Another note on the Internet/census issue

More comments on the Census Bureau's Mitre report, which helped justify the bureau's decision not to use the Internet to help with the count.

I got this note:

Some comments:

* Who on the Mitre team that reviewed Census' plans has worked at Google, Yahoo!, Ebay, Amazon, etc.? To what extent were they capable of evaluating the Census plan for the Internet versus a real-world plan to maximize Internet response using the type of approach these companies would use?

* Alternatively, can you imagine one of those companies turning to Mitre to evaluate their Internet plans? This is like asking Jiffy Lube to tune your Ferrari. Nice guys and they change the oil well, but they don't know diddly about high performance.

* Decrease in voluntary response rate: This is the one they will use to bludgeon the congressman with. "If you make me do the Internet and my initial response rates go down (and I'll make sure they do), you'd better be ready to pay BILLIONS (yep, with a B) more for a successful census." Do ya think that Ebay, Amazon and a raft of other Internet companies deal with this one every day?

* From a logic standpoint, if (per their argument) most people won't use the Internet, how does this argument affect paper responses? The only time this argument is valid is if they are getting massive Internet response and a breach occurs. Note that the back-end systems that process the paper responses could be hacked whether or not they take Internet responses.

* Also note that, for the 2000 census Web site, there was an air gap between the systems that took in the response forms and those that processed the data. Accepted forms were immediately encrypted with a one-way key and written to a file. Periodically, the file was written to a tape and hand carried (sneaker-net) to another system, where the files were decrypted with the reciprocal key. Hacking the back-end systems through that channel was a physical impossibility.

* Testing the Internet: Ah, the old "you didn't give me the money I asked for, so I couldn't do this." The most worn excuse in government.

* Moving on to benefits: "When the Internet is offered as an option on the mail-out census questionnaire, no increase in overall response was observed." Well, duh! When I get my paper bank statement, I don't rush to my computer to look at it there. I DO check my statement online constantly, because I think about it when I'm at my computer, and it's much faster than going to dig up a statement (that I can never find). What if Census did a massive campaign on Google, Yahoo! and a few others on the same day they mailed out the paper forms. Even better, what if the Internet wave came first, then a paper mail-out to houses that hadn't responded, and then NRFU?

* Note that Canada got an 18 percent Internet response rate versus the U.S. census at 6 or 7 percent. Do you think they asked, "What are we doing wrong?"

* Cost savings: I can make the numbers say what I want them to say. These numbers say, "If I assume the Internet response will fail and build a paper response big enough to handle the volume when it does, I don't save much money if it succeeds." Another "well, duh!"

* In section 6, the conclusions are clearly the ones Census wanted, not those created by an independent analysis. An independent analysis would have included extensive interviews and expertise from Web sites, as well as the folks in Canada that got 18 percent.