Estimation of Price Elasticity of Demand for Higher-class Car Travel on Commuter Trains

Abstract

Higher-class Car (seat upgrade) fares on local trains affect not only demand for higher-class travel, but also ordinary class demand on the same trains. It is therefore important to estimate the impact of price changes on the demand for Higher-class Cars, especially during weekdays' commuting time when overcrowding is at a high level. This research focuses on the two-stage pricing structure of Higher-class Car fares where prices jump by 210 JPY when the travel distance exceeds 50 km in the Tokyo area. Then, this research estimates the price elasticity using regression discontinuity design. The estimation results show that the elasticity is significantly larger than one, which means that the price sensitivity to demand is at a high level. Pricing decisions should be made carefully based on the estimation results of price elasticity and current overcrowding levels in Ordinary and Higher-class Cars.

1. Introduction

The spread of Covid-19 has promoted the introduction of new ways of working, such as teleworking and staggered commuting times, and has reduced the number of passengers on local trains in urban areas. On the other hand, West Japan Railway Company (JR West) expanded the routes covered by its Higher-class Car (seat upgrade), “A Seat,” for local trains in the Osaka area in October 2023. In addition, East Japan Railway Company (JR-EAST) plans to expand the routes covered by its Higher-class Car, “Green Car,” to local trains in the Tokyo area in the spring of 2025 [1]. Therefore, the demand for seat upgrades is expected to continue even after the Covid-19 pandemic. Note that “Green Car” is the name for the Higher-class Car used by the JR Group throughout Japan, not only on local trains but also on Shinkansen and other trains, while the “Green Car” in this article refers to the Green Car attached to local trains operated by JR East. Figure 1 shows the exterior and interior of the Green Car and Ordinary Car on the JR-EAST Joban Line.

Fig. 1　JR-EAST Green Car and Ordinary Car, Left: Exterior of Green Car and Ordinary Car, Center: Interior of Green Car, Right: Interior of Ordinary Car (Photographed by the author with prior permission from JR-EAST)

The pricing of seat upgrades is important both in terms of passenger satisfaction and railway operator revenue. For example, if the fare for the seat upgrades is reduced from the current level, passengers in Ordinary Cars will shift to higher-class cars. This in turn will reduce the comfort (in terms of quieter, non-crowded space) of the seat upgrades. On the other hand, if the fare for the seat upgrades increases, more passengers will opt to travel in ordinary class, causing overcrowding in Ordinary Cars. This shows how upgrade comfort levels and the level of crowding in the Ordinary Cars affect each other. Furthermore, changes in railway operator revenue depend on the relative size of the change in fares and the change in demand. It is therefore necessary to quantify the impact on demand of fare changes for seat upgrades in order to set appropriate fares. The impact of such fares on demand can be quantified using the indicator of price elasticity of demand. This index expresses the sensitivity of demand to price changes, and can be interpreted as a value that shows by approximately what percentage demand will decrease (increase) when the price increases (decreases) by 1%.

It is important to consider appropriate pricing based on the price elasticity of demand. However, Green Cars fares, the oldest type of seat upgrade, has been set at the same level as in the past, except for minor adjustments when the consumption tax was increased (in 2019, 2014 and 1997). Therefore, there is no guarantee that the current fare is appropriate. Note that Green Car fares was revised on March 16, 2024, but the goal is to promote a “more comprehensible fare system” and “IC and ticketless” [2].

There is no previous research that has estimated the price elasticity of demand for seat upgrades. There are a few previous studies that have estimated the price elasticity of demand for train fares in Japan [3, 4, 5], etc., but these use aggregated data, using for example, individual train lines. Therefore, all of the previous studies have room for more detailed analysis. Nevertheless, it is difficult to estimate the price elasticity of demand for rail because unlike air fares, rail fares do not fluctuate on a daily basis and are instead largely fixed.

An ideal way to estimate the price elasticity of demand for seat upgrades is to conduct a field experiment in which Green Car fares are randomly changed on a route-by-route or day-to-day basis. Figure 2 illustrates the results of such a field experiment in which more and less expensive fares were randomly changed on a daily basis for each route. The longer the distance traveled by train, the more likely the passenger is to choose a seat upgrade. With two pricing patterns like this, two curves can be obtained for the distance traveled on the train and the probability of choosing the seat upgrade. By quantifying the relationship between the difference in the two types of fares and the difference in the probability of choosing a seat upgrade, the price elasticity of demand can be estimated. However, conducting this type of field experiment is extremely difficult.

