The Efficiency of Document and Border Procedures for International Trade

Abstract

In many countries, document and border procedures create trading barriers, thereby impairing economic growth. These can range from insufficient transshipment facilities to unsupportive institutional arrangements. To address this, countries have taken reforms to improve their procedures by introducing electronic documentation systems, strengthening border infrastructure, and enhancing customs procedures. However, the efficiency of the document and border procedures in each country remains unclear, as well as how new reforms can affect these. This study investigated the efficiency of document and border procedures in each country, defined as the trade volume and value per required cost, time, and documents in the trading procedures. The efficiencies were calculated through a data envelopment analysis with cross-sectional data from 2019 and a window analysis with panel data from 2014 to 2019. The study found a positive change in export procedure efficiency after all three types of reforms were instituted in a country, but a positive change in import efficiency only after the introduction of electronic documentation. All countries were classified according to their document and border procedure strengths and weaknesses.

1. Introduction

With the recent economic globalization, international trade is playing a more important role than ever before. Typical international trade requires many steps to transport cargo from and to a trading partner country. First, the cargo is transported to the port or border in the country of export; second, it is cleared for export; third, it is shipped by sea or air; fourth, it is cleared for import, and fifth, it is transported from the port or border to the buyer or buyer’s customer [1]. At every port and border, international cargo operations must adhere to certain customs and trade procedures. In addition, there may be further procedures and controls along the way when in transit. In many countries, inefficient processes, unnecessary bureaucracy, and redundant procedures add to the time and cost of document and border procedures [2].

There are several studies investigating the influence of cargo transport procedures on trade. Martincus et al. found that there was a significant negative effect from customs-related delays on firms’ exports, such as a reduced number of shipments and buyers and reduced exports per buyer, in terms of both volume and value [3]. Djankov et al. found that each additional day a cargo was delayed from being shipped reduced trade by more than 1%, using gravity equation modeling with data from 98 countries on the timing of moving containerized products between the factory gate and ship [4]. Freund and Rocha found that a one-day increase in inland transit time reduced exports by 7% on average, and a one-day reduction in inland travel time translated into nearly a 1.5 percentage point decrease in all importing-country tariffs [5]. Furthermore, Portugal-Perez and Wilson found that improvements in border and transport efficiency, even halfway to the regional top performer’s efficiency level, can be substantial, through an analysis of their border and transport efficiency indicators, which include the number of days and documents to export and import [6]. Wang et al. also showed in their analysis exploring the relationship between green logistics and international trade with an augmented gravity model that the logistical performance of exporting and importing countries is positively correlated with trade volume [7].

It is not surprising, then, that several reforms have been put in place to improve the efficiency of border and document procedures. Some countries have enhanced customs administration and inspections. One of the reforms to enhance these procedures is the introduction of a Single Window. It is a single point of contact between foreign trade operators and the government to fulfill the requirements of import, export, and customs transit [8]. The Democratic Republic of the Congo, for instance, has reduced the time needed to export and import by implementing a national trade Single Window during 2018–2019 [9]. Porto et al. found from their analysis using a gravity model that the existence of a Single Window program improves countries’ trade performance [8]. In addition, the electronic systems for filing, processing, and exchanging trade information have been either introduced or improved in many countries. For example, India enhanced its electronic submission of documents by integrating several government agencies into an online system called e-Sanchit between 2018 and 2019 [9]. Implementing the latest technologies that limited the number of interactions between firms and border agencies was associated with both an increase in the number of firms and the volume of exports by user firms [10]. Other countries have implemented reforms that impact border infrastructure. For example, Saudi Arabia has made importing easier by upgrading its port infrastructure during 2018–2019 [9]. Inadequate infrastructure hinders smooth international trade as much as burdensome border procedures and poor logistical services [11].

However, how efficient the document and border procedures in each country are compared to that of other countries remains unclear. Besides, the impact of new reforms on its efficiency has not been clarified. Thus, this study evaluates the efficiency of the international trade document and border procedures of each country via data envelopment analysis (DEA). We also investigate the change in the efficiency of document and border procedures after conducting reforms. Our findings contribute to the literature by providing implications on how countries can make trading procedures more efficient.

The remainder of this paper is organized as follows. The methodology and data are explained in Section 2 and Section 3, respectively. The time-series change in the efficiency of document and border procedures, and the change after conducting reforms using the result of DEA window analysis with panel data, are analyzed in Section 4. This efficiency is further investigated, focusing on each country in 2019, in Section 5, and countries are classified according to their strengths and weaknesses in import document and border procedures with cross-sectional data. Benchmarks are also suggested for each country. Section 6 concludes our study and summarizes its limitations.

2. Methodology

2.1. Data Envelopment Analysis

In the discussion of efficiency measurement, Farrell was the first to provide definitions and a computational framework for efficiency [12]. Over the past several decades following his research, efficiency measurement has developed rapidly [13]. The DEA is one of the most important and useful approaches in this respect and has often been used to evaluate transport and transshipment efficiency. For example, Cullinane et al. applied not only the DEA but also the Stochastic Frontier Analysis to measure the efficiency of the container ports ranked in the global top 30 in 2001 [13]. Fanou and Wang evaluated the efficiency of the main transit corridors linking landlocked countries and seaports in Africa from 2008 to 2013 [14]. Xu and Ishiguro applied the DEA to cross-sectional data to evaluate the efficiency of Chinese and South Korean container terminals, including four automated and twenty traditional terminals [15]. However, to the best of our knowledge, no study has looked specifically at the efficiency of document and border procedures.

The approach includes two types of models: one assumes constant returns to scale (CRS), developed by Charnes et al., and the other assumes variable returns to scale (VRS), developed by Banker et al. [16,17]. Both models do not require a production function in advance to specify how to handle the observed data.

2.2. CRS Model

The CRS model assumes constant returns to scale of the production function, as its acronym suggests. Consider a set of $n$ decision-making units (DMUs), with the efficiency of DMU $k$, Charnes et al. measured this with the following equations, written as a linear program [16]:

$$\underset{u,v}{\max }{\theta }_k={\displaystyle \sum }_{r=1}^s{u}_r{y}_{rk}$$

(1)

s.t.

$${\displaystyle \sum }_{i=1}^m{v}_i{x}_{ik}=1$$

(2)

$$-{\displaystyle \sum }_{i=1}^m{v}_i{x}_{ij}+{\displaystyle \sum }_{r=1}^s{u}_r{y}_{rj} \le 0 \left(j=1, 2, \cdots , n\right)$$

(3)

$$v_i\ge 0 \left(i=1, 2,\dots ,m\right)$$

(4)

$$u_r\ge 0 \left(r=1, 2,\dots ,s\right)$$

(5)

where ${\theta }_k$ is the technical efficiency of the $k$th DMU among $n$ DMUs,$y_{rk}$ is the level of the $r$th output of the $k$th DMU,$x_{ik}$ is the level of the $i$th input of the $k$th DMU,$v_i$ is the weight given to input $i$ of the $k$th DMU,$u_r$ is the weight given to output $r$ of the $k$th DMU,$m$ is the number of inputs, and $s$ is the number of outputs.

