Respondent Acceptance of Web and Data Reporting for an Establishment Survey Richard Rosen, Louis Harrell, Hong Yu Division of Current Employment Statistics Bureau of Labor Statistics Third International Conference on Establishment Surveys Montreal, Quebec, Canada June 18-21, 2007
Current Employment Statistics (CES) Monthly survey of employment, payroll, and hours A Federal-State cooperative system A sample of 300,000 business establishments CES data are published after only days of collection Provides key economic indicators Employment by industry, state, and area Average hourly earnings Average weekly hours
Goals/Objectives Explore different methods of distributing and collecting CES reports Utilize existing technology Minimize respondent burden
Software Used off-the-shelf software SurveyTracker from Training Technologies Inc. The software package costs around $3800
Phases of collection Form design Distribution Respondent’s box Retrieval Export
Survey Tracker
Design
CES Form Design
Distribution
Respondent’s Box
Confirmation Screen
Response
Retrieval
Export The software exports by comma separated value format Read into SAS program and export into the format of choice
Advantages Ubiquitousness of The prevalence of in everyday life is such that almost anybody with a phone line and a minimal cost to sign up for internet service allow individuals to get an address Using off-the-shelf software By utilizing off-the-shelf software to design and distribute the software, it saves the effort of re-inventing the wheel Low marginal cost Sending s costs close to nothing Utilization of pre-existing infrastructure Utilization of pre-existing server as well as web server saves cost and time
Disadvantages No data edits Due to the security issues, no data edits will be performed by the SPAM blockers ISPs with aggressive SPAM blockers may flag the survey as SPAM Ill-formed data No easy way to account for respondent mistakes Reliance on software vendor Necessary to rely on the software vendor initiative to address the problems
Current Status Incorporated as one of the CES collection methods starting February 2007 Comparable to pre-existing collection methods Survey Tracker deemed to be not suitable, looking into another software package Inability to handle large database Compatibility issues Minimum requirement: Outlook and Internet Explorer Not all respondents who meet the minimum requirement can respond
Collection Rate Comparison AllTDEWeb February73% 66%74% March68%73%64%67% April69% *60%70% *CES experienced web server problems during April 2007
Collection Data Item Response Rate WWAEPRAEHRGMEPWPRHR 90%76%70%74%80%60%59%
CES and Web CES has been collecting data via Internet since 1996 (first Federal Agency) BLS Developed an Enterprise level Internet Facility to promote: Security and continuity of operations Uniform standards for Web pages Cross-survey participation
BLS Model Initial BLS Model Included: Digital Certificates Option to use an account number and password Password requirements are stringent Min. 8 chars. Including letters, numbers, upper/lower case CES has experienced declining Web response, looking for something new …
TDE vs Web Response Rates
Issues with Current Web Site Transition to Web was complicated Temporary Account/password Permanent account/password or digital certificate Ongoing Reporting Forgot account/password Non-renewal of certificate/not portable Contact change/turnover Lose continuity of knowledge on reporting Need to re-issue account/certificate
Web-Lite Features No temporary account/password No permanent account/password No digital certificate Access gained through CES Report ID already on the CES form More-”TDE Like”
Web-Lite Design Compromises No respondent identifying information displayed Cannot update contact information on-line No historical data shown Reduces visual continuity of reporting Limits edit capability Cannot directly link to other BLS Web surveys
Security Site remains secure: SSL 128 bit encryption Data stored on separate server Firewall protects other BLS resources Use of CAPTCHA
What is CAPTCHA? Completely Automated Public Turing test to tell Computers and Humans Apart. A trademarked acronym owned by Carnegie Mellon University. Used to determine whether a user is human or a robot. First developed in 1997, by Andrei Broder at Altavista.
Comparison of TDE and Web Response Rates, January 2004-March 2007
Sources of New Units Converted to Web Lite
Response Rate Comparison for TDE, Old Web, and Web Lite
Data Item Response Rates, Selected Items, March 2007 Women Workers Gross Month Earnings Total% Received DCC to Lite4,61152%45% State to Lite5153%27% TDE to Lite2,33162%47% Old Web to Lite34945%28% Old Web1,76044%25%
Conclusions If technical issues can be overcome, appears to be a viable collection method. Response rates are favorable Offers visual interface of web collection Research will continue on possible candidates for data collection software. Web offers on-line editing, and recent months had performance superior to TDE Web Lite offers improved item response rates