What's wrong with having residents fill out cards and count pregnant women. Complex solutions regularly bog down just like the Iowa count did. Count the votes and fax the numbers. Or call each county and get a total. Esoteric technologies don't apply to every situation.
2
The census is very important and we all need to be very precise when we speak about how it is implemented. The GOP is ltrying to achieve an ACCURATE count of white people and the elimination by undercounting of people of color. That is the whole point of the census while the GOP is in power.
2
Age 70, I have never done the census
1st 20 years we moved constantly, next 20 was not stable either
Age 40 to 70 we were easy to find
Nobody ever counted
1
@Lost In America
Not necessarily true. An enumerator might have gotten information from a neighbor. I worked for the Census Bureau and it is not uncommon to get reliable information without asking the family. The primary information is number of people in the household, gender, age and race.
4
Living in a town with ~80 people within a county with ~2000 people I can pretty much guarantee that the data aggregators already have all the information on us they need.
But it's especially important that our census population figures are accurate. It makes no difference for US Representative redistricting, since the state only gets one, but state legislature districts are also drawn from census data. And like it or not government aid is also based on census data.
It's going to be pretty obvious here if they start swapping data in and out, and very unpopular.
2
If all small towns disappear and we retrain their unemployed and underemployed denizens, plus giving them affordable housing in cities, they can then be morphed from Deplorables into Informed Citizens, not that many of the rest of us are.
We all then would only need the congress we had circa 1945- 1979, with a couple of improvements.
This algorithm has nothing to do with privacy and everything to do with voter suppression when the incorrect information will go to the states for redistricting.
Of course Republicans will pretend it is privacy they are so worried about but do nothing about Trump buying cell phone tracking data to target American's phones to see where they are at any given moment. That would be especially helpful in tracking immigrants.
https://www.independent.co.uk/news/world/americas/trump-ice-immigration-cellphone-location-data-report-surveillance-tracking-a9324221.html
3
Gerrymandering at it's finest.
3
The algorithm would seem to violate the Constitution since the census is a requirement precisely so an ACCURATE count can be obtained. Am I the only sane person left in this country?
18
@Jazz Paw No. There are at least 2 of us.
2
It makes sense to add "noise" to data to prevent reverse engineering.
But this does not have to be done in a uniformly random way. Areas with a low median income and areas with low population density will tend to need more government funding per person. So might places far from their state capital or a large city.
The algorithm could use these and similar factors to add mostly positive but still random "noise" to areas with greater need and mostly negative but still random noise elsewhere. Transparency about the algorithm is necessary for it to be accepted.
2
@Alan Good points, but federal funding is not the only important use of census data. An inflated population is also dangerous. It will make disease rates appear lower than they are, impacting the way public health interventions are prioritized and targeted. It will skew our understanding of whether disease rates are increasing or decreasing, which will limit our ability to detect and ultimately understand potential causes of disease such as environmental exposures.
4
This is a serious problem. However, it has become a problem because of the ability to apply technology to data, to locate and ID people from fragments of data, added to other data bases.
We need a census. We need to be able to guarantee anonymity, because it is the only way to get complete and accurate data that we need.
We also need that census to reveal to us the full pattern of people in our country, including urban areas, rural areas, and small towns. Yet if we let those who crunch data do so with this, and add to it what else they know or can find out, then the predicate of a safely anonymous census is defeated.
What we need, and always had, has run up against what some data miners are able to do to defeat our purposes.
They must be stopped. We can. Make it a crime, and enforce it. We already do that with tax returns, which were made Constitutional by being made secret with an extreme privilege. We need to do that here, and enforce it. We know we can, because we have for tax returns.
16
@Mark Thomason
Exactly what sorry of data mining do you think should be outlawed? Do you think using census data to determine the number of eligible voters or the number of U.S. citizens eligible for benefits should be a crime?
1
@William Case -- Good question. We would need to be very careful about the limits on what can be done with census data. But we need to do it anyway. The alternatives are to defeat the census or abuse the census until we defeat it anyway.
2
Call me a skeptic, but to state the algorithm is to protect privacy but yet it diminishes the key objective of the census is total hogwash. To undercount minority and rural populations is to deprive them of representation in government.
We’ve had gerrymandering, voter suppression and now a means to undercount minorities. The algorithm needs to be reworked or thrown out altogether. When has our government bent over backwards to protect our “privacy” anyways?
28
@Dawn Agree, it seems like too convenient of an excuse.
2
"Some weaknesses in the privacy algorithm’s implementation are only just now coming to light. A recent analysis by Randall Akee, a professor at the University of California, Los Angeles, found that on reservations with fewer than 5,000 people, the algorithm decreased the population of Native Americans by an average of 34 percent."
You think this is a weakness? A design flaw, maybe? I've got a bridge to sell you (or maybe a reservation).
20
Liberals should be BEGGING undocumented residents to participate in the census so we have more opportunities to acquire government funding. California harbors over 5 million undocumented immigrant workers, many of whom have children born in the US who are NOT COUNTED! California loses millions of dollars in federal funding because of their irrational fear of "being discovered." This is why California residents are being cheated and paying substantially higher taxes. Even NPR did a story on this and encouraged undocumented families to participate.
I'm tired of having to carry water for these people!
The message to the undocumented should be: Either participate in the census or you will be deported!