In Equation (2), $v_i{x}_{ik}$ is the virtual input value of the $i$th input variable. It represents the ratio of the input to total inputs. As a DMU assigns a greater weight to inputs that are relatively better than other DMUs, inputs that are performing well relative to other DMUs will also have higher virtual input values. Therefore, by looking at the virtual input values, we can extract the factors that are superior to other DMUs; in other words, their strengths.

The linear program in Equations (1)–(5) can be replaced by the dual problem. Here, $v_i$ and $u_r$ are also replaced by ${\lambda }_j$.

$$\underset{\theta ,\lambda }{\min } {\theta }_k$$

(6)

s.t.

$${\theta }_k{x}_{ik}-{\displaystyle \sum }_{j=1}^n{\lambda }_j{x}_{ij}\ge 0 \left(i=1, 2, \cdots , m\right)$$

(7)

$$-y_{rk}+{\displaystyle \sum }_{j=1}^n{\lambda }_j{y}_{rk}\ge 0 \left(r=1, 2, \cdots , s\right)$$

(8)

$${\lambda }_j\ge 0 \left(j=1, 2, \cdots , n\right)$$

(9)

The optimal $\theta$, denoted by ${\theta }^{*}$, satisfies $0<{\theta }^{*}\le 1$. $\theta$ = 1 indicates that the DMU is efficient, and lies on the efficiency frontier consisting of the set of efficient DMUs.

2.3. VRS Model

To account for variable returns to scale, Banker et al. extended the CRS model to obtain the VRS model [17]. The VRS model is defined by adding the following equation to the CRS equations.

$${\displaystyle \sum }_{i=1}^n{\lambda }_i=1$$

(10)

Compared with CRS, imposing the additional constraint causes the feasible region of VRS to become a subset of that of CRS, which means ${\theta }^{*}$ calculated with VRS is not less than ${\theta }^{*}$ calculated with CRS. The VRS model measures how a DMU utilizes the resources under exogenous environments. Thus, low efficiency in the VRS model implies that the DMU is inefficiently managing its resources.

2.4. Scale Efficiency

With both the efficiency values of the CRS and VRS models, the scale efficiency,$S{E}_k$, of the kth DMU is where the relationship between them is as follows:

$$S{E}_k=\frac{{\theta }_{CRS, k}^{*}}{{\theta }_{VRS,k}^{*}}$$

(11)

where ${\theta }_{CRS, k}^{*}$ is the technical efficiency of the $k$th DMU calculated in the CRS model, and ${\theta }_{VRS, k}^{*}$ is the technical efficiency of the $k$th DMU calculated in the VRS model. Scale efficiency indicates how efficiency is affected by the size of the scale. For scale efficiency, $S{E}_k$ is equal to one, whereas for scale inefficiency, which is caused by increasing or decreasing returns to scale, $S{E}_k$ is less than one. Scale inefficiency is caused by variable returns to scale. When the sum of the weights, ${\lambda }_i$ in Equations (2)–(4), is bigger than, equal to, and smaller than one, respectively, the law of increasing, constant, and decreasing returns to scale prevails, respectively.

2.5. Input- or Output-Oriented

DEA models can be distinguished according to whether they are input- or output-oriented. Input-oriented efficiency measures prevent any change in outputs, while the proportional reduction in inputs is considered, whereas output-oriented efficiency measures maintain constant inputs, while the proportion of the potential increase in output is investigated [13]. In this study, we use an input-oriented model to measure efficiency because attempts have been made to improve document and border procedures through reforms to reduce inputs such as cost, time, and the number of documents.

2.6. Window Analysis

The DEA window analysis is adapted to detect a trend in DMU over time. The trend can be detected by considering each DMU in a window as a completely different DMU and generalizing the notion of moving averages. In previous logistics studies, this methodology is often used to evaluate port efficiency. Cullinane et al. evaluated the efficiency of the world’s major container seaports from 1992 to 1999 by applying the DEA window analysis [18]. Pjevčević et al. also used it to determine the efficiency of ports and observe the possibility of changes in port efficiency in the years 2001–2008 [19].

Following Asmild et al., here, we considered $N$ DMUs $\left(n=1,\dots ,N\right)$ [20]. Each DMU had $r$ inputs and $s$ outputs during the observed periods $T \left(t=1,\dots ,T\right)$. An observation n in period t, $DM{U}_n^t$ had input vector $x_n^t=\left[\begin{array}{c}{x}_{1, n}^t\\ ⋮\\ x_{r, n}^t\end{array}\right]$ and an output vector $y_n^t=\left[\begin{array}{c}{y}_{1, n}^t\\ ⋮\\ y_{s, n}^t\end{array}\right]$. If the window with width $w, 1\le w\le T-k$, started at time $k, 1\le k\le T$, the matrices of inputs and outputs for the window analysis were given by:

$$X_{kw}^{}=\left[\begin{array}{cccc}{x}_1^k& x_2^k& \cdots & x_N^k\\ x_1^{k+1}& x_2^{k+1}& \cdots & x_N^{k+1}\\ ⋮& ⋮& \ddots & ⋮\\ x_1^{k+w}& x_2^{k+w}& \cdots & x_N^{k+w}\end{array}\right], Y_{kw}^{}=\left[\begin{array}{cccc}{y}_1^k& y_2^k& \cdots & y_N^k\\ y_1^{k+1}& y_2^{k+1}& \cdots & y_N^{k+1}\\ ⋮& ⋮& \ddots & ⋮\\ y_1^{k+w}& y_2^{k+w}& \cdots & y_N^{k+w}\end{array}\right]$$

(12)

Substituting inputs and outputs of $DM{U}_n^t$ into the VRS model produced the results of the DEA window analysis here. Then, the averaged efficiencies, averaged in each DMU, namely, each country, and each window were used to investigate the change in efficiency over the periods.

In this study, we conducted two types of analysis to estimate the efficiency of the document and border procedures for imports and exports. These were a panel data and window analysis over the 2014–2019 period, and a cross-sectional analysis in 2019.

3. Data

3.1. Definition of Trading Partner and Mode of Transport

In this analysis, we focused on the international trade between trading partners, as defined by the Doing Business Economy Profile of each country in 2020 [21]. The trading partner for importing cargo was defined as the country that imported the largest value (price times quantity) of auto parts (Harmonized System code, HS 8708). The trading partner for exporting cargo was defined as the country that was the largest purchaser of the product of the country’s comparative advantage, namely, its largest export value. The trading partners and export products for each economy are defined by collecting data on trade flows for the most recent four-year period from international databases, such as the United Nations Commodity Trade Statistics Database [2]. The mode of transport, land or maritime transport, was the one most widely used for the chosen export or import product and the trading partner, such as seaport or land border crossing [2].

