![Sample our Urban Studies journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110826/TOC-Subject-Sample-2021-Urban-Studies.jpg"})
Abstract
The spatial distribution of aging populations is commonly measured with either the aging population ratio or the aging population density. Used in isolation, however, these measures may fail to detect aging communities in certain types of urban or rural setting. This study uses both indices simultaneously to identify types and locations of aging communities more accurately. We investigate the spatial distribution of these communities using a standard correlation analysis and bivariate local spatial statistic analysis. Empirical analysis of geospatial data of the Aichi Prefecture in Japan suggests that using both indices allows us to capture different types of aging communities in diverse contexts (e.g. depopulated rural areas, pockets of aging communities in urban areas, and growing concentrations of aging population in the suburbs). The analysis uses data sets aggregated at different areal scales, confirming the generally stable nature of the outcome, despite some scale sensitivity.
Notes
1. 65-years-old is a widely accepted threshold for defining the aging population in developed countries. However, other figures are used in different contexts. For instance, the World Health Organization suggests the adoption of 60 years old to better account for population aging in countries in less developed countries.
2. These figures were collected by the United Nations (Citation2011) from several different sources and include estimated values. Therefore, the graph shows some unevenness in their trend, and some of its figures do not necessarily match those published by the statistical agencies of each country, but they nonetheless confirm the high aging population ratio of these nations.
3. (a) More developed regions comprise Europe, Northern America, Australia/New Zealand, and Japan. (b) Less developed regions comprise all regions of Africa, Asia (except Japan), Latin America, and the Caribbean plus Melanesia, Micronesia, and Polynesia. (c) The least developed countries, as defined by the United Nations General Assembly, included 49 countries: 34 in Africa, 9 in Asia, 5 in Oceania, and one in Latin America and the Caribbean (United Nations, Citation2011).
4. The aging population ratio for Japan published in the UN data (as shown in ) is much higher at 27.28%, and it does not match the figure of 24.1% (as of 1 October 2012) published by the Cabinet Office of Japan (2013). However, both data still confirm that Japan has an exceedingly high aging population ratio.
5. The generation born in the late 1940s who joined the labor market from the 1960s onward.
6. The outcome of choropleth mapping is known to change with the method and the threshold values used for classifying the data (Monmonier and de Blij, Citation1996). Rather than scrutinizing the difference in the nuances of the outcome, our study adopts the bin range commonly used for classifying the aging population ratio values in Japan: namely, 7–14% (aging communities), 14–21% (aged communities), 21–30% (highly aged communities), and 30% or more (very highly aged communities) (Cabinet Office of Japan, Citation2011; Ohno, Citation2008; Yano, Citation2007). An additional threshold of 50% was adopted from Ohno’s (Citation2008) work on Genkai Shuraku (critical settlements), which he defined as settlement units where the majority of the population is elderly. While Ohno (Citation2008) does not identify a convention for classifying the aging population density values, we determined them in accordance with the convention for the general population density: namely, 0–50, 50–100, 100–250, 250–500, 500–1,000, and more than 1,000 people/km2. As the bin range for the choropleth maps follows the convention, the outcome should reflect what is regularly seen in the geographical and planning contexts in Japan and would therefore provide an ideal base map with which to compare the outcomes from later analyses.
7. There are a number of ways to define the local neighbors, and these can be broadly classified into three categories: (1) those that use a fixed distance and identify all areas within that distance as the local neighbors, (2) those that specify the number of the closest local neighbors, and (3) those that focus on the geographical adjacency of the areas. As the size and shape of municipalities and districts vary widely—especially between those in the urban areas that are generally smaller and regularly shaped and those in the rural areas which tend to be larger and irregularly shaped—use of a fixed distance would result in a wide variance in the number of neighbors included in the calculation from area to area. Using geographical adjacency to define local neighbors also presents some challenge in the case of the 1 km-grid and the 500 m-grid data, as many grid cells have been excluded from the analysis due to the absence of residents or lack of data, which means that some grid cells would have very few or no adjacent cells as effective neighbors. To avoid creating an unnecessary bias, this study uses a fixed number of local neighbors for each location.
8. The grid squares of the Basic Grid Square data are determined by longitude and latitude. Therefore, these grids do not form a perfect square and their size is also slightly smaller than 1 km2. Naturally, the level of distortion increases at higher latitudes. Given that our study area lies at around 35°N and its area is relatively confined, the level of distortion is negligible.
9. Areal units with no population or no available data were excluded from the analysis. 362 units of the SA-level areas (2.8% of all SA-level areas in Aichi), 1,591 units (30.0%) of the 1 km-grids, and 10,464 units (51.1%) of the 500 m-grids had no residents. Of the areal units with residents, 322 units of the SA-level areas (2.5% of all SA-level areas in Aichi), 75 units (1.4%) of the 1 km-grids, and 238 units (1.7%) of the 500 m-grids were found to have very few residents, and the composition of their age groups was undisclosed for confidentiality reasons. Given that the units of the Grid Square data are mechanically assigned at a regular interval, the high proportion of the nonhabited units is somewhat inevitable, but has implications on the way we calculate the bivariate local Moran’s statistic, which relies on the value of local neighbors.