4
The census is one collection of data that was not meant to be private. With this algorithm, Republicans will be up to all kinds of mischief to harm the poor and blue states,
16
Exactly what privacy issues arise in the census since the name, and address of the individual is not published? It is time we hire (and elect) intelligent people to run our government.
15
Algorithm's have no place in taking a census. The bigger issue is counting illegal immigrants. Not having accurate counts of all people living in a community makes it hard to plan for the services the community needs as well as state and federal funding required to support them
10
Ah, the 21st century, where simply counting persons is no longer enough.
9
The census is a massive undertaking that costs $15 billion. We are told that, in one effort, the privacy algorithm couldn't effectively be reverse engineered. Let's not be ridiculous -- find some hot shot computer engineer from MIT or CalTech or Northeastern Southwestern State, give him or her a $10,000 bonus and solve the problem lickity split.
8
@michjas Hi, MICHJAS. I haven't heard the phrase "lickity split" since my mother died in 2000. Anyway, you don't need a hotshot at MIT or CAl Tech or Northeastern computer engineer to solve the problem of reverse engineering the census. My eighth grade nephew hacked into his school computer on a snow day at home recently, and shut down the homework assignment that day for the entire school system.
1
Scrambling census data won’t effect the number of congressional representatives or number of electoral votes allotted each state, but the scrambled data will be useless for drawing voting district maps. States would be free to base voting district maps on other data, such as the annual American Community Survey, which asks the citizenship question. (Almost all federal aid to states are distributed based on ACS data, not census data). States also could base voting district maps on the number of registered voters. Since only U.S. citizens can register to vote, foreign nationals and illegal immigrants would not be counted. Six states already already authorize alternate methods of estimating population.
https://www.ncsl.org/blog/2019/06/20/must-states-use-census-data-for-redistricting-not-always.aspx
3
But eliminating small towns is the point! The GOP wants no small towns to exist for voting purposes. So I will vote blue no matter who in November!
11
@Cookie
What's your reasoning? The big cities vote blue, the little rural towns vote red.
1
@Cookie Huh? Not questioning your voting decision, but I thought the GOP likes small, rural towns and not big cities and some suburbs. I doubt this algorithm is either Republican or Democratic.
1
Thank you NYT for covering this topic, though what we the people can do about it goodness only knows. We are being asked to respond on-line on census day. There will be no paper trail for those being counted that way. How do any of us know some of those on-line responses won't just disappear in some way? I will wait until I get a paper form in the mail because I like having a paper trail for the census, just like I like having paper ballots available in the real world after voting in an election.
10
According to US Census bureau information, the only personal aspect of the census will be the names of those counted, with the intent of avoiding duplications. The other seven questions break down along lines of race, age, sex, and employment status of participants, plus whether dwelling is rented or owned. Perhaps such a gathering of benign info will lead to more responses and an accurate count.
It is good to see so many respondents here concerned about that accuracy.
8
The constitution requires that all persons be counted. A census is potentially allowed to ask (as defined by Congress) race, age, gender and residency.
My personal opinion is that doing anything beyond counting people should not be allowed. But if I am legally required to answer those other four questions, I will. That's it. They can take their ACS survey and stuff it.
As to the census itself. why does it matter if it is de-anonymized by an algorithm? Anyone can already find out my race, age, gender and residency via a simple internet search. That is why I will answer those questions and nothing else.
So far as "randomizing" and scrambling the actual counts of people--- it's gerrymandering. Anyone can see that.
17
@Concernicus Yes, people often give way more information than what the census asks for when registering for a prize at the local boat show, for example. Why does this data need an algorithm at all?
"Only in America?" This is COMPLETE insanity!! There must be better ways of protecting people's privacy in census data, than making the results innacurate - extremely innacurate some cases. Why not simply use the raw data carefully collected by many volunteers, to do the official, government-related stuff, and release the "algorithmized" version to the public? Better yet, why not quickly develop an algorithm that preserves privacy but maintains the accuracy of the census data? If this goes down as the least accurate census EVER then as the author of this fine article has stated, the rate of compliance in future censuses will be dramatically reduced - and public confidence in the whole process will be undermined greatly.
12
Clearly, the collection of census data is critical to politics, economics, scholarship, history. Banking and health care institutions collect demographic data and keep it accurate and private without scrambling it, mixing it, and adulterating it. Why, then, does the census bureau need to falsify the data out of concerns for privacy? Doesn't this look like a consiracy to anyone else?
25
@Lady Grey
"Doesn't this look like a conspiracy to anyone else?"
Actually, it looks like a Republican conspiracy to just about every commenter, myself included.
I thought the purpose of the census was to count heads. That's all the Constitution requires. Everything else is something somebody decided to add along the way.
Those additions need to be dumped if the end result of the proposed census process is that heads are not counted accurately.
24
@Renee - We have altered much of the original purposes because reasons... and this isn't June 21, 1788 any longer.
5
It seems data could be anonymized through aggregation depending on need and the use for which it is being used. Raw population for purposes of allocating total representatives is ALL that is needed for states. A further detailed breakdown by race and above a voting age threshold is ALL that is needed to draw fair voting districts. The total number of people in Toksook Bay is ALL that's needed for them to receive their fair share of funds. If they need to provide more details like age breakdown, that does not need to be public info at the level of their village, but should be known to officials and others that NEED to know.
