Modal Mismatch and Accessibility Gap in Dhaka using a Time of Day Based Approach

Abstract

Transport accessibility is an area of growing global attention among transportation planners and policymakers. This paper aims to portray spatiotemporal variations of car and public bus accessibility in the context of a city in a developing country: Dhaka, Bangladesh. The public bus system in Dhaka is characterized by a semi-formal arrangement which means there is a lack of available data, for example, General Transit Feed Specification, which poses great difficulties in the study of accessibility. Given these limitations, we have presented the concept of major destinations to analyze spatiotemporal accessibility based on the simple understanding that trip purposes, time of day, and trip destinations are interlinked in an urban area, and different locations would attract a different number of trips based on the time of day. Using a spatial autocorrelation approach, we identified the statistically significant destination clusters in Dhaka by peak and off-peak hours. We measured accessibility to the major destinations using a cumulative opportunity-based metric followed by estimation of the Modal Accessibility Gap (MAG). The findings indicated that, regardless of the trip origins and time of day, dependence on public transport puts the users in a substantially disadvantageous position. From the policy perspective, we have suggested the introduction of a formal public transport system in Dhaka, particularly targeting the traffic analysis zones with higher MAG. Such an approach would lead to better resource usage while providing enhanced public transport services for both peak and off-peak hours and limiting dependence on cars.

A paradigm shift from mobility-geared to accessibility-driven transportation policy is increasingly gaining importance worldwide for the far-reaching impacts of accessibility on urban structure, the performance of transportation systems, and various social and economic issues ( 1 – 5 ). Studies on origin-based accessibility often measure the accessibility of a location by considering all reachable opportunities. Whereas the approach of considering all reachable opportunities of interest is a good point at which to start, it may be insufficient or even misleading when it comes to policy implications. This is especially true when employment opportunities are considered as the generalized “relevant opportunities” in accessibility estimation for ease of definition, calculation, and, more importantly, absence of reliable, complete data of other spatial opportunities. For example, people tend to make more non-work trips during the off-peak period compared with work trips. Therefore, a person residing near the central business district may have high accessibility in the conventional sense, but that may be of little significance for them during off-peak hours. Rather, another individual living close to shopping areas may enjoy comparative advantages during off-peak periods.

Added to that, in an urban area, different activity locations often cluster predominantly in different parts of the city. Mandatory and discretionary activities do not necessarily cluster in the same locations. Therefore, certain locations are expected to attract traffic during commuting hours but would not automatically attract traffic during non-commuting hours. This understanding is crucial especially from the policy standpoint.

The problem with the conventional approach of measuring accessibility is that it could potentially overestimate or underestimate accessibility by focusing more on theory instead of what a person could and would realistically expect to achieve in everyday life. Putting it another way, it often ignores that trip purpose, time of day, and trip destination are always interlinked in an urban area and cannot be isolated from one another.

Public transport is the subject of renewed interest as an essential means of sustainable urban development since growing automobile dependency around the world is associated with sprawl development, health and environmental consequences, equity issues, and so forth. Car and public transport accessibility differ considerably because of time–space constraints associated with public transport travel including ingress and egress time, waiting time, fixed routes, schedules, and so forth. While the car is usually unimodal in nature, public transport does not provide door-to-door services and requires feeder modes to complete a journey. Public transport is, therefore, inherently multimodal ( 6 ). Thus, by multimodality, we mean transfer between two or more modes as part of a journey between an origin and a destination ( 7 ).

This study is an attempt to address the interlink between trip purpose, time of day, and trip destination, leading to a more reliable illustration of what opportunities a city-dweller would realistically expect to have available at different times of the day by private car and public transport. We have chosen walking as the feeder mode of the public bus. To the best of our knowledge, no systematic study has been conducted so far to portray spatiotemporal variation of the car and multimodal public transport accessibility by offering an alternative way of addressing the significant limitations of data on the transit system and spatial opportunities. Such data limitations are quite common in developing countries like Bangladesh. The current study contributes to the existing research base in three ways. Firstly, it demonstrates that trip purpose, time of day, and trip destination are linked. Secondly, it proposes an alternative way of understanding spatiotemporal variation in accessibility even under a semi-formal and mostly unregulated public transport system. Lastly, it provides a measure for policymakers and urban planners to synthesize theory and practice for efficient, effective, and equitable infrastructure investment and management decisions. Compared with the conventional approach, the approach in this study will allow policymakers to better assess whether the city is actually moving toward accessibility-based planning and pinpoint areas for potential future investment to achieve equitable distribution of resources and opportunities, particularly in the context of developing countries.

Literature Review

The classic yet the simplest definition of accessibility probably dates back to 1959 when Hansen defined it as “the potential for interaction” ( 8 ). However simple that might sound, the conceptualization and measurement of accessibility have been far from simple for urban planners and transportation experts. An operational definition of accessibility refers to the ease with which spatially distributed opportunities may be reached from a given location using a particular transportation system ( 9 – 11 ).

Accessibility is a powerful concept because it incorporates the intricate interaction of four components: land use, transportation, temporal, and individual. Although the concept of accessibility is not new, it took a long time for urban planners and policymakers to embrace accessibility-based planning instead of the traditional mobility-focused planning. More importantly, there is still a premeditated preference for mobility-focused planning over accessibility-based planning worldwide. The need for accessibility-based planning arose from the widely accepted notion of transportation being a “derived demand.” The ultimate goal of transportation planning should, therefore, be accessibility, its means being mobility, proximity, and connectivity ( 12 ).

