according to the AP via wired, BUCEN is having trouble rolling out the house-to-house portion of the next decennial census. apparently the handhelds are too complicated for the average census-taker.

given that this is the government - and that the census won’t be taken for another two years - i’m willing to bet gold to little green martians that both the hardware and the software they’re using is already well into obsolescence and will be obsolete in 2010. however, i do think it makes sense to move census-taking into the digital age, even if it means losing the physical link to our past created by those paper records.

with that in mind, i’m offering BUCEN my consulting services, free of charge. (they should really take me up on the offer - before earning my law degree, my consulting rate was in the $100/hr neighborhood. now, well, heh.) here’s what I think they should do:

1. use iphones instead of those handheld computers. with the release of the SDK and firmware 2.0 coming out in june, it should not be too difficult to write an app to simplify the interview process. the iphones are smaller, most likely cheaper, and while the keypad interface isn’t as easy to use as a regular keyboard, the census isn’t exactly full of open-ended questions. the iphone also has a built-in camera, which brings me to the next item, data collection.

 

2. use a web-based survey instrument. this will guarantee everyone is running the exact same instrument. it will also allow the collected data to be automatically stored on centralized servers, eliminating the risk of lost data due to handheld hard drive malfunction or that ever-increasing source of government data loss, mobile computer theft.
    

   the instrument itself could make use of the iphone’s camera, using it as an optical scanner to input a unique identifier for each household. (bruji’s bookpedia already uses the isight camera this way, to facilitate data entry in their consumer-grade library database software.) the census-takers would be issued an ID card with a scannable code to be entered at the beginning of each session (or at the end of each questionnaire when the data are transmitted, as an additional security step), as well as sheets of bar codes, one for each questionnaire. using bar codes instead of numerical sequences reduces data entry error, and linking the bar codes issued to the census taker’s ID provides an added data verification check, allowing data to be analyzed for user error or data patterns that would suggest user fraud. (i’ve managed face-to-face interviewers. it happens.)

   all this brings me to the next point, transmitting the data.

3. use the white spaces spectrum. apparently there’s a debate raging on whether the white spaces spectrum should be opened up for wireless broadband access. due to conflicting reports, the feds don’t know if such use would interfere with broadcast television signals. the census provides an excellent opportunity for a controlled, nationwide impact study, while at the same time providing better data transmission rates than current cellular phone systems allow.

 <>

see that? easy-peasy. and in the spirit of full disclosure, this entire post was written using my iphone, although it may be edited on my laptop later today.