Cost Inefficiency of Japanese Railway Companies and Impacts of COVID-19 Pandemic and Digital Transformation

Abstract

The outbreak of the COVID-19 pandemic seriously affected railway businesses. The motivation of this study is to provide vital information to railway company management and policymakers by quantitatively assessing the cost efficiency of railway operations. We examine the efficiency of Japanese listed railway companies by applying stochastic frontier analysis to their operational and financial data from 2005 to 2020. Then, we classify the companies into four groups by cost efficiency levels and identify the characteristics of the best-practice companies. Furthermore, we analyze the factors influencing cost efficiency before and during the pandemic. Finally, we discuss the sustainable business practices and measures of digital transformation (DX) that can be applied to improve efficiency and survive severe events like the pandemic. From the results, we reveal that cost-efficient companies succeeded in securing profits through the creation of new services by proactive DX investments. The practical contributions of this study are threefold: quantifying the deterioration in efficiency due to the pandemic; identifying characteristics of best-practice companies; and examining the relationship between cost efficiency levels and concrete measures and investments for sustainable business practices. This study proposes a new analytical framework that combines conventional methods.

1. Introduction

The new coronavirus disease (COVID-19) pandemic started spreading rapidly worldwide in 2019. Therefore, the lives and economic activities of people were significantly affected at a global level (Watanabe 2020). The rapid increase in COVID-19 infections in major cities led to city lockdowns and curfews. In Japan, a state of emergency was declared to control human flow and prevent the spread of infection even during eating and drinking, which had a significant impact on economic activities. The spread of the COVID-19 pandemic has significantly impacted railway operations as well. The number of railway passengers decreased due to the decline in the number of commuters and business travelers to avoid crowded trains, the promotion of telework and online meetings, a sharp drop in inbound tourists visiting Japan, and the stagnation in travel demand because of the voluntary curfew on outings under the state of emergency. This affected not only railway businesses but also related businesses, including hotels, leisure businesses, and retail businesses at station buildings.

Before the COVID-19 pandemic, most Japanese listed railway companies were profitable until 2019; of the 26 listed railway companies, 25 were in a surplus and one was in a loss-making trend. However, in 2020, when the COVID-19 pandemic broke out, all railway companies recorded losses. In 2021, five companies were in surplus and twenty-one were in a loss-making trend.

Table 1. The number of Japan’s 26 listed railway companies with railways operating in profit surplus/loss-making.

Fiscal Year	The Number of Surplus Companies	The Number of Loss-Making Companies
2019	25	1
2020	0	26
2021	5	21

Data Source: Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan (2019a), Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan (2020a), Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan (2021a).

shows the number of Japan’s 26 listed railway companies with railways operating in a profit surplus/loss-making.

In response, Japanese railway companies have taken measures to reduce costs and increase revenues to protect their operations. Cost-cutting measures included reducing the number of regular trains, earlier last train times, reducing the number of staffed ticket counters at stations, shortening opening hours, and discontinuing local railway lines (Ishida 2020; The Asahi Shimbun 2021; The Asahi Shimbun 2022; Ichijo 2020).

Table 2. Change in passenger train kilometers (thousand kilometers) (from 2019 to 2021).

No.	Company	2019 (a)	2020 (b)	2021 (c)	Percentage in 2020 to 2019 (b)/(a)	Percentage in 2021 to 2019 (c)/(a)
1	JR-EAST	254,877	252,115	248,331	98.9%	97.4%
2	KEIO	14,713	14,582	14,404	99.1%	97.9%
3	KEISEI	13,933	13,524	13,804	97.1%	99.1%
4	SHIN-KEISEI	2465	2469	2683	100.2%	108.8%
5	TOBU	37,895	38,329	38,294	101.1%	101.1%
6	SEIBU	20,247	20,115	20,100	99.3%	99.3%
7	TOKYU	19,995	20,065	19,440	100.4%	97.2%
8	KEIKYU	15,823	15,761	15,546	99.6%	98.2%
9	ODAKYU	20,998	20,883	20,614	99.5%	98.2%
10	SOUTETSU	5052	5215	4941	103.2%	97.8%
11	CHICHIBU	2075	1649	1729	79.5%	83.3%
12	FUJIKYU	723	532	662	73.6%	91.6%
13	JR-CENTRAL	112,178	104,671	103,473	93.3%	92.2%
14	MEITETSU	40,075	39,522	38,685	98.6%	96.5%
15	JR-WEST	189,530	182,932	175,333	96.5%	92.5%
16	NANKAI	16,347	15,760	15,714	96.4%	96.1%
17	KINTETSU	56,638	55,820	53,302	98.6%	94.1%
18	KEIHAN	13,240	13,169	11,937	99.5%	90.2%
19	KEIFUKU	940	934	920	99.4%	97.9%
20	HANKYU	21,889	21,827	21,766	99.7%	99.4%
21	HANSHIN	8404	8410	8391	100.1%	99.8%
22	KOBE	4308	4154	4138	96.4%	96.1%
23	SANYO	6847	6871	6819	100.4%	99.6%
24	HIRODEN	4817	4817	4208	100.0%	87.4%
25	JR-KYUSHU	63,352	59,658	59,767	94.2%	94.3%
26	NISHITETSU	8686	8435	8302	97.1%	95.6%

Data Source: Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan (2019b), Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan (2020b), Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan (2021b).

shows changes in passenger train kilometers (thousand kilometers) from 2019 to 2021. Here, passenger train kilometers refer to the number of kilometers traveled by passenger trains operating on commercial lines, which facilitates an understanding of the number of trains in operation. Table 2 shows that most railway companies operated fewer trains in 2020 and 2021 compared with 2019.

Table 3. Earlier regularly scheduled last train times in Japanese major railway companies.

Company	Region	Start Year and Month	Number of Implemented Lines	Time to Move Up the Last Train (Minutes)	Source
JR-EAST	KANTO	March 2021	17	3–37	(JR-EAST 2020)
KEIO	KANTO	March 2021	2	10–30	(KEIO 2020)
KEISEI	KANTO	March 2021	5	10–20	(KEISEI 2020)
TOBU	KANTO	March 2021	3	9–15	(TOBU 2021)
SEIBU	KANTO	March 2021	10	20–30	(SEIBU 2020)
TOKYU	KANTO	March 2021	7	8–26	(TOKYU 2021)
KEIKYU	KANTO	March 2021	4	14–30	(KEIKYU 2020)
ODAKYU	KANTO	March 2021	3	7–23	(ODAKYU 2020)
SOTETSU	KANTO	March 2021	2	15–20	(SOTETSU 2021)
JR-CENTRAL	CHUKYO	March 2022	2	15–21	(JR-CENTRAL 2021)
MEITETSU	CHUKYO	May 2021	5	5–30	(MEITETSU 2021)
JR-WEST	KANSAI	March 2021	12	10–30	(JR-WEST 2020b)
NANKAI	KANSAI	May 2021	2	15–17	(NANKAI 2021)
KINTETSU	KANSAI	July 2021	14	8–29	(KINTETSU 2021)
KEIHAN	KANSAI	September 2021	5	13–21	(KEIHAN 2021)
KEIFUKU	KANSAI	March 2021	1	15	(KEIFUKU 2021)
HANKYU	KANSAI	March 2021	3	13–32	(HANKYU 2021)
HANSHIN	KANSAI	March 2021	1	10–14	(HANSHIN 2021)
KOBE	KANSAI	March 2021	2	15–21	(KOBE 2021)
JR-KYUSHU	FUKUOKA	March 2021	2	18–20	(JR-KYUSHU 2020)
NISHITETSU	FUKUOKA	March 2021	1	13–30	(NISHITETSU 2020)

shows the earlier regularly scheduled last train times on Japanese major railway companies.

Table 4. Examples of reduction in the number of stations with staffed ticket counters.

Company	Reduction in the Number of Stations with Staffed Ticket Counters	Source
JR-EAST	300 (2021–2025)	(JR-EAST 2021)
JR-WEST	160 (2020–2022)	(JR-WEST 2020a)
JR-KYUSHU	48 (2022)	(JR-KYUSHU 2021)
CHICHIBU	27 (2022)	(CHICHIBU 2022)
NISHITETSU	24 (2020) 9 (2022)	(NISHITETSU 2020) (NISHITETSU 2022)

shows examples of reductions in the number of stations with staffed ticket counters.

Table 5. Discontinued local lines by Japanese listed railway companies (since 2020).

Company	Region	Line Name	Operating Kilometers	Date of Abolition
JR-EAST	TOHOKU	Ofunato	43.7	1 April 2020
JR-EAST	TOHOKU	Kesennuma	55.3	1 April 2020

Data Source: Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan (2024).

shows the local lines discontinued by Japanese listed railroad companies after 2020. Meanwhile, measures to increase revenue included fare increases and raising the price of regular commuter passes, combined with the introduction of cheaper off-peak commuter passes (Ichijo 2021; The Asahi Shimbun 2023; Matsumoto 2023). This impact on railway operations was not just a temporary event but became the norm.

Subsequently, in 2021, the companies could be divided into profitable and loss-making companies. This division suggests that railway companies, particularly loss-making companies, needed to decrease operational inefficiency. Japanese listed railway companies are classified into the JR Group companies (JR), which were established by the privatization of Japanese National Railways (JNR) in 1987, and the original private railway companies (OPR). JR maintained profitability before the pandemic through an internal compensation mechanism, whereby profits in metropolitan areas and high-speed railway operations were used to cover the losses on regional local lines. The COVID-19 pandemic made this internal compensation mechanism infeasible because of the reduction in the number of passengers on high-speed railways and in metropolitan areas.

The influence of the pandemic has also been discussed abroad. Rothengatter et al. (2021) conducted an analysis of the impact of the COVID-19 pandemic on urban transport in air, railway, and bus transport. The study on people’s attitudes towards public transport use found that people were more aware that it was safer to use bicycles and private cars compared with pre-pandemic levels. In the air transport sector, a study by the International Civil Aviation Organization (ICAO) found that air transport worldwide would provide approximately 50% fewer seats, 2.9 billion fewer passengers, and economic losses of approximately USD 390 billion in 2020. In the railway sector, a study for the first half of 2020 revealed economic losses of approximately EUR 3.7 billion for Deutsche Bahn (Germany) and EUR 4 billion for SNCF (France), with passenger numbers down by around 21%. In 2020, the US Metropolitan Transportation Authority (MTA) estimated an economic loss of USD 12 billion and INR 1609 crore (one crore is INR 10 million) for the Delhi Metro in India. In Switzerland, railway and tram traffic fell by more than 80% (Rothengatter et al. 2021).