Fig. 2　Visualization of the field experiment

This study aims to estimate the price elasticity of demand for seat upgrades by identifying situations as if an experiment had been conducted, without actually conducting a field experiment. This study focuses on morning commuter trains in the Tokyo area, as these are the most crowded. As we will discuss in Section 3.3, it is common in Japan for employers to pay for their employees' commuter passes, since travel allowances are non-taxable up to a certain limit. Therefore, when estimating the price elasticity of demand for Green Cars for commuting, it is not necessary to consider the Regular fare, and I only need to quantify the impact on higher-class travel of changes in Green Car fares.

2. Method

2.1 Method approach

Green Car fares jump at a certain fare-calculation distance boundary. This study analyzes the Green Car fare jump before and after the boundary, considering it as a social experiment that changes the Green Car fare. Figure 3 depicts an illustration of how to estimate the price elasticity of demand for Green Car travel based on the boundary.

Fig. 3　Illustration of how price elasticity of demand for Green Car is estimated

This section explains the approach adopted for analyzing 2015 Green Car fares, because the analysis uses data from 2015. As shown in the upper part of Fig. 3, the first Green Car fare is 770 JPY (7.0 USD; 1 USD = 110 JPY) up to 50 km of the fare-calculation distance, but then jumps to 980 JPY (8.9 USD) at 51 km. This means that the fare jumps by 210 JPY (1.9 USD), even though the distance traveled is almost the same. Therefore, as shown in the lower part of Fig. 3, the probability of choosing Green Car travel should gradually increase as the fare-calculation distance increases up to 50 km, while the probability of choosing Green Car should drop at the 51 km boundary, where Green Car fare jumps by 210 JPY (1.9 USD). By quantifying the relationship between Green Car fares and the probability of choosing Green Cars, I estimate the impact of changes in Green Car fares on demand for Green Cars, i.e., the price elasticity of demand for Green Cars. Note that the fare-calculation distance is defined to the first decimal place for each station, however, unless otherwise specified, the fare-calculation distance in this article refers to the whole number rounded up to the first decimal place. This is because Green Car fares and other train fares are calculated based on the whole number of kilometers rounded up to the first decimal place. It should also be noted that Green Car fares in 2015 differ depending on whether a ticket is purchased on the train or in advance, the advance fare being cheaper. In this study, I assume that commuters purchase their Green Car ticket in advance, so the above shows the advance fares.

This method of quantifying causal relationships using jumps in levels before and after the boundary is known as a regression discontinuity design (hereafter, RDD). As mentioned above, there are very few examples of quantifying the impact of train fares on demand, because of the difficulty of conducting field experiments and the fact that railway fares do not vary from day to day. RDD-based analysis is expected to produce results with the same level of accuracy as field experiments when the conditions described in Section 2.2 are satisfied.

2.2 Overview of RDD and its application to this study

RDD is a method for estimating some effect by focusing on the event that the value of a certain continuous variable z is assigned to a separate group depending on whether it is lower or higher than a specific boundary value, and measuring the jump in the target variable y before and after that boundary. The condition for obtaining valid results using RDD is that no other variables other than the target variable y jump at the above-mentioned boundaries.

In this study, the continuous variable z above is the fare-calculation distance, and the dependent variable y is the probability of choosing a Green Car. By focusing on the fact that Green Car fares is either 770 JPY (7.0 USD) or 980 JPY (8.9 USD) depending on whether the fare-calculation distance is less than or more than 51 km, I estimate the price elasticity of demand for Green Cars by measuring the jump in the probability of choosing y around 51 km. The condition for obtaining a reasonable result here is that the variables other than the probability of choosing Green Cars y (time variables and attribute variables) do not jump around the fare-calculation distance of around 51 km. The advantage of RDD is that it can estimate causal relationships using a simple model with only continuous variables z and the jumping variables as explanatory variables. Note that while the results estimated using RDD provide highly reliable results (internal validity) around the boundary, there is always the question of how applicable they are outside the boundary (external validity).

2.3 RDD model in this study

This study employs a binomial logit model to formulate whether each passenger uses a Green Car or not. The binomial logit model is a statistical model commonly used to formalize choice behavior with two options. Equations (1) and (2) show the RDD model employed in this study.

