Quantifying the impact of COVID-19 on Chinese ports

Abstract

Purpose

The purpose of this study is to quantify port efficiency assessment indicators to analyze the impact of COVID-19 on Chinese One Belt One Road (OBOR) ports.

Design/methodology/approach

This study utilized a grey prediction model GM(1,1) to forecast five relevant indicators for each of the 17 OBOR ports both with and without COVID-19 background conditions. Additionally, the data envelopment analysis (DEA) efficiency assessment approach was used to analyze the impact of COVID-19 on port efficiency.

Findings

The results indicate that cargo and container throughput growth rates during the COVID-19 pandemic are reduced by 1.7 and 2.1%, respectively. There was also a noticeable reduction in technological efficiency (TE) as well as pure technological efficiency (PTE), while scale efficiency (SE) remained largely unaffected. Furthermore, the dynamic efficiency MI was mainly negatively impacted by changes in overall efficiency change (EFFCH), where pure efficiency change (PECH) less than one contributed significantly towards overall regression of port efficiencies during this period.

Originality/value

This paper is unique in its use of a combination of the grey prediction model and DEA efficiency assessment to quantify changes in important indicators during pandemic periods. This approach not only provides a quantitative understanding of the impact on port-level efficiency through numerical quantification but also offers readers an intuitive understanding.

Citation

Shang, F., Teng, X. and Park, M. (2023), "Quantifying the impact of COVID-19 on Chinese ports", Journal of International Logistics and Trade, Vol. 21 No. 4, pp. 255-272. https://doi.org/10.1108/JILT-03-2023-0020

1. Introduction

In late 2019, the COVID-19 pandemic outbreak led to a significant global economic downturn, causing a 3.3% decline in global Gross Domestic Product (GDP) in 2020. To curb the virus spread, countries implemented measures such as production shutdowns and travel restrictions, which halted economic activities and slowed down industries. The shipping industry, closely tied to foreign trade and financial conditions, faced particular vulnerability. Various sectors within the shipping economy, including production, transportation, storage and insurance, experienced varying degrees of impact. This resulted in a 3.8% decline in global seaborne trade to 11.5bn tons in 2020, comparable to the 4.0% decrease observed during the 2009 financial crisis. According to UNCTAD, maritime transport handles over 80% of the world's international cargo, making ports crucial hubs. China's coastal ports account for more than 90% of foreign trade imports and exports, particularly in energy and materials. However, due to reduced consumer demand, rising transportation costs, and a buildup of stagnant cargo, port throughput during the initial COVID-19 outbreak fell below expectations.

Figures 1 and 2 depict the throughput volume of major Chinese ports, showcasing notable fluctuations in cargo and container throughput in 2020. Particularly, February witnessed a significant decline in trade compared to the same period in 2019. This decrease can be attributed to the combined effects of the Chinese Lunar New Year and the COVID-19 pandemic, which led to the implementation of stringent policies by various countries to control port cargo transportation during this period. For a comparative view of cargo and container throughput growth rates between 2019 and 2020, please refer to Figures 3 and 4.

The Stringency Index, depicted in Figure 5, is a valuable tool provided by The Oxford COVID-19 Government Response Tracker (OxCGRT) for assessing the severity of pandemic control measures. This index comprises nine response metrics that gauge the level of intervention. As illustrated in Figure 5, China’s Stringency Index reached its peak in February 2020, indicating the implementation of rigorous control measures by the government during that time. Subsequently, the index gradually declined from March onwards. Consequently, the port throughput growth rate experienced a rebound in March. This observation suggests that the earlier decline in growth rate during the first two months of 2020 was likely attributed to the strict pandemic control measures enforced by the government.

The COVID-19 pandemic resulted in varying control policies across different regions of China, leading to delays in employees returning to work. Consequently, there was a shortage of labor supply, particularly at the upper end of the port logistics chain, which significantly impacted the handling capacity of ports. Moreover, nearly 100 countries implemented stringent preventive measures on Chinese ships or vessels calling at Chinese ports. These measures included mandatory quarantine periods and health certificate requirements upon entry into the port. Consequently, cargo transportation times were prolonged, port logistics costs escalated and route operation efficiency decreased, all of which contributed to a decline in container turnover rates. Furthermore, these measures increased the risk of chargebacks in international supply chains.

