Top Gear Race The Estimation of Societys Choice On Transportation Mode Using A Multinomial Logit Model.

Top Gear Race : The Estimation of Society’s Choice on Tr ansportation
Mode using a Multinomial Logit Model

I Wayan Sukadana1
Amr it a Nugr aheni Sar asw aty 2
I Gusti Ayu Putr i Anggar a Indr asw ari 3

1,2,3 Depar tment

of Development Economics, Faculty of Economics and
Business, Udayana Univer sity

Email : wayan.sukadana@gmail.com
amr itasar aswaty@gmail.com

Abstract

Bal i as par t of Indonesian gr owing economy is facing an incr easi ng demand of t r anspor tati on
infr ast ructur es. In or der t o r esponse hi gh demand of tr anspor tati on infr ast r uctur e, and
to pr event t he thr eat of tr affic congesti on, the gover nment of Bali pr ovides t he society
with a public tr anspor tation and a br and new toll r oad. These infr astr uctur es give an

oppor tunity to soci ety to decide between tr avel by t heir own car / motor cycl e or publ ic
tr anspor t. For the gover nment, i t is impor tant t o set effi cient t r anspor tati on
infr ast ructur es to escape fr om high cost economy which appear fr om tr affi c congestion.
This paper investi gates society’s choi ce on tr anspor t ati on mode in Denpasar City. We
esti mat e the passenger ’s choi ce data using a Mul tinomi al Logit model, with i ncome and
number of congestion spot as independent var i ables. The r esult shows that passenger
who has high value of t ime pr efer to tr avel by their own vehicle compar e to publ ic
tr anspor t. The r esult i s analogue to the Top Gear Race, w her e the car s most of the ti me
win the r ace against the publ ic t r anspor t. Finally, fr om the society’s point of vi ew, we can
say that the development of infr astr uctur es of public t ranspor t ation, which has been
done by t he gover nment , sti ll less attr active compar e to the pr i vate tr anspor tati on.

Key words: pr efer ence, tr anspor t ation infr astr uct ur e, lost mi nimi zati on, mul tinomial l ogit .

Abstrak

Bal i sebagai bagi an dar i per tumbuhan ekonomi Indonesia menghadapi peningkatan
per mint aan infr ast r uktur tr anspor t asi. Untuk mer espon tinggi nya per mintaan
infr ast ruktur tr anspor t asi , dan untuk mencegah ancaman kemacetan lalu li nt as,


pemer intah Bal i menyediakan t r ansport asi umum dan jalan tol bar u bagi masyar akat.
Infr astr uktur ini member ikan kesempatan kepada masyar akat untuk memutuskan
apakah melakukan per jalanan dengan mobi l mer eka sendir i / sepeda motor atau
angkutan umum. Bagi pemer intah, penting untuk mengatur infr astr uktur tr anspor tasi
yang efisien unt uk ter hindar dar i ekonomi biaya ti nggi yang muncul dar i kemacetan lalu
lintas. Makalah ini mengkaji pil ihan masyar akat pada moda tr anspor tasi di Kota
Denpasar . Data pilihan penumpang diper kir akan menggunakan model Mul tinomi al
Logit, dengan pendapat an dan jumlah lokasi kemacet an sebagai var i abel independen.
Hasil penel itian menunjukkan bahw a penumpang yang memi liki nil ai waktu tinggi lebih
memil ih untuk beper gian dengan kendar aan sendi r i di bandi ngkan dengan angkutan
umum. Hasil ini sesuai anal og Top Gear Race, di mana mobil-mobil pr ibadi sebagian
besar waktu memenangkan per lombaan melawan angkutan umum. Akhir nya, dar i sudut
pandang masyar akat, dapat di kat akan bahw a pembangunan infr astr uktur tr anspor t asi
publ ik, yang telah dilakukan oleh pemeri nt ah, masih kur ang menar ik dibandingkan
dengan kendar aan pr ibadi.