  

$$y_i^{\text{*}}=\alpha +{\mathit{\beta z}}_i+{\mathit{\gamma d}}_i+{\epsilon }_i$$(1)

  

$$y_i=\{\begin{array}{c}1(\mathrm{if}{y}_i^{\text{*}}>0)\\ 0(\mathrm{if}{y}_i^{\text{*}}\le 0)\end{array}$$(2)

The subscript i indicates an individual, and the objective variable y_i^* is the latent variable (utility) for choosing a Green Car. The explanatory variable z_i indicates the fare-calculation distance (before rounding to the first decimal place) that individual i traveled within the Green Car service area, and di is a dummy variable that takes the values 1 or 0, i.e. 1 when the fare-calculation distance is 51 km or more, and 0 when it is less than 51 km. By assuming that the error term ε_i follows an independent and identical Gumbel distribution, this model becomes a binomial logit model. The constant term is represented by α, while β and γ are parameters indicating the weight of each explanatory variable. As shown in (2), the dependent variable y_i takes the value of 1 or 0, where 1 indicates that individual i uses a Green Car and 0 indicates that it does not. In this study, the γ coefficient for di is the most important parameter because it represents the jump in demand for Green Cars.

The theoretical value of the probability of choosing a Green Car, Pr(y_i = 1), can be calculated with the estimated value of y^* as shown in (3).

  

$$\mathrm{Pr}(y_i=1)=\frac{\exp (\widehat{y^{\text{*}}})}{1+\exp (\widehat{y^{\text{*}}})}$$(3)

The price elasticity η in this study can be expressed as in (4), using (3) and Green Car fares p(d_i) depending on whether the fare-calculation distance is less than or more than 51 km.

  

$$\begin{array}{c}\eta =-\frac{\mathrm{Pr}(y_i=1\mid d_i=1,z_i=50.1)\mathrm{–}\mathrm{Pr}(y_i=1\mid d_i=0,z_i=50.0)}{\{\mathrm{Pr}(y_i=1\mid d_i=1,z_i=50.1)+\mathrm{Pr}(y_i=1\mid d_i=0,z_i=50.0)\}/2}\\ ÷\frac{p(d_i=1)-p(d_i=0)}{\{p(d_i=1)+p(d_i=0)\}/2}\end{array}$$(4)

p(d_i = 0) is 770 JPY (7.0 USD), and p(d_i = 1) is 980 JPY (8.9 USD). The first fraction term in (4) represents the rate of change in the probability of choosing a Green Car, and the second fraction term represents the rate of change in Green Car fares. The denominator used to calculate these rates of change is the midpoint between the pre- and post-change values, following Mankiw (2012) [6]. To calculate the probability of choosing a Green Car, I use the minimum fare-calculation distance (d_i = 1, z_i = 50.1) where Green Car fares is 980 JPY (8.9 USD) and the maximum fare-calculation distance (d_i = 0, z_i = 50.0) where Green Car fares is 770 JPY (7.0 USD), as the basis. Note that the price elasticity η is defined by multiplying by -1, so the estimated value should be positive.

Price elasticity η can be interpreted as the percentage decrease (increase) in demand when the price increases (decreases) by 1% in the vicinity of the fare-calculation distance of 51 km.

As already mentioned in Section 2.2, the condition for accurately estimating the price elasticity is that the variables other than the probability of choosing a Green Car do not jump. I will check this in Section 3.3.

3. Data

3.1 Overview of the Metropolitan Transportation Census

In this study, I use individual cross section data from the Metropolitan Transportation Census for the Tokyo area in 2015 (the latest year for which data are available before the Covid-19 pandemic).

The Metropolitan Transportation Census is conducted every five years by the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) to understand the actual use of railways and buses, and the target areas are the Tokyo, Nagoya and Osaka areas. The Metropolitan Transportation Census is made up of several surveys, but the main data used in this study is the individual cross section data for the Tokyo area of the “Survey of Commuters Using Railway Season Tickets, Ordinary Tickets, etc.” The survey was conducted over three days from Tuesday, November 17th to Thursday, November 19th, 2015.

