A comprehensive model to capture the preference for mass rapid transit in Dhaka

Dhaka, the capital of Bangladesh and one of the fastest growing megacities of the world, is already subjected to acute traffic congestion on a regular basis. Increasing the physical capacity to relieve congestion is however not feasible since already more than 70% of the area is built-up (Bari and Hasan, 2001). This has recently prompted the Government to prioritize the introduction of Mass Rapid Transit (MRT) options like Bus Rapid Transit (BRT) and Metro Rail in the city. Planning these MRT options however require rigorous mode choice models that can be used to predict ridership and quantify the willingness-to-pay (WTP) of the travellers. Though Dhaka is an old city (dating back to 16 th century), very few travel demand models have been developed for the city so far. Among the previous studies, four step travel demand models were developed in Dhaka Metropolitan Area Integrated Transport Study (DITS, 1993), Strategic Transport Plan (STP, 2005) and Dhaka Urban Transport Network Development Study (DHUTS, 2010) as well as by Habib (2002) and Hasan (2007) . However, in each case, the mode choice component was simplified and had grave limitations. In DITS, the mode choice model was simplified to a binomial choice model between private and public modes. In Habib (2002), an MNL model structure was adapted for the mode choice but the calibration results were counterintuitive with positive sign of the coefficients for time and cost parameters. In STP (2005), which is the most extensive travel demand model for Dhaka in recent years, a wide-scale household interview survey has been conducted for the first time. In the mode choice component of STP (2005), only two modes were considered i.e. Public Transport (PT) and Individualized Motorized Vehicles (IMV). In the IMV group, cars and taxis were grouped together overlooking their very different attributes (e.g. running cost, availability, accessibility, etc.) and the non-motorized vehicles (rickshaw) were not considered in spite of the fact that 37% of the person trips in Dhaka were made by rickshaw as reported in the same study (STP, 2005). Further, the STP model has adapted pre-set rules for determining choice-sets and ignored the heterogeneity among respondents. In Hasan (2007), a rule based choice model was adapted for car and a Multinomial Logit (MNL) model was adapted for the choice among rickshaw, auto-rickshaw, taxi and bus. Hasan’s model was based on STP data but the level-of-service (LOS) variables were updated using supplemental survey (for cost) and outputs of the software EMME/2 (for travel time). The potential measurement errors introduced in this process have however been ignored. In DHUTS (2010), a two step mode choice model has been developed where only two explanatory variables have been used: travel cost and Origin-Destination (O-D) shortest path distance (derived from network analysis). As evident from the description above, the existing mode choice models for Dhaka are based on pre set rules, ad-hoc choice-sets and network derived LOS values (without any correction for measurement errors). Further, the models are not robust enough to account for the new MRT, particularly since the LOS of MRT will vary significantly from the current modes. It may be noted that, the limitations of the available datasets played key roles behind the deficiencies of the previously developed mode choice models and this has prompted the current research where we present a comprehensive mode choice model which overcomes the limitations of the previous models and is robust enough to capture the preferences for the proposed MRT modes. In this paper, the STP data have been explored in detail, the key modeling issues have been identified and modeling approaches have been proposed to account for the data limitations. The improvements from the proposed approaches have been demonstrated by comparing the Value of Time (VOT) values. The rest of the paper is organized as follows: A short description of the Revealed Preference (RP) data highlighting the main limitations of the data and the description of the Stated Preference (SP) data collected as part of this research are presented first. This is followed by a description of the model framework. In the subsequent section, the estimation results of all the model components are presented which is followed by the VOT comparisons. The summary of findings and directions of future research are presented in the end.