2020 Census Data Products: Data Needs and Privacy Considerations Proceedings of a Workshop (2020) / Chapter Skim
Currently Skimming:
3 Geospatial Analyses of Social and Demographic Conditions
3 Geospatial Analyses of Social and Demographic Conditions
Pages 29-44
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 29... ...
as a geographic review, a set of exploratory spatial data investigations comparing the 2010 Demonstration Data Products (DDP) with the original 2010 Census publications.1 Their review included sweeping looks across various levels of geographic aggregation and 1 In closing their talk, Van Riper and Spielman said that all code and data related to this presentation would be posted at http://github.com/geoss/CNSTAT_DIFF_PRIVACY.
|
From page 30... ...
3.1.1 Residential Segregation Van Riper worked through a series of metrics comparing the 2010 DDP and the original 2010 Census publications regarding their implications for assessing residential segregation, which he said that researchers commonly study by comparing census tract-level population characteristics to the population distributions that prevail in the core-based (or, in older terminology, metropolitan) statistical areas.
|
From page 31... ...
The points Van Riper made in his presentation did not concern the meaning or policy interpretation of these segregation measures, but rather sought only to show whether the measures -- points on a scatterplot corresponding to corebased statistical areas, the computed score from 2010 Census Summary File 1 (SF1) on the x-axis and the scores from the 2010 DDP on the y -- generally fell on the 45-degree equality line.
|
From page 32... ...
Persons aged 65 years and over (d) Black householders Figure 3.1 Deviations of 10 percent or greater between 2010 Census published data and 2010 Demonstration Data Products by selected demographic groups and geographic levels.
|
From page 33... ...
3.1.3 Spatial Differences Spielman resumed the presentation with a series of histograms computed over all census tracts that show the distribution in "gains" and "losses" in population between the 2010 Census SF1 and the 2010 DDP, confirming impressions already reached in the earlier presentations: for total population, white population, and Hispanic population, distributions concentrated tightly around 0. As a geographer, though, he said he was naturally concerned with the possibilities for spatial patterning that might underlie these balanced histograms.
|
From page 34... ...
SOURCE: David Van Riper and Seth Spielman workshop presentation. SF1 data and the 2010 DDP compares to true spatial randomness, Spielman and Van Riper ran simulation studies: randomly assigning the calculated differences by tract to all of the pertinent tracts, computing a new Moran's I score, then repeating the random assignment 999 times.
|
From page 35... ...
What is clear from all of the figures is that the observed Moran's I scores are distinctly different from the spatially random simulations and so are "not consistent with spatial randomness." Again, exactly what the patterns are and how they might affect particular use cases of the census data awaits further investigation, but Spielman suggests that "certain places are gaining and certain places are losing" in a way that is "systematic within space" at the tract level, and that this may be problematic for local use cases of census data. 3.1.4 Closing Thoughts Van Riper closed the presentation by itemizing some thoughts on the process and products ahead that he had developed with Spielman in preparing for the workshop.
|
From page 36... ...
His personal opinion and preference is to assign a lot more value to the accuracy side of the balance than the privacy side, but he said that the tradeoffs need to be fully elaborated and considered, and that the final decision-making process should be "participatory," with data users having "a seat at the table" to give feedback to the Census Bureau. He warned that if the 2020 Census data end up emanating from an opaque, complicated, closed process that is not well understood by users (and, moreover, if those data run counter to local knowledge)
|
From page 37... ...
C • Q) O> C O· ro -25 • • • 100 1.000 10.000 100.000 1.000.000 SF1 population Figure 3.3 Change in real per capita allocation of state sales and use tax revenue in Tennessee between 2010 Summary File 1 and 2010 Demonstration Data Products by 2010 population, places in Tennessee.
|
From page 38... ...
census data. Nagle briefly displayed an analysis of change in total population, from 2010 Summary File 1 to the 2010 DDP for all incorporated places in the nation, a subset of the overall place geographic level that excludes the census-designated places that are defined for statistical purposes.
|
From page 39... ...
Pivoting to his originally planned topic, Nagle turned to another off-spine geographic level in school districts. In particular, he focused on school districts in his home county of Knox County, Tennessee.
|
From page 40... ...
Total population by place may be one statistic with sufficiently many statutory uses that it should be afforded additional accuracy protection, somehow. Off the cuff, Nagle suggested another possibility: constructing a parallel data product that does nothing but generate total population counts for the offspine geographies with a "government use," including incorporated places and American Indian and Alaska Native reservation lands.
|
From page 41... ...
At the least, there is less variation in the size of census tracts (due to the rules that govern their demarcation) than in other geographic levels, which partially suggests that Moran's I is picking up on a different kind of systematic effect.3 Hansen followed up by noting that the scatterplots of core-based statistical areas suggested some distinct outliers, and asked whether Van Riper and Spielman had investigated those outliers for any common features.
|
From page 42... ...
entered circulation, Spielman worried about the Census Bureau and the data user community repeating past mistakes. In the absence of good education on and orientation to the ACS data, people somewhat "ignored the changes in the data and just kind of did what they have always done." Nagle replied that he had been thinking of the temporal question rather than the spatial question -- that is, there are important time series questions in characterizing uncertainty in the long-term, across-decades shifts in allocation that localities may need to contend with, where "they could win the Census lottery one decade and lose the Census lottery the next decade." Building on Spielman's point, Hansen asked about the notion that had come up about presenting an uncertainty metric with census tabulations: "this, plus or minus something." Would that actually help or, as experienced with the ACS, would people "just interpret the middle point and away we go" again?
|
From page 43... ...
Nagle also reiterated that the broader data user community needs to take seriously the "challenge" of articulating their use cases to the Census Bureau. Van Riper said that he wanted to "push back" a little on Nagle's point, because "the American Indian and Alaska Native population" and "the total population residing in American Indian and Alaska Native areas" are not equivalent concepts -- but both are vitally important, and possibly both important to funding and other formulas.
|
From page 44... ...
The county uses the data on households with 7 or more persons to identify areas of overcrowding and to target housing affordability programs. Han said that, looking at just those two variables, the 2010 DDP data for census tracts are "not usable" -- all of the census tracts in the county "get either doubled or [halved]
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.