Passengers who receive the survey form describe and provide details of their first to third railway trips of the day. Here, ‘a trip’ refers to a series of movements between an origin (e.g. home) and a destination (e.g. workplace). The passengers describe the purpose of each trip, the origin and destination points, the lines used, the stations where they get on and off, the type of train used, and the departure and arrival times, etc. They also describe the time they start work and provide personal attributes such as gender and age. Note that respondents select train type from the following four options: each station stop, rapid, extra-fee service, and Shinkansen. ‘Extra-fee’ train service included Green Car travel, express trains, and special commuter liners. Therefore, when respondents selected ‘Extra-fee’ train service, for journeys in sections where several possible fee-paying train services were running side by side, it is necessary to distinguish local train Green Cars from other trains based on information such as the station where the respondent got on and off, and the time of day.

3.2 Analysis target sample

The analysis targets commuters who hold a commuter pass and whose purpose of the trip is commuting, and whose work starts between 8:00 and 10:00.

To formulate the choice between Ordinary Cars and Green Cars, the analysis targets lines where Green Cars are always attached to local trains. As of 2015, there were 9 lines (based on the line classification of the Metropolitan Transportation Census) where Green Cars are always attached to local trains. Basically, passengers need to pay Green Car fares once per train, but they can transfer to another train without paying the fare again as long as they do not exit the ticket gate. For this reason, the fare-calculation distance for each passenger's trip within the Green Car operating area is substituted for the model's explanatory variable z_i.

Even on sections where Green Cars are in operation, passengers who have the potential to bias the results must be excluded from the analysis. Even on sections where Green Cars operate, bias occurs at ODs where limited express trains or similar trains stop. For this reason, ODs where express trains or similar trains stop during the morning commuting hours are excluded from the analysis, based on the boarding and alighting stations within the Green Car operating area.

Table 1 shows the descriptive statistics for the samples analyzed. The fare-calculation distance is between 1 and 70 km, and the sample size is 17,136. Unknown responses are excluded from the calculation of descriptive statistics.

Table1　Descriptive statistics

Variables	Unit	Mean	Standard deviation	Min	Median	Max
zi : Fare-calculation distance	km	25.3	13.7	1.1	24.7	69.7
di : Dummy of 51 km or more	-	5.9%	0.24	0	0	1
yi : Dummy of Green Car choice	-	1.1%	0.10	0	0	1
Work start time	h:mm	8:56	0:25	8:00	9:00	10:00
Departure time	h:mm	7:11	0:42	4:00	7:10	9:53
Boarding time	h:mm	7:28	0:42	4:44	7:28	9:42
Alighting time	h:mm	8:17	0:38	5:25	8:17	10:00
Arrival time	h:mm	8:29	0:37	5:55	8:30	10:00
Female ratio※)	-	29.2%	0.45	0	0	1
Age	year	49.7	10.6	18	50	97

※) Descriptive statistics for a dummy variable that takes the value 1 if the passenger is female and 0 if the passenger is male.

3.3 Checking the conditions for the RDD

As already mentioned in Section 2.1, the condition for the RDD to be established is that the variables other than the probability of choosing the Green Car do not jump around the fare-calculation distance of 51 km. Figure 4 (a) to (f) show the average values and approximate curves (quadratic approximation) by fare-calculation distance for variables other than the probability of choosing a Green Car.

Fig. 4　Average values for time variables and personal attributes by fare-calculation distance

First, I interpret the time variables. The (a) starting time and (b) arrival time tend not to depend on the fare-calculation distance. The (c) departure time and (d) boarding time tend to be earlier as the fare-calculation distance increases. The reason for this is assumed to be that the longer the fare-calculation distance, the longer the commuting time, and so the need to leave earlier and get on earlier.

Secondly, I interpret the individual attribute variables. (e) The female ratio is the average of the female dummy, i.e. the ratio of women by fare-calculation distance. The female ratio tends to be lower for longer fare-calculation distances. The reason for this is assumed to be that women are more likely to be in part-time employment, and therefore have less incentive to spend time commuting than those in full-time employment. (f) Age tends to increase with longer fare-calculation distances. This is assumed to be due to the tendency for people to own houses as they get older. More specifically, they have an incentive to live in the suburbs where land prices are low in order to own a house, and the cost of moving becomes higher once they own a house.

