, it has drawn extensive research from a variety of disciplines, including operations research [7], management [8], and environmental science [9]. With the rapid growth of maritime transportation, economic researchers and policymakers face a significant challenge in synthesizing data timely, accurate, and effective to evaluate operational performance. A country’s maritime efficiency can be increased by optimizing its outputs from various inputs. Additionally, it can assist a country in enhancing its competitive advantages in global markets and improve the country’s resource utilization position relative to other countries [10]. In contrast, when a country’s factors of production are inefficient, the country’s economic resources are wasted. The unit’s cost will eventually rise if it is kept in the same conditions [11]. According to the UNCTAD-developed liner shipping connectivity index [12], four European countries are in the top ten most interconnected economies. Thus, containerized, dry bulk and gas cargoes were expected to drive 3.5% annual growth in international maritime trade between 2019 and 2024 [12]. On the other hand, uncertainty continues to be a dominant theme in the current maritime transport environment, with risks skewed to the downside.

From operational perspectives in maritime industry, it is feasible to distinguish between two distinct efficiency measurement methodologies that have been primarily used in the marine industry, namely parametric and non-parametric methods [13]. The parametric method entails developing an ad hoc function capable of inference by specifying the shape of the sample’s statistical distribution in terms of a finite set of parameters to be estimated. To accomplish this, the approach requires the fulfillment of certain hypotheses, including normality, homoskedasticity, independence, and an identical stochastic distribution of the error [14]. Parametric techniques are constrained by the requirement to provide extremely limiting hypotheses that are frequently difficult to justify. However, the distribution model is frequently unknown, or it is impossible to meet these tough hypotheses required for their validity [13]. Among numerous methods for parametric efficiency evaluations in the literature, the Data Envelopment Analysis (DEA) is a non-parametric technique for efficiency evaluation based on various outputs and inputs [15]. Therefore, the DEA approach has been adopted and extensively used by numerous researchers in the various research fields [16,17,18,19]. This is a performance evaluation process that has gained popularity in recent decades due to its suitability for testing various aspects of seashipping efficiency, its ability to conduct multidimensional comparisons [20], its ability to take into account multiple input and output variables concurrently, and the processing of which has resulted in the development of specialized calculation software. The model’s limit is defined by its sensitivity to the input data and parameters. Indeed, alternative outcomes can be reached by varying the input and output variables or by evaluating with fewer parameters.

Given the marine industry’s importance, the first research question arises whether the European countries are efficient in utilizing their geographical positions, seaports and shipping operations? Thus, a two-stage DEA approach is proposed to answer the second research question, “Does the proposed nonparametric DEA method have greater differentiation power when measuring the maritime transportation efficiency of European countries than the parametric method?” This study aims to shed light on the performance evaluation of European maritime industry. In first stage, the DEA Malmquist model is adopted to measure productivity changes of each DMUs. The Epsilon-Based Measure (EBM) model is utilized to assess the effectiveness and inefficiency of each European country’s maritime transportation system in the second stage.

As discussed above, this study provides some practical contributions as follows: First, this study proposes a two-stage DEA approach for an efficiency assessment of maritime transport through panel data of European countries from 2016 to 2019—which represents a first worldwide comparison and generates unique insights into different input-output variables in the context of maritime industry. Second, cross-country comparisons based on efficiency evaluations are rare but essential in terms of governance comparisons. Based on a comprehensive review from existing literature on maritime industry, previous studies applied the same approaches of DEA; Wang et al. [1] investigated DMU’s performance via six financial indicators from seaport firms’ financial statements. Thus, recent research [13] measured the performance efficiency of Vietnam’s top 18 seaports. On the contrary, efficiency scores have been linked to different determinants in this study (e.g., short sea shipping, energy consumption in transportation, containers, number and gross tonnage of vessels, passenger, gross weight of goods transported). As such, previous studies [1,13] focused on the regional scope of a nation. By investigating the cross-national maritime performance in European zone, this study has indicated the significant differentiation in the terms of geographical positions, infrastructure investment, productivity and maritime governance. Therefore, this study addresses a critical gap in the current body of knowledge by emphasizing the critical link between infrastructure and productivity, thereby supporting European governments in planning successful marine investments. Consequently, the findings of this study enable us to prioritize the maritime transport productivity of European countries. Measuring and quantifying the elements affecting the performance of maritime transport can assist managers, policymakers, and planners in boosting economic performance.

The remainder of the paper is divided into the following sections. Section 2 discusses maritime transportation research and the framework for evaluating the efficiency of maritime transportation. Section 3 contains an introduction to the proposed methods. Section 4 illustrates the proposed method through a case study involving European countries, and Section 5 presents conclusions and future research.

