International Journal of Innovative Technology and Exploring Engineering (IJITEE)

ISSN: 2278-3075, Volume-3, Issue-3, August 2013

Mode Choice Modelling For Work Trips in

Calicut City
Tushara T, Rajalakshmi P, Bino I Koshy

Abstract— Transportation modelling plays an important role The current study was to develop a comprehensive mode
in supporting transportation planning. Work trips are centre of choice model for Calicut City which will be capable to predict
focus of urban transportation planning and policy analysis. This modal shares of different modes of transport.
may causes congestion in peak hours in the urban transportation
The choice of a travel mode by the individual for
network. One of the important aspects of transportation modelling
is to predict the travel choice behaviour. The travel choice
home-to-work trip or even in case of other trips will involve
behaviour is also referred to as traveller mode choice, which is the complex decision making processes [3]. It will be influenced
most frequently modelled travel decision. It involves a specific by the attributes of the person and the characteristics of the
aspect of human behaviour dedicated to choice decisions. With a urban transport systems. A multinomial logit model with
model, as simplified representation of a part of reality provides a statistical data processing software SPSS was used for
better understanding and interpreting of these complex systems. explaining travel patterns and mode choice of employees
This paper investigates mode choice behaviour of employers in residing in Calicut city. It is widely used model in the discrete
Calicut city. A multinomial logit model (MNL) with statistical
choice model and it has many computational advantages. For
data processing software SPSS was used for explaining travel
patterns and mode choice of employees residing in Calicut city. a traveller in a City, the available modes and route choices for
MNL model was developed and identified the factors influencing a trip are sometimes numerous.
the mode choice of work trips.MNL is widely used model in the
discrete choice model and it has many computational advantages. II. SCOPE OF THE STUDY
Calicut is the one of the developing city in the state of
Index Terms— MNL model, Mode choice, Utility, work trips,
Kerala; the offices and main activity are located in the centre
employees.
of the city. Different types of modes are using for various
activity like work trips, school trips, leisure trips etc.
I. INTRODUCTION
Demands for both private and public transport have been
Transportation modelling plays an important role in increasing day by day. Without understanding the traveler‟s
supporting transportation planning. One of the major roles of needs and preferences it‟s very difficult to make solutions for
transportation modelling is to forecast travel demand based present conditions in the city. This study may be helpful to
on changes in the transportation system. There are different find mode choice behavior of work trips in the city. The
types of models that have been developed to create actual limited number of studies dealing with work trips suggests
travel patterns of people and existing demand conditions. The that these also play a significant part in the choice of transport
models are used to predict changes in travel pattern and mode for this type of trips.
utilization of the transportation system in response to changes This study focused on home-to-work trip and work to home
in land-use, demographics and socio-economic conditions [1]. trips. This group of travellers is one of the largest in number
Work trips are centre of focus of urban transportation and they may create congestion in peak hours. It enables to
planning and policy analysis. This may causes congestion in understand travel demand behavior to work trips and
peak hours in the urban transportation network [1]. One of the constraints that travelers face. The study identified the factors
important aspects of transportation modelling is to predict the effecting mode choice for work trips and also make model for
travel choice behaviour. The travel choice behaviour is also work trips in the city of Calicut.
referred to as traveller mode choice, which is the most
frequently modelled travel decision. It involves a specific III. OBJECTIVES
aspect of human behaviour dedicated to choice decisions.
The main objective of this thesis is,
With a model, as simplified representation of a part of reality
provides a better understanding and interpreting of these To identify the various variables that influences the mode
complex systems. Calicut city has selected for study for this choice behaviour of employees.
modelling purpose. This paper investigates mode choice To develop a model of travel mode choice for work trip
behaviour of a work trip for employers in Calicut city. Mode in Calicut City.
choice model is a principal component of travel demand
analysis [2]. The city consists of different categories of people IV. LITERATURE REVIEW
living in different zones and mode selection also depending As far as mode choice is concerned, the results highlight the
upon various factors like age, income, comfort [3]etc. importance of contextual and individual factors besides mode
characteristics (travel time, cost and comfort). The car
Manuscript received on August, 2013.
ownership decision is found to be mostly related to income
Ms. Tushara T, Department of Civil Engineering, Rajiv Gandhi Institute levels of the households, contextual constraints and location
of Technology, Kottayam, India. issues. The mode choice modelling to work has the central to
Ms. Rajalakshmi P, Department of Civil Engineering, Rajiv Gandhi the evaluation of the efforts to mitigate traffic congestion [4].
Institute of Technology, Kottayam, India.
Dr Bino I Koshy, Department of Civil Engineering, Rajiv Gandhi
Work trips in any urban area are always at the centre of focus
Institute of Technology, Kottayam, India. in urban transportation planning and policy analyses.