The price of a commuter pass increases by 1 km of fare-calculation distance. However, I suppose that this is not an issue for analysis since the cost of commuter passes is generally paid for by the company in Japan. This is because there is a tax-free allowance of 100,000 JPY (909.1 USD) per month for commuting expenses in Japan (as of 2015). There are Green Passes that allow you to use the Green Car every day, but employers are generally not expected to pay for Green Car fares. Therefore, Green Passes are not taken into account in the analysis. Furthermore, around the fare-calculation distance of 51 km, a Standard Green Car fare is generally cheaper than a Green Pass. There are two types of Green Pass: one-month and three-month. The three-month pass is cheaper per day. Assuming you use the three-month pass for 20 days a month, the fare is 1,564 JPY/day (14.2 USD/day) for a fare-calculation distance of 50 km, and 2,052 JPY/day (18.7 USD/day) for a fare-calculation distance of 51 km (as of 2015) [7] Therefore, there should be no problem with the analysis assuming that the Green Passes are not used very much.

Based on the above considerations, no jumps were observed at a fare-calculation distance of around 51 km. Therefore, the dummy variable parameter γ in the RDD model shown in Section 2.2 is assumed to indicate the change in utility due to changes in the fare of Green Cars.

4. Results of the analysis

4.1 Estimation results for the model parameters and the price elasticity

In this study, I estimate the parameters for three fare-calculation distances: (1) 1 to 70 km, (2) 11 to 70 km, and (3) 21 to 70 km, and calculate the price elasticity based on each of them. The reason for estimating the parameters for several fare-calculation distances is to check the robustness of the results. Note that the glm function (generalized linear model) of the statistical analysis software R (64-bit) version 3.6.1 was used to estimate the parameters using the maximum likelihood method.

Table 2 shows the results of the parameter estimation for the model and the price elasticity of demand for Green Cars. Fig. 5 depicts the relationship between the fare-calculation distance and the probability of choosing a Green Car for each estimated parameter.

Table 2　Estimation results of the parameters and the price elasticity

Parameters	Corresponding terms		(1)			(2)			(3)
α	1	Constant	-6.582	(-26.52)	***	-6.382	(-23.49)	***	-5.886	(-16.91)	***
β	zi	Fare-calculation distance	0.070	(9.57)	***	0.064	(8.09)	***	0.051	(5.18)	***
γ	di	Dummy of 51 km or more	-1.086	(-3.75)	***	-0.970	(-3.25)	***	-0.712	(-2.20)	**
Log-likelihood			-941.2			-923.0			-850.2
McFadden's pseudo-R-squared			0.058			0.041			0.019
Target of fare-calculation distance			1~70 km			11~70 km			21~60 km
Sample size			17,136			14,717			10,811
Pr(yi = 1 | di = 0)		Probability of Green Car choice at di = 0	4.3%			4.1%			3.5%
Pr(yi = 1 | di = 1)		Probability of Green Car choice at di = 1	1.5%			1.6%			1.8%
η		Price elasticity	4.013			3.647			2.761

Fig. 5　Average and theoretical values of the probability of choosing a Green Car, by fare-calculation distance

4.2 Interpretation

In all cases (1) to (3), γ is estimated to be statistically significant and negative, and the price elasticity is greater than 1. Therefore, the demand for Green Cars drops around the fare-calculation distance of 51 km, where Green Car fares jump by 210 JPY (1.9 USD). The price elasticity is also statistically significant and exceeds 1. Price elasticity exceeding 1 means that the rate of change in demand due to a change in price exceeds the rate of change in price, that is, the sensitivity of demand to price is high. Furthermore, a price elasticity value greater than 1 means that a reduction in price will increase demand by more than the rate of reduction in price, i.e. a reduction in price will increase revenue.

From the above, it can be interpreted that the demand for Green Cars is highly sensitive to fares, and that when Green Car fares is reduced, the demand for Green Cars increases by more than the percentage reduction in fares, resulting in an increase in fare revenue. This suggests that changes in Green Car fares can contribute to an increase in revenue for railway operators without large-scale investment in facilities such as quadruple tracks or additional vehicles. However, as already mentioned in Chapter 1, excessively low fares for Green Cars will lead to overcrowding and make it impossible to provide the kind of comfort in terms of space and quietness expected with the upgrade. Therefore, to set appropriate fares for Green Cars, careful consideration based on the estimated price elasticity and a comparison of the current level of congestion in Ordinary Cars and Green Cars is necessary. Note that since these values have been estimated from jumps in price at around 51km in fare-calculation distance, additional research is needed to determine whether the above interpretation can be extended to commuters other than those around 51 km.

