1 Competition i n long distance transport: Impacts on prices, frequencies, and
2 demand i n the Czech Republic
3 Hana Fitzová, Richard Kališ, Vilém Pařil, Marek Kasa
4
5 Hana Fitzová (corresponding author)
6 Masaryk University, Faculty of Economics and Administration
7 Lipová 41a
8 602 00, Brno
9 Czech Republic
10 Hana.Fitzova@econ.muni.cz, tel.: +420 54949 6904
11
12 Richard Kališ
13 University of Economics in Bratislava, Faculty of National Economy
14 Dolnozemská cesta 1
15 852 35, Bratislava
16 Slovakia
17 Richard.Kalis @ gmail.com
18
19 Vilém Pařil
20 Masaryk University, Faculty of Economics and Administration
21 Lipová 41a
22 602 00, Brno
23 Czech Republic
24 Vilem.Paril@econ.muni.cz
25
26 Marek Kasa
27 Železničná spoločnosť Slovensko, a. s.
28 Pri Bitúnku 2
29 040 01, Košice
30 Slovakia
31 Kasa.Marek@slovakrail.sk
32
1
33 Abstract
34 This article analyses the effect of different entry regulations on company conduct and traveller
35 behaviour. The paper presents a comprehensive case study of three railway markets with sig-
36 nificantly different entry policies using data on prices and frequencies together with a survey
37 conducted to obtain revealed preferences. The study employs data from the three main lines in
38 the Czech Republic. The two open access markets tended to provide significantly higher con-
39 nection frequencies than the line with regulated entry did. Surprisingly, low price variation
40 across the rail and bus markets suggests low monopoly power for the monopolised incumbent
41 and its uniform price strategy across markets with different entry regulations. On the other hand,
42 high price sensitivity among travellers confirms the importance of intramodal competition.
43
44 Keywords
45 Competition, Open access, Price strategy, Frequency, Elasticity
46 1 Introduction
47 Open access competition on railways is gradually becoming more widespread, especially after
48 the Fourth Railway Package of the European railway reforms (European Commission, 2016),
49 mostly in Central Europe. The market structure is slowly changing as is passenger behaviour.
50 This ongoing process of deregulation and market restructuring offers a unique opportunity to
51 compare markets with different levels of transformation. The general effect of competition en-
52 try on price level, quality of provided services, and frequency is unquestionable.
53 However, the railway industry has several specifics. Therefore, it is necessary to be careful
54 about the impacts of competitors on the market and demand. First, both intramodal and inter-
55 modal competition play an important role in transport behaviour. Therefore, regulated entry in
56 the case of one transportation mode can be partially offset by deregulation in other transport
57 modes. This is often the case for intercity bus and railroad competition. Furthermore, vertical
58 integration of railways, together with high fixed costs, can make entries socially undesirable.
59 Finally, from the traveller's perspective, rail services will always remain heterogeneous due to
60 such factors as the importance of departure times. Therefore, some non-zero market power al-
61 ways exists and may be challenging to regulate, if desired. For these reasons and some addi-
62 tional issues, open access for railroads remains the subject of ongoing discussion.
63 The aim of this paper is to analyse the effects of different types of railway competition on both
64 firms' conduct and travellers' behaviour using price and frequency information together with
65 elasticity analysis.
2
66 The Czech transportation market provides a unique opportunity for cross-section comparison
67 due to the variability of competition across different lines. The markets to compare include the
68 following long-distance transport routes: Prague-Brno, Prague-Ostrava, and Brno-Ostrava.
69 The Prague-Ostrava market has been a competitive open-access line since 2011 with the un-
70 precedented competition of three train providers (Czech Railways, RegioJet since 2011, and
71 LeoExpress since 2013). The Prague-Brno market has represented a mixed market since 2016
72 with both the incumbent providing public service obligation (PSO) services and open access
73 competitor RegioJet operating at its own risk (and also providing bus services via the parallel
74 D l motorway in competition mainly with FlixBus). Thus, this route is a case of intense inter-
75 modal coach and rail competition. The last relevant route, Brno-Ostrava, was operated as a
76 PSO by a state-owned company (Czech Railways). In December 2019, however, the incumbent
77 Czech Railways was replaced on the Brno-Ostrava line by RegioJet for the first time in history.
78 This change is unfortunately not captured in our data. Nevertheless, individual transportation
79 remains the only relevant alternative to the train on this route. Therefore, one needs to be careful
80 in any direct comparison of these routes as it lacks public alternatives1
. Moreover, this is the
81 only route in our analysis that is not connecting a city with the capital. This can have some
82 important implications for demand.
83 Regarding individual transportation, the Prague-Ostrava line has a very similar situation as
84 there is no direct motorway connecting the two cities. The competitive intermodal route from
85 Prague to Brno includes a relevant car alternative. According to our survey, however, 92% of
86 both car drivers and their passengers do not use any other transport alternative on these routes.
87 These travellers are thus the least flexible out of all relevant transport modes. Therefore, we
88 included only public transport in our further analysis
89 Table 1 summarises the main attributes of each transport market.
There are some bus services on the Brno-Ostrava line, but they usually have many more stops and a much longer
travel time compared to the train (more than 3 hours by bus compared to a little more than 2 hours by train).
According to our preliminary research based on focus groups, there are very few people using this mode of
transport on this line. Buses are used only for travelling shorter sections, such as to Olomouc.
3
90 Table 1 Markets and competition
Open access PSO/incumbent
Intramodal
competition
Intermodal
competition
Prague-Brno Yes (since 2016) Yes Yes Yes
Prague-Ostrava Yes (since 2011) No Yes No
Brno-Ostrava No Yes No No2
91 We utilised data on prices and frequencies from publicly available timetables and revealed pref-
92 erences obtained from a survey to compare markets with different entry setups and structures.
93 The case study analyses the three main lines in the Czech Republic with different regulatory
94 frameworks. The lines comprise: a monopolised market, a fully open access with three railway
95 competitors, and a mixed market including the incumbent contracted as a PSO and an entrant
96 in business at its own risk. This last market is well-known for its tough intramodal as well as
97 intermodal competition.
98 The rest of the paper is structured in the following way. First, we present a review of the liter-
99 ature on open-access entries and their effects on competition and travel behaviour. Then, we
100 clarify the research question and explain the methodology. Descriptive statistics of the con-
101 sumer survey follow. After that, we present elasticity estimations as a complementary analysis.
102 Finally, we discuss the results, list the research implications, and provide conclusions.
103 2 Literature review
104 Open-access competition on railways means a situation where the market is open, new entrants
105 can enter it, and operations are not subsidised from public resources. Experience with open-
106 access competition in passenger railway transport is quite limited - only a few countries have
107 experienced this type of competition. Head-on entries on principal railway routes can be found
108 only in Austria, the Czech Republic, Italy, Slovakia, and Sweden. Several case studies have
109 analysed free entries on railroads and their effect on the given market. Almost all of these stud-
110 ies have confirmed a positive effect from competition on prices - Cascetta & Coppola (2013),
111 Bergantino et al. (2015), Beria et al. (2016), and Desmaris & Croccolo (2018) for Italy; Tomeš
112 et al. (2016) for the Czech Republic; Kvizda & Solnička (2019) for Slovakia; Tomeš & Jandová
113 (2018) for the Czech Republic, Slovakia, and Austria; and Vigren (2017) for Sweden. Higher
114 service quality and product differentiation are among the other positive effects from competi-
115 tion as documented in all of the aforementioned studies. However, competition is not the only
116 factor determining prices, but also demand, capacity or willingness to pay (Beria & Bertolin,
2
As already mentioned, the relevance of intermodal competition is low.
4
117 2019). Competition also contributed to increased ridership (Fróidh & Nelldal, 2015). Accusa-
118 tions of unfair practices were not rare: price war in the Czech Republic (Tomeš et al., 2016),
119 Slovakia (Kvizda & Solnička, 2019), and Austria (Tomeš & Jandová, 2018) or political action
120 in Poland (Król et al., 2018). On the other hand, Bergantino et al. (2015) do not find evidence
121 of predatory pricing by the incumbent, and Desmaris & Croccolo (2018) show that there is no
122 blatant evidence of anti-competitive behaviour against the new operator in Italy. However,
123 lower prices meant slowly growing revenues, which has caused problems with long-term prof-
124 itability. Pressure on infrastructure capacity and the coexistence of open access and PSOs are
125 other significant problems (Tomeš & Jandová, 2018), especially in the Czech Republic.