3.2. Output Variables

Our output variables for both analyses, namely, our panel data analysis from 2014 to 2019 and the cross-sectional analysis in 2019, were $y_{mt}^t$, cargo volume for all commodities (MT) and $y_{nv}^t$, cargo value for all commodities (USD) transported from and to each country’s trading partner by major mode used for the chosen export or import product in year t. These annual data were collected from Global Trade Atlas Suite (GTAS) Forecasting provided by IHS Markit [22]. The major mode of transport cargo from and to each country’s trading partners was identified by comparing the volume of seaborne and overland cargo, based on trading with each country’s partner in 2019, following the definition of Doing Business reports [21].

3.3. Input Variables

We used different input variables for the panel data analysis from 2014 to 2019 and the cross-sectional analysis in 2019. However, the data sets of both analyses indicate the time and cost associated with each document and border procedure, respectively, and the number of documents. Many studies, including those by Jiang et al., Wiegmans and Janic, Fanou and Wang, Panagakos and Psaraftis, Moon et al., Kawasaki et al., and Regmi and Hanaoka, use cost and time as the indicators to reflect international cargo transport characteristics [14,23,24,25,26,27,28]. In addition, Martincus et al., Djankov et al., and Freund and Rocha have indicated the negative impact of a time delay on the trade flow [3,4,5]. The numbers of documents to export and import were also used to make aggregate trade facilitation indicators of border and transport efficiency by Portugal-Perez and Wilson [6].

The panel data analysis with the DEA window method used five variables as inputs for calculating the document and border procedure efficiency as shown in

Table 1. Output and input variables.

Category	Panel Data Analysis 2014–2019		Cross-Sectional Analysis 2019		Unit
Output	$y_{mt}^t$	Trade volume	$y_{mt}^{2019}$	Trade volume	[MT]
	$y_{nv}^t$	Trade value	$y_{nv}^{2019}$	Trade value	[USD]
Input	$x_{tb}^t$	Border procedures	$x_{tb1}^{2019}$	Clearance and inspections	Time [hours]
			$x_{tb2}^{2019}$	Port/border handling
	$x_{td}^t$	Documentary procedures	$x_{td}^{2019}$	Documentary procedures
	$x_{cb}^t$	Border procedures	$x_{cb1}^{2019}$	Clearance and inspections	Cost [USD]
			$x_{cb2}^{2019}$	Port/border handling
	$x_{cd}^t$	Documentary procedures	$x_{cd}^{2019}$	Documentary procedures
	$x_d^t$	Number of documents	$x_d^{2019}$	Number of documents	[No.]

. $x_{tb}^t\left[\mathrm{hours}\right] \mathrm{and} x_{cb}^t\left[\mathrm{USD}\right]$ capture the time and cost associated with the procedures at the ports or borders in year t. These include customs clearance and inspections by customs and inspections by other agencies, and cargo handling that takes place at the port or border. These procedures are mandatory for the shipment to cross the country’s borders. If all customs clearance and other inspections take place at the port or border at the same time, the time estimate for border procedures takes this into account. If some or all customs or other inspections take place at other locations, the time and cost for these procedures are added to the time and cost for those that take place at the port or border. $x_{td}^t \left[\mathrm{hours}\right] \mathrm{and} x_{cd}^t \left[\mathrm{USD}\right]$ indicate the time and cost for obtaining, preparing, and submitting documents during transport, clearance, inspections, and port or border handling at the origin, destination, and transit countries in year t. These variables cover all documents required by law and in practice, including the electronic submission of information. Finally, $x_d^t$ records the number of documents required by law, or as common practice, by relevant agencies for each import and export shipment in year t. For the country that has import or export cargo going through another country, the documents required by authorities in the transit country are also included.

In the cross-sectional analysis in 2019, we used seven variables as the inputs for calculating efficiency, as shown in Table 1. The three input variables of the panel data analysis were also used in the cross-sectional analysis in 2019. These were $x_{td}^{2019} \left[\mathrm{hours}\right] \mathrm{and} x_{cd}^{2019} \left[\mathrm{USD}\right]$: the time and cost for obtaining, preparing, and submitting documents, and $x_d^{2019}$: the number of documents required by relevant agencies. By contrast, $x_{tb}^{2019}$ and $x_{cb}^{2019}$ in the panel data analysis with the DEA window method were separated by customs clearance and inspections and cargo handling at the seaport or border crossing point. $x_{tb1}^{2019}\left[\mathrm{hours}\right]$ and $x_{cb1}^{2019}\left[\mathrm{USD}\right]$ were the time and cost for customs clearance and inspections by customs clearance and inspections by other agencies. $x_{tb2}^{2019}\left[\mathrm{hours}\right] \mathrm{and} x_{cb2}^{2019}\left[\mathrm{USD}\right]$ were the time and cost for cargo handling that took place at the port or border. These detailed data were not available for the years before 2018, and thus, could not be used in the panel data analysis from 2014 to 2019.

We collected these data for our input variables from the data in “Trading across borders” in each World Bank annual report, titled Doing Business, and the countrywide report, titled Economy Profile, published annually [21,29,30,31,32,33,34]. Note that Doing Business is a series that investigates the regulations that enhance business activity and those that constrain it [33]. Doing Business covers quantitative indicators on business regulations and the protection of property rights that can be compared across 190 countries—from Afghanistan to Zimbabwe, and over time [2]. In the category of “Trading across borders”, Doing Business records the time and costs associated with the logistical process of exporting and importing goods. Data in each report were current as of June or May 1 of the year before the annual report. The data in Doing Business are gathered through a questionnaire distributed to local freight forwarders, customs brokers, traders, and government agencies [2]. Questionnaire responses were verified through several rounds of follow-up communication with respondents as well as through third parties, and by consulting public sources [2].

3.4. Reforms

To investigate the change in efficiency in each country after reforms, we considered reforms that facilitated trade by implementing cost-effective, time-efficient, and transparent regulatory practices in each country. The data are collected from the record of reforms in Trading Across Borders compiled by the World Bank [9]. Each country’s reforms are classified into three groups, as shown in

Table 2. Number of countries that implemented reforms in each category.

Reform Category		2015–2016	2016–2017	2017–2018	Total
Enhanced customs administration and inspections	Export	1 (0)	11 (10)	16 (11)	28 (21)
	Import	1 (1)	11 (10)	15 (12)	27 (23)
Introduced or improved electronic submission and processing of documents	Export	21 (18)	18 (17)	21 (27)	60 (52)
	Import	25 (21)	16 (15)	19 (16)	60 (52)
Strengthened transport or port infrastructure	Export	1 (1)	11 (11)	7 (7)	19 (19)
	Import	3 (3)	9 (9)	10 (9)	22 (21)

Note: The number of countries excluding the country with missing data in certain years is mentioned in parentheses.