The probability of choosing a Green Car at a fare-calculation distance of 51 km (d_i = 1) was estimated to be between 1.5% and 1.8%, a difference of 0.3 points. The probability of choosing a Green Car at a fare-calculation distance of 50 km (d_i = 0) was estimated to be between 3.5% and 4.3%, a difference of 0.8 points, and the range was wider. The price elasticity was estimated to be between 2.761 and 4.013, with a somewhat wider range of 1.252 points. In addition, because the number of samples for fare-calculation distances of 51 km or more is small (5.9% of the total), the average probability of choosing a Green Car by fare-calculation distance is somewhat scattered. In order to solve these issues and obtain more stable estimation results, I need to ensure a sufficient sample size by using data from Metropolitan Transportation Censuses for years other than 2015, as well as data other than Metropolitan Transportation Census data.

5. Conclusions

5.1 Summary of this study

In this study, I have estimated the price elasticity of demand for Green Cars in the weekday morning commute by applying RDD, focusing on the jump in Green Car fares at the fare-calculation distance of 51 km. The analysis result shows that the estimated price elasticity of demand for Green Cars exceeds 1 with statistical significance. This suggests that the demand for Green Cars is highly sensitive to fares, and that a reduction in Green Car fares will increase the demand for Green Cars by more than the percentage reduction in fares and increase the railway operators' revenue. The appropriate price for Green Cars during the weekday morning commute needs to be carefully considered, comparing the current level of overcrowding in Ordinary Cars and Green Cars with the estimated price elasticity. I hope that the results of this research will be used in discussions on the appropriate fare setting for seat upgrades, including Green Cars, and the number of vehicles to be introduced.

In conclusion, although there are still issues to be addressed, such as the fact that the price elasticity was estimated over a somewhat wide range of values, I have established a basic method for estimating the price elasticity of demand for railways by applying RDD to railways for the first time.

5.2 Future work

Due to the limitations of the data sample used in this study, I focused on the weekday morning commute and estimated the price elasticity of demand for Green Cars by combining data from several lines. However, the actual price elasticity is expected to differ depending on the time of day, such as weekdays and holidays, morning and evening commutes, and off-peak hours and also depending on lines. Therefore, future work should involve using data from other years of the Metropolitan Transportation Census and other data on actual usage to eliminate the problem of data constraints and estimate price elasticity that differs by time of day and route. This will make it possible to consider more flexible fare systems, such as a system where the fare is higher during peak hours and lower during off-peak hours, or a system where the fare differs depending on the route.

Furthermore, the goal of future research is not only to estimate the price elasticity of demand for Green Cars, but also to expand it to the construction of a fare setting method for Green Cars that achieves an appropriate level of crowding from the perspective of both passenger satisfaction and operator revenue.

Finally, this paper is a slightly modified version of a paper [8] that was published in The Japanese Journal of Transportation Economics, No. 63.

References

• [1]  East Japan Railway Company, “中央線快速・青梅線でグリーン車サービスを開始します,” JR East News, September 10, 2024 (in Japanese).
• [2]  East Japan Railway Company, “首都圏の普通列車グリーン車の料金体系を見直します,” JR East News, December 15, 2023 (in Japanese).
• [3]  Kaneko, Y., Fukuda, A., Koda, J. and Chiwaki, Y., “Analysis of the Railway Fare Elasticity in the Tokyo Metropolitan Area,” Infrastructure Planning Review, Vol. 21, pp. 175-181, 2004 (in Japanese).
• [4]  Aoki, M., Suda, M. and Hayakawa, S., “Estimation and Comparative Analysis of Passenger Demand between Public-Private Railways and Private Railways in Japan,” The Japanese Journal of Transportation Economics, No. 49, pp. 161-170, 2006 (in Japanese).
• [5]  Fujita, T., “A Study on Railway Demand in Local Areas: Empirical Analysis by Panel Data,” The Japanese Journal of Transportation Economics, No. 62, pp. 45-52, 2019 (in Japanese).
• [6]  Mankiw, N. G., Principles of Economics (6^th Edition), Cengage Learning, 2012.
• [7]  Central Japan Railway Company, “旅客営業規則　別表第2号,” (as of November 22, 2017) (in Japanese).
• [8]  Matsumoto, R., “Estimation of the Price Elasticity of Local Train Green-Car Demand -Based on Regression Discontinuity Design-,” The Japanese Journal of Transportation Economics, No. 63, pp. 71-78, 2020 (in Japanese).

Author

Ryosuke MATSUMOTO Assistant Senior Researcher, Data Analytics Laboratory, Information and Communication Technology Division Research Areas: Transport Economics, Spatial Economics, Applied Econometrics