Accessibility has been found to impact actual travel behavior and mobility patterns, for example, trip generation, mode choice, travel time, trip frequency by different purposes, and so forth ( 13 ). Several studies indicate this latent relationship: higher accessibility is associated with lower vehicle-kilometers traveled (VKT), higher non-motorized transport (NMT), and lower automobile use, lower automobile ownership, shorter average non-work trip distance, and so forth. ( 11 , 14 , 15 ). Many studies have also been dedicated to accessibility and equity, evaluating how social and infrastructural benefits are shared by different socio-economic groups of people (3, 16 –18).

The crucial role of transport mode in accessibility has been highlighted in several studies. Studies have been conducted on the variation of private automobile and public transport accessibility and have led to the development of several concepts such as modal mismatch, transport poverty, and so forth. (1, 2, 10, 16, 19 –21). In the past few years, the dynamic nature and temporal variability of accessibility have also been studied since spatiotemporal variation in accessibility affects equity and offers scope for policy intervention ( 22 – 25 ). The multimodality of public transport has received significant attention in recent years. The emerging capabilities of GIS have made it easier, and almost imperative, to capture the first mile/last mile issue in contemporary studies ( 23 , 26 , 27 ). Yet such studies are still predominantly based on developed countries even while cities in developing countries are also becoming increasingly dependent on private automobiles. Growing automobile dependency, as well as the semi-formality of the public transport system, pose unique challenges in enhancing accessibility in many developing countries ( 28 ).

Despite extensive accessibility-related studies and integration of accessibility-based policies and plans, there is still considerable debate on how to put those theories in practice. Such concerns are even more crucial for developing countries like Bangladesh. This study is an attempt to explore spatiotemporal variation in walk-bus and private car accessibility in a context where public bus system runs on a semi-formal basis which means there is a lack of rich data, posing considerable challenges in both research and policymaking. The study is an attempt to address this research gap.

Context and Study Area

The study area includes 90 traffic analysis zones (TAZs) in the area covered by the Dhaka City Corporation as demarcated in the Dhaka Urban Transport Network Development Study (DHUTS) 2010 (Figure 1). This covers an area of 126.34 km² with a population density of approximately 55,200 per km² ( 29 ). The average area of the TAZs is approximately 1.5 km² with the smallest being approximately 0.2 km² and the largest approximately 7 km².

Figure 1.

Map of the study area.

Dhaka’s transportation sector has received increasing attention in recent years. The city consists of diversified transport modes. The DHUTS survey database reveals that walking and cycling account for 22.2% of the trips; rickshaws account for 44.8%; motorcycles for 1.6%; paratransit (CNG auto-rickshaw, Mishuk) for 5.3%; cars (including jeeps, microbuses) 3.5%; public buses 10.6%; with other modes (e.g., trucks, pick-ups, school vans, and so forth.) accounting for the remaining portion. The average travel time considering all modes as revealed by the dataset is around 31 min. For different modes, the average travel time stands as follows: walking, 14 min; bicycle, 23 min; car, 43 min; paratransit, 49 min; motorcycle, 32 min; and public bus, 56 min. Managing the efficient movement of these diversified modes is, however, far from simple. Over the last few years, Dhaka has been burdened with severe and increasingly worsening traffic congestion. High population density, rapid urbanization, a deteriorating public transport system, and increasing private car ownership and use have made the situation quite challenging for policymakers and urban planners.

Like many places around the world, Dhaka is experiencing growth in numbers of motorized vehicles. It experienced a twofold surge in registered motorized vehicles between 2011 and 2019. Private cars (including microbuses and jeeps) constituted a large portion (nearly 15%) of these newly registered vehicles in Dhaka in 2019, only lagging behind motorcycles. As of 2019, there are an estimated 23 private cars/1,000 persons in Dhaka ( 30 , 31 ). Although this number seems smaller than many other countries, a very low occupancy rate of cars causes wasteful road space usage ( 32 ). Although motorcycles have always constituted a substantial portion of total vehicle registration, available databases, planning documents, and survey reports indicate that their modal share is typically lower than most other motorized and non-motorized modes ( 31 , 33 ).

On the other hand, just about 2% of these new vehicle registrations were public buses ( 31 ). The bus routes, with substantial overlapping, have mostly expanded in the north–south direction and lack connectivity with east–west routes (Figure 1). Moreover, the bus system is still semi-formal with no specific schedule or regulation, not to mention the resulting lack of a regulated and well-maintained public transport database ( 34 ). It is customary to use comprehensive General Transit Feed Specification (GTFS) databases to evaluate accessibility in developed countries. But in developing countries like Bangladesh, there is still a lack of such information or public transit structure ( 5 , 23 , 35 , 36 ). Therefore, incorporating a temporal dimension is a challenging task in this context.

The average annual growth of cars has varied between 5% and 10% in the last five years ( 34 ). Nevertheless, the city planning authority of Dhaka claims that the modal share of cars would not increase beyond 8% by 2035 given the expected introduction of MRT/BRT and an enhanced bus system ( 34 ). Here, it is to be noted that the improvement of the public transport system has been considered a prerequisite if car use is to be kept at bay. Therefore, a comparison of public transport and private car accessibility in Dhaka could reflect the extent to which the city residents enjoy the benefits of public transport.