I can't see how the details that actually NEED to be public, aggregated statewide or by census tract, would allow data miners to cross reference the info with things like credit reports to provide information that is more detailed than the credit reports themselves. The "sophisticated techniques" have to be relying on more Census info than tract, whether an age is above 18, and race. Withhold all that other info and aggregate the rest. Data miners will then be so statistically uncertain about any detailed conclusions that their conclusions will be useless.
And yes, I also agree with those that say distribution of private info by private sector data collectors needs to be regulated.
9
I predict this is going to be a major disaster. With this process where is the raw data going to be kept? How will we know what the real census numbers are? Creating false results to protect privacy seems bizarre to me. What will they do if they report a community had say 500 people, but the community can prove 789 live there? At this point my faith in any on line system to conduct this is pretty shaky.
23
There's been a lot of horrifying news in recent weeks, but this article is the most alarming thing I've read. How can gerrymandering be identified and opposed if no one knows exactly where people live? How can manipulation of census data to serve political interests be identified and opposed?
If I have misunderstood this article, please clarify.
21
@Yellow Dog
Or is this "manipulation of census data to serve political interests"?
3
This rigging of data is unthinkable. Why will anyone agree to participate in surveys once this becomes widely known. Accuracy should not be thrown out for any reason. The Census Bureau needs to scrap this.
26
First, we are told that computers cannot make mistakes.
Second, we know that computer programmers are human thus can make mistakes.
Third, many/most of us don't know what an algorithm can do:
al·go·rithm: a process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer.
Fourth, Liberals won't believe a census that doesn't favor their constituents.......and Conservatives won't believe a census that doesn't identify citizens. Independents want to know how many citizens/legal residents/undocumented residents: it's important.
Fifth, most of us won't like the idea that a computer program makes changes - called 'machine learning' - changes that humans didn't make: "Generally speaking, when people talk about algorithms these days, they’re talking about something more specific, like the operations that power our social media news feeds. In one way or another, most of these systems are examples of a technology called machine learning. Instead of repeatedly processing a stable set of instructions, systems based on machine learning rewrite themselves as they work. It’s this that frightens some people, since it makes algorithms sound like they’re alive, possibly even sentient. (To be clear, they are neither.)"
Last - some of us will not trust what others of us will do and few of us will trust what we are told is 'machine learning'.
So we will be in court for years.
Paper and pens, please.
14
@Azalea Lover I am what I suppose you would call a liberal. I don't like a census that doesn't accurately report WHO lives WHERE. The census should go back to doing what it did in 1850-1880.
4
The US Census Bureau is using the new scrambling algorithm (called Differential Privacy by the way) because they believe that the current approach does not or may not protect privacy well enough. This statement from the article speaks to that point:
"Officials were able to correctly identify just 17 percent of the original 309 million records."
This statement can be interpreted two ways:
1. Officials made 17% correct identifications, and knew without checking which identifications were correct.
2. Officials made 17% correct identifications, but couldn't tell which were correct until they checked.
Which one it is makes a huge difference. The first protects privacy very well, the latter very poorly.
9
@Paul (US citizen) Hate to be under the operating knife with these procedures!!
2
Census data are extremely important in all kinds of economic, sociological, and political science research. They aren't just used to draw congressional districts. The Trump administration is already ignoring the real economic problems in the economy and claiming it's booming based only on the stock market. This will make it that much harder to identify the real economic problems we face-- student loan debt, stagnant wages, the gig economy, etc. Having flawed data for the next ten years will not help straighten things out.
20
Scholars of numerous types often use census material in their research.
What is described results in "fake data" and subsequently in fake scholarship with no real way of knowing this.
26
This worrying about the truth is just plain silly.
The truth went out the window with Trump.
Truth, lies, exaggerations, alternate facts...... this is the new way to MAGA. and we know for certain America only used to be great. Now it lies about that, and everything else.
And that's the truth!
26
@Doug Riemer Apparently, the truth went out the window in 2010. This 2020 census is just making the kids worse.
1
@Doug Riemer
If you think the truth went out the window with Trump, you are sadly misinformed.
Politicians tell us what they want us to hear - and tell us what we want to hear. TV "reporters" do the same thing.
Liberals tell us their views/truths; conservatives tell us their views/truths.
The truth is somewhere in the middle.
1
I understand the need to increase small localities' population numbers in order to keep results confidential. But why do that by siphoning off individuals from other small towns? Doesn't that just increase the latter towns' confidentiality issues? Why not transfer (statistically) individuals from America's largest cities, where the slight reduction would have little practical effect on the overall population levels?
30
What an odd predicament. Are we not exaggerating privacy concerns when this sort of census monstrosity results? Seems there are two possible fixes. First, a technical fix, wherein some wiz kid figures out a way, using quantum computing or whatever, to enter a secret code along with the data, that can later be used to unscramble the mess they have created. And second, a policy/political fix that results in hearings that weigh the relative importance of the various policy objectives - accuracy, privacy, etc. - and comes up with some kind of reasonable compromise. That's how government should, and can, work.
34
@James
Perhaps we could even have a law preventing corporations from using data without the the explicit and prior consent of the individual? What an odd idea!
124
@James Yes, good old fashioned political compromise is the way these problems were dealt with in the past. Unfortunately, thanks in part to social media, we are too polarized to accomplish this.
