|
 |
January 20, 2004 CENSUS PRIVACY....Genealogy is a hobby of mine, and among genealogists the main gripe about census information is that it's too private. The damn Census Bureau doesn't release census records for 70 long years, and us genealogists don't want to wait that long. Just kidding, of course. We really do understand the need to keep this stuff private and we're glad the government is serious about it. Except that they aren't. Via David Appell, who's really serious about hassling those poor revenoo-ers who show up at his door every few years, it turns out the government has been using census information. They've been using it along with airline passenger records to build their new CAPPS II passenger screening program. Now, I have to say that I kinda sympathize with the airlines who turned over passenger information to the feds for this program. Sure, they shouldn't have done it, but after 9/11 you have to figure that a lot of people wanted to cut government security programs a lot of slack — and understandably so. So maybe the airlines deserve some abuse, but also a bit of understanding for the position they were in. The Census Bureau is a whole different story. Black helicopter conspiracy theorists have been screeching for years that census information isn't really private, and of course the census folks have responded by swearing on stacks of Bibles that yes it is private. Every bit of it. Absolutely. But they were lying. Someone needs to be fired if this report turns out to be true. UPDATE: I've gotten a couple of emails from people who seem knowledgable about this stuff, and they suggest that the census data being used is probably not individual level records. Its usage for this project might still be improper, but it's only a misdemeanor (so to speak), not a felony. As one piece of compelling evidence for this, note that the Electronic Privacy Information Center, which obtained the documents describing the CAPPS II study in the first place, is not making a big deal out of this. If it were a serious problem, they would be. Posted by Kevin Drum at January 20, 2004 12:08 PM | TrackBackComments
well... The Census is private, and John Poindexter no longer runs DIA/DARPA. Right. This can and should be packaged (with Patriot Acts among other items) as a (somewhat important) campaign issue. For chrissakes, there must even be a bunch of right wing Ruby Ridge-types going bananas over this. Posted by: Alejandro Andreotti at January 20, 2004 12:25 PM | PERMALINKI think 1984 is finally here. I don't trust these shmucks. And once again, this will not make us safer, but it will sure help them to harrass 'undesirables'. You know, people who disagree with them. Posted by: four legs good at January 20, 2004 12:26 PM | PERMALINKFired? LOL If they got fired over this, imagine what would happen if the Saudis REALLY DID warn the Bushies about 9/11. Nobody's getting fired over this, or anything the BushCo has done. They're going to get away with all of it, from the looks of things. Welcome to America. Posted by: Monkey at January 20, 2004 12:28 PM | PERMALINKBushco has the great unwashed masses so frightened of terrorists that they have an open mandate to do whatever they please--and they do not even have to use the cover of national security anymore. Out a CIA operative? No big deal. Savage any critic of any policy position? Do it all the time. This issue is a non starter. Gawd I wish is wasn't so. Gawd I wish the know nothings would see the forest through the trees. Posted by: carbon4logic at January 20, 2004 12:34 PM | PERMALINK"...but after 9/11 you have to figure that a lot of people wanted to cut government security programs a lot of slack — and understandably so." If they had been doing their jobs, and using the then existing laws and procedures, we wouldn't be "cuttin them slack" today. We are not cutting them slack, they're cutting our throats! Posted by: Mooser at January 20, 2004 12:57 PM | PERMALINK"I think 1984 is finally here." It's long been my view that 1984 was far less about the future than it was about the past. The shit our 'leaders' pull nowadays makes Oceania look like New Zealand. Posted by: agrajag at January 20, 2004 01:05 PM | PERMALINKIt gets worse. Who mistrusts the census in the greatest numbers--Joe Schmoe, white suburban swing voter, or urban minorities and the homeless? And which way do they lean politically? And are the government programs to help them not funded according to census data? Christ, it's evil two ways. Posted by: Matt at January 20, 2004 01:13 PM | PERMALINK"Bushco has the great unwashed masses so frightened of terrorists" Don't forget the Democrats. If anything, they are trying to outflank Bush ON THE RIGHT on this issue, even including people like Al Sharpton who mentions this often. Go visit rising star John Edwards web site. You'll find "winning the war on terrorism" all over it. Listen to Minority Leader Barbara Boxer - she has the same line - they all do. They think Bush isn't doing ENOUGH to "fight terrorism," which in practical terms means they think he isn't doing ENOUGH spying on people, etc. Posted by: Eli Stephens at January 20, 2004 01:14 PM | PERMALINK
