Planning and Regulation

Route Planning

New Lines
Programme

Stated Preference Survey

Contents
1

BACKGROUND AND OBJECTIVES

Context

Stated Preference Survey

ORGANISATION OF THE SURVEYS

Fieldwork

Questionnaire

SAMPLE PROFILE

Presentation

Current Journey

Socio-Economic Profile

11

Attitude Towards the New Service

12

ANALYSIS OF THE MODE CHOICE SP EXERCISE

15

ANALYSIS OF THE ACCESS AND EGRESS SP EXERCISE

27

FIGURES
Figure 3.1

Journey Duration - Car

Figure 3.2

Mode of Access - Rail

Figure 3.3

Mode of Egress - Rail

Figure 3.4

Mode of Access Air

Figure 3.5

Mode of Egress Air

10

Figure 3.6

Car Availibility by Purpose and Mode

10

Figure 3.7

Gender Profile

11

Figure 3.8

Age Profile

11

Figure 3.9

Income Profile

12

Figure 3.10

Attitude Towards The Service

13

Figure 4.1

Example of Mode Choice SP Card (Air and Rail Users)

15

Figure 4.2

Example of Mode Choice SP Card (Car Users)

15

Figure 4.3

Multinomial Models

20

Figure 4.4

Single-Level Nested Model

21

Figure 4.5

Double-Level Nested Model

21

Figure 4.6

Classic and High-Speed Grouped Under Rail: Nested Model

21

Figure 5.1

Example of Access SP Choice

27

TABLES
Table 2.1

Stated Preference Exercises Seen per Respondent

Table 3.1

Number of Respondents by Journey Purpose

Table 3.2

Average One-Way Cost

Table 4.1

Rail And Air Respondents Choice Proportions: Varying Classic Rail InVehicle Time

16

Rail And Air Respondents Choice Proportions: Varying High Speed Rail
In-Vehicle Time

17

Table 4.3

Rail and Air Respondents Choice Proportions: Varying Classic Rail Cost

18

Table 4.4

Rail And Air Choice Proportions: Varying High Speed Rail Cost

18

Table 4.5

Car Respondents Choice Proportions: Varying Classic Rail IVT

19

Table 4.6

Car Respondents Choice Proportions: Varying High Speed Rail IVT

19

Table 4.7

Car Respondents Choice Proportions: Varying Classic Rail Cost

19

Table 4.8

Car Respondents Choice Proportions: Varying High Speed Rail Cost

19

Table 4.9

Trading Analysis Mode Choice SP

20

Table 4.10

Preferred Model - Model 10 Air-rail Sample Nested Structure

23

Table 4.11

Preferred Model - Model 12 - Car Sample Nested Structure

24

Table 5.1

Choice Proportions: Difference in Cost All Users

27

Table 5.2

Choice Proportions: Difference in Access Time - All Users

28

Table 5.3

Trading analysis - Access and Egress SP

28

Table 5.4

Rail Users Business Travellers

28

Table 5.5

Rail Users - Leisure Travellers

29

Table 5.6

Air Users - Business Travellers

29

Table 5.7

Air Users - Leisure Travellers

30

Table 5.8

Car Users - Business Travellers

30

Table 5.9

Car Users - Leisure Travellers

30

Table 4.2

APPENDICES
MAIN MODEL CALIBRATION
ADDITIONAL MODEL RESULTS
RAIL USER QUESTIONNAIRE - A
RAIL USER QUESTIONNAIRE - B
AIR USER QUESTIONNAIRE
CAR USER QUESTIONNAIRE AND SP CARDS

New Lines Programme: Stated Preference Survey

Background and Objectives

Context

1.1

Network Rail has commissioned Steer Davies Gleave to investigate whether there is
a strategic business case for building one or more new line(s) as additions to the
national rail network.

1.2

The study geography is bound by the five classic line route corridors that radiate in
an arc to the west and north of London. These are the:
I

Great Western Main Line (GWML)

Chiltern (Marylebone to Birmingham via High Wycombe)

West Coast Main Line (WCML)

Midland Main Line (MML)

East Coast Main Line (ECML)

1.3

The structured programme of work culminated in a recommendation as to whether

to progress to a more detailed investigation of one or more new lines as a next
phase of work.

1.4

The programme found that there is a case to take a new high speed line forward for
further investigation and that the preferred corridor for such a line would be the
London to North West & Scotland corridor.

Stated Preference Survey

1.5

As part of the New Lines project, Steer Davies Gleave has been commissioned to
undertake Stated Preference surveys to re-estimate relative importance given by
passengers to various characteristics of their journey such as in-vehicle time,
headway, mode and time or mode to access the rail station for the new service.

1.6

Initial weights for these parameters had been estimated as part of the research
undertaken in 2002 for the development of the Planet Strategic Model (PSM). The
aim of these surveys was to provide more up to date data to confirm, and
potentially re-estimate these parameters.

1.7

Stated Preference surveys were conducted for air, rail and car users between
November 2008 and February 2009. This report presents the surveys and their key
outputs. Chapter 2 presents how the surveys were organised, where and when they
took place. Chapter 3 details the analysis of the sample and finally Chapter 4
present the updated parameters from the Stated Preference exercises and compares
them with PSM parameters.

New Lines Programme: Stated Preference Survey

Organisation of the Surveys

2.1

The purpose of Stated Preference (SP) Studies is to assess how respondents choices
vary in different hypothetical situations. SP is a survey technique concerned with
measuring and understanding the preferences underlying peoples stated choices,
based on how they respond to hypothetical situations. In SP research respondents
are presented with choices between hypothetical but realistic alternatives, with
each alternative being described in terms of their characteristics or attributes.
By varying the values of these attributes in a carefully controlled way (using an
experimental design with appropriate statistical properties), we can learn about
how much importance people attach to each attribute on the basis of the choices
they make. It is also possible to estimate choice probabilities for given choice
alternatives similar to those studied.

2.2

The SP survey collected stated choice data that will be used to estimate discrete
choice models of transport choice and will form a key input to the demand model.
People eligible for the survey undertook a recent journey either by car, air, or rail
of at least two hours between any pair cities on each of these 3 routes:

2.3

London and Manchester (business and leisure trips); and

London and Glasgow (business and leisure trips).

People travelling by coach are out of the scope of this study.

Fieldwork
Pilot
2.4

The pilot phase was split into 2 phases: the first part, conducted on Thursday 13th
November and Friday 14th, surveyed rail and car users and the second part,
undertaken at the end of November on Thursday 27th and Friday 28th interviewed
air users. A second pilot took place on January, 15th at Manchester airport.

2.5

Rail interviews took place on-board Virgin trains between London and Manchester
and questionnaires were self-completed. Car interviews were undertaken at the
service station next to Northampton on the M1 and were face to face interviews.
The pilot phase of air interviews took place at Gatwick rail station and consisted of
face-to-face interviews.

2.6

Respondents for each mode answered a paper questionnaire that had been tailored
to each mode, although broadly similar.

2.7

Each respondent were shown two Stated Preference exercises. The first SP exercise
was a three way choice between 3 long-distance modes. For the second SP exercise,
all three answered the same access/egress exercise.

New Lines Programme: Stated Preference Survey

2.8

The following summarises each of the three interview types:

TABLE 2.1

STATED PREFERENCE EXERCISES SEEN PER RESPONDENT

SP1

Current main
mode

Mode choice

Classic Rail HSR

- Air

SP2

SP3

Mode choice

Access mode
choice

Classic Rail HSR

- Car

Rail PT - taxi

Rail

Air

Car

Main
2.9

The main phase was based on the pilot questionnaire. There were no changes to
either of the SP exercises however there were minor changes to the wording and
options for some of the extra questions. These changes were made to increase the
clarity and simplicity of the questionnaire.