7
@Neal
That horse bolted the barn years ago - you'd be threatening big data, and all the money and political influence they bring to bear.
7
Wow, as the resident of a small town I find this terrifying. We are growing, slowly, but the economics of small towns are very tough and we can't afford an algorithmic undercount. Surely there is a better way? Also, this feels like a recipe for gerrymandering, no matter if the state level count is accurate....
145
@Susan
The worst thing that can happen to a small town would be if it was growing.
The best thing that could happen to a small town was that it would continue declining.
1
@Susan: Good point! The Census Bureau acknowledges the importance of getting each state's total correct in order to correctly apportion the the number of Congressional districts. It seems that the states also need accurate data to apportion the allowed number of Congressional districts to ensure that cities, small towns, and rural areas have the correct representation per capita.
6
Let's be honest. The algorithm has nothing to do with privacy. Under counting the population benefits Republicans and Republicans currently control census administration. That's what the citizenship question was about as well. Population growth and demographic diversity can only harm the GOP politically. Hence, partisans seek to suppress the official record numerically. We're witnessing partisan gerrymandering in a different form.
177
@Andy That isn't quite right. I am not at all for this current administration, but small rural towns often lean republican, and these are the towns that are mostly likely to have large degrees of error (often reducing their count). The algorithm will keep the state populations invariant (i.e. set to their true counted value), so states will have an accurate count to the extent that they actually have their population respond to the census. This is because one of the mandated reasons for the census is congressional apportionment. The problem will be at smaller geographies and when you try to get a clearer picture of the characteristics of the population. It probably won't be too bad at the county level for most counties, but at smaller geographies, the data released regarding race and the age distribution of the population will likely be meaningless.
19
@Andy
So it's your position that when Obama conducted the 2010 census he intentionally overcounted rural districts that leaned Republican?
Currently, the Census Bureau is attempting to reduce the misrepresentations inherent in the Obama algorithms that misrepresent minority rural populations.
Why is it that Democrats always advocate for a more powerful federal government? Could it be that when Democrats are in control they like authoritarianism? But quickly attribute the motives of the party that advocates for a smaller federal government their own deceptions?
6
@ebmem all the evaluations of the 2010 census show that it was far more accurate than the 2000 census. It is generally considered to be a "good census." This is not Obama's doing nor is the current algorithim Trump's doing (I doubt he even pays much attention to it). This is largely due to the fact that there was sufficient funding for the 2010 Census (which is both Bush and Obama's doing - as preparing for the census is a decade long feat). The way error was introduced in the 2010 census (for privacy) is very similar to how it was done in 2000. It is not substantial and it is not sufficient for increasingly sophisticated data analytics.
The Census Bureau has been underfunded leading up to the 2020 census, so that may factor into the quality of this census, in addition to the algorithm. The Census Bureau is run by career demographers and statisticians, not politicians though politicians do have say in the level of funding the Bureau receives.
46
Generally small towns have group think. They work hard to control who succeeds (always those with power in local government) and who gets shut out. They do more brainwash work than other communities and they shrink because a few never take to the training. You are exposed in a small town and the people are in crazy cults believing in magic people in the sky. If you can live independently the rural life is good. If you need other people you will be a slave to lunatics.
They didn't build up their systems of control and exploitation just to be fair to outsiders and immigrants. New people must grovel before the feet of the priests and politicians or be ruined financially.
6
@Chris Your comment is rather insightful.
2
@Chris
Cameron County, PA (home) 75% Trump
NYC, NY 79% Clinton
Groupthink in small town America is a serious problem!!
2
Publish the true total number of persons in a census tract, but scramble the detailed census data according to the algorithm. Would that be O.K.?
8
Small towns are already dead.
2
The U.S. government is now going to falsify the Census, and the usual suspects will benefit. This is an incredibly important story. Why on earth is the NYT burying it away in the Opinion section? SMH.
20
Pencil and paper anyone?
5
@Diego By asking about using pencil and paper, you have said a lot. We have been through years of being told to computerize everything. Within a few years the computerized information is gone or lost. In contrast, written records and printed book on acid free paper last a hundred years. And handwriting can be photographed and converted to a computer font. I think that libraries have found that every ten years or so it is necessary to convert digital archives to the latest technology or be lost forever. Early records on 5 inch diskettes and audio cassettes are now lost. But papers in boxes in attics are safely preserved. One might surmise that there will be a comeback for letterpress printing and paper questionnaires because they are terribly resilient to degradation over long intervals. The importance of the paper-and-pencil is not completely gone.
14
@Diego
Yes on the pencil and paper........for the census and for voting.
Those who have advocated for voting by computer or so-called smartphone are not just wrong - they completely misunderstand the potential for manipulation of information.....and misinformation.
9
This is absolutely ridiculous, and anyone that sells this, or buys into it, can't think straight. I mean, you can still protect individuals by not putting a name to the data, and that is all that is necessary, really. This method, is an entire waste of money, actually, cancel the whole thing, as trying to brainwash any group of people into believing that this is how a census should be conducted is absolutely insane.
14
@MaryKayKlassen
Unfortunately by not putting a name to data collected about individuals and households, census enumerators could fabricate as many false records as a day's work would allow. Such a result would skew the data to an even greater extent than the census algorithm.