Kata kunci : pr efer ensi, infr astr uktur t r anspor tasi, lost minimization, mult inomial logi t .

INTRODUCTION


Since the new economic or der in the late 1960s, economic development has
been incr easing r apidly and so has the demand for transpor t. A decade
later , it became obvious that tr anspor tation on the w hole needed a lot of
attention. Lending agenci es, as w ell as the centr al gover nment made
funds available for highw ay and tr anspor tation pr ojects to impr ove its
role in developing the nation. (Sutomo et.al, 1993). It then turned out that
the supply of public transpor t failed to fulfill the mar ket demand, so
people began to collectively opt for pr ivate vehicles over public
transport ation.Now , for over the last decade, the aver age number of
public tr anspor tation user s has decr eased r apidly, par tly due to the
over all growth of the Indonesian economy, w hich has incr eased the
number of middle-income ear ner s and the mar ket’s vehicle affor dability
(Setiaw an, 2013)
On 2003, public tr anspor t user s accounted for 45 per cent of total tr ansportation
user s, and it r emained so until 2012. Fr om that point, motor cycles became mor e
ubiquitous as a r esult of their almost 60 per cent market share, as few er and
few er people took to public tr anspor tation. It is also wor th nothing that over the
same per iod, public tr anspor t only accounted for 1 per cent of total vehicles on
the r oad.
In the fast and moder n economy, demand of tr anspor t ation infr ast r uctur es ar e

getting higher and higher . Unfor tunately, this gr ow th was and still is not
concur r ently car r ied out w ith enough r esponsibility from the citizens. Vehicle
ow ner ship symbolizes pr estige and r elat es to socioeconomic status, w hile
quantity is valued over quality: the mor e vehicles one ow ns the better their
per ceived economic status. Gridlock and inconvenient public tr anspor t systems
ar e becoming a hindr ance to pr oductivity and a w inning str ategy to incr ease
citizens’ str ess levels. The cities that wer e initially developed for human
pur poses, ar e now tr ansfor ming into a container of vehicles, gr ow ing at a 12
per cent r at e annually (Setiaw an, 2013)

Bali as par t of Indonesian growing economy is also facing an incr easing
demand of tr anspor t ation infr ast r uctures. In or der to r esponse high
demand of tr anspor tation infr ast r uctur e, and to pr event the thr eat of
traffic congestion, the gover nment of Bali provides the society w ith a
public tr anspor tation and a br and new toll r oad. These infr ast ructur es
give an oppor tunity to society to decide between tr avel by their ow n
car / motor cycle or public tr anspor t. For the gover nment, it is impor tant to
set efficient t r anspor tation infr ast r uctur es t o escape fr om high cost
economy w hich appear fr om t r affic congestion.


Literature Review
To make an adequate measur es on this ar ea of study, it is important to have
an enough under standing of tr avel behavior and modal choice. Tr avel
behaviour is complex and in the other side modal choices ar e deter mined
by sever al factor s. Some studies have been done for analyzing the
transport ation mode choice. Studies that explor ed the deter minants of the
public tr anspor t demand conducted by Paulley et al. (2006) analysed
differ ent type of elasticities. They show that fares, quality of ser vice and
car ow ner ship ar e the most si gnificant var iables which influence public
transport demand. Meanw hile, Alpizar and Carlsson (2001) studied a
gr oup of policies aimed at discour aging the use of pr ivate transportation
dur ing peak hour s, both dir ectly and indir ectly, by incr easing the
attr activeness of the only available substitute, the bus. This is done using
a choice exper iment constr ucted to find the answ er to the follow ing basic
question: Given fixed house-to-w or k structures and no w or king hour
flexibility, by how much is the choice of tr avel mode for commuter s to
work sensitive to changes in tr avel time, changes in costs for each mode
and other ser vice attr ibutes?. The r esults show that in gener al, indicate
that


model

substitution

is sensitive to

the char acter istics and

per for mance of each mode. In par ticular , tr avel time for both modes and
travel cost for car are the most impor tant det er minants of mode choice.

