Lessons from Southern California Congestion Pricing Experiments

1
Lessons from Southern California Congestion
Pricing Experiments

• David Brownstone
• University of California, Irvine
• (with Ken Small, Tom Golob,
• Arindam Ghosh, Jia Yan, Seiji Steimetz)

2
SR 91 Toll Facility

• Privately funded 4 lanes in 8 miles of median in
  congested link between Riverside and L.A. /
  Orange County.
• Complex fixed time-of-day and week pricing
  schedule designed to maximize profit (revenue).
  Current maximum toll is 8.25, but carpools get
  50 discount.
• Connects to much longer carpool lanes on both
  sides of facility.

3
(No Transcript)
4
SR 91 Problems

• Politically very unpopular since private owners
  enforced a no compete clause with the state
  highway agency to prevent widening the free
  lanes.
• Private owners sold at a loss to local
  transportation agency, which is now stuck with
  paying off construction bonds.

5
I-15 HOT lanes

• Publicly-funded 2 reversible lanes where solo
  drivers can pay to use carpool lanes.
• Links northern inland San Diego county to central
  business district.
• Real-time congestion pricing designed to keep
  free-flow in HOT lanes.
• Toll changes at most every 6 minutes, current
  maximum about 4.00
• Changeable message signs give at least 30 seconds
  notice of current toll charge.

6
(No Transcript)
7
I 15 Popularity

• I 15 toll facility is politically popular.
• Revenues fund an express bus system that
  primarily serves domestic help for wealthy
  residents.
• I 15 corridor users are high-income (especially
  compared to SR91)
• I 15 toll facility being expanded 15 miles north.

8
Toll Facility Commonalities

• Both use radio transponders to collect tolls,
  which requires users to establish accounts.
• Bypass very congested links on typically long
  (30-40 minute) commutes. Maximum time savings is
  about 10 minutes.
• Use of toll facility is voluntary there is a
  free alternative following exactly the same route.

9
Questions

• Are these facilities better than just letting all
  vehicles use the new lanes?
• Are the charges set optimally?
• Do these toll facilities reduce vehicle
  emissions?
• Do these toll facilities increase safety?
• Can the answers to these questions be generalized
  to other areas (Governor interested in converting
  all carpool lanes to HOT lanes)

10
Speed-Flow Curve

• Relates speed to traffic flow (the number of cars
  per lane-hour)
• As speeds increase, flow increases until
  following distance starts to decrease below
  drivers safety margin. (tailgating is good for
  traffic flow!)
• Speed-flow curve is reverse C shaped.

11
Note that Fmax 1500-2000 cars per hour on a
freeway lane
12
(No Transcript)
13
Congestion

• Congestion occurs when enough vehicles enter the
  road so that each additional vehicle lowers the
  traffic flow.
• This happens at the lower side of the speed-flow
  curve.

14
Optimal Congestion

• Occurs when the marginal benefit of reducing
  congestion marginal cost of reducing
  congestion.
• Marginal cost is typically the cost of adding new
  road capacity 5-10 million per lane-mile
• Therefore optimal congestion is greater than zero.

15
Too Much Congestion?

• Does private marginal benefit of reducing
  congestion social marginal benefit?
• No individuals do not account for the effect of
  their deciding to take a trip on other travelers
  trip times.
• Therefore there is too much congestion.

16
Solution Congestion Pricing

• Charge commuters entering congested roads an
  amount equal to value of time (VOT) lost to all
  other road users by their entering the road.
• This will depend on the VOT of all other road
  users, time and day, plus road conditions.
• When road is not congested, the optimal
  congestion toll is zero.

17
Second Best Problems

• When there is a free alternative, then optimal
  congestion toll is below VOT of other road users.
• Due to increased congestion on free alternative.

18
Congestion Pricing Not Popular

• Electricity especially residential
• Telephones
• Disneyland
• Grocery and other retail stores
• Differences in VOT suggest a market for premium
  checkout lines

19
Answers - Are these facilities better than just
letting all vehicles use the new lanes?

• Potentially yes since studies of commuters VOT
  on these corridors find considerable variation
  depending partially on income and length of trip.

20
Are the charges set optimally?

• No, since both toll facilities set prices so that
  toll lanes are flowing freely. Recall that due
  to free alternative there should be some
  congestion in toll lanes!
• Setting optimal charges would be very hard to
  sell politically.

21
Do these toll facilities reduce vehicle
emissions?

• Slight reduction (about 5) relative to allowing
  free use of the toll lanes, but this is mostly
  due to setting tolls too high!
• There would be essentially no emission reductions
  with modern hybrid technology.
• If comparison is made to original system, then
  emissions are higher since toll facilities are
  new capacity!