. These were “Enhanced customs administration and inspections”, “Introduced or improved electronic submission and processing of documents”, and “Strengthened transport or port infrastructure”. First, the examples of the reform classified in the reform category “Enhanced customs administration and inspections” are as follows. In 2016–2017, El Salvador increased the number of customs officers for clearance and inspections, reducing border compliance time [9]. In 2018–2019, Ukraine made trading across borders easier by eliminating the verification requirement on auto-parts [9]. Second, the examples of the reform classified in the reform category “Introduced or improved electronic submission and processing of documents” are as follows. In 2015–2016, Argentina introduced a new Import Monitoring System, which reduced the time for import documentary compliance by 144 h [9]. In 2016–2017, Bolivia upgraded its automated customs system and reduced documentary compliance time to export [9]. Third, the examples of the reform classified in the reform category “Strengthened transport or port infrastructure” are as follows. In 2016–2017, Angola rehabilitated the Port of Luanda, improving handling processes and reducing border compliance time [9]. In 2018–2019, El Salvador made exporting easier by introducing an intermediate customs post in Santa Ana, reducing congestion at the Anguiatú border crossing [9]. Between 2015 and 2018, several countries such as Kosovo and the Kyrgyz Republic implemented reforms to enter a customs union or signed a trade agreement with a major trade partner for exports and imports. However, due to its limited sample size, a comparison of efficiency values before and after the reforms is not conducted.

3.5. Data Collection

Our first intention was to compare the efficiency of document and border procedures among countries all over the world. Thus, the sample comprised the world’s 190 countries, and we investigated their document and border procedures in the Doing business reports. However, we ended up excluding 22 countries, namely, the Dominican Republic; Eritrea; Finland; the Gambia; Iran; Islamic Rep.; Kiribati; Korea, Rep.; Kyrgyz Republic; Liechtenstein; Micronesia, Fed. Sts.; Niger; Nigeria; Samoa; Slovak Republic; Somalia; South Africa; Syrian Arab Republic; Tonga; Venezuela, RB; Yemen, Rep.; Zambia; and Zimbabwe. The required data for the input variables were not available in certain years. We also found missing data for the output variables in six countries out of the remaining 168 countries, namely, Azerbaijan, Kosovo, Montenegro, San Marino, São Tomé and Príncipe, West Bank, and Gaza. Thus, these were also excluded from our analysis. Thus, our final sample for analysis was a total of 162 countries.

Important statistics relating to the sample in 2019 are summarized in

Table 3. Summary sample statistics in 2019.

Export	Inputs							Output
	$x_{tb1}^{2019}$	$x_{tb2}^{2019}$	$x_{td}^{2019}$	$x_{cb1}^{2019}$	$x_{cb2}^{2019}$	$x_{cd}^{2019}$	$x_d^{2019}$	$y_{mt}^{2019}$	$y_{nv}^{2019}$
Mean	34	35	45	190	199	120	7.3	14,084,213	15,637,062,573
Median	10	24	24	134	170	85	7.0	991,376	1,086,768,344
S.D.	44	38	61	200	203	166	2.4	42,411,872	43,563,612,935
Kurtosis	2.9	11	19	13	3.4	64	-0.2	33.7	28.3
Skewness	1.7	2.5	3.4	2.8	1.4	6.7	0.6	5.5	5.0
Range	204	276	504	1500	1200	1800	14	332,287,612	316,065,646,203
Min.	0	0	1	0	0	0	4	36	41,562
Max.	204	276	504	1500	1200	1800	10	332,287,576	316,065,604,641
No. of Countries	162	162	162	162	162	162	162	162	162
Import	Inputs							Output
	$x_{tb1}^{2019}$	$x_{tb2}^{2019}$	$x_{td}^{2019}$	$x_{cb1}^{2019}$	$x_{cb2}^{2019}$	$x_{cd}^{2019}$	$x_d^{2019}$	$y_{mt}^{2019}$	$y_{nv}^{2019}$
Mean	47	46	54	225	220	157	7.3	6,604,451	14,174,380,757
Median	24	36	33	163	179	93	7.0	851,903	1,687,227,298
S.D.	60	52	66	248	240	192	2.4	18,564,114	40,472,822,572
Kurtosis	6.5	4.1	4.1	16	4.6	6.5	−0.5	50	30
Skewness	2.2	1.9	1.9	3.1	1.7	2.4	0.4	6	5
Range	360	267	360	2000	1476	1025	14	174,284,734	316,065,646,203
Min.	0	0	1	0	0	0	3	2640	10,433,061
Max.	360	267	360	2000	1476	1025	11	174,282,094	316,055,213,142
No. of Countries	162	162	162	162	162	162	162	162	162

Note: $x_{tb1}^{2019}\left[\mathrm{hours}\right], x_{cb1}^{2019}\left[\mathrm{USD}\right]$: the time and cost for customs clearance and inspections by customs clearance and inspections by other agencies. $x_{tb2}^{2019}\left[\mathrm{hours}\right], x_{cb2}^{2019}\left[\mathrm{USD}\right]$: the time and cost for handling that takes place at the port or border; $x_{td}^{2019} \left[\mathrm{hours}\right], x_{cd}^{2019} \left[\mathrm{USD}\right]$: the time and cost for obtaining, preparing, and submitting documents; $x_d^{2019}$: the number of documents required by relevant agencies; $y_{mt}^{2019}$[MT], $y_{nv}^{2019}$[USD]: the cargo volume and value of all commodities transported from and to each country’s trading partner by major mode used for the chosen export or import product.

. All variables’ distributions were positively skewed, as shown in the rows of Skewness in Table 3. This means that the mass of the distribution was concentrated on the small side. In addition, the cost for procedures at the port or border, namely, the sum of $x_{cb1}^{2019}$ and $x_{cb2}^{2019}$, was more expensive than that for document procedures $x_{cd}^{2019}$ in both export and import procedures. Comparing the input variables for export to that for import, all import input variables were greater than those for export except for the number of documents, as shown in the columns for the input variables in Table 3. This means that the import procedures for auto parts required more time and cost than the export procedures for the product of each country’s comparative advantage. Regarding the mean of $y_{mt}^{2019}$[MT], the cargo volume of all commodities traded with each country’s trading partner, that of imports was smaller than that of exports, as shown in the columns of output variables in Table 3. The same can be seen in the mean of $y_{nv}^{2019}$[USD], the cargo value of all commodities traded with each country’s trading partner, as shown in Table 3.

Figure 1 and Figure 2 show the ratio of each input and output variable’s mean compared with that in 2019. The time needed for the border and document procedures for both imports and exports, namely, $x_{td}^t \left[\mathrm{hours}\right]$ and $x_{tb}^t \left[\mathrm{USD}\right]$, respectively, decreased in this period, as shown in the figures. The figures indicate that the load for the border and document procedures decreased in terms of time. In contrast, the cost of border and document procedures for both imports and exports, $x_{cd}^t\left[\mathrm{hours}\right]$ and $x_{cd}^t\left[\mathrm{USD}\right],$ slightly decreased or did not change much in the observed periods, and the number of both import and export documents, $x_d^t$, increased over the years. Regarding the output variables, the import and export cargo volume for all commodities, $y_{mt}^t$[MT], consistently increased from 2016 to 2019. The cargo value of all commodities, $y_{nv}^t$[USD], for both imports and exports also increased after 2016 up to 2018, as shown in Figure 1 and Figure 2. This means that the volume and value of trade with the trading partner increased during these periods. Furthermore, both decreased up to 2016. The decline in trade value around 2015 was due to several factors, including an economic slowdown in China, a severe recession in Brazil, falling prices of oil and other commodities, and exchange rate volatility [35].

