Preferred citation style

Axhausen, K.W. (2016) How many cars are too many? A second attempt, distinguished transport lecture at the University of Hong Kong, Hong Kong, October 2016.

.

HK DTL 2016

How many cars are too many? A second attempt

KW Axhausen

IVTETHZürich

October 2016

Acknowledgments

A Loder for the mobility tool ownership work

G Sarlas and R Fuhrer for the work on Swiss wages/productivity

L Sun for the big data analysis

FCL M8 for the SG MATSim model

HK DTL 2016

Singapore everywhere ?

HK DTL 2016

Some numbers first

HK DTL 2016

Some SG numbers: Mode shares by income 2008

HK DTL 2016

0%# 20%# 40%# 60%# 80%# 100%#

no#income#

1k#

2k#

3k#

4k#

5k#

6k#

7k#

8k+#

Mode#share#of#trips#

Income#[kSG]#

MRT,#LRT#

Bus#

Company,#school,#shuKle#bus#

Car,#motorcycle,#light#truck#driver#

Car,#motorcycle,#light#truck#passenger#

Taxi#

Current problems in Singapore

HK DTL 2016

Bus speeds in Singapore by time of day (2012)

HK DTL 2016

Sun,

201

3

Headways along a bus line in Singapore (2012)

HK DTL 2016

Sun,

201

3

A model of Singapore‘s travel demand and traffic

HK DTL 2016

What type of model would be enough ?

HK DTL 2016

Would this be enough ?

HK DTL 2016

mMFDqcarqbusqrail

vcarvbusvrail

AcccarAccbusAccrail

ncar

nGA

NumberPop, Firm

ProductivityIncome

taxcar

taxGA

feePT

feecar

budgettransport

%capcar %capbus%caprail

taxincome

What do we know ?

HK DTL 2016

Access and productivity: Switzerland

HK DTL 2016

AcccarAccbusAccrail

ProductivityIncome

Population accessibility by public transport: 2010

15Axha

usen

et a

l. , 2

015

Income levels: 2010

16Axha

usen

et a

l. , 2

015

Grey: less then 20 observationsPink to purple: Low to high wages

Spatial error model – (some variables not shown)

17Axha

usen

et a

l. , 2

015

2000 2005 2010

Y: Ln mean salary Estimate Sig. Estimate Sig. Estimate Sig.

Intercept 6.43*** 7.07*** 6.89***Ln car accessibility 0.01** 0.02 *** 0.01**Ln public transport accessibility 0.01** 0.01*** 0.01*Ln number of local employed 0.02 *** 0.01*** 0.01***From outside Switzerland -0.11 *** -0.09*** -0.09***Average duration in-post 0.00* 0.01*** 0.01***Ln average age 0.36*** 0.24 *** 0.32 ***Men 0.17*** 0.07*** 0.13***lamda parameter 0.33*** 0.41*** 0.40 ***Nagelkerke pseudo-R-squared 0.693 0.665 0.623# observations 1448 2298 2229

Accessibility and mobility tools: Swiss case

HK DTL 2016

AcccarAccbusAccrail

ncar

nGA

Accessibility and car ownership in Switzerland

HK DTL 2016

Switzerland: general accessibility

HK DTL 2016

Switzerland: Probabilities by general accessibility

HK DTL 2016

Switzerland: Probabilities by log of income

HK DTL 2016

Switzerland: Conditional probabilities by log of income

HK DTL 2016

Mobility tools and use: Swiss case

HK DTL 2016

qcarqbusqrail

ncar

nGA

Travel, car and season-ticket ownership (CH, 1984-2010)

HK DTL 2016

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0 0.5 1.0 1.5

Trips,with,motorized,vehicles/day

Public,transport,trips/day

Mikrozensus,Schweiz,1984

Mikrozensus,Schweiz,1989

Mikrozensus,Schweiz,1994

Mikrozensus,Schweiz,2000

Mikrozensus,Schweiz,2005

Mikrozensus,Schweiz,2010

Vehicle,and,season,ticket,,No,vehicle,,but,season,ticketVehicle,,but,no,season,ticketNeither