Data and Methods

The DHUTS 2010 Household Interview Survey (HIS) data were collected from the Dhaka Transport Coordination Authority (DTCA). This database provided data from 1% samples of the households and all members in them within the study area ( 37 ). Finally, this database contained an array of socio-economic, demographic, and trip data on more than 13,000 households covering detailed information on all the trips made the day before they were interviewed. The trip database had information including trip origin and destination at TAZ level, trip purpose, travel time, mode, and so forth.

The percentages of trips attracted by purposes and different times of day indicate that the distribution of trip purposes varies by time of day as would be expected (Table 1). Because trip attraction is different for the TAZs for different times of the day, we identified the major destinations considering the temporal dimension. We aggregated the trips across TAZs for the peak and off-peak hours. The morning peak hour consisted of 8:00 to 10:30 a.m. and evening peak of 4:30 to 7:00 p.m. ( 38 ). The trip histogram generated from the database was also consistent with this information. The off-peak hours are not clearly defined in Dhaka city. Therefore, we chose 6:00 to 7:00 a.m., noon to 3:00 p.m., and 7:30 to 9:00 p.m. as off-peak hours.

Table 1.

Purposes of Attracted Trips by Time of Day

Trip purpose		Morning peak	Off-peak	Evening peak
Homeward		8.54%	63.02%	53.03%
Work		47.05%	3.80%	1.23%
Work (Self-employment)		3.59%	9.10%	6.37%
School		22.53%	4.42%	0.45%
Non-work	Shopping	4.75%	3.34%	17.07%
	Social	3.18%	2.88%	6.23%
	Recreation	0.87%	1.08%	4.06%
	Religious	0.18%	0.30%	0.41%
	Medical	1.51%	0.92%	2.29%
	Others	7.78%	11.13%	8.83%
	Non-work Total	18.27%	19.65%	38.89%
Total		100.00%	100.00%	100.00%

Applying first order queen contiguity on GeoDa, we then performed Univariate Local Moran’s I at p < 0.05 to estimate statistically significant destination clusters by time of day. We assumed that those TAZ clusters that attracted a significant number of trips were the activity zones of the city. The local Moran for a spatial unit i was calculated using Equation 1 ( 39 ).

I_{i} = (z_{i} / m_{2}) \sum_{j} w_{ij} z_{j}

(1)

where z_i, z_j are deviations from the mean and the summation over j will be based on neighboring values where $j \in J_{i}$ . w_ij are non-zero spatial weights when i and j are neighbors. Conventionally, w_ii = 0. In queen contiguity, neighbors include spatial units having shared vertices or shared edges ( 40 ).

The Local Indicator of Spatial Autocorrelation (LISA) cluster maps were generated in this process. The high-high clusters, in these maps, indicated high trip attracting TAZs surrounded by other high trip attracting TAZs and allowed us to identify TAZs with a higher concentration of trips relative to the mean number of trips. Unlike spatial outliers, the LISA cluster maps show only the core of the spatial clusters and not the actual clusters. The actual clusters include the neighbors of the cores ( 41 ). We employed this method to determine statistically significant major destination clusters.

The street network and building footprint data were obtained from OpenStreetMap (OSM) to create network datasets. We prepared two network datasets in ArcGIS, one for car and another multimodal one combining walk and bus. Bus route data collected from DTCA showed that there are 226 bus routes in Dhaka totaling in 194.87 km. Although these bus routes are approved by the Bangladesh Road Transport Authority (BRTA), buses run on these routes without any specific schedule and multiple privately-owned companies operate bus services on a single route, thereby creating a semi-formal system. While preparing the network datasets, we applied average car and bus speeds during peak and off-peak hours derived from field survey for every working day of a week for three functional categories of street (primary, secondary, local/residential). The bus stops were used as transfer points for switching between transport modes. Bus stop locations, as updated by Dhaka Metropolitan Police (DMP) in 2018, was collected through a GPS survey covering all bus routes within the study area. Street network data from OSM were used for the walking component of the multimodal network. We assumed walking speed of 4.8 km/h in the multimodal network dataset ( 23 ) and a total delay of 1 min for boarding and alighting a bus. Since there is no specific public transit timetable for Dhaka city, people tend to wait at the bus stop without knowing when the bus would arrive. For example, a bus might arrive at 8 a.m. at a stop on one day but might not arrive until 8:30 a.m. on the next day when traversing the same route. As a result, there is no way of predicting waiting time other than relying on people’s experiences. Therefore, the average waiting time for bus was collected through a primary household questionnaire survey of all members of 269 households chosen at random in 2017. The waiting times for all bus trips made by the household members on the workdays of a typical week were averaged out to feed into the multimodal network development stage. Finally, an average waiting time at the bus stop of 7.5 min was added to the multimodal network dataset.

Once the network datasets were prepared, travel times between the centroids of the building footprints and destination TAZs during peak and off-peak hours were estimated using the O-D Cost Matrix Solver in ArcGIS. Using three cut-off travel times of 30, 45, and 60 min, this procedure was performed for both transport modes: car and walk-bus. Average travel times for work and non-work trips to major destinations appeared to be around 53 min and 48 min, respectively. The average travel time ranged between 32 min and 54 min for different non-work purposes. Considering all trips, the average travel time was 57 min. Therefore, these three cut-off travel times of 30 min, 45 min, and 60 min seem reasonable for accessibility estimation in this study context.

