The Stockholm Congestion Charges forecasts and reality

1
The Stockholm Congestion Charges forecasts and
reality

• Jonas Eliasson
• Professor Transport Systems Analysis
• Centre for Transport Studies, Royal Institute of
  Technology
• WSP Analysis Strategy

Karin Brundell-Freij Ph D, Project Manager WSP
Analysis Strategy
2
The Stockholm congestion charges

• 10-20 SEK (1-2 ) per cordon crossing, depending
  on time of day
• No charge evenings or weekends
• Taxi, buses, green cars exempt
• Max 60 SEK/day

3
Trials and errors

• Tried quite a few system designs before arriving
  at final proposal
• Several did more damage than good created
  congestion in new places
• Unfortunately no systematic evaluation measure or
  mechanism

4
The SAMPERS model

• 5 separate regional models national
  long-distance model
• used Stockholm regional model
• Nested logit model (frequency, destination mode
  choice)
• six modes
• six trip purposes
• Distribution traffic fixed OD matrix
• Car OD matrix calibrated against traffic counts
• Car and PT assignment in EMME/2
• Departure time effects handled in simple, add-on
  model

5
Forecasts
6
Change in traffic flows (AM peak) overview
7
Forecasted effects on car traffic
average charged hours
-25
-13
-18
7
8
Travel times speed reductions in AM peak
After
Before
60 50 40 30 20
9
Time distribution of traffic across the cordon
No charges With charges
10
What did we believe?

• We thought that
• traffic would decrease, but less than the
  forecasted 25
• speeds would increase, primarily along the
  north-south axis
• but not visibly
• and mostly inside the cordon
• no change from the west
• Essinge traffic and congestion would increase
  significantly
• effects during peak hours would be much larger
  than between peaks

11
Evaluation results
12
Evaluation and appraisal

• Objectives of evaluation
• The evaluation program and process
• Selected trial results
• Volumes
• Travel behaviour
• Congestion and speed
• Environment
• Business and retail
• Economy (CBA)
• Perceptions and opinion
• Referendum
• Monitoring with a permanent system
• First results from the permanent system

13
Objectives for evaluation

• Basis for informed votes in the referendum
• Broad view, supplementary to personal experience
• Basis for modifications of the system, in case of
  a permanent implementation
• Databases for national and international research
  
• And, less officially, to
• communicate openness, transparency and
  procedural fairness. Potentially important for
  the outcome
• Counter campaigns based on hearsay and anecdotal
  evidence

14
Comprehensive and detailed evaluation

• Retail and visiting
• Taxi/deliveries/mobility service
• Workmen, driving schools
• Deliveries, garbage collecting
• Case study Two workplaces
• Childrens sports activities
• Cost-benefit analysis (equity effects in August)
• Regional economic analysis
• Log of special events

• Travel survey Stockholm county
• Travel survey Mälardalen
• Car traffic
• Cordon passages
• GPS travel times
• Queue lengths
• Public transit
• Parking
• Cycle and walk
• Traffic safety
• Perceived environment
• City life
• Air quality
• Emissions
• Noise

15
Basic design

• All aspects
• Before-and-after
• Problem due to delay of trial in most cases
  before- and after- studies were different
  seasons
• October 2004 and April 2006
• For a limited number of aspects
• Go-live (daily results first 2 weeks)
• Monthly reports
• Reference group (planning), Expert group
  (evaluating results)

16
The technology worked (not in official evaluation)
17
Goals essentially reached

• 10 - 15 less traffic to/from inner city
• Was 20-25
• Increased accessibility
• Queue times down 30-50 near/in inner city
• Essingeleden about the same
• Decreased emissions
• 14 less in inner city 2,5 in total county
• Inhabitants should perceive an improved urban
  environment
• Unclear difficult to define and measure

18
Traffic per day across the cordon
19
Less traffic also further out and inside the
cordon
Minor streets inside Larger streets
inside Arterials inside Veh. km. inside
cordon Across cordon Arterials outside
cordon Arterials far outside cordon Streets
outside cordon Ring roads
20
Where did the car drivers go?