The motivation of this study is to clarify the cost inefficiency of Japanese listed railway companies and identify the characteristics of the best-practice railway companies. Then, we analyze the factors influencing cost inefficiency before and during the pandemic. No studies have analyzed the production or cost efficiency of listed Japanese railway companies since the 2010s, the factors that affect cost-efficiency levels, and the impact of the COVID-19 pandemic on the cost inefficiency of Japanese railway companies. We intend to add knowledge about the cost inefficiency of the companies.

The objectives of this study are threefold. (1) We examine the cost inefficiency of listed Japanese railway companies by applying stochastic frontier analysis (SFA) to data from 2005 to 2020, when the COVID-19 pandemic began to influence the world. (2) We identify the characteristics of best-practice railway companies and examine the drivers of their cost inefficiencies. We also analyze the impacts of and differences from the crises that occurred before the COVID-19 pandemic, such as the financial crisis and the Great East Japan Earthquake, in terms of cost inefficiency. Finally, based on the empirical results, we discuss the sustainable business practices required by railway companies to survive after the COVID-19 pandemic.

The contributions of this study are a clarification of the cost inefficiency of listed Japanese railway companies, an investigation into the impact of their cost inefficiencies by the COVID-19 pandemic, and the identification of the characteristics of best-practice railway companies. Then we suggest the relationship between cost inefficiency and the implementation of efficiency improvement measures for each railway company. These insights are useful for policymakers and railway company managers from the perspective of the improvement of cost inefficiency for sustainable operations after the COVID-19 pandemic.

The remainder of this paper is organized as follows. Section 2 reviews previous studies that empirically analyze the efficiency of railway companies using SFA. Section 3 describes the data and methodology used in this study. Section 4 presents the results and discusses the implications for railway business and operations. Finally, Section 5 concludes the paper and provides future research directions.

2. Literature Review

This section reviews previous related studies by classifying them into five groups. The first group is SFA analysis methods, which includes the Battese and Coelli model used in this study and other recent analysis methods. The second group includes studies that analyzed the efficiency of the railway sector, and the third group includes studies that analyzed the efficiency of transport sectors other than the railway sector. We focus on SFA for the second group, but the first group also includes other methodologies, such as data envelopment analysis (DEA). Both methodologies are popular in efficiency analysis, and the same trend applies to the railway sector. Catalano et al. (2019) surveyed efficiency analysis methods used in previous studies with respect to passenger railway services. They found that three main methods were used: regression-based, parametric frontier (e.g., SFA), and nonparametric frontier methods (e.g., DEA). Similarly, Holvad (2020) conducted a literature review of efficiency analysis methods in the railway sector, focusing on frontier-based methods, including both parametric and nonparametric methods, and presented the characteristics of each analysis. The fourth group includes the impact of the COVID-19 pandemic and Digital Transformation (DX) on companies and businesses. The fifth group includes methods to improve efficiency in enterprises and businesses.

2.1. Stochastic Frontier Analysis (SFA)

This study applies SFA to data from Japanese railway companies to measure their cost inefficiencies. The reason for using SFA is that it takes into account the error term, which makes it less susceptible to the effect of data observation errors on efficiency values and easier to test hypotheses compared with DEA. SFA was introduced by Aigner et al. (1977) and Meeusen and Van Den Broeck (1977). Subsequently, several researchers developed various types of models and applications using these models. For instance, Battese and Coelli (1995) proposed a model to measure technical efficiency using a stochastic frontier production function based on panel data, while Battese and Coelli (1996) measured the technical inefficiency of Indian paddy farmers by applying the model to data collected over 10 years.

In this study, the SFA model of Battese and Coelli (1995) is applied. This model has strengths in analyzing efficiency using panel data and is a very popular and well-established method often used in empirical studies these days. Recent studies on technical efficiency using Battese and Coelli’s (1995) model include, for example, the following studies. Tsukamoto (2019) applied it to manufacturing data for 47 prefectures in Japan from 2002 to 2014; Khetrapal (2020) applied it to electricity distribution business data in India from 2006 to 2013; Zhang (2021) applied it to bus services data in the USA from 2008 to 2017; Zhang (2021) applied it to data from the manufacturing industry in Thailand from 2007 to 2017; and Amornkitvikai et al. (2024) applied it to data from the manufacturing industry in Thailand from 2007 to 2017. Srivastava et al. (2024) applied it to data from 2009 to 2019 for commercial banks in India to analyse profit efficiency; Jarboui and Alofaysan (2024) applied it to data from 2010 to 2019 for the top 20 global oil and gas companies by market capitalization and analyzed the impact of the energy transition on environmental and operational efficiency, applied to data from 2010 to 2019.

In addition, recent studies of SFA methods include, for example, Wheat et al. (2019), who proposed a new Student’s t-distribution for the error term and applied it to the analysis of the efficiency of highway maintenance in England; a study that proposed a new multi-product production function using a Multifactor model and applied it to the analysis of financial efficiency in Russian companies (Mitsel et al. 2021); a study proposing a semi-parametric model for estimating a stochastic frontier model with a single index structure, with an unknown link function and a linear index, and applied this to the analysis of the total energy demand of American states (Forchini and Theler 2023); a new proposal for a dependency between two error components using the Copula function and the application of asymmetry in random errors assigning a generalized logistic distribution, applied to the estimation of the production frontier of the Italian airport system and the cost efficiency of the Italian banking sector (Bonanno and Domma 2022); and a newly proposed Spatial Durbin Stochastic Frontier Model (SDF-STE) introducing spillover effects as determinants of technical efficiency which applied this to the analysis of efficiency in the Italian accommodation industry (Galli 2023).

Although this study does not propose a new analytical method as described above, its contribution is that it quantitatively assesses the efficiency of Japanese railway companies using recent data and identifies the characteristics and management measures of efficient companies, thereby providing suggestions for improving efficiency. Another contribution of this study is that it proposes a new framework for the entire series of analyses of railway companies by combining conventional methods.

2.2. Efficiency Analysis of the Railway Sector

Couto and Graham (2009) examined the performance of European railways from 1972 to 1999 using a stochastic cost frontier approach with the Translog function from 1972 to 1999. They found that inefficiencies were essentially explained by the excess capacity and overemployment of labor inputs. Asmild et al. (2009) examined railway operations in 23 European countries between 1995 and 2001 to determine whether railway policy reforms improved railway system efficiency. Specifically, they used a multi-directional efficiency analysis to investigate how railway reforms affected the efficiency of specific cost factors. In their study, the outputs included passenger train-kilometers and freight train-kilometers, while the discretionary inputs included staff costs and material costs, and the non-discretionary inputs included route length. The dummy variables of the reform were used, including accounting separation; complete separation; independent management from government; competitive tendering for passenger services; and market opening for freight transport. They indicated that these reforms generally improve efficiency, and in particular, accounting separation contributes to increased technical efficiency.

Friebel et al. (2010) assessed changes in technical efficiency as a result of the European railway reform using a stochastic production frontier with the Cobb–Douglas functional form, with a focus on passenger transport. They found that the reform positively affected output. Their study thus claimed that the railway sector seems to be sensitive to changes in the regulatory framework and to the way in which reforms are implemented.

The following studies analyzed the efficiency of Japanese railway operators. Mizutani et al. (2009) investigated the effectiveness of yardstick regulations by estimating variable cost frontier functions using a dataset of Japanese railway companies. In the variable cost functions, the estimated coefficient on the yardstick regulation dummy and that on the competitive pressure from outside the industry were statistically significant and negative, indicating that the introduction of yardstick regulations and competition tends to reduce railway companies’ variable costs. Fukui and Oda (2012) argued that the privatization of the JNR was a success for some JR companies (JR-EAST, JR-CENTRAL, and JR-WEST) but not for the others (JR-HOKKAIDO, JR-SHIKOKU, and JR-KYUSHU). Their study also pointed out that adjusting to regional profit disparities was challenging for both the JR companies and the government. Furthermore, one of the most serious flaws of the JNR reform was the lack of a concrete plan for the abolition of financially loss-making local lines. As such, while the JNR reform has been successful, continuing efforts are necessary to adapt to the changing social circumstances surrounding Japanese railways, such as the declining rural population.

Song and Shoji (2016) estimated a quantitative model to examine the relationship between diversification and transport investment for Japanese private railway companies. They revealed that investment in railway operations is influenced by diversification and that internal capital is likely to be used to invest in diversified operations, as well as transport operations. Le et al. (2022) analyzed the operational, cost, and revenue efficiencies of 18 lines operating in a Japanese metropolitan area using DEA. The quality of service for railway operations is reflected by operational efficiency by incorporating in-vehicle congestion into the efficiency assessment. The results indicated that a higher in-vehicle congestion rate leads to lower cost efficiency but higher revenue efficiency. This implies that lines with high in-vehicle congestion rates may not necessarily exhibit high financial performance. Kuramoto and Hirota (2008) analyzed the efficiency of third-sector railways using SFA as well as temporal changes in technical inefficiency using Battese and Coelli’s (1995) time-varying model. They found that the greater the proportion of private sector investment, the more efficient the management.

Wiegmans et al. (2018) analyzed the efficiency of road and railway freight networks in Canada, Europe, and the USA from 2000 to 2012 using SFA. They found that roads were more efficient than railway freight, suggesting that the liberalization of rail freight transport had no impact on efficiency and that the efficiency of the companies using the infrastructure had a significant impact. Whitmore et al. (2022) assessed the economic viability of public transport with shared automated mobility. Using Allegheny County, Pennsylvania, as a case study, the cost-efficiency analysis compared the direct operating costs of shared automated vehicles and automated shuttles with conventional transit buses, suggesting that they could be feasible at a lower cost than buses, on average.

2.3. Efficiency Analysis of Public Transport Other Than the Railway Sector

Holmgren (2013) analyzed the operational efficiency of public transport in Sweden by SFA using annual data from 1986 to 2009 for 26 counties. They indicated that the average cost efficiency decreased from 85.7% in the 1980s to 60.4% between 2000 and 2009, owing to the increased importance of route density and the stricter environmental and safety requirements. Filippini et al. (2015) directly compared the impact of competitive tendering and performance-based contract negotiations with the cost efficiency of Swiss bus services in 2009 using SFA. Their results indicated that the difference in cost efficiency between the two contracting regimes is not significant.