Accessibility Calculation

The Bangladesh Bureau of Statistics (BBS) collects data on the number of people involved in different economic activities, namely Total Persons Engaged (TPE) in permanent establishments, temporary establishments, and economic households ( 42 ). These data were collected from BBS for the year 2013. TPE across TAZs was used as a proxy variable for job opportunities available. It was assumed that TPE equals the number of jobs available and therefore, no vacancies exist. Using the origin-based cumulative-opportunity accessibility metric mentioned in Equation 2, we calculated public bus and car accessibility scores for all individual building footprints within the study area ( 9 ). These accessibility scores were then aggregated by averaging across each TAZ to generate a TAZ level accessibility score.

A_{i} = \sum_{j} O_{j} f (C_{ij})

(2)

f (C_{ij}) = {\begin{matrix} 1 & if C_{ij} \leq C_{ij}^{c} \\ 0 & if otherwise \end{matrix}

(2)

where

A_i = accessibility for building i,

O_j = number of employment opportunities in major destination TAZ j,

C_ij = travel time for a trip from i to j,

C_ij^c = cut-off travel time, and

f(C_ij) = impedance function measuring the spatial separation between i and j.

The TAZ level accessibility scores allow us easier visualization and understanding of accessibility scenario at the city scale and enable the synthesis of other TAZ level socio-economic, demographic, and travel data. Yet we chose to calculate accessibility scores at a disaggregated building footprint level first, and then aggregate them to deal with the Modifiable Areal Unit Problem (MAUP). Such an approach potentially reduces some likely aggregation errors arising from calculating accessibility directly on a larger spatial scale ( 43 ).

Modal Accessibility Gap Calculation

A simple ratio of car-bus accessibility at the TAZ level is the easiest and simplest way to calculate the accessibility gap. However, it potentially produces unusually large values, which are difficult to represent, compare, and interpret. The Modal Accessibility Gap (MAG) between car and walk-bus was, therefore, calculated using Equation 3 ( 19 ).

MAG = \frac{A_{c} - A_{p}}{A_{c} + A_{p}}

(3)

where A_p and A_c are population-weighted zonal accessibility indexes for public bus and car, respectively where

A_{p} = \sum_{i = 1} \frac{M_{i}}{M} A_{i}^{p}

(3)

A_{c} = \sum_{i = 1} \frac{M_{i}}{M} A_{i}^{c}

(3)

where M_i is the total population in TAZ i and M is the total population in the city.

This standardized index ranges from –1 to 1, thereby allowing easy and quick comparison. An extreme 1 indicates the outright advantage of the car over public transport and vice versa when the value is –1. MAG is 0 when accessibility by car and bus are equal. On the other hand, when a TAZ has no accessibility by either of the modes, the index cannot be computed mathematically. Conceptually, as well, there exists no “gap” in an actual sense. This standardized index has another advantage over the simple ratio approach that it allows measurement of the gap when a TAZ is not accessible by one of the modes. In such cases, the simple ratio approach would produce a misleading result: based on the denominator and numerator, the ratio would either be 0 or infinite. Thus, the MAG index is particularly suitable for our context because a good number of TAZs had car accessibility but no walk-bus accessibility.

Results and Discussion

Time of Day and Statistically Significant Major Destinations

Our identification of statistically significant major destinations by different times of the day confirms that morning peak, off-peak, and evening peak hour trips cluster in different locations within the city (Figure 2). Moreover, the earlier table (Table 1) and the major destination maps (Figure 2) indicate that trip purpose, time of day, and trip destinations are interlinked.

Figure 2.

Statistically significant major destinations by time of day (from left to right: morning peak, off-peak, and evening peak).

The statistically significant major destination cluster for the morning peak hour includes parts of Motijheel, Ramna, and Old Dhaka that contain the commercial hub of the city as well as several large secondary schools and higher secondary colleges well-known across the city. This confirms that a higher share of work and school trips are made around these hours. The off-peak hour cluster involves a mix of TAZs: some with residential characteristics such as Kalabagan, Mirpur, Ramna; shopping areas such as New Market; and the commercial-industrial area of Tejgaon. These areas have several primary and secondary schools which run on split shifts. The day-shift of these schools contributes to the off-peak school trips. Given the mix of trip purposes during the off-peak hour, such a cluster seems consistent. Homeward and non-work trips dominate the evening peak hour. Residential areas such as Mohammadpur, Mirpur, Dhanmondi (which host a wide variety of non-work activity locations such as shopping malls, restaurants, and food courts), along with shopping areas such as New Market provide enough evidence to explain the evening peak major destination cluster. Here, it is to be emphasized that these major destinations consist of spatial and temporal dimensions simultaneously.

Non-work trips constitute 23.6% of all the daily trips in Dhaka compared with 22.4% for work trips. Thus, discretionary activities, for example, healthcare, shopping, social relations, recreation, and so forth, are essential parts of human life. The identification of major destinations allowed us to consider both work and non-work trips.

A paired sample t-test was conducted to compare the difference in travel time and distance to major destinations and destinations other than the major destinations. The test revealed that there is a significant difference in both average travel time and distance between trips to major destinations and other destinations. The average travel time to major destinations is 4.48 min more than the travel time to other destinations (35.20 min and 30.72 min, respectively, t_13949.98 = 17.73, p < 0.001) and the average travel distance to major destinations is 0.65 km more than the travel distance to other destinations (3.25 km and 2.60 km respectively, t_14240.1 = 21.38, p < 0.001). This means that people travel further and for a longer time to reach the major destinations, suggesting the relative importance of the major destination clusters in offering crucial transport infrastructure services and facilities to the city dwellers.