Eli Stevens: No, I think not my American friend. Democrats think Bush is not fighting terrorism. They do not want him to spy on American people but to spy on actual terrorists. Thank you very much, Mr. Bush. You increase my family tenfold while decreasing sanity among your own people. I couldn't have paid you to be a better servant of terrorism. Posted by: Osama bin Laden at January 20, 2004 01:32 PM | PERMALINKOf course this is really all Clinton's fault. :P Posted by: Silence Dogood at January 20, 2004 01:45 PM | PERMALINKWow! Osama posts on this site! Now if Calpundit can find Osama, how come the White House can't? Eli, I have no idea what you're smoking. Most of the Dem candidates
have publicly denounced the naked fearmongering that has been pursued so
vigorously by Emperor Bush, Darth Cheney, and the rest (and is, of
course, the cornerstone of their reelection strategy). And the Democrats
have a valid beef about this Administration not fighting terrorism.
Spying on citizens, sowing chaos in the Middle East, and harassing 80
year old grandmothers at airports is a lot less effective than working
closely with other countries to track down terrorists. Oh, but I keep
forgetting, we hate other countries and don't need them anyway. i worked for the CENSUS in college as a clerk that reviewed forms and "corrected" them. i found detailed information on at least 4 people i knew rather well (the long form, with income, etc.). this in a census area of about 300,000 people. Posted by: razib at January 20, 2004 01:49 PM | PERMALINKThe problem prior to 9/11 was not enough people to evaluate the information we had gathered. We had a transmission talking about the attack the day before, but it took us two days to translate it. The Bush response was to gather more data while not sufficiently increasing the number of people needed to de-code it. If you're looking for a needle in a haystack, to don't bring in more hay. We are not focusing on terrorists, we are looking at everything. Bush is looking through the wrong end of the telescope. If these invasions of privacy had a possible use in locating terrorists I might understand it, but it is obvious that Bush doesn't have a clue. As a former intelligence analyst who went on to spend ten years in criminal investigations, I know what they should be doing, and I can tell you they aren't. Posted by: Bryan at January 20, 2004 01:55 PM | PERMALINKI still say the USA resembles Huxley's Brave New World way more than it does 1984. Posted by: nota bene at January 20, 2004 01:55 PM | PERMALINKI like how they take the same products and just keep re-packaging them with new names. Same crap, different box. Posted by: Susie Dow at January 20, 2004 02:03 PM | PERMALINKIn 2010, I'm telling the census bureau that I have 21 black hispanic children. Posted by: Mother Hubbard at January 20, 2004 02:04 PM | PERMALINKThis is nothing new...Census Block level data was used by the Feds to identify and roundup Japanese in the West during WWII. With that data, they could walk down a block and know that there would be 8 households that they needed to detain... I've just been waiting to see this happen again...and Iwork with this type of data every day of my life. Posted by: DCAl at January 20, 2004 02:12 PM | PERMALINKKevin: Worth noting, if you're going to be up in arms about CAPPS II, that your favorite D candidate lobbied for it. I don't mean to be snarky, but I do often find Clark supporters alternately saying his lobbying for Acxiom was no big deal, that the gov already has all this info anyway, but then getting infuriated by stuff like this. Posted by: emptywheel at January 20, 2004 02:12 PM | PERMALINKA census taker came to my house with the long form. They know
everything. They know how much I earn, what kind of health insurance I