Vigren (2016) measured the cost efficiency of public transport in Sweden using SFA and analyzed how different contracting factors affect cost efficiency. They found that service provision in densely populated areas is associated with lower cost efficiency and efficiency is also lower when a contract is awarded directly to a public transport operator. Holmgren (2018) examined how the output variables of public transportation should be defined and measured in efficiency studies. He compared the Cobb–Douglas production function with the three Translog functions by SFA using data from 1986 to 2015 in Sweden. He suggested that the choice of output variables is important when undertaking efficiency analysis and that a combination of both supply-oriented (vehicle kilometers) and demand-oriented (passenger kilometers) indicators should be used when analyzing public transport performance. Bergantino et al. (2021) applied the spatial stochastic frontier approach (SSFA) to analyze the impact of geographical distances of airports on technical efficiency. Data from 206 airports worldwide were used for the data. The results detected different effects on the efficiency levels of airports depending on the geographical distance. Karanki and Lim (2021) used two stochastic variable cost frontier models, the Battese and Coelli model (1995) and the True Random-Effects model (TRE) to measure the impact on the cost efficiency of airports. The analysis suggested that US airports could reduce operating costs by an average of 2.7%. Krljan et al. (2021) used the stochastic production frontier with control input variables in Cobb–Douglas and logarithmic function forms to analyze the technical efficiency of the North Adriatic Ports Association (NAPA), presenting that the technical efficiency of medium-sized NAPA terminals ranged from 65.24% to 93.92%, with a global average of 78.49. Chung and Chiou (2023) analyzed the technical efficiency of 20 bus companies and 890 route-year observations in Taiwan using stochastic meta-frontier models. They found that most bus routes were efficient in providing bus services but less so in attracting passengers, which indicates insufficient incentive designs in Taiwan’s current subsidy scheme. Bourjade and Muller-Vibes (2023) used the stochastic frontier approach proposed by Kumbhakar and Lovell (2000) using annual data from 134 airlines worldwide for the period 2007–2019) to analyze cost efficiency. The results indicated that airline-leasing plays a strategic role in improving the operational efficiency of airlines after the COVID-19 pandemic.

2.4. The Impact of the COVID-19 Pandemic and Digital Transformation

Yin et al. (2021) described the characteristics of the COVID-19 epidemic and identified vulnerabilities in preventing and containing the spread of the epidemic in the railway transport system. They showed that measures to prevent the spread included passenger services, information, staff, and equipment. Kawaguchi et al. (2022) analyzed the impact of the crisis on firms’ sales, employment, and working hours per employee for Japanese firms during the COVID-19 pandemic, and the role of the work-from-home system in mitigating the negative impact. The study found that adapting to the pandemic by increasing the number of remote-working employees helped companies to moderate the negative impact on sales. Maharjan and Kato (2023) conducted semi-structured interviews with five Japanese companies between November 2020 and February 2021 to understand the impact of the COVID-19 pandemic on different aspects of logistics and supply chain activities and the resilience strategies implemented. The results showed that the interviewed companies experienced both positive and negative impacts of the pandemic on their logistics and supply chain activities, and levels of resilience preparedness, response, and future planning differed between companies with different attributes, such as industry sector and organizational size.

Lee and Eom (2023) systematically reviewed the evidence on transportation responses to the COVID-19 pandemic and identified research trends. They indicated a strong relationship between the pandemic and mobility, with significant impacts on overall mobility decline, significant reductions in transportation use, changes in travel behavior, and improvements in transportation safety. Faber et al. (2023) estimated the relationship between telecommuting and travel behavior using panel data from the Netherlands Mobility Panel through 2017–2021. They found that working from home had a negative effect on commute travel time before and during the pandemic and a positive effect on leisure travel time only before the pandemic. Chauhan et al. (2024) conducted a nationwide online longitudinal survey of behavior during the COVID-19 pandemic in the United States. Survey items included commuting, long-distance travel, telecommuting, online learning, online shopping, pandemic experiences, attitudes, and demographic information. The survey was conducted three times with the same respondents to observe how their reactions to the pandemic changed over time. The survey results indicated that attitudes toward stay-at-home home and business shutdowns declined over time. Vašaničová and Bartók (2024) analyzed the relationship between the share of tourism in employment (TTEMPL) and the share of tourism in gross domestic product (TTGDP) between 2019 and 2021. They proposed a model for 117 countries using quantile regression (QR). The results of the QR showed that individual percentiles of TTGDP are more affected by TTEMPL than other percentiles of TTGDP, which is reflected in the changes in the regression coefficients. Furthermore, the COVID-19 pandemic showed that the tourism recession had a negative impact on TTEMPL and TTGDP.

Magoutas et al. (2024) analyzed how rapid progress in information and communications technology (ICT) is associated with economic growth in the European Union (EU). They conducted their analysis using structural equation modeling, utilizing data sets from the Digital Economy and Society Index 2022 (DESI), the Statistical Office of the European Union (EUROSTAT), and the Organization for Economic Cooperation and Development (OECD) for all 27 European Union member states. The results showed a positive correlation between ICT development and Gross Domestic Product (GDP) indices. They indicated the need to strengthen human capital and promote the growth of e-government technology. They suggested that this is crucial to strengthening the infrastructure that supports citizens and public enterprises in European countries. Lastauskaite and Krusinskas (2024) analyzed the impact of production digitalization investments on the financial performance (operating revenue) of European companies from 2013 to 2021. They conducted a fixed-effects panel regression analysis using a sample size of 5706 records from the Orbis database, covering 30 countries and 634 business units. The results suggest that Eastern European firms benefit more from production digitization than Western European firms. The results showed a tendency for firm costs of employees and intangible fixed assets to increase as investment in production digitization expands. It also indicated that larger firms tend to benefit more from these investments than smaller firms. Hussain et al. (2024) investigated the effects of transportation infrastructure, income inequality, and information and communication technology on transportation-oriented household expenditures. They selected a dataset covering 1997–2021 for 24 OECD countries and analyzed it using a cross-sectional autoregressive distributed lag (CS-ARDL) approach.

They showed a convex relationship between income inequality and household expenditures. The magnitude of information and communication technology (ICT) is negative and significant. They suggested that ICT reduces transportation-related household expenditures by allowing people to access real-time information and choose destinations and modes of travel from the comfort of their homes.

2.5. The Method to Improve Cost Inefficiency

Keuchel (2020) conducted an analysis of the ongoing process of traveler digitalization in Blumenberg. The study said the digitalization of the transport system results in the digitalization of transport information, reservations, and pricing, as well as the automated operation of vehicles, leading to an increase in the number of users, improved transport conditions, and safer transport. Zhang and Wu (2023) proposed a framework for examining the operation performance of public transport using traffic data, e.g., smart card data, bus basic operation data, traditional transportation survey data, and geographical data. They presented that the framework of these traffic data is able to improve public transportation systems and provide a method for further efficient operation. Altieri et al. (2022) examined the impact of railway business strategy on the growth of the population. Their results presented the lagged influence of business strategies with respect to the expansion of the railway network on the growth of population in the Tokyo Metropolitan Area from 1970 to 2015. In particular, they found that the growth of population was positively associated with private sector, diversified businesses and third sector railways, while it was negatively associated with public sector railways. This is because the private railway companies have positively influenced the population growth along the railway line as they jointly conducted railway deployment with urban development through business strategies from the 1970s to the 1990s. Meanwhile, public railways (Japanese National Railways and Subway: Teito Rapid Transit Authority) were not involved in urban development until the privatization of the Japanese National Railways in 1987, and therefore did not influence population growth along the railway lines.

Jansson et al. (2023) considered the challenges that the railway sector may face if automatic train operation is introduced on mainline railways. Automatic train operation enables unmanned operation, which reduces the number of crew members and reduces operating costs. They identified six categories of required driver action, illustrated some of the challenges of unmanned or undermanned train operations, and suggested the need to develop strategies not only for the operational aspects but also for technical operational management.

Fraczek and Urbanek (2021) discussed the reason for the digital payment development in public transportation and identified the influencing factors for the degree of digital ticketing and digital payments in the railway. They used a macroeconomic approach, which was applied to total data for European Union countries. The results from the regression analysis revealed the factors influencing digital payments in passenger transport, and indicated that the degree of digital payments in passengers depends on the degree of mobile device, as well as the degree of the provided electronic banking services. Chiscano and Darcy (2022) examined measures against COVID-19 for Spanish urban public transport, which should ensure safe journeys by intensifying the digitalization of services. They applied the method of service-dominant logic co-creation process for disabled persons via a comparative approach. They suggested that disabled persons are able to improve their experiences of public transport through a suitable combination of digital and non-digital resource allocation.

Erdei et al. (2023) proposed an operational concept that enables flexible transport concepts to meet ad hoc passenger demand, which is made more efficient by combining the travel demands of individual passengers. Bulková et al. (2023) proposed measures to improve the quality and efficiency of transport processes in railway passenger transport during a potential pandemic. They showed that introducing modern passenger equipment systems in rail passenger transport through process automation and digitalization will streamline and improve the quality of transport processes. Varotsis (2022) conducted a comprehensive review and identified six key factors influencing digital entrepreneurship and creative industries in tourism, such as the economic, socio-psychological, and other aspects of tourism.

2.6. Position of This Study

Considering the efforts and achievements of previous studies, our study makes the following original contributions to the literature. First, to the best of our knowledge, few studies have hitherto analyzed the production or cost efficiency of listed Japanese railway companies since the 2010s, as well as the impact of the declining population and the deteriorating efficiency of local lines. An exception is Tomikawa and Goto (2022), who analyzed four types of efficiency measures using recent data; however, their study did not include OPR but only JR.

Second, few previous studies have examined the factors that affect the cost efficiency levels of railway companies excepting, e.g., Asmild et al. (2009). In particular, this study uses variables such as the percentage of local lines, population density, railway usage rate, and number of transformer substations to explain inefficiency. Furthermore, we use dummy variables to identify the different impacts on efficiency caused by the individual characteristics of service areas, regional core cities, and local areas. These examinations identify correlations between demographics along railway lines, the railway infrastructure of substations, and cost inefficiency.

Third, few previous studies have analyzed the impact of the COVID-19 pandemic on the cost inefficiency of Japanese railway companies. This study thus provides useful insights for policymakers and railway company managers from the perspectives of cost inefficiency and sustainable operations against business uncertainties, such as that posed by the COVID-19 pandemic. By comparing the results with those of other railway companies, the corporate leaders of Japanese railway companies can identify the best-practice peers from which they can obtain hints to improve their management, not only under typical but also in critical business situations.

3. Methodology

3.1. Methodology

In this study, we use the stochastic frontier cost function approach that is based on Battese and Coelli (1996), which is one of the most popular models in empirical studies of SFA due to its characteristics with robust mathematical background and flexible capability for panel data applications. The model is described by Equation (1):

$$C_{it}=f\left(x_{itm}\right)V_{it}{U}_{it},$$

(1)

where $C_{it}$ is the cost for period $t(t=1,\dots ,T)$ of firm $i(i=1,\dots ,N)$, $x_{itm}$ is the production factor $m(m=1,\dots ,M)$ in period $t$ of firm $i$, $V_{it}$ is the error term, and $U_{it}$ is the inefficiency term.