106
 Mode Choice Modeling For Work Trips In Calicut City

Work trips in aggregation define peak versus

off-peak-period traffic flow in the urban transportation
network. Although activity based modelling practice extends
beyond peak-period travel, commuting activities are always at
the centre of all modelling approaches [5].
The importance of mode choice in transportation policy
analysis and decision making has led to a variety of methods
for predicting the effects of policy measures on travellers‟
mode choices [6]. Residential choice location is influenced by
many variables including socio-economic characteristics, life
cycle, location of work and other major activities such as
schools, shopping, family and friend, real estate values, and
characteristics of the residential and workplace area. Living
close to the workplace reduces vehicle kilometre of travel and
thus contribute to a more sustainable transportation system.
[7].
The choice of a travel mode by the individual during the Fig. 1 Calicut City
morning home-to-work trip or even in case of other trips B. Data Collection
could involve complex decision making processes largely Based on the previous literature review, the questionnaire
influenced by the attributes of the person and the was designed in order to fit the objectives of this study.
characteristics of the urban transport systems [8] However, the length of the questionnaire as well as the
accuracy associated with the responses of respondents was
V. METHODOLOGY also taken into account. Using an above-designed
The study area is generally regarded as the geographical questionnaire, the data collection method was implemented
region in which transport planning needs to be done. Private through a survey of face-to-face interview including two
transport has always dominated in the present scenario, it stages. The first stage was a pilot survey in order to improve
serves as the most reliable, comfortable and time-efficient the clarity, validity, and content of the questionnaire. In this
mode. Due to these reasons, most of the people forced to use stage, the questionnaires were delivered to twenty employees.
private vehicles, instead of using public transport. These After receiving their feedbacks, a final standard version of the
reasons have highly accounted for increasing road congestion, questionnaire was used in the next survey stage. The
environmental pollution and traffic hazards. Frame work of individual dataset was obtained through the direct interviews
the study as shown in Fig .2 with employees in Calicut city, only individuals were
interviewed.
A convenient sample of 514 employees was interviewed in
Calicut city. Random sampling method was adopted for
sampling technique. Data collected from different categories
of employees in Calicut city .Work trip was taken from 3
categories of employees i.e., Govt, private and self employees
as shown in Table I. The business activities are located at
centre of the city. These data may be identify the current
travel behaviour of and also reveals factors effecting the
selection of modes.
Table I Data collection details
Si Category Departments/offices
No
1 Govt. Labour Court, Pension Treasury,
Mining and Geology District
office, soil conservation, soil
survey department. Pathology
department, Medical College
Fig 2 Frame Work of the Study Kozhikode, SBI Kozhikode,
A. Study Area police station and SBI, Palayam.
2 Private Baby Memorial Hospital, Service
Calicut city is selected as the study area. Calicut is the third
Bank Chalappuram Aswini
largest city in Kerala and the hub of north Kerala with a
Diagnostic Services Calicut,
metropolitan population of 2,030,519 as per 2011 census.
KSFE, Kalyan kendra ,Gulf
Calicut is fast emerging as a prominent educational,
Bazar, Cochin Bakery, Focus
commercial and trade centre having well established
dentacare, Muthoot Bankers,
national/international trade routes. Calicut city alone accounts
Mahindra Motors, Kannankandy
for 40% of the urban population in the District and is
Kitchen Interiors and WITCO.
experiencing a rapid growth in urban population. Map of
Calicut city as shown in fig. 1 2 Self Shops at SM Street ,
Palayam,Kuthiravattom

107
 International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075, Volume-3, Issue-3, August 2013

C. Theoretical Foundation function of the systematic portion of the utility of all the
alternatives [8].
Multinomial Logit Model
To understand the relations between characteristics and VI. PRELIMINARY ANALYSIS
mode choice of the employees, a multinomial logit model Table II represents descriptive statistics on the
(MNL) is applied to distinguish the difference among the demographic profile of employees. 514 data collected, from
mode usage of these three categories of employees[7]. MNL this 14 are rejected due to missing values and other reasons,
is the simplest and most popular practical discrete choice 500 data were taken for the analysis of work trips.
model; it is based on the random utility theory that each option
Aj Ɛ A has associated a net utility Ujq for individual. It is Table II Descriptive statistics on the demographic profile of
assumed that Ujq has two components, employees
U jq V jq jq