have, even how many toilets in my house. Never again. Release of census information to ANYBODY is a violation of Title 13 and involves 5 years in prison and a 5,000 fine. to wit: See Title 13. http:///www4.law.cornell.edu/uscode/13/ch1schI.html. If all that was used was statistical information derived from Census reports that does not identify any individual, there is no violation of law BUT if Census information which identifies any individual, is voluntarily divulged to ANYONE voluntary compliance with the Census is severely undermined. If what is alleged is true (and do we really expect Ashcroft to investigate?) it is a very serious offense. Posted by: AJ at January 20, 2004 02:17 PM | PERMALINKThanks emptywheel - "CAPPS II, that your favorite D candidate (Clark) lobbied for it." I was going to make the same point. Check out this link... http://www.bop2004.org/bop2004/candidate.aspx?cid=12 I find myself wondering exactly what data was used here. You'd figure, if they're going to somehow profile individuals, they're going on (e.g.) which 'religion' box you checked, your country of origin, and other factors the likes of me is unaware of as having a (perceived) correlation to 'terrorism.' Of course, grosser data can be used. Live in Marin Country? Well, hey, that's where John Walker Lindh's from! Reside within a four-block radius of a mosque in Providence? Welcome to CAPPS-ville ... While I can speculate on that stuff (and not terribly responsibly, although I find the activity curiously engaging), what I'd really like to see from a FOIA inquiry is exactly what kind of data they got. If nothing else, it would tell us a lot about what they *think* they know about correlations between demographic data and 'terrorism'. Posted by: lazyman at January 20, 2004 02:38 PM | PERMALINKI can say that this will definitely make me think twice before responding to the Census next time around, regardless of penalties. Posted by: DavidNYC at January 20, 2004 02:47 PM | PERMALINKWe have seen that these types of systems don't work very well and are just waiting to be abused. Still, they are determined to do it. They say it was stopped, but in reality it just went under the radar. The movie Brazil is looking familier as well. Posted by: cn at January 20, 2004 02:50 PM | PERMALINKThis report is almost certaintly bad reporting by the Washington Times. (The fact that this report appeared there should have been a giveaway!) Executive summary: The NASA report uses **publicly available** Census data, not private data that can track individuals. I went to the relevant page on EPIC's website, and found a reference to this NASA study. The study (which *did* use the improperly released NWA data, by the way) had this to say about Census data: This data set contains the responses from the 1990 decennial Census in the United States. The data has information on both households and individuals. ..... Our data comes from the 5% State public use microdata samples and we used the short variable list [24]. Reference [24] is to the Minnesota Population Center Integrated Public Use Microdata Series. Here's what the Minnesota Population Center has to say about their dataset: Most population data - especially historical census data - have traditionally been available only in aggregated tabular form. The IPUMS is microdata, which means that it provides information about individual persons and households. This makes it possible for researchers to create tabulations tailored to their particular questions. Since the IPUMS includes nearly all the detail originally recorded by the census enumerations, users can construct a great variety of tabulations interrelating any desired set of variables. The flexibility offered by microdata is particularly important for historical research because the aggregate tabulations produced by the Census Bureau are often not comparable across time, and until recently the subject coverage of census publications was limited. Microdata do pose some limitations, however. For the period since 1940 census microdata are subject to strict confidentiality measures that limit their usefulness for some applications. The IPUMS samples for these years include no names, addresses or other potentially identifying information. To further ensure that no individuals can be identified, the Census Bureau limits the detail on place of residence, place of work, very high incomes, and several other variables. Most important, the microdata records for the period since 1940 identify no geographic areas with fewer than 100,000 inhabitants (250,000 in 1960 and 1970). Therefore the IPUMS is inappropriate for research that requires the identification of specific small geographic areas in those census years.(Emphasis added) So as far as I can tell, there is no evidence that the Census
department released inappropriate data to NASA for this project. It may
be that their confidentiality restrictions are insufficient, see the
study's description of outliers in the census dataset. But this dataset