2.10

The main phase was split into 2 phases:

I

Rail: 5th 12th December and 13th - 17th January 2009;

Car: 5th 12th December and 5th 25th January 2009; and

Air: the air fieldwork took place between Monday 19th January and Wednesday
4th February.

2.11

Rail and car interviews were undertaken in the same way. Air interviews were
carried out at Manchester airport: the interviews were undertaken airside.

2.12

The fieldwork was initially planned to take place late in the autumn with all modes
being interviewed at the same time. Unfortunately obtaining authorisations to
survey at the airports proved to be very difficult. BAA would not accept surveys at
any of their airports (Gatwick, Heathrow, Glasgow or Edinburgh).

2.13

An initial pilot phase was undertaken at Gatwick Rail station this would have been
combined with a programme of interviewing passengers accessing the airport by car.
In the meantime, the discussions with Manchester Airport were more fruitful and
eventually authorisations were granted to interview airside in December 2008.

2.14

It was decided to delay the fieldwork in order to interview airside rather than
conducting the interview at Gatwick rail station and Gatwick car park.
Unfortunately, this meant that the subsequent SP analysis was delayed, with results
becoming available in early March. However, this did provide sufficient time for the
core result (New lines modal preference) to be incorporated into the forecasts and
strategic business case analysis.

New Lines Programme: Stated Preference Survey

Questionnaire
2.15

The first part of the questionnaire asks the respondents to describe their current
journey and it includes questions on the following:
I

Journeys origin and destination;

Journey Purpose;

Journey duration;

Access and egress mode and cost;

Whether the respondent was travelling with luggage;

Whether the respondent was travelling with children;

Type of ticket the respondent was travelling on; and

Who was paying for the respondents journey.

2.16

The second part of the questionnaire describes the New Lines scheme briefly and
asks whether the respondent would choose the new service or not, and which
factors would influence their decision.

2.17

The third part of the questionnaire is the Stated Preference (SP) exercise, in which
the respondent is shown cards for the mode-choice and access-egress SP and is
asked to make decisions on varying attributes and their levels.

2.18

The final part of the survey asks the respondent whether there was a car available
for the journey, the frequency of similar journeys, and respondents socio-economic
profile. The survey signs off with a request to the respondent to give further
comments if they wish to.

New Lines Programme: Stated Preference Survey

Sample Profile
Presentation

3.1

The number of total respondents with their journey purpose and the number of
respondents excluded for every mode are given in table Table 3.1.
TABLE 3.1

NUMBER OF RESPONDENTS BY JOURNEY PURPOSE

Rail

Air

Car

Business

246

203

162

Leisure

207

48

191

Total (before
exclusion)

453

251

353

Excluded

Current Journey
3.2

Respondents were asked to describe their current journey or a recent journey that
they have undertaken in the corridor within the past month. The following
paragraphs detail how these journeys vary by mode and purpose.
Cost of the journey

3.3

Table 3.2 presents one-way costs. This does not include access costs to the station
or airports. Single way air or rail ticket could go up to 200 for business travellers.
TABLE 3.2

AVERAGE ONE-WAY COST

Car
Rail

Air

(excluding
parking costs)

Business

67

82

36

Leisure

27

48

38

All purposes

49

77

37

Journey duration
3.4

Figure 3.1 depicts the journey times in minutes for car distributed with respect to
the journey purpose. For car, most people had journeys between 4 hours and 4.5
hours.

New Lines Programme: Stated Preference Survey

FIGURE 3.1

JOURNEY DURATION - CAR

80%
60%

40

40%

Number of respondents

80

20%
0

0%
2h - 2h30 2h30 - 3h 3h - 3h30 3h30 - 4h 4h-4h30

4h30-5h

>=5h

Journey Duration

Mode of access and egress

3.5

Figure 3.2, Figure 3.3, Figure 3.4 and Figure 3.5 show the modes of access and
egress for air and rail with respect to their journey purpose. The average access
time for people who walked to the station was 13 minutes, which is reasonable.
Most of the respondents who travelled by rail used Public Transport (local or
regional rail) to access the station.
MODE OF ACCESS - RAIL

90

54%

80

48%

70

42%

60

36%

50

30%

40

24%

30

18%

20

12%

10

6%

business
%

Number of respondents

FIGURE 3.2

leisure
% Business
% Leisure

0%
On foot

Public Transport

Taxi

Car

Mode of access to the station

3.6

Most of the respondents who travelled by rail used taxi or public transport to access
their final destination.

New Lines Programme: Stated Preference Survey

MODE OF EGRESS - RAIL

60

60%

50

50%

40

40%

30

30%

20

20%

10

10%

business
%

Number of respondents

FIGURE 3.3

leisure
% Business
% Leisure

0%
On foot

Public Transport

Taxi

Car

Mode of access to the final destination

Most of the people accessed the airport via public transport. Two respondents said
they walked and indicated either a 5 or 10 minute walk to the airport.

Number of respondents

FIGURE 3.4

MODE OF ACCESS AIR

160

160%

140

140%

120

120%

100

100%

80

80%

60

60%

40

40%

20

20%

business
%

3.7

leisure
% Business
% Leisure

0%
On foot

Public Transport

Taxi

Car

Mode of access to the station

3.8

The modes used to reach the final destination varied in a balanced way between
Public Transport, Taxi and Car. 11 respondents said they walked and indicated
either a 5 or 10 minute walk from the airport.

New Lines Programme: Stated Preference Survey

MODE OF EGRESS AIR

70

70%

60

60%

50

50%

40

40%

30

30%

20

20%

10

10%

business
%

Number of respondents

FIGURE 3.5

leisure
% Business
% Leisure

0%
On foot

Public Transport

Taxi

Car

Mode of acces to the final destination

Journey description
3.9

Most of the air and rail respondents were travelling on their own (2/3 of the rail
respondents, 87% of the air users). Car respondents were more frequently travelling
with other people: a third of them were travelling alone, another third were
travelling with one person only and the remaining ones were travelling with at least
two other people.

3.10

Figure 3.6 shows car availability for respondents. This information is distributed
between Rail and Air with respect to the journey purposes. 15 air users and 90 rail
users did not own a car at all and are included.
CAR AVAILIBILITY BY PURPOSE AND MODE

180
160

45%
40%

140
120

35%
30%

100
80

25%
20%

60
40

15%
10%

20
0

5%
0%
Rail Business

Rail Leisure

Air Business

Mode and purpose

Air Leisure

Yes
%

Number of respondents

FIGURE 3.6

No
% Yes
% No

New Lines Programme: Stated Preference Survey

Socio-Economic Profile
Figure 3.7 shows the gender profile of the users with respect to the modes
travelled.

Number of respondents

FIGURE 3.7

GENDER PROFILE

300

60%

250

50%

200

40%

150

30%

100

20%

50

10%

Male
%

3.11

Female
%Male
%Female

0%
Rail

Air

Car

Mode

3.12

Figure 3.8 shows the age profile of the respondents distributed with respect to the
modes used. For rail, the highest number of respondents belonged to the age group
of 25 to 34 years old whereas for air and car, the most respondents came from the
age group of 35 to 44 years and 45-54 years, respectively.
FIGURE 3.8

AGE PROFILE
90%

140

16-24 years old

80%

120

25-34 years old

35-44 years old

100

45-54 years old

60%

55-64 years old

80

50%

60

40%

Number of respondents

70%

65 or older
16-24 years old %
25-34 years old %

30%
40
20%

35-44 years old %

45-54 years old %
55-64 years old %

20

10%

65 or older %

0%
Rail

Air

Car

Mode

11

New Lines Programme: Stated Preference Survey

3.13

Figure 3.9 shows the income profile of the respondents with respect to the modes
used. The car user category was the one with the largest proportion of people who
did not state their income.
FIGURE 3.9

INCOME PROFILE
70%

250

60%
up to 15,199

200

15,600 - 25,999
26,000 - 31,199

50%

36,400 - 46,799
46,800 - 51,999

150
40%

Over 52,000
%

Number of respondents

31,200 - 36,399

Not Stated
up to 15,199 %

30%
100

15,600 - 25,999 %
26,000 - 31,199 %
31,200 - 36,399 %
36,400 - 46,799 %

20%

46,800 - 51,999 %
Over 52,000 %

50

Not Stated %
10%

0%
Rail

Air

Car

Mode

Attitude Towards the New Service

3.14

Figure 3.10 shows if people were willing to use the new proposed service. This
question can be used to flag out people who are definitely against the service and
can be used in the Stated Preference analysis. As it can be seen, rail users are most
enthusiastic about it.