Accessibility Gap and Modal Mismatch

Comparison of Car and Bus Accessibility

The difference in car and walk-bus accessibility is staggering (Table 2). On average, traveling 30 min by car would make, at least, an estimated 13% of the jobs in the major destinations reachable regardless of time of day. An hour of travel by car would substantially increase this number to an estimate of at least 55.08% of the jobs being accessible within that time. By contrast, at most, an estimated 12.84% of the jobs can be reached within 60 min of travel by walk-bus, even lower than the jobs reachable within 30 min of travel by car.

Table 2.

Percentage of Jobs Reachable by Car and Walk-Bus at Different Cut-off Travel Time

Time of day	Travel time (minutes)	Mode	% Estimated reachable jobs
Morning peak	30	Walk-Bus	4.84
Morning peak	30	Car	53.28
	45	Walk-Bus	9.13
	45	Car	70.21
	60	Walk-Bus	12.84
	60	Car	81.19
Off-peak	30	Walk-Bus	2.34
	30	Car	30.15
	45	Walk-Bus	4.89
	45	Car	58.29
	60	Walk-Bus	9.21
	60	Car	80.29
Evening peak	30	Walk-Bus	2.22
Evening peak	30	Car	13.00
	45	Walk-Bus	4.68
	45	Car	32.96
	60	Walk-Bus	8.41
	60	Car	55.08

Proximity to Major Destinations and Accessibility

Being close to major destinations yields better accessibility for both car and bus users, and understandably so. However, proximity to the major destinations is more crucial for bus users than car users. The scatter plots in Figure 3 show the difference between car and walk-bus accessibility of the TAZs with the average distance from the major destinations. Every TAZ has a pair of points in each of the scatter plots: one for the car (in orange) and another for the walk-bus (in green). All the scatter plots tell a similar story.

Figure 3.

Accessibility scores by average distance from major destinations at 30, 45, and 60 min cut-off travel time for different times of the day.

Walk-bus accessibility is noticeably lower compared with car accessibility, even at small average distances from the major destinations. The entire city lacks accessibility to jobs by walk-bus compared with car. For several TAZs, walk-bus accessibility scores are so low that the car is the prevailing alternative. Moreover, compared with the car, accessibility by walk-bus falls to zero at a markedly smaller average distance from the major destination clusters.

However, the most important observation is that even car users who are a long way from the major destinations still enjoy substantially better accessibility compared with bus users residing relatively closer. Proximity to the major destinations evidently portrays disparate benefits for car and bus users. The gap is so wide that no matter how close bus users live to the major destinations, they are extremely unlikely to enjoy the levels of accessibility enjoyed by car users. This echoes several previous studies which found that travel modes play a relatively more important role than location in determining accessibility ( 1 , 12 , 44 , 45 ). This supports the classic “modal mismatch” theory from both a spatial and temporal perspective: regardless of where a person resides or when a trip is made, dependence on public transport puts them at a substantial disadvantage.

Modal Accessibility Gap

To confirm our observations and findings of difference in accessibility by car and bus, we tested whether the difference is statistically significant. Our data did not follow normal distribution because of a considerable number of TAZs with accessibility scores of 0, especially at lower cut-off travel times and by walk-bus. Therefore, we performed a non-parametric sign test for each of the nine pairs of accessibility score sets, that is, for all the times of the day and cut-off travel times. For all the pairs, we found that car accessibility was significantly greater than walk-bus accessibility (p = 0.000 for each pair).

Next, we attempted to look more closely at the accessibility gap at the TAZ level. We calculated the MAG using Equation 3.

Average MAG values are close to 1 indicating extremely poor accessibility by walk-bus throughout the day. The gap between walk-bus and car accessibility tends to decrease with increasing cut-off travel time (the only exception is 30- and 45-min cut-off travel times during evening peak) (Table 3).

Table 3.

Average Modal Accessibility Gap between Car and Walk-Bus

	30 min	45 min	60 min
Morning peak	0.88	0.84	0.81
Off-peak	0.82	0.82	0.78
Evening peak	0.69	0.72	0.71

This is likely because the effect of walking and waiting time for public transit do not necessarily change with different cut-off travel times. Public transit travel is highly constrained by specific scheduling and routes. Moreover, the first mile/last mile transit access involves a considerable portion of total travel time ( 26 ). These are, in a nutshell, reasons for comparatively lower public transit accessibility than that enjoyed by the car in any city. Therefore, as the effect of walking and waiting times do not always increase with increasing travel time, the difference between public bus and car accessibility tends to narrow.

In general, MAG increases as the TAZs spread outward from the major destination clusters regardless of time of day (Figures 4–6). This is consistent with our discussion so far that proximity to major destinations has a substantially different effect on car and bus. During off and evening peak, with increasing travel time, MAG increases for some TAZs (Figures 5 and 6). This indicates that having more reachable opportunities with increasing travel time does not always reduce the accessibility gap for individual TAZs. Rather it may sometimes widen the gap by simultaneously making a disproportionately higher number of opportunities available by car. Since major destination clusters vary by time of day, living in proximity to morning peak hour cluster may make the journey to work easier for walk-bus riders, but not the non-work trip makers. A similar thing can be said for other times of the day.

Figure 4.

Modal Accessibility Gap (MAG) during morning peak at cut-off travel times of: (left) 30 min, (middle) 45 min, and (right) 60 min.

Figure 5.