This article aims to assess the potential impact of the COVID-19 outbreak on Chinese ports. The evaluation is conducted on two levels. Firstly, it quantifies the potential changes in port trade volumes by analyzing cargo and container throughput as representative metrics. Secondly, it quantifies the potential changes in port efficiency through a system of port infrastructure indicators, taking into account the presence or absence of COVID-19. The study calculates the degree of change in efficiency values for each of the 17 Chinese ports along the Belt and Road by combining actual and predicted throughput values. The objective of this research is to provide valuable insights into the effects of the pandemic on the port industry and offer guidance for future port construction in the post-pandemic era.

The following sections comprise a literature review, an explanation of the quadratic residual-correction GM (1,1), SBM-data envelopment analysis (DEA) and Malmquist Index models used in this study. These models are then applied to evaluate 17 OBOR ports in China. Finally, the findings are summarized, and the limitations of this research are discussed.

2. Literature review

2.1 Impact of the COVID-19 on ports

With the outbreak and ongoing spread of COVID-19, quarantine policies implemented by various countries have had significant negative impacts on port activities, including the suspension of shipping routes, entry denials, and disruptions in transportation. Taking the Port of Barcelona as an example, during the lockdown measures implemented from March to June 2020, there was a significant decrease in maritime traffic. The number of port calls reduced by nearly 30% compared to the pre-lockdown period (Mujal-Colilles et al., 2022). The gradual shutdown of economic activities in ports, as they are essential for trade between countries, has prompted policymakers and scholars in different local governments to be concerned about the development of various indicators and changes in the overall efficiency of ports. Wang et al. (2022) identified key factors contributing to the decline in port efficiency during the pandemic, including a shortage of port personnel, resulting in longer loading and unloading times. Delays in customs clearance and slow delivery by shippers also played a role. Moreover, the disruption of road and rail networks caused congestion and cargo backlog in intermodal connections. To address the uncertainties, Russell et al. (2022) proposed a conceptual framework focusing on four dimensions: coastal interfaces, port platforms, inland interfaces and the overall system scope. This framework aims to enhance the flexibility of container port logistics and improve their resilience in response to such challenges.

Xu et al. (2021a) examined the relationship between the severity of the pandemic, urban economy and port activities using cargo throughput as a key indicator. They found that the pandemic had a greater negative impact on imports compared to exports. While government control measures had a limited effect on port throughput, they discovered a negative correlation between the implementation of restrictive policies and exports. On the other hand, an increase in import volume partially sustained port vitality (Xu et al., 2021b).

Zhou et al. (2022a) employed a system dynamics model to assess the potential economic losses in ports by constructing a maritime supply chain that includes subsystems such as shipping, ports, transportation, manufacturing and society. They found that in a two-year recovery period, the revenue generated by the shipping industry in Shanghai Port far exceeded that of the port services sector. Besides the port handling services, Zhou et al. (2022b) identified economic losses in terms of port debts and anchorage fees. Therefore, enhancing handling efficiency is crucial for the economic recovery of the port.

2.2 Grey prediction model GM (1,1)

Ceylan (2021) devised a dual hybrid model that combined grey prediction with a rolling mechanism and particle swarm optimization (PSO) to forecast short-term confirmed COVID-19 cases in Germany, Turkey and the USA. In order to assess the accuracy of the results, a non-linear autoregressive neural network was also developed. It was determined that the optimized parameters of the GM (1,1) grey modeling using the particle swarm optimization algorithm yielded more robust and effective predictions.

Saxena (2021) made modifications to the GM (1,1) model and utilized it to forecast the spread of the coronavirus within India. The model incorporated data from overlapping periods of infected cases to predict the transmission of the virus in different regions of the country. This innovative approach demonstrated superior prediction accuracy compared to IOGMs, GM (1,1) and NGM(1,1,k) models. Additionally, the study found that increasing the duration of the overlap period led to a reduction in prediction errors.

Li et al. (2023) discussed the challenges posed by the sudden outbreak of COVID-19 on global public health and textile export trade. The article proposes a novel forecasting model, BODGM (1,1), to predict the impact of pandemic-induced uncertainty on the volatility of cotton exports in China. The model outperforms other models in terms of accuracy, especially with limited data.

2.3 Port efficiency analysis

DEA is a non-parametric method used to measure the relative efficiency of multiple decision-making units (DMUs) by comparing their input and output variables. In the context of port operations, the DEA model helps evaluate how effectively ports utilize their resources to generate desired and undesired outputs.