7
@Pdf
Your point is absolutely correct. It would easily be possible to sit at home and fabricate records while being paid to collect census data.
2
Do we do nothing as a country that is what it appears?
5
The algorithm strikes again, now as concerns the system for counting us Americans, a system described by former USCB Director Professor Kenneth Prewitt as archaic and in need of replacement.
An introduction to Prewitt's recommendations is available in his August 2013 OpEd @ https://nyti.ms/13GYjwo
That OpEd was written at the request of a New York Times Editor, name unknown, who found Prewitt's just published book interesting, which it indeed is.
The contents of that book bear directly on what is reported on in this column. In that book, he proposes that classification by race/ethnicity gradually be phased out. During this phasing out process the collection of more and better socio-economic data would be phased in.
As a citizen of the USA and Sweden - and as a manuscript reviewer and translator for Swedish medical researchers, I am quite familiar with the US system and with the superiority of the Swedish system as a basis for medical and other research.
It appears from this single article that the Census Bureau may be taking the census in a different direction by using the so-called privacy algorithm.
I write this preliminary comment to call attention to the need for serious discussion of all aspects of the system since I have a strong commitment to Professor Prewitt's position that the race/ethnicity question must be removed and that a much higher quality system be developed.
Only-NeverInSweden.blogspot.com
Citizen US SE
13
@Larry Lundgren What is the title of book?
How is it that our government finds more and more ways for the ever-growing number of public-sector gig and career knowledge personnel to be less and less accountable for any sort of tangible outcome – let alone measurable outcome.
In a city that's added pre-K and teacher's aides to an already too-high bill for K-12 education, the average competency scores for high school grads – assuming students even get that far – is abysmal.
So, of course, they first set out to kill all the high-achievement schools. To put a finer point on it, to kill high scholastic achievement of any sort. This isn't limited to STEM. Candidates for prosecutors now campaign on not prosecuting, if elected.
We now are so knee-deep in junk science and science-fiction for so much of our future's brightness, that an idiot proposal like this gets serious discussion. Perhaps we should install facial-recognition cameras everywhere – but intentionally mix up the names. And – to be more fair and just – mistakenly recognize people we don't like, as people doing bad things.
We’ve actually seen our future, in this regard. We should have realized that when our government deployed a sad embarrassment of a sad joke for our health care IT infrastructure, they were just protecting our privacy.
6
Oh, yeah, I trust Trump’s census bureau to change around the numbers and just ‘trust’ them that it is accurate.
12
We should get over this hysterical obsession about privacy. It's not that important. It's not important at all.
In traditional villages there was no privacy. Everyone knew everything about everyone. No one seemed to think there was anything wrong with that.
2
@Jonathan Katz "In traditional villages there was no privacy "... Then civilisation developed and cities emerged, along with knowledge and education.
In small villages people were uneducated and racist (that comes with lack of knowledge and education). This was less of an issue at the time because most people never met outsiders. The control that came with the lack of privacy was a major issue for many people but they couldn't do much about. it.
2
It is definitely not shocking that the US government might do a poor job with the census to the detriment of minority groups.
8
"Fearing that data brokers using new statistical techniques could de-anonymize the published population totals" - to what detail are these data published? Why publish them? Can they be protected from FOIA? Let the big-data people generate their own census for their own purposes and keep the decennial census for its purpose.
4
This is dishonest in the extreme. Must be a republican idea, Who thinks it stands a chance of fending off greedy corporation and their data teams for more than a few years while condemning small towns to inequitable treatment.
Yes, our government is failing us.
4
@kirk
Doubt it's a Republican idea. Expect it's a computer programmers' idea.
They trust the information put into the computer, and trust the algorithm that manipulates the data.
3
Scrambling algorithms to misrepresent how many people live in a given place? Algorithms that diminish the number of Native Americans? Why are we not applying to all populations of color and other outliers, so we can steal some more of their fair share of the American pie? It's such a perfect, beautiful plan, it could only have been conceived by an agency trying to curry favor with Trump and the White Rights Republican Party.
8
Is there no one that can finally take charge of this madness? Facebook chooses our President. Fox News decides which pizza place will be attacked by armed vigilantes who believe insane rumors. Young men in Macedonia make a living as Trump trolls. SWAT departments are dispatched to innocent people's homes sometimes as pranks sometimes as hate crimes.
And now the census has to be skewed in order to protect Americans privacy? If only we had an actual President. Or a Senate the majority party of which was not made up of the worst of the worst. If only there was someone looking out for us. Instead our President is just another Internet Predator looking to prey on the most vulnerable. And Republican politicians are probably the ones making the vicious prank calls.
14
@Brother Shuyun
If you think it's only Republican politicians and their employees who are doing this, you are not keeping up.
See: Iowa caucuses.
1
@Azalea Lover
Rumor has it (may be just a rumor, but sounds plausible) that some Trumpers got the phone number in Iowa and jammed the lines.
1
"Imaginary people will be added to some locations and real people will be removed from others."
Oh. My. God.
I would distrust this if the Democrats suggested it. In the hands of Republicans it reeks of brimstone and doom.
You know it, I know it.
30
Sounds like this algorithm was designed by the same people that did the Iowa Caucus App. I don't trust it. Let's not enforce messed up technology on something so important.
9
Why does the Census Bureau need our credit scores?