This includes a measurable part Vjq and a random part Ɛ jq. Demographic Variable Category Frequency Percentage (%)
The measurable part is Vjq considered as a function of 0 18-35 231 46.2
measured attributes x; thus Vjq is often formulated as a linear 1 36-45 165 33
combination of x, such as the following: Age 2 Over 45 104 20.8
Vj X 0 Female 163 32.6
jq kj jkq
Gender 1 Male 337 67.4
Where the parameter θ are often assumed to be the same. 0 Govt. 113 22.6
Based on the hypothesis of rational choice, Probability of 1 Private 242 48.4
alternative i chosen by individual q can be formulated as: Occupation 2 Self 145 29
Pn e Vn /( e Vm ) 0 Low 1000-15000 42 8.4
m M Income 1 Medium 15001-45000 339 67.8
Pn is probability that the individual select the mode n Level 2 High >45000 119 23.8
Vn is utility of mode n
Vm is utility of any mode A. Preliminary analysis result
M is set of all available travelers mode The age level of employees was from 18 to 65 and the
The closed form of the MNL is to estimate (maximum average age in the sample was about 45. Of the three groups,
likelihood estimation procedure), interpret and use. The the 18-35 age range occupied the highest proportion (46.2%),
MNL model was developed. For development of the models, while the second and the third accounted for 33% and
the following choices were considered in the universal choice 20.80%, respectively. From the 500 employees, there were
set: car, two-wheeler, bus, auto and walking. Two-wheeler 163 female employees (approximately 32.6%) while male
was taken as base mode. TIME (total travel time in Minutes) employees accounted for slightly above 67.4%.
and COST (total travel cost) were the obtained for each of the About the occupation of employees, out of 500
five modes. Travel time and travel cost represent mode respondents, 113 employees (22.6%) worked in government
related attributes; all other things being equal, a faster mode offices, 242 worked in private sector (48.4%) and 145 were in
of travel is more likely to be chosen than a slower mode and a business(29%) sector. Finally, monthly individual income of
less expensive mode is more likely to be chosen than a costlier employees in the sample is also displayed in Table 2. The
mode. income level between employees was from 1000 to 150000
Multinomial logit models are used to model relationships and the average income in the sample was about 75000. Of
between a polytomous response variable and a set of regressor three groups, the 15001- 45000 (medium income) income
variables. The term “multinomial logit model” includes, in a range occupied the highest proportion (67.8%), while the first
broad sense, a variety of models. The cumulative logit model (1000-15000) and the third (>45000) accounted 8.4% and
is used when the response of an individual unit is restricted to 23.8%, respectively.
one of a finite number of ordinal values[8]. 4.2.1 Travel Mode Chosen by Employees
The specific assumptions that lead to the Multinomial Logit The employees were chosen set of alternatives, such as:
Model are (1) the error components are extreme-value (or car, two-wheeler, bus, auto rickshaw and walk. They were
Gumbel) distributed, (2) the error components are identically coded as follows: 1 =car, 2 = two-wheeler, 3 = bus, 4 = auto
and independently distributed across alternatives, and (3) the and 5 = walk. In the model, each employee must choose one
error components are identically and independently mode, of the five travel modes. Two-wheeler is chosen as the
distributed across observations/individuals[8]. reference alternative because majority of employees choosing
Assumptions are discussed below, this mode than other mode for work trips. Fig 9 shows mode
The most common assumption for error distributions in the split of employees in city. The distribution of mode split
statistical and modeling is that errors are distributed normally. shown in Table III.
There are good theoretical and practical reasons for using the Table III Mode split
normal distribution for many modeling applications. The
Gumbel distribution is selected because it has computational Mode Frequency Percent
advantages in a context where maximization is important, Car 46 9.2
closely approximates the normal distribution and produces a Two wheeler 313 62.6
closed-form probabilistic choice model. The three Bus 130 26
assumptions, taken together, lead to the mathematical
Auto 10 2
structure known as the Multinomial Logit Model (MNL),
which gives the choice probabilities of each alternative as a Walk 1 0.2