is available to anyone who wants it, including you and me, and has at
least been somewhat trimmed to avoid privacy violations. It wasn't clear to me from the linked article that the government was using personally-identifiable census information at all in that study. About 3 years after each census, the Census Bureau releases a "Public Use Microdata File," with observations on 1 or 5% of households in the country (taken from the 1-in-6 households who get the long form). All names, addresses, social security numbers, etc. are stripped off the records; incomes and house values are topcoded (i.e. if Bill Gates is in the sample, his income will be reported as "Above $200,000" or some such figure); and geographic identification is carefully edited so that no area of smaller than 100,000 people is identifiable. All of this is designed to ensure that there is no way to identify any individual from the data, which are made freely available to researchers. It sounded to me as if this was the data set used in the study mentioned. Does anyone have any reason to think otherwise? I don't really see any privacy problems with such a data set. To the best of my knowledge, no one has successfully broken the confidentiality restrictions, and much of the work that the Census Bureau does after each census is to ensure that this remains true. Of course, I make my living analyzing data sets like this, so I have an interest in defending them. (Before 9-11, I heard claims from people at the Census Bureau that they were under pressure, largely from conservatives, to reduce the amount of information collected and reported, because it was considered an intrusion. George W. Bush, for example, encouraged people not to fill out the long form if they were sent it.) None of this, of course, is intended to defend the CAPPS system, the use of airline passenger records, or the "no fly list." I'm just not sure NASA's use of census information was a similar abuse. Posted by: Jesse R. at January 20, 2004 03:03 PM | PERMALINKAlex R. had the same idea I did, and is a faster typist. Thanks, and sorry for the cross-posting... Posted by: Jesse R. at January 20, 2004 03:04 PM | PERMALINKThe folks over at http://www.dontspyon.us have a lot to say on CAPPS II and census data...70% of the people in the so-called microdata set could be IDed through data mining! Posted by: Ann Onymous at January 20, 2004 03:06 PM | PERMALINKAnn Onymous, I don't want to get in a flame war over this, and I don't know if anything I can say can convince you that the public use data is safe, but I think the statistic you cite is very, very misleading. I clicked through to the paper that that statistic is linked to. The result was that for 70% of the observations in the public use sample, there were no other observations with exactly the same values for each variable. Thus, if I knew what my friend Joe Schmoe answered to each of the census questions, and I knew that he had been randomly selected to get a long form, and I knew that he his long form had been randomly selected for the public use sample, I could find his record in the public use sample. There are two important reasons why this is not a problem. First, notice that in order to find Joe Schmoe's record, I had to know every answer he had provided on the census. Once I know all that, it isn't very useful to find his census record, since I already had all the information it provides. (There is a response to this, that some records are unique on a subset of the census questions; this would allow me to find out the answers to the other questions if I knew Joe's answers to that subset.) The second reason not to worry about this is that there is no way to know whether Joe Schmoe's information was randomly selected into the public use sample in the first place. Because the public use sample contains only 5% of the population, there are nineteen people whose records were not selected. The chance that an individual's response is unique in the population is much, much lower than in the public use sample. So even if I know enough about Joe Schmoe's census responses to match only a single public use record, I have no confidence that this record belongs to Joe. If you want to worry about personally identifiable information, worry about credit reporting bureaus. They have much more info about you than does the Census Bureau, and are much less careful about releasing it. I don't think there's much cause for concern with the census data. Posted by: Jesse R. at January 20, 2004 03:35 PM | PERMALINKThis is the internet at it's worst. The moonie newspaper botches a
story and 25 people start ranting about 1984. Then Alex R. and Jesse R.