Krmac and Mansouri Kaleibar (2022) conducted a comprehensive review of the literature on the application of the DEA model in port efficiency assessment over the past three decades, up until October 2021. The earliest studies on port efficiency evaluation date back to 1993 (Roll and Hayuth, 1993). With the wide-ranging adoption of the DEA model, it has undergone continuous improvement and upgrading. Notable advancements include DEA models for measuring the economic efficiency of ports (Itoh, 2002; Cullinane and Wang, 2007; Al-Eraqi et al., 2010; Bray et al., 2014; Morales-Fusco et al., 2016; Zarbi et al., 2019; da Costa et al., 2021). Environmental DEA models have been considered port environmental factors (Chin and Low, 2010; Chang, 2013; Na et al., 2017; Castellano et al., 2020; Quintano et al., 2021). Moreover, integrated objective efficiency DEA models have been introduced to assess both port and maritime transportation objectives (Sharma and Yu, 2009; Lozano et al., 2011; Yuen et al., 2013; Figueiredo De Oliveira and Cariou, 2015; Schøyen et al., 2018; Nguyen et al., 2020; Kong and Liu, 2021). These extended methods provide researchers with diverse measurement perspectives, enabling more precise evaluations of port efficiency and performance across various research objectives, while also offering valuable decision support.

Based on the literature review, it is apparent that hybrid research methods combining predictive modeling and efficiency evaluation are uncommon. However, Wang et al. (2020) conducted a two-year assessment of future productivity and performance in e-commerce companies by integrating the Grey model with the Malmquist-I-C DEA model. Notably, there is a lack of studies that apply this combined approach to evaluate future efficiency in the context of ports. This observation has motivated me to pursue research that hybrid method for assessing port efficiency. Furthermore, this paper aims to provide a comprehensive analysis of the strengths and weaknesses of port development in the context of the pandemic environment. It intends to identify the development trends of different variables in ports, compared to normal conditions, and propose timely improvements. Ultimately, the objective is to minimize the adverse effects of uncertainty caused by the pandemic on port operations.

3. Methodology

3.1 Residual-correction GM (1,1)

This section provides an overview of the modeling mechanism and calculation process of the GM(1,1) method. In addition, it describes how the residuals between the actual and predicted values are utilized to construct the residual prediction series. Finally, the initial GM (1,1) results are refined by incorporating the residual prediction results to generate the final prediction outcomes.

3.1.1 Grey prediction model GM (1,1)

Grey system theory offers several advantages, such as the ability to extract valid information from data with a small base and poor information. Additionally, it can accurately describe and predict the characteristics and properties of uncertain systems, making it useful in prediction, decision-making, control analysis, evaluation and optimization analysis. Grey prediction models, which are based on grey system theory, inherit these advantages and can adapt to a wide range of models and small data sample sizes. In contrast to statistical analysis models, grey prediction models are not limited to sample data that must follow a specific probability distribution. Furthermore, unlike time series models, grey prediction models do not require various correlation tests and analyses of the original modeled data.

GM (1,1) is a widely applied form of the grey prediction model, which is based on a differential equation of order one and has only one variable. The modeling process for GM (1,1) is as follows:

Collect non-negative raw data:

(1)X(0)=(x(0)(1),x(0)(2),…,x(0)(n))

Perform a first-order summation of the original data to generate a monotonically increasing series:

(2)X(1)=(x(1)(1),x(1)(2)…,x(1)(n))

which

(3)x(1)(k)=∑i=1kx(0)(1),k=1,2,…,n

Let x⁽¹⁾ be the 1-AGO (first-order accumulated generating operation) sequence of x⁽⁰⁾. x⁽¹⁾ can weaken the volatility of the original data sequence, resulting in a smoother accumulated data series.

Then, z(1) is obtained by calculating the adjacent averages of the first-order accumulative sequence x(1):

(4)z(1)=(z(1)(1),z(1)(2),…,z(1)(n))

which

(5)z(1)(k)=0.5x(1)(k)+0.5x(1)(k−1),k=2,…,n

Next, we establish the GM (1,1) first-order differential linear equation, which has the basic form:

(6)x(0)(k)+az(1)(k)=b,k=2,…,n

After the whitening, the form becomes:

(7)dx(1)(t)dt+ax(1)(t)=b

where a is the development coefficient, and b is the amount of grey effect. This equation can be solved using the least-squares method to obtain the estimated values of a and b, denoted as

(8)a^=[ab]=(B⊺B)−1B⊺y

where B is the data matrix, and Y is the parameter vector.

