2008 OECD Survey on Household Environmental Behaviour: transport behaviour

Transcription

1 OECD Conference on Household Behaviour and Environmental Policy 3-4 June Paris, France 2008 OECD Survey on Household Environmental Behaviour: transport behaviour Alejandro Guevara-Sanginés José Alberto Lara-Pulido Universidad Iberoamericana, Ciudad de México June 4, 2009

2 Contents 1. Introduction - Rationale 2. Theoretical and empirical background 3. Data 4. Estimation 5. Policy implications and conclusions

3 1. Rationale The transport sector contributes with around 20% of total GG emissions in countries which belong to the UNFCCC. CO 2 emissions from transport are expected to more than double in the period to 2050 (Stern, 2006) In order to adopt cleaner transport policies is fundamental to understand how people take their transport decisions. An empirical analysis of the determinants of both, transport mode choice decisions and public transport use is presented in this work.

4 1. Rationale: Commuting is a joint decision involving destination, mode and car ownership Transport decisions We want to reduce the choice of more polluting options Destination Mode choice (& intensity of use) Mode availability/ ownership Joint decision Car Public Transport Other (Walking, cycling)

5 2. Theoretical and empirical background Since this is quite a complex task, different authors have approached to the issue in different ways: Train (1980) estimates the joint probability of having a certain level of car-ownership and choosing a certain transport mode. Johansson-Stenman (2002) use the Heckman s selection model to account for self-selection on driving distance and public transport use decisions. Bhat and Koppelman (1993) estimate a simultaneous system of equations that include car ownership, taking income and employment as endogenous variables. Guliano and Dargay (2005) specify a recursive system of equations which can be estimated independently to model travel decisions and car ownership.

6 3. Data Commuting mode to and from work OECD(10) 5% 21% 37% 38% Australia 1% 15% 35% 49% Canada 1% 20% 35% 44% Czech Republic 4% 26% 33% 36% France 1% 15% 31% 52% bicycle Italy 3% 12% 38% 47% public car Mexico 1% 22% 32% 45% walk Netherlands 10% 22% 27% 41% Norway 2% 14% 29% 55% South Korea 1% 27% 32% 41% Sweden 10% 20% 30% 41%

7 3. Data Characteristics of public transport Public transport characteristics (Order of preferences): 70% 60% 50% 40% 30% 20% 10% 0% Rapidity Rapidity Reliability Convenience and comfort. Safety (Important for Mexico) 70% 60% 50% 40% 30% 20% 10% 0% Not at all likely Not very likely Quite likely Very likely Security Not at all likely Not very likely Quite likely Very likely

8 4. Estimation The main objective of our work is to model mode choice and public transport use. Respectively, the dependent variables analysed are: 1. Mode choice: a vector indicating whether or not a person chose certain modes of transport to commute to a certain destination. Modes: walking, car, public transport, bicycle (non-mutually exclusive). Destinations: commuting to and from work, visiting family and friends, shopping, education, sports and cultural activities, and travelling to undertake usual professional activities. 2. Public transport use: measured as monthly expenditures in public transport as a share of income.

9 Empirical results (1): Continuous variables

10 Prob prof educ w ork visit sport Income is an important determinant of mode choice shop educ prof w ork 0.10 sport 0.05 visit shop ,564 2,923 5,461 10,202 19,062 35,615 66, ,329 Source: 2008 OECD Survey on Household Environmental Behaviour Incom e (Euros) Significant Not significant

11 Prob educ w ork prof sport visit shop Age reflects a life-cycle effect the probability of commuting by public transport educ prof w ork sport visit shop Source: 2008 OECD Survey on Household Environmental Behaviour Age Significant Not significant

12 Prob educ prof w ork The number of children has only a weak effect on mode choice (economies of scale?) prof 0.15 sport educ w ork visit shop sport visit shop Source: 2008 OECD Survey on Household Environmental Behaviour Number of children Significant Not significant

13 Prob Yet small, environmental 0.25 concerns do have an effect on transport decisions 0.20 educ educ prof w ork wprof ork sport sport visit 0.05 visit shop shop Source: 2008 OECD Survey on Household Environmental Behaviour Environmental concerns (index) Significant Not significant

14 Prob A similar pattern is 0.30 observed with respect to environmentally-friendly 0.25 habits 0.20 prof educ 0.15 w ork 0.10 sport visit 0.05 shop educ prof w ork sport visit shop Source: 2008 OECD Survey on Household Environmental Behaviour Recycling/reusing/reducing habits (index) Significant Not significant

15 Prob Finally, attitudes to 0.30 environmental issues also have an influence on 0.25 mode choice 0.20 educ prof w ork w ork educ prof sport sport visit 0.05 shop visit shop Source: 2008 OECD Survey on Household Environmental Behaviour Attitudes to environmental issues (index) Significant Not significant

16 Empirical results (2): Discrete variables

17 Living alone Change in predicted probability (base=married) 0.2 People who live alone are less likely to drive walk car public bicycle not signif work visit shop educ sport prof Source: 2008 OECD Survey on Household Environmental Behaviour

18 Employed full time Change in predicted probability walk car public bicycle not signif The inverse pattern is observed for people employed full time work visit shop educ sport prof Source: 2008 OECD Survey on Household Environmental Behaviour

19 Post-secondary education or highest Change in predicted probability There is no consensus across studies which should be the effect of schooling on mode choice walk car public bicycle not signif work visit shop educ sport prof Source: 2008 OECD Survey on Household Environmental Behaviour

20 Male that earns most in household Change in predicted probability Gender does not determine mode choice; however there are differences for someone who is a male AND earns the most in household walk car public bicycle not signif work visit shop educ sport prof Source: 2008 OECD Survey on Household Environmental Behaviour

