Mode Choice of Shopping Travel for Older People-V4b

7/23/2019 Mode Choice of Shopping Travel for Older People-V4b

1/29

Fengming Su

Dr Jan-Dirk Schmcker

Professor Michael G. H. Bell

Centre for Transort Stu!ies

"merial College# $on!on

%&

1


2/29

M'D( CH'"C( 'F '$D() P('P$( B(F')( *+D *FT() SH'PP"+G ,

* ST%D "TH $'+D'+ D*T*

Fengming Su# Jan-Dirk Schmcker an! Michael G. H. Bell

Centre for Transort Stu!ies# "merial College# $on!on# %&

[email protected]

*/stract

With the population aging in many countries, older peoples travel is recently getting more

attention in the transportation literature. However our understanding of factors influencing

their mode choice is still limited. In this research the focus is on mode choice for shopping

trips as these are the most frequent trips of older people. The study is not limited to shopping

trips, but also investigates the mode choice of the trips after shopping trips. Two types of

modelsthe multinomial logit model and the nested logit model are fitted to !T" data to

estimate the effects of various factors on the mode choice of older people for shopping travel.

The appropriateness of these models and the implications of the findings are discussed.

0 "ntro!uction

In the past transportation mode choice research has focused on commuting #$hat %&&'($hat

) "ardesai *++(-e alma ) /ochat %&&&(alma ) /ochat *+++("chwanen *++0("chwanen

) 1o2htarian *++3(Titheridge ) Hall *++4, but not a lot of research on shopping travel has

been done. It might be that in transportation research shopping travel is not perceived to be

as important to analyse as commuting, because shopping travel is not generally carried out

on a daily basis. 5urther, shopping travel has less time and location constraints, all of which

ma2e it more difficult to analyse. However recently shopping travel is getting more attention

2
mailto:[email protected]:[email protected]


3/29

#$awa ) 6hosh %&&&($hat %&&($hat %&&7b($hat, "iva2umar, ) !8hausen *++9(:ubu2cu

*++%(;im ) ar2 %&&'(/obinson ) ected to rise to *9 per

cent by *+9%. ?specially the population group of the @oldest old is growingA eople aged 73

and over have grown by 70 per cent between %&7% and *++0, to over %.% million #Bffice for

Cational "tatistics *++34. This increase of the older population gives new impetus to the need

to better understand older peoples travel behaviour.

5or older people, shopping travel has probably a more important meaning than for younger

people. Blder people ma2e less wor2 related travel and the ma>ority of older peoples travel is

for shopping #0D of going out trips for the over 3s are for shopping according to the ondon

!rea Travel "urvey *++% dataset4.5ew studies have loo2ed at this aspect. "ome research on

older peoples mode choice was carried out not distinguishing trip purposes but therefore

often overloo2ing the specific characteristics of shopping trips #Wilds, %&70( ;im, *++0( etc.4

The current paper e8amines older peoples mode choice for shopping travel. It allows for the

fact that mode choice is often not determined by the single trip but by the trip chain. Therefore

the analysis of mode choice for tours #and not single trips4 including shopping trips is the

focus of this paper. 1ost tours are composed of two trips( if there are more than * trips, it is

complicated to consider all the mode choice combinations in the tours. In this paper, in order

to simplify the analysis, only mode choices of shopping trips and the trips after shopping trips

are loo2ed into.

3


4/29

1 $iterature )e2ie3

! variety of past research has been done on shopping travel and older peoples travel. !t first,

some studies defined shopping travel so as to distinguish it from travel for other purposes.

6iuliano et al. # *++94 defined E"hopping tripsF as trips for purchasing commodities and

windowGshopping. $arber # %&&34 defined a shopping trip as a trip to any retail centre,

irrespective of the sie and type of the store or shop, and whether a purchase is made or not.

In the hierarchy of a household threeGway classification of activities #6olob ) 1cCally %&&'4,

shopping does not have a fi8ed time and distance, unli2e subsistence activities such as wor2

or education trips. In this paper, shopping is defined as a maintenance activity. -espite less

strict time constraints, most households have some regularity regarding shopping travel

behaviour in order to cover their daily needs. !lthough not as frequent as wor2 or education

trips, shopping trips still comprise a large portion of all household trips. This highlights the

importance of shopping travel.