126 The findings of these case studies are in line with the modelled situation for a duopoly market
127 in Broman & Eliasson, 2019. These authors found equilibrium with one dominant firm holding
128 more than two thirds of the market. Such asymmetry stems from natural differentiation of com-
129 panies through the heterogeneity of departure times. However, such an outcome is still prefer-
130 able with respect to overall welfare compared to a profit-maximising monopoly. On the other
131 hand, Wheat et al. (2018) found cost disadvantages for firms operating on open-access markets.
132 This stemmed from both comparable costs for franchised operators and the loss the advantage
133 to profit from increasing returns to scale common to monopolised markets. This was partially
134 confirmed in a market with three competitors as described in Tomeš et al. (2016). All three of
135 the operators remained unprofitable after opening competition on the Czech Prague-Ostrava
136 line.
137 In addition to intramodal (railway) competition, intermodal competition between rail and
138 coaches is also worth investigating, especially due to a parallel highway on the Prague-Brno
139 route in the Czech Republic. Fare (Finez, 2014; Paulley et al., 2006), travel time or speed (Beh-
140 rens & Pels, 2012; Fróidh, 2008), comfort (Fróidh & Bystrom, 2013, Allard & Moura, 2018),
141 safety (Si et al., 2009), frequency (Raturi & Verma, 2019, Paulley et al., 2006), income (Toro-
142 Gonzalez, et al., 2020), the opportunity to work (Varghese & Jana, 2018), congestion (Droes &
143 Ritvield, 2015), capacity (Daly et al., 2014), and station availability and parking (Eagling &
144 Ryley, 2015; Pagliara et al., 2012) have been among the most important factors that influence
145 passengers' choices. Yen et al. (2018) mentioned trip characteristics, socio-demographic char-
146 acteristics, frequency, the need for transfers, and easier accessibility. Frequency, transfers, traf-
147 fic congestion, and shortages of parking spaces were discussed by Ben-Akiva & Morikawa
148 (2002). Attitudes and perceptions have also affected the way individuals choose between dif-
149 ferent transport modes, in addition to price and product differentiation (Bahamonde-Birke et
150 al, 2014).
5
151 Rail and bus intermodal competition led to price decrease on routes with intermodal competi-
152 tion compared to monopolistic routes (Gremm, 2018). In Aarhaug et al. (2018), competition
153 from low-cost air carriers was significant for long-distance coach lines, whereas improved road
154 infrastructure and rail services led to increased competition from private cars and rail for shorter
155 coach lines. Moreover, Beria et al. (2018) showed that intermodal competition matters with
156 results that bus routes overlapping with rail PSO are priced less, but interestingly this happens
157 also for high-speed lines. This shows that the two markets are not independent. The level of
158 intermodal long-distance passenger competition in France is high among coaches, BlaBlaCar,
159 highspeed rail, and also low-cost airlines (Crozet & Guihery, 2018). New deregulated bus ser-
160 vices represent only about 2% of long-distance transport. However, intramodal competition is
161 very strong. Burgdorf et al. (2018) analyse long-distance bus services in Germany. They show
162 that price, speed, reliability, convenience, and the carriage of luggage are the most important
163 determinants of modal choice.
164 As it is clear from the rich body of literature, competition in long-distance transport and espe-
165 daily in railway brings some indisputable benefits for passengers and transport system in gen-
166 eral. Most of the aforementioned papers analysed the effects using mostly individual case stud-
167 ies of a single route or by analysing partial characteristics of the examined markets.
168 Our paper contributes to the existing empirical literature on competition and regulatory effects
169 in long-distance passenger transportation by providing a robust analysis of the effects of differ-
170 ent regulatory regimes and competition. We collected a vast amount of data with respect to
171 prices, frequencies, and departure times for three different markets with different regulatory
172 regimes to analyse the firm's conducts. Moreover, we carried out two focus groups and addi-
173 tional surveys to understand passenger's behaviour with respect to varying conditions on ex-
174 amined markets. Furthermore, we used signalling data from mobile operators to verify our sur-
175 vey sample composition.
176
177 3 Research question, data, and methods
178 This section presents the research question and a methodology adequate to answer it.
179 3.1 Research question
180 The regulatory framework represents an important element in long-distance transport markets.
181 Therefore, the corresponding research question is to determine what effects bring open access
182 in railway compared to more traditional PSO services and how different institutional frame-
183 works influence behaviour of intermodal and intramodal competitors together with travellers
6
184 on different markets. To answer this question, we carried out a survey to collect data on pas-
185 sengers' mobility choices. In addition, we gathered data on prices and frequencies and then
186 carried out elasticity analysis using discrete choice models.
187 3.2 Frequency and price data collection
188 The first methodological step was based on frequency and price data collection for the relevant
189 transport markets (the selected three routes) during the same time: from 9 November to 22
190 November 2019. We collected standard ticket purchases without any special tariffs or discounts
191 (from relevant webpages). The data set consists of more than 12,000 bus and train connections
192 and ticket prices on the relevant routes provided by Czech Railways, RegioJet, LeoExpress, and
193 FlixBus.
194 Further, we merged these data with the mobility survey to adjust for possible different impacts
195 from special tariffs (for students, seniors) in comparison with standard tickets. The level of
196 senior and student discounts on public transport tickets is guaranteed at 75% by the Czech Min-
197 istry of Transport (excluding taxis and relatively new car alternatives such as car sharing and
198 carpooling).
199
200 3.3 Passenger survey data collection
201 As a preliminary launching research step before the mobility survey, we conducted two focus
202 groups for the Prague-Brno transport market. We focused on the modal choice between bus
203 and train, including the respondents' relationships or loyalty to a service provider despite var-
204 ying conditions - price, frequency, age category, and other socio-demographic features. These
205 focus groups brought some preliminary results that helped in designing the main mobility sur-
206 vey.
207 The mobility survey aimed to identify the factors determining the use of a particular transport
208 service by inspecting the preferences of passengers. The results of the survey served as feedback
209 on the opinions, attitudes, and reasons on the basis of which passengers "choose" or "do not
210 choose" a specific mode of transport and a specific company. The survey was carried out via
211 systematic sampling, arranging the study population in accordance with selected routes and
212 modes during October and November 2019. Interviewers carried out the data collection via
213 face-to-face paper and pen interviews localised in trains and buses or at train stations, bus stops,
214 or motorway rest areas. In the case of bus and train passengers, there were two phases of field
215 data collection. In the first phase, the form and content of the questionnaire were verified by a
7
216 pilot survey of 50 bus and train passengers. In the second phase, the survey was conducted
217 using an optimised questionnaire.
218 In the case of modal choice focused on car passengers, there were also two phases, but with a
219 slightly different design. The data collection was followed by verification with 10% of car re-
220 spondents. Randomly selected respondents were queried through telephone inquiries and e-mail
221 correspondence on the basis of screening questions to select respondents who had been on a car
222 journey on a relevant route (Prague-Brno, Brno-Ostrava, or Prague-Ostrava) within the previ-
223 ous 14 days.
224 Our research comprises data for three train operators and two bus operators. Table 2 summarises
225 the passenger survey data set, including all complete and relevant interviews. The original sam-
226 pie was even larger with 1887 respondents (see Appendix 1), but due to error answers, the final
227 sample includes only 1521 responses; thus, the error rate was less than 20 % of all sample.
228 Table 2 Number of observations for different modes and operators
Mode Company Prague-Brno Prague-Ostrava Brno-Ostrava
Czech Railways 238 229 131
Train RegioJet
LeoExpress
166 191
108
—
Bus
RegioJet
FlixBus
172
183 : :
Car - - - 103
229 We also conducted a parallel study complementary to our key research in 2019 (pre-covid pe-
230 riod) in accordance with our long-term research goals, which was aimed at mobile operator's
231 signalling big data on population mobility derived from SIM card movements (Ficek, 2019).