2. Literature Review

The DEA technique is a widely used technique for assessing the efficiency of decision-making units (DMUs). In 1978, Charnes, Cooper, and Rhodes (CCR) introduced DEA analysis [21]. Nevertheless, concepts for assessing efficiency have existed for a long length of time. Farrell developed the production possibility frontier as a criterion for assessing the efficacy of enterprises operating in the same industry in 1957. The production possibility frontier comprises two components: resource allocation efficiency and total technological efficiency [22]. The CCR technique utilized a non-parametric methodology to generate a production possibility frontier curve based on DMU data collection [21]. Then, the efficiency of the DMUs was calculated and compared using various mathematical programming techniques. In 1984, Banker, Charnes, and Cooper (BCC) extended the CCR model by calculating variable returns to scale (VRS) scenarios [23]. The BCC model that resulted enabled a more detailed examination of DMU efficiency. Tone [24] suggested a slacks-based measure of efficiency (SBM) that directly incorporates the slacks’ goal function reflecting the units’ input excess and output shortage. Thus, the SBM model is non-radial, as its inputs and outputs do not require simultaneous verification [25], whereas the EBM model accounts for both radial and non-radial components, allowing for a more precise evaluation of efficiency. The EBM model accurately reflects the unit’s efficiency under evaluation and the difference between its radial and non-radial components [26]. The DEA Malmquist model is a highly valuable modification of the conventional DEA model for measuring the productivity of DMUs, with the Malmquist productivity index (MPI) equal to the product of the catch-up index and the frontier-shift index [27].

In recent years, researchers have focused their attention on transportation sustainability measurement. For instance, Qiu et al. [28] proposed a fuzzy evaluation approach to determine Wenzhou, China’s transportation sustainability. Additional methods have been widely adopted, including Analytic Hierarchical Process (AHP) [29]. Notably, the evaluation process for AHP or PCA is frequently extremely subjective, as the weights for each indicator must be accurately established in advance. DEA [21] has also been offered as a method for assessing transportation sustainability. For example, Yu and Lin [30] suggested the multi-activity network DEA model for measuring railway performance efficiency by considering both production and consumption technologies and then comparing the European commercial transport industry’s CO2-sensitive productivity growth. Shiau and Jhang [31] deployed a DEA technique based on rough set theory to assess Taiwan’s transportation sustainability. Lin [32] employed a generalized DEA to assess the relative energy efficiency of rail, road, aircraft, and water transport, and extended the DEA model to forecast future energy consumption in China’s transportation sector. Wu [33] applied DEA to assess Chinese transportation networks’ energy and environmental performance. Zhang and Song [34] compared the super-efficiency DEA model and the traditional DEA model to assess railways, highways, and aviation’s coordinated growth in Beijing-Tianjin-Hebei. Zhou and Hu [35] suggested a network DEA model with unwanted outputs and shared resources for evaluating the sustainability of China’s eastern railway. Stefaniec et al. [36] suggested a network DEA approach based on the triple bottom line for assessing inland transportation in China, considering sustainability’s social, economic, and environmental components. To assess the regional transport sustainability efficiency in Shaanxi province, China, Tian et al. [37] developed super-efficiency slacks-based measure DEA (SBM-DEA) model with weighting preference. Jose-Kandappassery et al. [38] created a non-radial DEA model to evaluate the sustainability of European transportation networks by assessing a wide variety of economic, environmental, and social (EES) aspects. It sees evaluation as an efficiency analysis method and disregards subjectivity in computing indicator weights when using DEA models.

Table 1. A summary of previous study methodologies and problem features.

Authors/Year	Inputs	Outputs	Methodology	Applied Areas
Yu and Lin (2008)	Number of employees Length of lines Number of passenger cars Number of freight cars	Passenger train-km Freight train-km Purchasing power parity Population density	Multi-activity network DEA	Railway firms in the world
Shiau and Jhang (2010)	Land take by road infrastructure Fossil fuel consumption by transport system	Vehicle-Kilometer (VKM) traveled by private mode VKM traveled by transit VKM traveled by truck	CCR-DEA RST	Transport systems in EU
Lin et al. (2015)	Energy	Passenger-kilometers Freight ton-kilometers	CCR-DEA	Transport Modes in China
Wu et al. (2016)	Cargo tonnage Energy Capital Highway mileage	Freight turnover volume CO2	CCR-DEA	Transportation systems in China
Zhang and Song (2017)	Traffic infrastructure investment Number of employees	Freight traffic Passenger traffic	SE-DEA	Beijing-Tianjin-Hebei region in China
Zhou and Hu (2017)	Capital Land Labor	Railway mileage Railway density Passenger turnover Freight turnover Average salary growth	Undesirable outputs-DEA Shared inputs-DEA	Railway transport of China’s eastern
Stefaniec et al. (2020)	Vehicles Capital Employment Energy consumption	Accessibility Traffic casualties Value-added Turnover Green energy usage CO2 emissions	Undesirable-DEA	Inland transport of China
Tian et al. (2020)	Length of transport routes Transport density Capital investment Staff input Expenditure in transportation Input of vehicles Land take by transportation Fossil fuel consumption	Quality of transport routes On-time rate of vehicles Accessibility to rural area Speed of vehicles Education level Traffic accident Output value Income level Passenger capacity Freight capacity Passenger-Kilometers (PKM) Ton-kilometers (TKM) Carbon emissions Air pollution emissions Traffic noise pollution	Super SBM-DEA	Transport in Shaanxi province, China
Jose-Kandappassery et al. (2021)	Infrastructure spending	Carbon Dioxide emissions Energy consumption Gross Value Added	Super SBM-DEA	Transport Systems in Europe

summarizes the related DEA works.

According to the aforementioned literature assessment, no research has been undertaken to evaluate maritime transport in European countries using the DEA Malmquist and EBM models. As a result, this shortcoming attracted our prominence to conduct this study. By examining economic, social, and environmental sustainability and identifying factors affecting maritime transportation networks in the European Union, this study seeks to fill a gap in the existing literature on transportation system efficiency (EU). The DEA methodology is an efficient and practical method for determining the most efficient DMUs in this evaluation from a collection of DMUs. The model is specifically used to respond to five inputs (short sea shipping, energy consumption in transport, labor force, containers, number and gross tonnage of vessels). Outputs include passengers and the gross weight of commodities moved. The study’s insights can aid governments and policymakers in evaluating and enhancing the operational efficiency of maritime transportation and associated industries.