(9)B=[−z(1)(2) 1−z(1)(3) 1⋮−z(1)(n) 1],Y=[x(0)(2)x(0)(3)⋮x(0)(n)]

The time response series corresponding to the whitening differential equation can be expressed as follows:

(10)x^(1)(k+1)=(x^(1)(1)−ba)e−ak+ba,k=2,3,…,n

The final step is to perform cumulative reduction to obtain the predicted results:

(11)x^(0)(k+1)=x^(1)(k+1)−x^(1)(k),k=2,3,…,n

After obtaining the predicted results, the accuracy of the model is evaluated using the residual test method. This is based on the residual series of the GM (1,1):

(12)ε(k)=x(0)(k)−x^(0)(k)

The relative error can be calculated as:

(13)εk=|ε(k)|x(0)(k)

The average relative error of the series is calculated as:

(14)ε(a)=1n−1∑k=2n|ε(k)|

Modeling accuracy can be evaluated using various metrics:

(15)p=(1−ε(a))×100%

(16)Emae=1n∑i=1n|x−xi^|

(17)Emape=1n∑i=1n|x−xi^|xi^×100%

(18)Emse=1n∑i=1n(x−xi^)2

3.1.2 Quadratic residual-correction GM (1,1)

To further improve the fitting accuracy, the initial prediction results of the GM (1,1) model can be corrected using the residual forecasted series. The specific process is as follows:

Set the sequence of residuals of X⁽¹⁾ as:

(19)ε(0)=(ε(0)(1),ε(0)(2),…,ε(0)(n))

(20)ε(0)(k)=x(1)(k)−x^(1)(k),k=1,2,…,n

If there exists k0 such that

(21){∀k>k0,ε(0)(k)>0n−k0≥4 or {∀k>k0,ε(0)(k)≤0n−k0≥4

Then |ε(0)(k0)|,|ε(0)(k0+1)|,…,|ε(0)(n−m)| is deemed to be the model able residual tail, expressed as:

(22)ε(0)=(ε(0)(k0),ε(0)(k0+1),…,ε(0)(n))

The ε(0) of GM(1,1) is established, and the corresponding restored time response equation is obtained as follows:

(23)ε^(0)(k+1)=(−aε)(ε(0)(k0)−bεaε)e−aε(k−k0),k>k0

Eventually, the x^(1) is corrected using with ε^(0). The corrected time response equation is:

(24)x^(1)(k+1){(x(0)(1)−ba)e−ak+ba ,k<k0(x(0)(1)−ba)e−ak+ba± aε(ε(0)(k0)−bεaε)−aε(k−k0),k≥k0

In the residual-corrected GM(1,1) model, the sign of the residual correction value, denoted by ε^(0), should be consistent with the sign of the ε(0) residual tails.

(25)ε^(0)(k+1)=(−aε)(ε(0)(k0)−bεaε)e−aε(k−k0),k>k0

Finally, the accuracy of the predicted results with quadratic residual correction is evaluated again through testing.

3.2 Port efficiency: SBM-DEA model

DEA is a method used for multi-input and output analysis. It is a systematic approach for evaluating the relative efficiency of inputs and outputs. The traditional DEA models, such as CCR and BBC, are radial models that do not take into account the “slack” part. To address this, Tone (2001) proposed the Slack-Based Measure (SBM) model, which includes the slack variables directly in the objective function.

Suppose there are n DMUs, each with two vectors: input vector and output vector, represented as X=(xij)∈Rm×n,Y=(yij)∈Rs×n, where X > 0, Y > 0. The production possibility set (PPS) can be defined as P={(x,y)| x≥Xλ,y≤Yλ,λ≥0}, where λ is the non-negative vector.

(26)ρ*=minimize1−1m∑i=1m si−xik1+1q∑r=1q sr+yrk

S.T.

(27)Xλ+s−=xk

(28)Yλ−s+=yk

(29)λ,s−,s+≥0

Let s− and s+ denote the slack values of inputs and outputs, and 0 ≤ ρ* ≤1. ρ* <1 means the DMU is inefficient. This model is a non-linear programming model, but it can be transformed into a linear programming model as:

(30)min⁡ρ*=t−1m∑i=1m Si−xik

S.T.

