Dear CTPP Listserv members,
In June 2021 the CTPP Oversight Board submitted a request to the Census Bureau for the next CTPP package based on 2017-2021 American Community Survey (ACS) survey data. Census Bureau staff, including the Disclosure Review Board (DRB), ACS staff, and Journey-to-Work subject matter experts responded in January 2022 with a memo outlining disclosure concerns and delineating proposed rules for this CTPP special tabulation. The impact of these rules will affect the CTPP package and the Table Subcommittee of the AASHTO Oversight Board is working hard to try and preserve a quality product.
The Census Bureau's concerns, and proposed rules include:
1. Reduce the overall size of the package
2. Standardization of Universes, Variables, and Categories wherever possible to match standard Census ACS products
3. Perturbation techniques must be employed
4. Any flow must be subject to an unweighted minimum of 3 persons. This means 3 people had to respond to the ACS that they live in one geography and work in the other.
5. Non-residence geographies must have 50 unweighted cases to avoid suppression. This means that part 2 (workplace) data at small geography will be in jeopardy of being suppressed
6. Mean and Aggregate tables must have a minimum of 3 to be tabulated
As a result of the memo and its implications, the Table Subcommittee is developing a response to the Census Bureau to determine a path forward. We have been working with industry experts on that response. To date, our strategy involves:
1. Eliminating block groups in parts 2 and 3 and much of part 1. (A small list of part 1 tables at the block group level is TBD)
2. Universes, variables, and categories/cohorts will be standardized to the extent possible
3. Developing a sound justification for keeping CTPP-specific universes, variables, or categories
4. Expecting perturbation decisions to be data driven and originate at the CB
5. Documenting the loss of data that occurs when the rule of 3 for flows is implemented
6. Mean and Aggregate tables with less than 3 are not very meaningful and are a disclosure risk
Where we have significant issues are the 50 unweighted cases at the workplace that renders much of the package useless for parts 2 and 3. We are exploring options such as not having the requirement for a small set of tables or employing either perturbation or synthesizing techniques to supplement what the DRB will approve.
The work is ongoing to respond to the Census Bureau. If you have concerns about what is going, please reach out to Arash, Penelope or me.
Clara, Arash, and Penelope
The Census Bureau released the five-year, 2016-2020 American Community Survey tables today. The r-package TIDYCENSUS worked without a hitch this morning.
In case folks are interested in using TIDYCENSUS to examine the three sets of non-overlapping periods: 2006-2010, 2011-2015 and 2016-2020, I’ve updated my R script for all states, counties, places and the US. This is just for the means of transportation to work, plus a few key demographic variables.
https://gist.github.com/chuckpurvis <https://gist.github.com/chuckpurvis>
Important to note is that the 2016-2020 ACS is based on 2020 Census geography. The older datasets are based on 2010 Census geography. This isn’t a big deal if your geographic areas haven’t changed, 2010 to 2020, but I’d be careful when comparing older (pre-2020) to newest (2016-2020) data, especially for smaller geographies such as tracts, block groups and places.
My next step would be to do some count checks of geographies included in Census 2020 PL 94-171 vs the 2016-2020 ACS. Check to make sure that the geographies (block groups, tracts, places, counties) match between the Decennial 2020 and ACS 2016-2020.
And to reiterate, there is no single year ACS data for 2020. Those tables are using the “experimental weights” and are only published at the nation and state level.
Happy St Patrick’s Day to all!
Chuck Purvis,
Hayward, California
Dear Colleagues,
I am beyond excited to announce that the Census Data for Transportation Planning Conference scheduled for June 7 - 9, 2022 in Reno, NV has opened for registration!
This conference promises to be heavily technical, interactive and peer to peer, featuring many workshops and sessions on hot topics (equity, workplace geo-coding, tools for data access, geography wrangling, and more), the latest research, and policy implications for census and census derived data.
Of course, we will include a healthy dose of social activity as well!
Early bird pricing is available until April, 29.
See the conference agenda and register today at: https://cvent.me/l7X2ma
Please don't hesitate to reach out to me with any questions. I can't wait to see you all.
Penelope Weinberger
She/They
CTPP Program Manager
AASHTO
Ctpp.transportation.org
Hello:
I am still a little confused why using overlapping ACS datasets is not good practice. Can someone explain it to me?
Thank you.
