Price elasticities of travel demand - review and meta analysis

Institute for Transport StudiesFACULTY OF ENVIRONMENT

PRICE ELASTICITIES OF TRAVEL DEMAND

REVIEW AND META-ANALYSIS

Professor Mark Wardman

MOTIVATIONS

• Why Review?

Price elasticities are important for policy and forecasting

Benchmarking and new insights

Several previous reviews and update needed

• Why Quantify Relationships (Meta-Analysis)?

Conclusions beyond a single study (eg, over time, method)

Exploit results outside public domain

Model can provide elasticity estimates where no evidence exists

Classic literature reviews emphasise means

BACKGROUND: PREVIOUS REVIEWS

Numerous conventional reviews of price elasticities:

TRRL (1980), Goodwin (1992), Oum et al. (1992), Graham and Glaister

(2004), TRL et al. (2004), Wallis (2004), Litman (2010)

Goodwin (1992) - international evidence

50 bus fare elasticities from 21 studies

92 rail fare elasticities from 22 studies

120 car cost elasticities from 13 studies

Oum et al. (1992)

60 international studies of passenger demand

TRL et al. (2004) – local and suburban

33 UK studies and 11 Non UK studies of bus fare elasticities

35 UK and 20 Non UK rail fare elasticities

BACKGROUND: PREVIOUS META-ANALYSES

A few previous meta-analyses covering price elasticities

Hensher (2008) international evidence (price, time, headway)

Meta model 319 observations from 39 studies

Holmgren (2007) range of public domain elasticities

81 for public transport price

Wardman and Shires (2003)

104 UK studies 1968-2002 and 902 PT price elasticities

DATA COVERAGE

• 167 UK studies 1957 to 2010 yielding 1633 price elasticities

Published as: Wardman, M. (2014) Price Elasticities of Surface Travel Demand: A Meta-analysis of UK Evidence. Journal of Transport Economics and Policy, 48 (3), pp.367-384

• Follows on from major meta-analyses of:

1749 UK values of time from 226 studies

Published as: Abrantes, P.A.L. and Wardman, M. (2011) Meta-Analysis of UK Values of Time: An Update. Transportation Research A 45 (1), pp. 1-17

427 UK time elasticities from 69 studies

Published as: Wardman, M. (2012) Review and Meta-Analysis of U.K. Time Elasticities of Travel Demand. Transportation 39 (3), pp. 465-490

• List independent variables

EXPLANATORY VARIABLES

• Multiple observations (elasticities) per study where appropriate

• Key segmenting (e.g. possible influential) variables are:

Type of data and sample size

Mode and cost numeraire

Periodicity (dynamic, static, indeterminate), time period, lag structure

Level of aggregation and model type

Distance, region and type of flow

Journey purpose and/or ticket type

Year

Source

SOME SUMMARY MEASURES

Car: 77 values from 27 studiesRail: 1140 values from 102 studiesBus: 377 values from 60 studiesUnderground: 39 values from 11 studies

Times Series: Short Run 25%, Long Run 24%, Static 27% Choice: RP 4% SP 7%Other 13%

Urban 38% Inter-Urban 50% All 12%

1957-1970 4%1971-1980 21%1981-1990 19%1991-2000 47%2001-2010 9%

SUMMARY ELASTICITIES (I)

  CAR RAIL BUS

Time Series/Panel SR -0.30 (0.07) -0.69 (0.02) -0.45 (0.03)

Time Series/Panel LR -0.66 (0.13) -1.11 (0.03) -0.62 (0.04)

Time Series/Panel Static -0.35 (0.18) -0.73 (0.02) -0.36 (0.02)

Cross Sectional -0.61 (0.20) -1.01 (0.14) -0.50 (0.04)

SP Choice -0.20 (0.08) -0.89 (0.07) -0.58 (0.12)

RP Choice -0.13 (0.02) -0.57 (0.09) -0.30 (0.03)

