340 likes | 483 Views
Kenneth E. Wallen a,b,c , Adam C. Landon a,b , Gerard T. Kyle a,b,c , Michael A. Schuett a,c , Jeremy Leitz d , & Ken Kurzawski d. Sampling Efficacy & Bias in Mode of Response for Survey-Based Research.
E N D
Kenneth E. Wallena,b,c, Adam C. Landona,b, Gerard T. Kylea,b,c, Michael A. Schuetta,c, Jeremy Leitzd, & Ken Kurzawskid Sampling Efficacy & Bias in Mode of Response for Survey-Based Research a Department of Recreation, Park and Tourism Sciences , Texas A&M University b Human Dimensions of Natural Resources Lab , Texas A&M University c Applied Biodiversity Science NSF-IGERT Program , Texas A&M University d Texas Parks and Wildlife Department
Background ‘’89 ‘90 ‘93 ‘97 ‘01 ‘04 ‘08 ‘12
Response Rates • Why do we care about response rates? • Need (desire) to reduce survey error… the difference between the estimate using the collected data and the true value of the variables in the population • Non-response error… emerges when those who do not respond differ from those who respond • Mail questionnaires are costly… printing, postage & labor
Trends in Response Rates • All modes of collection area are on the decline • Mixed-mode designs (mail/web) have demonstrated success for reducing non-response error • Respondents to mail and web-based collection modes are different demographically and with their use of technology • Mixed-mode benefit is twofold… opportunity in preferred manner & remind respondent of the opportunity to respond
Texas Recreational Anglers • Heterogeneous • Demographics • Motivations • Preferred resources • Target species • Avidity • Expenditures • Imperative to reduce non-response error to obtain data that is reflective of the angling population to make informed decisions that will impact this population
Study Purpose • Compare three modes of questionnaire administration in terms of: • Response rate • Respondent characteristics
Survey Methodology • Data Collection • Sample drawn randomly from TPWD database of licensed fresh/saltwater anglers • Three modes of contact commencing the last week of September, 2012 • Mixed-mode (n=4,000) • Email-Only (n=4,000) • Combination (n=1,000) • Web-based questionnaire was designed/administered thru Qualtrix with the URL: WWW.TPWD-SURVEY.ORG
Mixed-Mode • Initial contact: Invitation letter with web push • One week follow-up: Thank you/reminder postcard with web push • Two week follow-up: Second contact letter to non-respondents with web push • Three week follow-up: Survey packet to non-respondents… cover letter (with web push), questionnaire, and postage-paid reply envelope
Email-Only • Four email invitations, one week apart • Similar to the cover letter sent to the Mixed-Mode group, the email invitations outlined the study purpose and invited respondents to complete the questionnaire online • Both a URL to be entered into respondents’ browser and hyperlinked icon (“Take Survey”) were provided for respondents to access the questionnaire
Combination • A combination of contacts replicating the “Mixed Mode” and “Email Only” methods • Eight invitations (4 hardcopy, 4 email invitations) were sent to respondents, one week apart • The sending of email and postal invitations were synchronized to arrive simultaneously; i.e., email invitations were sent approximately two days following the mailing of hard copies
Survey Response • Effective Response Rates Mixed-mode 20.0% (697/3486) Email-only 29.9% (784/2685) Combination 63.4% (407/640)
Survey Response • 53.8% of Mixed-mode respondents completed hard copies of the questionnaire • 29.5% of Combination respondents completed hard copies
Survey Analysis • Multinomial Logistic Regression Model (logit) • Dependent variable: survey response mode • Null-model Approach • Independent Variables • Socio-demographics, motivations, & avidity • Reference group: Mixed-mode Ø = Mixed-mode 1 = Combination 2 = Email only Age + gender + income + race + ethnicity + motivation + avidity
Motives Notes. 1Mean score value is on a scale ranging from 1 (not important) to 5 (extremely important). Like superscripts indicate significant differences at p ≤ 0.01 (Scheffe post-hoc analysis). ** p-value < 0.01 ***p-value < 0.001
Discussion – Response Rates • Combination (surface mail & email) yielded strongest response rate • Lowest non-response error? • Respondents (or non-respondents) are reluctant to go from the paper invitation to their computer or device
Discussion - Socio-demographics • Socio-demographic variation • Variations in age • Web-based respondents slightly higher household incomes • Men more inclined than women to complete online
Discussion – Motivation & Avidity Motivation • On items where there was significant variation web-based respondents considered these facets more important • Avidity • Some indication that hard copy respondents most avid
Next Steps • 2015 Survey of licensed Texas Anglers • Mixed-Mode – Mail survey packet with a web push • Incentive – “lifetime license”
Acknowledgements • Texas Parks and Wildlife Department • Robin Riechers • John Taylor