2
@Jacquie The Census Bureau isn't collecting our credit scores. From the article:
"The more the algorithm muddles the results, the more difficult it will be, for example, for a data scientist to combine a set of addresses and credit scores with census results to learn the age and race of people living on a certain block."
The "set of addresses and credit scores" would be obtained from data sets *other than* the Census, and combined with census data (which contain age and race - something that may not be present in the data set from which this hypothetical data scientist got the credit scores and addresses).
So, the idea here is that the algorithm would scramble up the census results in such a way as to prevent that data scientist from matching up "addresses and credit scores" with "age and race" data on the same people.
7
You can have privacy or you can have transparency. You can have safety or you can have maximum freedom. You cannot have both. So many or our problems now stem from this conflict. A lot of people make a fetish of hunter-gatherer societies and small rural villages. The virtues of those societies are based on everyone knowing everyone else.
5
A valuable article:
Our "representation," as currently designed (with minimal available contributions/participation from citizens) continuously evolves to disguise how we are manipulated and channeled into the slots which the dominant culture constructs and adapts to disguise our actual servitude.
4
@Barry McKenna It looks like voter suppression since results will go to the states for redistricting.
5
This is a very, very, very bad idea. It makes the census results inherently questionable, it makes the census bureau appear even more political, it opens the results up to computerized manipulation and adjustment by the party in power, it will force judges when examining lawsuits to consider arguments that the census is wrong, or skewed or inaccurate. Privacy is a legitimate concern but this seems like the wrong way to address it.
45
I have tracked families name by name in every federal census back to the beginning. Will this privacy algorithm end that? When binary is introduced will we know how many males and females were in a family? As Charley Brown would say "Good grief!"
1
@Bill D No. These privacy protections do not affect historical records. After 72 years census privacy protections expire and the unredacted data is released.
8
Didn’t our NSA develop math that models Metadata, data that produces information about other data? Statistics from telephone use, internet use, USPS use, credit cards, banking statistics, visits to hospitals, visits to merchants, electricity use, school attendance, & water use are meaningful data. This data could check the initial census results. If our government is using reality modeling algorithms to establish population estimates, can’t other data provide “flags” that require human review and inspection?
If the famous Bell Curve is so “fat” it’s showing less than 1 sigma reliability, then we are back to driving the 1986 Yugo, the 1978 Dodge Omni. Somebody has to speak up. It won’t be me. I got a C in that course.
6
This article is concerningly dismissive of privacy.
It says "just 17 percent of the original 309 million records" were correctly identified. That's 52 million people whose privacy was compromised! And the Census Bureau have done further experiments that correctly identified 46% of Americans - that's 142 million people whose data is implicitly revealed.
The Census Bureau is rightly concerned about Americans' privacy.
2
@Thomas
This is one thing I wanted to understand better. It makes every difference whether the correct identification was known to be correct or not. If I can get 17% correct, but don't know which of my predictions are correct and which are not, then privacy is very well protected. If on the other hand I somehow know when I got a correct prediction, then that is very poor privacy.
From other things I've read (and written ... see https://medium.com/@francis_49362/dear-differential-privacy-put-up-or-shut-up-48ff255ec35), I think it is the former (good privacy). But this is an important point and I wish it was made clear.
2
@Thomas
Why are we so worried about privacy on the census data when private companies legally collect and sell much more sensitive data without any restrictions?
For example, if you use your Krogers card while purchasing groceries, Krogers has a complete account of everything that goes into your stomach at home. Think your health insurance company is interested in that data?
Do you buy gas with a credit card? Both your credit card and gasoline companies know how much you drive, where you are, and can (and do) use the data for profit.
I think the issue here is that the census is a government function, and we don't trust government, especially with Republicans in charge.
Well, if democrats living in mostly republican areas refused to participate in the census (small fine, but not often applied), that could decrease republican representation and help even out the problem of the electoral college and senate (2 senators per state no matter population). Couldn't it?
8
First of all, the count is the count, no more, no less, just the actual number of people in a given place on a given day.
No enhancements, no manipulations, no algorithms, no anything else, just the actual number of people in a given place on a given day.
Any effort to deviate from that fact is fiction and is prohibited.
Everyone has an angle and angles are not allowed.
That is what a census MUST be.
15
@S Butler Love it!
2
Political ideology is sure to skew the results even more than otherwise. Nothing in this country remains politically neutral anymore and Republicans have even begun to scratch the surface in how far they will go to hold power.
Shelby v. Holder opened the floodgates in voting manipulation, and the Iowa meltdown shows how much damage can result from technology gone awry, inadvertently or intentionally. The "black box" of algorithms can provide the perfect cover for Republican mischief.
We already have Russian meddling, Trump's naked manipulation of government, Republican power grabs such as stolen votes in NC, its advantage in digital tech, plus the "normal" distortions of gerrymandering, restrictions on voter access, the electoral college, local political mischief, the right wing propaganda machine, etc. Then there is probably biggest problem of all: voter apathy.
Soon, there will be one crack too many before the already battered dam holding up our democracy (what little of it we have) comes crashing down. The census could be that one crack.
9
So in the guise of privacy they are erasing the ability of researchers, let alone actual inhabitants, feeling the census is an accurate count or provides usable information.
I grow more disillusioned with this nation by the day. I’m 75 and it’s feeling to me like life in the USA is no longer worth staying alive for.