21 Residence is a detached house Change in predicted probability It is assumed that those living in detached houses are more likely to live in outlying areas [of the CBD] (Nolan, 2002) walk car public bicycle not signif work visit shop educ sport prof Source: 2008 OECD Survey on Household Environmental Behaviour

22 Residence is in an urban municipality Change in predicted probability People in urban areas have to travel shorter distances, thus they are less likely to choose car walk car public bicycle not signif work visit shop educ sport prof Source: 2008 OECD Survey on Household Environmental Behaviour

23 There is no public transport station near residence Change in predicted probability With non-accesibility to public transport, people neither walk nor cycle they use cars walk car public bicycle not signif work visit shop educ sport prof Source: 2008 OECD Survey on Household Environmental Behaviour

24 Fixed effect by country Predicted probability of commute by public transport 50% 45% 40% 35% 30% 25% 20% 15% 10% work visit shop educ sport prof not signif. 5% 0% ne fr it cz swe nw au mx ko ca Source: 2008 OECD Survey on Household Environmental Behaviour People in Korea are the most likely to choose public transport.

25 4. Estimation Public transport use Public transport use Public transport was analysed specifying a Heckman selection model. First, the probability of choosing public transport is estimated. Then, the intensity of public transport use is estimated. (Measured in terms of public transport monthly expenditures as a share of income). Results from estimations are:

26 Variable Intensity of use Probability of selection lincome *** aprice *** retired *** female educated *** envpurch_index *** envattid_index ** *** encncrn_index ** Urban ** *** suburban ** *** children *** age *** age *** car *** comnopub *** can *** nld *** fra *** ita ** czr *** swe * * nor *** ** aus *** constant *** /athrho /lnsigma *** rho sigma lambda LR test of indep. eqns. (rho = 0): chi2(1) = 0.05 Number of obs 9,242 Censores obs 6,368 Uncensored obs 2,874 Log Likelihood Prob > chi2 (Wald) 0 Prob > chi2 = 0.816

27 % Probability of commuting by public transport vs. fuel price 29.5% 29.0% 28.5% Assumptions: -Average fuel price OECD(9) = Constant fuel efficiency factor = 7 km/ltr 28.0% 27.5% 27.0% 26.5% 26.0% 25.5% 1.4% Implicit Fuel price ( ) Fuel prices sources: (EU Countries, 2008), (Canada, 2009), (South Korea, 2007)

28 5. Policy implications and conclusions The way in which people make their commuting decisions is dependent on a number of socio demographic features such as age and income, but also on the way any given trip occurs in terms of rapidity, comfort, personal safety, etc., and on more permanent choices, such as the level of car ownership. Given these facts, changing the way in which people commute is not easy. 48 years old is related to the least prob. of commuting by public transport Prob of commuting by public transport is reduced by 2% when annual income increase by 1,000 euros An increase of 1 euro of price of fuel increase prob. Of commuting by public transport in 1.4%

29 5. Policy implications and conclusions While useful, it is claimed that traditional pricing measures have a limited scope mainly because of the rebound effect (Dargay, 2008). In this context, our results show that the probability of commuting by public transport is inelastic to fuel prices. However, there are other policies that can increase the use of public transport: Invest in accesibility, comfort and rapidity. Adopt an origin-destination approach to identify travel patterns Targeting younger and older people. Differentiatted fees. Implementing soft policies Public campaigns, educational programs.

30 Thank You

31 Independent variable Reference Car (Commuting mode) Income Age Male Education Children (# no.) Accessibility to public transport City size Attitude to env. Country Simma and Axhausen (2004) + Austria Golob and Hensher (1998)? + +/ 0 Australia Dargay (2005) GB Guevara and Lara (2008) + / OECD countries Public transport use Abreu e Silva et al. (2006)* + Portugal Johansson-Stenman (2002)* / Sweden Guevara and Lara (2008) + 0 (+) Public transport use (Commuting mode) 10 OECD countries Golob and Hensher (1998) + /+ /+ + + Australia Dieleman et al. (2002) + NL Guevara and Lara (2008) 0 / OECD countries * Measured as kilometres travelled.

32 4. Estimation Mode choice For each destination a conditional logit is specified. This model estimates the probability of choosing a specific mode of transport. P j e x j J i 1 e j x i i Estimation of the β is carried out by maximum likelihood. The set X includes socio-demographic characteristics, attitudes toward environment issues, characteristics of the trips and attitudes to public transport. Main results are presented next:

33 Fixed effect by country Predicted probability of walking 45% 40% 35% 30% 25% 20% 15% 10% work visit shop educ sport prof not signif. 5% 0% ne fr it cz swe nw au mx ko ca Source: 2008 OECD Survey on Household Environmental Behaviour

34 Fixed effect by country Predicted probability of commute by car 100% 90% 80% 70% 60% 50% 40% work visit shop educ sport prof not signif. 30% ne fr it cz swe nw au mx ko ca Source: 2008 OECD Survey on Household Environmental Behaviour

35 Fixed effect by country Predicted probability of cicyling 45% 40% 35% 30% 25% 20% 15% 10% 5% The Netherlands bicycle effect. work visit shop educ sport prof not signif. 0% ne fr it cz swe nw au mx ko ca Source: 2008 OECD Survey on Household Environmental Behaviour

36 Prob Number of adults effect is somewhat weak, yet significant and positive. -Some empirical shows that household size (a proxy of number of adults) has a negative effect, educ prof w ork sport visit shop however -When the number of children and being single dummy are excluded, a negative effect of household size is found probably indicating that this variable is a proxy of the latter educ prof w ork sport visit shop Source: 2008 OECD Survey on Household Environmental Behaviour Number of adults Significant Not significant