MODE CHOICE ANALYSIS BASED ON STATED AND REVEALED PREFERENCES FOR HOME BASED WORK TRIPS IN DELHI

Abstract

Delhi being the centre of socio-economic and political activities of India is growing at an alarming rate. With the growth of population and activities the transport demand has also escalated phenomenally. The road based mass transit system has failed to meet this demand efficiently and there is a need to introduce a Mass Rapid Transit System (MRTS) in the present scene. Mode choice analysis, an important step in four stage modelling is being utilised for analysing demand of innovative modes i.e. the mode which is non existent currently but being planned for the future. Conventionally mode choice analysis has been undertaken within the domain of revealed preferences where the commuter has actually considered the modes in making a choice decision. In the revealed preference approach the models are based on the data obtained by direct observation of travel behaviour. A comparison of chosen travel alternatives and the rejected alternatives reveals the preferences of the commuter. This approach being based on the existing choice behaviour has been found to be inadequate for predicting the demand of an innovative mode which is yet to be introduced. On the other hand stated preference approach which has been used frequently for estimating market demand of consumer products, stands as a viable alternative to revealed preference. The introduction of Mass Rapid Transit System (MRTS) depends on the feasibility of the system in terms of its market demand. The system being highly cost intensive it is essential to determine the demand for the system accurately. In this context the comparison between revealed preference and stated preference approach shall be useful in devising a viable technique for demand estimation. STATED PREFERENCE TECHNIQUE (SPT) Stated Preference Technique (SPT) refers to a family of techniques which use individual respondents statement about their preferences in a set of transport options, to estimate the utility functions. The common feature of this particular set of stated preference techniques is their use of experimental designs to construct a number of hypothetical alternatives which are designated as scenario. Individuals are asked to indicate how they would respond if these situations faced them in reality (that is, state their preferences towards the choice offered). The researcher has complete control over the factors included in the hypothetical scenario (which could include in-vehicle travel time, travel cost etc.). This enables to wide range of situations to be investigated, which may not be easily measured when observations of actual behaviour ("real life" choices) are used. OBJECTIVES The main objectives of this investigation are : (i) To study the mode choice behaviour of commuters for work trips in Delhi based on revealed preference approach, (ii) To develop an experimental design for study of commuter behaviour under hypothetical MRTS scenario, (iii) To analyse the demand of MRTS using revealed preference models and compare it with the stated preferenceestimates. MODELLING CHOICE BEHAVIOUR The economic theory of travel behaviour is founded on an assumption that each individual has an utility which can be partitioned into two components, a representative utility and a random utility. Individual traveller is assumed to be a deterministic utility maximiser. Based on this theory of utility maximisation and adopting a Weibull distribution (i i) for the random component the logit models of choice are developed. Logit models are generally estimated by adopting maximum likelihood estimation technique. Coefficient of determination , percent correctly estimated and goodness-of-fit test are the measures that explain how well a fitted logit model explains the variation in choice in an estimation sample. DESIGNING STATED PREFERENCE SURVEY Most of the efforts within stated preference approach draw upon "experimental design" which is usually orthogonal. In developing an experimental design the attributes are varied at different levels, when every possible levels of attributes are used it is knows as "full factorial design". When the number of options presented to the respondents becomes high, there is a strong likelihood that respondents will loose interest in the survey process, so increasing response error. Likewise inclusion of large number of attributes and many levels may lead to some items being ignored by respondents. It has been found that 9 to 16 options are acceptable. The following strategies can be adopted to achieve a sound experimental design, (i) remove those options that will "dominate" or be 'dominated" by all other options in the choice set (ii) separate the options into "blocks" so that the full choice set is completed by groups of respondents, each corresponding to a different sub-set of options (iii) carry out a series of experiments with each individual, offering different attributes, but with at least one attribute common to all, to enable comparison (iv) use "fractional factorial" design (v) define attributes in terms of differences between alternatives. (iii) These approaches can be used separately or in conjunction with each other. DEMAND OF INNOVATIVE MODE A mode choice model specified generically using revealed preferences can be directly used for predicting the demand of an innovative mode. This innovative mode can be a Mass Rapid Transit System(MRTS) or Sky Train or any other future alternative mode. The only requirement is that it should be a distinct alternative to satisfy multinomial choice formulation. The utility function of each mode is described purely in terms of influencing variables of choice. After a model is calibrated the estimated coefficients can be used directly to develop the utility of the innovative mode. This utility is simply added to the denominator in the probability expression for finding the probability of the innovative mode. Such an introduction is possible because incorporation of additional alternative cannot change the relative odds with which the previous alternatives are selected. SCOPE OF WORK The following investigations were envisaged, (i) Review of