232 Based on a very different trajectory of road and railway lines among Prague, Brno, and Ostrava
233 we are able to provide quite precise modal share estimates. These results are furthermore sup-
234 ported by other relevant data about train crews (from Czech Railways company), the capacity
235 of buses (the survey described above in chapter 3.3), from transport Census (Ministry of
236 Transport, 2016) and from road cargo line haul toll monitoring (Ministry of Transport, 2019).
237 Our big data estimates show modal and company transport shares strongly in accordance with
238 our survey sample - see Appendix 2. There are more bus companies operating on relevant mar-
239 kets, but RegioJet and Flixbus represent about 84 % of all provided bus capacities.
240 3.4 Elasticity analysis
241 We used a standardised methodology of discrete choice models with respect to different num-
242 bers of alternatives. All of our models were based on logistic regressions.
8
243 - Brno-Ostrava (BRQ-OSR): binary logistic regression
244 - Prague-Ostrava (PRG-OSR): McFadden's conditional logistic regression
245 - Prague-Brno (PRG-BRQ): nested multinomial logistic regression
246 The monopolised line Brno-Ostrava had the simplest model. The single alternative train (the
247 choice) was predicted relative to individual car transportation, which was the unchosen alter-
248 native in this case. Therefore, the model is closed in the sense of travelling, and we did not
249 consider outside alternatives, comprising people currently not travelling at all. For an introduc-
250 tion to discrete choice models, see Greene (2009), and for an application, see Tomes & Fitzova
251 (2019).
252 The Prague-Ostrava market was modelled with McFadden's conditional logistic regression
253 (McFadden, 1973). The three alternatives (trains) compete within the line; therefore, a binary
254 choice is no longer relevant. There are two options to model such a market. The more common
255 option is multinomial logistic regression, which is focused on the individual unit and uses only
256 the individual's characteristics to explain a choice, and the less common is conditional logistic
257 regression (Hoffman & Duncan, 1988). In the second option, there are two forms of independ-
258 ent variables: alternative specific (varying across and within cases, e.g. price) and case-specific
259 (constant within cases, e.g. travel purpose). Our explanatory variables included both types, and
260 so we used the conditional option.
261 The last market, Brno-Prague, is well known for its intense intramodal competition. The well-
262 known problem of the independence of irrelevant alternatives (see McFadden, 1974) can be
263 solved by grouping similar alternatives into groups or nests. In our case, there are two bus
264 alternatives and two train alternatives. This methodological approach was standardised based
265 on the available literature, such as Koppelman & Bhat (2006) and applications in Forinash &
266 Koppelman (1993) and Polydoropoulou & Ben-Akiva (2001).
267 Finally, the elasticities were calculated the same way across different models. First, the original
268 fitted values and adjusted fitted values were calculated. In the case of price elasticities, all ob-
269 servations were adjusted by increasing price and frequency by 1 %. The individual elasticity for
270 the given mode and company was then calculated by subtracting the original and adjusted fitted
271 value. The market elasticities for transport mode were calculated as the average across individ-
272 ual elasticities.
273 4 Exploratory data analysis
274 The following chapter shows the overall context of the passenger long-distance transport mar-
275 ket in the Czech Republic. It provides an exploratory analysis focused on the frequency of con-
276 nections, prices and our survey sample design.
9
4.1 Long-distance transport markets in the Czech Republic
The research focused on long-distance transport lines connecting three key metropolitan areas
(OECD, 2020) - Prague, Brno, and Ostrava. Brno and Ostrava are centres of the Jihomoravský
and Moravskoslezský NUTS 3 regions (see Figure IChyba! Nenalezen zdroj odkazů.), while
Prague is a NUTS 3 region itself and also the regional centre for the neighbouring Středočeský
region. As we are interested in long-distance travel, we define the entire NUTS3 regions as the
relevant area.
Figure 1 Metropolitan population
Legend 1 Legend 2
Railway corridor Metropolitan population
Sources: OECD, 2020; ArcCRver.3.3, 2020; own processing
Table 3 shows the population, GDP per capita, and the unemployment rate for the metropolitan
areas of interest.
Table 3 Metropolitan stats - population density, GDP per capita, and unemployment rate
Relevant NUTS 3 regions Population density GDP per capita ;EUR) Unemployment rate
2007 2017 change 2007 2017 change 2007 2017 change
Prague and Středočeský 207.06 228.68 +10.4% 580,468 715,428 +23.3% 2.79 3.25 +0.47pp
Jihomoravský 157.82 164.08 +4.0% 337,998 450,135 +33.2% 5.61 5.05 -0.56pp
Moravskoslezský 230.21 222.49 -3.4% 307,168 392,827 +27.9% 7.84 6.42 -1.42pp
290 Sources: OECD, 2020; Pařil & Viturka, 2020, own processing.
291 Table 3 shows significant long-term changes in metropolitan structure in the Czech Republic.
292 The metropolization process in Prague and Středočeský region surrounding the Czech capital
293 is obvious regarding increasing population density in these two regions driven by suburbs in
294 Prague's neighbourhood. Continuing suburbanization and metroplization process occur in Ji-
295 homoravský region surrounding Brno. However, in Moravskoslezský region with the capital of
10
296 Ostrava, the decrease of population density exhibits the long-term depopulation process. Fur-
297 thermore, Prague and its metropolitan area have a significant economic performance with a
298 very low unemployment rate that reflects the hegemon economic position of the Czech capital.
299 4.2 Connection frequency
300 In addition to prices, the frequency of connections is another essential variable influenced by
301 the level of intramodal and intermodal competition as well as the openness of the railway mar-
302 kets. Table 4 provides the average number of connections per day over a week.
303 Table 4 Connections per day
Mode Company Prague-Brno Prague-Ostrava Brno-Ostrava
Bus
FlixBus
RegioJet
18
22
Czech Railways 32 18 16
Train RegioJet
LeoExpress
10 12
7
Total 82 37 16
304 Chyba! Nenalezen zdroj odkazii. shows the saturation of the routes during workdays. Every
305 mark represents departure time of a single connection. The colours distinguish direction within
306 the route. There were clear and significant differences among the markets. On average, there
307 was a train or bus connection every 20 minutes between Brno and Prague and every 32 minutes
308 between Prague and Ostrava but only every 74 minutes between Brno and Ostrava.
11
309 Figure 2 Saturation of routes
•
•
• • A • • • • • • • • • • • • • •
• • 4» • »•••»••• • • • • • • ••• ••
<38> <38> O O <S> O
• •• •
i—
00:00
1
05:30
1
11:00
Departure time
1
16:30
Czech Railways RegioJet train O Leo Express • FlixBus
1 —
22:00
RegioJet bus
310
311 Note: Blue marks represent the directions from Brno to Ostrava, from Prague to Ostrava, and from Prague to
312 Brno. Black marks represent return trips accordingly. Different symbols show different companies within the
313 markets.
314 First, observe the frequency on the Prague-Brno line. The number of current connections
315 reaches the limit of the existing capacity for train connections. There are slightly more train
316 connections within the duopoly Prague-Brno market than there are on the Prague-Ostrava route
317 with three competitors. However, the frequencies for Czech Railways on particular markets
318 differed significantly. There were 18 trains on the Prague-Ostrava route, which is very similar
319 to the monopolised market of Brno-Ostrava, but only roughly half of the capacity of the most
320 competitive market, Prague-Brno.
321 In addition to higher peak frequencies, higher off-peak times frequencies are also connected to
322 a higher level of competition. Services on the Brno-Ostrava line were provided only 16 hours
323 per day during the workday compared to more than 21 hours for the other two analysed markets
324 (or 19 hours for the Prague-Brno line excluding bus alternatives).
325 In summary, competition seems to have played an essential role regarding the frequency in both
326 the number of connections and the proportion of the day serviced by public service operators.