New Lines Programme: Stated Preference Survey

FIGURE 3.10 ATTITUDE TOWARDS THE SERVICE
250

70%

Yes - regularly
Yes - occasionally

50%

No

150

40%
%

Number of respondents

60%
200

30%

100

Don't know
Yes - regularly %
Yes - occasionally %

20%
50
10%
0

No %
Don't know %

0%
Rail

Air

Car

Mode

13

New Lines Programme: Stated Preference Survey

Analysis of the mode choice SP exercise

4.1

Each respondent saw two Stated Preference exercises: the first one was concerned
with mode choice and offered a choice between:
I

Classic Rail High Speed Rail Air for rail and air users

Classic Rail High Speed Rail Car for car users

4.2

It was considered that showing all four modes in each survey would lead to an overly
complex survey that would be onerous to complete.

4.3

Figure 4.1 and Figure 4.2 present one example of SP choices presented to the
respondent. Full questionnaires are available in Appendices C to F.
FIGURE 4.1

EXAMPLE OF MODE CHOICE SP CARD (AIR AND RAIL USERS)

SP1-2
Existing Rail Service

Service Frequency
Journey Time
One-way Cost

High Speed Rail

Air

every 20 mins

every 20 mins

every 90 mins

3 h 30 mins

1 h 40 mins

2 h 15 mins

80

95

90

Which of these options do you prefer?

SP1-2

FIGURE 4.2

EXAMPLE OF MODE CHOICE SP CARD (CAR USERS)

SP2-2
Existing Rail Service

Service Frequency
Journey Time
One-way Cost

High Speed Rail

Car

every 20 mins

every 60 mins

3 h 55 min

2 h 30 min

4 h 30 min

65

80

40

Which of these options do you prefer?

4.4

The second exercise dealt with access mode to the rail station, assuming that the
respondent would use the new rail service and valuation of access time.

15

New Lines Programme: Stated Preference Survey

4.5

This chapter presents the results of the mode choice exercise, and their application
in the demand forecasting framework. In the following chapter, the key findings
from the access SP exercise are presented.
Choice proportions

4.6

The choice proportions are useful to understand if the respondents understood the
Stated Preference exercises and reacted to the proposed trade-offs in a sensible
way, i.e. the higher cost a service is, the fewer people chose it.

4.7

The following tables show the percentage of people choosing each option when a
change in journey time and cost occurs and thus the sensitivities to these changes.
People were sensitive to changes in in-vehicle time or one-way cost but not to
changes in headway which is hence not mentioned here. The level of headway
presented to the respondent was between 20 and 60 minutes and was thus covering
a wide range. Only rail and high-speed rail headway levels were varying. Air
headway level was assumed to be constant to 1h30 minutes.

4.8

Each table presents the percentage of respondents choosing one of the three
options when either classic rail in-vehicle time varies or when high speed rail invehicle time varies. Table 4.1 and Table 4.2 show the choice proportions when
either Classic Rail in-vehicle time varies or High Speed Rail in-vehicle time varies.

TABLE 4.1

RAIL AND AIR RESPONDENTS CHOICE PROPORTIONS: VARYING CLASSIC RAIL

IN-VEHICLE TIME
Rail users
%
choosing
Classic
Rail

%
choosing
High
Speed
Rail

%
choosing
Air

%
choosing
Classic
Rail

%
choosing
High
Speed
Rail

%
choosing
Air

150

50%

46%

4%

12%

51%

37%

210

25%

66%

9%

5%

54%

41%

IVT
(mins)

4.9

Air users

As the journey time of classic rail increase from 150 to 210 minutes, the percentage
of people choosing classic rail is halved and is distributed across high speed rail and
air. For a 20 minute change in the journey time of high speed rail, the same
phenomenon occurs but not as severe as classic rail.

New Lines Programme: Stated Preference Survey

4.10

As the journey time of classic rail increase from 150 to 210 minutes, the percentage
of people choosing classic rail is halved and is distributed across high speed rail and
air. For a 20 minute change in the journey time of high speed rail, the same
phenomenon occurs but not as severe as classic rail.

TABLE 4.2

RAIL AND AIR RESPONDENTS CHOICE PROPORTIONS: VARYING HIGH SPEED

RAIL IN-VEHICLE TIME
Rail users

4.11

Air users

IVT
(mins)

%
choosing
Classic
Rail

%
choosing
High
Speed
Rail

%
choosing
Air

%
choosing
Classic
Rail

%
choosing
High
Speed
Rail

%
choosing
Air

100

30%

65%

5%

6%

59%

35%

120

44%

48%

8%

10%

46%

44%

Again it is seen from Table 4.3 and Table 4.4 that as the cost increases for classic
rail or high speed rail, respondents chose to move to different modes. If the cost
difference is low the trading is also lower, whereas if the cost difference is high
e.g., from 95 to 130 for high speed rail, a higher level of trading occurs between
high speed rail and other modes.

17

New Lines Programme: Stated Preference Survey

TABLE 4.3

RAIL AND AIR RESPONDENTS CHOICE PROPORTIONS: VARYING CLASSIC RAIL

COST
Rail users

% choosing
Air

%
choosing
Classic
Rail

%
choosing
High
Speed
Rail

%
choosing
Air

46%

5%

14%

50%

36%

39%

54%

7%

9%

51%

40%

30%

63%

7%

5%

54%

41%

% choosing
Classic
Rail

% choosing
High
Speed Rail

60

50%

70
80

Cost
()

TABLE 4.4

Air users

RAIL AND AIR CHOICE PROPORTIONS: VARYING HIGH SPEED RAIL COST
Rail users

Air users
%
choosing
High
Speed
Rail

%
choosing
Air

% choosing
Classic
Rail

% choosing
High
Speed Rail

% choosing
Air

%
choosing
Classic
Rail

90

26%

64%

10%

6%

52%

42%

95

30%

66%

3%

7%

57%

57%

130

64%

28%

8%

14%

43%

43%

Cost
()

4.12

For air users, when the cost for classic rail increases to 70, 9% of the people said
they would use classic rail (see Table 4.3) whereas 51% said they would use high
speed rail. However, when the price rises to 80, only 4% choose to use classic rail
whereas the percentage of respondents choosing high speed rail goes up to 54%.
People who chose air are more or less insensitive to price or IVT changes in classic
rail.

4.13

Table 4.5, Table 4.6, Table 4.7 and Table 4.8 show the choice proportions for car
users. They illustrate the fact that for car users, when the cost increases from 50
to 60 for people who chose classic rail, the percentage of people who said they
would choose classic rail falls from 20% to 12%, while the percentage of people who
chose high speed rail rose from 26% to 32%.