Following Battese and Coelli’s (1996) model, time-varying cost inefficiency term $U_{it}$ is described by Equation (2):

$$U_{it}=f\left(z_{it}\right),$$

(2)

where $z_{it}$ represents the factors that explain firm $i$’s inefficiency in period $t$. We use three input factors such that $m$ in Equation (1) represents labor, fuel, and capital, and the variables in the estimation model are logarithmic.

The cost frontier function is formulated as a Cobb–Douglas type, according to Coelli (1996), and Equation (1) becomes Equation (3):

$$ln\left({\displaystyle \frac{C_{it}}{R_{it}}}\right)={\beta }_0+{\beta }_{1}lnln{Q}_{it}+{\beta }_{2}lnln\left({\displaystyle \frac{W_{it}}{R_{it}}}\right)+{\beta }_{3}lnln\left({\displaystyle \frac{E_{it}}{R_{it}}}\right)+V_{it}+U_{it},$$

(3)

where the explanatory variable are passenger kilometers for output $Q_{it}$ and the prices of production factors for labor ($W_{it}$), fuel ($E_{it}$), and capital ($R_{it}$). Labor price $W_{it}$ is calculated from wages divided by the number of employees, capital price $R_{it}$ is the depreciation cost divided by the fixed assets, and fuel price $E_{it}$ is the electricity cost divided by the number of car-kilometers. Explained variable $C_{it}$ is railway operating expenditure. $V_{it}$ is a random error term with a distribution of i.i.d. $N\left(0,{\sigma }_v^2\right)$ and $U_{it}$ represents cost inefficiency, which is a random variable following a non-negative truncation of normal distribution $N^{+}\left({\mu }_{it},{\sigma }_u^2\right)$. Note that the cost function is assumed to satisfy first-order homogeneity with respect to input prices, so that it is divided by capital price.

The estimation model for the average cost inefficiency is specified in Equation (4):

$${\mu }_{it}={\delta }_0+\sum _{j=1}{\delta }_j{Z}_{jit}.$$

(4)

$Z_{jit}$ is the $j$ th variable $(j=1,\dots ,J)$ representing the characteristics of firm $i$ in period $t$ and δ is the vector of estimated coefficients. The concrete cost inefficiency equation estimated in this study is described by Equation (5):

$$\begin{array}{cc}\hfill {\mu }_{it}& ={\delta }_0+{\sum }_{j=1}{\delta }_j{Z}_{jit}\hfill \\ & ={\delta }_0\hfill \\ & \hspace{1em}{+\delta }_1{\left(Percentageoflocallinesoperatingkilometers\right)}_{it}\hfill \\ & \hspace{1em}+{\delta }_2{\left(Numberoftransformersubstaions\right)}_{it}\hfill \\ & \hspace{1em}+{\delta }_3{\left(Populationdensityofrailwayoperatingarea\right)}_{it}\hfill \\ & \hspace{1em}+{\delta }_4{\left(Railwayusagerate\right)}_{it}\hfill \\ & \hspace{1em}+{\delta }_5\left(Dummyofregionalcorecity\right)\hfill \\ & \hspace{1em}+{\delta }_6\left(Dummyoflocalareas\right)\hfill \end{array}$$

(5)

The estimated parameters and variables for inefficiency in Equation (5) are as follows:

(i)

${\delta }_1$ (percentage of local lines’ operating kilometers): to examine whether inefficiency increases with a higher percentage of local routes and fewer users.

(ii)

${\delta }_2$ (number of transformer substations): to examine whether the cost of operating equipment represented by transformer substations, including overhead line facilities and other ground infrastructure required for electric railways, affects inefficiency.

(iii)

${\delta }_3$ (population density of railway operating areas: to examine whether population density in relation to the number of railway passengers affects inefficiency.

(iv)

${\delta }_4$ (railway usage rate): to examine whether railway usage rate affects inefficiency.

(v)

${\delta }_5$ (dummy variable of regional core city) and ${\delta }_6$ (dummy variable of local areas): to examine whether there are differences in inefficiency according to city size and location.

The parameters in the cost frontier model were then obtained using maximum likelihood estimation.

Cost inefficiency measure ${\mu }_{it}$ ranges as $1\le {\mu }_{it}$, where 1 represents the status of full efficiency. If ${\mu }_{it}$ is larger than 1, it indicates inefficiency. The larger the value, the more inefficient the system.

3.2. Data

This study examined 26 listed Japanese railway companies.

Table 6. Summary of 26 Japanese listed railway companies.

No.	Company Name	Business Type	Service Area
01	JR-EAST	JR	KANTO, TOHOKU, CHUBU (Yamanashi, Nigata, Nagano, Shizuoka)
02	KEIO	OPR	KANTO (Tokyo)
03	KEISEI	OPR	KANTO (Tokyo, Chiba)
04	SHIN-KEISEI	OPR	KANTO (Chiba)
05	TOBU	OPR	KANTO (Tokyo, Saitama, Gunma, Tochigi, Chiba)
06	SEIBU	OPR	KANTO (Tokyo, Saitama)
07	TOKYU	OPR	KANTO (Tokyo, Kanagawa)
08	KEIKYU	OPR	KANTO (Tokyo, Kanagawa)
09	ODAKYU	OPR	KANTO (Tokyo, Kanagawa)
10	SOTETSU	OPR	KANTO (Kanagawa)
11	CHICHIBU	OPR	KANTO (Saitama)
12	FUJIKYU	OPR	CHUBU (Yamanashi)
13	JR-CENTRAL	JR	KANTO (Tokyo, Kanagawa), CHUBU (Shizuoka, Nagano, Yamanashi), CHUKYO (Aichi, Gifu, Mie), KANSAI (Shiga, Kyoto, Osaka)
14	MEITETSU	OPR	CHUKYO (Aichi, Gifu, Mie)
15	JR-WEST	JR	CHUBU (Nigata, Nagano), HOKURIKU, KANSAI, CHUGOKU, KYUSHU (Fukuoka)
16	NANKAI	OPR	KANSAI (Osaka, Wakayama)
17	KINTETSU	OPR	CHUKYO (Aichi Mie), KANSAI (Osaka, Kyoto, Nara)
18	KEIHAN	OPR	KANSAI (Osaka, Kyoto)
19	KEIFUKU	OPR	HOKURIKU (Fukui), KANSAI (Kyoto)
20	HANKYU	OPR	KANSAI (Osaka, Kyoto, Hyogo)
21	HANSHIN	OPR	KANSAI (Osaka, Hyogo)
22	KOBE	OPR	KANSAI (Hyogo)
23	SANYO	OPR	KANSAI (Hyogo)
24	HIRODEN	OPR	CHUGOKU (Hiroshima)
25	JR-KYUSHU	JR	KYUSHU
26	NISHITETSU	OPR	KYUSHU (Fukuoka)

Note: Service areas are described by combinations of regions, such as KANTO and TOHOKU, and the prefectures in the region are shown between parentheses, such as Tokyo and Chiba.

summarizes the 26 listed railway companies in Japan and Figure 1 shows a map of service areas for each company.

This study uses a panel dataset comprising 26 Japanese companies from 2005 to 2020 (Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan 2005–2020). For analysis, we used Japan’s fiscal year from April to March. This period includes several events that deteriorated the business environment, such as the financial crisis from 2007 to 2008 and the Great East Japan Earthquake of 2011, and we compare those impacts with those of the COVID-19 pandemic after 2020.

The sources of the data used in this study are “Annual Report on Railway Statistics” and “Inter-Regional Travel Survey in Japan” published by the Ministry of Land, Infrastructure, Transport and Tourism, Japan; the “National Census” by the Ministry of Internal Affairs and Communication; and the annual securities reports of each railway company. The variables used for the inefficiency factors are as follows:

(i)

Percentage of operating kilometers of local lines: We used data from the Annual Report on Railway Statistics. Specifically, this is the ratio of all local lines’ operating kilometers, calculated as an average daily transit capacity (transit density) of local lines with less than 2000 passengers/day to the total operating kilometers. Here, the limit of 2000 passengers/day is determined by Japan National Railways as a guideline for discontinuing railway lines.

(ii)

Number of transformer substations: We used data from the annual securities reports of each railway company. This number represents the total number of traction substations owned by each railway company.

(iii)

Population density of railway operating areas: We used census data for the population and the area of each prefecture. As the census is conducted every 5 years, the most recent census data are used for several years. For example, 2005 data were used from 2005 to 2009 and 2010 data were used from 2010 to 2014. Specifically, we used the population density of the prefecture in which each railway company operates. If the operating area extended over several prefectures, the population density was calculated using the average value.

(iv)

Railway usage rate. We used data from the inter-prefectural passenger table of the Inter-Regional Travel Survey in Japan. However, as we could not access data on private car use, we used the percentage of railway users relative to that of the entire public transport system. Specifically, we calculated the rate for the four JR companies (JR-EAST, JR-CENTRAL, JR-WEST, and JR-KYUSHU) and for the OPR companies using Equation (6):

$$\begin{array}{l}Railwayusagerate\left(JRorOPR\right)\\ ={\displaystyle \frac{\sum \left(TotalnumberofJRorOPRpassengersineachprefectureofoperatingarea\right)}{\sum \left(Totalnumberofpublictransportpassengersineachprefectureofoperatingarea\right)}}\end{array}$$

(6)

(v)

Dummy of regional core city: It takes 1 if the operating area includes an ordinance-designated city, excepting three metropolitan areas (Tokyo, Kanagawa, Saitama, Chiba, Osaka, Kyoto, Hyogo, Nara, Wakayama, Aichi, Gifu, and Mie prefectures), that is, Fukuoka, Kumamoto, Hiroshima, Okayama, Shizuoka, Hamamatsu, Niigata, or Sendai City, and 0 otherwise.

(vi)

Dummy of local areas: It takes 1 if the three metropolitan areas and regional core city are not included in the operating area, and 0 otherwise.

The software used for the analyses was FRONTIER Version 4.1 for the SFA and IBM SPSS Ver. 29 for the descriptive statistics.

3.3. Flow of Analysis

We can summarize the flow of the analyses in this study as follows. Figure 2 shows the methodology flow. First, we measured the cost inefficiencies of railway companies from 2005 to 2020 by estimating the stochastic frontier cost function and identified the railway companies with the highest and lowest cost inefficiency. Then, we examined the differences in the cost inefficiency of each railway company between 2019 and 2020, that is, before and after the start of the COVID-19 pandemic. Second, we rank listed Japanese railway companies based on cost inefficiency and profile characteristics of the companies with the lowest inefficiency (best practice). Third, we compared the cost inefficiency between the best practice railway companies in 2019 and 2020, that is, before and after the start of the COVID-19 pandemic. Fourth, we examined the relationship between cost inefficiency and the implementation of efficiency measures for each railway company. This is because it is important to examine what efficiency measures influence costs in companies for improving cost efficiency.