They also conclude that a pr ogr am aimed at reducing congestion and
pollution dur ing peak hour s should focus on incr easing the cost of pr ivate
transport and providing faster and mor e r eliable public tr ansport.

The opposition between car and public tr anspor t use and w ays to encour age
people to take public tr anspor t that pr ovided by the gover ment can be
found in many empirical r esear chs such as; Hensher (1998) and Meyer
(1999) demonstr ate that, in ter m of t r anspor t demand management
(TDM), the best action is to incr ease the pr ice of car for individual use and

to reduce the overall attractiveness of the car . Nakamur a and Hayashi
(2013) define thr ee str ategies for low-carbon ur ban tr anspor t: ‘avoid’ (i.e.
reducing unnecessar y tr avel demand), ‘shift’ (i.e. encour aging modal shift
in favor of public tr anspor t use) and ‘i mpr ove’ (i.e. impr oving fuel
economy and emission intensity).
The method that usually uses to estimate tr anspor tation mode choice is the
logit model (McCar thy (2001) and also Gebeyehu and Takano (2007).
Depar t fr om those studies, w e also try to estimate the society’s pr efer ence
tow ar d the tr anspor tation mode. This study estimates the impact of
monetar y expense and value of time on the society’s choices

METHODOLOGY

Model of Society’s Choice

In this paper we assume the society is facing a lost minimization pr oblem of
two alternatives to tr avel, w hich ar e their own car / motor cycle or the
public transpor tation. Befor e w e ar r ive at the lost minimization problem,

let w e consider a simple model of society’s choice, w hich adapt from Shy

(1995) w ith a slight modification. We assume that the passenger s, w ho
travel fr om Denpasar to Bukit Jimbar an Campus, have tw o possible
alter natives for getting there; each passenger can use either to tr avel by
their ow n vehicle or choose to use the public tr ansport ation. For the
passenger s who tr avel by their own vehicle, we assume deal w ith tC tr avel
time. This tr avel time depends on the tr affic ( congestion) and ther efor e
depends on the number of all passengers w ho decide to tr avel with their
ow n vehicle. Formally, w e can w r ite this t ravel time as;

tC  α  βnC , wher e 0  α 1 and β  0

(1)

The par ameter α measur es the time that spend by passenger , w hich is
independent from congestion, such as the time it takes to star t and heat a
car , to check the oil, dust, and so on. The par ameter β measures the effect
of congestion on tr avel time, w hich depends on the quality of r oad,
number of lanes, and tr affic lights. Similar to tr avel time by using pr ivate
vehicle, let we consider a tP for the t r avel time by using public
transport ation, and w e can w r ite for mally as follow s;


tP  φ ϕnC , w her e 0  φ 1 and ϕ  0

(2)

The par ameter φ measures the time that spend by passenger, w hich is
independent fr om congestion, such as the time it takes to get to the bus
station or stop. The par ameter ϕ measures the effect of congestion on
travel time, which depends on the quality of r oad, number of lanes, and
traffic lights. The inter pr et ation of parameter ϕ almost identic w ith β ,
because public tr anspor tation has almost same r oute w ith pr ivate
transport ation.

We assume N passenger s w ill fall into nC passengers w ho chose to t r avel with
pr ivate vehicle and n P passenger s w ho chose to tr avel by public
transport ation, so nC  nP  N . Than, we denote v as the value of time and
define the lost function for the passenger s w ho use the pr ivate vehicle as;

(α  βnC ) γ
(3)

Meanw hile the lost function for passenger s w ho use public tr anspor t is as
follows;

(φ ϕnC )  λ
(4)
The lost-minimization functions of the two alter natives ar e functions of both
value of time and a monetar y expense.