New Lines Programme: Stated Preference Survey

TABLE 4.5

CAR RESPONDENTS CHOICE PROPORTIONS: VARYING CLASSIC RAIL IVT

% choosing Classic
Rail

% choosing High
Speed Rail

% choosing Car

175

11%

30%

59%

180

21%

25%

53%

210

12%

34%

54%

235

9%

37%

54%

IVT (minutes)

TABLE 4.6

CAR RESPONDENTS CHOICE PROPORTIONS: VARYING HIGH SPEED RAIL IVT

% choosing Classic
Rail

% choosing High
Speed Rail

% choosing Car

120

11%

30%

59%

150

14%

32%

54%

IVT (minutes)

TABLE 4.7

CAR RESPONDENTS CHOICE PROPORTIONS: VARYING CLASSIC RAIL COST

% choosing Classic
Rail

% choosing High
Speed Rail

% choosing Car

50

20%

26%

53%

60

12%

32%

57%

65

11%

35%

55%

Cost ()

TABLE 4.8

CAR RESPONDENTS CHOICE PROPORTIONS: VARYING HIGH SPEED RAIL COST

% choosing Classic
Rail

% choosing High
Speed Rail

% choosing Car

70

10%

43%

47%

80

17%

26%

57%

90

16%

16%

69%

Cost ()

Trading analysis
4.14

A respondent is said to have traded if he chose at least once two different modes
and that he did not always choose his current or preferred mode. The following
table shows the trading proportions with respect to journey purpose.

19

New Lines Programme: Stated Preference Survey

4.15

For example, almost 84% of leisure travellers on rail always traded. Trading level for
car users is lower than for the other modes: this was expected since people
travelling by car are usually doing so either for practical purposes or because they
like it and are more reticent to switch to other modes.

4.16

It is important to note that the people who did not trade either always chose their
current mode or another one. For instance, amongst the air users who did not trade,
half of them always chose high speed rail and half of them always chose air. This is
a sign that the Stated Preference exercise is not too biased towards one mode.
TABLE 4.9

TRADING ANALYSIS

Mode

MODE CHOICE SP
Business

Leisure

Rail

86%

84%

Air

65%

72%

Car

58%

50%

Model calibration
4.17

Several models were tested which included both Multinomial and Nested logit
models.

4.18

The multinomial structure illustrated in Figure 4.3 assumes that the three modes
are independent. Two different models were calibrated, one including air and rail
respondents offered a choice between classic rail, high-speed rail and air, the other
offered a choice between two rail options and car
FIGURE 4.3

Classic Rail

4.19

MULTINOMIAL MODELS

High
Speed
Rail

Air

Classic Rail

High
Speed
Rail

Car

The nested logit models fell into three main categories. First, a single level nested
logit model (see Figure 4.4) where classic and high-speed rails were distinguished
between and grouped under one category namely, rail. This model assumes that the
two rail options are not independent and share common characteristics. A similar
but separate model was calibrated for car users, distinguishing car from the rail
nest. Our preferred models come from this structure. The nest parameter for car
was significantly different than 1. For the air-rail choice, the parameter was quite
close to 1, which indicates that the air option is quite similar to the rail ones for this
mode choice exercise.

New Lines Programme: Stated Preference Survey

FIGURE 4.4

SINGLE-LEVEL NESTED MODEL

Car

Air
Classic Rail
4.20

Classic Rail

High Speed Rail

Second, a double-level nested logit model (shown in Figure 4.5) whereby car and
public transport were differentiated with air and rail coming under the public
transport branch; additionally, rail was further divided into a nest with high-speed
and classic rail acting as two different nodes.
FIGURE 4.5

DOUBLE-LEVEL NESTED MODEL

Air

Classic Rail

4.21

High Speed Rail

Car

High Speed Rail

Third, a model whereby distinction between classic and high-speed rail was not
made and both were grouped under Rail (shown in Figure 4.6).
FIGURE 4.6

CLASSIC AND HIGH-SPEED GROUPED UNDER RAIL: NESTED MODEL

Car
Rail

4.22

Air

A generic example of the utility function for classic rail is as follows:

Classic Rail = BETA_COST * cost_classic_rail + BETA_HWAY * headway_classic_rail +
BETA_JT * time_classic_rail + mode_constant

21

New Lines Programme: Stated Preference Survey

4.23

The above example shows that the utility function consists of the cost, the invehicle-time (IVT) and the headway. Various models were tested either by adding an
alternative specific constant and having the same BETA_COST, BETA_HWAY and
BETA_JT for all modes or by having the relative weight on in-vehicle time vary per
mode and not having a mode constant. This latter is equivalent to assuming that the
mode constant is not a one-off value but that the impact of the mode is continuous
as the journey time varies.
Model results

4.24

A large number of models were tested: Table 4.10 and Table 4.11 give the results
from the two preferred models i.e. models 10 and 12 for the air-rail and car samples
respectively. These are the final results that were used for further analysis. Both
are nested structures. It should be noted here that if the nest parameter is close to
1 then the nest is not very strong. The figures in italics indicate insignificant values.

4.25

The value of travel time savings for classic rail users travelling on business purpose
that is very close to the PSM value. It should be noted that the value of travel time
savings on high speed rail is lower than the one for classic rail, which was expected.
It is lower than the PSM value to which a large mode constant was associated. In the
absence of any mode constant, the value of time for High Speed Rail is smaller.

New Lines Programme: Stated Preference Survey

TABLE 4.10 PREFERRED MODEL - MODEL 10 AIR-RAIL SAMPLE
NESTED STRUCTURE

A
Cl

Hi

Relative
value

PSM Values
(inflated to 2008)

-8.47

39.6/h

40/h

-0.016

-7.46

25.2/h

11/h

High Speed Rail IVT

(Business)

-0.019

-7.09

30.2/h

51/h

High Speed Rail IVT

(Leisure)

-0.0145

-5.68

23.3/h

14/h

Air IVT (Business)

-0.0225

-10.83

36.1/h

51/h

Air IVT (Leisure)

-0.022

-10.68

35.3/h

14/h

Parameter

Value

T-stat

Cost

-0.037

-8.71

Classic Rail IVT

(Business)

-0.025

Classic Rail IVT

(Leisure)

Headway (Business)

Headway (Leisure)

Nest parameter

-0.012

-6.4

-0.003

-1.61

1.19

8.36

19.3/h

5.3/h

22/h
for
Classic
Rail
and
HSR

8/h
for CR

12/h
for
Air

2/h
for
HSR
and
Air

23

New Lines Programme: Stated Preference Survey

TABLE 4.11 PREFERRED MODEL - MODEL 12 - CAR SAMPLE
NESTED STRUCTURE

A
Cl

Hi

Relative value

PSM Values
(inflated to 2008)

-5.43

24/h

40/h

-0.007

-4.37

20

11/h

High Speed Rail IVT

(Business)

-0.008

-3.59

19/h

51/h

High Speed Rail IVT

(Leisure)

-0.006

-2.89

15/h

14/h

Car IVT (Business)

-0.009

-6.54

23/h

40/h

Car IVT (Leisure)

-0.004

-3.15

11/h

14/h

Parameter

Value

T-stat

Cost

-0.0242

-9.7

Classic Rail IVT

(Business)

-0.0095

Classic Rail IVT

(Leisure)

Headway (Business)

Headway (Leisure)

Nest parameter

-0.005

-3.82

-0.002

-1.76

3.39

7.5

12/h

22/h
for
Classic
Rail
and
HSR

12/h
for
Air

6/h

8/h
for
Classic
Rail

2/h
for
HSR
and
Air

4.26

For the air and rail sample, (see Table 4.10) the relative values for IVT for business
travellers on Classic Rail calibrated on the collected sample are of the same order
or magnitude as the PSM parameters for business people (30/h to 40/h compared
to the PSM values (40/h to 50/h). However the leisure parameters are
significantly higher (20/h - 35/h) than the PSM values (11/h to 14/h).

