US Census is back on paper

Back on April 3rd, U.S. Secretary of Commerce Carlos M. Gutierrez cancelled the Field Data Collection Automation (FDCA) program that would have conducted the 2010 Census on wireless handheld devices. The main issue seems to be larger than expected budgets, both for the FDCA project and for the overall census. In March, Engadget reported various problems that were encountered during the 2007 test of the PDA’s. Evidently the handheld systems was difficult to use and had problems transmitting the necessary data during the test.

According to this Computer World interview, the survey was going to be a “short-form questionnaire, only seven questions.” I’m guessing the questionnaire is the “Non-Reponse Follow-Up (NRFU) used to used to collect data from nonrepondents to the mailed questionnaires” which is now being moved to paper according to the U.S. Census Press Release. According to the press release the handhelds will still be used for “Address canvassing (AdCan) the compiling of household addresses in preparation for the Census mailout.”

The hardware is a customized version of High Tech Computer Corp’s handhelds running Windows Mobile 5.0. The device does look beautiful and evidently even includes fingerprint authentication.

Office-Based Patient Interview

Data Collection Outcomes Comparing Paper Forms With PDA Forms in an Office-Based Patient Survey
by James M. Galliher, Thomas V. Stewart, Paramod K. Pathak, James J. Werner, L. Miriam Dickinson, and John M. Hickner
Annals of Family Medicine, March 2008; 6: 154 - 160.

In this project, a nurse or medical assistant from each of 22 medical practices interviewed 60 patients, 30 with a paper form and 30 with a PDA. Data was collected between August 2002 and September 2003 and interviewer training was conducted with individual phone meetings. The project used Sony Clie PDA’s running Palm OS and used Pendragon Forms to program the survey. It appears the PDA survey used skip patterns, had some manadatory responses and did data validation on the responses.

The authors reported that the technical difficulties with using PDA’s for this application made it so that they would prefer to use tablet computers or mobile phones in the future.

The original plan was to transmit the data as it was collected from the medical practices to the Pendragon Secure Server. Unfortunately the communication was blocked by firewalls and the PDA’s were instead sent by mail to complete the data collection. It appears that the secure comminication provided by Pendragon Software is a proprietary format over TCP/IP.

The return rate was better for paper that for the PDA’s (94% vs. 82%), but this doesn’t take into staffing issues and thefts. Of the 570 expect returned forms the the response rate was 537 for paper and 466 for PDA’s. Two of the PDA’s were stolen and for both paper and PDA there were some staffing issues that effected the return rate. It appears that for the 15 of 19 sites that didn’t have stolen PDA’s or staffing issues, that there were still more returned paper forms than electronic forms (446 vs. 430) which is only a 3.7% difference.

The error rate was better for PDA’s than for paper (3% vs. 35%). This difference was mostly because of the significantly lower rate of omissions on the PDA because of the designated mandatory responses.

This study didn’t compare the time involved in the administration of the surveys by the research team, but they did report that the staff spent many hours navigating computer security issues and with assisting with the setup of the PDA’s.

Tri-modal Data Collection for Singapore Census

General Household Survey 2005: Innovations in Fieldwork Operations
by Mr Edmond Lee and Mr Tan Kian Heng
Social & Population Statistics Division
Singapore Department of Statistics
Statistics Singapore Newsletter, September 2006, 8-10.

For the large-scale 2005 General Household Survey, the Singapore Department of Statistics used an innovative approach which relied on: Internet enumeration, Computer Assisted Telepone Interviewing (CATI) as well fieldwork. Data was collected from 90,000 dwelling units over 12 - 15 weeks. Household chose to submit surveys via the internet or by telephone interview. If the survey was not complete by the due date, field interviews visited the households. For even more efficiency, if a household completed the survey after the due date, an automated systems sent an SMS message to the appropriate field interviewer.

The mobile surveys included branching as well as data consistency and security support. The article doesn’t specify what kind of software was used for programming the mobile survey, but it seems most likely that the software was custom developed for internal use.