3.6. Sample Size

The DEA literature has suggested that there should be a sufficient number of observations in comparison with the number of factors, as shown in the following equations.

$$n > 2 \ast \left(m+s\right)$$

(13)

$$n > 3 \ast \left(m+s\right)$$

(14)

$$n > 2 \ast m\ast s$$

(15)

where $n$ is the number of DMUs, $m$ is the number of input variables, and s is the number of output variables.

Golany and Roll reference and apply Equation (13); Banker et al. and Friedman and Sinuany-Stern, Equation (14); and Dyson et al., Equation (15) [36,37,38,39]. In addition, Pedraja-Chaparro et al. mention that the DEA loses its discrimination power in terms of the number of efficient and inefficient units when the value of $n/\left(m+p\right)$ is too small [40].

Our cross-sectional analysis in 2019 had the largest number of input variables (seven) and output variables (two) and the smallest number of DMUs (164). Even in this circumstance, all the cutoff thresholds were satisfied. Thus, we used these variables and DMUs in the following analyses. Regarding the window length of the panel data analysis in the DEA window method, Asmild et al. identified that it should be as small as possible to minimize the unfairness of comparison over time, but still large enough to include a sufficient sample size [20]. Thus, we defined the length of the window as two years.

4. Panel Data Analysis

4.1. Properties of Window Analysis

The following equations from Cooper and Seiford [41] can be used to study the properties of window analysis. Here, we introduce the symbols, $n$: Number of DMUs; $k$: number of periods; and $p$: length of the window $\left(p\le k\right)$.

Number of windows: $w,$

$$w=k-p+1=6-2+1=5$$

(16)

Number of different DMUs,

$$npw=164\ast 2\ast 5=1640$$

(17)

Thus, the DEA model is applied to 1640 different data points to obtain the efficiency scores.

4.2. Change in Efficiency from 2014 to 2019

The time-series change in the efficiency of the document and border procedures, calculated with the window DEA-VRS analysis, is shown in Figure 3 and Figure 4. The efficiency in 2014–2015 was the lowest among the investigated periods for both imports and exports. The mean and median efficiency in export document and border procedures increased year by year, while that for import document and border procedures peaked in 2015–2016. This means that more countries have been becoming concentrated on the high efficiency side only for export document and border procedures.

Note: the x marks and midline in the boxes represent the mean and median of efficiency value in each period respectively. The bottom and top edges of each shaded rectangle represent the first and third quartile of efficiency values in each period, respectively. The vertical line extends from the top and bottom of the rectangle to indicate the maximum and minimum values of efficiency in each period, respectively.

4.3. Change in Efficiency after Reforms

We compared the efficiencies of the document and border procedures calculated with the DEA-VRS model, before and after the three types of reforms, namely, “Enhanced customs administration and inspections”, “Introduced or improved electronic submission and processing of documents”, and “Strengthened transport or port infrastructure”. The efficiency just before the period of reform was compared with that just after the period of reform. For example, if a country reformed its document and border procedures during 2015–2016, we compared this country’s efficiency in 2014–2015 with that in 2016–2017. As a statistical test, this study used the paired samples t-test.

Regarding the export document and border procedures, the efficiency value before and after all three reforms differed statistically at less than 0.1% significance, as shown in

Table 4. The change in the efficiency of the DEA-VRS model after reforms.

	Customs Administration and Inspections		Electronic Submission and Processing of Documents		Transport or Port Infrastructure
Export	Before †	After ‡	Before †	After ‡	Before †	After ‡
Mean	0.39	0.57	0.38	0.52	0.41	0.55
Median	0.38	0.62	0.38	0.50	0.39	0.57
S.D.	0.09	0.13	0.15	0.18	0.14	0.16
No. of Countries	21	21	52	52	19	19
p-value	p < 0.001 **		p < 0.001 **		p < 0.001 **
Import	Before †	After ‡	Before †	After ‡	Before †	After ‡
Mean	0.36	0.42	0.35	0.40	0.40	0.39
Median	0.36	0.38	0.32	0.38	0.36	0.38
S.D.	0.16	0.18	0.13	0.13	0.13	0.11
No. of Countries	23	23	52	52	21	21
p-value	p = 0.778		p = 0.002 *		p = 0.8076

Note: †: The efficiency just before the period of reform, ‡: The efficiency just after the period of reform, *: p-value < 0.01, **: p-value < 0.001.

. A comparison of the mean and median of the efficiency value in Table 4 revealed that the export document and border procedures after each reform had a higher efficiency value than that before each reform. Additionally, the efficiency value of import document and border procedures changed only after introducing or improving the electronic submission and processing of documents, differing statistically at less than 1% significance, as shown in Table 4. Its mean and median efficiency value after this reform was also higher than that before this reform. This result shows that the efficiency of export document and border procedures improved after all three reforms while the efficiency of import document and border procedures improved only after the introduction of or improvement in the electronic submission and processing of documents. These results are consistent with the findings of Porto et al., Carballo et al., and Lanz et al. [8,10,11]. In other words, the efficiencies of import document and border procedures did not differ statistically before and after the reforms for enhancing customs administration and inspections and strengthening transport or port infrastructure, even at a 5% significance level, as shown in Table 4.

5. Cross-Sectional Analysis

To investigate the import document and border procedure efficiency by country, we conducted further analyses with the 2019 data, focusing on differences among the countries and the categories of inputs regarding import procedures. The inputs in these categories can be improved via additional reforms.

Table 5. The efficiency of import document and border procedures in 2019.