actually do some homework, and it turns out that this sinister database
is just the PUMS (public use microdata sample)data, which anyone can buy
from the census bureau on DVD for $70. As Jesse R. explained, PUMS is a
sample of raw census records with name and address stripped off, and
grouped within large geographic areas. There is no realistic privacy
threat from this data. Academics and market researchers have been using
this stuff for years. Whatever the merits of CAPPS II, the census
privacy angle is not an issue. Kevin should correct his original post. 1984 was never meant to be futurist. It was commentary on the 'present' of Orwell's experience... it was a critique of Totalitarianism. He was responding to the excesses of the Communist Bloc. Brave New World, on the other hand, was meant to be predictive, and was frighteningly apt. I always found Orwell the better writer (in terms of the language, the story structure, etc.), but Huxley was a freaking visionary. Unfortunately for us. Oh, and Mike D., I kind of think this is an example of the Internet at it's best. The moonies botch a story, and within a miraculously short period of time, 25 people, after ranting and commenting, come up with very pertinent data that deflates it. Posted by: Christian at January 20, 2004 03:57 PM | PERMALINK"Under the US Constitution, the Census is designed for ONE primary purpose: to assign electoral and congressional representatives according to each state's population. If that goes out because people refuse to cooperate with the Census, Democracy is endangered." Gee. THAT would stop 'em. Ya, sure. Mm-hm. You bet. Posted by: DoublePositive at January 20, 2004 05:43 PM | PERMALINKI'm with Christian: Jesse R. sounds like he actually knows something about Census data. I, on the other hand, was just suspicious of the fact that EPIC didn't mention the use of Census data at all, and also happen to know how to use Google... The great thing about the Internet is that I could spend 15 minutes
with Google and figure out exactly what the Wash Times reporters could
have if they had bothered... I have census IPUMS microdata on my desk computer right now, and I can assure you it is useless for anything except broad demographic stuff. I'm sure that is what they were using it for...among the circles I move in there would have been a big fuss if anyone had wanted to break census confidentiality for "homeland security" reasons...believe me, the Census Bureau would have gone to the mat to prevent it. Individually identifiable ensus microdata would not be that useful for homeland security purposes anyway...the long form is only collected from a small minority of the population and at the individual level one always wonders about its accuracy anyway. If you want to worry about your privacy, worry about your credit records, health records, and how easy it is to track your INTERNET USE... Posted by: MQ at January 20, 2004 07:59 PM | PERMALINKI've read the NASA paper by Bay & Schwavaher linked from the EPIC website. The Washington Times has completely misread the workd(shocking!). The authors' objective is to develop techniques to detect outliers (i.e. points that look very different from the rest of the dataset). To illustate the effectiveness of their method, they used five completely different and separate datasets: (1) color histograms from Corel digital photos (2) airline passenger data (3) KDDCUP computer network TCP binary intrustion data (4) 5% State sample data from the 1990 Census (5) a synthetic data set generated from a 30-dimensional normal distribution. It is standard to use multiple datasets in these exercises because of the No Free Lunch Theorem --- just because a method works well for one problem does not mean that it will work well for others, so you should show the performance in a number of different situations. Nowhere is it implied that any of these five data sets were ever mixed and matched. These were just five separate exercises, with five different conclusions. The Washington Times is being stupid here (shocking!) Posted by: eswn at January 21, 2004 07:42 AM | PERMALINKThis is the most amusing blog I've seen. Thanks. I needed something encouraging after the giant gas-bag attack of last evening. Posted by: kickboygeorge2hellout at January 21, 2004 09:47 AM | PERMALINKExcept the entire PURPOSE of data mining algorithms is to beat the mix-n-match phenomenon by finding telling features and functions of features that are true for class X and not for class Y. So you can't argue both sides here: either the data CANNOT find individual characteristics of someone, and you just have a normal distribution over a sample of 100k people--in which case, data