14
@TheraP The way I understand it from Census Bureau presentations I have been to, is that the real count will be held super protected, and will be released at the usual 70+ year mark. There is a worry with current methods that individual people will be identifiable, and it is a priority of the Bureau that this not happen.
I still don't get how the "mixing records" thing will work, though, and this article didn't clear it up for me.
3
@TheraP It is not the job of the census to feed data to researchers or provide usable information. The only thing required of the census is an enumeration. Congress has decided to distribute federal largess through arcane formulae based on information gathered by the census, by force of law, that is not necessary for apportionment of representatives.
4
Scrambling census data won’t effect the number of congressional representatives or number of electoral votes allotted each state, but the scrambled data will be useless for drawing voting district maps. States would be free to base voting district maps on other data, such as the annual American Community Survey, which asks the citizenship question. (Almost all federal aid to states are distributed based on ACS data, not census data). States also could base voting district maps on the number of registered voters. Since only U.S. citizens can register to vote, foreign nationals and illegal immigrants would not be counted. Six states already already authorize alternate methods of estimating population.
7
Having seen how untested and essentially forever unreliable data skewing algorithms are being thought of is literally a bad solution being jammed into a problem which should have a less nonsensical solution.
I'm not sure a better way would be to enact a law, not an administrative rule, that any person or company which analyzes census data for the purpose of unscrambling it has a mandatory felony which cannot be expunged and a minimum of say 5 years in prison. And the company forfeits all assets it has with mandatory liquidation.
Draconian? Sure, but it might be the thing that gets Google or Facebook or some little known data mining factory who's only reason for doing so it to profit off data.
We have laws protecting national secrets or impinge upon national security. This is no different. The outcome of head counts should be just that. All the extra tripe that people want to add to the census is not required, and likely no longer of the value (at the expense of loss of privacy) that it once was.
Big data has gotten away with far too much for far too long.
7
So we are trying to fix something (privacy) of factual data by adding fake data. I know, let's have a first count, then ask residents if they want to change where they want to live (and have people lobby them to join them in another country) and then have a second count. Then let's assign weight to where people live, so rural communities have more sway than urban areas.
Oh wait, we cannot do that, Iowa already tried that method with their Democrat Caucus system and it failed.
13
"government agencies won’t have access to the original, unmodified census counts when they distribute that money"
The whole point of the census is accurate counts which especially impact smaller communities who struggle with local budgets.
Kudos to the NYT for publicizing this, perhaps it will lead to a common sense result.
58
@Scott Newton Common sense isn't common anymore.
2
@Scott Newton
Hmm. The Obama administration used their 2010 data to strip funding from rural hospitals, which were mostly in districts that voted Republican even though the welfare beneficiaries who needed the funding were mostly Democrats.
Kudos to the NYT for suppressing any evidence of bad Obama policy.
Where was the article in 2012, after the results of the 2010 census were known?
1
@ebmem Funding stripped form hospitals was supposed to be replaced by ACA insured patients who paid the tab. States chose not to protect their own by not implementing the ACA medicaid expansion and not accepting Federal money. Blame the 2010 census and Obama Admin for the loss of funding; blame the state for failing to protect it's own citizens for political reasons. Apparently accepting federal aid for a hospital is conservative, and insuring the people who were being funded is not.
4
Straight numbers do obviously not count for much in the USA.
Perhaps it is so, that the US Mail has better knowledge of the citizens than the census people?
How else can they distribute letters and packages without fail?
34
@Bjarte Rundereim Because the address is on the package (and if someone moves and wants mail forwarded they lest the Post Office know). What a silly question. How does your postal service in Norway work?
1
Is this mess of a scrambling algorithm thought to be the only conceivable way to protect privacy, or something? Why is the Census Bureau committed to it? Was this algorithm specifically a requirement in the legislation, or something?
Readers here, presumably amateurs at this sort of thing, instantly came up with what sound like reasonable solutions to the problem of getting accurate over-all numbers. We're not looking at the management of the Bureau, are we?
19
Government has let data brokers, such as credit reporting businesses, and social media companies such as Facebook, Google and Amazon run amok collecting insecure, think Equifax, detailed files on individuals. These files seem to be able to be cross referenced by bad actors who have not followed the so called privacy rules the private companies claim to follow, think Cambridge Analytica. Now we have undermined trust in the census by having to introduce uncertainty into the census data to protect our privacy from the data brokers who will attempt to cross reference it. This bodes badly for those who see good government, powered by accurate census data, as a counterweight to the ever growing power of private companies and the politicians that are kept in office through their financial backing. I fear it is too late. Our privacy of our personal data is like a dish broken into hundreds of shards. But those shards can and are being virtually reassembled by the data brokers to create an accurate and saleable picture of who we are.
29
@ws
The 2010 Census, run by the Obama administration, was using this privacy algorithm, but it was not newsworthy because it made it possible for urban bureaucrats to confuse data about rural areas.
Now, it's newsworthy, that the Census Bureau is attempting to meet Congress's demand for privacy while decreasing the data distortion.
3
@ebmem That's not what the article says. The official 2010 census results were not reported using this algorithm. Rather, part of the testing for the algorithm involved running the 2010 data through it and comparing the result to the actual, official results.