108
 Mode Choice Modeling For Work Trips In Calicut City

From this chart Two wheelers are the most prominently

(62.8%) used for work trips, this may due to less consumption
of fuel and less space needed for parking.26% of employees
are choosing bus as their mode because of low cost and
followed by car (9.2%).The percentage of mode selection of
auto and walk is very less (<5%), hence this data excluded for
the analysis .Mode split as shown in fig 3

Fig.6 Income wise mode selection

Fig.7 shows details of vehicle ownership of employees.
This chart shows 56.6 % of employees having one car and
84.6% having one two-wheeler.

Fig .3 Mode split

Fig. 4 shows relationship between gender and mode choice
behaviour of employees. In this analysis 55% of male
employees are selecting two- wheeler as regular mode for
work trips. This may due to less cost of fuel and less space for
parking places of their offices. But 19.4 % of females are
preferring bus as their regular mode for work trips because of
less bus fare.
Fig .7 Details of Vehicle ownership

In the city, Two wheelers are the most prominently (62.8%)

used mode for work trips, followed by bus and car and very
less (<5%) are selecting auto and walk for their trip purpose.
55% of male employees are selecting two- wheeler and 19.4
% of females are preferring bus as regular mode for work
trips. In the city, 56.6 % of employees are having car and
84.6% having two-wheeler ownership.
Fig .4 Gender wise mode selection
The age wise mode selection of employees is shown in VII. MODEL FORMULATION AND VALIDATION
fig.5.From the age group 18-35 is choosing two- wheeler In the formulation of MNL model, three categories of mode
(32.2%) as their mode for work trip and followed by bus and usage, namely: ie bus, car and two wheeler would serve as the
car. alternatives that be adopted by employees. On the other hand,
explanatory variables are (1) Age (numerical variables,
recorded in integer); (2) Gender (dummy variable, male: 1;
female: 0); (3) Marital status (dummy variable, married: 1;
unmarried: 0); (4) Employment status (dummy variable,
Govt: 0; Private :1 and self : 2); and (5) Head of the house 6)
vehicle license ownership (Two-wheeler ,car) ,these are
applied into the model to formulate the measurable utility .
In this study, how employee‟s mode usage would change as
they are getting older is our major concern. Therefore, age is
selected as an explanatory variable. Besides, dummy
Fig.5 Age wise mode selection variables are also introduced into this model. The variable
The employees are divided in to three different categories “gender” is designed to tell if there is any difference of mode
based on income. The grouping started with the lowest choice between men and women. The variable “employment
income level, i.e. less than15000 and high income level was status” is designed to realize if these respondents have to
greater than 45000. Fig. 6 income wise mode selection of commute routinely or not. The last variable “vehicle driver
employees in the city. Considering the Medium income level license ownership” is introduced to realize if the employees
people, 46.2% of employees are using two wheeler and 18.2% are allowed to drive own vehicle or bus for work trip.
are selecting bus as their mode and very less number of people
choosing car as a mode (2.2%).Lower income groups were A. Preliminary analysis result
choosing bus and two- wheeler as their mode for work trips. The software used for modelling is SPSS (Statistical
Package for the Social Sciences).