4. Results and Discussion

4.1. Estimation Results for Stochastic Cost Frontier Function

Table 7. Descriptive statistics for the 26 Japanese listed railway companies (from 2005 to 2020).

Variable Name	Minimum Value	Maximum Value	Average Value	Standard Deviation
Railway operating expenditure (million yen)	1147	1,715,178	170,278	348,194
Passenger-kilometer (million passenger kilometers)	18	137,598	13,406	26,987
Labor price	1194	9905	5577	1533
Capital price	0.01	36.86	0.15	1.80
Fuel price	16.70	65.63	33.99	9.29
Number of personnel	97	54,697	5680	10,535
Power cost (million yen)	46	71,577	8167	14,380
Depreciation cost (million yen)	105	298,807	34,033	66,131
Salary cost (million yen)	262	297,516	32,717	58,231
Car kilometer (million passenger kilometers)	0.89	2343	275	509
Percentage of local lines operating kilometers	0.00	0.54	0.07	0.14
Number of substations	1	339	43.92	74.07
Population density	181	6403	1824	1662
Railway usage rate	0.06	0.66	0.48	0.14
Dummy of regional core city	0.00	1.00	-	-
Dummy of local areas	0.00	1.00	-	-

shows the descriptive statistics of 26 Japanese listed railway companies to better understand their characteristics. The effective data size is 416, as the data cover 26 companies over 16 years in the form of balanced panel data.

Table 8. Estimation results for the stochastic frontier cost function of the 26 Japanese listed railway companies (from 2005 to 2020).

Variable		Coefficient	Standard Error	t-Value
β0	Constant term	0.0155	0.1153	0.1344
β1	Passenger-kilometers	0.2477 ***	0.0378	6.5461
β2	Labor price/Capital price	0.7876 ***	0.0362	21.7366
β3	Fuel price/Capital price	0.8416 ***	0.0532	15.8068
δ0	Constant term	0.2795 ***	0.093	3.0052
δ1	Percentage of local lines operating kilometers	−0.0001	0.0003	−0.4658
δ2	Number of substations	−0.0001 ***	0.0000	−15.145
δ3	Population density	−0.6801 ***	0.2457	−2.7674
δ4	Railway (JR/OPR) usage rate	0.2107 ***	0.0538	3.914
δ5	Dummy of regional core city	−0.3352 ***	0.0731	−4.5863
δ6	Dummy of local areas	0.4352 ***	0.0743	5.851
	lnsigma2	0.0317 ***	0.0547	5.7998
	Γ	0.8396 ***	0.0350	23.9845
	loglikelihood	411.9217	-	-
	Number of observations	416	-	-
	LR test of the one-sided error	213.8084	-	-

Notes: 1: lnsigma2 stands for $lnln{\sigma }^2=lnln\left({{\sigma }_u}^2+{{\sigma }_v}^2\right),$ where ${{\sigma }_u}^2$ is the variance of the cost inefficiency term and ${{\sigma }_v}^2$ the variance of the random error term. $\gamma$ represents the proportion of the cost inefficiency to the whole error term in the variance. 2: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

presents the results of the stochastic cost frontier function estimation.

The estimated parameters of the stochastic cost frontier function ($\beta$) are all positive and significant at the 1% level, which is consistent with the monotonicity condition of the cost function for input prices. Meanwhile, the estimated parameters of inefficiency ($\delta$) indicate that cost inefficiency increases if the parameter is positive and decreases if it is negative.

The estimated coefficients for the cost frontier function, passenger kilometers, railway operating expenditure, labor price, and fuel price are positive and significant at the 1% level. For the inefficiency term, the number of substations, population density, railway (JR/OPR) usage rate, the dummy of the regional core city, and the dummy of local areas were significant at the 1% level, whereas the percentage of local lines operating in kilometers was not significant. Of the significant variables, the number of substations, population density, and the dummy of the regional core city had negative coefficients, while the railway (JR/OPR) usage rate and the dummy of local areas had positive ones.

Because cost inefficiency decreases when the number of variables with significant and negative estimated coefficients increase, we expect that an increase in the number of substations, population density, and dummy variables of the regional core city would decrease cost inefficiency. Intuitively, an increase in the number of substations would lead to a larger cost inefficiency, owing to operating and maintenance costs; however, in this study, it contributes to reducing cost inefficiency, which probably represents a greater increase in outputs. For example, from 2005 to 2020, some railway companies increased the number of transformer substations by opening new lines or increasing transport capacity, which may have positively affected reducing cost inefficiency by increasing the output values of passenger kilometers. Meanwhile, we can intuitively understand that cost inefficiency decreases as population density increases because of the economies of scale.

4.2. Estimation Results of Cost Inefficiency

Table 9. Cost inefficiency from the stochastic cost frontier functions of railway companies (from 2005 to 2020).

No.	Company	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	Average
	Event				Financial crisis	Financial crisis		Great East Japan Earthquake								Before COVID-19	During COVID-19
1	JR-EAST	1.0796	1.0968	1.0780	1.0524	1.0556	1.0495	1.0491	1.0413	1.0373	1.0380	1.0536	1.0591	1.0530	1.0447	1.0504	1.1949	1.0646
2	KEIO	1.0155	1.0164	1.0173	1.0178	1.0251	1.0216	1.0197	1.0168	1.0145	1.0138	1.0147	1.0180	1.0173	1.0155	1.0159	1.0434	1.0190
3	KEISEI	1.0405	1.0408	1.0407	1.0317	1.0451	1.0527	1.0522	1.0422	1.0331	1.0320	1.0383	1.0477	1.0441	1.0361	1.0426	1.2047	1.0515
4	SHIN-KEISEI	1.0269	1.0275	1.0288	1.0271	1.0365	1.0357	1.0308	1.0233	1.0192	1.0181	1.0204	1.0265	1.0254	1.0236	1.0251	1.0834	1.0299
5	TOBU	1.0354	1.0356	1.0360	1.0320	1.0255	1.0430	1.0360	1.0299	1.0214	1.0202	1.0223	1.0276	1.0263	1.0234	1.0324	1.0880	1.0334
6	SEIBU	1.0315	1.0251	1.0280	1.0241	1.0318	1.0288	1.0263	1.0182	1.0161	1.0163	1.0177	1.0220	1.0206	1.0185	1.0211	1.0732	1.0262
7	TOKYU	1.0255	1.0324	1.0269	1.0266	1.0403	1.0327	1.0354	1.0259	1.0220	1.0192	1.0201	1.0283	1.0247	1.0223	1.0141	1.0386	1.0272
8	KEIKYU	1.0213	1.0213	1.0244	1.0213	1.0327	1.0306	1.0303	1.0248	1.0196	1.0190	1.0204	1.0283	1.0264	1.0210	1.0242	1.0964	1.0289
9	ODAKYU	1.0288	1.0274	1.0277	1.0248	1.0374	1.0326	1.0277	1.0208	1.0179	1.0165	1.0182	1.0248	1.0223	1.0203	1.0212	1.0762	1.0278
10	SOTETSU	1.0304	1.0300	1.0339	1.0286	1.0135	1.0279	1.0273	1.0209	1.0183	1.0182	1.0209	1.0262	1.0239	1.0214	1.0200	1.1153	1.0298
11	CHICHIBU	1.0540	1.0492	1.0458	1.0352	1.0462	1.0397	1.0395	1.0326	1.0287	1.0258	1.0301	1.0414	1.0372	1.0345	1.0365	1.1493	1.0454
12	FUJIKYU	1.1173	1.1389	1.1115	1.0781	1.1257	1.1085	1.0944	1.0620	1.0479	1.0589	1.0790	1.1203	1.1153	1.1266	1.2375	1.6133	1.1397
13	JR-CENTRAL	1.0796	1.0854	1.0832	1.0791	1.1284	1.1137	1.0897	1.0759	1.0473	1.0421	1.0449	1.0558	1.0460	1.0371	1.0491	1.3425	1.0875
14	MEITETSU	1.1182	1.1235	1.1372	1.1163	1.1543	1.1383	1.1148	1.0978	1.0848	1.0623	1.0709	1.0954	1.0775	1.0638	1.0564	1.2413	1.1096
15	JR-WEST	1.0760	1.0845	1.0884	1.0749	1.1079	1.0982	1.0888	1.0781	1.0550	1.0499	1.0570	1.0691	1.0583	1.0579	1.0447	1.1337	1.0764
16	NANKAI	1.0465	1.0709	1.0743	1.0747	1.1196	1.0982	1.0979	1.0884	1.0527	1.0492	1.0479	1.0607	1.0609	1.0869	1.0989	1.2808	1.0880
17	KINTETSU	1.0723	1.0598	1.1335	1.1257	1.1500	1.1456	1.1340	1.1117	1.0772	1.0688	1.0718	1.0910	1.0798	1.0812	1.0873	1.2979	1.1117
18	KEIHAN	1.0381	1.0414	1.0442	1.0435	1.0518	1.0534	1.0500	1.0467	1.0314	1.0282	1.0285	1.0313	1.0291	1.0360	1.0278	1.0745	1.0410
19	KEIFUKU	1.0369	1.0358	1.0476	1.0348	1.0414	1.0384	1.0390	1.0389	1.0346	1.0353	1.0337	1.0483	1.0521	1.0535	1.0566	1.2368	1.0540
20	HANKYU	1.0304	1.0356	1.0357	1.0307	1.0416	1.0406	1.0364	1.0321	1.0225	1.0212	1.0214	1.0238	1.0239	1.0290	1.0325	1.1303	1.0367
21	HANSHIN	1.0458	1.0828	1.0763	1.0739	1.1172	1.1207	1.1006	1.0880	1.0612	1.0440	1.0455	1.0478	1.0444	1.0566	1.0345	1.1345	1.0734
22	KOBE	1.0609	1.0741	1.0746	1.0646	1.0831	1.0725	1.0587	1.0622	1.0410	1.0382	1.0420	1.0531	1.0523	1.0572	1.0519	1.1765	1.0664
23	SANYO	1.1793	1.1678	1.1723	1.1409	1.1720	1.1172	1.1068	1.1240	1.0564	1.0538	1.0632	1.0864	1.0699	1.0734	1.0931	1.1961	1.1170
24	HIRODEN	1.5361	1.5372	1.5750	1.5197	1.5978	1.5986	1.5677	1.5770	1.5630	1.5448	1.5907	1.6176	1.5878	1.6036	1.5201	2.1363	1.6046
25	JR-KYUSHU	1.0822	1.0500	1.0783	1.0647	1.0902	1.0875	1.0631	1.0809	1.0515	1.0482	1.1075	1.0366	1.0338	1.0303	1.0308	1.1050	1.0650
26	NISHITETSU	1.0493	1.0551	1.0956	1.0958	1.1058	1.1165	1.1045	1.0905	1.0606	1.0608	1.0649	1.0821	1.0728	1.0597	1.0495	1.1813	1.0841
	Average	1.0753	1.0787	1.0852	1.0746	1.0953	1.0901	1.0816	1.0750	1.0590	1.0555	1.0633	1.0719	1.0664	1.0667	1.0682	1.2094	1.0823

Note: Bold italics indicate the most efficient railway company.

presents the cost inefficiencies of railway companies measured based on the estimated stochastic cost frontier functions.

