7. A Focus on Prediction Intervals
• Travel demand forecasts have prediction intervals
1. What do these prediction intervals look like?
2. What does empirical evidence tell us?
7
An estimate of an interval in which future observations will fall,
with a certain probability,
given what has already been observed.
8. A Focus on Prediction Intervals
• Travel demand forecasts have prediction intervals
1. What do these prediction intervals look like?
2. What does empirical evidence tell us?
8
An estimate of an interval in which future observations will fall,
with a certain probability,
given what has already been observed.
20. Conclusions
• Toll road traffic forecasting errors...
– are common
– are commonly large
– are symmetrically distributed around a mean < 1.0
• ‘optimism bias’
20
23. Literature Review
• JP Morgan (1997)
– 14 new toll roads in the US
– Only 1 exceeded its revenue prediction
– 3 missed the mark by 25%
– For 4, revenue was less than 30% of forecast
• Flyvbjerg et al (2005)
– 183 tolled and non-tolled roads
– For ½, difference between actual and forecast was ± 20%
– For ¼, difference was over ± 40%
– No improvement in accuracy over 30 years
23
24. Literature Review
• US Transportation Research Board (2006)
– 26 toll highways in the US
– “Most of the results demonstrate an underperformance”
– “Even with the availability of updated forecasts, only a small number
of projections are within 10% of the actual revenues”
• Vassallo (2007)
– 14 toll roads in Spain
– “On average, actual traffic was substantially overestimated (by
approximately 35%)”
24
25. Literature Review
• Li & Hensher (2009)
– Toll roads, bridges and tunnels in Australia
– “On average, the actual traffic level of these five toll roads*
is 45 percent lower than predicted”
• Bain (2009)
– 27 toll roads for the European Investment Bank
– Sample included some ‘shadow’ toll roads
– Accuracy range: -67% to +32%
– Clear evidence of error and skew (bias)
25
* The M2, M7, Cross City Tunnel, Lane Cove Tunnel & EastLink
26. Conclusions
• Toll road traffic forecasting errors...
– are common
– are commonly large
– are symmetrically distributed, but around a mean < 1.0
• ‘optimism bias’
• From Bain, 2009 (weak) and Flyvbjerg et al, 2005 (strong)
– Toll and toll-free roads differ
• In terms of bias (tendency for over-forecasting)
– Toll and toll-free roads are similar
• In terms of error (size of standard deviation)
26
29. Forecast Accuracy in the UK (POPE*)
UK Highways Agency Dataset (2012)
(n = 55)
Actual/Forecast Traffic
* Post-Opening Project Evaluation (POPE) of Major Schemes
29
30. Conclusions
• “73% of actuals within ± 15% forecasts”
– Varies by scheme type
• On-line schemes: 82% within ± 15%
• Bypass schemes: 67% within ± 15%
• Junction schemes: 67% within ± 15%
• Contrasted with toll road research findings
– Mean is different (≈ 1.0 cf. 0.77)
• Absence of systematic bias
– SD is not that different (0.22 cf. 0.26)
• Still significant error range
• HA: “¼ forecasts” are out by > 15%
– From HA raw data I calculate closer to ⅓
• Note (of importance): average age of forecast ≈ 5 - 10 years
30
33. Summary
• Compiled inventory of traditional (4-step) model short-comings, flaws
and limitations
– Many technical, detailed and mode-specific:
– Simplified assumptions about human behaviour
– Simplified assumptions about transport supply
– Other assumptions adopted for computational convenience
33
Models - irrespective of mathematical sophistication - remain
crude and imperfect representations
of complex and dynamic interactions
between transport demand and supply
34. Let’s Pretend...
• Let’s pretend that transport demand models are perfect
– Complete and correct in terms of design, architecture,
specification (explanatory variables & inter-relationships) etc.
– 100% accurate in terms of data used to construct and
estimate the base-year model
• In forecasting mode we need other inputs
– Forecasts of the explanatory (growth) variables themselves
34
37. Forecasting Inputs
• Forecasts of population are a key input for many
(most?) transport demand models
• Population forecasting should be relatively easy
– We know the population today
– There is a limited set of influences
• Births
• Deaths
• Migration 37
42. US Counties Are Actually Quite Large
• Average counties/state = 62
• Average population/county = 146,850
• But traffic modelling zones:
– Populations of around 1,000 - 3,000
– 70+
times smaller
42
43. US Counties Are Actually Quite Large
• Average counties/state = 62
• Average population/county = 146,850
• But traffic modelling zones:
– Populations of around 1,000 - 3,000
– 70+
times smaller
43
49. Our ‘Zone’ of Interest
These are 10-year forecast MAPEs.
Deeper horizon MAPEs will be larger!
49
50. Conclusions 1
• Population forecasts have sizeable error ranges associated with them
• These error ranges increase as the forecasting horizon increases
– Linear relationship?