3. Methods

3.1. Proposed Research Framework

The efficiency of maritime transport in European countries is analyzed for the period 2016–2019 using a two-stage DEA model of Malmquist and EBM. The research process is separated into two parts, as depicted in Figure 1. The DEA Malmquist model assesses total productivity change due to technical efficiency improvements (catch-up) and technology investment (frontier-shifting) using selected input and output variables in the first stage. Prior to implementing the Malmquist model, Pearson correlation is used to guarantee that the dataset is homogeneous and isotonic. The second stage is computing the DMUs’ efficiency and inefficiency scores using the EBM model. If the EBM model is applicable, the diversity indices and affinity coefficients are validated.

3.2. DEA Malmquist Model

The Pearson correlation test [39] is used to determine whether selected variables have positive associations before using DEA models via Equation (1):

$$r_{uv}=\frac{{{\displaystyle \sum }}_{i=1}^n\left(u_0^i-\overline{u}\right)\left(v_0^i-\overline{v}\right) }{\sqrt{{{\displaystyle \sum }}_{i=1}^n{\left(u_0^i-\overline{u}\right)}^2{{\displaystyle \sum }}_{i=1}^n{\left(v_0^i-\overline{v}\right)}^2}}$$

(1)

where $r_{uv}$ denoted as the correlation coefficient value, n is the sample size; $v_0^i$ is the ith input and $u_0^i$ the ith output; $\overline{u}$ and $\overline{v}$ are the avarage values for DMUs_.

DEA-based Malmquist productivity index [40] measures the productivity of DMUs change over time period $t$ to period $t+1$. $T{E}_0^t\left(v_0^t,u_0^t\right)$ denotes denotes the proportional reduction in observed inputs required to produce the specified output level in time period t. $T{E}_0^{t+1}\left(v_0^{t+1},u_0^{t+1}\right)$ is the technical efficiency (catch-up) score for DMUs in time period t + 1, shown in Equation (2):

$$\mathrm{Technical} \mathrm{efficiency} \left(T{E}_0\right)=\frac{T{E}_0^{t+1}\left(v_0^{t+1},u_0^{t+1}\right)}{T{E}_0^t\left(v_0^t,u_0^t\right)}$$

(2)

$F{S}_0$ measures technology efficiency (frontier-shift) between time period t and t + 1 using Equation (3):

$$\mathrm{Technological} \mathrm{efficiency} \left(F{S}_0\right)=\sqrt{\frac{T{E}_0^t\left(v_0^{t+1},u_0^{t+1}\right)}{T{E}_0^{t+1}\left(v_0^{t+1},u_0^{t+1}\right)} x\frac{T{E}_0^t\left(v_0^t,u_0^t\right)}{T{E}_0^{t+1}\left(v_0^t,u_0^t\right)}}$$

(3)

Total productivity (MPI) is calculated using Equation (4):

$$\mathrm{MPI}=\left(T{E}_0\right)\mathrm{X} \left(F{S}_0\right)=\frac{T{E}_0^{t+1}\left(v_0^{t+1},u_0^{t+1}\right)}{T{E}_0^t\left(v_0^t,u_0^t\right)}x\sqrt{\frac{T{E}_0^t\left(v_0^{t+1},u_0^{t+1}\right)}{T{E}_0^{t+1}\left(v_0^{t+1},u_0^{t+1}\right)} x\frac{T{E}_0^t\left(v_0^t,u_0^t\right)}{T{E}_0^{t+1}\left(v_0^t,u_0^t\right)}}$$

(4)

It is well noted that $\mathrm{MPI}=\left(T{E}_0\right)\mathrm{X} \left(F{S}_0\right)$ implies into three scenarios:

• MPI value > 1 indicates that the productivity gain is from the combined effect of average technology progress and relative efficiency improvement.
• MPI value = 1 indicates that the productivity is constant.
• MPI value < 1 indicates that the productivity loss is from the combined effect of average technology regress and relative efficiency improvement.

3.3. DEA Epsilon-Based Measure Efficiency

The DEA EBM-I-C model [41] contains two parameters, one scalar and one vector, which are defined by an affinity index with respect to the inputs and outputs. These two factors are defined to combine the radial and nonradial models into a single model for evaluating the efficacy of DMUs. The EBM model considers $n$ DMUs ($j=1, 2, \dots , n$) including $v_0^i$inputs ($i=1, 2, \dots , m$) and $u_0^i$outputs ($i=1, 2, \dots , s$). $X=\left\{x_{ij}\right\}\in R^{m\times n}$ and $Y=\left\{y_{rj}\right\}\in R^{s\times n}$ are denoted as non-negative matrices.