12
327 4.3 Pricing strategy assessment
328 The analysis focused on bus and train tickets provided by the incumbent Czech Railways (CD)
329 and its key competitors RegioJet (RJ), LeoExpress (Leo), and FlixBus (Flix). Figure 3 presents
330 standard ticket pricing box plots for the relevant transport markets, taking into account route,
331 mode, and provider. The lines inside the inner part of the boxes represent the medians, the
332 crosses show the means, and the dots are outliers.
333 Figure 3 Standard (anytime single) prices (euro cents per km)
8
6
5J
2
Prague-Brno Prague-Ostrava Brno-Ostrava
• —
• =
- 1 \- ^ « — 1
1
•
FlixBus RJ bus RJ train CD RJ train CD Leo CD
334
335 The route from Prague to Ostrava exhibited the greatest price volatility. This is quite an im-
336 portant finding because this passenger transport market is the only fully open train market in
337 the Czech Republic, with three competitors - the incumbent Czech Railways and the two other
338 competitors RegioJet and LeoExpress. The price strategy of LeoExpress was the most flexible.
339 LeoExpress prices varied between 1 and 7 euro cents per km while those of the incumbent and
340 RegioJet were mostly 3.5-4 euro cents per km.
341 The lowest volatility was observed on the Prague-Brno route for RegioJet buses and trains and
342 also the incumbents' trains, especially when compared to the much more flexible prices of
343 FlixBus.
344 Figure 4 depicts prices on the relevant transport markets including passenger discounts. These
345 discounts are based primarily on the aforementioned 75% discounts for seniors and students.
346 However, there are many other options provided by particular operators. Czech Railways pro-
347 vided several kinds of loyalty cards: IN25 (25% discounts on each ticket with an initial deposit
13
348 of CZK 450/EUR 17 per year),3
FN50 (50% discounts with an initial deposit of CZK 2,990/EUR
349 112 per year); FN100 (100% discounts with an initial deposit of CZK 19,990/EUR 752 per
350 year); and INbusiness (100% discounts including first-class tickets with an initial deposit of
351 CZK 35,000/EUR 1,316 per year). RegioJet offered credit tickets (with an initial deposit of at
352 least CZK 300/EUR 11 granting discounts on each ticket of around CZK 5/EUR 0.19).
353 Figure 4 Tariff prices including classes and discounts (euro cents per km)
8
6
2
Prague-Brno Prague-Ostrava Brno-Ostrava
i
I
1
X
X
X
X X
X
X
I
;
FlixBus RJ bus RJ train CD RJ train CD Leo CD
354
355 Operators offered different types of services. Czech Railways' fast trains offered second class,
356 first class, and business class. RegioJet offered four types of services: Low cost, Standard, Re-
357 lax, and Business. LeoExpress provided Economy, Business, and Premium tickets (however,
358 premium passengers were very rare as the price was 3 to 5 times as much; therefore, this class
359 was excluded from the sample).
360 The inclusion of tariffs, discounts, and classes had a substantial impact on the minimum prices
361 and the first quartile prices, which were much lower (see Figure 4) than in the previous case.
362 This fact might have been caused by the composition of the population sample (see Figure 7).
363 LeoExpress prices showed the narrowest interquartile range and the highest final ticket price
364 per kilometre bought by passengers (the mean was about 2 euro cents per kilometre higher than
365 those for other modes and providers) as students and seniors represented a minimal share of the
366 passengers. In summary, almost one third of passengers took advantage of 75% discounts. Sim-
367 ilar level of prices has been observed in Slovakia, where seniors and students have been able to
368 travel even with 100% discounts since 2014. Available data from 2018 shows that the share of
3
There were also IN cards for 3 months and 3 years.
14
369 100% discount holders including seniors and students on the passenger rail market in Slovakia
370 was more than 37% of all rail passengers in Slovakia (ZSSK, 2019). This fact highlights the
371 important position of senior and student passengers in long-distance domestic public transport.
372 4.4 Survey passenger characteristics
373 This section describes the main features of the data set regardless of route. First, we analysed
374 the age structure of the travellers, especially the share of students and seniors, which is im-
375 portant due to the 75% discount for these groups. Adult travellers formed 67% of the sample,
376 students 25%, and seniors 8%. However, there were significant differences among the operators
377 - RegioJet attracted significantly more students (31%). The average and median ages were 38
378 and the inner 50% of the travellers were between 26 and 50 years old. Trains were the most
379 common means of transport for seniors, and slightly less so for adults and students.
380 Figure 5 depicts the frequency of travelling for different modes and operators. Most passengers
381 did not travel multiple times per week. Most of the travellers travelled 1-3 times per month or
382 less. The structure of the frequency was similar across various modes and operators. However,
383 occasional travellers (once per year) chose private operators (RegioJet, LeoExpress) more fre-
384 quently than the incumbent, which had the highest share of regular travellers.
385 Figure 5 Passengers' frequency of travelling
TRAIN BUS
1.00-
0.75-
0.50-
0.25-
0.00Frequency
(%)
I 2-3x a week
once a week
I 1 -3x a month
I 2-10 a year
I once a year
CD Leo RJ RJ Flix
386 Figure 6 shows the types of tickets bought. Most passengers used economy class (Economy or
387 Standard in RegioJet and Economy in LeoExpress). The share of economy tickets was highest
388 for the incumbent. RegioJet provided two types of higher comfort tickets (Relax and Business).
15
389 Figure 6 Train ticket types
1.00-
0.75-
0.50-
0.25-
0.00Class
type (%)
I CD Economy
Leo Economy
I RJ Economy
I RJ Standard
I CD First class
I Leo Business
RJ Relax
RJ Business
CD Leo RJ
390 Furthermore, about 24% of Czech Railways travellers used an IN card, which enabled users to
391 buy tickets at a discount (e.g. a yearly IN25 card provides a discount of 25% for CZK 450 per
392 year). Similarly, 23% of RegioJet passengers used RegioJet credit tickets, which also enabled
393 buying cheaper tickets but was free of charge.
394 Figure 7 shows the type of economic activity the passengers of the various transport modes and
395 operators were engaged in. Employees (51%) and businessmen (12%) formed the largest part
396 of the travellers, but their proportions were slightly different for the various modes and opera-
397 tors. Businessmen represented 25% of all LeoExpress passengers. Students represent another
398 large group (24%); they substantially preferred RegioJet.
16
399 Figure 7 Economic activity
TRAIN BUS
1.00- ^^m^^m^^m ^^^m^^^m
Activity type (%)
I Employee (private)
I Employee (public)
I Businessman
I Parental
I Unemployed
Pensioner
Student
400 In summary, the structure of the passengers' economic activity played an essential role. Seniors
401 and students preferred trains, with students preferring the higher quality services provided es-
402 pecially by RegioJet trains.
403 5 Models and elasticity assessment
404 This section presents the results of the model estimations and computed elasticities. In our re-
405 search, the elasticity analysis represents a supplementary means of identifying the relationship
406 between the level of competition on the markets and behaviour using revealed preferences ob-
407 tained through a survey of all analysed markets.
408 5.1 Model estimations
409 Due to different specification of the models, the predictors differed slightly. In general, we
410 controlled for both mode and company-specific variables, including ticket price and frequency.
411 Elasticities were calculated with respect to these variables. The specification of Model 3 did
412 not enable the use of frequency. The existing alternative to Czech Railways between Brno and
413 Ostrava was only individual car transportation, which is not consistent with any frequency in-
414 formation.
415 Second, the socio-demographic characteristics were specific to individuals and did not vary
416 across modes or companies. The variables of travel purpose and passenger travel frequency
417 were used in all three models. In addition, in the Prague-Brno model, the highest completed
0.75-
0.50-
0.25-
0.00CD
Leo RJ RJ Flix
17
418 education was used to better distinguish among the available options.4
Last, the need to change
419 was captured as a dummy variable at the departure, the origin, or both. Detailed variable de-
420 scriptions for all three models can be found in Appendix 3.
421 Table 5 presents estimations for the final models. The interpretation varied slightly across mod-
422 els, and therefore only the sign is interpreted here. Observe that in all three models the unob-
423 served utility (or probability of choice) was lower with increasing price. The impact was sig-
424 nificantly different from zero. Connection frequency was, on the other hand, positively corre-
425 lated with the probability of choice, e.g. the utility of consumption. This parameter was, how-
426 ever, not significantly different from zero in the case of Model 1. This could be interpreted as
427 being related to the already high number of connections per day between Brno and Prague.