mining for a pattern to detect an outlier is USELESS--or you think that by using various feature detection algorithms that the census didn't think to aggregate over, or by intersecting such aggregation, you can get back to a TINY sample, the sample you want. Which is it? could they or couldn't they detect an outlier? if they couldn't pick out a set of 100 from 100k, then their algorithms are worthless. If they could, well, that's a lot of info about 100. A lot. They are working with a bunch of features here. Enough to distinguish 100 from 100k by some metric! but it's not going to be a perfect classifer. even if it had correct classification 99 percent of the time in both directions, it'll miss 1 in 100 terrrorists (oops) and it'll red flag 999 innocents. but they know an awful lot about those people they think fit in that category. speaking of data aggregation, do you know how few people are in your Zip+4 area? Do you know that level of info is enough, to determine your apartment building? The problem is NOT that an individual s/w data miner learned who you were. The problem is that they are going to write algorithms that are going to select people ---ACTUAL people--because they data they used was from Actual People. They are going to find that there's a correlation between the black guy in that zip+4 and that terrorist--and gee, there aren't all that many black guys in the building, but there might be 2, instead of 1. No? there are no actual people, because they used aggregated data in the first place? Then THEIR ALGORITHM ISN:T GOING TO FIND A THREAT, because they don't have enough information to KNOW anything anyway. There are attempts to "sanitize" data to make it not contain the personal, identifiable characteristics. Except that doesn't work very well. Because the truth is, they want the personal, identifiable characteristics. If they don't, then they are only computing likelihoods about terrorists anyway--and that's useless unless they intend to intensively search every passenger in a very wide sample set. They won't do that. If all they are showing is that they can find outliers in geographic aggregate census data, well then they wasted a whole lot of money on this project. Outlier detection already exists, and there's millions of possible data bases they could have used. If their point was to find outliers to classify possible threats, then their algorithm needs to do better than bring forth a class of people of size 100k. Posted by: Foo at February 16, 2004 02:54 AM | PERMALINK5473 You can buy viagra from this site :http://www.ed.greatnow.com Posted by: Viagra at August 7, 2004 08:28 PM | PERMALINK1669 Why is Texas holdem so darn popular all the sudden? http://www.texas-holdem.greatnow.com Best XXX Sites - 7829 ok you can play online poker at this address : http://www.play-online-poker.greatnow.com Posted by: online poker at August 10, 2004 08:41 AM | PERMALINK6019 ok you can play online poker at this address : http://www.play-online-poker.greatnow.com Posted by: online poker at August 10, 2004 08:46 AM | PERMALINK4578 get cialis online from this site http://www.cialis.owns1.com Posted by: cialis at August 10, 2004 11:52 PM | PERMALINK8251 Keep it up! Try Viagra once and youll see. http://viagra.levitra-i.com Posted by: Viagra at August 14, 2004 11:57 AM | PERMALINK2890 Get your online poker fix at http://www.onlinepoker-dot.com Posted by: poker at August 15, 2004 04:57 PM | PERMALINK5938 black jack is hot hot hot! get your blackjack at http://www.blackjack-dot.com Posted by: play blackjack at August 17, 2004 09:13 AM | PERMALINK2624 black jack is hot hot hot! get your blackjack at http://www.blackjack-dot.com Posted by: blackjack at August 17, 2004 09:20 AM | PERMALINK6593 so theres Krankenversicherung and then there is 6549 Its great to experiance the awesome power of debt consolidation so hury and consolidate debt through http://www.debtconsolidation.greatnow.com pronto Posted by: debt consolidation at August 18, 2004 11:58 PM | PERMALINK4478 http://www.exoticdvds.co.uk for Posted by viagra at 24.08. 2004 11:16:56 for http://www.viaga-viagra.greatnow.com 8056 check out the hot blackjack at http://www.blackjack-p.com here you can play blackjack online all you want! So everyone ~SMURKLE~ Posted by: blackjack at August 24, 2004 04:47 AM | PERMALINK8413 Herie http://blaja.web-cialis.com is online for all your black jack needs. We also have your blackjack needs met as well ;-) Posted by: blackjack at August 24, 2004 08:12 PM | PERMALINK4859 check out http://texhold.levitra-i.com for texas hold em online action boodrow Posted by: texas hold em at August 26, 2004 05:09 AM | PERMALINK |
|
|
Powered by Movable Type 2.63 
|
||||
