Transport Modelling Forum - session PT modelling and data Stairway to Heaven – Rail Growth for Ever???? Birmingham Tuesday 18th November 2014
Who am I? • Technical Director, Rail Planning, Mott MacDonald • 25 years in the “industry” – Aecom, Atkins, Motts • Undertaken a lot of research into rail demand drivers • Liked and trusted by DfT for knowledge about rail demand and the north • Not a Northerner (unfortunately!) • Like trains
What I’m going to tell you about today • What’s been happening to rail demand • Why might this be? • Two key studies • Summing up • Questions
What has been happening - 1? Chart 1.1b Passenger kilometres Great Britain annual data 1947 to 2007 (billions) 60
40
30
20
10
2005
2000
1995
1990
1985
1980
1975
1970
1965
1960
1955
1950
0 1947
Passenger kilometres (billions)
50
What has been happening - 2? • Rail has continued to grow during the longest recession in recent times - PDFH would not predict this • Two possible reasons: – New forces coming to play; – Revised theoretical behaviours.
How do we predict it for rail? g
Broken down into a series of markets:
p
GDPnew POPnew × × exp(n( NCnew − NCbase))× I E = GDP POP base base FUELCOSTnew FUELCOSTbase
f
CARTIMEnew × CARTIMEbase b
c
BUSCOSTnew × BUSCOSTbase a
b
BUSTIMEnew × BUSTIMEbase
BUSHEADnew AIRCOSTnew AIRHEADnew × × BUSHEAD AIRCOST AIRHEAD base base base
r
t
×
•
London Travelcard (LTC) flows
•
Southeast to/from LTC Area
•
Within-Southeast (excluding LTC)
•
Non-London and Southeast (LSE) to/from LTC Area
•
Inter-urban flows over 20 miles
•
Non-PTE less than 20 miles
•
PTE less than 20 miles
•
Airport flows
Source: PDFH Chapter B1
6
But we haven’t been very good at it of late – an Intercity route
But we haven’t been very good at it of late – West Midlands PTE <20 miles
When are we happy or sad? Segment
Good at predicting?
Around London
☺
Inter-city Regional Inter-urban outside London Seasons Full fare Reduced fares
☺
But why?
New forces impacting on rail – some ideas •
Company car taxation changes
•
Lower levels of learning to drive
•
London congestion charge – once off boost
•
Continuing car congestion and operating cost growth (a perception?)
New forces impacting on rail – more ideas •
Regional city structural change – big impact up north
•
Part versus full time employment
•
Students commuting to college from parental home
•
Car no longer an “item” of identity – but social media is!
•
Internet retailing on rail – more savvy ticket selection
•
Yield management on rail – demand up, yield down!
City
2003
2004
2005
2006
2007
CAGR
Leeds
66%
68%
68%
73%
76%
3.30%
Liverpool
61%
62%
63%
68%
64%
1.30%
Manchester
58%
61%
63%
64%
70%
4.60%
Newcastle
62%
61%
57%
59%
58%
-1.40%
Sheffield
66%
67%
67%
68%
69%
1.10%
Source: DfT Northern HLOS Growth Study – % white collar employment, ABI employment SIC codes J to O within city catchment
Research work into rail demand drivers • ATOC External Impacts Study (MM/Southampton TRG) • DfT Northern HLOS Growth Study (MM/Southampton TRG) • ATOC Additional Rolling Stock Study (MM/ITS Leeds) • PTEG Rail in the North Study (MVA/MM) • Segmentation of Rail Passenger Demand using Geographically Weighted Regression (Southampton TRG) • On the Move: Making sense of car and train travel trends in Britain (UCL, Imperial College, ITS) • Regional Flows Study (MVA) • Recession Impacts (SDG)
DfT Northern HLOS Growth Study
•
Trying to infill gap between best local PDFH predicted and observed demand – full and seasons only
•
Best PDFH predicted using localised input data including more relevant attraction station catchments
•
Consider use of other parameters which influence growth (beyond PDFH) •
Backcasting undertaken on six corridors in North of England agreed with client steering group
How bad are we? - season ticket Observed and predicted journeys for selected corridor - SEASON-TICKET trips 180.00 170.00 160.00
Index (2002 = 100)
150.00 140.00 Base PDFH Sector ABI employment Sector em't + MVA elasts Observed
130.00 120.00 110.00 100.00 90.00 80.00 2002
2003
2004
2005
2006
Railway year ended 31 March
2007
2008
2009
How bad are we?– full fare Observed and predicted journeys for selected corridor - FULL-FARE trips 160.00
150.00
Index (2002 = 100)
140.00
130.00 Base PDFH Sector ABI employment Sector em't + MVA elasts Observed
120.00
110.00
Note: For full and reduced fares, altering the employment measure has no effect, as employment is only assumed in PDFH to be linked to seasonticket journeys. Therefore the two predicted trends are identical for these ticket types.
100.00
90.00
80.00 2002
2003
2004
2005
2006
Railway year ended 31 March
2007
2008
2009
What can we see? •
Still a big gap evident
•
MVA GVA elasticities not tenable in downturn
•
Need to infill the gap using additional variables: – Structural change measure: ratio of service sector to total employment (jobs) in the city centres – Real parking cost changes over time – Origin-end descriptor to capture propensity to travel by rail (household incomes)
How we did it! • Ran SPSS regression to estimate parameters to explain the difference between best PDFH and observed • Assessed following new parameters: – Origin and destination structural change (becoming like the south!) – Car parking demand and supply – Household incomes – Crowding – Train service delivery compared to plan Observed yr 2 Observed yr1
PDFH yr 2 CarParkCos t y 2 = * PDFH yr1 CarParkCos t yr1
λ
Detailed employment information Employment Annual Growth Rates 2003 - 2007 7.0%
growth % p.a.
