RTI International is a trade name of Research Triangle Institute th Street, NW ■ Suite 750 ■ Washington, DC, USA Phone A Revolution in Survey Research Karol Krotki Washington Statistical Society and DC-AAPOR Washington, DC 20 May, 2009
2 Challenges – Cell Phones Close to 20% household are cell-phone-only (CPO) Mapping no longer 1-1 One individual can own several numbers Several individuals can share the same number More difficult to link number to geographical location Data collection Safety Compensation Separate frame
3 Challenges – Response Rates OMB used to insist on 80% Now 60% is difficult to achieve Decrease in response rates results in increase in survey costs Decrease in response rates leads in increased risk of nonresponse bias
4 Challenges – New Technologies Traditional approaches Landline phone endangered species In-person too expensive Internet/Web Smart phones Other
5 Future Directions – Multiple Frames Cell phone and landline phone Address lists (see below) and phone lists Internet data bases Commercial lists Administrative data/large data bases External sources of information Disclosure control/confidentiality Data linkage
6 Future Directions – Address-Based Sample Complete list of addresses Process Select sample of addresses Match as many addresses with telephone numbers as possible Use multiple modes (mail, telephone, in-person, internet) to collect data Nielsen Moved from telephone to ABS last November “RDD is dead”
7 Future Directions – Internet Surveys Coverage and selection bias Harris – large base Frame bias Membership bias “Professional respondents” Is it possible to model to overcome shortcomings? Knowledge Networks
8 Future Directions – Multi-Mode Offer respondents a choice Increase cooperation Select optimal mode as a function of: survey topic target population(s) However, beware of mode effects Adjust mode during field period
9 Future Directions – GIS Frame building Sample selection Screening Data collection
10 Future Directions - GIS
11 Future Directions – Rolling Samples Increase sample size by combining samples over time E.g., American Community Survey (ACS) Sampling and sampling errors more complex
12 Future Directions – Nonresponse Bias Gauge survey quality using nonresponse bias rather than response rates Correlation need not be high Challenges How to measure bias? How to standardize measurement process?
13 Future Directions – Responsive Design Adaptive/responsive surveys Requires multi-mode, flexible approach Requires monitoring during data collection Adjust strategies “mid-stream” Cumulating knowledge Sampling Interviewing
14 Conclusions Surveys of the future will be more complex Need to be flexible Need to be creative Future looks very interesting, albeit not very clear