analysis of rail travel time and fare differences between london and the north
TRANSCRIPT
An exploratory analysis of rail travel time and fare differences between
London and the North using publicly available datasets.
The Institute For Transport Studies – The University of LeedsDr Andrew Mark Tomlinson
23rd April 2015
Class 87 at Crewe, April 1977 3 x Class 86’s at Preston, April 1977
Class 08 shunter waiting for work at Preston, April 1977
Presentation Aims
• To introduce and raise awareness of two useful rail datasets
• To outline the content of the datasets and the difficulties associated with using them
• To demonstrate the use of the datasets in an example problem
• To report on leading edge research
Station Usage Data• Shows Passenger Entries/Exits/Interchanges
• Differentiates between Peak, Off-Peak and Seasons
• 1997 onwards• http://orr.gov.uk/statistics/published-stats/station-usage-estimates
• Excel format, with notes on methodology
• Better estimate of total passenger trips compared to ORR headline figure (1332.5M vs 1600M)
Leeds Station: Total Entries 1998 - 2014
UK Centres of Gravity
(using rail station usage data)
• UK Population
• Rail Station
• Rail Passengers
Median Method: Number North=Number South &Number West=Number East
UK Rail Passengers: Centre of Gravity
(London Paddington)
UK Rail Passengers: Centre of Gravity
(London Marylebone)
How does the daily commute differ between London and the North?
• Fare Paid
• Journey Time
Examined using publicly available datasets:
• ATOC Timetable Data
• ATOC Fares Data
Timetable Dataset• UK timetable available in electronic form from ATOC
(http://data.atoc.org/how-to)
– Text based fixed format files defined according to CIF End User Specification (www.atoc.org/clientfiles/files/RSPDocuments/20070801.pdf)
• Stations (nodes)– identified by CRS (3-letter) code and TIPLOC (timing point location)– geocoded to within 500m using Easting/Northing pair
• Services (links)– Header record:
• validity, days of operation, head-code, power-type, speed, class, TOC
– Details records (one per pair of adjacent stations):• Arrival/departure times, allowances, special instructions/activities
• Problems– Missing interchange times for large stations?
• Stations + Service records create a 3 dimensional network (x, y, and time).
• Traversing this network yields all routes and timings between two points
Timetable Dataset ExampleService Header
Service UID Y52133
From 14/12/2014
To 10/05/2015
Days Run 0000001
Head-code 2M63
Power Type DMU
Speed 075
Timing Load A
Train Class S
TOC (X Header) NT
Station Arrive Depart
HUD 10:15
SWT 10:22 10:22
MSN 10:27 10:28
GFD 10:36 10:36
MSL 10:41 10:41
SWT 10:45 10:46
AHN 10:50 10:50
MCV 11:04
• 2,953 station records, • 70,166 train service headers (period December 2014 – May 2015)• 837,007 train service movements (between pairs of stations)
Fares Dataset
• All UK rail fares available in electronic form from ATOC– Text based fixed format files– Uses a mix of CRS and NLC codes to identify locations– Comprehensive description of each table and field available
(http://data.atoc.org/sites/all/themes/atoc/files/SP0035.pdf) – Split into standard fares and non-derivable, TOC specific and Advance
purchase fares– Useful other information: restrictions, discounts, rounding, rail cards,
rovers, supplements
• Problems– Dataset very large
• standard fares alone can be imported into Access• Importing other fares cause Access 2GB limit to be exceeded
– No information about how to query the data• Reverse engineering + Validation
• Standalone Advantix Traveller application also available (much faster than the web)
Finding a Fare
Origin
Destination
CRS: HUDNLC: 8437
CRS: LDSNLC: 8487
One-way fare
Two-way fare