"hopping travel is different from other 2inds of travel purpose for several reasonsA 5irstly, it

does not have strict time constraints as long as it is within the time window of shop opening

hours and can satisfy the maintenance requirements. "econdly, after shopping, generally

there would be a load to carry, which increases travellers difficulties. $oth of these aspects

might influence the mode choice.

There have been some studies on mode choice of shopping travel. 1ode choices of shopping

travel are diverse, and car driving is in general found to be the most frequently used mode.

#6ould, 6olob, ) $arwise %&&74. In the =; for e8ample although some shopping trips are

made by foot, mass transit, or bicycle, the car accounts for more than '0D of shopping trips.

#-epartment of Transport %&&4. Trips for shopping are one of the ma>or uses of the

4


5/29

household vehicle. In the =;, %*D of all mileage and *+D of all car >ourneys are underta2en

for shopping #-epartment of Transport %&&4. These statistics should be interpreted with care.

5irst, shopping trips are often combined with other types of travel, li2e the trip home from wor2

#$hat %&&'b4. In some travel statistics, trips ta2en for multiple purposes are not crossG

classified. "econdly, many shopping trips involve multiple stops, and there is some confusion

as to whether these should be categoried as separate shopping trips.

It is found that shopping mode choice is correlated with socioGeconomic variables, and

shopping mode choice is not the same across different population sectors. The socioG

economic variables influencing shopping mode choice include gender, employment status,

age etc. Handy # %&&4 compared shopping mode choice between men and women, and

women from different household types. Household socioGdemographic variables are found to

have more significant effects on mode choice of shopping than gender. $hat # %&&7a4 found

that employed individuals are more li2ely to drive alone for shopping trips than unemployed

individuals. :ompared to younger people those aged +G3 are also more li2ely to drive alone

for shopping, as was pointed by "chmc2er et al # *++4, but this trend reverses sharply for

individuals over 3 years of age. eople who are older than 3 are more li2ely to use public

transport or wal2ing instead. Whereas "chmc2er et al ignored mode choice for shopping

trips chains, this researchloo2s into older peoples mode choice for shopping trips and trips

following shopping trips on the same tour.

1ost past research on older peoples mode choice does not focus on shopping travel. 1any

factors have significant effects, including transport service characteristics. Wilds et al # %&704

concluded that older passenger perception of the reliability of dialGaGride and bus transit and

the accessibility of bus transit are primary factors that affect the decision for wor2ing trips and

shopping trips. !lsnih # *++94 identified the role of conventional and specialised public

5


6/29

transport as an alternative to the automobile in meeting mobility and accessibility needs of

older people. The effects of socioGeconomic variables and trip characteristics are also

significant. :hen et al # *++94 found that car availability, fare and group type affect mode

preference between paratransit services. ;im et al # *++04 e8amined not only the effects of

neighbourhood and trip characteristics but also personal and household characteristics on the

mode choice of older people. !lthough automobile and public transit have received much

attention, this does not mean wal2ing is not important. This paper loo2s at the importance of

wal2ing in shopping trips for older people.

!ccording to ondon data, around '&D of shopping trips are shorter than 9.3 2m #!T",

*++%4. /ietveld # *+++4 pointed out that the share of shortGdistance trips is large #about 3+D

of all trips are shorter than 9.3 2m4 and in this range wal2ing and bi2ing indeed have modal

shares of more than 3+D. We can e8pect a modal share of more than 3+D for wal2ing and

bi2ing in shopping trips from !T" data. $ut according to the data, a lot of wal2ing only

occurred within a trip chain. The main mode choice for the ma>ority of the trip chains #here

defined as a sequence of trips starting or ending at the home, namely a homeGbased tour4 is

the automobile or public transit( wal2ing is used as a supplement to move from one shop to

another shop nearby. In this paper, analysis of mode choices of trips for shopping and after

shopping accounts for these situations.