Modal Accessibility Gap (MAG) during off-peak at cut-off travel times of: (left) 30 min, (middle) 45 min, and (right) 60 min.

Figure 6.

Modal Accessibility Gap (MAG) during evening peak at cut-off travel times of: (left) 30 min, (middle) 45 min, and (right) 60 min.

On the flip side of the coin, private car gives substantially more freedom, flexibility, and ease in overcoming the friction associated with travel ( 46 ). The unsurprising result is greater freedom and ease in reaching both work and non-work activities. Public bus users are, therefore, disadvantaged throughout the day more as a result of the mode they use than their location of residence or time of travel.

Conclusion and Policy Implications

In this study, we tried to highlight the potential of considering the temporal dimension in accessibility analysis in the context of a public transport-specific data-deprived developing country city. Nevertheless, the importance of the temporal dimension does not diminish under such settings, given the prevalence of a higher share of daily non-work trips compared with work trips. We add to the growing body of accessibility studies by finding an alternative way to integrate the spatial and temporal dimension in accessibility evaluation.

While interpreting the results, it should be borne in mind that the statistically significant major destinations are different for different times of the day. Thus, the number of available opportunities is also different. A change in accessibility for a certain TAZ from one time of day to another could be attributed to the difference in travel speed, difference in total employment opportunities available in the major destinations, or a combination of both. Therefore, care should be taken while comparing results across different times of the day.

To the best of our knowledge, studies on accessibility addressing the interlink of trip purpose, time of day, and trip destination in an urban area are rare ( 46 ). Yet considering this interlink and multimodality of public transport simultaneously is even rarer. Our approach of identifying major destinations by different times of day is particularly important to get an understanding not just of work trips but also of accessibility scenarios for several discretionary activities such as healthcare, shopping, social relations, recreation, and so forth. Down that line, this approach further leads to an insight that only a handful of locations are equally accessible (or inaccessible) to car and walk-bus users all day long. Locations that enjoy better accessibility during peak working hours are likely to experience poor accessibility by both walk-bus and car when it comes to opportunities such as healthcare or shopping. The paucity of accessibility, however, is substantially more prominent for walk-bus.

The approach outlined in this paper was chosen to partly hint at the implications of national transport policies in developing country city like Dhaka. Transport planning here, more often than not, has an underlying preference for road construction, thereby focusing on faster car movement ( 33 , 34 ). Such mobility-oriented investments often cater to only a small share of the population, ignoring the needs of the majority who are dependent on public transport or non-motorized modes.

However, when a large amount of public money is spent on providing services and building new infrastructure, it is imperative to ensure that the money is used in the right place, at the right time, and for the people with limited options who need it the most. The major destinations lead us to those locations where people go the most. Those destinations, therefore, need to be served well by public transport as a priority to achieve greater social equity and to reduce environmental pollution and congestion.

This study illustrated that walk-bus accessibility was significantly lower than car accessibility at any time of day in Dhaka. The MAG between car and walk-bus is striking. Absence of a formal, reliable public bus system followed by the lack of east–west connectivity contributes to the staggering accessibility gap between car and bus. The government has recently undertaken the Bus Route Rationalization initiative which aims at scrapping the unnecessary overlapping bus routes and bringing the bus routes under a franchise system with all buses operating on the same route under a single company. Our approach of identifying the major destination clusters could be used in this initiative for delineating the bus route network.

The MAG maps demonstrate an increasing trend of MAG values moving outward from the major destination clusters. This underscores the importance of improving public transport accessibility to major destinations in the TAZs with higher MAG, lying relatively far from the major destinations. The time of day approach has the potential to shape a well-connected bus route with a reliable schedule system, not only for mandatory trips but also for discretionary trips which are essential for the city. New investments in the public transport sector should take into account how the major destinations could be well-served throughout the day, thereby reducing accessibility gaps between car and bus. Planners and policymakers could be benefited greatly by using the approach and findings from this study to develop a smart, GPS-enabled formal public transport system which does not exist in the city today. For example, higher public bus frequencies during the morning peak could be initiated to morning-peak major destination clusters, particularly from the TAZs with high MAG. A relatively lower frequency might be acceptable to off-peak or evening peak clusters during this period. In doing so, TAZs with higher MAG should be focused so that they achieve a better level of public transport accessibility. Such an approach would lead to better resource usage while providing enhanced public transport services. Improved public transport services for peak and off-peak hours would in effect benefit the socially disadvantaged groups of the population who have to depend on public transport on a daily basis. These measures, to be successful, have to be supplemented with reliable and effective public transport design strategies including first mile/last mile connectivity. Future research might consider shedding light on this premise.

The study has encountered limitations in data. In the context of Bangladesh, finding a reliable and up-to-date database is extremely challenging. We had to depend on the DHUTS 2010 survey data since these are the most consistent data available for Dhaka. Also, we had to depend on proxy job opportunity data, assuming that it would serve as a representative base for all facilities. A more accurate picture could be achieved if a reliable database of other relevant facilities was available for Dhaka. However, we maintain that this study could serve as a solid starting point to understand spatiotemporal variation in accessibility with limited data availability. Accessibility scores have been weighted by the residential population for all times of the day because of data unavailability of the nighttime population. Future studies could consider using population by time of day for a more precise result. The methodology adopted here did not reflect individuals’ characteristics, for example, age, gender, physical capability, and so forth. Similarly, we did not take into account the monetary costs involved, for example, the ability to own and maintain a car or pay for public transit. The only cost considered here is the travel time spent. But we acknowledge that an individual’s characteristics influence accessibility to a large extent and future research could explore these issues.

