Comparing Populations across Changing Census Geographies
By Amy Kracker Selzer, College of William and Mary
The ability to link demographic data to geographic units provides an opportunity to map the changing spatial dimensions of social dynamics. Census data generally provide one of the most complete and widely available sources of corresponding spatial and demographic data. However, a frequent issue with this type of analysis is that the geographic boundaries of the spatial units at which census data is available change from one round of data collection to the next. This can make it difficult to compare the spatial distribution of demographic characteristics over time. This post discusses one potential way to address this problem based on my work comparing South African census data from 1996 and 2001.
StatsSA – the South African agency that conducts the census – made the 1996 census data available at a number of geographic scales ranging from the national level down to the enumeration area (EA) - the smallest geographic aggregation of data made publicly available. The size of the enumeration area can be roughly equated to the block level in the US census. For the 2001 census, the smallest level of aggregation made publicly available was the subplace level, a geographic unit that generally corresponds to neighborhood boundaries. Additionally, the boundaries of the subplaces in the 1996 and 2001 censuses did not always correspond to one another. If the 1996 EAs all fell neatly within boundaries of 2001 subplaces, they could simply be aggregated to the subplace level. While this is generally the case, at times, the 1996 EAs crossed the boundaries of one or more 2001 subplaces and sometimes covered a larger geographic area than subplaces. This lack of geographic consistency makes it difficult to compare spatial patterns between these two years.
In order to address this issue, a set of common geographic boundaries was established for both years of data. Because the smallest available unit for 2001 (the subplace) is generally larger in area than the smallest geographic unit for 1996 (the EA), the 2001 subplace was used as the baseline. Because EAs are the smaller of the two units, they can more accurately be aggregated up to the 2001 subplace level. The issue then becomes aggregating the 1996 EA level tabular data into the boundaries of the 2001 subplaces. ArcGIS offers a data partition tool that reassigns data to alternate spatial units based on the proportion of overlap between polygons, making it possible to distribute the 1996 EA data into the 2001 subplace boundaries. The data partition tool will overlay the 2001 subplace boundaries on top of the 1996 EAs and redistribute the 1996 data within the boundaries of the 2001 subplaces based on the proportion of area that overlaps between the two layers. Essentially, if 50% of the area of EA 1 overlaps with subplace A, 50% of the population of the EA will be assigned to subplace A while the remainder is similarly divided among the other overlapping subplace(s). The partition tool relies on the assumption that the population is evenly distributed throughout the EA, which is unlikely to be true but the generally small size of the EAs helps to minimize the associated distortion.
/methods/upload_files?prefix=20170626200049/components/com_communityanswers/assets/uploader/plupload.flash.swf/components/com_communityanswers/assets/uploader/plupload.silverlight.xappdf,doc,docx,xls,xlsx,odt,rtf,txt,odf,csv,dbf,odp,uop,pptx,ppt,pps,ppsx,eps,gif,jpg,jpeg,jpe,png,ps,svg,tif,tiff,bmp,cdfClick here/methods/verify_captchaContribute Your ResponseCancelYesNoError!ConfirmSuccess!Click here/methods/subscribe_qn/37-comparing-populations-across-census-geographies/methods/unsubscribe_qn/37-comparing-populations-across-census-geographiesYou have successfully subscribed to this topic. You should now receive emails when new respones are posted.You have successfully unsubscribed to this topic.Subscribe To TopicUnsubscribe From Topic/methods/get_answerReport ThisSubmitInformation!