4 Data

The analysis in this paper is based on an interim version of the London Area Travel Survey

2001#!T"4, provided by Transport for ondon #Tf4. !T" has data on ',*3* individuals in

*&,&'9 households based on home interviews throughout the 6reater ondon !uthority and

6


7/29

some neighbouring districts.%The survey includes four main datasets for each individualA

Household information( personal information( details on vehicles owned by the household(

and trip details of all trips done on one wee2day. !ll interviews were done on a personal basis

and the respondents were as2ed to fill in a oneGday travel diary. In total, %',039 trips were

made by respondents. !T" data only includes trips in and around ondon and does not

include holiday trips leaving the region. 5rom the interim !T" dataset records all tours made

by persons aged 3 or older #&%+& tour records made by *3% individuals4 were e8tracted.

5rom these people, only complete home based tours are used for the analysis, which means

only older people whose first trip is from home and last trip is to home on the day are

included. "ince this analysis is for shopping travel, only older people tours with at least one

shopping trip are suitable for the analysis. This reduces the data set to 9&%3 older people and

39& tours. In total these tours consist of %9%* trips of which 3%7% are shopping trips.

The second dataset used in this analysis is the new Inde8 of 1ultiple -eprivation *++0 #I1-

*++04, which is a "uper Butput !rea #"B!4 level measure of multiple deprivation and is made

up of seven "B! level -omain Indices. The model of multiple deprivation which underpins the

I1- *++0 is based on the idea of distinct dimensions of deprivation which can be recognised

and measured separately and are e8perienced by individuals living in an area. eople may be

counted as EdeprivedF in one or more of the domains depending on the number of types of

deprivations that they e8perience. The overall I1- is conceptualised as a weighted area level

aggregation of these specific dimensions of deprivation.

The new I1- *++0 contains measures of deprivation which relate to income, employment,

health and disability, education, s2ills and training Ebarriers to housing and servicesF and

finally Eliving environment and crimeF. There are further two supplementary indicesA EIncome

1The Greater London Authorit includes the 33 London !oroughs.

"


8/29

-eprivation !ffecting :hildrenF and EIncome -eprivation !ffecting Blder eopleF which are

subsets of the Income -omain. The latter one is used in the following analysis.

The deprivation data is at the "B! level. $ut the data from !T" is at the level of three digit

postcode. In order to correlate the two datasets, the I1- indices for "B! are aggregated to

arrive at indices for the larger three digit postcode level weighting by population as followsA

=ji j

ii

j

PCSOA PC

SOASOA

PCPOP

POPIMDIMD

#

wherej is a threeGdigit postcode area and iare the "B! areas within the postcode areaj.

!fter correlating these databases, records with missing variables are e8cluded.

5 Descriti2e statistics

eople aged between 93 and 30 are most li2ely to have a >ob and dependent children. Those

aged between 33 and 0 are in many cases still in full time employment but their lifecycle has

changed as often their children are now living independent lives. In this study, this age group

is referred to as the EpreGolderF. Blder people, here defined as anyone aged 3 and over are

compared with these two sections. revious research further indicated that it is often useful to

divide older people into two sectionsA Eyounger oldF #3G'04 and Eolder oldF people #'3J4

because at an age of around '3 often a decline in health limits the feasible activities #!lsnih

and Hensher, *++%4

5or shopping travel, mode choice across different population sections is different. 5igure %

shows that older people are more li2ely to go shopping as a car passenger or by bus rather

than driving themselves.

$


9/29

Figure 1 Modal share across population sections

The difference of mode choice between older and younger people for their shopping travel

shows clearly that the travel demand of older people could not be ta2en as the same for

younger people. !nalysis on this specific population sector is necessary. In the following

model estimation, older people #3 or above4s shopping mode choice and tour type are

loo2ed into.

:onsidering that people might not use the same mode when travelling to or from their

shopping so>ourn means that there are in total the %% alternatives shown in Table %. The table

indicates that most trips before and after shopping are made by the same transport mode, in

which the most frequent alternative for older people is wal2ing, with public transport following.

Wal2ing is more li2ely to be used with another transport mode. When different modes are

used before and after shopping, the combination is more li2ely to be wal2 and public

transport. This is reasonable, since if a car is used, the traveller would be reluctant to wal2.