109
 International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075, Volume-3, Issue-3, August 2013

SPSS Statistics is a software package used for statistical variable (modal choice) can be explained by the estimated
analysis and is now officially named "IBM SPSS Statistics". MNL regression model. Thus, it can be concluded that the
SPSS is a comprehensive and flexible statistical analysis and MNL model fits the sample data well.
data management solution. SPSS can take data from almost The Cox and Snell R2 measure operates like R2, with higher
any type of file and use them to generate tabulated reports, values indicating greater model fit. This measure is limited in
charts, and plots of distributions and trends, descriptive that it cannot reach the maximum value of 1. So Nagelkerke
statistics, and conduct complex statistical analyses. The proposed a modification that had the range from 0 to 1.
software is used for developing models like: Nagelkerke‟s measure is relied for indicating the strength of
1. Multinomial Logit models the relationship.
2. Nested Logit models
3. Random parameter logit models Table IV Pseudo R- Square
4. Probit models
5. Artificial Neural Network models. Pseudo R-Square
B. Multinomial logit model Cox and Snell 0.735
Multinomial logit models are used to model relationships Nagelkerke 0.896
between a polytomous response variable and a set of McFadden 0.773
regression variables. Logistic regression can be extended to
handle responses that are polytomous, i.e. taking r > 2 Likelihood ratio test (Table V) indicates the contribution of
categories. (The word polychotomous is sometimes used, but the variable to the overall relationship between the dependent
this word does not exist) .When analyzing a polytomous and individual independent variable in differentiating
response, it is important to note whether the response is between the groups specified by the dependent variable. The
ordinal (consisting of ordered categories) or nominal likelihood ratio test is a hypothesis test that the variable
(consisting of unordered categories) [8]. Some types of contributes to the reduction in error measured by the -2 log
models are appropriate only for ordinal responses; other likelihood statistic In this model, the variables time, cost,
models may be used whether the response is ordinal or waiting time, income, two-wheeler ownership, age and gender
nominal. If the response is ordinal, we do not necessarily have contributes significantly to explaining the mode choice.
to take the ordering into account, but it often helps if we do. Table V Likelihood ratio test
Using the natural ordering can of the above models; the
MNL is the simplest and most popular model [9]. The Effect Model Likelihood Ratio Tests
important assumption of this model is that the random Fitting
components of the utilities of various alternatives are Criteria
independently and identically distributed (IID). This -2 Log Chi-Square df Sig.
hypothesis creates several advantages in applications because Likelihood
calculation of choice probabilities, presentation, and of Reduced
interpretation of the parameters are easy [1] However, Model
because the IID assumption is often violated (i.e., the
Intercept 191.328 0 0 .
similarities amongst alternatives in a choice set are ignored) in
practice, the applications of the MNL are limited (McFadden, Time 379.188 187.86 2 0.000
1978) [10]. To overcome part of this limitation, researchers Cost 413.623 222.295 2 0.000
have proposed an application of the NMNL model in which Waiting time 201.296 9.967 2 0.007
the random components of the utility function are correlated Income 207.439 16.111 2 0.000
within each of the groupings [8] The MNL model is TW
lead to a simpler ownership 207.184 15.856 2 0.000
more economical model and Age 201.291 9.963 4 0.041
Increase the power to detect relationships with other
Gender 224.043 32.715 2 0.000
variables.
C. Model result Table VI shows the model fitting information about the
The discrete choice data was analyzed using the SPSS 16.0 model. This is the Likelihood Ratio (LR) Chi-Square test that
MNL program. 514 samples were collected, from this 25 at least one of the predictors' regression coefficients is not
samples are rejected due to missing values and other reasons, equal to zero in the model. The LR Chi-Square statistic can be
489 samples were taken for model formulation. The program calculated by -2xL(null model) - (-2xL(fitted model)) =
ran with different models using various attributes to ascertain 841.213 – 191.328 = 649.885 where L(null model) is from
the essential attributes to the model. Of the attributes selected, the log likelihood with just the response variable in the model
two separate models (with the intercept only and with all the (Intercept Only) and L(fitted model) is the log likelihood
coefficients) were run using the same MNL analysis. from the final iteration with all the parameters. This test is
Estimation results of the MNL model are displayed in analogous to the F-test for R2 value in multiple regressions
Table VIII. According to the results obtained, the relationship which test whether or not the improvement in the model
between the modal choice variable and the set of independent associated with the additional variables is statistically
variables is significant. The MNL model is much better than significant. In this model, chi-square value of 649.885 has
only a constant model. Further, it has a relatively high significance (0.000) and is less than 0.001, so there is a
pseudo-R2 value of 0.896 (Table IV), indicating that significant relationship between the dependent variable and
approximately 89.6% of the variation in the dependent the set of independent variables