(31)XΛ+S−−txk=0

(32)YΛ−S+−tyk=0

(33)t+1q∑r=1q sr+yrk=1

(34)λ,S−,S+≥0

3.3 Malmquist Index

The Malmquist Index (MI) is a useful tool for analyzing changes in the production technology and scale of each decision unit over time. In traditional DEA models, decision units are placed on a uniform production frontier for efficiency assessment and comparison, making it difficult to analyze the contribution of changes in production technology and scale to changes in efficiency. The MI method, introduced by Färe et al., used the ratio of distance functions to measure the total factor productivity change (TFPCH) from period t to period t + 1. This change can be decomposed into an index of technical efficiency change (EFFCH) and technical progress change (TECH). Furthermore, the EFFCH can be further decomposed into scale efficiency change (SECH) and pure technological progress change (TECH). The MI can be defined as:

MI=EFFCH×TECH=SECH×PTECH×TECH

(35)MI(xt+1,yt+1,xt,yt)=Dt(xt,yt)Dt+1(xt+1,yt+1)×Dt+1(xt+1,yt+1)/vrsDt(xt,yt)/vrs×[Dt(xt,yt)Dt+1(xt,yt)×Dt(xt+1,yt+1)Dt+1(xt,yt)]12

An MI value greater than 1 indicates an increasing trend in total factor productivity (TFP), while a value less than 1 indicates a decreasing trend in TFP. A value equal to 1 indicates that TFP remains constant over the two time periods.

4. Application: the case of Chinese 17 OBOR ports

4.1 Data and variable

This paper focuses on studying 17 Chinese ports located along the Belt and Road Initiative (BRI). The research collects data on five crucial indicators to assess these ports. The indicators include the number of production berths, berths over 10,000 tons, berth length, cargo throughput and container throughput. The Ministry of Transport of China and the China Port Statistical Yearbook serve as the sources of data for this study. In order to provide a visual representation of the ports' geographical locations, Figure 6 presents a map, while Table 1 displays statistical results for each indicator.

4.2 Residual-correction GM (1,1) prediction results

Each indicator was initially forecasted using the GM(1,1) model, and the predictions were divided into two scenarios. In scenario 1, the development of each indicator was projected from 2018 to 2025, with the modeling years spanning from 2010 to 2017. This scenario assumed no occurrence of COVID-19. To validate the accuracy of the prediction results, data from 2011 to 2019 were used, and this forecast is referred to as the “no-COVID forecast result”.

In scenario 2, the GM (1,1) model was applied considering the presence of the pandemic. Since the model requires only 4 to 8 data points to construct the original dataset for future prediction, the data from 2016 to 2020 were selected to complement the prediction until 2025. The actual data from 2017 to 2020 were utilized for accuracy testing in this scenario.

To enhance the GM (1,1) prediction results, quadratic residual correction was employed. Table 2 provides the fit accuracy outcomes for both the GM (1,1) and the residual-improved GM (1,1) models.

Due to the extensive nature of the results, we will focus on presenting the forecasted results for cargo and container indicators. Figures 7 and 8 show that the quadratic residual-corrected grey model provides a better fit than the traditional GM(1,1) model. The mean absolute percentage error (MAPE) for the conventional GM(1,1) model in predicting accuracy without the COVID-19 background has a mean value of 4.05%, whereas the residual-corrected MAPE is 2.88%. Moreover, in the context of COVID-19, the MAPE improved by 1.06%, from 2.88% to 1.93%.

Figures 9 and 10 depict the throughput volume at the seventeen OBOR ports from 2020 to 2025, illustrating both scenarios with and without COVID-19. In both cases, there is an overall upward trend in throughput volume over the years, but the growth rates differ.

In the scenario without COVID-19, the container throughput exhibits a five-year growth rate of 58.51mn TEU, with an average annual growth rate of approximately 5.7%. However, in the scenario considering the impact of the pandemic, the increase in container throughput is approximately 32.37mn TEU, indicating a decrease in growth rate by approximately 2.1% compared to the situation without COVID-19.

Furthermore, in the absence of COVID-19, the average annual growth in cargo throughput declined from 5.2% to 3.5%. The occurrence of the pandemic had a direct impact on trade throughput due to the implementation of COVID-19 control measures during quarantine periods. These measures resulted in significant declines in both global and domestic trade volumes during the initial stages of lockdowns. Upstream activities, including consumption impulses and productivity levels, also experienced substantial contraction, affecting dock operations and leading to lower trends compared to previous years, even within similar time frames.