%


10/29

Table 1 Combined mode choice to and from shopping

!fter loo2ing at mode choices, the relationship between mode choice and tour type will be

investigated. "everal terms describing travel activities are used in this paper. The most

elemental unit of travel activity is the sojourn, which is a visit to a place remote from home.

The activity which an individual pursues during a so>ourn can be either wor2 or nonGwor2(

multiple activities may be pursued >ointly on a single so>ourn. The travel which carries an

individual between his home and a so>ourn or between temporally consecutive so>ourns is

called a trip. ! set of consecutive trips which begin and end at an individuals home is called a

tour#!lder ) $enG!2iva %&'7($owman ) $enG!2iva *+++4. If a tour is composed of * or more

activity locations, it is defined as a trip chain. There might be one or more than one EtourF per

day. !ccording to trip number and travel purpose, the type of tour is decided. In section , tour

type is included in the model in the form of three dummy variablesA shop only tour or not, two

trips per tour or not, and travel purpose after shopping.

5igure *, 5igure 9, and 5igure 0show the modal share made by older people distinguished by

type of tour. The results indicate that tour type has significant effects on mode choice. "impler

the tour type is #less purpose, less stops4, mode choice less li2ely changes.

1&


11/29

Figure 2 Modal share of different tour types-1


11


12/29


6 Descrition of Discrete Choice Mo!els

3.% 1C

$ierlaire # %&&'4 provides a detailed review of discrete choice models. -iscrete choice theory

has played an important role in transportation analysis for the last *3 years. 5undamental to

this is the random utility functionA

njnjnj VU += KKKKKKKKKKKKKKKKKKKKKKKKKKK#%4

-ifferent assumptions about the random and deterministic terms nj and njV respectively

will produce specific choice models. The deterministic part of the utility function is e8pressed

as a linear combination of attributesA

njnjnj xV = K.........................................................................................................#*4

wherenj

x

is a vector of independent e8planatory variables for individual nand alternativej,

and nj is a vector of parameters to be estimated.

=nder the random utility interpretation the alternative with the highest utility is assumed to be

chosen. Therefore, the probability that alternative ifrom choice set : is chosen by decisionG

ma2er nis equal to

12


13/29

( ) ( )CjUUPCiP njni = ' KKKKKKKKKKKKKKKKKK..#94

If the error terms are independently and identically 6umbel distributed, with location

parameter + and scale parameter , according to the multinomial logit model the probability

that a given individual will choose alternative iis given by

=

Cj

V

V

Cj

i

e

eiP

()KKKKKKKKKKKKKKKKKKKKKKKKKK#04

3.* Cest ogit 1odels

The nested logit model is an e8tension of the multinomial logit model designed to capture

correlations among alternatives. It is based on the e8clusive and e8haustive partitioning of the

choice set Cinto several nests kC

. The utility function of each alternativei

is composed of a

term specific to the alternative, a term associated with the nestkC

Vand a random termA

iCii kVVU ++=

KKKKKKKKKKKKKKKKKKKKKKKK#34

The nested logit model is obtained by assuming that the unobserved utility i has a

cumulative distribution #Train *++*4A

=

=

K

k Ci k

kieF1

*e+p()

KKKKKKKKKKKKKKKKKKKK..#4

5or any two alternativesj

andm

in nest kC

,j

is correlated with m

. 5or any two

alternatives in different nests, the unobserved portion of utility is uncorrelated. The parameter

kis a measure of the degree of independence in the unobserved utility among the

13


14/29

alternatives in nestk

. ! higher value ofk

means greater independence and less correlation.

When

k

L % for all k, representing independence among all the alternatives in all nests, the

nested logit model reduces to the multinomial logit model. The value ofk

must be within a

particular range for the model to be consistent with utility ma8imiing behaviour. If kk is

between ero and one, the model is consistent with utility ma8imiation for all possible values

of the e8planatory variables. kk *= , so scale parameter 1>k .