428 Model 1 contains both mode-specific as well as mode and company-specific variables. Moreo-
429 ver, the variables are always interpreted in comparison to a benchmark, which is in all three
430 cases the omitted option. For example, the statistically significant variable One_day_travel re-
431 fers to trips shorter than one full day. These consumers had lower utility for using the bus as
432 compared to the train. As can be seen, the constants were significantly lower for both bus op-
433 tions compared to Czech Railways trains. On the other hand, in the case of RegioJet trains, the
434 constant utility was higher but not statistically significant. The constant utility could with some
435 caution be interpreted as unobserved comfort. Lastly, the parameter Tau in the case of the nested
436 version of the discrete choice model captures the dissimilarity between options (companies) in
437 the specified groups (modes). Tau is always lower than one, although for the bus mode it was
438 not significantly different from one, which can be interpreted as showing a high dissimilarity
439 between bus alternatives.
4
This was used mainly due to groups of students using bus and private train companies more often than
others.
18
440 Table 5 Estimation results
Variable Company/mode Model 1 Model 2 Model 3
BRQ-PRG PRG-OSR OSR-BRQ
Ln_price -1.575** -0.565*** —1749***
(0.73) (0.212) (0.435)
Ln_frequency 0.35
(0.38)
1.176***
(0.278)
01d_X_public 0.607***
(0.203)
One_day_travel Bus -0.704***
(0.165)
Origin_change Bus -0.299*
(0.166)
Constant FlixBus bus -1.952**
(0.85)
Constant RegioJet bus -1.851**
(0.836)
Constant RegioJet train 0.249
(0.463)
Tau Bus 0.983
(0.778)
Tau Train 0.603**
(0.239)
Origin_change LeoExpress Train -0.275
(0.368)
Destination_change LeoExpress Train -2.965***
(0.684)
Weekend LeoExpress Train -0.435
(0.329)
Constant LeoExpress Train 0.718
(0.485)
Origin_change RegioJet Train 0.403
(0.376)
Destination_change RegioJet Train -0.833*
(0.427)
Weekend RegioJet Train -0.084
(0.262)
Constant RegioJet Train 0.005
(0.426)
Change CD Train 1.419***
(0.519)
Constant CD Train 1.895**
(0.953)
Number of observations 3,036 1,584 205
Note: standard errors in parentheses; asterisks (***, **, and *) correspond to the significance level (1%,
5%, and 10%, respectively). The number of observations is calculated as the number of passengers from
the survey multiplied by available alternatives, i.e., we have four alternatives for PRG-BRQ and 759
surveyed passengers. For connection OSR-BRQ, we excluded travellers from abroad.
441 Model 2 explains the Prague-Ostrava market. In both cases, the constant was not significantly
442 higher in comparison to the state-owned provider. Except for the variable Destination_change,
443 all of the other predictors were not significantly different from zero. For both private compa-
444 nies, consumers had significantly lower utility when there was a change in destination.
19
445 Again, Model 3 does not show the binary variables for travel purpose and travel frequency. The
446 need to change at the destination or origin had a positive effect on the traveller's utility, which
447 runs contrary to expectations. This can be explained by the poor alternative options. Therefore,
448 the network of the incumbent with integrated regional trains can be fully utilised for fast trains.
449 The predictive power of the models was calculated as the share of the correct choice prediction
450 compared to all predictions for the given mode and company (see Table 6). In other words, the
451 highest probability5
for the chosen alternative was compared to the consumer's actual choice.
452 If the prediction was the same as the real choice, the fit was right, and if it was not the same,
453 the fit was not right. Last, the sum of all correct predictions was divided by the total number of
454 chosen observations for the given mode and company.
455 Table 6 Predictive power of models
Company Mode Model 1 Model 2 Model 3
BRQ-PRG PRG-OSR OSR-BRQ
FlixBus Bus 23.5%
RegioJet Bus 27.9%
Czech Railways Train 62.2%
RegioJet Train 42.2%
LeoExpress Train 21.3%
RegioJet Train 48.7%
Czech Railways Train 68.6%
Czech Railways Train 87.8%
(choice == 0) Other 75.7%
456 Model 1 correctly predicted 62.2% of all choices by travellers that used Czech Railways trains
457 on Prague-Brno connections and 42.2% for those using the second train company, RegioJet.
458 However, the model correctly predicted only 23.5% of FlixBus choices and 27.9% of RegioJet
459 bus choices. Overall, the prediction power for bus modes was lower. Similarly, the fit of Model
460 2 was better for companies with higher market shares, namely Czech Railways and RegioJet
461 trains. However, the fit was rather poor for LeoExpress trains. The best performance was pro-
462 vided by Model 3, which was able to correctly predict 87.8% of all choices for Czech Railways
463 trains.6
464 5.2 Elasticities with respect to price and frequency
465 The coefficients for prices and frequencies from Table 5 provide little explanatory value due to
466 unobserved and individual-specific utility. Therefore, the market demand elasticities for travel
467 alternatives were estimated.
5
For Model 3 and standard logistic regression, the threshold of 50% was used to distinguish the chosen alternative.
6
In the case of Model 3, McFadden's R2
was also calculated as 0.41.
20
468
469 Table 7 gives the calculated elasticities with respect to price for PRG-BRQ and PRG-OSR
470 routes. Table 8 shows result for the specific long-distance transport market BRQ-OSR.
471 Table 7 Elasticities with respect to price on PRG-BRQ and PRG-OSR markets
BRQ-PRG
Company FlixBus RegioJet bus RegioJet train Czech Railways
FlixBus -1.12 0.47 0.32 0.30
RegioJet bus 0.44 -1.15 0.30 0.28
RegioJet train 0.29 0.29 -1.68 0.74
Czech Railways 0.39 0.39 1.05 -1.33
PRG-OSR
Czech Railways LeoExpress train RegioJet train
Czech Railways -0.28 0.20 0.22
LeoExpress train 0.10 -0.40 0.11
RegioJet train 0.18 0.20 -0.33
472 The interpretation of the calculated elasticities with respect to price is as follows: a 1% increase
473 in price for FlixBus would lead to a 1.12% decrease in FlixBus's share. Furthermore, the 1.12%
474 decrease in FlixBus's share would lead to a 0.44% increase in RegioJet's bus share, a 0.29%
475 increase in RegioJet's train share, and a 0.39% increase in Czech Railways' share.
476 Table 8 Elasticities with respect to price on BRQ-OSR market
BRQ-OSR*
Czech Railways
Czech Railways -0.38
477 *Note: The route BRQ-OSR is a different case as it does create a connection with the capital city, and further,
478 there is no alternative (competition) on this route. Thus, the computed elasticity for this route is not determined
479 for direct comparison with the results from the other two routes; however, it is still a relevant market connecting
480 two important municipalities and, in many ways, comparable to the other two markets. Therefore, we provide an
481 estimation of price elasticity also for this third market.
482 In terms of elasticity analysis, three interesting results are observable. First, the existence of
483 intermodal competition seems to have had a stronger effect on elasticity of demand than another
484 intramodal competitor did. Moreover, the results do not provide any sign of brand loyalty on
485 the part of consumers. Furthermore, the market power of Czech Railways within the Brno-
486 Ostrava connection seems to have been low with respect to price. Even though elasticity anal-
487 ysis is not stand-alone proof of any of these conclusions and each one requires separate inves-
488 tigation, the direct- and cross-price elasticities are essential indicators of the findings.
489 Table 9 presents elasticities with respect to frequency, i.e. percentual changes of connections
490 (or, in other words, capacity). A 1% increase for FlixBus connections on the Prague-Brno line
491 would increase the company's market share by 0.25%. Such an increase would, in term, reduce
492 the market shares of RegioJet buses by 0.10%, RegioJet trains by 0.07%, and Czech Railways
21
493 by 0.09%. In this case, the elasticities for the Prague-Brno market are not statistically different
494 from zero. As mentioned previously, the elasticities for the Brno-Ostrava market were not es-
495 timated due to the model specification.