In conclusion, this study highlighted the importance of addressing the accessibility gap between car and public bus. It also demonstrated the possibility of developing a formal, demand-sensitive public transport system in Dhaka and the importance of a rich database which is critical for effective policymaking and planning as well as putting those policies and plans into practice. Kenya, for example, has developed their GTFS database and is using that to achieve the transportation-related Sustainable Development Goals (SDG) ( 47 ). The public transport sector of Dhaka needs a total transformation as well: from routes to schedules and everything related to the system. The city planning authority of Dhaka acknowledges that a well-developed public bus system is a prerequisite to control private car use. Our study supplements that by showing the dire need to improve public bus accessibility in Dhaka and outlines an approach to do so in a demand-sensitive manner.

Footnotes

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: Paromita Nakshi, Anindya Kishore Debnath; data collection: Paromita Nakshi, Anindya Kishore Debnath; analysis and interpretation of results: Paromita Nakshi; draft manuscript preparation: Paromita Nakshi, Anindya Kishore Debnath. Both authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Data Accessibility Statement

OpenStreetMap data were collected from . All other datasets were directly collected from the respective government organizations as mentioned in the article.

References

Kawabata

Job Access and Employment Among Low-Skilled Autoless Workers in US Metropolitan Areas. Environment and Planning A: Economy and Space, Vol. 35, No. 9, 2003, pp. 1651–1668.

Grengs

Job Accessibility and the Modal Mismatch in Detroit. Journal of Transport Geography, Vol. 18, No. 1, 2010, pp. 42–54.

Deboosere

El-Geneidy

Evaluating Equity and Accessibility to Jobs by Public Transport Across Canada. Journal of Transport Geography, Vol. 73, 2018, pp. 54–63.

Kawabata

Shen

Job Accessibility as an Indicator of Auto-Oriented Urban Structure: A Comparison of Boston and Los Angeles with Tokyo. Environment and Planning B: Planning and Design, Vol. 33, No. 1, 2006, pp. 115–130.

Cui

Boisjoly

El-Geneidy

Levinson

Accessibility and the Journey to Work Through the Lens of Equity. Journal of Transport Geography, Vol. 74, 2019, pp. 269–277.

Krygsman

Dijst

Multimodal Trips in The Netherlands: Microlevel Individual Attributes and Residential Context. Transportation Research Record: Journal of the Transportation Research Board, 2001. 1753: 11–19.

Clifton

Muhs

C. D.

Capturing and Representing Multimodal Trips in Travel Surveys. Transportation Research Record: Journal of the Transportation Research Board, 2012. 2285: 74–83.

Hansen

W. G.

Accessibility and Residential Growth. Doctoral dissertation. Massachusetts Institute of Technology, MA, 1959.

Shen

Location Characteristics of Inner-City Neighborhoods and Employment Accessibility of Low-Wage Workers. Environment and Planning B: Planning and Design, Vol. 25, No. 3, 1998, pp. 345–365.

10.

Kawabata

Takahashi

Spatial Dimensions of Job Accessibility by Commuting Time and Mode in Tokyo Metropolitan Area. Theory and Applications of GIS, Vol. 13, No. 2, 2005, pp. 31–40.

11.

Handy

S. L.

Regional Versus Local Accessibility: Neo-Traditional Development and its Implications for Non-Work Travel. Built Environment, Vol. 18, No. 4, 1992, pp. 253–267.

12.

Grengs

Levine

Shen

Intermetropolitan Comparison of Transportation Accessibility: Sorting out Mobility and Proximity in San Francisco and Washington, D.C. Journal of Planning Education and Research, Vol. 29, No. 4, 2010, pp. 427–443.

13.

Vickerman

R. W.

Accessibility, Attraction, and Potential: A Review of Some Concepts and their Use in Determining Mobility. Environment and Planning A: Economy and Space, Vol. 6, No. 6, 1974, pp. 675–691.

14.

Hanson

Schwab

Accessibility and Intraurban Travel. Environment and Planning A: Economy and Space, Vol. 19, No. 6, 1987, pp. 735–748.

15.

Kockelman

K. M.

Travel Behavior as Function of Accessibility, Land Use Mixing, and Land Use Balance: Evidence from San Francisco Bay Area. Transportation Research Record: Journal of the Transportation Research Board, 1997. 1607: 116–125.

16.

Preston

Rajé

Accessibility, Mobility and Transport-Related Social Exclusion. Journal of Transport Geography, Vol. 15, No. 3, 2007, pp. 151–160.

17.

Allen

Farber

Sizing up Transport Poverty: A National Scale Accounting of Low-Income Households Suffering from Inaccessibility in Canada, and What to do about it. Transport Policy, Vol. 74, 2019, pp. 214–223.

18.

Páez

Farber

Mercado

Roorda

Morency

Jobs and the Single Parent: An Analysis of Accessibility to Employment in Toronto. Urban Geography, Vol. 34, No. 6, 2013, pp. 815–842.

19.

Kwok

R. C.

Yeh

A. G.

The Use of Modal Accessibility Gap as an Indicator for Sustainable Transport Development. Environment and Planning A: Economy and Space, Vol. 36, No. 5, 2004, pp. 921–936.

20.

Niedzielski

Śleszyński

Analyzing Accessibility by Transport Mode in Warsaw, Poland. Geographia Polonica, Vol. 81, No. 2, 2008, pp. 61–78.