4.27

It is known that a Stated Preference design is tailored to capture some value of time
better than others, i.e. the ones in the core part of the range of values of time
underlying each SP card. In this particular case, our SP design could have captured
values as low as the PSM leisure value of time. The fact that the leisure values of
time are higher than expected is thus not a side effect of our design.

New Lines Programme: Stated Preference Survey

4.28

These issues were investigated by calibrating different models that are presented in
the appendices of this document, but this did not improve on the model calibration.

4.29

For the car sample, the value of time for High Speed Rail leisure travellers, is very
similar to the inflated PSM values. The valuation of travel time savings by rail for
car users is lower than the one for rail users (Business travellers: 24/h instead of
39/h, leisure travellers: 20/h instead of 25).

4.30

From these preferred models it was noted that the value of time for leisure was
higher than expected, which led us to test several more models. These are set out
in Appendix A.
Use of Stated Preference analysis in demand modelling

4.31

4.32

As discussed in Chapter 2, the results of the preferred model (cf. Table 4.10) were
used to derive values of time for high-speed rail segmented by purpose, relative to
classic rail in-vehicle time. Revealed Preference calibration was not updated from
the PSM research and thus the nest parameters and classic rail, air and car values of
time were kept unchanged from the PSM research. However, the relative values
employed were:
I

a business high speed rail value of time being 25% lower than classic rail value of
time;

a leisure high speed rail value of time being 8% lower than classic rail; and

commuter values were unchanged from PSM.

More detail on the reasons underlying this selection of parameters is provided in the
demand and revenue model report.

25

New Lines Programme: Stated Preference Survey

New Lines Programme: Stated Preference Survey

Analysis of the Access and Egress SP exercise

5.1

The same SP exercise was presented to all respondents. Respondents were asked to
compare ways of accessing the rail station, if they were to use the new rail service.
Figure 5.1 presents an example of the SP card.
FIGURE 5.1

EXAMPLE OF ACCESS SP CHOICE

Option A

Option B

Rail

Rail

30 min

40 min

Mode of Transport

Journey time

Cost

Which of these options would you prefer?

SP3

Analysis of the Choice Proportions

5.2

The access-egress SP was different in design from the mode-choice SP. The left
hand card, or Option A, always represented a shorter journey but more expensive.
The right hand card, or Option B, always represented a longer journey but cheaper
than Option A. As the cost difference increases, more respondents tend to choose
the cheaper option. It is seen from Table 5.1 that when there was a difference of 2
for option A, 79% people chose the more expensive option. When the difference was
increased to 8, only 44% chose the more expensive option.
TABLE 5.1

CHOICE PROPORTIONS: DIFFERENCE IN COST ALL USERS

Difference between Left

card and Right card ()

5.3

Option A:

Option B:

quicker, more expensive

longer, cheaper

79%

21%

54%

46%

44%

56%

Table 5.2 presents the results for the whole sample. When segmented by existing
mode, it is to be noted that rail users were the most sensitive to changes in access
time.

27

New Lines Programme: Stated Preference Survey

TABLE 5.2

CHOICE PROPORTIONS: DIFFERENCE IN ACCESS TIME - ALL USERS

Difference between Left

card and Right card (min
of access)

Option A:

Option B:

quicker, more expensive

longer, cheaper

10

49%

51%

15

56%

44%

25

72%

28%

Trading analysis
5.4

Table 5.3 shows the trading proportions for different modes. These are very high
and indicate that most of the respondents chose different modes of access
according to their characteristics.
TABLE 5.3

TRADING ANALYSIS - ACCESS AND EGRESS SP

Mode

% of traders per current mode

Rail

93%

Air

90%

Car

92%

Model calibration
5.5

Table 5.4 to Table 5.9 show the results of the logit models for different modes with
respect to journey purpose. The constant indicates that, all things being equal,
access by rail is preferred to access by bus by an equivalent of 10 minutes of access
time. Taxi is valued slightly better than rail by business people travelling by train on
their long-distance journey.

TABLE 5.4

5.6

RAIL USERS BUSINESS TRAVELLERS

Parameter

Value

Z Value

Relative value

Cost

-0.32

-16.37

IVT

-0.09

-18.24

18/h

Rail vs. bus ASC

0.95

9.72

10 minutes

Taxi vs. bus ASC

1.66

7.08

12 minutes

People travelling for leisure purpose value access by rail slightly better than access
by taxi. It can be seen that the monetary value of access time for leisure travellers
is lower than business travellers by almost 5.

New Lines Programme: Stated Preference Survey

TABLE 5.5

5.7

RAIL USERS - LEISURE TRAVELLERS

Parameter

Value

Z Value

Relative value

Cost

-0.38

-15.0

IVT

-0.077

-11.92

12/h

Rail vs. bus ASC

0.97

8.83

12.5 minutes

Taxi vs. bus ASC

0.79

4.6

10 minutes

Air travellers have a much higher valuation of access time but do not value much
the difference of mode of access, the important factor being journey time.

TABLE 5.6

AIR USERS - BUSINESS TRAVELLERS

Parameter

Value

Z Value

Relative value

Cost

-0.15

-8.5

IVT

-0.08

-14.3

35/h

Rail vs. bus ASC

0.6

5.3

6.5 minutes

Taxi vs. bus ASC

0.7

4.05

8 minutes

29

New Lines Programme: Stated Preference Survey

TABLE 5.7

5.8

AIR USERS - LEISURE TRAVELLERS

Parameter

Value

Z Value

Relative value

Cost

-0.02

-5.3

IVT

-0.06

-5.9

18.3/h

Rail vs. bus ASC

0.03

1.5

6 minutes

Taxi vs. bus ASC

0.5

1.3

8 minutes

Car travellers, especially the ones travelling for leisure purposes, value very
differently on access by bus or rail than access by taxi to the rail station.

TABLE 5.8

CAR USERS - BUSINESS TRAVELLERS

Parameter

Value

Z Value

Cost

-0.15

-5.1

IVT

-0.07

-7.5

29/h

Rail vs. bus ASC

0.23

1.03

3 minutes

Taxi vs. bus ASC

0.9

3.0

12 minutes

Value

Z Value

Relative value

Cost

-0.2

-7.7

IVT

-0.04

-5.7

12/h

Rail vs. bus ASC

0.3

1.3

7 minutes

Taxi vs. bus ASC

1.1

2.8

24 minutes

TABLE 5.9

Relative value

CAR USERS - LEISURE TRAVELLERS

Parameter

APPENDIX
A
MAIN MODEL CALIBRATION

Appendix A

A1.

MODE CHOICE SP RESULTS

As discussed in Chapter 4, different models were calibrated in order to see how the
relative weight to journey time and frequency would vary. This Appendix presents
models whose structures vary substantially from the preferred model. Appendix B
presents the results of some additional models whose structures are closer to the
preferred model.
The findings from these models are in-line with results obtained with the preferred
model. When it has not been the case, it was mainly because the model failed to
calibrate in a satisfactory manner.
Models tested included

A2.

Estimating the relative weight to generalised journey time, rather than invehicle time only;

Segmenting the results by route (Glasgow/Manchester);

Having a discrete representation of headway;

Combining both rail options into a single one;

Calibrating the models on the whole sample (air-rail-car users);

Estimating the relative weight to a non-linear impact of change in in-vehicle

time;

Estimating the variation amongst respondents.