110
 Mode Choice Modeling For Work Trips In Calicut City

Table VI Model Fitting Information A. Interpretation of result

1) Car Relative to Two- Wheeler
Model Model Fitting Likelihood Ratio Tests
a) Cost and Time
Criteria
-2 Log Chi-Square df Sig. The multinomial log-odds of cost preferring car relative to
Likelihood two -wheeler would be expected to increase by 0.166 units
Intercept 841.213
while holding all other variables in the model constant, but the
Only time decreases by 0.211 units. The Wald test statistic for the
Final 191.328 649.885 16 0 predictor cost is 34.358 an associated p-value of 0.000., the
regression coefficient for cost has been found to be
The data clearly indicated that the said attributes were
statistically significant, the Wald test statistic for the
indeed viable and provide the best fit to the data. The null
predictor time is 16.741 with an associated p-value of 0.000,
model serves as a benchmark against which we compare the
which is less than 0.05, and hence it is significant. Since the
fit of the final choice model and because the null model is
utility of a mode decreases as the mode becomes more
more complete model with other mode choices, a likelihood
expensive. Hence it would significant to accept the null
ratio test statistic is valid (Table VII)
hypothesis and conclude that the commuters are less likely to
Table VII Model goodness-of-fit
prefer car as their mode than two-wheeler.
Goodness-of-Fit
b) Income
Chi-Square Df Significance The multinomial log-odds of income preferring car relative
Pearson 1258.508 794 0.000 to two -wheeler would be expected to increase by 2.689 units
Deviance 191.328 794 1.000 while holding all other variables in the model constant.
The parameter estimates of the minimum acceptable model P-value of income is 0.000, hence it is significant.
are given in Table VIII. The degrees of freedom (df) of the c) Age
chi-square distribution used to test the LR Chi-Square statistic The multinomial log-odds for [Age=0] and [Age=1]
and is defined by the number of predictors in the model (seven relative to [Age=2] are -2.131 and - 2.311 preferring the
predictors in two models). The term „Significance‟ is the reference mode, which means the employees who has age
probability getting a LR test statistic being as extreme as, or greater than 45 are more likely to prefer car as their work
more so, than the observed statistic under the null hypothesis; purposes. It may be due to the convenience.
the null hypothesis is that all of the regression coefficients in The p- value of Two-wheeler ownership, waiting time and
the model are equal to zero. In other words, this is the gender are greater than 0.05, hence these are not significant
probability of obtaining this chi-square statistic (649.885), or
one more extreme, if there is in fact no effect of the predictor 2) Bus Relative to Two- Wheeler
variables. This p-value is compared to a specified alpha level, a) Cost and Time
our willingness to accept a type I error, which is typically set The multinomial log-odds of cost preferring bus relative to
at 0.05 or 0.01. The small p-value from the LR test, <0.00001, two -wheeler would be expected to decreases by 0.438 units
would lead us to conclude that at least one of the regression while holding all other variables in the model constant, but the
coefficients in the model is not equal to zero. The parameter time increases by .319 units. The Wald test statistic for cost
of the chi-square distribution used to test the null hypothesis is and time are 34.793 and 34.534 with an associated p-value are
defined by the degrees of freedom in the prior column. 0.000, which is less than 0.05, and hence it is significant.
Table VIII Parameter estimate b) Waiting time
95% Confidence The multinomial log-odds of waiting time preferring bus
Interval for Exp(B)
relative to two -wheeler would be expected to increases by
Std. Lower Upper
MODE B Error Wald df Sig. Bound Bound 0.244 units while holding all other variables in the model
Intercept -5.213 1.44 13.096 1 0 constant, The Wald test statistic for waiting time is 10.346
Time -0.211 0.052 16.741 1 0 0.732 0.896
Cost 0.166 0.028 34.358 1 0 1.117 1.248
with an associated p-value are 0.001, which is less than 0.05,
Waiting and hence it is significant.
time 0.193 0.204 0.895 1 0.344 0.813 1.807
Income 2.689 0.775 12.033 1 0.001 3.221 67.275
c) Vehicle Ownership
TW Two-Wheeler- the coefficient of two wheeler owner ships
ownership 0.25 0.653 0.147 1 0.702 0.357 4.615
is -3.179 of preferring bus relative to two- wheeler would be
[Age=0] -2.131 0.894 5.685 1 0.017 0.021 0.684
[Age=1] -2.311 0.919 6.317 1 0.012 0.016 0.601 expected to decrease in 3.179 units while holding all other
[Age=2] 0 . . 0 . . . variables in the model constant. The commuter who are
[Gender=0] 1.22 0.758 2.589 1 0.108 0.766 14.982
[Gender=1] 0 . . 0 . . .
having two-wheeler, they may less likely to prefer bus than
1 two- wheeler. The Wald test statistic for two –wheeler owner
Intercept -0.512 1.215 0.178 1 0.673 ship is 12.20 with an associated p-value of 0.000, hence it
Time 0.319 0.052 37.534 1 0 1.243 1.524
Cost -0.438 0.074 34.793 1 0 0.558 0.747
would significant to accept the null hypothesis and conclude
Waiting that the commuters are less likely to prefer bus as their mode
time 0.244 0.076 10.346 1 0.001 1.1 1.481
Income -0.311 0.506 0.378 1 0.538 0.272 1.975
than two-wheeler.
TW d) Gender
ownership -3.179 0.91 12.207 1 0 0.007 0.248
[Age=0] -0.079 0.857 0.009 1 0.927 0.172 4.953
The coefficient of female relative to male is 2.700 units,
[Age=1] 0.425 0.858 0.245 1 0.621 0.285 8.217 which is higher for preferring the reference mode, given all
[Age=2] 0 . . 0 . . . other predictor variables in the model are held constant. In
[Gender=0] 2.7 0.506 28.49 1 0 5.522 40.121
3 [Gender=1] 0 . . 0 . . .
other words, females are more likely than male to prefer bus
a. The reference category is: 2.(Two-wheeler) for their work trip purposes than reference mode.
b. This parameter is set to zero because it is redundant. The coefficient of Intercept is -0.512 with an associated

