Page 368 - 360.revista de Alta Velocidad - Nº 6

Yang, Haoran. Dobruszkes, Frédéric. Wang, Jiao’e. Dijst, Martin.

nodes can be translated into the external interaction (Robinson, 2005), which means
the derived linkages of people, information, and service from node attributes cannot
reflect in which direction and to what extent flows are actually produced by people (Neal,
2010). Therefore, a better alternative approach is based on actual corporeal flows in the first
transport infrastructure layer by means of either schedule data (the supply side) or actual
passenger data (the demand side).. Airline scheduled seats have been used to investigate
the structure of world cities at the global scale (e.g., Smith & Timberlake 2001; Choi et
al. 2006; Derudder & Witlox 2005) and inter-regional airline transport linkages in Europe
(Derudder and Witlox, 2009; Van Nuffel et al., 2010), the USA (Derudder et al., 2013) and
China (Lao et al., 2016; Ma and Timberlake, 2008). In contrast, only a few scholars have
considered HSR travel to investigate interactions between cities. For instance, Zhang et
al. (2016) used the HSR time schedule data to approximate the actual passenger flows to
uncover the relationships of cities in the Yangzi River Delta (YRD) region in China and Hall
& Pain (2006) used the scheduled train flows to identify the polycentric urban regions in
Europe.

However, both airline and HSR data raise several issues. First, it is common to consider
supply-related data (typically the number of seats offered between two cities, or sometimes
train frequencies or seatkm’s). The rationale for supply-side data relies on the fact it
says something about carriers’ strategies that are expected to draw networks according
to existing and potential interactions between places served. However, the supply is by
definition larger or equal to the demand, so at best it can be considered as a proxy for
actual flows of people (Neal, 2014). Second, supply or demand data are usually given at the
individual legs of trips rather than the trip as a whole. For instance, if air or rail passengers
travel from A to B where they connect to C, usual figures would count the number of
(airline or HSR) seats or passengers between A and B and between B and C but not between
A and C via B. As a result, transfers distort the picture of actual intercity relationships
(Derudder et al., 2010; Derudder and Witlox, 2008, 2005). As far as air travel is concerned,
some researchers have addressed this issue by using the so-called MIDT dataset, which is
based on actual origins/destinations air travellers flew from/to (Derudder et al., 2007).
However, information is based on bookings made through global distribution systems (GDS).
It means that those travellers who directly book on airlines’ websites are not included.
This could arguably lead to biases, for instance, an underestimation of people flying by
1
low-cost airlines .
Finally, HSR timetables are difficult to convert into the number of seats for two reasons.
First, many HSR routes are served by heterogeneous rolling stock (e.g., shorter vs. longer
trains or single- vs. double-deck trains). This means that if a train operator would pursue
a high-frequency strategy (that is, the operation of frequent services but with likely less
capacity per train), the alleged interactions between cities derived from HSR frequency
would be biased. Second, provided the number of HSR seats is nevertheless available
(e.g., from the train operator or thanks to homogeneous rolling stock), one still needs
to consider that most high-speed trains (HSTs) call at several intermediate stations. This
involves uncertainties on how seats are split between the various city-pairs thus served.
For instance, if a Beijing (A) to Shanghai (D) HSR service calls at Jinan (B) and Nanjing (C),
then seats are potentially sold for A-B, A-C, A-D, B-C, B-D and C-D city-pairs. Either the
train operator pre-allocates seats to all pairs or the actual bookings make the split change
in real time. But in both cases, this information is usually not available to researchers. It
is thus not surprising that Yang et al. (2017) found that the scheduled train flows actually

1 In Europe for instance, European low-cost airlines have long kept out of GDS to avoid extra costs.

366 360.revista de alta velocidad