[cid:image001.png@01D83956.276DFD40]
David Heller, PP/AICP
Program Manager - Systems Performance and Subregional Programs
South Jersey Transportation Planning Organization
782 S Brewster Road, Unit B6
Vineland, New Jersey 08361
(856) 794-1941 | www.sjtpo.org
From: Benjamin Gruswitz <bgruswitz(a)dvrpc.org>
Sent: Tuesday, March 15, 2022 4:59 PM
To: The Census Transportation Products Program Community of Practice/Users discussion and news list <ctpp(a)listserv.transportation.org>
Subject: [CTPP News] Re: CTPP commuter flows in strong MCD states (Vermont test case)
Yes, Chuck, this is a data advantage for strong MCD regions--the workplace allocation is complete at this subcounty level that covers all areas of each county (as opposed to place, which doesn't have county-wide coverage). And our TAZs nest within our municipal boundaries, so fitting their flows to the MCD total is a good way to go for adjustments. The only issue in our region is that Philadelphia is both a county and MCD, so we don't get a subcounty control for our TAZ workplace fitting within our high pop/high employment urban center the way we do for our smallest boroughs and townships (we have one borough with a population of 10 and employment of 25).
Thanks for always pointing us to good resources and encouraging our experienced and burgeoning R users to explore CTPP data with that toolset!
Ben
Working from Home | 301.655.3170
Ben Gruswitz, AICP | Manager, Socioeconomic & Land Use Analytics
(Pronouns: he/him)
Delaware Valley Regional Planning Commission
190 N Independence Mall West, 8th Floor
Philadelphia, PA 19106-1520<https://www.google.com/maps/place/Delaware+Valley+Regional+Planning+Commiss…>
215.238.2882 | www.dvrpc.org<https://www.dvrpc.org/>
Subscribe<https://app.e2ma.net/app2/audience/signup/1808352/1403728/> | Facebook<http://www.facebook.com/DVRPC> | Twitter<http://www.twitter.com/DVRPC> | Instagram<http://www.instagram.com/DVRPC> | LinkedIn<http://www.linkedin.com/company/delaware-valley-regional-planning-commission> | YouTube<https://www.youtube.com/channel/UCEU8UI5_iGkVypHP93b5jLA>
[DVRPC]<https://www.dvrpc.org/>
On Tue, Mar 15, 2022 at 4:05 PM Charles Purvis <clpurvis(a)att.net<mailto:clpurvis@att.net>> wrote:
Being a west coaster, I rarely dabble in MCDs - Minor Civil Divisions, or NECTAs (New England City and Town Areas). I thought this deserves some exploration.
I created a new version of my R-package CTPPr script that pulls in intra-state Vermont total commuters: county-to-county, tract-to-tract, and MCD-to-MCD. I’ve shared my Vermont code on my GIST GITHUB. I screwed up yesterday, and had the other scripts in “secret” mode. Oops, sorry. I’ve made the correction.
https://gist.github.com/chuckpurvis
There are 14 counties in Vermont, 184 census tracts, and 255 MCDs (towns) in Vermont. The 255 MCDs are “wall-to-wall” coverage of the entire state (i.e., no lingering unincorporated “balance of county” areas.) I was surprised that there are fewer census tracts than MCDs in Vermont, but I had some notion that the MCD-to-MCD flow data could be quite valuable (in certain states!)
According to the CTPPr documentation, probably the official CTPP documentation, as well, there are MCD-to-MCD commuter flows for the twelve “strong MCD” states.
From some random US Treasury document:
"Since the government services provided by MCDs differ greatly by state, the Census Bureau refers to
twelve states with MCDs that generally provide a wide range of general government services as “strong-
MCD” states. In these states, MCDs are generally are treated as municipalities according to state statutes
and codes. In eight other states, MCDs typically play less of a governmental role and provide more limited
government services, even though they are still active governments (“weak-MCD” states). The twelve
strong-MCD states are Connecticut, Maine, Massachusetts, Michigan, Minnesota, New Hampshire, New
Jersey, New York, Pennsylvania, Rhode Island, Vermont, and Wisconsin. The eight weak-MCD states are
Illinois, Indiana, Kansas, Missouri, Nebraska, North Dakota, Ohio, and South Dakota."
Here are the highlights of this Vermont test case:
Total Workers, Intra-State, Vermont:
County-to-County = 298,422 total workers
MCD-to-MCD = 299,415 total workers
tract-to-tract = 214,970 total workers.
The county-to-county and MCD-to-MCD totals for Vermont should be very, very close, since they both have the “standard allocation procedures” that the Census Bureau uses to impute missing workplace to the county and place level. I’m pretty sure the difference between county-to-county and MCD-to-MCD is rounding issues? Can never tell.
The tract-to-tract file does not have the standard allocation procedures applied: it’s the raw data, rounded of course. If I were Vermont, I’d stick with MCD-to-MCD flows as the best bet for controls. Adjust/factor any of the TAZ-to-TAZ flow data to MCD-to-MCD.
Happy Ides of March,
Chuck Purvis
Hayward, California
_______________________________________________
CTPP mailing list -- ctpp(a)listserv.transportation.org<mailto:ctpp@listserv.transportation.org>
To unsubscribe send an email to ctpp-leave(a)listserv.transportation.org<mailto:ctpp-leave@listserv.transportation.org>