literature on mode choice analysis based on revealed and stated preference approaches, (ii) Developing binary and multinomial logit model for work trips in Delhi, (iiideveloping a stated preference design and conducting surveys for commuter response, (iv) To utilise the logit model for estimating demand of hypothetical Mass Rapid Transit System (MRTS) (v) Analysis of stated preference survey to estimate switching (iv) characteristics of commuters in the presence of Mass Rapid Transit System (MRTS). (vi) Comparison of revealed preference and stated preference approach in the demand estimation of Mass Rapid Transit System (MRTS). SURVEY DESIGN AND FIELD STUDIES Two separate questionnaires were developed for studying the revealed preferences and stated preferences of commuters. The revealed preference questionnaire included personal information, household information and work trip information of the commuters interviewed. The work trip information was designed carefully to have perfect information about the actual choices available to the commuter. In the level-of-service variables category in vehicle travel time, out-of vehicle travel time and travel cost were included. The stated preference questionnaire was a choice based design where seven different scenario of Mass Rapid Transit System (MRTS) were presented to the commuters. To give an impression about MRTS various photographs of MRTS in Calculta (Metro Railway) were also shown to the interviewees. The commuters were asked to make a choice between their present mode and the particular MRTS scenario. After conducting a pilot survey the questionnaire was suitably modified. The final surveys were conducted at different employment locations in Delhi. REVEALED PREFERENCE APPROACH Logit being computationally less complex was used for modelling the choice behaviour. The transportation system variables considered were in-vehicle travel time (IVTT), out-of-vehicle travel time (OVTT) and travel cost (COST). These attributes have been included linearly in the utility function. The socio-economic variables e.g. age, sex, income, (v) vehicle ownership were used to stratify the sample. The specification of alternatives for each commuter plays an important role in the modelling of choice behaviour. Based upon the vehicle ownership four homogeneous groups of commuters were identified with similar choice alternatives. The level-of-service variables were specified generically without a alternative specific constant for developing disaggregate mode choice models in view of the use of model for estimating the demand of innovative mode. STATED PREFERENCE APPROACH To compare the relative merit of revealed and stated preference approach,the variables used in revealed preference modelling have been used in developing stated preference design. The variables namely ^ in-vehicle travel time (IVTT), out-of-vehicle travel time (OVTT) and travel cost (COST) were varied at three levels. The values of these three variables have been derived from the operating characteristics of metro rail in Calcutta, the only example of MRTS in India. REVEALED PREFERENCE MODELS After developing multinomial logit models for different segments of commuters the set of models based on vehicle ownership criterion were found to be the best, based on their goodness-of-fit test, signs of coefficients and percent correctly estimated figures. The above segments are (i) commuters without owning vehicle who have chartered bus and DTC bus as only two choices, (ii) commuters owning motor cycle/scooter who have three choices i.e. own vehicle, chartered bus and DTC bus (iii) commuters owning car who have two choices i.e. car and chartered bus and (iv) commuters owning both motor cycle and car are those who have the global choice set. All the coefficients in the above models were found to have negative sign which is compatible with the general notion i.e. (vi) 4 with increase in the time and cost variables the utility of mode decreases. STATED PREFERENCE ANALYSIS The stated preference survey was analysed to estimate the switching characteristics of different commuter segments. It was observed that commuters presently using chartered bus are reluctant to switchover to MRTS compared to DTC bus users. Again at higher cost variations commuters of low income group have reservations in shifting. Switching over to the new mode for smaller trip length is low. The stated preference data was also used to develop binary logit models for binary choice set i.e. the existing mode and MRTS for commuter segment. The commuters segments are car user, motor cycle/scooter users, chartered bus users and DTC bus users. These models have been used to estimate demand of MRTS. COMPARISON OF REVEALED PREFERENCE AND STATED PREFERENCE APPROACHES The revealed preference model has been used for estimating the probabilities of various MRTS options. Based on the computation of revised probabilities if the MRTS probability is maximum then the commuter is predicted to have switched over to MRTS. Thus, the switching behaviour of all the commuters have been predicted. After predicting the disaggregate switching behaviour the results have been regrouped into different commuter segments. The switching behaviour of different commuter segments from stated preference models and predicted from revealed preference modelling have been compared. For individual commuter segment it was observed that switching behaviour of commuters is over-estimated by revealed preference methods. In achi-square test the significance of both the results were also compared. (vi i) CONCLUSIONS (i) Stated Preference Approach is the integrated form of sequentially developed output formats based on logical process of decision making of individual choice-makers, (ii) Stated Preference Approach gives a conservative estimation of demand of innovative mode, (iii) Chi-square test conducted to compare the estimated choice probability of MRTS by the stated and revealed preference approaches indicate that results of stated preference was significant at 10% level, (iv) Stated Preference Approach is a relatively efficient technique being less dependent upon rigorous mathematical modelling , computations and a huge data base. (v