A cross-sectional comparison of throughput growth rates from 2021 to 2025 indicates that the trend of throughput growth in the presence of COVID-19 is significantly weaker than the normal trend. This suggests that the impact of the pandemic will become more pronounced over time. The difference in throughput between the pre-pandemic and post-pandemic periods is projected to increase annually as the control measures and quarantine policies continue to impact port operations.

In this study, the initial year of the outbreak (2020) was used to forecast the indicators for the subsequent years. This is because 2020 represented the peak impact of the pandemic on China, with the highest level of control and management measures and the most intense quarantine policies in place. The predictions for the following years were based on a similar level of pandemic control as that of 2020. As a result, the degree of potential throughput loss will inevitably be linked to the high level of control measures implemented during the pandemic.

In theory, the difference between the throughput differential and the growth rate in the context of COVID-19 is expected to exhibit an upward and then downward trend, following a typical development pattern. However, due to the highly contagious nature, significant impact and prolonged incubation period of the coronavirus, the Chinese government has implemented stringent control and quarantine measures in heavily affected areas. This has led to a long-term COVID-control plan with the objective of achieving zero infections within China.

The implementation of this long-term prevention and control program has hindered the development of the economic base, thereby limiting the progress in port construction, trade activities and other related operations. Consequently, it will be challenging to restore most of China's ports to their normal conditions in the short term, given the ongoing efforts to prevent new infections and achieve complete control over the virus.

4.3 SBM-DEA results

Figure 11 presents the efficiency scores of 17 Chinese OBOR ports from 2020 to 2025.

Figure 11 shows that without COVID-19, the average values of total efficiency (TE) and partial technical efficiency (PTE) were consistently higher. However, in the presence of the pandemic, TE decreased, reaching a minimum value in 2023 before rebounding in 2024. The largest difference between the two scenarios was observed in 2023.

Without COVID-19, the trend of PTE was opposite to that of TE, gradually decreasing from 0.794 to 0.761. However, with the presence of the pandemic, PTE showed a slight rebound after reaching a minimum value in 2022, and then remained relatively constant.

Scale efficiency (SE) increased significantly by 12.2% over five years in the absence of the pandemic, reaching a value of 0.849. However, under the influence of the pandemic, SE exhibited a volatile trend, fluctuating around the SE in normal conditions, with a peak efficiency of 0.847 in 2024.

Therefore, COVID-19 not only reduced most efficiency values but also fundamentally altered the efficiency profile, with the most significant changes observed in TE and PTE.

4.4 Malmquist productivity index results

In this paper, we will analyze dynamic efficiency in three different timelines.

Firstly, we will examine the normal development of port dynamic efficiency without the assumption of COVID-19, as illustrated in Figure 12:

In the absence of COVID-19, the MI values show an overall increase, indicating an annual improvement in port efficiency. However, in the presence of COVID-19, the MI values decline, primarily due to a decrease in PECH. Figure 13 depicts this scenario.

Figure 13 shows that the COVID-19 outbreak directly affects the average efficiency of ports, leading to a decline in the MI as the pandemic continues. This decline is primarily attributed to a decrease in EFFCH, mainly driven by a decline in PECH from 0.954 to 0.848. In contrast, SECH initially declines but remains above 1. Moreover, TECH exhibits a consistent upward trend above 1, indicating that the pandemic has not hindered technological advancements in the port industry.

In the third timeline, the dynamic efficiency with and without the pandemic was compared for the same year t. This comparison provides a clearer indication of the impact of COVID-19 on the port's efficiency and the potential loss of DMU for each efficiency value. The results are presented in Figure 14.

The changes in efficiency were more significant for the same year compared to the inter-period results. In all cases, MI was less than 1, indicating that the coronavirus had a severe impact on the overall efficiency of ports. Although TECH gradually rebounded, the degradation effect of EFFCH was more substantial, leading to a progressive decrease in MI. The most negatively affected efficiency value by the COVID-19 outbreak was PECH, directly related to the slowdown in production activities under control measures such as quarantine. However, SECH showed a significant increase, which was the opposite of PECH.