StationCluster
StationCluster
+ Group Stations (Bradford Stations)
+ Ticket Type: return/single, anytime/off-peak, first/standard
+ Route + Restrictions: Via / Not Via, Valid / Not Valid
Four Northern Cities
Leeds: WYPTE -LDS
Sheffield: SYPTE -SHF
Manchester: GMPTE - MAN
Liverpool: MPTE - LIV
London: TfL -LON
Model Specification(s)Attribute Value
Model Type Linear (OLS)
2 x Models 1. Destination LDS + MAN + SHF + LIV2. Destination LON
2 x Dependant variables
A. One way fare to centre, £ (Anytime day return/2)B. Travel Time to centre, minutes (including waiting time)
Independent variables
Variable A (Fare) Variable B (Time)
Model 1 Model 2 Models 1 + 2
• Cartesian Distance (km)
• Is Not in PTE (Dummy)
• Is City X (dummy)
• CartesianDistance (km)
• Cartesian Distance (km)
• Is Not Direct (Dummy)
Filter Origin >5 km, Not HS1 station, Journey Time < 90 minutes, Day Return
Data Points LON: 427, LDS: 119, LIV: 177, MAN: 228, SHF: 111
Results Model A (Fare)
• Fares increase (almost) linearly with distance
• Access charge becomes less significant as distance increases
• Fares within the ‘home’ PTE region cheaper than those outside PTE region
Model 1 (North) Model 2 (London)
n 635 427
Adjusted R2 0.80 0.84
Standard Error 1.04 1.32
B Std. Err B Std. Err
Constant: Access Charge (£) 1.01 0.105 1.00 0.137
Distance (£/km) 0.12** 0.005 0.25** 0.005
Not in PTE (£) 1.37** 0.126
Is MAN (£) 0.44** 0.087
Results Model B (Travel Time)
• Fit not that good• Travel time increases (approximately) linearly with distance• Overall journey times are shorter in London• Impact of changes more significant in North
Model 1 (North) Model 2 (London)
n 635 427
Adjusted R2 0.69 0.55 !
Standard Error 10.7 8.5
B Std. Err B Std. Err
Constant (minutes) 8.45 1.02 11.82 0.88
Distance (minutes/km) 1.11** 0.04 0.78** 0.03
Change needed (minutes) 6.73** 0.98 1.56 0.96
What proportion of fares difference can be attributed to time savings?
• Model rephrased to include Value of Journey time @ £6.81/hour (commuting VOT, 2014)
• Difference suggests that time saving benefits represent 25%-30% of fare premium paid by Londoners
• Some value could also be attached to other quality attributes (The Hated Pacers!)
Model 1 (North) Model 2 (London)
n 635 427
Adjusted R2 0.79 0.84
Standard Error 2.14 1.82
B Std. Err B Std. Err
Constant (£) 2.14 0.217 2.47 0.188
Distance (£/km) 0.26** 0.01 0.35** 0.007
Not in PTE (£) 2.28** 0.26
London vs North
Attribute Winner Notes
Journey Times London (≈ 20km/hfaster)
Effect of ChangingTrains
London (fewer and less disruptive)
London: 103 (24%) average wait 7.2 minutesNorth: 238 (38%) average wait 12.6 minutes
Fares North (≈ £0.12/kmcheaper)
VOT benefits account for 25% of difference
Fare Boundaries London (fewer/none)
PTE boundaries create artificial barriers, impose financial penalty on cross boundary travel (compare with VRR in Germany)
Day Return Tickets London (availablefrom all origins)
London: 427 (99.8%) out of 428North: 635 (91.2%) out of 696
Thirsk-Leeds, Preston-Manchester
Longer Distance Commuting
London Limited opportunities for commuting from >50km in North
Further Uses
• To create a repository of all timetables and fares data going forward
• To study evolving service patterns in order to write a narrative around the changing nature of passenger rail travel/industry
• Combine:
– fares, timetable and station entry/exit data
– population and employment data
To reverse engineer/synthesise a public OD trip matrix
How does the cost of car commuting compare to rail ?
AA (July 2014), ‘Average’ Petrol car • Fixed costs £3,678• Running costs £0.13/km (@ £1.09/litre)• Commuting assumed 5 days/week for 46 weeks/year• Excludes parking costs and values of difference in journey time