Bold italics indicate the most cost-efficient (lowest cost inefficiency) railway company in each year and the average one. KEIO presents the lowest cost inefficiency during 2005–2008 and 2010–2018, SOTETSU in 2009, and TOKYU in 2019 and 2020. On average, from 2005 to 2020, KEIO had the lowest cost inefficiency (1.019), indicating there is still room for a 1.9% cost reduction compared with the least inefficiency with a unity measure. In terms of annual averages for all companies, the lowest cost inefficiency was recorded in 2014 at 1.0555, and the most inefficient year with the largest cost inefficiency was 2020 with 1.2094. In addition, cost inefficiency significantly worsened by 0.1412 points before the COVID-19 pandemic (1.0682) in 2019 compared with that during the pandemic (1.2094) in 2020.

The second highest cost inefficiency was 1.0953 in 2009, a year affected by the financial crisis. In 2011, the cost inefficiency was 1.0816, which is better than the 2010 value of 1.0901; thus, this can be seen as a relatively mild impact of the Great East Japan Earthquake. This is because the areas directly affected by the Great East Japan Earthquake were limited to the TOHOKU region and JR-EAST was the only railway company affected. In other words, the earthquake did not have a significant impact on the operations of the other railway companies.

In Table 9, many KANTO region railway companies, as classified in

Table 10. Highest and lowest cost inefficiencies for each company from 2005 to 2020.

No.	Company	Region	“FY 2005–FY 2019”				FY 2020	Average Value for “FY 2005–FY2020”	Order
			Lowest (Best) Cost Inefficiency	Fiscal Year	Highest (Worst) Cost Inefficiency	Fiscal Year	Highest Cost Inefficiency During Whole Period
1	JR-EAST	TOHOKU KANTO CHUBU	1.0373	(2013)	1.0968	(2006)	1.1949	1.0646	14
2	KEIO	KANTO	1.0138	(2014)	1.0251	(2009)	1.0434	1.0190	1
3	KEISEI	KANTO	1.0317	(2008)	1.0527	(2010)	1.2047	1.0515	12
4	SHIN-KEISEI	KANTO	1.0181	(2014)	1.0365	(2009)	1.0834	1.0299	7
5	TOBU	KANTO	1.0202	(2014)	1.0360	(2007)	1.0880	1.0334	8
6	SEIBU	KANTO	1.0161	(2013)	1.0318	(2009)	1.0732	1.0262	2
7	TOKYU	KANTO	1.0192	(2014)	1.0403	(2009)	1.0386	1.0272	3
8	KEIKYU	KANTO	1.0190	(2014)	1.0327	(2009)	1.0964	1.0289	5
9	ODAKYU	KANTO	1.0165	(2014)	1.0374	(2009)	1.0762	1.0278	4
10	SOTETSU	KANTO	1.0135	(2009)	1.0339	(2007)	1.1153	1.0298	6
11	CHICHIBU	KANTO	1.0258	(2014)	1.0540	(2005)	1.1493	1.0454	11
12	FUJIKYU	CHUBU	1.0479	(2013)	1.2375	(2019)	1.5977	1.1397	25
13	JR-CENTRAL	KANTO CHUBU GHUKYO KANSAI	1.0371	(2018)	1.1284	(2009)	1.3425	1.0875	20
14	MEITETSU	CHUKYO	1.0564	(2019)	1.1543	(2009)	1.2413	1.1096	22
15	JR-WEST	HOKURIKU KANSAI CHUGOKU KYUSHU	1.0447	(2019)	1.1079	(2009)	1.1337	1.0764	18
16	NANKAI	KANSAI	1.0479	(2015)	1.1196	(2009)	1.2808	1.0880	21
17	KINTETSU	KANSAI	1.0598	(2006)	1.1500	(2009)	1.2979	1.1117	23
18	KEIHAN	KANSAI	1.0278	(2019)	1.0534	(2010)	1.0745	1.0410	10
19	KEIFUKU	KANSAI HOKURIKU	1.0337	(2015)	1.0476	(2007)	1.2368	1.0540	13
20	HANKYU	KANSAI	1.0212	(2014)	1.0416	(2009)	1.1303	1.0367	9
21	HANSHIN	KANSAI	1.0345	(2019)	1.1207	(2010)	1.1345	1.0734	17
22	KOBE	KANSAI	1.0382	(2014)	1.0831	(2009)	1.1765	1.0664	16
23	SANYO	KANSAI	1.0538	(2014)	1.1723	(2007)	1.1961	1.1170	24
24	HIRODEN	CHUGOKU	1.5201	(2019)	1.5978	(2009)	2.1363	1.6046	26
25	JR-KYUSHU	KYUSHU	1.0303	(2018)	1.0902	(2009)	1.1050	1.0650	15
26	NISHITETSU	KYUSHU	1.0495	(2019)	1.1165	(2010)	1.1813	1.0841	19
	Average		1.0513		1.1038		1.2088	1.0823

, ranked higher in terms of cost efficiency. This is consistent with the hypothesis that the densely populated areas of the KANTO region, particularly Tokyo, have high railway usage rates. Specifically, KEIO ranks best in terms of cost efficiency (lowest cost inefficiency) because its operating area covers only Tokyo, which provides the benefit of an efficient railway business in the most densely populated area. However, in Table 9, the reason for the lower cost efficiency (higher cost inefficiency) of the four JR companies can be attributed to the fact that these companies also own local lines in rural areas, which makes them less efficient than OPR companies operating in the three metropolitan areas.

Next, we examine the major external factors (financial crisis, Great East Japan Earthquake, and COVID-19 pandemic) that greatly influenced the business environment and cost inefficiency of railway companies. Table 10 presents the lowest and highest cost inefficiency measures (the best and worst efficiency) and the corresponding years for each company from 2005 to 2019; the highest cost inefficiency (worst) during the entire period (2005 to 2020) is concentrated in 2020, and the average cost inefficiencies for the entire period (2005 to 2020) for each railway company are also shown.

Furthermore,

Table 11. Distribution of the number of railway companies from 2005 to 2020.

	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
The lowest (best) inefficiency value	0	1	0	1	1	0	0	0	3	11	3	0	0	2	5	0
The second highest (worst) inefficiency value	1	1	5	0	14	4	0	0	0	0	0	0	0	0	1	0
The highest (worst) inefficiency value	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	26

shows the number of companies in each year that recorded the lowest and the first and second highest cost inefficiencies (the best and the first and second worst efficiency) from 2005 to 2020.

Table 11 presents two findings. First, eleven companies recorded the lowest cost inefficiency (best efficiency) in 2014 and five companies did so in 2019. This can be partly explained by the influence of fare revisions following the consumption tax increase from 5% to 8% in 2014 and the additional increase to 10% in 2019. Second, 14 companies recorded the second highest cost inefficiency (second worst efficiency) in 2009. This finding suggests that cost efficiency deteriorated because of the financial crisis. The highest cost inefficiency (worst efficiency) for all companies is concentrated in 2020, suggesting that the COVID-19 pandemic has influenced the cost inefficiency of railway companies more severely compared with the financial crisis.

4.3. Comparison of Cost Inefficiency Before and during the COVID-19 Pandemic

Figure 3 shows the cost inefficiency of the 26 companies before the COVID-19 pandemic (averages from 2005 to 2019) and during it (2020).

From the figure, the value of cost inefficiency is higher for all railway companies in 2020, although the level of the difference varies by company. Here, 12: FUJIKYU and 24: HIRODEN had higher values than the other companies. Because the service areas of these companies are outside the three metropolitan areas, they were mainly used by tourists of the World Heritage Sites: Mt. Fuji for FUJIKYU and Miyajima for HIRODEN. Due to the COVID-19 pandemic, the number of inbound tourists decreased sharply due to entry restrictions into Japan. These companies were greatly affected by the decline in the number of passengers and their cost inefficiency increased.

Next, we identify the characteristics of best-practice railway companies.

Table 12. Comparison of profiles for the best-practice and other railway companies.

	Average of Best Practice Railway Companies			Average of the Other Railway Companies
	Before COVID-19 Pandemic in 2019	During COVID-19 Pandemic in 2020 (Year on Year Rate)	Growth Rate	Before COVID-19 Pandemic in 2019	During COVID-19 Pandemic in 2020 (Year on Year Rate)	Growth Rate
Operating kilometers (km)	133.6	133.6	100.0%	1140	1134	99.4%
Number of stations	83	84	100.1%	306	304	99.4%
Number of rolling stocks	1028	1022	99.5%	1984	1984	100.0%
Number of transported passengers (thousand passengers)	597,630	421,460	70.5%	710,490	505,988	71.2%
Passenger-kilometers (million passenger-kilometers)	7541	5092	67.5%	18,624	10,660	57.2%
Car kilometer (thousand kilometers)	136,728	137,251	100.4%	371,850	362,859	97.6%
Number of employees in the railway sector	2520	2535	100.6%	7204	7140	99.1%

Note: The growth rate is calculated from 2019 to 2020. Data source: “Annual Report on Railway Statistics” (2019, 2020) from the Ministry of Land, Infrastructure, Transport and Tourism, Japan.

presents a comparison of the profiles of the top 10 railway companies (KEIO, SEIBU, TOKYU, ODAKYU, KEIKYU, SOTETSU, SHIN-KEISI, TOBU, HANKYU, and KEIHAN) and of others before and during the pandemic. We call the top 10 railway companies the best-practice railway companies. The profiles include operating kilometers, number of stations, number of rolling stocks, number of transported passengers, passenger kilometers, car-kilometers, and number of employees in the railway sector.

Here, we consider 2019 as the year before the COVID-19 pandemic and 2020 as the year during the COVID-19 pandemic. We found that the other railway companies had higher values than the best-practice railway companies for all variables. The significant difference between the best-practice railway companies and the other ones is observed in passenger-kilometers, which decreased by 67.5% in 2020 compared with 2019 for the former railway companies and 57.2% for the latter ones. Passenger-kilometers refer to the total length of a passenger’s trip by train. This indicates that passengers made fewer long-distance trips because they refrained from inter-prefectural travel during the pandemic.

We then analyzed the changes in cost inefficiency due to the COVID-19 pandemic.

Table 13. Difference between the cost inefficiency of each railway company in 2019 and 2020.