Economy	CRS	Rank	VRS	SE	Return	Economy	CRS	Rank	VRS	SE	Return
Canada	1.00	1	1.00	1.00	-	Georgia	0.00	36	0.76	0.00	drs
France	1.00	1	1.00	1.00	-	Iceland	0.00	38	0.75	0.00	drs
Mexico	1.00	1	1.00	1.00	-	Singapore	0.02	39	0.73	0.03	drs
Netherlands	1.00	1	1.00	1.00	-	Japan	0.57	40	0.73	0.79	drs
United States	1.00	1	1.00	1.00	-	Botswana	0.03	41	0.72	0.04	drs
Poland	0.67	1	1.00	0.67	drs	Eswatini	0.01	42	0.72	0.02	drs
Austria	0.66	1	1.00	0.66	drs	Taiwan, China	0.02	43	0.71	0.03	drs
Italy	0.66	1	1.00	0.66	drs	New Zealand	0.03	44	0.66	0.04	drs
Belgium	0.53	1	1.00	0.53	drs	Switzerland	0.30	45	0.65	0.47	drs
Germany	0.51	1	1.00	0.51	drs	Moldova	0.00	46	0.65	0.00	drs
Czech Republic	0.44	1	1.00	0.44	drs	Norway	0.03	47	0.64	0.05	drs
Kazakhstan	0.41	1	1.00	0.41	drs	North Macedonia	0.01	48	0.62	0.01	drs
Spain	0.36	1	1.00	0.36	drs	Namibia	0.01	49	0.62	0.02	drs
Sweden	0.21	1	1.00	0.21	drs	Australia	0.06	50	0.60	0.11	drs
Denmark	0.18	1	1.00	0.18	drs	Palau	0.00	51	0.60	0.00	drs
Romania	0.16	1	1.00	0.16	drs	Bosnia and Herzegovina	0.00	52	0.59	0.01	drs
Luxembourg	0.12	1	1.00	0.12	drs	Bahamas, The	0.02	53	0.57	0.03	drs
Portugal	0.10	1	1.00	0.10	drs	Bhutan	0.01	54	0.54	0.02	drs
Greece	0.05	1	1.00	0.05	drs	Serbia	0.01	54	0.54	0.01	drs
Bulgaria	0.04	1	1.00	0.04	drs	Mauritius	0.00	56	0.53	0.00	drs
Croatia	0.04	1	1.00	0.04	drs	Rwanda	0.00	57	0.52	0.00	drs
Lithuania	0.03	1	1.00	0.03	drs	Iraq	0.04	58	0.50	0.08	drs
Malta	0.02	1	1.00	0.02	drs	Philippines	0.04	58	0.50	0.07	drs
Estonia	0.02	1	1.00	0.02	drs	Marshall Islands	0.02	58	0.50	0.03	drs
Latvia	0.02	1	1.00	0.02	drs	Mongolia	0.01	58	0.50	0.03	drs
Slovenia	0.01	1	1.00	0.01	drs	Uruguay	0.01	58	0.50	0.02	drs
Lesotho	0.01	27	0.98	0.01	drs	Ecuador	0.01	58	0.50	0.02	drs
Armenia	0.02	28	0.95	0.02	drs	Antigua and Barbuda	0.00	58	0.50	0.01	drs
Hungary	0.21	29	0.80	0.26	drs	Costa Rica	0.00	58	0.50	0.00	drs
Belarus	0.45	30	0.80	0.56	drs	St. Lucia	0.00	58	0.50	0.00	drs
Ireland	0.24	31	0.79	0.30	drs	Solomon Islands	0.00	58	0.50	0.00	drs
Panama	0.05	32	0.79	0.06	drs	Nicaragua	0.00	58	0.50	0.00	drs
Hong Kong SAR	0.09	33	0.78	0.11	drs	Oman	0.01	69	0.49	0.01	drs
United Kingdom	0.48	34	0.77	0.62	drs	China	0.39	70	0.46	0.83	drs
Turkey	0.08	35	0.76	0.10	drs	Malaysia	0.03	71	0.44	0.07	drs
Cyprus	0.00	36	0.76	0.01	drs	UAE	0.02	72	0.44	0.04	drs
Albania	0.00	73	0.43	0.00	drs	Maldives	0.00	91	0.38	0.00	drs
Guatemala	0.07	74	0.43	0.16	drs	St. Kitts and Nevis	0.00	91	0.38	0.00	drs
Brazil	0.05	74	0.43	0.12	drs	Bolivia	0.00	91	0.38	0.00	drs
Uzbekistan	0.03	74	0.43	0.06	drs	Benin	0.00	91	0.38	0.00	drs
Saudi Arabia	0.02	74	0.43	0.04	drs	Mauritania	0.00	91	0.38	0.00	drs
Israel	0.01	74	0.43	0.03	drs	Comoros	0.00	91	0.38	0.00	drs
Barbados	0.00	74	0.43	0.01	drs	Lao PDR	0.02	114	0.35	0.06	drs
East Timor	0.00	74	0.43	0.00	drs	Russia	0.05	115	0.34	0.13	drs
Bahrain	0.00	74	0.43	0.00	drs	Colombia	0.10	116	0.33	0.29	drs
Brunei	0.00	74	0.43	0.00	drs	Indonesia	0.04	117	0.33	0.11	drs
Grenada	0.00	74	0.43	0.00	drs	Cambodia	0.03	117	0.33	0.09	drs
Equatorial Guinea	0.00	74	0.43	0.00	drs	Peru	0.03	117	0.33	0.09	drs
St. Vincent and the Grenadines	0.00	74	0.43	0.00	drs	Argentina	0.02	117	0.33	0.06	drs
Vanuatu	0.00	74	0.43	0.00	drs	Morocco	0.02	117	0.33	0.05	drs
Seychelles	0.00	74	0.43	0.00	drs	Haiti	0.01	117	0.33	0.02	drs
Thailand	0.08	88	0.42	0.20	drs	Kuwait	0.00	117	0.33	0.01	drs
Ukraine	0.13	89	0.38	0.34	drs	Senegal	0.00	117	0.33	0.01	drs
Chile	0.10	90	0.38	0.28	drs	Madagascar	0.00	117	0.33	0.01	drs
Vietnam	0.05	91	0.38	0.13	drs	Cape Verde	0.00	117	0.33	0.01	drs
Honduras	0.03	91	0.38	0.09	drs	Congo, D.R.	0.00	117	0.33	0.01	drs
Jamaica	0.02	91	0.38	0.05	drs	Lebanon	0.00	117	0.33	0.00	drs
Paraguay	0.01	91	0.38	0.03	drs	Gabon	0.00	117	0.33	0.00	drs
Libya	0.01	91	0.38	0.02	drs	Congo, Rep.	0.00	117	0.33	0.00	drs
Tunisia	0.01	91	0.38	0.02	drs	Chad	0.00	117	0.33	0.00	drs
Papua New Guinea	0.01	91	0.38	0.02	drs	Burkina Faso	0.00	117	0.33	0.00	drs
Guinea	0.00	91	0.38	0.01	drs	Tajikistan	0.00	133	0.33	0.01	drs
Djibouti	0.00	91	0.38	0.01	drs	El Salvador	0.02	134	0.32	0.06	drs
Qatar	0.00	91	0.38	0.01	drs	Ethiopia	0.00	135	0.32	0.00	drs
Guyana	0.00	91	0.38	0.01	drs	Malawi	0.00	136	0.31	0.00	drs
Sri Lanka	0.00	91	0.38	0.01	drs	Bangladesh	0.07	137	0.30	0.23	drs
Belize	0.00	91	0.38	0.01	drs	Nepal	0.07	138	0.30	0.23	drs
Cameroon	0.00	91	0.38	0.00	drs	India	0.05	138	0.30	0.16	drs
Jordan	0.00	91	0.38	0.00	drs	Myanmar	0.04	138	0.30	0.12	drs
Suriname	0.00	91	0.38	0.00	drs	Tanzania	0.01	138	0.30	0.02	drs
Guinea-Bissau	0.00	91	0.38	0.00	drs	Pakistan	0.00	138	0.30	0.01	drs
Angola	0.00	138	0.30	0.01	drs	Ghana	0.00	151	0.27	0.00	drs
Kenya	0.00	138	0.30	0.01	drs	Togo	0.00	151	0.27	0.00	drs
Dominica	0.00	138	0.30	0.01	drs	Afghanistan	0.00	151	0.27	0.00	drs
Fiji	0.00	138	0.30	0.01	drs	Central African Rep.	0.00	151	0.27	0.00	drs
Mali	0.00	138	0.30	0.00	drs	Burundi	0.00	157	0.25	0.00	drs
Sierra Leone	0.00	138	0.30	0.00	drs	Cote d’Ivoire	0.00	158	0.23	0.01	drs
Sudan	0.00	138	0.30	0.00	drs	Liberia	0.00	158	0.23	0.01	drs
Mozambique	0.05	150	0.29	0.16	drs	Trinidad and Tobago	0.00	158	0.23	0.00	drs
Algeria	0.02	151	0.27	0.09	drs	South Sudan	0.00	158	0.23	0.00	drs
Egypt, Arab Rep.	0.00	151	0.27	0.01	drs	Uganda	0.00	162	0.21	0.00	drs