5ollowing $enG!2iva and erman # %&7'4, we can e8press the nested logit model choice

probability as a product of marginal and conditional choice probabilities. 5or the twoGlevel

choice problem, the nested logit model is written asA

=+

+

==K

l

IV

IV

Cj

V

V

CCiiClllC

kkkC

k

kj

ki

kkk

e

e

e

ePPP

1

*

*

* ##

KKKKKKKKKKKKKKKK#'4

where

=

k

kj

Cj

V

k eI *

lnKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK#74

andkkI

is the e8pected utility that a decisionGma2er receives from the choice among the

alternatives in nestkC .

14


15/29

7 (mirical *nal8sis

7.0 Mo!el attri/utes

The overall ob>ective of this study is to understand the effects of different factors on older

peoples shopping travel mode choice. 5or the calibration of the various models discussed in

this article, the estimation software $IB6?1? #$ierlaire, *++94 was used. This estimation tool

can be used for all types of closedGform as well as mi8ed 6?< model structures. 5urthermore,

the program can accommodate nonGlinear utility functions.

:ombined mode choice of trips before and after shopping is the dependent variable. There

are mainly five sets of independent variables included in the model estimation. 5irst of all,

travel cost and travel time of the two trips before and after shopping( secondly, tour type

characteristics e8plained by stop number and travel purpose etc( thirdly, transportation

service system characteristics( fourthly, individual and household socioGeconomic variables

and fifthly, the Inde8 of 1ultiple -eprivation and the separate Indices of -eprivation.

5rom the e8tracted data, trips made by car driving, public transport, car passenger, and wal2

are ta2en separately to estimate linear regression models to analye travel time by various

transport modes #Table *4. If a mode is not chose, travel time is calculated by the following

formulas.

Table 2 Calculated travel time by different modes

The above results are used to estimate the travel times of the unchosen alternatives. The

travel cost for these unchosen alternatives is then estimated as followsA

15


16/29

When older people have a freedom pass the trip cost is assumed to be + otherwise

the public transport cost is ta2en as M%. The 5reedom ass allows free travel on

ondonNs public transport for people aged + or over and for people with certain

disabilities who live permanently in a ondon borough. "ingle fare of public transport

is 7+p for bus and M%.3 pounds for underground. In the following models bus and

underground are not distinguished so that a public transport cost of M% is assumed.

It is assumed that car passenger costs are free since "chmc2er et al #*++4 find

with the same data set that this is a reaonsable assumption

5urther following "chmc2er et al # *++4, the marginal car costs are estimated at

%*p per mile #L '.00 pence O 2m4 plus associated par2ing cost. !T" contains data on

the nature of par2ing for chosen trips. 5or all the missing car trips and for the

unchosen alternatives, the charge for par2ing in both Inner ondon and Buter ondon

is estimated at M9 and M% respectively. In summary, the total cost for driving is

estimated asA '.0pence P travel distance #2m4 J par2ing cost.

Trip characteristics included in the model are travel purpose after shopping #going home or

not4, travel purpose in a tour #shop only or not4, simple tour or not. The number of tours on the

day surveyed was included in a model, but the results are not significant. Transportation

service system characteristics include bus accessibility, bus served destination, bus headway,

underground accessibility, underground served destination, underground headway. These

public transportation system characteristics are only included in mode choices involving public

transport. Household and individual socioGeconomic variables included in model estimation

are age, travel disability, household income #over M*3+++ or not4, household location #Inner

ondon or not4, gender #male and female4, household vehicle access. 5ollowing findings from

!lsnih and Hensher #*++94 the EyoungerGoldF#+G'04 and the EolderGoldF #'3J4 are

distinguished and a dummy variable for olderGold is introduced.

16


17/29

7.1 )esults from multinomial logit mo!el

5irstly a 1C model was first fitted to the data( the estimation results for this model are

reported inTable 0. Trips made by driving only #to and from the shopping place4 are the

reference category. !ll mode specific constants are negative, which means, all else equal,

ma2ing the whole >ourney by driving is the preferred mode. :oefficients of travel cost of

shopping trips, and travel cost of the trips after shopping are significantly negative, which is

e8pected. There are other models estimated including travel cost and travel time specific to

different mode choices. $ut the coefficients for travel time and travel cost are not significant

and quite similar to each other, and the models do not have big improvements comparing final

values of logGli2elihood. Transportation service system characteristics are only specific to

alternatives involving public transport. The model further shows that higher bus accessibility

has a positive effect on choosing public transport, and longer bus headways have negative

effects on public transport choice.