We assume that ther e are a lar ge number of passenger s wi shing to go
Udayana Univer sity on Jimbar an, so each passenger ignor es their
marginal effect on congestion. Hence, each passenger take n C as given and
minimizes;

min L ≡ min L , L
C

P

(5)


C,P

Therefor e, if in equilibr ium passenger s use both tr anspor t ation methods, then nC
must satisfy;

v(α  βnC ) γ  v(φ ϕnC )  λ
(6)
Assuming that N is sufficiency large, so that not all passenger s use the same
transport ation mode, the equilibr ium allocation of passenger s betw een
the two tr anspor t ation methods is given by;

nCe 

(λ −γ )  v(φ − α)

v(β −ϕ)
(7)
The equilibr ium number of passenger s who use pr ivate vehicle nC incr eases
with the cost of public transpor tation (the ticket) λ and decr ease with an
increase in toll-r oad ticket, γ . For a given value of time, v , the n C increase
with time it takes to get to the bus station or stop, intuitively w e can
assume the α

, is ver y

small. The equation (7) also give an inter esting

interpr etation, the number of pr ivate vehicle’s driver is decr ease with the
decr ease of the travel time using public transpor tation ϕ .

RESULT AND DISCUSSION

Sample and Sur vey Result

The samples ar e the students and employee of Udayana Univer sity w ho
travel for study or work in Bukit Jimbar an Campus fr om Denpasar greater
ar ea ( Sar bagita). We classified the mode of tr ansport ation into thr ee
categor y; “public” for the tr ans-sar bagita, “motor ” for motor cycle, and
“car” for car . We select the samples r andomly fr om the student and
employee w ho have an exper ience in using t hose thr ee t ransport ation
modes. This setting is ver y impor tant in gaining gr eater data sets that will
use in Multinomial Logit Model later on. We classified the per son w ho
usually uses the public transpor tation into samples that chose public
transport ation and, w e coded it w ith 1 and 0 for others. We also applied
the same method for others mode.

To collect the data, w e have distributed 100 questionnair es, but ther e ar e 8
retur n questionnair es that not give clear infor mation, and we decide to
dr op it. Table 1. show the frequencies and percent ages of the mode
chosen by r espondent.

Table 1. Frequencies and Percentages of the Chosen Mode

mode

Freq.

Percent

Cum.

public

30

32.61

32.61

car

27

29.35

61.96

motor

35

38.04

100.00

Total

92

100.00

The data on Table 1, show that r espondents ar e distr ibuted almost equally
among the alternatives. Motor cycle appear s to be the most popular mode,
compar e to car or public mode. This popular ity may because of
motor cycle have the shor test tr avel time among the alter natives as show n
in Table 2.

Table 2. Mean of Tr avel Time by Mode

mode

mean(tmotor)

mean(tcar)

mean(tpublic)

public

32

48

120

car

29

44

107

motor

29

44

105

Respondent tend to choose motorcycle to tr avel when they face mor e
congestion spot. Table 3, show us that the aver age number of congestion
spot that face by r espondent w ho choose car and public tr ansportation
does not differ so much. This figur e tell us that people tend to indiffer ent
in choosing between car or public transpor t ation when they face less
congestion spot.

Table 3. Mean of Congestion Spot by Mode

mode

N(scong)

mean(scong
)

sd(scong)

public

30

2.5

.5085476

car

27

2.22222

.4236593

motor

35

3.51429

.8178677

Fr om Table 4, w e can see that r espondent w ho has lower income choose
public mode. Meanw hile, for the highest income tend to choose car.

Table 4. Mean of Respondent by Mode in Thousand Rupiah

mode

N(income)

mean(income
)

sd(income)

public

30

1420

99.65458

car

27

3814.81

845.6694

motor

35

1728.57

388.4899

Econometr ic Appr oach and Multinomial Logit Estimation

We estimate the passenger ’s choice data using a Multinomial Logit model,
with income and number of congestion spot as independent var iables.
Fir st, w e estimate the impact of income on mode choice. This setting t ry
to analyze monet ar y expense impact on society’s choice tow ard the
transport ation mode.

Table 5. Multinomial Logit Estimation; Monetar y Expense Impact

Multinomial logistic regression

Numberobs

Log likelihood = -39.354221

mode

Coef.