No.	Company	The Difference in Cost Inefficiency between FY 2020 and FY 2019	No	Company	The Difference in Cost Inefficiency between FY 2020 and FY 2019	No	Company	The Difference in Cost Inefficiency between FY 2020 and FY 2019
1	JR EAST	0.1445	11	CHICHIBU	0.1128	21	HANSHIN	0.1000
2	KEIO	0.0275	12	FUJIKYU	0.3758	22	KOBE	0.1246
3	KEISEI	0.1621	13	JR CENTRAL	0.2934	23	SANYO	0.1030
4	SHIN-KEISEI	0.0583	14	MEITETSU	0.1849	24	HIRODEN	0.6162
5	TOBU	0.0556	15	JR WEST	0.0890	25	JR KYUSHU	0.0742
6	SEIBU	0.0521	16	NANKAI	0.1819	26	NISHITETSU	0.1318
7	TOKYU	0.0245	17	KINTETSU	0.2106		Average	0.1398
8	KEIKYU	0.0722	18	KEIHAN	0.0467		Minimum	0.0181
9	ODAKYU	0.0550	19	KEIFUKU	0.1802		Maximum	0.6088
10	SOTETSU	0.0953	20	HANKYU	0.0978		Standard deviation	0.1240

shows the difference in the cost inefficiency of each railway company between 2019 and 2020, corresponding to the first research concern, that is, “How much did the cost inefficiency of railway companies decrease during the COVID-19 pandemic?”.

Table 13 presents the deterioration in cost inefficiency (increase in values), where the minimum value was 0.0245 for TOKYU, the maximum value was 0.6162 for HIRODEN, the mean value was 0.1398, and the standard deviation was 0.1240. These results reveal the changes in the cost inefficiency of Japanese listed railway companies before and during the COVID-19 pandemic and the scale of the impact on railway companies’ efficiency.

A comparison was then made between the cost inefficiency values of Kuramoto and Hirota (2008), on which this study is based, and the efficiency values calculated in this study. In the study by Kuramoto and Hirota (2008), the analysis focuses on third-sector railways. Third-sector railways are railway companies jointly owned by the public sector and private companies. As such, they do not include the 26 listed railway companies covered by this study. For this reason, a comparison was made with the profiles of the railway companies analyzed by Kuramoto and Hirota (2008), which are closer to the inefficiency values. The results of the Kuramoto and Hirota (2008) analysis confirm the SEMBOKU: inefficiency value of 1.09, TOYO-KOSOKU: inefficiency value of 1.15, and OSAKA-KOSOKU: inefficiency value of 1.78, which are within the range of inefficiency values in this study. For ease of comparison, a simplified passenger-kilometer (a)/number of employees (b) value was calculated to calculate the output per employee (passenger-kilometers). Kuramoto and Hirota (2008) found that SEMBOKU is the least inefficient third-sector railway company in their study.

Table 14. Comparison of the cost inefficiency and profiles for this study and previous studies (Kuramoto and Hirota (2008).

Order	No.	Company	Region	Cost Inefficiency of Average Value for “FY 2005–FY2020”	Operating Kilometers (km)	Number of Stations	Number of Rolling Stocks	Passenger-Kilometers (Million Passenger-Kilometers) (a)	Number of Employees in the Railway Sector (b)	(a)/(b)
19	26	NISHITETSU	KYUSHU	1.0841	106.1	72	311	1574	600	2.62
20	13	JR-CENTRAL	KANTO, CHUBU, GHUKYO, KANSAI	1.0875	1970.8	405	4827	63,427	18282	3.46
21	16	NANKAI	KANSAI	1.0880	154.8	100	696	3922	2195	1.78
-	-	SEMBOKU	KANSAI	1.09※	14.3	6	112	441	258	1.70
22	14	MEITETSU	CHUKYO	1.1096	444.2	275	1070	7260	4085	1.77
23	17	KINTETSU	KANSAI	1.1117	501.1	286	1933	10,590	7226	1.46
24	23	SANYO	KANSAI	1.1170	63.2	49	236	891	715	1.24
25	12	FUJIKYU	CHUBU	1.1397	26.6	18	33	48	254	0.18
-	-	TOYO-KOSOKU	KANTO	1.15※	16.2	9	110	552	304	1.81
26	24	HIRODEN	CHUGOKU	1.6046	35.1	82	300	197	1728	0.11
-	-	OSAKA-KOSOKU	KANSAI	1.78※	28.0	18	88	311	241	1.29

Note: The value of cost inefficiency with ※ of SEMBOKU, TOYO-KOSOKU, and OSAKA-KOSOKU is cited from Kuramoto and Hirota (2008).

shows that SEMBOKU is close to the trend in output per employee in this study, but the other third-sector railway companies do not fit the trend. This is considered to mean that SEMBOKU has achieved efficiency levels close to those of private railway companies, while TOYO-KOSOKU and OSAKA-KOSOKU are not at the level of private railway companies in terms of inefficiency and require further efficiency improvements.

It follows from this that the results of this study direct the analysis of Kuramoto and Hirota (2008) and confirm that the efficiency of private listed railway companies is higher than the efficiency of third-sector railway companies.

4.4. Measures for Efficiency Improvement

Here, we discuss methods to decrease cost inefficiencies during and after the COVID-19 pandemic. We propose the following measures to improve cost inefficiency: For capital cost, companies can divest unneeded assets and business units by reviewing capital investment plans. Fuel costs can be reduced by reducing the number of railway cars while maintaining transportation capacity through efficient train operation. Labor costs can be reduced by reducing the number of staff through automation and mechanization by investing in information and digital technologies. For example, a reduction in the number of ticket counters at stations, a shift to ticketless smartphone applications (apps), online sales of travel products, and driver-only train operations can improve cost efficiency. Through these operations, railway companies need to adjust and elaborate on operating costs C_it and output (passenger kilometers) Q_it to secure profits above the break-even point.

It should be noted that the percentage of local lines was not significant in the cost frontier estimation results in Table 8, which is probably due to the fact that only four JR companies own local lines, meaning the impacts were diluted in the total average. However, issues related to local lines have become more serious, particularly since the COVID-19 pandemic. This is because JR was able to make operational profits in the railway sector through internal mutual aid among service areas, whereby the surplus in metropolitan areas covered the losses on local lines. However, after the pandemic, it became difficult for JR companies to cover the deficit on local lines with profits from metropolitan areas, as the number of passengers in metropolitan areas decreased sharply. If the number of railway users is extremely low and railway operations are inefficient for local lines, it may be effective to switch the mode of transport from transit to buses and tramcars, which can operate more efficiently in relation to the number of users.

Next, we examined the relationship between cost inefficiency and the implementation of efficiency measures for each railway company. This is because it is important to examine what efficiency measures influence costs in companies for improving cost efficiency. For example, Sharma and Chauhan (2022) addressed timetable rationalization in an urban rail transit system, for which we need to consider human intervention to meet rising passenger demand in the form of ridership, and lowering the operational expenditure in terms of energy consumption during train operation. To explore the influence on cost and efficiency measures for Japanese railway companies, we surveyed IR information for each company in 2022 and organized the efficiency improvement measures into five groups (

Table 15. Survey items on the efficiency improvement measures.

Categorization of Efficiency Improvement Measures	Items of Efficiency Improvement Measures
Operation	Autonomous driving Driver-only operation (long trainset: more than three cars) Driver-only operation (short trainset: less than two cars) Station remote control systems (ticket counter•ticket vending machines)
Streamline facilities	Wireless railway car control systems Overhead contact line-less
Ticketing	Prepaid transportation IC cards Mobile prepaid transportation IC cards Smartphone applications Ticketless QR code (payments, tickets) Credit card payments Electric money
Energy-saving	Energy-saving railcars (Variable Voltage Variable Frequency (VVVF) inverter control, regenerative brake) Battery powered railcar Fuel cell railcar Hybrid railcar (combined Internal-combustion engine and energy storage mechanism such as storage battery or flywheel or catenary and battery-powered hybrid railcar) Energy-saving facilities (LED lighting, co-generation system, thermal barrier/insulation system, air conditioning energy-saving operation, energy-saving escalator or elevator, automatic power control by motion sensor, high-efficiency substation facilities, etc.) Renewable energy (solar power, wind power, geothermal power, biomass, hydrogen energy, etc.) Carbon-neutral plan
Peak shift(reduce the number of trains operating during peak periods)	Providing information according to congestion level in real time Commuter pass for off-peak use Fare ticket or coupon ticket for off-peak use Fare of off-peak use Award points for off-peak use (each railway company’s membership point service)

).

Based on Table 15, Figure 4 depicts the relationship between cost inefficiency and the degree of implementation of efficiency improvement measures for each company.

The left vertical axis uses a cumulative bar graph to represent the number of efficiency improvement measures that have already been implemented, those scheduled to be implemented by the end of the fiscal year, and those undergoing demonstration testing. Specifically, the implemented items are given 1 point and 0.5 points for those currently being tested or scheduled for implementation, respectively. The cumulative bar graph comprises efficiency improvement measure categories (i.e., operation, streamlined facilities, ticketing, energy saving, and peak shift), allowing the user to visually grasp the implementation status of efficiency improvement measures for each category.

The vertical axis on the right-hand side shows cost inefficiency. On the horizontal axis, railway companies are listed from left to right in terms of decreasing cost efficiency. The cost inefficiency was averaged for each railway company from FY 2005 to FY 2020.

In Figure 4, the entire group is divided into four subgroups according to cost inefficiencies. Group 1 is defined as a cost inefficiency between 1.00 and higher but less than 1.03, Group 2 as between 1.03 and higher but less than 1.06, Group 3 as between 1.06 and higher but less than 1.09, and Group 4 as 1.09 or more.

Looking at the cost inefficiency and number of efficiency improvement measures taken by each company in a specific group, the dotted lines indicate an approximate trend from left to right for companies with the largest number of efficiency improvement measures to the smallest. This indicates that companies that implement more efficiency improvement measures have higher cost efficiency in each group. In this study, we define the dotted line connecting the values of efficiency improvement measures as the “Trend Line of the implementation of efficiency improvement Measures” (TLM).

In Figure 4, the TLM for each group shows a decreasing trend from lower (left) to higher inefficiency (right), although some companies violate this trend and show much lower degrees of TLM implementation. Specifically, these companies are KEIO, SHIN-KEISEI, JR-KYUSHU, KOBE, HANSHIN, NISHITETSU, JR-CENTRAL, MEITETSU, SANYO, and FUJIKYU.