111
 International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075, Volume-3, Issue-3, August 2013

p-value of 0.673., hence it would not significant to accept the transportation system provides inefficient, unproductive, and
null hypothesis and conclude that the commuters are less unsafe level of services. To maintain the economic viability of
likely to prefer bus as their mode than two-wheeler. this city and to keep its environment sustainable, an energy
The p- value for age and income are greater than 0.05, and efficient mass transportation system is imperative.
hence these are not significant Results of the multinomial logit model analysis
The empirical utility functions for different mode are: demonstrated that these five factors (age, gender, income time
V (CAR) = -5.213 – 0.211(TT) +0.166(TC) +2.689 (INC) - and cost) were strongly related to considerations of the
2.131(AGE=0) - 2.311(AGE=1) employees of their travel. The sign of each coefficient
V (BUS) = -0.512 +0.319 (T.T) -0.438 (T.C) +2.44 (WT) - generated in the MNL model is convincible and consistent
3.179(TW) +2.700(GEN=0) with the hypothesis. In terms of the hypothesis the mode
The prediction success of the model is given in Table IX. choice of workers affected mainly by time and cost.
The validation run was also made with the same model In Calicut the Two -wheelers are used by most of road users
specification. The model was calibrated for data set of 326 for their trip purposes. Thus empirical results in this study do
cases and validated using hold-out sample of 163. identify the relations between mode choice and characteristics
of employees in Calicut city.
Table IXClassification (Prediction success)
B. Conclusion
Observed Predicted
1 2 3 Percent A mode choice model was developed for work trip for
Correct Calicut city. Age, gender, income, time and cost are proved to
1 62 14 0 81.60% be the significant factors that influence the mode usage of the
2 10 165 10 89.20% employees, and the model of employee‟s mode choice
formulated in this study is also convincible. This work trip is
3 1 15 98 86.00%
essentially an attempt to identify the various variables that
Overall 19.50% 51.70% 28.80% 86.70%
influences the mode choice behaviour of employees. The
Percentage
investigation was conducted through a comprehensive
analysis of geographic, transportation, and personal statistics,
The result of validation is aggregated in Table X. The
principally with the aid of the logit model.
overall correctly predicted percentage indicates the percent of
Findings from this study can be summarized as follows:
cases where the actual choices and the predicted choices of
Main factors influencing work trip of employees are age,
the individuals match. The present model is 92% accurate; the
gender, income, waiting time, two-wheeler ownership,
difference is only 5.3%, which is excellent or good.
time and cost.
Employees (62.8%) would choose two-wheeler for their
Table X Classification (Validation Success)
trip.
Observed Predicted
Bus is chosen by females who are not having driving
1 2 3 Percent
licence of car or Two-wheeler.
Correct
Differences were found between female and male for
1 22 6 0 78.60%
selecting the mode for their work trips, the female
2 2 225 4 97.40%
(19.4%) tend to choose bus but male (55%)are preferring
3 0 14 53 79.10%
two wheeler for their work trip.
Overall 7.40% 75.20% 17.50% 92.00%
The age group 18-35 is choosing two- wheeler (32.2%)
Percentage
for their work trip purposes and followed by bus and car.
Age group (>45) are choosing car for their work trip
VIII. CONCLUSION AND DISCUSSION
purposes, it may be due to the comfort of the journey.
A. Discussion The Medium income level people ,46.2% of employees
are using two wheeler and 18.2% are selecting bus as
The Multinomial logit model for mode choice analysis of their mode and very less number of people choosing car
work trips was developed. The significant deterministic as a mode, i.e. less than 10%.Lower income groups were
variables include the total travel time and travel cost for each choosing bus and two- wheeler for their work trips.
mode of transport. The utility ranking and probability of
Regarding vehicle ownership, 56.6 % of employees
choosing modes, the two-wheeler is always at the top level
having one car and 84.6% having one two-wheeler.
because of it less fuel consumption, followed by the bus. The
Auto and Walk are found to be less (<5%) popular
utility for buses necessarily does not indicate users‟
among the working groups.
willingness to prefer the service, rather it may represent that
people are forced to use that mode of transport due to
IX. ACKNOWLEDGMENT
unavailability of suitable alternatives before them. It implies
that the improvement of this system will attract people who The authors owe thanks to all people for their support during
choose their mode as private car. the making of the paper. We would also like to thank the
The business activities and offices are locating in the centre faculty of Department of Civil Engineering at Rajiv Institute
of the city, these leads traffic congestion and other Technology, Kottayam for their kind patronage and
environmental problems. The existing transportation system encouragement that made this paper possible
is a major bottleneck for the development of the city.
Unplanned urbanization, especially poor transportation REFERENCES
planning and lower land utilization efficiency, has turned the [1] N. A Saad and R. B William, “Modeling Saudi –Arabia Bahrin
city into a dangerous urban jungle. The existing road-based corridor mode choice.”Trans. Geo. , Vol 3, No.4,1995,PP .349-353.