Note: CRS: efficiency from the CRS-DEA; VRS: efficiency from the VRS-DEA; SE: scale efficiency; Return: return to scale; irs: increasing return to scale; crs: constant return to scale; drs: decreasing return to scale.

shows the results of the efficiency evaluation of the import document and border procedures in 2019 using the DEA approach. The results reveal that among 161 DMUs, five were judged as efficient under the CRS model. These were Canada, France, Mexico, the Netherlands, and the United States. These countries included all the member states of the North American Free Trade Agreement. Under the VRS model, 26 DMUs were identified as efficient. Beyond the previous five mentioned as efficient DMUs in the CRS model, Poland, Austria, Italy, Belgium, Germany, the Czech Republic, Kazakhstan, Spain, Sweden, Denmark, Romania, Luxembourg, Portugal, Greece, Bulgaria, Croatia, Lithuania, Malta, Estonia, Latvia, and Slovenia also reflected efficiency in the DEA VRS model. Countries with large cargo volumes and value showed high efficiency in the CRS model. All the efficient countries in terms of constant returns to scale were among the top seven importers in terms of volume and top five importers in terms of value in the world in 2019. In contrast, the countries with a relatively small total volume and value cargo import, such as Malta and Luxembourg, were also in the efficient group in the VRS model.

We also compared the efficiency value in the VRS model with the scale efficiency value, among the 146 inefficient countries under the CRS model. China and Japan showed smaller efficiency values in the VRS model than their scale efficiency values, as shown in Table 5. The implication is that these countries utilized their input factors less efficiently. Thus, a decrease in input or increase in output would be more worth considering in those countries than adapting their production scales. In contrast, the result shows that the remaining 135 countries were inefficient mainly due to an inappropriate production scale, because their scale efficiency was much lower than the efficiency values in the VRS model. For those countries, adapting their production scales would be much more effective than changing the inputs or outputs.

Further, we investigated the virtual input values to examine the influence of inputs on the efficiency of import document and border procedures. The composition of the virtual inputs of each country is shown in

Table 6. The virtual input composition of import document and border procedures in 2019.

Group	Country	VRS	$x_{td}^{2019}$	$x_{tb1}^{2019}$	$x_{tb2}^{2019}$	$x_{cd}^{2019}$	$x_{cb1}^{2019}$	$x_{cb2}^{2019}$	$x_d^{2019}$
1	France	1.00	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Netherlands	1.00	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Kazakhstan	1.00	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Austria	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Belgium	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Bulgaria	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Croatia	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Czech Republic	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Denmark	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Estonia	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Germany	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Greece	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Italy	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Latvia	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Lithuania	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Luxembourg	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Poland	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Portugal	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Romania	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Slovenia	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Spain	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Sweden	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.98
1	Mexico	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.97
1	Lesotho	0.98	0.00	0.00	0.00	0.00	0.00	0.00	0.95
1	Hungary	0.80	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Belarus	0.80	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Ireland	0.79	0.00	0.00	0.00	0.00	0.00	0.00	0.97
1	New Zealand	0.66	0.00	0.00	0.00	0.00	0.00	0.00	0.97
1	Switzerland	0.65	0.00	0.00	0.00	0.00	0.00	0.00	0.95
1	Moldova	0.65	0.00	0.00	0.00	0.00	0.00	0.00	0.95
1	Palau	0.60	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Antigua and Barbuda	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Costa Rica	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Ecuador	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Iraq	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Marshall Islands	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Mongolia	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Nicaragua	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Philippines	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Solomon Islands	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	St. Lucia	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Uruguay	0.50	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	China	0.46	0.00	0.00	0.00	0.00	0.00	0.00	0.97
1	Bahrain	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Barbados	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Brazil	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Brunei	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Equatorial Guinea	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Grenada	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Guatemala	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Israel	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Saudi Arabia	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Seychelles	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	St. Vincent and the Grenadines	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	East Timor	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Uzbekistan	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Vanuatu	0.43	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Ukraine	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Chile	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Belize	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Benin	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Bolivia	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Cameroon	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Comoros	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Djibouti	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Guinea	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Guinea-Bissau	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Guyana	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Honduras	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Jamaica	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Jordan	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Libya	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Maldives	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Mauritania	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Papua New Guinea	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Paraguay	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Qatar	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Sri Lanka	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	St. Kitts and Nevis	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Suriname	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Tunisia	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Vietnam	0.38	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Russia	0.34	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Colombia	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Argentina	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Burkina Faso	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Cape Verde	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Cambodia	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Chad	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Congo, D.R.	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Congo, Rep.	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Gabon	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Haiti	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Indonesia	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Kuwait	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Lebanon	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Madagascar	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Morocco	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Peru	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Senegal	0.33	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Bangladesh	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Angola	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Dominica	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Fiji	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	India	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Kenya	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Mali	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Myanmar	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Nepal	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Pakistan	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Sierra Leone	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Sudan	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Tanzania	0.30	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Afghanistan	0.27	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Algeria	0.27	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Central African Rep.	0.27	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Egypt, Arab Rep.	0.27	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Ghana	0.27	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Togo	0.27	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Burundi	0.25	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Cote d’Ivoire	0.23	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Liberia	0.23	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	South Sudan	0.23	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Trinidad and Tobago	0.23	0.00	0.00	0.00	0.00	0.00	0.00	1.00
1	Uganda	0.21	0.00	0.00	0.00	0.00	0.00	0.00	1.00
2	Armenia	0.95	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Panama	0.79	0.20	0.00	0.00	0.00	0.00	0.00	0.80
2	Hong Kong SAR	0.78	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	United Kingdom	0.77	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Turkey	0.76	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Cyprus	0.76	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Georgia	0.76	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Iceland	0.75	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Singapore	0.73	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Japan	0.73	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Botswana	0.72	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Eswatini	0.72	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Taiwan, China	0.71	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Norway	0.64	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	North Macedonia	0.62	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Namibia	0.62	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Australia	0.60	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	Bahamas, The	0.57	0.20	0.00	0.00	0.00	0.00	0.00	0.80
2	Serbia	0.54	0.08	0.00	0.00	0.00	0.00	0.00	0.92
2	Bhutan	0.54	0.20	0.00	0.00	0.00	0.00	0.00	0.80
2	Mauritius	0.53	0.20	0.00	0.00	0.00	0.00	0.00	0.80
2	Oman	0.49	0.20	0.00	0.00	0.00	0.00	0.00	0.80
2	Malaysia	0.44	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	UAE	0.44	0.30	0.00	0.00	0.00	0.00	0.00	0.70
2	Thailand	0.42	0.10	0.00	0.00	0.00	0.00	0.00	0.90
2	El Salvador	0.32	0.20	0.00	0.00	0.00	0.00	0.00	0.80
2	Mozambique	0.29	0.20	0.00	0.00	0.00	0.00	0.00	0.80
3	Bosnia and Herzegovina	0.59	0.00	0.00	0.10	0.00	0.00	0.00	0.90
3	Rwanda	0.52	0.00	0.00	0.20	0.00	0.00	0.00	0.80
3	Albania	0.43	0.00	0.00	0.10	0.00	0.00	0.00	0.90
3	Lao PDR	0.35	0.00	0.00	0.10	0.00	0.00	0.00	0.90
3	Tajikistan	0.33	0.00	0.00	0.10	0.00	0.00	0.00	0.90
3	Ethiopia	0.32	0.00	0.00	0.10	0.00	0.00	0.00	0.90
3	Malawi	0.31	0.00	0.00	0.20	0.00	0.00	0.00	0.80
4	Canada	1.00	0.00	0.00	0.00	0.60	0.00	0.00	0.40
4	United States	1.00	0.00	0.00	0.00	0.20	0.00	0.00	0.80
5	Malta	1.00	1.00	0.00	0.00	0.00	0.00	0.00	0.00

