This study examines interfirm alliances between Korean and Japanese companies, and characteris-tics of linkages between Seoul and Tokyo. The data were obtained from personal interviews with top and middle executives of 45 large firms based in Tokyo, which are associated with Seoul-based firms.
The linkages between these capital cities have been reinforced by various factors, especially by the alliances between big business groups (so-called giobjibdan or chaebol in Korea, and kigyoshudan or keiretsu in Japan) which are located in Tokyo and Seoul. The most important features of the alli-ances are: (1) Tokyo-and Seoul-based business groups have strong intrafirm networks with member and affiliated firms of the group consisting of top managers' committees; (2) the intrafirm networks play an intermediate role in the alliances both formally and informally; and (3) they are spatially con-centrated in the Tokyo and Seoul metropolitan areas.
Intraf irm networks of big business groups are classified into two levels: one is a network at the firm level, which forms relationships between member and affiliated firms of the group, consisting of top managers' committees at the national level, and of those between Tokyo-and Seoul-based head-quarters and overseas offices at the international level; and the other is a network of personal connec-tions between top and middle executives.
The results are as follows: (1) intrafirm networks are important mechanisms associated with inter-firm alliances; and (2) historical and social relationships are more important than economic factors suggested by transaction costs theory to explain alliance relationships between the firms studied.
We may conclude that networks of big business groups agglomerated in Tokyo and Seoul have led to the diversification of the urban functions of capital cities, and are primary factors in creating gaps between metropolitan and non-metropolitan cities in Korea and Japan.

View full abstract

The objective of this study is to use Digital Map 2500 (Spatial Data Framework, SDF) produced by the Geographical Survey Institute of Japan and the Geographical Information System (GIS) to develop a spatial analysis tool for testing spatial autocorrelations among spatial entities. In the test-ing of spatial autocorrelation, the area under study is subdivided into several unit zones, especially administrative districts with irregular spatial shapes. Problems arise in the acquisition of topological data that are necessary to calculate the weighting matrix {w_ij} which represents linkages based on the spatial configuration of irregular zones. In previous studies testing spatial autocorrelation using the GIS, the weighting matrix {w_ij} was calculated using the topological data structure of GIS such as Arc/Info. In this study, the topological data extracted from the SDF are used for calculating the weight-ing matrix {w_ij}. This study also shows that the SDF is functional and effective data fundament for spa-tial analysis with GIS technology.
This paper addresses the following points:
1. There are two types of linkage of the GIS with spatial analysis software. The first is the inter-face type which involves the interfacing of user-developed algorithms via macro languages provided by the GIS. The second type integrates the GIS with spatial analysis software, which involves file trans-fers between the GIS and other software. The latter type is used in this study, so that flexible and general spatial analysis tools can be developed and integrated with several GIS, both with and with-out topological data structures.
2. A topological vector data model provides an explicit representation of spatial relations among spa-tial entities, i. e., points, lines, or areas. In this study, the topological vector data for spatial autocorrela-tion testing and the graphic data for presenting and measuring administrative districts are drawn up using the data structure of the SDF, such as map grids and tree structures of directories and data sources. The weighting matrix {w_ij} is constructed using the topological vector data.
3. Some aspects of spatial autocorrelation are observed on a choropleth of population density in Ma-tsudo City, Chiba Prefecture. This paper aims to use Moren's I statistic with binary and generalized weighting matrices statistically to test the observed spatial autocorrelation among spatial entities.
This study reveals that in testing spatial autocorrelation in irregular zones, the use of a binary weighting matrix or a generalized one depends on the size, shape, and spatial arrangement of observa-tional zones.

View full abstract