GENERALISED JOURNEY TIME

This section, discusses the results of the models where instead of the In-Vehicle-Time
(IVT), Generalized Journey Time (GJT) was measured. In other words, the access and
egress costs and times were added to the utilities.
The models in tables A1.1 and A1.2, made the best use of the available information.
The respondents characteristics for access and egress to its current mode were
added into the estimation. Since the access and egress characteristics for the other
mode (e.g. access time to the airport for rail user) was not available, the average
value for people who chose this mode (e.g. average access time to the airport on all
air users) was used. It is recognised that this may be biased; people travelling by air
may do so because their origin or destination is closer to the airport than to the rail
station. However more detailed information is not available at this stage. The
addition of access and egress time introduces a different way of estimating the
journey times and costs. Results are in-line with other models for business travellers
but are considered too high for leisure travellers.

Appendix

APPENDIX: TABLE A2.1 MODEL1: GJT BUSINESS TRAVELLERS

MULTINOMIAL STRUCTURE

Parameter

Value

T-stat

Relative value

Cost

-0.015

-9.02

Classic Rail GJT

-0.010

-15.27

43/h

High Speed Rail

GJT

-0.005

-5.89

21/h

Air GJT

-0.007

-7.61

27/h

Headway

-0.009

-5.64

39/h

APPENDIX: TABLE A2.2 MODEL 1: GJT - LEISURE TRAVELLERS

MULTINOMIAL STRUCTURE

Parameter

Value

T-stat

Relative value

Cost

-0.03

-13.71

Classic Rail GJT

-0.005

-9.62

10/h

High Speed Rail

GJT

-0.0015

-2.64

3/h

Air GJT

-0.005

-5.7

10/h

Headway

-0.002

-0.94

4/h

Appendix A

In the models in tables A1.3 and A1.4, we decided to represent access times and costs
in a similar way for the respondents current mode and the alternative one by using
the average value for people who chose this mode (e.g. average access time to the
airport on all air users). Results are very close to the models based on in-vehicle time
only.
APPENDIX: TABLE A2.3 MODEL 2: GJT - BUSINESS TRAVELLERS
MULTINOMIAL STRUCTURE

Parameter

Value

T-stat

Relative value

Cost

-0.04

-17.42

Classic Rail GJT

-0.028

-18.55

38/h

High Speed Rail

GJT

-0.022

-11.85

32/h

Air GJT

-0.018

-11.01

26/h

Headway

-0.013

-7.32

18/h

For leisure travellers, the relative weight on headway is not statistically significant at
a 5% level and almost significant at 10% level.

APPENDIX: TABLE A2.4 MODEL 2: GJT - LEISURE TRAVELLERS

MULTINOMIAL STRUCTURE

A3.

Parameter

Value

T-stat

Relative value

Cost

-0.05

-15.17

Classic Rail GJT

-0.179

-11.4

23/h

High Speed Rail

GJT

-0.016

-7.71

21/h

Air GJT

-0.015

-8.09

20/h

Headway

-0.004

-1.63

5/h

SEGMENTATION PER ROUTE

Table A1.5 shows the impact of journey route since the value of time may vary
according to the duration of the travel. There were two routes on which respondents
were interviewed: one was the shorter on to Manchester and surrounding areas and
the other was the longer one to Glasgow. Rail and car users on both routes were
interviewed. Only air users travelling on the Manchester route were interviewed.

Appendix

Tables A1.5 and A1.6 show the same model disaggregated into the two routes. The
results are in-line with expectations: shorter journeys tend to have a larger value of
time than longer journeys.

APPENDIX: TABLE A3.1 MODEL 9C: MANCHESTER ROUTE

MULTINOMIAL STRUCTURE
Parameter

Value

T-stat

Relative value

Cost

-0.044

-22.9

Classic Rail IVT (Business)

-0.03

-20.52

41/h

Classic Rail IVT (Leisure)

-0.02

-12.9

27/h

High Speed Rail IVT

(Business)

-03023

-10.64

31/h

High Speed Rail IVT (Leisure)

-0.0187

-7.61

26/h

Air IVT (Business)

-0.025

-13.3

34/h

Air IVT (Leisure)

-0.024

-11.29

33/h

Headway (Business)

-0.014

-7.54

19/h

Headway (Leisure)

-0.004

-1.65

5/h

Table A1.6 presents the results for the Glasgow route. It should be noted that rail
passengers travelling on leisure on this route had a very low valuation of travel time
savings. This is again in-line with expectations since, given the journey time
difference between air and rail mode on this route, you would not expect people who
value travel time savings highly to travel by train.
APPENDIX: TABLE A3.2 MODEL 9C: GLASGOW ROUTE
MULTINOMIAL STRUCTURE

Parameter

Value

T-stat

Cost

-0.044

-22.9

Classic Rail IVT (Business)

0.012

6.44

25/h

Classic Rail IVT (Leisure)

0.0122

6.05

10/h

High Speed Rail IVT

(Business)

0.014

5.05

12/h

High Speed Rail IVT (Leisure)

0.017

5.02

3/h

Air IVT (Business)

Relative value

No air travellers to Glasgow

Appendix A

Parameter

Value

Air IVT (Leisure)

A4.

T-stat

Relative value

No air travellers to Glasgow

Headway (Business)

-0.014

-7.54

19/h

Headway (Leisure)

-0.004

-1.65

5/h

ALTERNATIVE REPRESENTATION OF HEADWAY

The preferred model estimates the relative weight per minute of headway. This was
not found to be statistically significant for leisure trips. We subsequently tried to
calibrate a discrete level of headway, i.e. the relative weight of a 30-minute
headway compared to the relative weight of 60-minute headway.
Three discrete levels of headway were presented in the SP exercise (20 minute, 30
minute and 60 minute headway) for the two rail options. Air headway was presented
as constant (90 minutes). The base level in the results presented below is 20 minutes
and we present the valuation
It could be noted that headway of 30 minutes compared to 20 minutes for leisure
travellers was not significant. The Table A1.7 describes the results from the model.
APPENDIX: TABLE A4.1 MODEL 9B: REPRESENTATION OF HEADWAY
MULTINOMIAL STRUCTURE

Appendix

Parameter

Value

T-stat

Relative value

Cost

-0.05

-22.8

Classic Rail IVT (Business)

-0.03

-20.03

40/h

Classic Rail IVT (Leisure)

-0.02

-12.1

25/h

High Speed Rail IVT

(Business)

-0.022

-10.1

30/h

High Speed Rail IVT (Leisure)

-0.016

-6.87

23/h

Air IVT (Business)

-0.03

-17.5

43/h

Air IVT (Leisure)

-0.03

-12.4

34/h

Headway Rail - 30 minutes

(Business)

-0.14

-2.20

20/h

Headway Rail 60 minutes

(Business)

-0.54

-7.43

19/h

Headway Rail 60 minutes

(Leisure)

-0.134

-1.6

5/h

A5.

NEST WITH ONE RAIL OPTION

The model presented in this section was the nested structure with classic and highspeed rail combined into a single category rail. This model is the one in which the 2
rail options are considered as one option. In order to calibrate this model, we had to
duplicate the two rail choices per respondent (as one hsr-air choice and one classic
rail air choice). This is not a conventional approach and the results produced were
not considered robust.

Car
Rail

Air

A multinomial model with the same variables was also tested: Tables A1.8 and A1.9
respectively present the results of the nested and of the multinomial models.
In Model 13a, it can be seen that the coefficient of cost does not have the right sign
on it i.e., it is positive when it should be negative. This is highly unrealistic.
APPENDIX: TABLE A5.1 MODEL 13A: CLASSIC AND HIGH SPEED RAIL COMBINED
NESTED STRUCTURE
Parameter

Value

T-stat

Cost

0.02

21.23

Classic Rail IVT (Business)

-0.004

-6.57

12/h

Classic Rail IVT (Leisure)

0.0016

2.97

5/h

Air IVT (Business)

-0.024

-13.55

76.5

Air IVT (Leisure)

-0.02

-11.5

63/h

Car IVT (Business)

-0.018

-21.3

60/h

Car IVT (Leisure)

-0.018

-21.3

60/h

Headway (Business)

0.0053

3.91

17.2/h

Headway (Leisure)

0.00125

0.92

4/h

1.95

14.5

Nest Parameter

Relative value

Appendix A

In Model 13b above, it can be seen that journey times have a very high relative
weight.
APPENDIX: TABLE A5.2 MODEL 13B: CLASSIC AND HIGH SPEED RAIL COMBINED
MULTINOMIAL STRUCTURE

A6.

