New Preprint: Postal Code Geocoding Error and Exposure Misclassification in Canada: A National Validation of PCCF+ against national address registers

Abstract

Population health research in Canada frequently assigns environmental and area-based exposures to individuals using residential postal codes. Of the several tools available for this process, the most common is the Postal Code Conversion File Plus (PCCF+), which uses population-weighted random allocation for postal codes with multiple potential geographic locations. However, the positional error of this method, and its consequences for assignment of environmental exposures or area-based deprivation, have not been comprehensively measured against a national address register. This study geocoded 12.4 million residential addresses from the National Address Register (NAR) with PCCF+ Version 8D and compared each assigned PCCF+ representative point with a reference location given by the NAR building point or, where that was absent, a matched Open Database of Addresses point or the NAR blockface point. Positional error, dissemination area (DA) agreement, growing-season greenness from the MODIS Enhanced Vegetation Index (EVI), and quintiles of the Material and Social Deprivation Index were compared across postal delivery modes and categories of rurality. The median error was 65 m for urban addresses served by letter carriers and 3.1 km for rural postal codes, and DA agreement was 92% and 30%, respectively. Greenness assigned from PCCF+ was nearly unbiased for urban houses (mean absolute difference 0.013) but was systematically higher than at the true location for rural postal codes (bias +0.036; r = 0.51). The material deprivation quintile was misclassified for 5.0% of urban addresses and 46.2% of rural postal-code addresses, with errors distributed symmetrically. Misclassification depended far more on delivery mode than on rurality: rural postal codes were misclassified at similar rates (about 44% to 49%) in metropolitan areas, census agglomerations and most metropolitan influenced zones, and less often in areas with no metropolitan influence. Error measures and two correction methods were developed from the register: a stratum correction based on the probability of each assigned deprivation quintile given the true quintile, by province, Statistical Area Classification (SAC) and delivery mode, and, for record-level data, a postal-code correction based on the distribution of true quintiles within each postal code. Delivery mode explained about twice as much of the error as province and SAC combined, and error measures stratified only by province and SAC misstated misclassification by 20 to 40 percentage points for populations with atypical delivery-mode mixes. In simulated cohorts with a true rate ratio of 2.0, naive estimates were attenuated by 5% nationally and by up to 31% in rural Saskatchewan, whereas the postal-code correction reduced bias to under 3% with 92–100% interval coverage across all cohort types. Exposure misclassification from postal code geocoding is concentrated among rural addresses and attenuated deprivation gradients in simulated cohorts, and it can be corrected using national address data.

Preprint Citation

Peters P.A. & T. McGaughey. Postal Code Geocoding Error and Exposure Misclassification in Canada: A National Validation of PCCF+ against national address registers. Preprints [Preprint]; 2026. Available from https://doi.org/10.20944/preprints202610.1204.v1

GitGub Citation

Peters PA, McGaughey T. Postal Code Geocoding Error and Exposure Misclassification in Canada: code and derived tables. [Internet] GitHub; 2026. https://doi.org/10.5281/zenodo.23564814