Bulletin of geography. Socio–economic series

W. Zhang, B. Derudder, J. Wang, F. Witlox / Bulletin of Geography. Socio-economic Series / 31 (2016): 145–160

155

Fig. 4. The 15 largest inter-city links in the original network and the transformed net-

work (up figure: The original network; down figure: The transformed network)

Source: Own studies

 - 10.1515/bog-2016-0010

Downloaded from De Gruyter Online at 09/23/2016 11:26:34AM

via Universiteit Antwerpen

---

W. Zhang, B. Derudder, J. Wang, F. Witlox  / Bulletin of Geography. Socio-economic Series / 31 (2016): 145–160

156

Fig. 5. The rank-size distributions of cities’ degree centralities in the original network and the transformed network

Source: Own studies

5.  A benchmark test using the data 

on Weibo users’ inter-city movements

In order to evaluate the validity of our method, we 

need compare the transformed network to a measure 

of actual flows of people. Due to the difficulty of 

finding a corresponding database of flows of HSR 

passengers, here we utilize a database of Weibo

4

-us-

ers’ inter-city movements, which represents a specif-

ic part of tangible flows of people. It can be argued 

that the flows of HSR passengers and Weibo-users’ 

inter-city movements have similar characteristics. 

The reason is that they serve relatively similar user 

markets: the market of HSR is mainly oriented to 

business travel and leisure tourism of citizens with 

certain economic means (Wu et al., 2013; also see 

Zheng and Kahn, 2013: ‘poor rural migrants would 

not choose bullet trains’); and most of social media 

users are young adults who have certain econom-

ic capacities that include the use of smartphones. 

Both the collection and subsequent transformation 

of Weibo data follow the methodology developed in 

Zhang et al. (2015); here we summarize the main 

tenets.

Similar to other social media services (such as 

Facebook and Gowalla), Weibo users are allowed 

to share their location through a mobile application 

that is commonly known as a geo-tagged server, thus 

generating massive location records contributed 

by millions users. We transform the geo-tagged 

information into individual travelling trajectories 

by connecting users’ registered place and their 

visited cities. In practice, we employ an Application 

 - 10.1515/bog-2016-0010

Downloaded from De Gruyter Online at 09/23/2016 11:26:34AM

via Universiteit Antwerpen

---

W. Zhang, B. Derudder, J. Wang, F. Witlox / Bulletin of Geography. Socio-economic Series / 31 (2016): 145–160

157

Programming Interface (API) provided by Weibo 

to crawl all Weibo-users’ travel records submitted 

within the YRD region from March to August 2014. 

This dataset contains 3 million inter-city footprints; 

each record represents a directional inter-city flow 

of a person. Finally, the directional network of 

Weibo users’ flows was converted to an undirected 

one by combining opposite directional links.

By means of a Pearson correlation measure, we 

first compare the similarity between both networks 

(the transformed network and the original network) 

and the benchmark network of Weibo-users’ inter-

city movements in terms of cities’ connectivities. 

The correlation coefficients show that, in gener-

al, there is a more similar relationship between the 

transformed network and the benchmark network 

(r = 0.87 at the 0.001 significance level) which ex-

ceeds the coefficient for the original network 

(r = 0.76 at the 0.001 significance level). We also 

plot the rank-size distribution of cities’ degree cen-

tralities in the Weibo-users’ movements network 

(fig. 5) to compare networks’ structural similarity. 

An intuitive sense is that the curve of the Weibo-

users’ network is closer with the curve of the 

transformed network. We mathematically compute 

the flattening ratio of the Weibo-users’ network 

(0.29), which is indeed closer with the flattening 

ratio (0.23) of the transformed network.

6. Conclusion

The purpose of this paper has above all been 

methodological: we propose to rethink some of the 

discrepancies between physical infrastructure net-

works and actual flows occurring in these networks, 

focusing on the lens of the railway system. We did 

so by (1) assessing some limitations in commonly 

used measures of inter-city rail connections and (2) 

setting out an alternative approach to approximat-

ing passenger flows in railway networks.

Previous empirical research on measuring 

inter-city linkages through the lens of railway 

infrastructures has tended to use proxy strategies, 

where (1) measuring the potential for interactions 

by train and (2) measuring the volume of trains 

making inter-city connections stand out as the 

two main strands due to the lack of data on actual 

passenger flows. However, the method of measuring 

the potential for interactions only mirrors the quality 

or efficiency of train transport infrastructures itself 

rather than considering the ‘direct demand’ for inter-

city linkages. And, the proxy of using the volume of 

trains structurally predetermines a flatter structure 

in the urban hierarchy than warranted.

This research has shown that ‘dwell time’ at 

train stations reflects the length of the alighting 

and boarding process, and we use this insight to 

estimate actual interaction through the application 

of a bimodal network projection function. The em-

pirical application to the high-speed railway (HSR) 

network within the Yangtze River Delta (YRD) re-

gion revealed that the transformed network varies 

from the original network to a large extent: (i) the 

positions of transit cities along main transport cor-

ridors in the YRD urban system are drove down 

while some arguable sub-central cities stand out; (ii) 

inter-city connectivities tend to be more hierarchi-

cal; and (iii) the flattening ratio has decreased to 

60% of the original flattening ratio. Moreover, the 

validity of our method has been evaluated through 

a comparative analysis with Weibo-users’ inter-city 

movements, verifying that the transformed network 

more parallels tangible flows of people.

We believe our paper makes two contributions 

to the literature. The first is to remind researchers 

to re-examine the validity of proxy strategies when 

measuring inter-city transport flows. With the ex-

ception of recent research on airline networks in 

the context of the world city literature, relative-

ly limited attention has been paid to the degree to 

which these infrastructure networks reflect the ac-

tual flows they undergird. In this regard, this article 

offers empirical evidence for the structural deter-

minism of using train networks per se, as these 

tend to result in flatter networks. Second, the cen-

tral contribution of this paper has been to set out 

an alternative method of approximating actual flows 

in railway networks, which permits practical appli-

cations in simulating flows of railway passengers 

in other cases.

Apart from empirical applications in other cases, 

further research issues also include: discussing other 

modes of constructing equations, discovering alter-

native perspectives to approximating actual flows in 

railway networks, investigating the biases between 

the infrastructure provision and corporeal flows in 

 - 10.1515/bog-2016-0010

Downloaded from De Gruyter Online at 09/23/2016 11:26:34AM

via Universiteit Antwerpen

---