21.

Kawabata

Spatiotemporal Dimensions of Modal Accessibility Disparity in Boston and San Francisco. Environment and Planning A: Economy and Space, Vol. 41, No. 1, 2009, pp. 183–198.

22.

Owen

Levinson

D. M.

Modeling the Commute Mode Share of Transit using Continuous Accessibility to Jobs. Transportation Research Part A: Policy and Practice, Vol. 74, 2015, pp. 110–122.

23.

Farber

Morang

M. Z.

Widener

M. J.

Temporal Variability in Transit-Based Accessibility to Supermarkets. Applied Geography, Vol. 53, 2014, pp. 149–159.

24.

Neutens

Schwanen

Witlox

De Maeyer

Evaluating the Temporal Organization of Public Service Provision using Space-Time Accessibility Analysis. Urban Geography, Vol. 31, No. 8, 2010, pp. 1039–1064.

25.

Cui

Levinson

Accessibility and the Ring of Unreliability. Transportmetrica A: Transport Science, Vol. 14, No. 1–2, 2018, pp. 4–21.

26.

Boarnet

M. G.

Giuliano

Hou

Shin

E. J.

First/Last Mile Transit Access as an Equity Planning Issue. Transportation Research Part A: Policy and Practice, Vol. 103, 2017, pp. 296–310.

27.

Mavoa

Witten

McCreanor

O’Sullivan

GIS Based Destination Accessibility Via Public Transit and Walking in Auckland, New Zealand. Journal of Transport Geography, Vol. 20, No. 1, 2012, pp. 15–22.

28.

Campbell

K. B.

Rising

J. A.

Klopp

J. M.

Mbilo

J. M.

Accessibility Across Transport Modes and Residential Developments in Nairobi. Journal of Transport Geography, Vol. 74, 2019, pp. 77–90.

29.

Bangladesh Population and Housing Census 2011. Urban Area Report. Bangladesh Bureau of Statistics, 2014.

30.

Demographia. Demographia World Urban Areas. 15th Annual Edition. 2019.

31.

Bangladesh Road Transport Authority. Number of Registered Motor Vehicles in Dhaka (Yearwise). 2019. http://www.brta.gov.bd/.

32.

Siddique

M. A. B.

Choudhury

C. F.

Modelling the Behavioural Response to Congestion Pricing in Dhaka, Bangladesh. Transportation in Developing Economies, Vol. 23, No. 23, 2017, pp. 1–13.

33.

Japan International Cooperation Agency, Dhaka Transport Coordination Authority. The Project on the Revision and Updating of the Strategic Transport Plan for Dhaka. Urban Transport Policy, 2015.

34.

Rajdhani Unnayan Katripakkha. Draft Dhaka Structure Plan 2016-2035. n.d.

35.

Widener

M. J.

Minaker

Farber

Allen

Vitali

Coleman

P. C.

Cook

How do Changes in the Daily Food and Transportation Environments Affect Grocery Store Accessibility?

Applied Geography, Vol. 83, 2017, pp. 46–62.

36.

Stępniak

Pritchard

J. P.

Geurs

K. T.

Goliszek

The Impact of Temporal Resolution on Public Transport Accessibility Measurement: Review and Case Study in Poland. Journal of Transport Geography, Vol. 75, 2019, pp. 8–24.

37.

Preparatory Survey Report on Dhaka Urban Transport Network Development Study (DHUTS) in Bangladesh. Japan International Cooperation Agency, 2010.

38.

Nippon Koei Co. Ltd. Study on Dhaka Mass Rapid Transit East-West Line Project in Bangladesh. Nippon Koei Co. Ltd, 2015.

39.

Anselin

Local Indicators of Spatial Association—LISA. Geographical Analysis, Vol. 27, No. 2, 1995, pp. 93–115.

40.

Anselin

Contiguity-Based Spatial Weights. 2018. https://geodacenter.github.io/workbook/4a_contig_weights/lab4a.html.

41.

Anselin

Local Spatial Autocorrelation (1). 2019. https://geodacenter.github.io/workbook/6a_local_auto/lab6a.html.

42.

Economic Census 2013 (District Report: Dhaka). Bangladesh Bureau of Statistics, 2016.

43.

Apparicio

Abdelmajid

Riva

Shearmur

Comparing Alternative Approaches to Measuring the Geographical Accessibility of Urban Health Services: Distance Types and Aggregation-Error Issues. International Journal of Health Geographics, Vol. 7, No. 1, 2008, p. 7.

44.

Hess

D. B.

Access to Employment for Adults in Poverty in the Buffalo-Niagara Region. Urban Studies, Vol. 42, No. 7, 2016, pp. 1177–200.

45.

Ong

P. M.

Miller

Spatial and Transportation Mismatch in Los Angeles. Journal of Planning Education and Research, Vol. 25, No. 1, 2005, pp. 43–56.

46.

Currie

Gap Analysis of Public Transport Needs: Measuring Spatial Distribution of Public Transport Needs and Identifying Gaps in the Quality of Public Transport Provision. Transportation Research Record: Journal of the Transportation Research Board, 2004. 1895: 137–146.

47.

Fried

Tun

T. H.

Klopp

J. M.

Welle

Measuring the Sustainable Development Goal (SDG) Transport Target and Accessibility of Nairobi’s Matatus. Transportation Research Record: Journal of the Transportation Research Board, 2020. 2674: 196–207.