Privacy concerns are breaking the Census • Hacker News Board

Re: Privacy concerns are breaking the Census

#41

Post by jefftk » Wed, Jan 26, 2022, 5:59 PM UTC

The language of "do not disclose the information reported by, or on behalf of, any particular respondent" really does look to me like it requires Differential Privacy. Otherwise the different aggregations will reveal information for some participants.

It's unfortunate that this means the Census isn't accurately presenting information needed for fair representation, but I don't think this is something that can be fixed by attempting to shame the Census for putting out "inaccurate information"?

Re: Privacy concerns are breaking the Census

#42

Post by droidist2 » Wed, Jan 26, 2022, 5:59 PM UTC

HedgingPointer wrote: post #24

It seems as thought this author has never read any materials on inference control... Ross Anderson's Security Engineering Chapter 9 is all about this https://www.cl.cam.ac.uk/~rja14/Papers/SEv2-c09.pdf (Anderson releases the textbook free online) This is a fairly well developed field (i.e., in medicine there highly screened "Medical Statisticians" who give data to research groups only after it's properly anonymized.…

> They know beforehand what the researchers are interested in, and introduce noise that effects all the measures _except_ the ones the researchers said they were interested in looking at beforehand.

Isn't there a danger in the census statisticians doing the same thing? That they'll introduce noise that affects looking at the data in ways that they didn't anticipate or allow for. Or even that hinders drawing conclusions that go against current dogma.

Re: Privacy concerns are breaking the Census

#43

Post by giaour » Wed, Jan 26, 2022, 6:10 PM UTC

This was an extremely lazy article. It doesn't seem like Yglesias made any effort to engage with the Census's rationale for using differential privacy, with the Bureau's historic methods for privacy preservation (e.g., "swapping," a different way to deliberately introduce errors that has been in use in Census data products for decades), or with what might have changed since 2010 that might justify a new approach to privacy. Even though congress hasn't updated Title 13 in the past decade, there have been significant advances in statistical methodology that reduced or eliminated the effectiveness of the Bureau's previous methods of privacy preservation.

Re: Privacy concerns are breaking the Census

#44

Post by parkingrift » Wed, Jan 26, 2022, 6:18 PM UTC

Can someone explain to me why we need the census at all? Why are we actively pretending that the federal government doesn't have precise estimates for where people live?

Re: Privacy concerns are breaking the Census

#45

Post by willis936 » Wed, Jan 26, 2022, 6:18 PM UTC

tharne wrote: post #27

The Census has really lost the thread. The original mission was simply to do a headcount to determine how many representatives and electoral votes each state was entitled to. One thing that annoyed me in particular was their creepy obsession around race. Even once you answer, "White", there are several more questions aimed at determining precisely what subcategory of white person you are. If anyone at the census is l…

You're misrepresenting what a census is and why they are done (yes, including historically since the US began and even before).

You might not like the idea of being accounted for, but that negates exactly zero benefit to society that census brings. How can the rulemakers make informed decisions if they're blind?

Re: Privacy concerns are breaking the Census

#46

Post by ironman1478 » Wed, Jan 26, 2022, 6:19 PM UTC

tharne wrote: post #31

ironman1478 wrote: post #28

Earlier quoted context omitted.

This is important for understanding how gerrymandered a district is.

> This is important for understanding how gerrymandered a district is. That's a bit circular. One of the main reasons that districts are so easy to gerrymander is precisely because there is so much rich census data that can be used to predict voting patterns.

Its probably possible to gerrymander a state without this information? Sure, it would be harder (and probably less effective) but you could use proxy metrics to figure out the distribution of race within a state. I imagine I could use housing data as a proxy for example, given that property values are correlated with race. I don't have exact proof of that last statement, but it feels right.

Also, I did some reading. Apparently the census has been asking these questions since 1790 https://www.census.gov/acs/www/about/why-we-ask-each-questio.... Though, I imagine the nature of the question has changed over time.

Re: Privacy concerns are breaking the Census

#47

Post by spicyusername » Wed, Jan 26, 2022, 6:19 PM UTC

parkingrift wrote: post #44

Can someone explain to me why we need the census at all? Why are we actively pretending that the federal government doesn't have precise estimates for where people live?

> have precise estimates for where people live?

The census is how they have that estimate...

Re: Privacy concerns are breaking the Census

#48

Post by ironman1478 » Wed, Jan 26, 2022, 6:21 PM UTC

caymanjim wrote: post #36

ironman1478 wrote: post #28

Earlier quoted context omitted.

This is important for understanding how gerrymandered a district is.

That's not the purpose of the census, though. That's a political and social concern, and people can debate the value of that data and who should collect it, but it's not part of the Constitutional mandate to count the number of people.

https://www.census.gov/programs-surveys/decennial-census/abo....

Unless I misunderstood your question and you meant to say that the race portion specifically is not.

Re: Privacy concerns are breaking the Census

#49

Post by brimble » Wed, Jan 26, 2022, 6:38 PM UTC

parkingrift wrote: post #44

Can someone explain to me why we need the census at all? Why are we actively pretending that the federal government doesn't have precise estimates for where people live?

The way we live in a complete surveillance state except when it would in any way provide convenience, at which point we have to pretend we don't, is something I find irritating on a pretty regular basis.

Re: Privacy concerns are breaking the Census

#50

Post by brimble » Wed, Jan 26, 2022, 6:40 PM UTC

tharne wrote: post #31

ironman1478 wrote: post #28

Earlier quoted context omitted.

This is important for understanding how gerrymandered a district is.

> This is important for understanding how gerrymandered a district is. That's a bit circular. One of the main reasons that districts are so easy to gerrymander is precisely because there is so much rich census data that can be used to predict voting patterns.

Parties and campaigns already buy & maintain huge private databases on this stuff. They wouldn't hesitate to buy slightly more to have that info, because gerrymandering is so powerful.

Watchdogs and groups trying to fight gerrymandering (racially-tilted or otherwise) usually don't have as many resources. It's harder for the public to keep an eye on this without public data, while keeping it private-only barely inconveniences the most relevant bad actors.