$$\begin{gathered} {\delta }^{*}=\underset{\theta ,\lambda ,s^{-}}{Min}\theta -{\epsilon }_x{{\displaystyle \sum }}_{i=1}^m\frac{w_i^{-}{s}_i^{-}}{x_{io}} \\ \mathrm{subject} \mathrm{to} \\ {\displaystyle \sum }_{j=1}^n{x}_{ij}{\lambda }_j=\theta x_{io}-s_i^{-},i=1,\dots ,\mathrm{m} \\ {\displaystyle \sum }_{j=1}^n{y}_{rj}{\lambda }_j\ge y_{ro},r=1,\dots ,\mathrm{s} \\ {\lambda }_j\ge 0, j=1, 2,\dots ,n \\ {s}_i^{-}\ge 0,i=1,2,\dots ,\mathrm{m} \end{gathered}$$

(5)

where ${\lambda }_j$ denotes the intensive vector of DMU, the subscript “$o$” denotes the DMU is under evaluation, $s_i^{-}$ and $w_i^{-}$ denote the amount of non-radial slack and relative weight in the ith input, a key parameter ${\epsilon }_x$ which depends on the degree of dispersion of inputs, and $\theta$ denotes the radial properties.

$DIV\left(u_0^i , v_0^i \right)$ is denoted as diversity index and $AFF\left(u_0^i ,v_0^i \right)$is denoted as affinity index. The EBM model need to satisfy the conditions, which are $0\le DIV\left(u_0^i ,v_0^i \right)=DIV\left(u_0^i ,v_0^i \right) \le 0.5$, and $0\le AFF\left(u_0^i ,v_0^i \right)=1-2DIV\left(u_0^i ,v_0^i \right) \le 1$.

4. Empirical Analysis

4.1. Data Collection

The datasets were collected from 24 DMUs between 2016 and 2019. The short sea shipping, energy consumption, containers, labor force, number and gross tonnage of vessels, passenger, gross weight of goods transported are all collected in the databases of the European statistics website [42]. The labor force is detailed in The WorldBank report [43] (

Table 2. Selected output and input variables.

Variables	Definitions	Units
Short sea shipping	The maritime transport of goods over relatively short distances, as opposed to the intercontinental cross-ocean deep-sea shipping	Thousand tons
Energy consumption in transport	The total energy utilized by all forms of transport: road (buses, trucks, etc.), railway (trains, metro, etc.), marine, airway, and pipeline transport.	Thousand tons of oil equivalent
Labor force	All people in a country or region who are legally permitted to work.	Thousands of persons
Containers	The type of cargo classification deals with containers that are moved between the vessel and the port by being lifted on or lifted off. This involves the use of specialized equipment to attach to the fittings on the container to allow such movements	TEU
Number and gross tonnage of vessels	The number of vessels and overall size of a ship are determined as a function of the molded volume of all enclosed areas on board.	Number
Passenger	There are numbers of passengers moving between a port of embarkation and a port of disembarkation. These are the quantities that define the amount of maritime passenger transport carried out	Thousand passengers
Gross weight of goods transported	The tonnage of goods carried, including packaging, excludes containers’ tare weight or Ro-Ro units.	Thousand tons

).

Table 3. Statistical description.

Year	Statistics	I1	I2	I3	I4	I5	O1	O2
2016	Max	286,148	56,557	43,058,646	15,408	456,759	67,273	588,230
	Min	3483	200	215,527	115	2138	3	3781
	Average	111,133	11,980	9,604,118	4003	88,294	15,745	155,451
	SD	101,132	15,578	12,077,618	5106	126,772	20,156	160,120
2017	Max	316,196	57,247	43,288,289	16,013	470,424	595,807	470,424
	Min	3707	207	223,541	124	2021	4100	2021
	Average	113,040	12,227	9,694,461	4186	89,325	160,970	89,325
	SD	103,912	15,773	12,200,068	5269	129,214	166,301	129,214
2018	Max	313,739	55,470	43,562,285	17,210	478,567	85,382	604,000
	Min	4351	231	236,718	134	1878	0	4559
	Average	115,485	12,272	9,761,664	4518	96,135	17,440	164,650
	SD	105,458	15,604	12,310,614	5650	139,843	23,406	168,672
2019	Max	320,701	56,220	43,871,267	17,408	472,540	86,530	607,500
	Min	4324	246	249,860	133	1649	0	5195
	Average	117,264	12,359	9,803,084	4542	97,529	17,771	165,886
	SD	107,922	15,729	12,368,825	5648	143,296	23,816	170,661

Note: calculated by the authors.

presents a descriptive analysis of maritime transportation’s input and output variables for 2016–2019. Regarding the results from

Table A1. Correlation coefficient.