• Half of them changed to public transport
• Mainly work and school trips
• The other half changed in less traceable ways
• Fewer trips trip chaining
• Other destinations
• but not
• car-pooling
• work at home
• changed departure times
• Switch to non-charged by-pass (Essingeleden)

21
30-50 less time in queues morning peak
Delay time, AM peak
Access roads, inbound
Access roads, outbound
Inner streets
Arterials inside northbound southbound
22
Even larger effect during evening peak
Delay time, PM peak
Inner streets
Arterials inside northbound southbound
Access roads, outbound
Access roads inbound
23
Substantial reductions of emissions

• 10-14 less emissions in the inner city
• positive long-term health effects
• 2-3 less CO2 emissions in the county

Inner city 10-14 reduction
County 2-3 reduction
24
More positive attitude towards congestion charges

• Inhabitants of Stockholm municipality that
  regarded the trial to be a good/bad decision

Good
Bad
25
Business and retail

• Taxi, delivery services, craftsmen
• Dependent on a well functioning road transprt
  system
• Observed and appreciated travel time reductions
• Administrative burden
• No significant effects on retail
• Most shopping trips are evening/ week-end and
  non-car trips
• Models Marginal effects on land use, real estate
  prices and regional economy
• Negligible effects on household purchasing power

26
Cost-benefit analysis of the congestion charges
Payback time 4 years.
27
Referendum results
28
One third of county inhabitants did not vote

• Small majority for the charges in Stockholm city
• Small majority against the charges in Stockholm
  county
• One third lived in municipalites without
  referendum

29
First effects of permanent system same effects
on volumes
30
But less effects on congestion (disturbances)
31
Afternoon peak
32
Actual outcome compared to forecast
33
Overall picture fairly correct
34
Average daily effects close to predictions
forecast
Minor streets inside Larger streets
inside Arterials inside Veh. km. inside
cordon Across cordon Arterials outside
cordon Arterials far outside cordon Streets
outside cordon Ring roads
-18
-13
-25
-6
35
Essinge bypass increased less than expected (0-5)
2006 2005
forecast
7
Gröndalsbron
Frösunda
Midsommar-kransen
36
Time distribution different than prediction

• Traffic decrease largest during PM peak.
• Less decrease during AM peak
• Evening/night traffic decreased as well

37
Revenues less than forecasted

• Forecast 890 mill. SEK/year
• Outcome 760 mill. SEK/year (incl. lateness
  penalties etc.)
• 14 less than expected
• More exempted vehicles than we thought
• total 30 - assumed 17
• Lidingö 8-9 (predicted 9)
• taxi 8-9 (assumed 4)
• green cars 4 (assumed negligible)
• other exemptions 8-10 (assumed 4)

38
Effects on travel times(AM peak)
Change in travel time ()
39
Travel time decreases - examples

40
Predictions outcome summary

• Average daily decrease in traffic flows fairly
  correct
• Less increase on the Essinge bypass
• Relative effects AM/mid-day/PM completely wrong
• Missed the decrease of evening/night traffic
• Overestimated change in departure times
• Travel time decreases much larger, more visible
  and occurred in a much larger area than expected
• but perhaps we should have expected it!
• Revenues less than predicted
• No measurable retail effect as predicted
• Didnt know distributional impacts over time,
  i.e. was it the same people paying each day or
  was the cost spread across a larger group?

41
What caused the prediction errors ?

• More chose to go through the city centre, less on
  the Essinge bypass
• Values of time higher than we thought
• Travel time gains in the city centre larger than
  predicted
• Larger effect PM peak and mid-day than predicted
• too simplified trip timing distribution
• didnt account for vanishing return trips
• Why was AM peak effect overestimated?
• less cost sensitive car drivers?
• more people needed their vehicle?
• more people had arrival time restrictions?
• Revenue overestimation
• almost no data on vehicle types across the cordon
  taxis, buses, green cars

42
Problems with travel times in static network
models

• Link travel times are assumed to be independent
• in reality, queues block links and junctions
  upstreams
• Volume-delay-functions too generous
• too showrt travel time when congestion is very
  high
• cannot handle traffic volumes above max. capacity
• Dynamic (simulation) models necessary when
  congestion is high
• upstreams queue propagation
• queue buildup/dissipation
• also handles departure times

43
Theres nothing more practical than a good theory.