Go back to Literature Dumping Ground
2.1: Definition of Travel Time Inaccuracy (TTI)
Travel time inaccuracy (TTI) in this paper strictly refers to how “scheduled levels of service might differ from what is delivered to the population”
(
Braga et al., 2023). That means that actual service may be worse than
scheduled but it can also be better. Travel time variability (TTV), on the other hand, looks at the variability of travel times between an origin and a
destination. This variability could manifest between two modes of transport that start their journey at the same time, or between different times of the
day, or over different days (
Noland & Polak, 2002;
Mazloumi et al., 2009).
While the causes of TTV and TTI may be the same, and TTI could induce TTV,
Chen and Botta (2026) are clear that both terms by itself are ultimately separate
issues. This is because a service that consistently runs five minutes late every day is a representation of TTI, not TTV. Meanwhile, a punctual service
that has different frequencies between morning peak hours and off-peak hours does not have TTI but produces TTV.
As will be expounded later in Section 4, this paper will look at how service delivery deviates from the schedule across four Wednesdays in June 2026.
Thus, it will look at both TTI and TTV.
2.2: Use of Real-Time Data in Accessibility Studies
Google Transit Feed Specification, later
General Transit Feed Specification (GTFS), is a static dataset comprising of at least seven
relational text files that transport agencies or operators use to share information about their scheduled services, such as which routes do they operate
(found in
routes.txt), which specific times would they run the routes (found in
trips.txt) and at what specific times would the buses call
at the specific stops along the routes that they are servicing (found in
stop_times.txt)
(
GTFS, n.d.). There is also a real-time extension called
GTFS-RealTime (GTFS-RT). Unlike its static counterpart, GTFS-RT data is provided in binary files that needs prior decoding using Google's Protocol
Buffers and provides the buses' live locations and the trips that they are serving
(
GTFS, n.d.). Some GTFS-RT feeds would also provide information about
the next stops to which these buses are heading and the estimated arrival times, but the feeds provided by UK public transport operators are known to
have left this out
(
BODS User Forum, 2025;
Webb et al., 2025). This means that static GTFS data is favoured over the dynamic
GTFS-RT data in academic circles because it is easier to access
(pun intended) and interpret the former, which is also more complete.
The path dependency of preferring GTFS data to GTFS-RT is further cemented by the advent of
R5, a Java-based routing model, and its wrappers in
R and
Python languages (
r5r and
r5py respectively), which requires public transport input to be in the format of the former
dataset.
R5's appeal lies in its ability to produce travel time estimates for multiple origins and destinations quickly
(
Conway et al., 2017), meaning that it is a useful tool for transportation researchers to
measure accessibility. It has been applied for various contexts around the world such as Brussels, Santiago and Mexico City, to name a few
(
Verbeken et al., 2025;
Marin-Flores et al., 2026;
Verduzco-Torres & McArthur, 2026). In the UK, it has been used to generate
nationwide accessibility indicator datasets (
Verduzco-Torres & McArthur, 2024),
evaluating transport infrastructure investments (
Lee & van Dijk, 2026) and
exploring healthcare accessibility inequality (
Chen & Botta, 2025).
However, static GTFS data does not reflect the dynamic nature of public transport operations. In reality, delays could arise at any part of the route due
to multiple reasons such as long boarding times, stopping at successive traffic signals, or short distances between successive stops
(
El-Geneidy et al., 2011;
Yongha et al., 2020;
Kaewunruen et al., 2021). These delays may cascade down the route, knocking
services off their schedules and cause TTI and/or TTV. Meanwhile, static data may also incorporate schedule padding, where extra time is added to
accommodate for potential delays such as the ones mentioned earlier
(
Wessel & Farber, 2019). That means that findings from accessibility studies based
on static GTFS data alone are inaccurate because they could either be overestimations or underestimations. It was
Wessel et al (2017) who first developed a method to convert GTFS-RT data
into retrospective GTFS datasets and then demonstrated its utility in exploring how TTI impacted cumulative accessibility to jobs. This paved the way for
other researchers to apply and adapt their methods to measure TTI/TTV-impacted accessibility in different contexts
(Braga et al., 2023;
Javanmard et al., 2025).
In the UK, the viability of real-time data utilisation in accessibility research increased after the introduction of the
Bus Open Data Service (BODS) in 2021, which mandated all bus
operators, among other things, to share their live location data. However, research using real-time data has been limited because there was no stable
archive of BODS GTFS-RT feeds until 2025 (
Open Innovations, 2025).
Nevertheless, there is a growing interest to utilise real-time data for accessibility research in the UK, evidenced by the recognition of how TTI could
be one of the reasons behind UK cities' lagging productivity vis-à-vis their similarly sized European counterparts
(
Strange, 2024), and the development of
code to convert BODS GTFS-RT feeds into retrospective GTFS dataset along the lines of Wessel et al
(
Open Innovations, 2024; Chen & Botta, 2026).
2.3: Accessibility Metrics in Transport Appraisal
Lorem ipsum, will be updated soon!