Mapping the spatial pattern of the uncertain data in urban areas: The disadvantaged predict global nonresponse rate in the National Household Survey

• Published in: The Canadian geographer 2020-03, Vol.64 (1), p.79-104
• Main Authors: Bell, Scott, Sidloski, Michaela, Shah, Tayyab Ikram
• Format: Article
• Language: English
• Subjects: Bias; Cartography; Configurations; Cross-sectional studies; cross‐sectional survey; Educational attainment; Employment; enquête ponctuelle; fiabilité de l'enquête; Households; Housing; Indigenous peoples; Inequality; Mapping; National Census response rates; NGOs; Nongovernmental organizations; Polls & surveys; privation socio‐urbaine; Regression analysis; Response bias; Spatial analysis; spatial statistics; statistiques spatiales; survey reliability; taux de réponse au recensement national; Uncertainty; Unemployment; Urban areas; urban social deprivation; Variables
• ISSN: 0008-3658; 1541-0064
• DOI: 10.1111/cag.12556

High levels of survey nonresponse potentially produce unreliable data due to the often indeterminable possibility of such data being subject to nonresponse bias. In this paper, spatial patterns of global nonresponse rate are analyzed in order to identify whether systemic bias exists across urban spaces with regard to survey nonresponse. Forward stepwise regression is used in combination with spatial regression analysis to build models enabling the prediction of global nonresponse rates in the voluntary 2011 National Household Survey based on explanatory employment, housing, income, and other variables within 11 Canadian cities. The modelling process underscores the inequity of global nonresponse rates; places with high unemployment, high rates of rental properties, a higher proportion of Aboriginal residents, and lower educational attainment have lower compliance with the voluntary survey. Such a pattern has the potential to dramatically influence the ability of government, non‐governmental organizations, and other service providers to address the needs of residents of such urban areas. Key Messages Global nonresponse rate in the 2011 National Household Survey varies greatly between and within urban spaces. Linear regressions are used to produce models predicting global nonresponse rates in 11 Canadian cities, based on socioeconomic and demographic variables. Areas in which variables representing social disadvantage are more prevalent show relatively higher global nonresponse rates than areas in which these variables are less prevalent. Cartographier la configuration spatiale des données incertaines en milieu urbain: le taux de non‐réponses des défavorisés dans l'Enquête nationale auprès des ménages Des niveaux élevés de non‐réponses à l'enquête nationale auprès des ménages peuvent produire des données non fiables en raison du possible biais des non‐réponses. À l'intérieur de ce texte, les configurations spatiales du taux global de non‐réponses sont analysées dans le but de déterminer s'il existe un biais systémique dans les espaces urbains en ce qui a trait aux non‐réponses à l'enquête. Ainsi, une régression progressive est utilisée en combinaison avec une analyse de régression spatiale pour établir des modèles permettant la prévision de taux globaux de non‐réponses à l'Enquête nationale à participation volontaire auprès des ménages de 2011. Ces traitements se concentrent sur les variables explicatives de l'emploi, du logement, du revenu et sur d'autres v