If the traveller went home after shopping then driving J wal2ing, public transport J wal2ing,

car passenger J wal2ing are less li2ely mode choices, instead wal2ing J driving, wal2ing J

public transport, wal2ing J car passenger are more li2ely to happen. This is reasonable, as if

a person used car or public transport for going out, then they par2ed the car or got off from

the public transport for so>ourn !, after that they wal2ed from ! to a shop, another stop in the

trip chain, after shopping the last trip going home are quite li2ely to change bac2 to the

original transport mode. "o wal2ing is less li2ely to be the mode choice of going home if

public or private transport is used in the same tour. If the tour is not only composed of

shopping travel, driving is more li2ely to be used. The e8planation might be that tours with

more purposes are more complicated and therefore driving is more convenient for the

1"


18/29

traveller. If the tours are simple, i.e. only two trips, older people are more li2ely to choose

public transport or car. If the tour has more trips, wal2ing is more li2ely to be used to connect

different so>ourns.

If the traveller has a travel disability, public transport is less li2ely to be chosen which confirms

the results of "chmc2er et al #*++4. Households with higher income are less li2ely to use

public transport. 1ales are less li2ely to use public transport and travel as a car passenger

#car passenger J wal2, wal2 J car passenger4. !s e8pected the model indicates that having

car access means older people have more chances to be a car passenger #car passenger J

wal2, wal2 J car passenger4.

In this section, the I1- score and each domain score including one subGdomain score is

added into the multinomial logit model separately to estimate the influence of deprivation on

older peoples mode choice. The inclusion of I1- or domain score does not improve the

model significantly( final logGli2elihood was only improved less than %+ units after the inclusion

of another & variables. The coefficients of I1- or domain score are not significant either.

$ut most of the signs of the coefficients are reasonable.

Blder people from more deprived areas #higher deprivation scores4 are less li2ely to

choose the >oint choice of private transport and wal2ing. These older people have

less chance of using private transport.

Blder people from areas with less income, lower employment rate, less accessible

service facilities, and worse living environments are less li2ely to choose the >oint

choice of private transport and wal2ing. They have less chance of using private

transport and the environment does not allow them to use wal2ing more.

1$


19/29

Blder people from areas with higher health and disability deprivation scores are less

li2ely to use public transport and wal2 #wal2ing before andOor after shopping4 but are

more li2ely to be a car passenger. Blder people from areas with more crime

deprivation are less li2ely to wal2.

$ut the Income -eprivation !ffecting Blder eople Inde8which is a subset of the Income

-omainhas significant effects on older peoples mode choice #Table 34. Blder people from

areas with more deprived income affecting older people are more li2ely to use public

transport.

Table 3 Multinomial Logit Model iagnosis

1%


20/29

Table 4 Multinomial Logit Model !stimation "esult #bold number sho$s %&' significance level( )talian number sho$s %*' significance level+

2&


21/29

Table & !ffects of the )ncome eprivation ,ffecting lder .eople )nde/

21


22/29

7.4 +este! logit mo!el

To account for the potential e8istence of hidden correlation between some of the alternatives,

several separate nested logit models were estimated with the same data. The framewor2 of

the two models with significant nests and best model fit is shown in #5igure 34A

%. 1ode choice of trips before and after shopping not changed #E"ame 1odeF4(

Cest !A ublic or private transport mode before shopping and wal2ing after shopping

#Cest EWal2 from shopF4(

Cest $A Wal2ing to shopping center and change to a private or public transport mode

after shopping #Cest EWal2 to shopF4

*. "ame mode choice before and after shopping #E"ame 1odeF4(

Cest !A -ifferent mode choices before and after shopping #Cest E-ifferent 1odeF4

The first and second nesting resulted in structural parameters that are all greater than %. The

results are shown inTable ,Table ', Table 7,andTable &. With the first nested logit model

structure, the nesting parameters ta2e a value of %.0+ and %.00 #both significantly different

from % according to the tGtest *.*+ and *.04, implying a correlation between the unobserved

utilities around +.0& and +.3*. With the second nested logit model structure, the nesting

parameter ta2es a value %.9+ #significantly from % according to tGtest 9.+'4, which implies a

low correlation of around +.0%. !side from a difference in scale, the substantive results of the

two models and the multinomial logit model are very similar. 5inally, in terms of model fit, the

results do not show a very significant increase in logGli2elihood.