496 Table 9 Elasticities with respect to frequency
BRQ-PRG
Company FlixBus Bus RegioJet Bus RegioJet train Czech Railways
FlixBus Bus 0.25 -0.10 -0.07 -0.07
RegioJet Bus -0.10 0.26 -0.07 -0.06
RegioJet train -0.07 -0.07 0.37 -0.16
Czech Railways -0.09 -0.09 -0.24 0.30
OSR-PRG
Czech Railways LeoExpress train RegioJet train
Czech Railways 0.58 -0.42 -0.46
LeoExpress train -0.20 0.84 -0.24
RegioJet train -0.38 -0.42 0.69
497 In general, the lower the frequency was, the higher the elasticity of change was. Therefore, even
498 though we cannot estimate elasticity with respect to frequency for the Brno-Ostrava line, the
499 expectation would be findings of a highly elastic market.
500 6 Findings and discussion
501 The results for prices and frequencies for different entry setups provide, at first sight, a surpris-
502 ing contradiction. First, we observe little variability in price across markets with significantly
503 different market structures. On the other hand, the findings for frequencies are in line with
504 expectations.
505 There is almost no geographic price discrimination from the incumbent Czech Railways, i.e. it
506 employs a uniform price strategy across the analysed markets. Together with conditions based
507 on PSOs, this leaves very little space for price manoeuvres by the incumbent on the monopo-
508 lised Brno-Ostrava market. 7
Therefore, we did not observe higher prices (on average) within
509 the monopolised market compared to those on the more competitive lines. Moreover, there is a
510 documented effect in the previous literature of competition on price level through, for example,
511 two observed price wars on the Prague-Brno and Prague-Ostrava lines shortly after open access
512 was introduced.
Connection Brno-Ostrava is provided under a netto contract. This obligation is determined such that it transfers
some risk on the provider since the Ministry of Transport does not fix the prices. The price regulation only
determines maximum prices; these can be increased by 20% for a maximum of 20% of connections per day. This
allows some flexibility for price determination in otherwise very stable price strategies of Czech Railways
company.
22
513 However, the connection frequency is a different story. There was great variation in the number
514 of connections across markets. This is in line with previous findings of equilibria with one
515 dominant firm (most likely the incumbent) and smaller entrants. Not only was the number of
516 connections per day higher in case of open access lines, but also more of the day was served.
517 This suggests that players tended to compete for both peak and off-peak times.
518 The elasticity analysis of market demand both confirmed previous findings and opened new
519 questions. First, the lack of differences in price elasticities between the monopolised market
520 and the market with three railway companies confirmed the results of the price analysis itself
521 (i.e. little price discrimination and overall low market power on the monopolised Brno-Ostrava
522 market).
523 There was, however, a striking difference in elasticities with respect to price between the Pra-
524 gue-Brno and Prague-Ostrava markets. Three possible reasons are discussed here. First, the
525 documented price war on the Prague-Brno line occurred before the survey, which may have
526 contributed to travellers' price sensitivity. Second, the intense competition on this railway line
527 was even intensified by equally tough competition between bus alternatives. Finally, there were,
528 on average, 82 connections between Brno and Prague during a regular workday. This specific
529 connection has very limited market power, which normally stems from departure time differ-
530 entiation. Therefore, market demand was highly elastic to price even for negligible changes.
531 Lastly, the market structure enabled us to test for consumers' brand loyalty. However, we did
532 not find any tendency to prefer a brand alternative to a mode alternative from a competitor. As
533 mentioned before, however, this is based on our supplementary elasticity analysis, which suf-
534 fers from several shortcomings.
535 7 Research implications
536 A cross-section comparison of markets with varying entry regulations and competition did not
537 show variability in price level, but did show a clear positive impact of entry on the connection
538 frequency. The monopolised PSO market did not encourage higher connection frequencies in
539 comparison to the more competitive lines. Moreover, consumers' price sensitivity was higher
540 for relevant intermodal competition. The low level of price variability between markets may
541 have had several causes; we tend to believe that one of the most important was the low monop-
542 oly power of the incumbent Czech Railways. Therefore, the market power of a sole player could
543 be reasonably reduced by combining PSOs and competition on other markets, as in the case of
544 the Brno-Ostrava line.
545 On the other hand, if a high frequency of connections is desired, it has not been achieved with-
546 out intense competition in the Czech Republic. On average, there is a train every 30 minutes on
23
547 competitive lines compared to a gap of more than 70 minutes on the monopolised connection.
548 Last, if possible, not only intramodal alternatives should be encouraged, but intermodality as
549 well. The closeness of substitution can play an important role in maintaining low market power.
550 This was the case for the high elasticity within the Prague-Brno line and therefore the expected
551 low market power for firms. As expected, consumers clearly benefited from competition. If
552 maximising consumer surplus is the aim for regulators and managers, competition or at least
553 some kind of competitive tendering should always be preferred.
554 Several questions remain open and provide space for further research. The effect of free entry
555 on the consumer side of welfare seems to be unquestionable. However, overall welfare is still
556 an issue. Moreover, even on markets with a relatively high number of firms, the firms remained
557 naturally differentiated with some local monopoly power due to departure time heterogeneity.
558 The question arises of how high the real market power of such firms is and what the regulatory
559 policies are on the open-access market.
560 8 Conclusions
561 We have provided a comprehensive analysis of the effects of different entry regulations on
562 competition and travellers' decision processes. To this end, we studied three different Czech
563 lines: Prague-Brno, Prague-Ostrava, and Brno-Ostrava. These differed significantly in both
564 railway entry policies and market structure. Moreover, the intense railway competition on the
565 Prague-Brno market was further intensified by intramodal competition with bus alternatives.
566 Our findings are in line with the existing literature describing a positive effect from railway
567 competition for consumers. However, we could not find significant price level differences
568 across markets with varying entry regulations. We believe that this is an effect of the low mo-
569 nopoly power of the incumbent Czech Railways. It seems that the incumbent's pricing policy
570 does not distinguish between routes. Given the estimated price elasticity, the policy is certainly
571 not profit maximising on these routes. The pricing strategy is based on what is fair rather than
572 what is profitable, and this strategy makes it seem unacceptable to discriminate by route. On
573 the other hand, we identified a clear relationship between entry setup and increased connection
574 frequency. Both findings are further supported by estimated market demand and firm-specific
575 elasticities. In the case of price sensitivity, there seems to have been a significant effect from
576 intramodal competition. Furthermore, intense competition increased the number of connections
577 per day significantly. On the other hand, the connection frequency was much lower on the mo-
578 nopolised PSO market. Even though we were not able to estimate elasticity there directly, we
579 expect sensitivity to market demand with respect to frequency to be reasonably high.
24
580 In summary, the paper contributes to the existing empirical literature on the competition and
581 regulatory effects in long-distance passenger transportation. The paper has the potential to pro-
582 vide new arguments for ongoing policy discussion on trade-offs between open access regime
583 and more traditional regulation on railways. Moreover, by collecting a vast amount of supple-
584 mentary data on prices and frequencies, together with conducted surveys, we fulfil the existing
585 gap in the academic literature on comparisons of markets under different regulatory regimes.
586 Acknowledgements
587 This article is the output of the project "New Mobility - High-Speed Transport Systems and
588 Transport-Related Human Behaviour", Reg. No. CZ.02.1.01/0.0/0.0/16_026/0008430, co-
589 financed by the Operational Programme Research, Development and Education.
590 References
591 Aarhaug, J., Farstad, E., Fearnley, N., & Halse, A. H. (2018). Express coaches: An up-hill battle after
592 liberalization? Research in Transportation Economics, 72 SI, 82-91. https://doi.org/10.1016/j.re-
593 trec.2018.07.031
594 Allard, R. F., & Moura, F. (2018). Effect of transport transfer quality on intercity passenger
595 mode choice. Transportation Research Part A: Policy and Practice, 109, 89-107.
