Elena Kotyrlo (p)

Makoto Tsukai

Observing the dramatic increase in the flow of migrants since the 2000s, one can hypothesise that emergent access to the Internet and mobile communications has contributed to it. Indeed, modern information and communication technologies (ICTs) help potential migrants overcome the strongest barrier to migration – incomplete information. Studies have found that ICTs assist migration decision-making and support people with information when crossing borders. An alternative view suggests that ICTs enhance capacities to maintain working and family life across long distances and act as a substitute for movement. This is due to the role of ICTs in the growth of independence from a fixed place of work, lessening information, emotional and social losses of migration, strengthening the cultural identity of migrants, and facilitating remittances to non-migrated family members. Empirical evidence for 191 countries for the period 1995–2015 supports this latter suggestion of a negative link between ICT development and migration flows. The model is specified as a dynamic panel data model that allows estimation of the medium- and long-term effects of ICTs on migration. A system generalised method of moments estimator and bias-correction for the FE estimator are employed to account for potential endogeneity of ICT development to countries’ socio-economic conditions.

Makoto Tsukai (p), Satoko Ohno , Yuta Tsukano

Elena Kotyrlo

In Japan, the expanded suburban area has often lower density, suffering from inefficient infrastructure performance with higher maintenance cost. In order to induce proper land use, the necessity to understand the structure of metropolitan area is increased.
Understanding the structure of metropolitan area with its dynamics requires the comprehensive analysis not only in built environment, but also demographic, industrial or commercial characteristics.
This study purposes to clarify the applicability of topic model to geographical characteristics, named geographic topic model. Using the proposed model, Kitakyuusyuu and Fukuoka metropolitan areas are analyzed to compare the geographical characteristics of both cities and its change between two cross sections.
Topic model is used for a topic estimation problem from the documents with huge number of words. The advantage of output by topic model over by the conventional factor analysis is positive estimation of parameters, while negative estimates often appears in conventional approaches. Positive estimates of parameters make much easier interpretation for topic, and mesh to topic relationship is also clear since spatial aggregation of topic is possible.
The input information to topic model is called Bag-of-Words (BOW), which counts the number of vocabulary for each document. This study proposes to discretize the attribute distribution by using multiple dummy variables, in order to make BOW dataset.
The BOW in this study consists of 34 attributes collected form national census, economic census and office and company statistics. The number of meshes in the target area is totally 2113 for a cross section, by dropping unused area. In order to common topics in two cross sections, the datasets in 2000 and 2010 are stacked, so then totally 4226 samples are inputted. The comparison of spatial distribution of topics showed that in Fukuoka, CBD area is expanded, and the suburban areas are dynamically developed. On the other hand in Kitakyusyu, CBD area becomes small and low-density suburban area is increased. In terms of public transport access, Fukuoka can successfully attract the agglomeration around the stations, while Kitakyuusyuu failed to attract the agglomeration, but makes low density land use around the station.
Therefore, the application of geographical topic model, the performance of proposed model was fairly good.