Std. Err.

=

92

LR chi2(2)

=

122.38

Prob >

=

0.0000

Pseudo

=

0.6086

z

P>|z|

1.78

0.07
4

[95%

Conf. Interval]

Car

inc

73.23428

41.03145

7.18
5889

153.6545

_cons

1199
.714

61.28633

-569.214

321.6897

-1.77

0.07
7

9.903107

3.05575

3.24

0.00
1

3.913946

15.89227

0.00
1

116.
0253

28.6
288
3

Motor

inc

_cons

(mode==public

-72.32705

is the
base

22.29542

-3.24

outcom
e)

Table 5, tell us that the incr ease in income w ill induce the society to leave the
public tr ansportation. When society’s income incr eases they will pr efer
motor cycle compar e to public tr anspor tation. This phenomenon is causes
by the tr avel time using motorcycle is shor ter compar e to public
transport ation. When, their income incr eases mor e, and then they pr efer
car compar e to motor cycle. Although the tr avel time using motorcycle is
shor ter than car , using car gives people mor e comfor t, mor e over if the
congestion spot is less.

The income affect society’s choice of tr anspor tation mode. It is in line with
Mukala’s and Chunchu’s r esearch (2011). They developed var ious choice
models based on the st ated prefer ence dat a for modeling the inter city
mode choice behavior in India. The result found that the income and total
travel cost play an impor tant r ole in the mode choice decision for the
intercity tr anspor t.
On the other hand, car availability gives people mor e comfor t compar e to
motor cycle or public tr anspor t. This r esult in line with Nurdeen, et.al
(2007). They found car availability is ther efor e a major factor that
deter mines the choice of inter city tr anspor t mode. Resistance to
sw itching w as obser ved among r espondents w ho have one vehicle

available, w hile r espondents w ho have tw o or mor e vehicles w er e less
resistant to mode change.

The second estimation is dedicated to test the value of time tow ard society’s
the number of congestion spot as the proxy of value of time. The
estimation r esult is as show n in Table 6.

Table 6. Estimation Result

ltinomial logistic regression

Numberobs

Table 6. Mult i nomi al Logit Est imat ion;

g likelihood

-70.34808

mode

Coef.

Std. Err.

scong

1.2
156
37

.5816799

_cons

2.75329
2

1.380804

scong

3.38163
6

1.032338

_cons

9.7
987
06

z

=

92

LR chi2(2)

=

60.39

Prob >

=

0.0000

Pseudo

=

0.3003

P>|z|

[95%

Conf. Interval]

2
.
0
9

0.03
7

2.35
5709

.075
565
8

1.99

0.04
6

.0469648

5.459619

3.28

0.00
1

1.358291

5.404982

0.00
1

15.8
0218

3.79
523
5

tor

-

ode==public

the
bas
e

3.063051

3
.
2
0

outcom
e)

Table 6, tell us that an incr ease in the number of congestion spot faced by
society will causes people to use motor cycle. Inter estingly, the incr ease in

the number of congestion spot makes people to sw itch fr om using car to
public transpor tation. This r esult suggest us that if the gover nment w ant
to incr ease the number

of public transportation costumer , the

gover nment have to decr ease the time tr avel of this public transportation
so can compete w ith motorcycle. It is in line with other r esear ch that
conducted by Alpizar and Car lsson (2001) for r educing travel time.

CONCLUSION

The Multinomial logit estimation shows that passenger who has high value of
time pr efer to chose t ravel by their ow n vehicle, w hich is their
motor cycle, compar e to car or public tr ansport. The r eason is, because it
mor e pr acticable and no waiting time. The result is analogue to the Top
Gear Race, wher e the car s or motor cycle most of the time w in the r ace
against the public tr ansport. Finally, from the society’s point of view , we
can say that the development of infr astr uctur es of public tr anspor tation,
which has been done by the government, still less attr active compar e to
the pr ivate tr anspor t ation. How ever , the gover nment still has an
oppor tunity to w in the hear t of society by increasing the number of bus to
reduce the w aiting time. Although it is hard to compete with the
motor cycle tr avel time and simplicity, the public transportation still may
win in comfor t aspect.

