Thanks for pointing this out, Ed.

I downloaded the redistricting dataset (PL94) and ran some comparisons to 2010 SF1 numbers. The first thing that jumps out to me is that housing and population data are apparently adjusted independently of one another. Here’s a sample block group from the city of Chelsea, MA:

In this example, the significant reduction in occupied housing units (households) is not matched by a reduction in population. The changes to housing occupancy rate and average household size are significant.

Looking at all 2,438 block groups in our 97-city MPO region:

·         211 block groups (8.7%) have occupied housing unit changes of at least +/- 100.

·         Only 3 block groups have population changes of at least +/- 100.

·         The shifts in households serve to fill some block groups at the expense of others. The number of block groups with 100% housing unit occupancy increases from 14 to 1,319. The number of block groups with less than 75% occupancy increases from 21 to 166.

·         222 block groups have a change in average household size of more than +/- 0.5.

The changes at the TAZ level are more significant given that our TAZs are loosely based on block boundaries (there are a number of split blocks) and the block data is more volatile. Over 300 out of 2,727 TAZs contain fewer than 10 blocks.

Here are two examples from Boston and Chelsea:

In the Boston example, the increase in household population results in an average household size of 22.5!

Looking at the 1,901 TAZs within our MPO region:

·         116 TAZs (10.5%) have occupied housing unit changes of at least +/- 100.

·         144 TAZs have population changes of at least +/- 100.

·         The number of TAZs with 100% housing unit occupancy increases from 64 to 815. The number of block groups with less than 75% occupancy increases from 41 to 102.

·         222 TAZs have a change in average household size of more than +/- 0.5.

Another point of concern is with minority population data. Here are a couple of examples from Boston:

For our MPO region:

·         The minority population percentage in both datasets is 28.2%. TAZs with minority population percentages higher than this are flagged as minority TAZs.

·         116 of 1,901 TAZs (including both seen in the example) would have a change in minority status: 46 from minority to non-minority and 70 from non-minority to minority.

·         287 TAZs have a change in minority population percentage of at least +/- 5%.

The changes are not as significant at the block group level.

We will be redrawing TAZ boundaries following the 2020 Census, and the starting point for TAZs will likely be block groups. Given the volatility of the block-level data after application of differential privacy, we’ll require data from some other source where we need to divide block groups into multiple TAZs. I think the block group population numbers we’ll get post-differential privacy will be workable. But the household numbers are a problem. We can’t be confident in our trip generation results if we don’t know how many households are in each TAZ.

Since the 2006-2010 ACS was first released in 2011, we have used population and household counts from the 2010 Census as the controls for normalizing ACS statistics. We envision doing the same with 2020 Census and 2016-2020 ACS data when they are available. But, again, we need accurate population and household controls.

Another area where the minority population numbers are a concern is in the Environmental Justice (EJ) component of project scoring for our Transportation Improvement Program. We create a half mile buffer around project locations and summarize block-level population data for the buffers in order to score each project. The uncertainty in the minority population statistics, and in other EJ population statistics drawn from the ACS, could raise questions as to whether our universe of projects is distributed equitably.

I look forward to hearing from others who have an opportunity to look at this data.

Paul Reim

Paul Reim  |  Chief Planner

CENTRAL TRANSPORTATION PLANNING STAFF

857.702.3688  | preim@ctps.org

www.ctps.org/bostonmpo

This may be of interest to some:
-----------------------------------------------------------------------
Yesterday the Census Bureau released the 2010 Demonstration Data Products to help data users understand how differential privacy may or may not impact data products they are used to receiving. The products include the 2010 Demonstration Public Law 94-171 (P.L. 94-171) Redistricting Data Summary File and the Demographic and Housing Demonstration File, which is similar to Summary File 1.
 
Documentation and resources related to these products are available on the Census Bureau’s website:
https://www.census.gov/programs-surveys/decennial-census/2020-census/planning-management/2020-census-data-products/2010-demonstration-data-products.html
 
The Census Bureau is working on an FAQ document to provide additional information about these data. You can also direct questions to the Census Bureau via email:
dcmd.2010.demonstration.data.products@census.gov.

 

-- 

Ed Christopher

Transportation Planning Consultant

708-269-5237