Parameter

Value

T-stat

Relative value

Cost

-0.03

-18.6

Classic Rail IVT (Business)

-0.03

-29.1

59/h

Classic Rail IVT (Leisure)

-0.02

-15.4

31/h

Air IVT (Business)

-0.03

-25.3

62/h

Air IVT (Leisure)

-0.02

-15.2

43/h

Car IVT (Business)

-0.02

-26.2

45/h

Car IVT (Leisure)

-0.01

-11.1

19/h

Headway (Business)

-0.01

-4.8

18/h

Headway (Leisure)

-0.004

-1.96

8/h

MODEL FOR ALL RESPONDENTS (AIR-RAIL-CAR)

We calibrated two models on the whole data set (one for leisure travellers and the
other for business travellers). The relative weight on journey time is assumed to be
the same for all modes but alternative specific constants were estimated.
Tables A1.10 and A1.11 will look at the double-nested structures for both business
and leisure travellers.

APPENDIX: TABLE A6.1 MODEL 14: COEFFIECIENTS OF IVT COMBINED FOR ALL
MODES, BUSINESS TRAVELLERS NESTED STRUCTURE

Appendix

Parameter

Value

T-stat

Relative value

Cost

-0.02

-6.33

IVT

-0.01

-5.9

37/h

Headway

-0.006

-5.1

19/h

Air ASC

0.174

1.55

-15 minutes

Car ASC

0.5

3.64

-46 minutes

High Speed Rail ASC

0.4

5.09

-30 minutes

Parameter

Value

T-stat

Nest Parameter Rail

2.3

6.01

Nest Parameter
PT

1.94

4.9

Relative value

APPENDIX: TABLE A6.2 MODEL 14: COEFFICENTS OF IVT COMBINED FOR ALL
MODES LEISURE TRAVELLERS NESTED STRUCTURE

A7.

Parameter

Value

T-stat

Relative value

Cost

-0.03

-9.04

IVT

-0.01

-9.04

21/h

Headway

-0.002

-2.1

5/h

Air ASC

0.6

-2.3

67 minutes

Car ASC

5.21E+08

-107 minutes

High Speed Rail ASC

0.2

3.17

-22 minutes

Nest Parameter Rail

2.0

9.0

Nest Parameter
PT

1.4

4.4

NON-LINEAR VALUATION OF TRAVEL TIME SAVINGS

We calibrated a non-linear model in which time was the only variable that was nonlinear (log-formulation). All parameters are strongly significant at a 5% level. Even
the weight on headway for leisure people is significant.
APPENDIX: TABLE A7.1 MODEL 14: TIME IS NON-LINEAR MULTINOMIAL STRUCTURE

Parameter

Value

T-stat

Cost

-0.05

-23.71

Classic Rail IVT (Business)

-4.2

-17.27

Classic Rail IVT (Leisure)

-2.9

-11.01

High Speed Rail IVT (Business)

-4.3

-18.03

Appendix A

Parameter

Value

T-stat

High Speed Rail IVT (Leisure)

-2.8

-11.0

Air IVT (Business)

-4.4

-18.1

Air IVT (Leisure)

-2.8

-10.7

Headway (Business)

-0.5

-7.4

Headway (Leisure)

-0.18

-2.0

Tables A1.13 and A1.14 show the monetary impact of 30 minute reduction in journey
times and how this impact varies according to the different modes and different
travellers. It can be seen that the value of time is significantly lower for leisure
travellers with comparison to the business travellers for the same modes. It can also
be seen that as the existing IVT increases from 2 to 4 hours, the value of time
decreases. This is consistent for all travellers on all modes.
APPENDIX: TABLE A7.2 MONETARY IMPACT () OF A REDUCTION OF 30 MINUTES
IN JOURNEY TIME BUSINESS TRAVELLERS

IVT

Classic Rail

High Speed Rail

Air

2hours 15
minutes

23

23

24

3 hours

17

17

4 hours 30
minutes

12

12

APPENDIX: TABLE A7.3 MONETARY IMPACT () OF A REDUCTION OF 30 MINUTES

IN JOURNEY TIME LEISURE TRAVELLERS
IVT

Appendix

Classic Rail

High Speed Rail

Air

2hours 15
minutes

16

15.5

15

3 hours

12

11

4 hours 30
minutes

A8.

TASTE VARIATION
Taste variation aims at understanding how differently respondents value travel time
savings. Taste variation models assume an underlying distribution of variation and
estimate average and variance of the sample for this distribution. We calibrated two
models: one for which the variation was assumed to be on the cost weight and the
other for which we assumed that the variation was on the time weight. In both cases
we assumed a normal distribution.
When we try the variation on journey time, we see that the only significant variation
is on classic rail in-vehicle time. The variation per respondent on high speed rail and
air journey times does not come out as statistically significant.

APPENDIX: TABLE A8.1 TASTE VARIATION ON JOURNEY TIME PARAMETERS

Parameter

Value

T-stat

Cost

-0.06

-10.9

Average: -0.04

Average: -9.7

Std. Dev.: 0.01

Std. Dev.: 5.8

Average: --0.03

Average: -8.3

Std. Dev.: 0.002

Std. Dev. 0.8

Average: -0.03

Average: --9.6

Std. Dev.: 0.004

Std. Dev. 1.2

-0.01

--6.3

Classic Rail IVT

High Speed Rail IVT

Air IVT

Headway

Relative value

37/h

30/h

34/h

13/h

The variation per respondent on cost comes out as statistically significant.

APPENDIX: TABLE A8.2 TASTE VARIATINON ON COST PARAMETER

Parameter

Value

T-stat

Average: -0.04

Average:-18.4

Std. Dev. 0.02

Std. Dev.: 3.9

Classic Rail IVT

-0.02

-18.1

34/h

High Speed Rail IVT

-0.02

-10.7

27/h

Air IVT

-0.02

-14.5

34/h

Headway

-0.009

-5.5

12/h

Cost

Relative value

Appendix A

Some tables in Appendix B describe in detail the results of the a few more model
runs, varying the use of alternative specific constant and the use of different weights
on journey time per respondent. The results are in-line with the models presented so
far.

Appendix

APPENDIX
B
ADDITIONAL MODEL RESULTS

Appendix B

B1.

MODE CHOICE SP RESULTS

The models presented below vary very slightly from the preferred models. The
differences are
I

the structure of the logit model (multinomial or nested)

the presence or absence of alternative specific constant (denoted ASC)

a value of travel time constant for all modes or varying per mode.

The results are very close to the one of the preferred models.

