Cultural Versus Objective Distances: The DBS-EM Approach
详细信息 查看全文
- 作者：Massimo Mucciardi ; Gustavo De Santis
- 关键词：Clusters ; Cultural distance ; Euclidean distance ; EM algorithm
- 刊名：Social Indicators Research
- 出版年：2017
- 出版时间：February 2017
- 年：2017
- 卷：130
- 期：3
- 页码：867-882
- 全文大小：
- 刊物类别：Humanities, Social Sciences and Law
- 刊物主题：Sociology, general; Quality of Life Research; Microeconomics; Public Health; Human Geography; Quality of Life Research;
- 出版者：Springer Netherlands
- ISSN：1573-0921
- 卷排序：130
文摘
In this paper we extend and improve a recently proposed method for the measurement of the cultural distance between strata. In the original version, strata meant countries: respondents from different countries were clustered on the basis of their answers to a set of questions and their resulting distribution among the K clusters thus formed was used to calculate the “position” of each country as a point in a K-dimensional space. The proposed improvements are as follows. First, the notion of “strata” is enlarged: not only geographic units, but also gender, age, education, religious attitudes and rural/urban residence. Second, clustering is now based on EM, or Expectation Maximization, which automatically determines the optimal number of clusters, thus overcoming one of the major limitations of the previous version of the method. Third, since this optimal number of clusters turns out to be small, a principal component analysis is used to capture most of the variability and draw a very telling, two-dimensional representation of how (culturally) distant strata are from one another. Fourth, since two types of distances between strata can be computed, a cultural and an “objective” one (e.g., kilometers between regions or years between age groups), their correlation can be calculated. On our Istat (Indagine multiscopo, Aspetti della vita quotidiana, Rome, 2013) data, expectations are confirmed: the farther strata are, the greater their cultural distance. The same happens for the (rural/urban) type of commune of residence. Religion, instead, is rarely, and gender is never, associated to any measurable cultural difference.