TP -1.23 (0.10) -0.52 (0.06)

Urban -0.10 (0.03) -0.76 (0.04) -0.44 (0.04)

Inter-Urban -0.19 (0.04) -0.88 (0.02) -0.69 (0.11)

All Distances -0.39 (0.06) -0.93 (0.04) -0.72 (0.19)

SUMMARY ELASTICITIES (II)

  CAR RAIL BUS

Business -0.13 (0.05) -0.62 (0.06) -0.56 (0.14)

Commute -0.16 (0.04) -0.82 (0.19) -0.35 (0.04)

Peak -0.12 (0.0) -0.34 (0.03)

Leisure -0.13 (0.02) -1.05 (0.12) -0.41 (0.04)

Off Peak -0.55 (0.04)

Season   -0.71 (0.03)  

1st Class   -0.76 (0.06)  

Standard Class   -0.94 (0.03)  

Both 1st and Standard   -0.82 (0.02)  

Full Fare (Peak)   -0.69 (0.06)  

Reduced (Off-Peak)   -0.97 (0.06)  

SUMMARY ELASTICITIES (III)

  CAR RAIL BUS

1957-1970 -0.65 (0.39) -0.33 (0.02)

1971-1980 -0.16 (0.04) -0.81 (0.02) -0.43 (0.03)

1981-1990 -0.47 (0.09) -0.91 (0.04) -0.51 (0.04)

1991-2000 -0.25 (0.07) -0.85 (0.02) -0.48 (0.02)

2001-2010 -0.13 (0.02) -1.03 (0.11) -0.53 (0.05)

MODEL FORM

The meta-model takes the form:

jk

q

kjk

p

ji

Zn

ii eX

1

11

1

R2 goodness of fit of 0.64 seems reasonable. No outliers removed. Weighting made little difference 63 ‘study specific’ fixed effects

META ANALYSIS RESULTS (I)

Variable Coeff (t) Effect Variable Coeff (t) Effect

Data Type

Base=Static-Four   LR-Quarter(Quarter) n.s. -

Static-Quarter n.s. - LR-Quarter(Annual) n.s. -

Static-Annual 0.388 (7.2) +47.4% LR-Annual 0.544 (9.2) +72.3%

SR-Four -0.144 (2.9) -13.4% Monitor n.s. -

SR-Quarter(Quarter) -0.729 (4.3) -51.8% Cross-Rail 0.483 (2.4) +62.1%

SR-Quarter(Annual) n.s. - Cross-Bus 0.223 (2.0) +25.0%

SR-Annual n.s. - RP Choice n.s. -

LR-Four(Four) 0.345 (6.9) +41.2% SP 0.698 (8.5) +101.0%

LR-Four(Annual) n.s. - TP-Rail 0.793 (8.5) +121.0%

LR-Four(FourAnnual) 0.539 (3.7) +71.4% TP-Bus 0.291 (1.6) +33.8%

META ANALYSIS RESULTS (II)

Variable Coeff ( t) Effect Variable Coeff (t) Effect

Distance Category Area Type

Base=Urban Base=NonUrban

InterUrban-Rail n.s. - UrbanPTE-Rail -0.521 (8.3) -40.6%

InterUrban-Bus 0.228 (1.7) +25.6% UrbanPTE-Bus -0.260 (3.9) -22.9%

InterUrban-CarTrips 0.571 (3.8) +77.0% UrbanNonPTE-Rail n.s. -

InterNonLondon-Rail -0.075 (2.2) -7.2% UrbanNonPTE-Bus -0.206 (2.7) -18.6%

      London-Rail -0.466 (3.0) -37.2%

      London-Bus n.s. -

META ANALYSIS RESULTS (III)