And this comparison showed that many low-population counties were dramatically over- or under-counted by the algorithm, compared to the actual 2010 count.
7
I am working on my and my family's ancestry. The Census of various counties is very important in doing this work. Undercounting, or lack of census records are quite crippling. Some of my ancestors came from Mississippi, whose census records are sketchy until 1900. I cannot find when my ancestors arrived in MS and from where, got married, or how many children they had. They are all white. Was this a racial problem and only certain counties were added prior to that? Accurate census records are essential to future generations.
13
@Joyce Benkarski
Thank you for writing the comment that immediately popped into my head. My grandmother, God love her, had a way of "Trumping" the truth -- her version of facts were not always 100% spot on. It wasn't until a decade after we lost her that my mom and I figured out that she had lied about her husband's, my grandfather's, age. We figured it out by accident using census information.
12
@Joyce Benkarski
Accurate census records are not needed.
The more inaccurate the better for this country.
@Joyce Benkarski After 72 years privacy protections expire and the unredacted census records will be released. The new privacy protections will not affect historical records at all.
3
It looks like the algorithmically determined variation in county population obeys a normal distribution. That is, there are as many counties who are credited with more people than they have as there are counties whose population who are credited with less people than they have. That would imply that no group or municipality would be given preferential treatment; there would be as many losers as winners. Further, one would expect the population numbers to change much more for smaller communities than larger ones as any migration causes a greater percentage change in smaller communities (when 100 people leave Toksook Bay it's a major change in population, whereas in any big city a loss of 100 people would not be noticeable). We should certainly strive to make the census correct, but it does not appear that your belief that algorithms will hurt all small towns more than larger ones is, on average, correct.
@Dr B, While, “on average”, the data may be correct, the data is used to allocate funding, determine congressional representation, etc. If the scrambled data show Toksook Bay lost 100 people, when they may in fact have gained 100 people, this would be a major blow to the community. The census is supposed to be an accurate counting of the people in the country. Close enough won’t cut it.
6
I’m curious why the counts are not completely before the privacy algorithms are applied.
8
Sorry, scrambling the data to create anonymity should be challenged in court immediately as false data.
82
@tom Get with it, man! It's not false data, it's ALTERNATIVE data. Don't you know we are now firmly ensconced in Trump World?
7
Why have the census at all if we intentionally falsify the data?
75
@Laurie
The enumeration of the population is a constitutional requirement (Article I, Sec. 2, (3) assigned to the legislative branch of government. It must be done.
I worked on the 2010 census and we were told to make the census as accurate as we could. I did not know about the massaging of the results by algorithms. Seems contradictory to me.
The politicization of the results is especially fraught in this time of polarization and squabbles about drawing political district boundaries (un)fairly reflect conflicts of interest.
55
The fact that the government is being forced to skew its own census results to protect the privacy of its people from consumer marketers is just plain crazy! Why can't we legislate the protection of our privacy instead of mucking about with the census results?
166
@kj
Follow the Money. Who financially supports the majority Party? The people who grab your data from the government and bury you in sales calls.
4
@kj Yes, this is just another example of how unregulated tech is destroying our republic and our personal lives.
10
@kj - Criminalize all non-governmental de-anonymization. Huge fines ($1,000 per record?) or jail terms (1 week per record?) might make Zuckerberg think twice. Oh, and put perps in the jug as flight risks pending trial.
17
How about the Census Bureau accurately reporting the total number of people in each location but scrambling the other data (age, sex, etc.)?
So we would know that 1,256 people lived in X town, but the results would show that there are 600 men and 656 women, instead of the real number of 625 and 631? And the median income would be reported as $46,200, rather than the real $55,000?
16
@John Graybeard Yes! The only thing required of the census in the US Constitution is an "enumeration". That is all, nothing more! The other stuff they ask about has been added to the census by sociologists and marketers. It is not needed for apportionment.
7
@John Graybeard
You have just explained how it is that Democrats are able to report that women are paid less than men and rural residents are poor. The algorithm is designed with a bias toward the Democrat agenda.
1
When state legislatures establish congressional district boundaries and city councils establish ward and precinct boundaries, how can anyone reach a fair conclusion on whether the boundaries were fairly drawn or mount an effective challenge based on census data? That is the issue.
Just imagine yourself being the judge trying a gerrymandering case. How long would you listen to evidence based on an algorithm that was designed to misstate the true population of a city, county or congressional district? Would you be able to judge whether the map was drawn to erode minority voting rights?
The real impact of fake census data is to enshrine the gerrymander and voter suppression. An accurate census is to important to our federal system of representative government to taint with any algorithm designed to undermine public confidence in the accuracy and fairness of the census.
93
For the US census to continue to support the fake science of race is deplorable. On the privacy side, Big Data companies most likely have better maps and numbers on how many people, where they live, what they spend, if they work, age, kids, pets, health conditions, and a thousand other things that the census doesn't even think to ask. So this blue the data by throwing in fake numbers is wrong.
23
@Sierra Morgan Most companies and government entities actually rely on the census for population data. The census, historically, has been the most thorough data on the total population. However, I would not be surprised if we saw big data sweeping in an filling this newly developed gap in data availability. Great way for someone to make quite a bit of money!
6
We should keep the census results out of the hands of lawyers and legislators to stop gerrymandering. They have proven themselves to be unable to use the data appropriately.
44