• These error ranges increase as the study area decreases
– Non-linear inverse relationship
• The distributional characteristics of absolute percent forecast errors
appear stable over time (Smith & Sincich, 1988)
– Past errors can be used to estimate CIs for current forecasts
50
51. Conclusions 2
• Population is one of the more predictable variables commonly used to
explain traffic growth
• Try forecasting employment
• ...and allocating it to the correct zones
– Evidence from the US suggests that employment projections can be
(nearly) twice as inaccurate as population forecasts
Transportation Research Board, 2009
• Try forecasting GDP, income or fuel price!
51
54. DfT Research & Guidance
• WebTAG Unit 3.15.5
– The Treatment of Uncertainty in Model Forecasting
– Use a range about the core scenario growth forecast of:
± 2.5% * √n (n = number of years ahead)
– Formula estimated from national traffic forecast performance
– Functional form is intuitively appealing
• If error variance increases linearly with time...
• ...standard deviation should vary with the square root of the forecast horizon
– Local forecasts will have a (much?) wider range
• NRTF97
– For total traffic at the level of a GOR...the uncertainty should widen to
about ± 25% at the 35th
year
• “± 25% at GOR level feels narrow compared to ± 15% (Year 36) at the national
level”
• “The range for individual area types/links will be greater than GOR level (> ± 25%)”
54
55. DfT Research & Guidance
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
55
56. DfT Research & Guidance
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional
56
57. DfT Research & Guidance
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional ± 4.3% * √n
57
GOR: ± 25% = √35 * 4.3%
58. DfT Research & Guidance
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional > ± 4.3% * √n
58
59. DfT Research & Guidance
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional > ± 4.3% * √n
Local ?
59
60. DfT Research & Guidance
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional > ± 4.3% * √n
Local >> ± 4.3% * √n
60
63. Traffic Forecasting Accuracy Survey
Survey Respondents:
• International responses (11 countries)
• Consultants/modelling practitioners
– President
– Managing director
– Director of transport planning
• Government officials
– Transport modelling manager
– Senior transport & economics advisor
– Traffic & toll modelling manager
• Academics/researchers
– 4 professors
• Including one of the authors of ‘Modelling Transport’
– Senior lecturers
– Deputy director, centre for transport studies
63
64. Forecasting Accuracy Survey Results
Forecast Horizon Traffic Forecasting Accuracy
Existing Road New Road
Next Day
1 Year
5 Years
20 Years
64
65. Forecasting Accuracy Survey Results
Forecast Horizon Traffic Forecasting Accuracy
Existing Road New Road
Next Day ± 7.5% n/a
1 Year ± 12.5% ± 17.5%
5 Years ± 20% ± 27.5%
20 Years ± 42.5% ± 47.5%
65
97. Prediction Intervals
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
97
98. Prediction Intervals
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional > ± 4.3% * √n ?
98
99. Prediction Intervals
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional
Local
99
100. Prediction Intervals
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional
Local ± 7.5% * √n
100
101. Prediction Intervals
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional
Local ± 7.5% * √n ± 8% ± 24% ± 38%
101
102. Prediction Intervals
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional ± 5% * √n ?
Local ± 7.5% * √n ± 8% ± 24% ± 38%
102
103. Prediction Intervals
Study Level 90% Confidence Traffic Forecast Horizon
1 Year 10 Years 25 Years
National ± 2.5% * √n ± 3% ± 8% ± 13%
Regional ± 5% * √n ± 5% ± 16% ± 25%
Local ± 7.5% * √n ± 8% ± 24% ± 38%
103
104. So This is What Local Traffic
Forecasts Should Look Like...
109. 2015 Research Update
• The slide deck presented here dates from late 2013
• Since then Bain has been…
– Adding to the existing research steams:
• Forecasting errors - new studies measuring toll road forecasting errors (eg. EU
Court of Auditors)
– Consistent findings
• Population forecasts - turning from past errors to published forecast ranges
– Wide!
• Practitioner expectations - A US equivalent of my ‘ask the practitioners’ survey
has been conducted
– Broadly consistent findings
– Adding 3 new research streams (more petals!)…
109
110. New Research (WIP)
8. Same project, different consultants
• Huge variance!
9. Same project, different models
• Different models → different forecasts
10. Compare & contrast with Monte Carlo simulation
• Output ranges not dissimilar to my predictive intervals
110
All that remains is to publish the paper
Always the difficult part!!
111. New Research (WIP)
8. Same project, different consultants
• Huge variance!
9. Same project, different models
• Different models → different forecasts
10. Compare & contrast with Monte Carlo simulation
• Output ranges not dissimilar to my predictive intervals
111
All that remains is to publish the paper
Always the difficult part!!
128. • Into receivership in Jan. 2010.
• Cost $1.1bn+ but sold for $630m.
• $160m legal action against traffic
forecasters launched in Sept. 2009.
128
Lesson 4
136. Lesson 6
• Shares initially fell to 20% of value. Now worth $0.
• Into receivership in Feb. 2011 owing $1.3bn to banks.
• $150m class action lawsuit against traffic forecasters.
• $2bn legal action by receivers against traffic forecasters.