22


23/29

Table 0 ested logit model 1 diagnosis

Table ested logit model 2 diagnosis

23


24/29

5igure 3A 5ramewor2s of two nested logit models

,ri-e

.T

.assenger

/alk

ame Mode5 est 6al7 from shop5

est 6al7 to shop5

.T

.T

,ri-e ,

ri-e

.assenger

.assenger

.T

.T

.T

,ri-e

,ri-e

,ri-e

.assenger

.assenger

.assenger

/alk

ame Mode5

estifferent Mode5

24


25/29

Table 8 !stimation "esult of ested Logit Model 1 #bold number sho$s %&' significance level( )talian number sho$s %*' significance level+

25


26/29

Table % !stimation "esult of ested Logit Model 2 #bold number sho$s %&' significance level( )talian number sho$s %*' significance level+

26


27/29

9 Conclusions

In this article, models of mode choice of shopping travel appear to produce some interesting

results. "ome significant findings from the study are summarised belowA :oefficients for travel cost have the e8pected negative sign. The coefficients for travel

time are close to +, which might also be not surprising given that the data is only for

people older than +. However further wor2 should loo2 at speed, price per unit

distance #poundsO2m4 and distances travelled for shopping.

$etter public transportation service especially better bus services could improve older

peoples public transport use. If the bus services have better accessibility and higher

frequency, older people are more li2ely to choose mode choice involving the use of

public transport. 5rom another aspect, current traditional public transport service

could not satisfy older peoples travel demand, but the improvement of the service is

quite costly. "o special transport services are another method to fulfil older peoples

travel requirements.

The pattern of shopping tour has significant effects on older peoples travel mode

choice. We also could say that mode choice has significant effects on older peoples

shopping tour pattern. If older people have more mode choice fle8ibility, they could

choose different shopping tour pattern more fle8ibly. If older people want to travel in

more tour patterns, they need have access to more mode choices.

Blder people differ within the population section. With various socioGeconomic

variables, they have different travel demand( they have different mode choice

preference. It is not wise to treat them as same, giving same transport services.

roviding corresponding transport services according to their different transport

demands are more reasonable. The following points outline mode choice preferences

of older people with different socioGdemographic variables clarified by the model

estimation.

Blder people with higher income are less li2ely to use public transport.

Blder people living in inner ondon are less li2ely to be a car passenger or combine

wal2 and public transport. !s e8pected, older people having access to vehicles are more li2ely to be car

passengers.

1ale older people have less chance of using public transport or being car passenger.

Blder people from more deprived areas are more li2ely to use public transport.

2"


28/29

/eferences

!lder, T. ) $enG!2iva, 1. %&'7, Qa theoretical and empirical model of trip chaining behaviorQ,Transportation research B, vol. %9$, pp. *09G*3'.

!lsnih, /. ) Hensher, -. !. *++9, QThe mobility and accessibility e8pectations of seniors in anaging populationQ, Transportation Research Part A: Policy and Practice , vol. 9', no. %+, pp.&+9G&%.

$arber, 6. %&&3, Q!ggregate :haracteristics of =rban Travel,Q in The eo!raphy o" #r$anTransportation, ". Hanson, ed., The 6uilfor ress, Cew Ror2, pp. 7%G&&.

$awa, ;. ) 6hosh, !. %&&&, Q! 1odel of Household 6rocery "hopping $ehaviorQ, %arketin!Letters, vol. %+, no. *, pp. %0&G%+.

$enG!2iva, 1. ) erman, ". %&7', &iscrete Choice Analysis1IT ress.

$hat, :. /. %&&, Q! haardGbased duration model of shopping activity with nonparametric

baseline specification and nonparametric control for unobserved heterogeneityQ,Transportation research B, vol. 9+, no. 3, pp. %7&G*+'.