5. Conclusion

The study employed a modified GM (1,1) model to predict five key indicators for 17 major Chinese OBOR ports under both pandemic and non-pandemic conditions. The results indicated that the COVID-19 had a negative impact on the volume and growth rate of port throughput. Without the pandemic, cargo and container throughput were projected to grow at rates of 5.2 and 5.7% respectively from 2020 to 2025. However, due to the Covid, these growth rates significantly decreased to only 3.5 and 3.6%, respectively. Additionally, a quantitative analysis of port efficiency using SBM-DEA and MI revealed that the pandemic shock resulted in a significant decrease in TE and PTE compared to the scenario without the pandemic.

Nevertheless, the study found that the SE remained consistent throughout the entire study period. As for dynamic efficiency, when the no-Covid norm was applied for MI, the results were greater than 1, indicating a gradual improvement in integrated efficiency over time. However, when considering the impact of Covid, MI exceeded one only during the initial and final phases. When comparing indices with and without Covid in the same year, it was observed that MI was less than one primarily due to the inefficiency in EFFCH.

It is important to acknowledge that the duration of the pandemic's impact on the study was relatively short. Consequently, certain indicators affected by the Covid may not fully reflect the current situation. The most substantial impact was observed in 2020. As we transition into a post-pandemic period, it is anticipated that Covid control measures in China will become more targeted and region-specific. Therefore, when evaluating each indicator in this paper, it is vital to consider the intensity and timing of the Covid control measures implemented in the cities where the ports are situated.

Figures

Figure 1

Total cargo throughput of Chinese major ports (unit: 10,000 Ton)

Figure 2

Total container throughput of major Chinese ports (unit: 10,000 TEU)

Figure 3

Cargo throughput growth rate of Chinese major ports

Figure 4

Container throughput growth rate of Chinese major ports

Figure 5

2020 Chinese Stringency Index for COVID-19 control

Figure 6

Map of Chinese 17 OBOR ports

Figure 7

Predicted results for container throughput under no-COVID impact (10,000 TEU)

Figure 8

Predicted results for cargo throughput with COVID-19 (10,000 Ton)

Figure 9

Predicted cargo throughput from 2021 to 2025 with and without COVID-19

Figure 10

Predicted container throughput for 2021–2025 with and without COVID-19

Figure 11

Comparison of efficiency results with and without COVID-19

Figure 12

MI and decomposition index results under the no-Covid assumption

Figure 13

MI and decomposition index results for no-Covid year t and with-Covid year t+1

Figure 14

MI and decomposition index results in the same year with and without COVID-19

Table 1

Statistics of the data

Input	X1	Number of berths for production (unit)
	X2	Number of berths for production of 10,000 tons and above(unit)
	X3	Length of berths for production (m)
Output	Y1	Cargo throughput (10,000Ton)
	Y2	Container throughput (10,000TEU)

Source(s): Authors’ own work

Table 2

Accuracy results for GM (1,1) and quadratic residual-correction GM (1,1)

	MSE		MAD		APE		MAPE
Without COVID-19	GM(1,1)	Residual-corrected GM(1,1)	GM(1,1)	Residual-corrected GM(1,1)	GM(1,1)	Residual-corrected GM(1,1)	GM(1,1)	Residual-corrected GM(1,1)
Production berth number	50	38	5	4	493.65%	395.40%	3.63%	2.91%
10,000 tons berth number	10	9	2	2	518.26%	434.27%	3.80%	3.19%
Berth length	1,265,657	1,171,324	727	632	482.00%	409.96%	3.54%	3.01%
Cargo throughput	1,582,688	1,332,596	919	757	501.03%	390.74%	3.68%	2.87%
Container throughput	1,260	914	21	17	760.42%	328.00%	5.59%	2.41%
With COVID-19	GM(1,1)	Residual-corrected GM(1,1)	GM(1,1)	Residual-corrected GM(1,1)	GM(1,1)	Residual-corrected GM(1,1)	GM(1,1)	Residual-corrected GM(1,1)
Production berth number	56	38	4	3	224.40%	153.51%	3.30%	2.26%
10,000 tons berth number	4	1	1	1	167.46%	98.02%	2.46%	1.44%
Berth length	6,12,497	4,29,938	526	336	176.64%	122.15%	2.60%	1.80%
Cargo throughput	1,465,365	7,35,915	687	430	175.53%	120.11%	2.58%	1.77%
Container throughput	1,762	1,364	25	17	235.76%	163.44%	3.47%	2.40%

Source(s): Authors’ own work