Year		I1	I2	I3	I4	I5	O1	O2
2016	I1	1.000	0.544	0.547	0.669	0.266	0.280	0.862
	I2	0.544	1.000	0.974	0.761	0.240	0.366	0.635
	I3	0.547	0.974	1.000	0.754	0.205	0.292	0.644
	I4	0.669	0.761	0.754	1.000	0.292	0.301	0.874
	I5	0.266	0.240	0.205	0.292	1.000	0.931	0.227
	O1	0.280	0.366	0.292	0.301	0.931	1.000	0.222
	O2	0.862	0.635	0.644	0.874	0.227	0.222	1.000
2017	I1	1.000	0.538	0.561	0.665	0.253	0.858	0.253
	I2	0.538	1.000	0.976	0.743	0.225	0.633	0.225
	I3	0.561	0.976	1.000	0.745	0.202	0.652	0.202
	I4	0.665	0.743	0.745	1.000	0.275	0.895	0.275
	I5	0.253	0.225	0.202	0.275	1.000	0.222	1.000
	O1	0.858	0.633	0.652	0.895	0.222	1.000	0.222
	O2	0.253	0.225	0.202	0.275	1.000	0.222	1.000
2018	I1	1.000	0.562	0.574	0.698	0.274	0.323	0.866
	I2	0.562	1.000	0.976	0.743	0.226	0.356	0.645
	I3	0.574	0.976	1.000	0.736	0.189	0.289	0.651
	I4	0.698	0.743	0.736	1.000	0.303	0.329	0.904
	I5	0.274	0.226	0.189	0.303	1.000	0.935	0.236
	O1	0.323	0.356	0.289	0.329	0.935	1.000	0.259
	O2	0.866	0.645	0.651	0.904	0.236	0.259	1.000
2019	I1	1.000	0.559	0.583	0.710	0.286	0.309	0.890
	I2	0.559	1.000	0.974	0.723	0.240	0.354	0.636
	I3	0.583	0.974	1.000	0.725	0.196	0.283	0.653
	I4	0.710	0.723	0.725	1.000	0.287	0.297	0.899
	I5	0.286	0.240	0.196	0.287	1.000	0.946	0.248
	O1	0.309	0.354	0.283	0.297	0.946	1.000	0.256
	O2	0.890	0.636	0.653	0.899	0.248	0.256	1.000

Note: calculated by the authors.

in the Appendix A, positive and significant correlation coefficients range from 0.189 to 0.976. Therefore, the dataset can be used in the second procedure to develop the DEA Malmquist model.

4.2. DEA Malmquist Model Results

Table 4. DEA Malmquist analysis results.

	2016–2017			2017–2018			2018–2019
DMU	MPI	Catch-Up	Frontier Shift	MPI	Catch-Up	Frontier Shift	MPI	Catch-Up	Frontier Shift
Belgium	1.537	1.181	1.301	0.730	0.924	0.789	0.999	0.997	1.002
Bulgaria	1.389	1.013	1.371	0.580	0.615	0.942	1.258	1.600	0.786
Denmark	2.716	1.485	1.829	0.360	0.667	0.539	0.996	0.971	1.025
Germany	1.318	1.161	1.136	0.789	0.877	0.899	1.021	1.020	1.000
Estonia	1.423	0.895	1.589	0.875	1.090	0.803	1.031	1.032	0.999
Ireland	1.380	1.137	1.214	0.756	0.879	0.860	0.971	0.981	0.990
Greece	3.104	1.814	1.711	0.334	0.580	0.576	1.008	0.987	1.022
Spain	1.260	1.024	1.230	0.806	0.977	0.825	0.520	0.525	0.991
France	1.007	0.623	1.618	0.818	1.042	0.785	0.980	0.972	1.009
Croatia	2.962	1.885	1.571	0.342	0.543	0.631	0.952	0.935	1.018
Italy	2.419	1.350	1.791	0.437	0.770	0.567	1.023	1.018	1.005
Cyprus	1.181	1.230	0.960	0.686	0.601	1.142	1.064	1.063	1.001
Latvia	1.301	0.919	1.415	0.762	1.052	0.724	0.966	0.964	1.002
Lithuania	1.431	1.033	1.384	0.628	0.739	0.850	0.997	1.012	0.986
Malta	0.890	0.402	2.213	1.109	2.475	0.448	0.977	0.942	1.037
The Netherlands	1.434	1.019	1.408	0.721	0.978	0.736	1.012	1.020	0.993
Poland	1.493	1.306	1.143	0.746	0.843	0.885	1.008	1.004	1.003
Portugal	1.552	1.276	1.217	0.658	0.793	0.829	0.979	0.977	1.002
Romania	1.452	1.053	1.379	0.986	1.674	0.589	1.133	1.022	1.108
Slovenia	1.595	1.213	1.316	1.008	1.666	0.605	0.850	0.996	0.853
Finland	0.983	0.611	1.610	1.069	1.624	0.658	0.996	0.994	1.002
Sweden	1.254	1.067	1.176	0.824	0.956	0.862	0.951	0.953	0.998
Norway	1.427	1.058	1.349	0.737	0.952	0.774	1.015	1.019	0.996
Turkey	1.571	1.257	1.250	0.705	0.815	0.865	1.025	1.020	1.004

Note: calculated by the authors.

contains the MPI results, representing the degree of productivity change between 2016 and 2017. Almost all DMUs, except for Malta and Finland, have an MPI greater than 1, indicating that most DMUs have experienced positive productivity growth over time. Greece has the highest Malmquist productivity index, having greatly caught up with practices, while the technology frontier has advanced by 71%, indicating Greece’s success in adopting new technologies. Croatia (2.962) has the second-highest MPI score, with a catch-up best practices value of 1.885 and a frontier shift value of 1.571, suggesting improvement in both domains.

Except for Estonia, France, Latvia, Malta, and Finland, the majority of DMUs have catch-up values more than 1, resulting in a large reduction in their Malmquist productivity index. Furthermore, with the exception of Cyprus, the majority of DMUs have border shift values greater than one. This illustrates that technological innovation has accelerated productivity growth at the majority of DMUs.

