Innovation diffusion in a centralized decision-making setting should be examined in terms of the dynamic facilities location problem (Brown, 1981). This paper is concerned with the spatial diffusion of a non-profit motivated innovation in the above setting, taking the example of radio stations in Japan before World War II. Especially the following point will be discussed here: whether radio stations spread hierarchically or not, when a certain optimal location strategy is implemented.If people are evenly distributed, efficiency and equity may be congruent. In the real world, however, the discrepancy between efficiency and equity is caused by various conditions, so that the optimal location pattern under the efficiency strategy will differ from that under the equity strategy. This paper asks which strategy could produce a hierarchical diffusion of new facilities. If the objective of the optimizing model is to minimize the total travel distance, facilities are oriented to densely populated areas. In addition, high income areas tend to be chosen under the efficiency strategy (Morrill, 1974; Morrill and Symons, 1977). Consequently, facilities would be preferentially located in larger cities with high population densities and high income levels. This discussion results in a working hypothesis that new facilities may spread hierarchically under the efficiency strategy.Experimental approach to the hierarchical diffusion of new facilities To investigate the above hypothesis theoretically, the following simulation was attempted by using Tornqvist's model. We shall consider a set of cities in a hypothetical region. They are arranged in a 9×9 mesh of 81 cells. With regard to city-size distribution, two types of distribution are considered: 1) rank-size distribution where the largest city with population of 9, 000, 000 is followed by the 2nd and lower ranking cities to fit such a rank-size curve as Pr=9, 000, 000/r; 2) hierarchical distribution where the largest city with population of 9, 000, 000 is followed by the 2nd to 9th ranking cities, the 10th to 27th ranking cities, and the 28th and lower ranking cities to fit such rank-size curves as Pr=5, 000, 000/r, Pr=3, 000, 000/r and Pr=1, 000, 000/r respectively. With regard to the spatial distribution of population, the largest city is always fixed at the center of a region, and other cities are randomly arranged in such a way that population distribution shows either significantly positive spatial autocorrelation or no spatial autocorrelation, regardless of city-size distribution.After the four hypothetical regions are constructed by the combination of city-size distribution and spatial distribution of population, facilities are located in each region under the following rules: 1) the total of 9 new facilities are located during the 9 unit periods according to the efficiency strategy; 2) only one facility is located during each unit period; 3) the first facility is always given to the largest city, and then 8 facilities are sequentially located one by one.For each region, 10 different spatial distributions of population are produced by random numbers, and the total of 40 simulation runs is repeated. Then the significant difference of means of correlation coefficients between the period of facility location and population was tested for each pair of the four regions. Results (Table 2) are summarized as follows: 1) though there are statistically significant differences of means of correlation coefficients between all the pairs of regions, hierarchical diffusion is not present in all the regions even under the efficiency strategy; 2) a hierarchical diffusion tendency often emerges in regions where populations are not spatially autocorrelated, which suggests that spatial distribution of population is more critical than city-size distribution.
Read full abstract