Measuring Ethnic Clustering and Exposure with the Q Statistic: An Exploratory Analysis of Irish, Germans, and Yankees in 1880 Newark

Abstract

The study of population patterns has animated a large body of urban social research over the years. An important part of this literature is concerned with the identification and measurement of segregation patterns. Recently, emphatic calls have been made to develop measures that are better able to capture the geography of population patterns. The objective of this article is to demonstrate the application of the Q statistic, developed for the analysis of spatial association of qualitative variables, to the detection of ethnic clustering and exposure patterns. The application is to historical data from 1880 Newark in the United States, with individuals classified by ethnicity and geocoded by place of residence. Three ethnic groups, termed Irish, Germans, and Yankees, are considered. Exploratory analysis with the Q statistic identifies significant differences in the tendency of individuals and building occupancy to cluster by ethnicity. In particular, there is evidence of a strong affinity within ethnic clusters and some intermingling between Yankee and Irish residents. In contrast, the exposure of Germans to individuals of other groups is found to be more limited.

A través del tiempo, el estudio de los patrones de población ha construido un abultado cuerpo de investigación social urbana. Buena parte de esta literatura se ocupa de la identificación y medida de los patrones de segregación. Recientemente, son notables los llamados enfáticos que propenden por medidas más efectivas que capten la geografía de los patrones de población. El propósito de este artículo es demostrar la aplicación de la estadística Q, desarrollada para analizar la asociación espacial de variables cualitativas, con la cual detectar la agrupación étnica y patrones de exposición. La aplicación se hace a datos históricos de la Newark de 1880 en los Estados Unidos, clasificando los individuos por etnicidad y geocodificados por lugar de residencia. Se tomaron en cuenta tres grupos étnicos, denominados irlandeses, alemanes yanquis. El análisis exploratorio con la estadística $Q$ identifica diferencias significativas en la tendencia de los individuos y ocupación de edificaciones a agruparse por etnicidad. En particular, existe evidencia de una fuerte afinidad al interior de los agrupamientos étnicos, lo mismo que de cierta mezcla de residentes yanquis e irlandeses. Por contraste, se ha encontrado que la exposición de los alemanes ante individuos de otros grupos es más limitada.

Key Words:

• clustering
• exposure
• Q statistic
• segregation
• spatial association

关键词:

• 集群
• 曝光
• Q统计
• 隔离
• 空间关联

Palabras clave:

• agrupamiento
• exposición
• estadística Q
• segregación
• asociación espacial

Acknowledgments

The authors express their appreciation to three anonymous reviewers for their valuable comments on previous versions of this article. The work of Manuel Ruiz was partially supported by MCI (Ministerio de Ciencia e Innovación) grant MTM2009–07373, Fundación Séneca of Región de Murcia and by grant 861–2009–2010 from the Social Sciences and Humanities Research Council of Canada. Fernando López received financial support from the project ECO-2009–10534-ECON of Ministerio de Ciencia y Tecnología and from the project 11897/PHCS/09 of Fundación Séneca de la Región de Murcia. John Logan was supported by research grants from the National Science Foundation (0647584) and the National Institutes of Health (1R01HD049493–01A2). The usual disclaimer applies: The authors alone are responsible for the contents of this article.

Notes

1. MATLAB code to calculate and test Q is available as supplementary material that accompanies Ruiz, López, and Páez (2010). The code can also be downloaded at http://www.science.mcmaster.ca/geo/faculty/paez/publications.html#journals

2. As noted earlier, decreasing overlap degree reduces the risk of false positives but also the power of the statistic. The application is therefore very conservative. For thoroughness, we calculated the statistic using r = 2, 3, and 4. The statistic is highly significant and rejects the null hypothesis of randomness in every case. As well, the relative frequency of symbols, and their significance, does not display undue variations. Detailed results for this sensitivity analysis are available from the authors.

3. We also calculated the statistic using r = 2. The results hold.

4. We also calculated the statistic using r = 2 and 3. The results hold.