For the census data nerds, Alabama's challenge to differential privacy is 👀
Differential privacy is essentially the fuzzing of data in order to prevent someone from being able to infer individuals' census responses from the released aggregated data. An analogy is perhaps figuring out the individual numbers in a Sudoku puzzle
Why this matters to the census is the Census Bureau stated it interprets the confidentiality provisions of individuals' responses to the census found in Title 13 of the U.S. code to require data fuzzing
Differential privacy is a particular method of data fuzzing. In theory, essentially random noise is injected into the data, kind of like the margin of error arising from sampling in a poll
But adding random noise to data can create zero or negative population numbers. What Alabama should also challenge is what the Census Bureau calls the "post-processing" to ensure population numbers can't be zero or negative
The differential privacy/post-processing data fuzzing is potentially creating problems that might result in broad bipartisan support for Alabama's lawsuit. These problems aren't too evident when you're working with these data at the state, congressional districts, counties, etc.
But when you drill down to the local level, the data fuzzing is creating implausible counts of children, shifts of minority population out of dense racial/ethnic communities, nonsensical housing occupancy rates, Tribal areas with zero population, and on and on
The Census Bureau claims they fixed these problems that became evident late in 2019, but the post-processing adjustment is a black box and the Census Bureau has not demonstrated what improvements they have made to fix data fuzzing problems
I would add the Census Bureau was rightfully concerned another federal agency like ICE could do a re-identification attack on the aggregate citizenship data to target the deporting of people, but since the Census Bureau will no longer release citizenship data, the issue has waned

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Michael McDonald

Michael McDonald Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @ElectProject

11 Mar
Oh great. The last time California held a recall election, there were 135 candidates that ran to replace Gray Davis
Just think, if Trump ran there would be a good chance the vote would be so fractured he could win
I shudder to think, but Trump might just run for California Governor

- He can scam, I mean, raise more money
- He gets to hold hate fests, I mean, campaign rallies
- He can rail on how bad big state Democrats are
- If he loses, he can just claim it was vote fraud
Read 4 tweets
9 Mar
My report on voting in the 2020 general election co-authored with @NpVOTE is out!
One of the important relationships in the America Goes to the Polls report is percent mail ballot usage vs. turnout rate. There is a clear positive trend. Since states have many laws that affect mail balloting, usage is a better measure than looking at one law in isolation
I do not necessarily think that mail balloting increased turnout in the 2020 election, rather that in an exceptionally high turnout election held during a pandemic, mail balloting offered a safe way for people who wished to practice social distancing to participate
Read 5 tweets
12 Feb
The Census Bureau is now saying the data for redistricting will not be released until September. The delay now seriously threatens the drawing of new districts before the 2022 elections and is sure to spawn lawsuits as states seek to extend legal deadlines nytimes.com/2021/02/11/us/…
The pandemic and the Trump administration's rush to short-circuit the count must have done a real number on the census
The release of these census data for redistricting continues to be a moving target. Just a week or so ago, census was saying these data would be released on July 30. If I were in charge of redistricting, I would not be certain of even this new September release date
Read 6 tweets
9 Feb
Not a surprise, but early voting in Virginia shot up in the 2020 general election elections.virginia.gov/media/formswar…
Virginia new voter registrations were down slightly in 2020 from 2016. An increase in new online registrations almost made up for a decrease in new paper registrations
Virginia extended its deadline for the receipt of mail ballots (postmarked by Election Day). 10,901 mail ballots were received in the extended period. 2,063 mail ballots were received after the deadline and were rejected
Read 4 tweets
24 Jan
There really is a simple answer to this problem: don’t invite liars on to your programs. Republicans are now in the minority. There is no news value of having them on to spout and mainstream conspiracy theories
Join us again next week for another episode of outrage theater
I think the media is broken. And that is it. There is no both sides about the election outcome. Inviting guests on to shows to spout conspiracy theories has downstream effects that are destroying peoples' lives, all for a few more clicks and eyeballs
Read 4 tweets
16 Jan
North Carolina party registration changes over past week

Rep -> something else
3,007

Dem -> something else
667

Most of these changes are to Unaffiliated

Rep -> Unaffiliated
2,650

Dem -> Unaffiliated
484
These statewide changes appear to confirm anecdotes of persons changing their party registration newsweek.com/swing-states-r…
Hard to gauge the significance of these party registration changes. On one hand, they're a small proportion of the 8.2 million NC registered voters. On the other, changing party registration is not a common thing, so they may represent stronger undercurrents
Read 14 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!