This indicates that the number of efficiency improvement measures implemented was lower than that of TLM for these companies. This suggests a favorable business environment, in which lower cost inefficiencies have been realized even without implementing many efficiency improvement measures. Therefore, if these companies implement more efficiency improvement measures in the future, their cost inefficiency will further decrease, suggesting they may belong to a higher rank or group. For example, SHIN-KEISEI (7th) and MEITETSU (22nd) could move from Group 2 to Group 1 and from Group 4 to Group 3, respectively, if they implement more efficiency increasing measures.

Furthermore, when we focus on the categories of efficiency improvement measures (operation, streamline facilities, ticketing, energy saving, and peak shift) in Figure 4, almost all listed Japanese railway companies are implementing ticketing and promoting the digitalization of contact with customers, including the introduction of smartphone apps, ticketless service, and a variety of payment methods.

However, the measures that differ in terms of efficiency include operations, peak shifts, energy savings, and streamlined facilities. Companies that implement many of these methods tend to achieve higher efficiencies. For example, the top two companies in Group 3 are JR-EAST and JR-KYUSHU, for which many efficiency improvement measures have been implemented.

Next, we analyze the relationship between digital transformation (DX) investments and cost inefficiency. According to Durand et al.’s (2023) investigation of the current situation and issues regarding the digitization of public transportation from the passenger’s perspective, DX increasingly influences not only the public transport sector but also other sectors in various ways. Since the onset of the COVID-19 pandemic, many companies have been making efforts to transform their internal environments and business models through DX and create new value from them. Therefore, we focus on DX investments, particularly efficiency improvement investments, and examine their relationship with cost inefficiencies.

Specifically, the details of each railway company’s DX investment in 2022 were investigated using the following methodology. (1) From the financial results briefing materials of each railway company, we identified items related to DX among the capital investments planned by each company in 2022. We defined DX investment as items that include the keywords “DX”, “Digital”, “System”, “Electronic”, “Network”, “Database”, “Platform”, “ICT”, “MaaS”, “App”, “Smart”, “Virtual”, “Automation”, “Robot”, “AI”, and “Remote.” (2) We extracted relevant capital investments as DX investments. (3) For railway companies that did not have financial results, briefing materials, materials from general shareholders’ meetings, and press releases of business plans for 2022 were used. The reason financial results briefing materials are used for (1) and (2) is because each railway company’s priority capital investments are clearly described in these materials.

Table 16. DX investment items of Japanese listed railway companies in 2022 (Operation, Streamline facilities).

DX Investment Category				Operation					Streamline Facilities
Order	Group	No.	Company Name	Autonomous Driving•Driver Only Operation	Smart Maintenance	Customer Service Automation	Operational Efficiency by DX	Asset Management	Wireless Railway Car Control Systems
1	1	02	KEIO		○	○		○
2		06	SEIBU			○	○	○	○
3		09	ODAKYU	○	○			○
4		08	KEIKYU			○	○	○
5		05	TOBU	○		○	○
6		07	TOKYU	○	○	○			○
7		10	SOTETSU		○	○	○	○
8	2	04	SHIN-KEISEI
9		20	HANKYU				○
10		11	CHICHIBU
11		18	KEIHAN	○	○	○	○
12		01	JR-EAST	○	○	○	○		○
13		19	KEIFUKU
14		03	KEISEI
15		25	JR-KYUSHU	○	○	○	○	○
16		15	JR-WEST	○	○	○		○	○
17	3	21	HANSHIN				○
18		13	JR-CENTRAL
19		22	KOBE
20		16	NANKAI	○		○
21		26	NISHITESU				○	○
22	4	17	KINTETSU			○	○		○
23		14	MEITETSU	○	○	○		○
24		23	SANYO				○
25		12	FUJIKYU
26		24	HIRODEN				○

,

Table 17. DX investment items of Japanese listed railway companies in 2022 (Ticketing, Customer database platform).

DX Investment Category				Ticketing					Streamline Facilities
Order	Group	No.	Company Name	Prepaid Transportation IC Cards	Cashless	Ticketless•QR Code	Smartphone Payment	Credit Card Touch Payments	Common System Infrastructure Development	Apps ※1	MaaS ※2	Point Service ※3
1	1	02	KEIO									○
2		06	SEIBU			○			○	○	○
3		09	ODAKYU			○	○		○		○	○
4		08	KEIKYU		○				○		○
5		05	TOBU			○				○	○	○
6		07	TOKYU						○	○		○
7		10	SOTETSU		○				○	○	○	○
8	2	04	SHIN-KEISEI							○
9		20	HANKYU			○			○	○	○	○
10		11	CHICHIBU			○				○
11		18	KEIHAN			○			○		○	○
12		01	JR-EAST	○		○			○	○	○	○
13		19	KEIFUKU		○					○
14		03	KEISEI			○
15		25	JR-KYUSHU			○			○	○	○
16		15	JR-WEST	○	○		○		○	○	○	○
17	3	21	HANSHIN			○			○	○	○	○
18		13	JR-CENTRAL			○				○	○	○
19		22	KOBE
20		16	NANKAI					○			○	○
21		26	NISHITETSU			○		○	○	○	○
22	4	17	KINTETSU						○	○	○	○
23		14	MEITETSU							○	○
24		23	SANYO								○
25		12	FUJIKYU						○
26		24	HIRODEN			○	○

※1: Applications (Apps). ※2: Mobility as a Service (MaaS). ※3: Railway loyalty program.

and

Table 18. DX investment items of Japanese listed railway companies in 2022 (Service, New business).

DX Investment Category				Service						New Business
Order	Group	No.	Company Name	Marketing Analysis	Travel Web Reservation	Banking Services	Real-Digital Hybrid	EC Mall	Digital Service ※4	Community Symbiosis Business	Digital Business	Open Innovation
1	1	02	KEIO			○				○		○
2		06	SEIBU	○						○
3		09	ODAKYU	○	○					○
4		08	KEIKYU							○
5		05	TOBU	○				○		○
6		07	TOKYU	○						○		○
7		10	SOTETSU						○
8	2	04	SHIN-KEISEI
9		20	HANKYU	○	○			○
10		11	CHICHIBU
11		18	KEIHAN	○				○		○		○
12		01	JR-EAST	○	○	○	○	○		○		○
13		19	KEIFUKU							○
14		03	KEISEI
15		25	JR-KYUSHU							○
16		15	JR-WEST	○							○	○
17	3	21	HANSHIN	○	○
18		13	JR-CENTRAL		○
19		22	KOBE							○
20		16	NANKAI		○				○	○		○
21		26	NISHITETSU					○	○	○	○
22	4	17	KINTETSU		○		○		○	○	○
23		14	MEITETSU							○	○	○
24		23	SANYO
25		12	FUJIKYU
26		24	HIRODEN

※4: Digital service includes VR shop, Metaverse, Smart logistics.

show the results of the survey of DX investment targets for each railway company in 2022. Measures marked with “○” indicate DX investment targets.

The classification of the DX investment measures is based on Table 15. The three DX investment categories of “Operations”, “Streamlines Facilities”, and “Ticketing” are from the efficiency improvement measures in Table 15. Moreover, three new categories, that is, “Customer database platform”, “Service”, and “New Business” are added which are deeply related to business model reform and establishment. Consequently, there are six categories. The purpose of DX investment is to provide a reasonable classification, considering its characteristics of DX investment, which serve as a foundation for companies to build new business models that adapt to future technological and social changes.

We then analyze the characteristics of DX investment targets for each of the four groups, as shown in Figure 4. Group 1 is characterized by an even distribution of investment in operations (smart maintenance, customer service automation, operational efficiency by DX, Asset management), streamlined facilities, ticketing (cashless, ticket, smartphone payments), customer database platforms (common system infrastructure development, apps, Mobility as a Service (MaaS), point services), services (marketing analysis), and new business (regional symbiosis business, open innovation). The relatively low investment in overall services for Group 1 can be attributed to the fact that these companies have already invested in this area and are focusing on creating new services using customer databases by investing in operational efficiency and marketing.

Compared with Group 1, Group 2 shows no investment in operations (asset management) or ticketing; meanwhile, the emphasis is on investments in ticketless services and e-commerce malls. This suggests that services are still in the process of being digitized for the companies in Group 2.

Compared with Groups 1 and 2, Group 3 shows more investment in ticketing (prepaid transportation IC cards), services (travel web reservation, digital service), and new business (digital business), indicating that they focus on existing services and not on marketing analysis.

Group 4 shows less investment in ticketing, services, and new business than Groups 1, 2, and 3.

Considering that almost all Japanese listed railway companies have invested in customer database platforms, it can be inferred that companies with low-cost inefficiency (high-cost efficiency) have effectively linked their DX investments to earnings. In other words, these companies succeeded in securing profits by creating new services in cooperation with local communities. They thus effectively use their customer databases through proactive DX investments in marketing and asset management.

5. Conclusions

This study applied a stochastic frontier analysis using a Cobb–Douglas cost function to the data on 26 listed Japanese railway companies from 2005 to 2020 to understand their cost inefficiencies before and after the COVID-19 pandemic. We also analyzed the changes in cost inefficiency during the pandemic. We found that the best-practice railway companies are KEIO and TOKYU. The main reason for their good performance is that both operate in and near the Tokyo metropolitan area, which is densely populated. Most Japanese railway companies showed the lowest cost inefficiency (most efficient) in 2014 and the highest in 2020 (least efficient), that is, during the COVID-19 pandemic. The second-worst occurred in 2009, when affected by the financial crisis. However, we did not observe a significant impact of the 2011 Great East Japan Earthquake. This is because no railway company was influenced by the earthquake in this operating area, except for JR-EAST. The deterioration of cost inefficiency due to the pandemic varied depending on the railway company. We found that cost efficient companies succeeded in securing profits through the creation of new services in cooperation with the local community by making effective use of customer databases through proactive DX investments in marketing and asset management.

The results of this research can be used by railway company management and railway policymakers in the formulation of management strategies to improve cost inefficiencies in the event of high-cost inefficiencies due to a deteriorating business environment. This study aimed to provide railway company managers with a quantitative understanding of their company’s management and useful information for developing management strategies to improve efficiency. To this end, the following specific analyses were conducted. (i) Quantified the deterioration in the efficiency of Japanese private railway companies due to the COVID-19 pandemic. (ii) Presented the characteristics of the size and profit status of the best-practice railway companies. (iii) Identified the relationship between cost efficiency and efficiency measures and investments by railway companies.

The limitations of this study are as follows. First, the data used in this study were from railway companies in Japan. It is necessary to use other nations/areas’ data to examine the generality and validity of results of this study. Second, this study should benefit from conducting an extended factor analysis to integrate variables of efficiency measures. Third, the population density data used in this study were at the prefectural level. For a more detailed analysis, it is necessary to use data at the municipal level. Fourth, the efficiency improvement measures and DX investment survey items were obtained from a broader range of available data. More focused and appropriate evaluation items need to be selected depending on the target of the analysis. These are all future tasks of this study.