596 https ://doi.org/10.1016/i .tra.2018.01.018
597 ArcCR ver.3.3. (2020) ESRI ARC GIS data set for the Czech Republic. ARCDATA PRAHA: Prague.
598 https://www.arcdata.cz/produkty/geograficka-data/arccr-500 (Accessed: 21 July 2020)
599 Bahamonde-Birke, F., Kunert, U., Link, H., & Ortuzar, J. (2014). Liberalization of the Interurban Coach
600 Market in Germany: Do Attitudes and Perceptions Drive the Choice between Rail and Coach?.
601 https://papers.ssrn.com/sol3/papers.cfm?abstract id=2506615 (Accessed: 21 July 2020)
602 Behrens, C , & Pels, E. (2012). Intermodal competition in the London-Paris passenger market: High-
603 Speed Rail and air transport. Journal of Urban Economics, 71(3), 278-288.
604 https://doi.Org/10.1016/i.iue.2011.12.005
605 Ben-Akiva, M., & Morikawa, T. (2002). Comparing ridership attraction of rail and bus. Transport Pol-
606 icy, 9(2), 107-116. https://doi.org/l0.1016/S0967-070X(02)00009-4
607 Bergantino, A. S., Capozza, C , & Capurso, M. (2015). The impact of open access on intra-and inter-
608 modal rail competition. A national level analysis in Italy. Transport Policy, 39, 77-86.
609 https://doi.Org/10.1016/i.tranpol.2015.01.008
610 Beria, P., Redondi, R., & Malighetti, P. (2016). The effect of open access competition on average rail
611 prices. The case of Milan - Ancona. Journal of rail transport planning and management, 6(3), 271-
612 283. https://doi.Org/10.1016/i.irtpm.2016.09.001
25
613 Beria, P., Nistri, D., & Laurino, A. (2018). Intercity coach liberalisation in Italy: Fares determinants in
614 an evolving market, Research in Transportation Economics, 69, 260-269,
615 https://doi.Org/10.1016/i.retrec.2018.07.029
616 Beria, P., & Bertolin, A. (2019). Evolving long-distance passenger services. Market concentration, fares
617 and specialisation patterns in Italy. Research in Transportation Economics, 74, 77-92.
618 https ://doi.org/10.1016/i .retrec.2019.01.004
619 Broman, E., & Eliasson, J. (2019). Welfare effects of open access competition on railway markets.
620 Transportation Research Part A: Policy and Practice, 129, 72-91.
621 https://doi.Org/10.1016/i.tra.2019.07.005
622 Burgdorf, C , Eisenkopf, A., & Knorr, A. (2018). User acceptance of long distance bus services in Ger-
623 many. Research in Transportation Economics, 69, 270-283. https://doi.org/10.1016/j.re-
624 trec.2018.07.023
625 Cascetta, E., & Coppola, P. (2013). Competition on fast track: an analysis of the first competitive market
626 for HSR services. Procedia - Social and Behavioral Sciences, 111, 176-185.
627 https://doi.Org/10.1016/i.sbspro.2014.01.050
628 Crozet, Y., & Guihery, L. (2018). Deregulation of long distance coach services in France. Research in
629 Transportation Economics, 69, 284-289. https://doi.Org/10.1016/i.retrec.2018.07.021
630 Daly, H. E.; Ramea, K., Chiodi, A., Yeh, S., Gargiulo, M., & Gallachoir, B. O. (2014). Incorporating
631 travel behaviour and travel time into TIMES energy system models. Applied Energy, 135, 429-439.
632 https://doi.Org/10.1016/i.apenergy.2014.08.051
633 Desmaris, C , & Croccolo, F. (2018). The HSR competition in Italy: How are the regulatory design and
634 practices concerned?. Research in Transportation Economics, 69, 290-299.
635 https ://doi.org/10.1016/i .retrec.2018.05.004
636 Droes, M. I., & Rietveld, P. (2015). Rail-based public transport and urban spatial structure: The interplay
637 between network design, congestion and urban form. Transportation Research Part-B Methodogical,
638 81 (Part 2), 421-439. https://doi.Org/10.1016/i.trb.2015.07.004
639 Eagling, J., & Ryley, T. (2015). An investigation into the feasibility of increasing rail use as an alter-
640 native to the car. Transportation Planning and Technology, 38(5), 552-568.
641 https://doi.org/l 0.1080/03081060.2015.1039234
642 European Commission. Mobility and transport - Railway packages [online]. EC, (2016). https://ec.eu-
643 ropa.eu/transport/modes/rail/packages/2013 en (Accessed 21 July 2020)
644 Greene, W. (2009). Discrete choice modeling. Palgrave Handbook ofEconometrics (473-556). London:
645 Palgrave Macmillan.
26
646 Ficek, M. (2020). Handover documentation of Big Data contract research in the framework of project
647 New mobility - high-speed transport systems and transport behavior of the population
648 (CZ.02.1.01/0.0/0.0/16_026/0008430), CE Traffic, Prague, p. 87.
649 Finez, J. (2014). Fare setting by the French National Railway Company (SNCF), a social history of
650 pricing. From uniform fares to yield management (1938-2012). Revue Francaise de sociologie,
651 55(1), 5-39. https://doi.org/10.3917/rfs.551.0005
652 Forinash, C. V., & Koppelman, F. S. (1993). Application and interpretation of nested logit models of
653 intercity mode choice. Transportation Research Record, 1413, 98-106.
654 Froidh, O. (2008). Perspectives for a future high-speed train in the Swedish domestic travel market.
655 Journal of Transport Geography, 16(4), 268-277. https://doi.Org/10.1016/i.itrangeo.2007.09.005
656 Froidh, O., & Bystrom, C. (2013). Competition on the tracks - Passengers' response to deregulation of
657 interregional rail services. Transportation Research Part A: Policy and Practice, 56, 1-10.
658 https://doi.Org/10.1016/i.tra.2013.09.001
659 Froidh, O., & Nelldal, B. L. (2015). The impact of market opening on the supply of interregional train
660 services. Journal of Transport Geography, 46, 189-200.
661 https://doi.Org/10.1016/i.itrangeo.2015.06.017
662 Gremm, C. (2018). The effect of intermodal competition on the pricing behaviour of a railway company:
663 Evidence from the German case. Research in Transportation Economics, 12, 49-64.
664 https://doi.org/l0.1016/i .retrec.2018.11.004
665 Hoffman, S. D., & Duncan, G J. (1988). Multinomial and conditional logit discrete-choice models in
666 demography. Demography, 25(3), 415-427. https://doi.org/10.2307/2061541
667 Koppelman, F. S., & Bhat, C. (2006). A Self Instructing Course in Mode Choice Modeling: Multinomial
668 and Nested Logit Models. Washington, D.C.: Federal Transit Administration.
669 Krol, M., Taczanowski, J., & Kolos, A. (2018). The rise and fall of Interregio. Extensive open-access
670 passenger rail competition in Poland. Research in Transportation Economics, 72, 37-48.
671 https://doi.Org/10.1016/i.retrec.2018.06.008
672 Kvizda, M. & Solnicka, J. (2019). Open access passenger rail competition in Slovakia - Experience
673 from the Bratislava-Kosice line. Journal of Rail Transport Planning & Management, 12, 100143.
674 https://doi.Org/10.1016/i.irtpm.2019.100143
675 McFadden, D. (1973). Conditional logit analysis of qualitative choice behavior. In P. Zarembka (ed.),
676 Frontiers in Econometrics (105-142). New York: Academic Press.
677 McFadden, D. (1974). The measurement of urban travel demand." Journal of Public Economics, 3(4),
678 303-328. https://doi.org/10.1016/0047-2727(74)90003-6
679 Ministry of Transport (2016). Transport Census 2016. Available online: https://www.rsd.cz/wps/por-
680 tal/web/Silnice-a-dalnice/Scitani-dopravy
27
681 Ministry of Transport (2019). Highway Toll Data 2019.
682 OECD (2020). OECD.stat Metropolitan areas - Metropolitan statistics 2001-2018.
683 https://stats.oecd.org/Index.aspx?DataSetCode=CITIES (Accessed 2 July 2020).
684 Pagliara, F., Manuel Vassallo, J., & Roman, C. (2012). High-speed rail versus air transportation: Case
685 study of Madrid-Barcelona, Spain. Transportation Research Record, 2289(1), 10-17.