Three DMUs, Estonia, France, and Latvia, have catchup indices of 0.895, 0.623, and 0.919, respectively, with a frontier shift greater than one. This result demonstrates that while the technical efficiency of these DMUs decreased in 2017, their productivity increased as a result of technological advancement. Additionally, while Malta and Finland have catch-up indices of 0.402 and 0.611, respectively, and Malta’s frontier shift reaches a peak of 2.213 and Finland’s frontier shift reaches 1.610, the productivity performance of these two DMUs remains the worst, indicating that the technical efficiency of these two DMUs decreased significantly in 2017, having a significant impact on productivity efficiency. Moreover, Table 4 shows the DEA Malmquist analysis results for 2017–2018. Three terminals Malta, Slovenia, and Finland—reached the best positive productivity growth across this period, of 10.9%, 0.8% and 6.9%, respectively. These DMUs are recorded with productivity growth in contrast to 2016–2017. Malta, Slovenia, and Finland recorded positive technical efficiency but lost ground in frontier technology.

Only seven terminals, including Estonia, France, Latvia, Malta, Romania, Slovenia, Finland, achieved technical efficiency. In contrast, only Cyprus achieved technological efficiency during this period, proving that these DMUs attained productivity efficiency largely based on technical efficiency. Among DMUs, Malta got the best in terms of catch-up performance, with a score of 2.475, but the worst in terms of frontier shift scores, with a score of less than 1. This reflects an increase in technological efficiency but a significant decrease in technology utilization.

Three DMUs, including Estonia, France, and Latvia, achieved MPI values less than 1. Notably, these three DMUs received a high MPI score of 1.423, 1.007, and 1.301 in 2016–2017; however, this increase in productivity reversed in 2017–2018, with Estonia’s productivity decreasing by 54.8%, France’s by 18.9%, and Latvia’s by 53.9%, respectively, indicating that these DMUs did not achieve productivity efficiency during this period. This decline in production is entirely related to regressive frontier changes, as they have greatly increased their catch-up scores during this period. Also, Finland’s productivity increased remarkably, rising by 8.7% compared to 2016–2017, helping Finland rise to second place in productivity efficiency, while this DMU has almost the worst performance among DMUs in the period 2016–2017. This was due to reduced frontier shift, falling to 0.658, while a progressive push in the catch-up index of 1.624.

After a period of fluctuation and a decline in total productivity, 11 out of 24 DMUs have a Malmquist index greater than one for the 2018–2019 period, indicating that these DMUs performed well in terms of total factor productivity. It is particularly important to mention Bulgaria, which has consistently outperformed over the periods with an MPI score of 1.258. In comparison, the remaining DMUs did not perform well in terms of productivity, with Spain scoring the lowest at 0.520 on the MPI.

11 out of the 24 DMUs achieved technical efficiency in the last research period, while 15 DMUs attained technological efficiency. In particular, the significant enhancement of Bulgaria must be considered, given that the preceding period was among the group with the lowest technical index. Bulgaria achieved the highest technical performance of 1.600 during this era, while the technological performance demonstrated the lowest efficiency performance of 0.786. This reflects progress in terms of technical efficiency but a significant lag in terms of technological advancement. However, following a period of poor technological performance, most countries focused their efforts in the 2018–2019 era on investing in operational technology to increase their efficacy. It is necessary to mention Romania, which achieved tremendous growth in comparison to other DMUs with a frontier shift score of 1.108, indicating a strong development in the application of high technology to this country, while Romania was at the bottom of the DMUs with the worst performance in technology efficiency from 2017 to 2018, indicating that a significant effort is being made to improve Romania’s technology efficiency in this period.

4.3. DEA-EBM Results

The DEA Malmquist model results reveal the current operational profile of maritime transport in 24 EU member states after examining total productivity changes. This stage applies the EBM model to identify the efficiency and inefficiency ratings of 24 DMUs between 2016 and 2019. This article makes use of an input-oriented model with a constant rate of return to scale (EBM-I-C). EBM techniques assess the variety of production alternatives defined by the affinity matrix constructed from observable input and output variables.

The diversity index and affinity index for the EBM model are shown in Table A1 and

Table A2. Diversity matrix.

Year	Inputs	I1	I2	I3	I4	I5
2016	Short sea shipping (I1)	0.000	0.219	0.197	0.212	0.220
	Energy consumption in transport (I2)	0.219	0.000	0.198	0.187	0.229
	Labor force (I3)	0.197	0.198	0.000	0.175	0.227
	Containers (I4)	0.212	0.187	0.175	0.000	0.207
	Number and gross tonnage of vessels (I5)	0.220	0.229	0.227	0.207	0.000
2017	Short sea shipping (I1)	0.000	0.231	0.197	0.218	0.224
	Energy consumption in transport (I2)	0.231	0.000	0.185	0.224	0.203
	Labor force (I3)	0.197	0.185	0.000	0.201	0.201
	Containers (I4)	0.218	0.224	0.201	0.000	0.206
	Number and gross tonnage of vessels (I5)	0.224	0.203	0.201	0.206	0.000
2018	Short sea shipping (I1)	0.000	0.248	0.212	0.202	0.215
	Energy consumption in transport (I2)	0.248	0.000	0.156	0.212	0.220
	Labor force (I3)	0.212	0.156	0.000	0.206	0.201
	Containers (I4)	0.202	0.212	0.206	0.000	0.202
	Number and gross tonnage of vessels (I5)	0.215	0.220	0.201	0.202	0.000
2019	Short sea shipping (I1)	0.000	0.230	0.218	0.209	0.212
	Energy consumption in transport (I2)	0.230	0.000	0.154	0.221	0.219
	Labor force (I3)	0.218	0.154	0.000	0.209	0.202
	Containers (I4)	0.209	0.221	0.209	0.000	0.210
	Number and gross tonnage of vessels (I5)	0.212	0.219	0.202	0.210	0.000

Note: calculated by the authors.