112
 Mode Choice Modeling For Work Trips In Calicut City

[2] F. Koppelman and C. Bhat , “A self instructing course in mode choice

modeling: Multinomial and Nested Logit Models.” U.S. Dept. Of
Trans,2006.
[3] Gang Liu, “A behavioral model of work-trip mode choice in Shanghai”
Discussion Papers No. 444, 2006, Norway, Research Department.
[4] A. De Palma and D. Rochat, “Mode choices for trips to work in
Geneva: An empirical analysis.”Trans. Geogr.8, 2000, 43-51.
[5] M. K Nurul Habib, “Modeling commuting mode choice jointly with
work start time,”Transportation Research Part A 46 ,2012,pp 33–47
[6] J. Prashker, Y. Shiftan , P. Hershkovitch –Sarusi, “Residential choice
location, gender and the commute trip to work in Tel Aviv,” Journal of
Transport Geography 16, 2008, 332–341.
[7] D.B Madan and R Groenhout , “Modelling Travel Mode Choices For
The Sydney Work Trip,” Journal Of Transport Economics and Policy,
May 1987.
[8] H. L Chang and W. U Shuen-Cheng, “Exploring The Mode Choice in
Daily Travel Behaviour Of The Elderly In Taiwan,” Journal of the
Eastern Asia Society for Transportation Studies, Vol. 6, 2005, pp.
1818 – 1832.
[9] M. Ben-Akiva , “Discrete Choice Analysis,”2010.
[10] K. Train and M. C Fadden, “The Goods/Leisure Trade Off And
Disaggregate Work Trip Mode Choice Models” Transportation
Research, Vol. 12.1978.

Ms. Tushara T has obtained her B.Tech in

Civil Engineering from Karunya Institute of
Technology, Coimbatore. Presently she is
pursuing her M.Tech in Transportation
Engineering from Rajiv Gandhi Institute of
Technology, Kottayam. Her area of interest
is in Transportation planning.

Ms. Rajalakshmi has obtained her B.Tech

in Civil Engineering from TKM College of
Engineering,, Kollam. Presently she is
pursuing her M.Tech in Transportation
Engineering from Rajiv Gandhi Institute of
Technology, Kottayam. Her area of interest
is in Transportation planning.

Dr. Bino I Koshy is working as Professor

in the Department of Civil Engineering at
Rajiv Gandhi Institute of Technology,
Kottayam. He received his Doctorate and
M.Tech. Degree in Transportation
Engineering from IIT Madras and B.Tech
degree in Civil Engineering from Kerala
University. His area of interest is
transportation planning.

113