686 https://doi.org/10.3141/2289-02
687 Pafil, V., & Viturka, M. (2020). Assessment of priorities of construction of high-speed rail in the Czech
688 Republic in terms of impacts on internal and external integration. Review ofEconomic Perspectives,
689 20(2), 217-241. https://doi.org/10.2478/revecp-2020-0010
690 Paulley, N., Balcombe, R., Mackett, R., Titheridge, H., Preston, J., Wardman, M.,... & White, P. (2006).
691 The demand for public transport: The effects of fares, quality of service, income and car ownership.
692 Transport policy, 13(4), 295-306. https://doi.Org/10.1016/i.tranpol.2005.12.004
693 Polydoropoulou, A., & Ben-Akiva, M. (2001). Combined revealed and stated preference nested logit
694 access and mode choice model for multiple mass transit technologies. Transportation Research Rec-
695 ord, 1771(1), 38-45. https://doi.org/10.3141/1771-05
696 Raturi, V., & Verma, A. (2019). Competition between high speed rail and conventional transport modes:
697 Market entry game analysis on Indian corridors. Networks & Spatial Economics, 19(3), 763-790.
698 https://doi.org/l0.1080/03081060.2020.1701666.
699 Si, B., Zhong, M., & Gao, Z. (2009). Bilevel programming for evaluating revenue strategy of railway
700 passenger transport under multimodal market competition. Transportation Research Record,
701 2117(1), 1-6. https://doi.org/10.3141/2117-01
702 Tomes, Z., Kvizda, M., Jandova, M., & Rederer, V. (2016). Open access passenger rail competition in
703 the Czech Republic. Transport policy, 47, 203-211. https://doi.Org/10.1016/i.tranpol.2016.02.003
704 Tomes, Z., & Jandova, M. (2018). Open access passenger rail services in Central Europe. Research in
705 Transportation Economics, 72, 74-81. https://doi.org/10.1016/j.retrec.2018.10.002
706 Tomes, Z., & Fitzova, H. (2019). Does the incumbent have an advantage in open access passenger rail
707 competition? A case study on the Prague-Brno line. Journal of Rail Transport Planning and Man-
708 agement, 12, 100140. https://doi.Org/10.1016/i.irtpm.2019.100140
709 Toro-Gonzalez, D. Cantillo, V., & Cantillo-Garcia, V. (2020). Factors influencing demand for public
710 transport in Colombia. Research in Transportation Business & Management, 36, 100514.
711 https://doi.Org/10.1016/j.rtbm.2020.100514
712 Varghese, V., & Jana, A. (2018). Impact of ICT on multitasking during travel and the value of travel
713 time savings: Empirical evidences from Mumbai, India. Travel Behaviour and Society, 12, 11-22.
714 https://doi.Org/10.1016/i.tbs.2018.03.003
28
715 Vigren, A. (2017). Competition in Swedish passenger railway: Entry in an open access market and its
716 effect on prices. Economics of Transportation, 11-12, 49-59. https://doi.org/10.1016/j.eco-
717 tra.2017.10.005
718 Wheat, P., Smith, A. S., & Rasmussen, T. (2018). Can competition for and in the market co-exist in
719 terms of delivering cost efficient services? Evidence from open access train operators and their fran-
720 chised counterparts in Britain. Transportation Research Part A: Policy and Practice, 113, 114-124.
721 https://doi.Org/10.1016/i.tra.2018.03.004
722 Yen, B. T. H., Mulley, C , & Tseng, W.-C. (2018). Inter-modal competition in an urbanised area: Heavy
723 rail and busways. Research in Transportation Economics, 69, 77-85. https://doi.org/10.1016/j .re-
724 trec.2018.04.007
725 ZSSK (2019). Železničná spoločnosť Slovensko, a. s., department of long-distance and international
726 transport services and products. Internal data on the number of passengers using senior and student
727 discounts.
29
728 Appendix 1 Original passenger survey sample and questionnaire
Czech Railways 312 263 142
RegioJet 202
Bus
FlixBus 200
Car - 129 18 103
729 Note: This table shows original structure ofour passenger survey.
730 Source: passenger survey - see chapter 3.3, authors
731 Questions:
732 - Interviewer:
733 o Mode
734 o Company
735 o Time
736 o Location
737 - Interviewed:
738 o Where do you travel from and where (municipality, region)?
739 o How long are you driving?
740 o Do you use fare type?
741 o What type of fare do you have?
742 o What class are you currently traveling in (train only)?
743 o To what extent do you usually use individual transport modes or carriers when
744 traveling on the selected routes (Prague, Brno, Ostrava)?
745 o To what extent do the following reasons affect your choice of transport mode /
746 specific carrier?
747 • High frequency of connections (car availability)
748 • Possibility to use a customer (discount) card
749 • Price
750 • Speed
751 • Safety of operation (accident)
752 • Personal feeling of safety in the coach/vehicle (attack)
753 • Reliability / Less delayed connections
754 • Comfort (place for feet, etc.)
755 • Level of services (refreshments, Wi-fi, steward, magazines, etc.)
756 • Possibility to work in a vehicle
757 • Other: ...
758 o Gender
759 o Age
760 o Education
761 o Employment
762 o Driver's license holder
763
30
764 Appendix 2 Modal & Intercompany Share
765 Sample of survey modal and company shares
Mode Company Prague-Brno Prague-Ostrava Brno-Ostrava
Czech Railways 31% 43% 56%
Train RegioJet 22% 36% 0%
LeoExpress 0% 20% 0%
Bus
RegioJet 23% 0% 0%
Bus
FlixBus 24% 0% 0%
Car - - - 44%
766 Note: This table shows relative share structure ofour passenger survey.
767 Source: passenger survey - see chapter 3.3, authorsBig data verification modal and company shares
768 based on average modal occupancy
Mode Company Prague-Brno Prague-Ostrava Brno-Ostrava
Train
Czech Railways
RegioJet
LeoExpress
39%
24%
0%
53%
31%
15%
45%
0%
0%
Bus
RegioJet
FlixBus
18%
19%
0%
0%
0%
0%
Car - - - 55%
Note: This table shows relative share structure of big data model on modal split
general loadfactor according to each mode oftransport.
on relevant routes based
Source: SIM card's movement big data (Ficek, 2020)~Big data verification modal and company shai
based on specific company occupancy in rail (CD, RegioJet a LeoExpress)
Mode Company Prague-Brno Prague-Ostrava Brno-Ostrava
Czech Railways 39% 53% 45%
Train RegioJet 24% 31% 0%
LeoExpress 0% 15% 0%
Bus
RegioJet
FlixBus
18%
19%
0%
0%
0%
0%
Car - - - 55%
773 Note: This table shows relative share structure of big data model on modal split on relevant routes based on
11A general loadfactor according to each mode oftransport and each individual company on train market.
775 Source: SIM card's movement big data (Ficek, 2020)
31
776
777
778 Appendix 3 Variable description
Individual Alternative
specific specific Variable description
Ln_price
Ln_frequency
01d_X_public
One_day_travel
Origin_change
Destination_change
Change
Weekend
Travel purpose fixed effects
Education fixed effects
Travel frequency fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Log of price in EUR, adjusted for
Yes discounts
Yes Log of number of connections per day
Interaction variable 1 for people born
before 1977 and using Czech Railways
Yes at the same time, otherwise 0.
1 if travel duration < one day,
No otherwise 0
1 if origin municipality for traveller is
not equal to resident municipality for
No traveller, otherwise 0
1 if destination municipality for
traveller is not equal to resident
No municipality for traveller, otherwise 0
1 if destination or origin municipality
for traveller is not equal to resident
No municipality for traveller, otherwise 0
1 if day of travel is Saturday or
No Sunday, otherwise 0
A) Business trip; B) travel to work; C)
study; D) family, friends; E) tourism -
1 day; F) tourism - overnight; G)
No private affairs; H) other
Elementary, secondary without
qualifications, secondary with
No qualifications, tertiary
4+ times per week, 2-3 times per
week, once per week, 1-3 times per
month, 2-10 times per year, once per
No year
779
32