, respectively, over the period 2016–2019. The values of the diversity and affinity matrices range from 0 to 0.248 and 0.505 to 1, respectively. These figures correspond to the model’s specifications. Thus, EBM can be utilized to assess the efficacy/inefficacy of DMUs. Following that,

Table 5. Weight to input/output and epsilon for EBM model.

Period	Weight to Input/Output					Epsilon
	Short Sea Shipping	Energy Consumption in Transport	Labor Force	Containers	Number and Gross Tonnage of Vessels
2016	0.197	0.199	0.204	0.206	0.192	0.414
2017	0.195	0.199	0.207	0.198	0.200	0.418
2018	0.194	0.199	0.207	0.201	0.199	0.414
2019	0.195	0.202	0.207	0.198	0.199	0.416

Note: calculated by the authors.

calculates the input/output and epsilon weights for the EBM model. It demonstrates that over the 2016–2019 assessment period, the epsilon for the EBM model is positive, a critical parameter for combining radial and non-radial properties (i.e., the values range from 0.414 to 0.418).

The relative ranking efficiency and inefficiency of 24 DMUs between 2016 and 2019 are calculated using the EBM model’s epsilon and the weight to input/output ratio, as shown in

Table 6. Efficiency score of EBM model.

DMUs	2016	2017	2018	2019
Belgium	0.773	0.903	0.859	0.853
Bulgaria	0.865	0.908	0.827	0.854
Denmark	0.750	0.965	0.763	0.758
Germany	0.665	0.788	0.672	0.692
Estonia	1	1	1	1
Ireland	0.671	0.834	0.678	0.656
Greece	0.675	0.967	0.690	0.686
Spain	0.837	0.970	0.875	0.795
France	0.901	0.845	0.857	0.840
Croatia	1	1	1	0.971
Italy	0.574	0.820	0.592	0.629
Cyprus	0.518	0.787	0.440	0.446
Latvia	1	1	1	1
Lithuania	0.952	1	0.888	0.895
Malta	1	1	1	1
The Netherlands	1	1	1	1
Poland	0.536	0.792	0.595	0.588
Portugal	0.688	0.832	0.687	0.678
Romania	0.770	0.834	0.886	0.940
Slovenia	0.723	0.835	0.860	0.845
Finland	1	0.861	1	1
Sweden	0.845	0.904	0.824	0.788
Norway	1	1	1	1
Turkey	0.683	0.840	0.675	0.672

Note: calculate by the authors.

and Figure 2. Estonia, Latvia, The Netherlands, and Norway all attain a high level of efficiency, with a score of 1 and no slack, for the 2016–2019 period. Additionally, Estonia, Croatia, Latvia, Malta, The Netherlands, and Norway achieve a high efficiency score during a three-year period, from 2016 to 2017. Lithuania fared admirably in 2017, but then struggled in 2018 and 2019. In comparison, Finland had a bad result in 2017 before making a serious attempt to improve in later years. Belgium, Bulgaria, Denmark, Germany, Ireland, Greece, Spain, France, Italy, Cyprus, Poland, Portugal, Romania, and Slovenia all have inefficient energy systems. Italy, Cyprus, and Poland, in particular, had the lowest efficiency scores for the 2016–2019 timeframe evaluated. It is critical to demonstrate that these DMUs with an EBM efficiency score less than 1 (particularly the top three with the lowest score) should monitor proactively by increasing technical efficiency, i.e., maximizing their available resources such as capital, labor, equipment supplies, and concentrating on technological investment effect, i.e., technology development. This implies the likelihood of more profitable investment in countries with an EBM efficiency score of one or above.

4.4. Discussions

Regardless of the novelty of previous research, most of them have focused on the impact of transportation infrastructure on economic growth in a single country, either at the national or regional level [1,44]. In light of the DEA models’ results in Table 4 and Table 6, this study sheds light on the sea shipping industry’s performance in 24 EU countries from 2016 to 2019. The marine industry, which encompasses ports, shipping, and logistics, has adopted environmental rules and laws to mitigate negative environmental consequences and carbon emissions [45].

Our findings are aligned with the reports of UNCTAD [46], Estonia, Croatia, Latvia, Malta, The Netherlands, and Norway achieved a high maritime transport efficiency during 2017. However, notice the contrast between our results and the reports [4,47], the UNCTAD liner shipping connectivity index indicated that Belgium, Germany, United Kingdom are the best-connected countries among global leaders. To be more specific, the liner shipping connectivity index reflects both changes in demand and decisions taken by carriers, which in turn depend on their strategic vessel deployment and responses to port investments and reforms in the container ports of countries.

Among the selected input variables, short sea shipping is considered one of the most sustainable and economically competitive means of transport compared to road. The EU countries support short sea shipping by identifying priority areas for EU action to boost the maritime sector’s competitiveness and environmental performance [48]. Our results are consistent with the study of [49], which indicated that European governments had taken political and technological steps to increase energy efficiency and reduce energy usage in the maritime sector. Besides, energy consumption has become a major environmental issue for European administrations. Policymakers implementing measures promoting growth and steering economies toward sustainable development and macroeconomic and political stability may increase maritime transport efficiency. To boost marine efficiency, EU countries must develop public policies that support maritime diversification, development of shipbuilding, maintenance activities, and port activities, even in labor-intensive industries [50]. Countries that do not efficiently manage their resources confront political restrictions as a result of limited investment, weak productivity growth, a volatile economic environment, and insufficient infrastructure [10].