APPENDIX: TABLE B1.1 MODEL 1: AIR-RAIL SAMPLE, BUSINESS TRAVELLERS

MULTINOMIAL STRUCTURE, NO ASC, MODAL VOT

Parameter

Value

T-stat

Relative value

Cost

-0.0422

-17.2

Classic Rail IVT

-0.028

-19.6

41/h

High Speed Rail IVT

-0.022

-10.2

31/h

Air IVT

-0.024

-13.4

35/h

Headway

-0.013

-7.3

19/h

APPENDIX: TABLE B1.2 MODEL 2: AIR-RAIL SAMPLE, LEISURE TRAVELLERS

MULTINOMIAL STRUCTURE, NO ASC, MODAL VALUE OF TIME

Appendix

Parameter

Value

T-stat

Relative value

Cost

-0.046

-15.1

Classic Rail IVT

-0.019

-12.1

24/h

High Speed Rail IVT

-0.017

-6.9

22 /h

Air IVT

-0.024

-11.1

31 /h

Headway

-0.004

-1.6

5/h

APPENDIX: TABLE B1.3 MODEL 3: AIR-RAIL SAMPLE, BUSINESS TRAVELLERS

MULTINOMAIL STRUCTURE, 2 ASC'S, FIXED VALUE OF TIME

Parameter

Value

T-stat

Cost

-0.045

-17.8

In vehicle time

-0.026

-16.73

35/h

High Speed Rail

ASC

1.01

7.99

39 minutes

Air ASC

0.734

4.96

28 minutes

-0.0135

-7.38

18/h

Headway

Relative value

APPENDIX: TABLE B1.4 MODEL 4: AIR-RAIL SAMPLE, LEISURE TRAVELLERS

MULTINOMIAL STRUCTURE, 2 ASC'S, FIXED VALUE OF TIME

Parameter

Value

T-stat

Relative value

Cost

-0.047

-15.49

In vehicle time

-0.017

-10.67

22/h

High Speed Rail

ASC

0.354

2.54

20 minutes

Air ASC

-0.603

-3.52

37 minutes

Headway

-0.004

-1.63

5/h

Appendix B

APPENDIX: TABLE B1.5 MODEL 5: AIR-RAIL SAMPLE, BUSINESS

NESTED STRUCTURE, NO ASC, MODAL VALE OF TIME
Parameter

Value

T-stat

Relative value

Cost

-0.038

-17.3

Classic Rail IVT

-0.026

-7.27

41/h

High Speed Rail IVT

-0.02

-6.48

31/h

Air IVT

-0.02

-9.84

36/h

Headway

-0.01

-5.96

19/h

Nest parameter
Rail

1.12

6.88

APPENDIX: TABLE B1.6 MODEL 6: AIR-RAIL SAMPLE, LEISURE TRAVELLERS NESTED

STRUCTURE, NO ASC, MODAL VALUE OF TIME

Parameter

Value

T-stat

Cost

-0.033

-3.62

Classic Rail IVT

-0.015

-3.59

27/h

High Speed Rail IVT

-0.014

-3.23

25/h

Air IVT

-0.02

-6.48

37/h

Headway

-0.004

-1.45

7/h

1.36

3.5

Nest parameter

Appendix

Relative value

APPENDIX: TABLE B1.7 MODEL 6B: AIR-RAIL SAMPLE, LEISURE TRAVELLERS

NESTED STRUCTURE, NO ASC, MODAL VALUE OF TIME

Parameter

Value

T-stat

Cost

-0.042

-3.49

Classic Rail IVT

-0.015

-3.28

21/h

High Speed Rail IVT

-0.013

-2.78

19/h

Air IVT

-0.024

-6.47

34/h

Headway

-0.001

-0.35

1/h

1.23

3.33

Nest parameter

Relative value

APPENDIX: TABLE B1.8 MODEL 7: AIR-RAIL SAMPLE, BUSIINES TRAVELLERS

NESTED STRUCTURE, 2 ASC'S, FIXED VALUE OF TIME
Parameter

Value

T-stat

Cost

-0.04

-7.31

In vehicle time

-0.023

-7.36

35/h

High Speed Rail

ASC

0.877

5.11

38 minutes

Air ASC

0.493

1.8

21 minutes

Headway

-0.012

-7.31

19/h

1.15

7.05

Nest parameter

Relative value

Appendix B

APPENDIX: TABLE B1.9 MODEL 8: AIR-RAIL SAMPLE, LEISURE TRAVELLERS

NESTED STRUCTURE, 2 ASC'S, FIXED VALUE OF TIME
Parameter

Value

T-stat

Cost

-0.0375

-5

In vehicle time

-0.014

-4.73

22/h

High Speed Rail

ASC

0.268

2.17

19 minutes

Air ASC

-0.857

-3.8

62 minutes

Headway

-0.003

-1.48

5/h

1.31

4.78

Nest parameter

APPENDIX: TABLE B1.10

Appendix

Relative value

MODEL 9: AIR-RAIL SAMPLE, BUSINESS AND LEISURE

TRAVELLERS MULTINOMIAL STRUCTURE, NO ASC, MODAL
VALUE OF TIME

Parameter

Value

T-stat

Relative value

Cost

-0.044

-22.9

Classic Rail IVT (Business)

-0.029

-20.1

39.6/h

Classic Rail IVT (Leisure)

-0.018

-12.2

25.2/h

High Speed Rail IVT (Business)

-0.022

-10.1

29.9/h

High Speed Rail IVT (Leisure)

-0.017

-7.0

23.1 /h

Air IVT (Business)

-0.024

-13.1

33.7/h

Air IVT (Leisure)

-0.023

-11.0

32.2/h

Headway (Business)

-0.012

-6.4

19.3/h

Headway (Leisure)

-0.004

-1.7

5.3/h

APPENDIX: TABLE B1.11

MODEL 10: AIR-RAIL SAMPLE, BUSINESS AND LEISURE

TRAVELLERS NESTED STRUCTURE, NO ASC, MODAL VALUE
OF TIME

Parameter

Value

T-stat

Cost

-0.037

-8.71

Classic Rail IVT (Business)

-0.025

-8.47

39.6/h

Classic Rail IVT (Leisure)

-0.016

-7.46

25.2/h

High Speed Rail IVT

(Business)

-0.019

-7.09

30.2/h

High Speed Rail IVT (Leisure)

-0.0145

-5.68

23.3/h

Air IVT (Business)

-0.0225

-10.83

36.1/h

Air IVT (Leisure)

-0.022

-10.68

35.3/h

Headway (Business)

-0.012

-6.4

19.3/h

Headway (Leisure)

-0.003

-1.61

5.3/h

1.19

8.36

Nest parameter

APPENDIX: TABLE B1.12

Parameter

Relative value

MODEL 11: CAR SAMPLE, BUSINESS AND LEISURE

TRAVELLERS MULTINOMIAL STRUCTURE, NO ASC, MODAL
VALUE OF TIME
Value

T-stat

Relative value

Cost

-0.0315

-12.97

Classic Rail IVT (Business)

-0.025

-5.88

47.6/h

Classic Rail IVT (Leisure)

-0.021

-1.73

39/h

High Speed Rail IVT

(Business)

-0.023

-3.85

43/h

High Speed Rail IVT (Leisure)

-0.0214

-3.41

40.8/h

Car IVT (Business)

-0.019

-5.63

36/h

Car IVT (Leisure)

-0.0131

-3.68

25/h

Headway (Business)

-0.010

-3.14

19.4/h

Headway (Leisure)

-0.006

-1.73

10.9/h

APPENDIX: TABLE B1.13

Parameter

Value

T-stat

Cost

-0.0242

-9.7

Classic Rail IVT (Business)

-0.0095

-5.43

23.6/h

Classic Rail IVT (Leisure)

-0.007

-4.37

19.7

High Speed Rail IVT

(Business)

-0.008

-3.59

18.8/h

High Speed Rail IVT (Leisure)

-0.006

-2.89

15.3/h

Car IVT (Business)

-0.009

-6.54

22.9/h

Car IVT (Leisure)

-0.004

-3.15

10.6/h

Headway (Business)

-0.005

-3.82

12.4/h

Headway (Leisure)

-0.002

-1.76

5.9/h

3.39

7.5

Nest parameter

Appendix

MODEL 12: CAR SAMPLE, BUSINESS AND LEISURE

TRAVELLERS NESTED STRUCTURE, NO ASC, MODAL VALUE
OF TIME
Relative value