Variable Coeff (t) Effect Variable Coeff (t) Effect

Mode and Ticket     Purpose    

Base=Bus     Base=Commute-Peak    

Car-Trips -1.496 (8.4) -77.6% Business -0.574 (6.4) -43.7%

Car-Km -0.744 (4.9) -52.5% Leisure-OffPeak 0.270 (3.5) +31.0%

Car-FuelCons -0.557 (1.9) -42.7% NonComm n.s. -

Rail-Seasons n.s. - NonEB n.s. -

Rail-First -0.232 (3.2) -20.7% All n.s. -

Rail-Standard n.s. - Car Cost    

Rail-Both -0.121 (3.2) -11.4% Base=Fuel    

Rail-Full -0.158 (1.8) -14.6% FuelParking 0.154 (1.5) +16.6%

Rail-Reduced 0.128 (2.0) +13.7% Total 0.568 (3.1) +76.5%

UG n.s. -      

META ANALYSIS RESULTS (IV)

Variable Coeff (t) Effect Variable Coeff (t) Effect

Price Index     Car Ownership    

LnPrice-Rail-NonSeason

0.179 (8.0) 0.179 LnCars-LeisKmFuel -1.240 (3.0) -1.240

LnPrice-Bus 0.068 (3.0) 0.068 Elasticity Type    

LnPrice-Car n.s. - Base=FullElasticity    

LnPrice-UG n.s. - ModeChoice-Leisure -0.276 (1.9) -24.1%

Study Source          

Base=Published          

Unpublished -0.099 (2.5) -9.4%      

IMPLIED RAIL ELASTICITIES

    ShortPTE

ShortLondon

ShortOther

InterLondon

InterNonLon

Season SR -0.18 -0.19 -0.31 -0.31 -0.28

Static -0.31 -0.33 -0.52 -0.52 -0.48

LR -0.36 -0.38 -0.61 -0.61 -0.57

PDFH LR -0.40 (-0.60) -0.60 -0.40 (-0.70) -0.60 (-0.75) -0.60 (-0.90)

Non SeasonFirst and Standard

SR -0.37 -0.39 -0.62 -0.62 -0.57

Static -0.63 -0.66 -1.05 -1.05 -0.98

LR -0.73 -0.77 -1.23 -1.23 -1.14

PDFH LR -0.70 (-0.85) -0.80 -0.70 (-1.00) -1.05 -1.20 (-1.00)

First LR       -1.10 -1.02

Full LR -0.70 -0.74 -1.19 -1.19 -1.10

Reduced LR -0.94 -0.99 -1.58 -1.58 -1.47

IMPLIED CAR ELASTICITIES

    Trips Urban Trips Inter

Km Fuel

Business SR -0.04 -0.07    

  Static -0.07 -0.12    

  LR -0.12 -0.21    

  DfT LR     -0.10  

CommuteandAll

SR -0.07 -0.12 -0.12 -0.15

Static -0.12 -0.21 -0.21 -0.25

LR -0.21 -0.36 -0.37 -0.45

DfT LR     -0.30  

Leisure SR -0.08 -0.14    

  Static -0.13 -0.23    

  LR -0.23 -0.41    

  DfT LR     -0.40  

IMPLIED BUS ELASTICITIES

    PTE Urban London & Rural

Inter

Commute SR -0.32 -0.34 -0.42 -

Static -0.55 -0.58 -0.71 -

LR -0.64 -0.68 -0.83 -

Leisure SR -0.42 -0.45 -0.55 -0.69

Static -0.72 -0.76 -0.93 -1.17

LR -0.84 -0.89 -1.09 -1.37

Total LR -0.77 -0.82 -1.00 -

CONCLUSIONS

• Price elasticities critical in transport planning, policy and pricing

• Largest meta-analysis of price elasticities ever undertaken

• Summarises a number of key relationships

• Considerable practical value

• Methodological insights

• How do price elasticity meta-analysis findings fit with:

time elasticity meta-analysis evidence?

value of time meta-analysis evidence?

• Would be nice to extend to other countries!