RECOMMENDATION
The possibility of separ ating public tr anspor t by cr eating a parallel ser vice
that pr ovides a faster ser vice and more comfor t vehicle, is one potential
alter native to det r act society’s from private t r anspor tation. It is alr eady
conduct with Sar bagita’s Progr amme but it still need mor e effor t to satisfy
the society’s. Our study sheds light on the featur es r equir ed by that
system if it is to att ract t r aveler s from pr ivate modes. Specifically, special

emphasis has to be put on redesigning r outes and exclusive bus lanes, and
pr oviding tr affic pr iority for buses, faster connections betw een routes,
and mor e fr equent and reliable depar tur es, among other measur es
intended to r educe travel time w ith public tr anspor t.

REFERENCES

Alpizar, F., and Carlsson, F. 2001. Policy Implications and Analysis of the
Deter minants of Tr avel Mode Choice: An Application of Choice Exper iment s
to Metr opolitan Costa Rica. Wor king Paper s in Economi cs no. 56.
Depar t ment of Economics, Götebor g Univer sity. Sweden.

Gebeyehu,M and Takano, S. 2007. Diagnostic Evaluation of Public Tr ansportation
Mode Choice in Addis Ababa. Jour nal of Publi c Tr anspor t at ion, Vol. 10, No. 4,
2007.pp 27-50

Hensher , D.A. 1998. The imbalance between car and public transport use in
ur ban Australia: w hy does it exist?. Transpor t Policy, 5, pp. 193-204

McCarthy. P. S. 2001. Tr anspor tat ion Economics, Theory and Pract ice; A Case
St udy Appr oach. Massachusett s: Blackw ellpublisher s Inc.

Meyer, M.D. 1999. Demand management as an element of t r anspor tation policy:
using car r ots and stickts to influence t r avel behavior. Tr anspor t ation
Resear ch Par t A, 33, pp. 575-599

Mukala, P. K. and M. Chunchu. 2011. Mode choice modelling for inter city
tr anspor t ation in India: A case of Guwahati to five met r o cities.
Int er nat i onal Jour nal of Eart h Sciences and Engineer ing , vol. 04, no. 06 SPL,
pp. 364-374

Nakamur a, K. and Hayashi, Y. 2013. Str ategies and instr uments for low-car bon

ur ban tr anspor t: an inter nation r eview on tr ends and effect s. Transpor t
Policy, 29, pp. 264-274

Nur deen. A., R. A. O. Rahmat, and A. Ismail. 2007. Modeling of t ransportation
behavior for coer cive measur es for car driving in Kuala Lumpur . ARPN
Jour nal of Engineer ing and Applied Sciences, Vol. 2, no.2, pp.18-24

Paulley, N., Balcombe, R., Mackett, R., Tither idge, H., Pr eston, J., War dman, M.,
Shir es, J., White, P. 2006, The demand for public tr anspor t: the effects of
far es, quality of service, income and car ow ner ship. Transpor t Policy, 13, pp.
295-306

Setiaw an, R. 2013. Fixing Indonesia’s Public Transport Woes. Comment ar y
Ar t icle on The Jakar t a Globe. Nov. 4 th. 2013. Retr ieved July 27 2014 fr om
http:/ / w w w .thejakartaglobe.com/ opinion/ commentary/ fixing-indonesiaspublic-tr anspor t-woes/

Shy, Oz. 1995. Indust r ial Organization Theory and Applicat ions. Cambr idge:
MIT
Pr ess.

Sutomo, H., Dikun, S., Tumew u, W., 1993. Tr anspor t Problems, Policies and
Cur r ent Resear ch and Education in Indonesia : An Over view . IATSS
Resear ch.
Vol.
17
No.
1.pp
43-51