Fee and car ownership

HK DTL 2016

qcarfeecar

Singapore: COE Category B prices 2001 - 2013

HK DTL 2016 Quota12501000750500250

CO

E/M

ean

inco

me

1 .4000

1.2000

1.0000

.8000

.6000

.4000

.2000

.0000

3.001.50.50

Growth rate

Page 1

28

CH: Quality- and inflation adjusted car pricesSo

urce

: nac

hFr

ei (2

005)

0

200

400

600

800

1000

1200

1400

1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

Qualitätsbereinigter  Preisindex

 (2004  =  100)  [%]

Frei,  2004

Raff  und  Trajtenberg,  1990

HK DTL 2016

CH: Car always available by sex

29

0

20

40

60

80

100

0 10 20 30 40 50 60 70 80 90 100

Dirv

ing

Lice

nse

Ow

ners

hip

[%]

Average Cohort Age [Years]

Before 1910!

1910-29!

1930-39!

1940-49!1950-59!1960-69!

1970-79!

1980-89!

Men Women!

1990-99

Fleet size and speeds

HK DTL 2016

MFDqcar vcar

Macroscopic fundamental diagram MFD (Yokohama)

HK DTL 2016

Ger

olim

inis

and

Dag

anzo

, 200

8

Fleet size and speeds

HK DTL 2016

mMFDqcarqbusqrail

vcarvbusvrail

3d MFD (Zürich, FCD & loops) City centre

HK DTL 2016

Lode

r et a

l., 20

16

3d MFD (Zürich, FCD & loops) city centre

HK DTL 2016

Lode

r et a

l., 20

16

3d MFD (Zürich, FCD & loops) city centre - max speed

HK DTL 2016

Lode

r et a

l., 20

16

What we kind of know and what we don’t know

HK DTL 2016

What we kind of know

HK DTL 2016

qcarqbusqrail

feePT

feecar

What we kind of know

HK DTL 2016

taxcar

taxGA

feePT

feecar

budgettransport

taxincome

What we don’t know

HK DTL 2016

budgettransport

%capcar %capbus%caprail

What we don’t know

HK DTL 2016

mMFD

%capcar %capbus%caprail

What we don’t know

HK DTL 2016

vcarvbusvrail

AcccarAccbusAccrail

NumberPop, Firm

Further research questions

HK DTL 2016

• Close knowledge gaps for Switzerland

• Replicate results beyond Switzerland

• Add estimates of externalities

• Closed form optimisation model • For desired speed (accessibility) level• For welfare maximisation

Questions ?

HK DTL 2016

Appendix

HK DTL 2016

Estimation of models – Spatial error model – full model

45

Year 2000 Year 2005 Year 2010Independent Variable: Ln mean salary Estimate Pr(>|t|) Estimate Pr(>|t|) Estimate Pr(>|t|)Intercept 6.432 *** 7.068 *** 6.887 ***Ln car accessibility 0.010 ** 0.019 *** 0.011 **Ln public transport accessibility 0.014 ** 0.011 *** 0.012 *Ln number of local employed 0.016 *** 0.010 *** 0.013 ***Commuter from outside Switzerland -0.114 *** -0.094 *** -0.093 ***Short residence permit -0.236 *** -0.134 *** -0.226 ***Average duration in-post 0.003 * 0.008 *** 0.005 ***Ln average age 0.364 *** 0.237 *** 0.322 ***Men 0.169 *** 0.067 *** 0.132 ***Tertiary education 0.834 *** 0.663 *** 0.541 ***Professional training 0.553 *** 0.216 *** 0.324 ***Further vocational training 0.228 *** 0.171 *** 0.231 ***Teaching degree 0.197 ** 0.205 *** 0.321 ***Highschool diploma 0.601 *** 0.179 * 0.258 **Vocational training 0.074 *** 0.030 . 0.021Positions with highest demands 0.420 *** 0.385 *** 0.409 ***Positions with qualified indep. work 0.199 *** 0.246 *** 0.247 ***Positions with professional skills 0.135 *** 0.195 *** 0.140 ***Working (3rd sector) 0.214 *** 0.152 *** 0.056 .Working (other private sector) -0.096 *** -0.099 *** -0.059 ***Working (manufacturing) -0.226 *** -0.252 *** -0.107 ***Working (FIRE) 0.146 *** 0.006 0.085 ***Working (hotel, restaurants) -0.127 *** -0.132 *** -0.111 ***lamda parameter 0.331 *** 0.411 *** 0.402 ***AIC -2731 -4754 -4234AIC ols -2676 -4651 -4143Nagelkerke pseudo-R-squared 0.693 0.665 0.623Residuals' spatial autocorrelation -0.009 -0.009 -0.007OLS residuals' spatial autocorrelation 0.113 *** 0.103 *** 0.097 ***# observations 1448 2298 2229Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Axha