Note: VRS: the efficiency calculated with the DEA-VRS model. $x_{tb2}^{2019}\left[\mathrm{hours}\right]$: the time for handling that takes place at its port or border; $x_{td}^{2019} \left[\mathrm{hours}\right], x_{cd}^{2019} \left[\mathrm{USD}\right]$: the time and cost for obtaining, preparing, and submitting documents; $x_d^{2019}$: the number of documents required by relevant agencies; $x_{tb2}^{2019}\left[\mathrm{hours}\right]$: the time for handling that takes place at the port or border; and $x_{cd}^{2019} \left[\mathrm{USD}\right]$: the cost for obtaining, preparing, and submitting documents. However, Malta’s virtual input comprised only $x_{td}^{2019} \left[\mathrm{hours}\right]$: the time for obtaining, preparing, and submitting documents. These input variables reflect the strengths of each country’s import document and border procedures. When a country plans to take steps to improve its procedures, other higher efficiency countries in the same group can be one of the benchmarks.

. A variable whose cell is colored darker red indicates that it has a higher proportion. $x_d^{2019}$ represents the number of documents required by relevant agencies and has the largest ratio among all input variables in most countries, except that of Canada and Malta. However, all countries did not have values in $x_{tb1}^{2019}\left[\mathrm{hours}\right]$ and $x_{cb1}^{2019}\left[\mathrm{USD}\right]$, the time and cost for customs clearance and inspections by customs clearance and inspections by other agencies, or in $x_{cb2}^{2019}\left[\mathrm{USD}\right]$, the cost for handling that takes place at the port or border. In our analysis of the changes in efficiency after reforms, the efficiencies for import document and border procedures did not differ statistically before and after the reforms enhancing customs administration and inspections and strengthening transport or port infrastructure, as shown in Table 4. This may be caused by the relationship between the types of input variables and the types of reforms. Namely, $x_{tb1}^{2019}\left[\mathrm{hours}\right]$ and $x_{cb1}^{2019}\left[\mathrm{USD}\right]$, the time and cost for customs clearance and inspections by customs clearance and inspections by other agencies, may be reduced by enhancing customs administration and inspections. In addition, $x_{cb2}^{2019}\left[\mathrm{USD}\right]$, the cost for cargo handling at the port or border, may decrease when transport or port infrastructure has been upgraded.

We classified all the countries into five groups based on their virtual input value composition. The first group’s virtual inputs consisted mostly of $x_d^{2019}$, the number of documents required by relevant agencies. Most countries were part of this group. The implication is that in terms of inputs, these countries had strength in the number of documents rather than in the other inputs in our study. The countries classified in Groups 2, 3, and 4 had strengths not only in $x_d^{2019}$—the number of documents required by relevant agencies—but also in $x_{td}^{2019}$—the time for obtaining, preparing, and submitting documents

6. Conclusions

Our study conducted a worldwide evaluation of the efficiency of each country’s document and border procedures using both panel and cross-sectional data. Looking at panel data from 2014 to 2019, we investigated time-series changes using the DEA window method. We also considered changes after the three types of reforms, namely, introducing electronic documentation systems, strengthening border infrastructure, and enhancing customs procedures. We found the following.

Regarding the efficiency of export document and border procedures, more countries have been becoming concentrated on the highly efficient side of the distribution, but the efficiency distribution for import document and border procedures did not change much. With respect to changes after the three reforms, export document and border procedures after these reforms became more efficient than before the reforms. This is consistent with the findings in previous studies, which investigated the impact of each reform on trade. However, only the introduction of electronic documentation systems made a difference in the efficiency of import document and border procedures before and after the reforms.

To investigate the import document and border procedure efficiency by country, we conducted a further analysis with cross-sectional data in 2019, consisting of more detailed segmented data. Five countries, namely, Canada, France, Mexico, the Netherlands, and the United States, were judged to be efficient assuming constant returns to scale. Their economies were relatively larger than those of the other countries analyzed. All countries were classified into five groups by the strengths and weaknesses of their document and border procedures, evidenced by the composition of their virtual input values. They include the group composed of countries with strengths mainly in the required number of documents, groups with strengths in the required time for documents or border handling or the required cost of documents in addition to the required number of documents, and others. A country with high efficiency in each group can be one of the benchmarks when another country plans to take steps to improve its procedures. Our results have important implications for countries seeking to expand their imports by helping decision-makers find practical solutions to promote trade efficiency.

However, this study has several limitations. First, there are several data limitations, including the input variables being only for trade with each country’s major trading partners and the lack of details on the reforms implemented in each. Second, although we identified the strengths and weaknesses of the document and border procedures in each country, the DEA does not provide specific suggestions regarding the most efficient ways to improve these procedures. Third, our analysis did not determine how much reform can improve the efficiency of document and border procedures. This is because a single type of reform may improve multiple inputs or have a negative impact on other input variables. Future studies should try to address these limitations. Specifically, what and how much of the input variables dealt with in this study can be changed by each type of reform to improve the document and border procedure efficiency should be analyzed. In addition, what reforms are appropriate for each country considering its characteristics should be investigated. Based on these, practical suggestions can be made to the governments of each country to improve their document and border procedures.