6.0% 5.0%
TEMPRO V5.4
4.0%
ABI - Local Authority Level (All sectors)
3.0%
ABI - City Centre Zones (All Sectors)
2.0% ABI - City Centre Zones (Sectors J-O)
1.0% 0.0% Leeds
Liverpool
Manchester
Newcastle
Sheffield
-1.0% HLOS City
Sectors J-O are financial, business, white collar other, public sector i.e. office based
What is structural change? City
2003
2004
2005
2006
2007
CAGR
Leeds
66%
68%
68%
73%
76%
3.3%
Liverpool
61%
62%
63%
68%
64%
1.3%
Manchester
58%
61%
63%
64%
70%
4.6%
New castle
62%
61%
57%
59%
58%
-1.4%
Sheffield
66%
67%
67%
68%
69%
1.1%
Ratio white collar to total employment in city centres (ABI data)
Regression results Season •
Best model explains gap by real increases in car parking cost, with elasticity of 1.090
•
This means an average +3% real increase p.a. in parking cost leads to season demand growing by +3.3% pa
•
Are we really valuing the ratio of white collar workers to long stay car parking spaces??
•
Scrubland parking space taken over by offices?
Full •
Best model from growth in proportion of white collar workers with elasticity of 0.721
•
This means that a change from 71% to 73% (i.e. +2.8%, recent values for central Manchester) leads to a +2.0% uplift in full demand
•
A more responsive market to change?
•
Full trips are partly business so expect a function of white collar jobs
How well did we fit? – North West corridor #1 Manchester Rochdale - Full + Season 1.7 1.6
Index 2002=1
1.5 1.4 1.3 1.2 1.1 1.0 0.9 2002
2003
2004
2005
2006
2007
2008
Year Observed Regression Results
Local PDFH Regression Results (Minus One)
Aggregate PDFH Regression Results (Plus One)
2009
How well did we fit? - North West corridor #2 Manchester CLC - Full + Season 1.6
1.5
Index 2002=1
1.4
1.3
1.2
1.1
1.0
0.9 2002
2003 Observed Regression Results
2004
2005
2006
Year Local PDFH Regression Results (Minus One)
2007
2008
Aggregate PDFH Regression Results (Plus One)
2009
How well did we fit? – Yorkshire corridor #1 Leeds Calder - Full + Season 1.5
1.4
Index 2002=1
1.3
1.2
1.1
1.0
0.9 2002
2003 Observed "Regression Results"
2004
2005
2006
Local PDFH Year Regression Results (Minus One)
2007
2008
Aggregate PDFH Regression Results (Plus One)
2009
How well did we fit? – Yorkshire corridor #2 Leeds Harrogate - Full + Season 2.5
Index 2002=1
2.0
1.5
1.0
0.5 2002
2003
2004
2005
2006
2007
2008
Year Observed Regression Results
Local PDFH Regression Results (Minus One)
Aggregate PDFH Regression Results (Plus One)
2009
ATOC Rail Demand and External Impacts Study •
To better understand how external factors influence rail demand and how this may change in the future –
•
what has been continuing to drive rail demand?
Is there an identifiable mechanism linking lagged economic effects to lagged impacts on rail demand? –
will the economic downturn hit rail in time?
•
Is there a (theoretical) need to revise PDFH v5.1 to cover the impact of external factors, either incrementally or a major change in approach? –
can we get PDFH much better?
All built around RUDD dataset • Extended TOAD data set with additional variables • Covers 20,778 O-D flows • Some new flows drop into dataset in 2007 (e.g. Luton Airport Parkway) • Covers period 1994-2012 • Ticket type based flows aggregated up to F, R, S • So 20,778 * 18 * 3 = 1,122,012 data entries! • Does not include PTE, Oyster infills • Exogenous data at NUTS1 and NUTS3 level
Variables tested •
Destination GVA (@ NUTS1 and NUTS3)
•
Origin population (@ NUTS1 and NUTS3)
•
Destination employment type (@ NUTS1 and NUTS3)
•
Car ownership (zero, mean and 2+)
•
Car fuel cost
•
Car journey time
•
Rail fare
•
Rail GJT
•
Rail reliability (PPM)
•
Lags
28
What has been found….hot off the press •
LSOA-based catchments for population gave clear improvement over NUTS 3-based data
•
LSOA-based catchments for employment gave an improvement, but estimating reliable parameters problematic
•
Including both origin and destination population improves model performance
•
Including average PPM improves model performance
•
Parameters for FT/PT and service sector employment very small
•
Segmented analysis suggests factors determining rail demand differ for trips to/from London •
•
Central London employment influences trips in both directions
Lagged impacts not clear
29
What has been found….in summary •
More disaggregate data helps a lot
•
Directionality a clear problem – predict bidirectionally for long distance/inter-urban markets
•
Full fare works better with employment as a driver
•
Reliability helps predictions
•
Revenue/yield a real problem
•
Apex tickets/single leg pricing have upset the apple-cart!
30
But the proof is in the pudding
31
Combined model results – full fare With Reliability
Without Reliability
32
Combined model results – reduced fare With Reliability
Without Reliability
33
Combined model results – season With Reliability
Without Reliability
34
Combined model results – Interurban >20 miles (without reliability)
35
Summing up • The rail industry thinks that it has the tools to predict rail demand, but…………. • The tools have failed to predict the explosion in rail since 1995 and riding out the recession • Many different reasons for demand growth – structural change up north, apexs, reliability, etc • Motts have had a leading role in the last 8 years in helping to explain why this has all happened • Single leg pricing messes it all up! • Will it continue or will we go back to how we were with BR?
www.mottmac.com
[email protected]