Maritime transport comprises a diverse variety of activities and, when combined with port operations and logistic hubs, has a major impact on the development of maritime sectors and trade, stimulating economic growth and job creation. As previously stated, maritime transport and related activities (shipbuilding, maintenance, and port operations) generate approximately 40% of the blue economy’s value added and employ approximately 24% of its workforce [50]. In our opinion, investments in maritime port infrastructure boost port facilities, resulting in improved handling capacity and a good effect on the volume of commodities that may be processed in ports. Likewise, this may necessitate the hiring of additional employees, resulting in a decrease in unemployment. On the other hand, in the absence of significant green transition measures, the more intensive maritime transport is, the more pollution it generates.

5. Conclusions, Limitations and Future Studies

5.1. Conclusions

The DEA approach is regarded as an appropriate tool for evaluating the relative efficiency of DMUs with a large number of inputs and outputs because it avoids making subjective assumptions about the links between inputs and outcomes. Between 2016 and 2019, the operational efficiency of maritime transport in 24 European Union member countries was evaluated using DEA Malmquist and EBM models. It incorporates short sea shipping, energy usage during transport, containerization, and vessels’ number and gross tonnage. The passengers and gross weight of goods transported are the two outputs in this study. As a result, Estonia, Croatia, Latvia, Malta, The Netherlands and Norway rank among the countries with the highest efficiency scores for maritime transportation. In general, the productivity efficiency of the maritime transport system in the 24 EU member states has tended to decrease. This study may be valuable for stakeholders, i.e., governments and organizations, in studying and improving the operational efficiency of maritime transportation and a variety of other sectors in terms of managerial implications. The use of a DEA-based approach in this study assists in identifying countries that have not been maximizing the use of their resources in maritime transportation (those with low efficiency). As a result, there is a significant opportunity for them to boost the efficiency of their transportation system, which may be accomplished by more efficient utilization of their existing resources rather than resource expansion. The cause of the shift can be traced back to political or economic circumstances.

The study presented here adds knowledge and practical applications to the subject of maritime performance measurement. The findings are essential because they will assist marine operators in better understanding and controlling critical indicators utilized in maritime transportation operations and processes. As a result, operators’ performance in technical and technological sectors may be enhanced. Due to rising competition in maritime logistics, technology has risen to prominence [51], despite the fact that maritime transit is critical to the global economy, accounting for more than 90% of worldwide trade. Subsequently, marine, like the majority of other industries, is undergoing rapid transformation as a result of a number of technical advances aimed at making operations more environmentally friendly, cost effective, and efficient. Meanwhile, the COVID-19 pandemic has upended maritime transport, presenting unprecedented problems for industry specialists. In light of this, we must maximize the potential of technology in order to develop a more sustainable shipping industry [52]. Therefore, the managerial implications of this study could assist managers in developing long-term strategies that positively contribute to broader systemic resilience.

Briefly stated, the following are the most significant contributions made by this study: First, this study proposes a two-stage DEA approach for an efficiency assessment of maritime transport through panel data of European countries from 2016 to 2019—which represents a first worldwide comparison and generates unique insights into different input-output variables. Cross-country comparisons based on efficiency evaluations are rare but essential in terms of governance comparisons. Second, efficiency scores have been linked to six critical indicators (e.g., short sea shipping, energy consumption in transportation, containers, number and gross tonnage of vessels, passenger, gross weight of goods transported), filling an important gap in current literature and thus highlighting an important problem (i.e., the link between infrastructure and productivity), assisting European governments in planning effective maritime investments. Finally, the findings of this study enable us to create a ranking of European countries with the best maritime transport productivity. Furthermore, measuring maritime transport performance and identifying the factors that influence it can aid government stakeholders in improving economic performance.

5.2. Limitations and Future Studies

Although this novel two-stage DEA strategy offers stakeholders unique insights into the maritime sector. This study has some limitations, and much additional work needs to be done in the future.

First, the EU, like many other countries, has imposed lock-down measures and restricted migration in response to the pandemic. The country’s lockdown reduced demand across the board, producing problems in transportation networks, including the marine sector, particularly the port and shipping sector. It is obvious that this study did not consider the COVID-19 impacts to reduce the volatility. In this study, the authors investigated the maritime transport performance of European countries based on the dataset from 2016 to 2019. With further data accessible and valid through 2021, the maritime industry’s overall efficiency under COVID-19 impacts may be assessed in future research.

Second, this study employed a combination of DEA Malmquist and EBM approaches with all “desired” input and output variables. However, it is also worthwhile to investigate the issue of output combination with undesirable output such as environmental dimensions including greenhouse gas emissions, energy consumption, and traffic noise pollution.

Third, the proposed DEA approach is not the only way to assess the relative efficiency of maritime operations in the EU, and it cannot forecast future performance. As a result, future research should integrate DEA with the Grey prediction model GM (1,1) [53], the DEA resampling predicting technique [54], or machine learning predicting techniques [55] and multi-criteria decision-making (MCDM) models [56,57] for evaluating and forecasting to provide more detailed information to governments and policymakers.