22
Do these toll facilities increase safety?

• No, since free alternative is so congested that
  it is hard to get in a serious accident.
• The worst accidents are more frequent when road
  is near capacity. This suggests that congestion
  pricing may reduce safety, but not by a very
  large amount.

23
Can the answers to these questions be generalized
to other areas?

• Probably yes.
• Key VOT results are similar for both SR 91 and I
  15 corridor in spite of large demographic
  differences and different survey modes (mail
  versus telephone).
• Model fit on I 15 data can accurately predict SR
  91 results.

24
How Should Revenue be Used?

• Optimally should be used to reduce income taxes
  (reduces labor market distortions).
• Political feasibility seems to require recycling
  back to corridor users. I 15 bus service lowers
  cost of domestic help to same people paying
  tolls!
• Revenue will not be sufficient to build toll
  lanes.

25
Value of time

• Marginal Benefit of reducing congestion is the
  value to road users of the time saved (VOT).
• For commercial users this is typically the value
  of time saved (wage) plus the inventory costs of
  goods delayed in transit.
• For work commute trips, VOT should be the
  opportunity cost of time wage.
• But majority of freeway trips are not work trips
  or commercial trips.

26
VOT for Commute Trips

• What is opportunity cost of commuting time?
• Work time?
• Home production?
• Leisure?
• VOT may not equal wage, and may vary across
  people and trips.

27
VOT Variability

• Unless VOT varies across commuters, things like
  the 91 toll lanes cannot be efficient (because of
  free alternative).
• We would expect that VOT for work commute trips
  increases with wage, but perhaps not linearly.

28
How to Measure Commute VOT

• Revealed Preference infer from commuters
  choices
• Mode Choice (bus, walk, bike, car)
• Route choice (especially toll vs. free)
• Stated Preference ask commuters to respond to
  hypothetical experiments where they can pay to
  avoid congestion.

29
Revealed Preference

• Reflects actual choice behavior but
• Very few situations where commuters can pay to
  reduce travel time usually bundled with
  something else like carpooling
• Hard to measure time savings and cost for actual
  trips since there is a lot of variability in most
  commute trip times

30
Stated Preference

• Example Suppose you could pay 10 to reduce the
  time it took you to drive to school this morning
  by 10 minutes. Would you take this option? __Yes
  __No
• Worse alternative How much would you pay to
  reduce your trip time by 10 minutes?
• Much cheaper than revealed preference studies.

31
Stated Preference Problems

• Respondent may not understand question, or may
  assume that the mechanism to reduce travel time
  is bundled with something else.
• Strategic Behavior respondent may lie to
  influence public policy they may overstate VOT
  to get more roads built
• This may explain urban rail results

32
Results from LA HOT lane studies

• Revealed preference from SR91 and I15 HOT lanes
  finds VOT about 20/ hour of work commute trip
  time saved. Varies with income and trip distance
• Stated preference studies find VOT estimates
  about 10/ hour

33
Identification Issues

• Use floating car to measure time savings.
• Assume commuters know distribution of time
  savings and toll level, but not actual time
  savings.
• SR91 variation in /time saved from fixed toll
  schedule and carpool discount. I15 variation from
  HOT bypass ramp at key congested intersection.

34
I15 HOT Lane Time Savings
35
 
 
 
 
 
36
 
 
37
SP/RP differences

• Difference persists across many different model
  specifications and very different SP collection
  methods (mail and CATI).
• Respondents are familiar with electronic toll
  collection and toll facilities.
• Respondents systematically perceive twice the
  actual time savings (more for women!).

38
Safety

• HOT choice bundled with perceived increased
  safety, and this might explain RP/SP differences
• Actually safer on regular lanes (due to slow
  speed)
• Steimetz (2004) models extra effort to avoid
  collisions in congested traffic, and finds this
  only accounts for 33 of RP/SP differences

39
Value of Reliability

• Commuters dont like uncertain travel time.
• Cost of being late higher than being early, so
  measure uncertainty by 90th 50th percentile of
  time savings distribution.
• Reliability is closely correlated with signaling
  function of real-time congestion tolls, so we
  cant identify VOR from I15 data.

40
 
 
41
VOR

• VOT accounts for 2/3 and VOR 1/3 of the service
  quality differential between free and express
  lanes on SR91.
• Women have roughly twice the VOR as men, which
  explains why women more likely to take toll roads
  on both SR91 and I15.