136
145. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
145
146. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
146
147. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
147
148. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
148
149. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
149
150. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
150
151. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
151
152. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
152
153. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
153
154. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
154
155. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
155
156. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
156
157. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
157
158. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
158
159. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
159
160. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
160
161. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
161
162. Why? 17 Contributory Factors...
• Public sector focus = maximum upfront value extraction from concession sale
• Big ticket, greenfield projects that could not be developed incrementally (tunnels)
• Lots of money chasing few assets
• Contractor and investment bank-led bidding consortia
• Small construction industry: limited competition (and higher prices)
• Aggressive financial structures: high leverage, massive debt
• Equity packaged and sold-down to third party investors
• Traffic consultants on big success fees (and extensive use of disclaimers)
• Traffic consultants promised further work
• High toll tariffs (reflecting high initial capex)
• Modest time savings (especially off-peak)
• Traffic calming on surface streets not delivered
• Traffic demand models with all the dials turned up to 11
• Limited period modelling and reliance on (massive) expansion/annualisation factors
• Key links (including target) over-capacity from Day 1. Busiest facilities in the world!
• Traffic & revenue auditor dispensed with (or remit considerably reduced)
• Winning bidders way out front (runners-up bunched around half of the winning forecast)
162
163. Analysis of Contributory Factors
• 4 main categories:
– Model-related
– Project-related
– Skewed incentives
– Gaming the forecasts
• Gaming the forecasts
– For strategic, commercial or political gain
– Forecasts are inherently contestable, so easy/obvious target!
163
164. Analysis of Contributory Factors
• 4 main categories:
– Model-related
– Project-related
– Skewed incentives
– Gaming the forecasts
• Gaming the forecasts
– For strategic, commercial or political gain
– Forecasts are inherently contestable, so easy/obvious target!
164
165. Analysis of Contributory Factors
• 4 main categories:
– Model-related
– Project-related
– Skewed incentives
– Gaming the forecasts
• Gaming the forecasts
– For strategic, commercial, institutional or political gain
– Forecasts are inherently contestable, so obvious/easy target!
165
166. Who’s the Bad Guy?
• The private sector is often painted as the bad guy
– ...but the public sector has its moments!
• Recent commission (review public sector forecasts)
– Initial feedback - forecasts = too high - not well received
– Suspicions that BCR was being propped-up by the (high) forecasts
– My suggested numbers removed from final report
– Final report not made public
– Business case not made public
– Forecasts
• Public sector consultants: 65-70,000 vehicles/day
• Independent review team: 35-55,000 vehicles/day
• Politician: 80-100,000 vehicles/day!!
– Project opponents vilified in the press
• Despite focus on process, not project
• Accusations of “analysis paralysis”
• Public sector forecasts regularly regarded as ‘floor’
– ...or hurdle for private sector to beat
– Often = the root cause of the problem
166
167. Who’s the Bad Guy?
• The private sector is often painted as the bad guy
– ...but the public sector has its moments!
• Recent commission (review public sector forecasts)
– Initial feedback - forecasts = too high - not well received
– Suspicions that BCR was being propped-up by the (high) forecasts
– My suggested numbers removed from final report
– Final report not made public
– Business case not made public
– Forecasts
• Public sector consultants: 65-70,000 vehicles/day
• Independent review team: 35-55,000 vehicles/day
• Politician: 80-100,000 vehicles/day!!
– Project opponents vilified in the press
• Despite focus on process, not project
• Accusations of “analysis paralysis”
• Public sector forecasts regularly regarded as ‘floor’
– ...or hurdle for private sector to beat
– Often = the root cause of the problem
167
168. Who’s the Bad Guy?
• The private sector is often painted as the bad guy
– ...but the public sector has its moments!
• Recent commission (review public sector forecasts)
– Initial feedback - forecasts = too high - not well received
– Suspicions that BCR was being propped-up by the (high) forecasts
– My suggested numbers removed from final report
– Final report not made public
– Business case not made public
– Forecasts
• Public sector consultants: 65-70,000 vehicles/day
• Independent review team: 35-55,000 vehicles/day
• Politician: 80-100,000 vehicles/day!!
– Project opponents vilified in the press
• Despite focus on process, not project
• Accusations of “analysis paralysis”
• Public sector forecasts regularly regarded as ‘floor’
– ...or hurdle for private sector to beat
– Often = the root cause of the problem
168
169. Who’s the Bad Guy?
• The private sector is often painted as the bad guy
– ...but the public sector has its moments!
• Recent commission (review public sector forecasts)
– Initial feedback - forecasts = too high - not well received
– Suspicions that BCR was being propped-up by the (high) forecasts
– My suggested numbers removed from final report
– Final report not made public
– Business case not made public
– Forecasts
• Public sector consultants: 65-70,000 vehicles/day
• Independent review team: 35-55,000 vehicles/day
• Politician: 80-100,000 vehicles/day!!
– Project opponents vilified in the press
• Despite focus on process, not project
• Accusations of “analysis paralysis”
• Public sector forecasts regularly regarded as ‘floor’
– ...or hurdle for private sector to beat
– Often = the root cause of the problem
169
170. Solution?
• Disclosure and transparency?
• Adjust/correct the incentives?
• Checks and balances to reduce the scope for influence?
• Take the pressure off the forecasts?
• ...or a mix of all 4?
• What do you think?
170