usen

et a

l. , 2

015

Model formulation 1/2

HK DTL 2016

ℒ(𝜶) = 𝛿'𝜙) 𝛽+𝑥-+,𝛽/𝑥-/,𝛽)𝑥-);𝑷𝟑 𝑑𝒙5 +   1 − 𝛿 : 𝜙/ 𝛽+𝑥-+,𝛽/𝑥-/;𝑷𝟐 𝑑𝒙5

𝒙𝒖𝒑

𝒙𝒍𝒐𝒘

𝒙AB

𝒙𝒍𝒐𝒘

Case Choice Probability1 None 𝑃+ =  Φ/(−𝑥+𝛽+;−𝑥/𝛽/;𝚸/)2 Car & no ticket 𝑃/ =  Φ/(−𝑥+𝛽+;𝑥/𝛽/;𝚸/)3 Car & local ticket 𝑃) =  Φ)(𝑥+𝛽+;𝑥/𝛽/ −𝑥) 𝛽);𝚸))4 Car & GA 𝑃F =  Φ)(𝑥+𝛽+;𝑥/𝛽/;𝑥)𝛽);𝚸))5 No car & local ticket 𝑃F =  Φ)(𝑥+𝛽+;−𝑥/𝛽/;−𝑥)𝛽);𝚸))6 No car & GA 𝑃G =  Φ)(𝑥+𝛽+;−𝑥/𝛽/;𝑥)𝛽);𝚸))

Choice environment

Likelihood function

Estimation method: • Maximum simulated likelihood in Stata using Newton Raphson technique• Using draws to compute the integral

Model formulation 2/2

HK DTL 2016

𝛿 Sample  selection  dummy,  equal  to  1  if  observation holds  season  ticket

ΦH N-­dimensional  cumulative  distribution  function  of  the  normal  distribution

𝜙H N-­dimensional  probability  density  function  of  the  normal  distribution

 𝛽 Parameters  of  the  model

 Σ Symmetric  correlation  matrix  with  typical  elements  𝜌KL  and  𝜌KK = 1.    The  same correlations  appear  in  both  Σ/ and  Σ) by  using  their  Cholesky decomposition  and  estimating  the  Cholesky factors  in  the  model

 𝛼 Parameter  vector  to  be estimated  that  contains  all  𝛽 and Choleskyfactors  of  Σ

𝒙NO,PQR Upper  and  lower  limits of  integration  domain,  determined  by  values of  each  observation

Switzerland: Ownership models (1/2)

HK DTL 2016

Season-ticket

owner

Car available

Age -0.059 *** 0.099 ***Age squared 0.052 *** -0.088 ***Male -0.132 *** 0.439 ***Working 0.066 *** 0.258 ***University level education 0.146 *** -0.054 **Log of monthly household income 0.075 *** 0.391 ***Center of agglomeration 0.132 *** -0.22 ***Constant 0.052 -6.039 ***

Switzerland: Ownership models (2/2)

HK DTL 2016

Season-ticket

owner

Car available

Local access to public transport: E -0.474 *** 0.505 ***Local access to public transport: D -0.348 *** 0.384 ***Local access to public transport: C -0.253 *** 0.286 ***Local access to public transport: B -0.097 *** 0.154 ***General accessibility 0.089 *** -0.028 ***Surplus public transport acc. -0.005 *** -0.066 ***Surplus workplace accessibility 0.729 *** -0.527 ***

Switzerland: GA given season ticket (2/2)

HK DTL 2016

Generalabonnement

Secondary residence 0.302 ***Log of monthly household income 0.128 ***Self-reported distance [1000km] 0.005 ***Constant -2.188 ***

Error correlations

Car available GASeason ticket -0.44 0.62Car available -0.24