$hat, :. /. %&&7, Q!nalysis of travel mode and departure time choice for urban shoppingtripsQ, Transportation Research Part B: %ethodolo!ical, vol. 9*, no. , pp. 9%G9'%.

$hat, :. /. %&&', QWor2 Travel 1ode :hoice and Cumber of ConGwor2 :ommute "topsQ,Transportation Research Part B: %ethodolo!ical, vol. 9%, no. %, pp. 0%G30.

$hat, :. /. ) "ardesai, /. *++, QThe impact of stopGma2ing and travel time reliability oncommute mode choiceQ, Transportation Research Part B: %ethodolo!ical, vol. 0+, no. &, pp.'+&G'9+.

$hat, :. /., "iva2umar, !., ) !8hausen, ;. W. *++9, Q!n analysis of the impact of informationand communication technologies on nonGmaintenance shopping activitiesQ, TransportationResearch Part B: %ethodolo!ical, vol. 9', no. %+, pp. 73'G77%.

$ierlaire, 1. %&&', &iscrete choice 'odels.

$owman, S. . ) $enG!2iva, 1. ?. *+++, QactivityGbased disaggregate travel demand modelsystem with activity schedulesQ, Transportation research A, vol. 93, pp. %G*7.

:hen, I. W., 6ross, 5., echeu8, ;., ) Sovanis, . . *++9, QEstimating Model Preferencebetween ITS-Enhances and Existing Paratransit ServicesQ, Transportation ResearchRecord.

:ubu2cu, ;. *++%, Factors Afecting Shopping Trip Generation Rates in MetropolitanAreas.

-e alma, !. ) /ochat, -. %&&&, Q=nderstanding individual travel decisionsA results from acommuters survey in 6enevaQ, Transportation, vol. *, no. 9, pp. *9G*7%.

-epartment of Transport.Transport Statistics. %&&./ef TypeA -ata 5ile

6iuliano, 6., Hu, H., ) ee, ;. *++9, Travel Patterns o" the (lderly: the Role or Land useMETRAS PR!"E#T $$-%.

6olob, T. ) 1cCally, 1. %&&', Q! 1odel of !ctivity articipation and Travel Interactions

$etween Household HeadsQ, Transportation research B, vol. 9%, pp. %''G%&0.

2$


29/29

6ould, S., 6olob, T. 5., ) $arwise, . &h' do people drive to shop( )*t*re travel andtelecomm*nications tradeo+s. :enter for !ctivity "ystems !nalysis . %&&7./ef TypeA ?lectronic :itation

Handy, ". %&&, Qon-wor, travel of women patterns. perceptions and preferencesQ,The WomenNs Travel Issues "econd Cational :onference.

Hess, "., $ierlaire, 1., ) ola2, S. W. Q/evelopment and Application of a Mixed #ross-ested 0ogit ModelQ, E*ropean Transport #onference.

;im, $. -. ) ar2, ;. %&&', Q"tudying patterns of consumerNs grocery shopping tripQ, )ournalo" Retailin!, vol. '9, no. 0, pp. 3+%G3%'.

;im, ". ) =lfarsson, 6. 5. *++0, QThe Travel 1ode :hoice of the ?lderlyA ?ffects of ersonal,Household, Ceighborhood, and Trip :haracteristicsQ, Transportation Research Record.

1c5adden, -. ) Train, ;. *+++, QMixed M0 Models for /iscrete ResponseQ, )ournal o"Applied (cono'etrics, vol. %3, pp. 00'G0'+.

Bffice for Cational "tatistics. *++3./ef TypeA Internet :ommunication

alma, !. d. ) /ochat, -. *+++, Q1ode choices for trips to wor2 in 6enevaA an empiricalanalysisQ, )ournal o" Transport eo!raphy, vol. 7, no. %, pp. 09G3%.

/ietveld, . *+++, QConGmotorised modes in transport systemsA a multimodal chainperspective for The CetherlandsQ, Transportation Research Part &: Transport and(nviron'ent, vol. 3, no. %, pp. 9%G9.

/obinson, /.

Documents

Mode Choice of Shopping Travel for Older People-V4b