Definition• Method <strong>of</strong> data collection or mode <strong>of</strong> administration sometimes refer to• Method for contacting sample member• In-person, telephone, mail, internet• Means by which question is administered to respondent• Interviewer, other aural, text• Medium on which response is recorded• Paper, computer• Method <strong>and</strong> mode are also closely associated with sample framePrincipal modes <strong>of</strong> administering questions <strong>and</strong>recording responsesQuestion<strong>Administration</strong>Interviewer(IAQ)Interviewer(IAQ)Voice <strong>and</strong>textFace-t<strong>of</strong>aceRecordingResponsesPaperPAPITelephone PATIComputer(CAI,CASIC)CAPICATIA-CASIText SAQ CASIP Mixed mode surveys may combine more than one means <strong>of</strong> delivery or morethan one means <strong>of</strong> question administration or responseP Other modes – TDE, direct to disk, web-based surveys, etc. – have beenstudied less.Computer-assisted survey information collection(CASIC) or Computer-assisted interviewing (CAI)• Technologies that are available change rapidly• Task performed by technology include• Manage sample• Present survey questions• Record answers• Edit data• Exchange dataImplications <strong>of</strong> the proliferation <strong>of</strong> methods <strong>and</strong>modes• Need to ask what reference to mode refers to <strong>and</strong> to be explicit in yourown work• Research about method or mode effects• Need to be explicit about what “mode effect” refers to (e.g., frame,method <strong>of</strong> contact, method <strong>of</strong> administration)• Literature does not (yet) cover all variations• Need theory to inform expectations about effects• Mixed-mode surveys or hybrid designs• Increasingly common• Effect <strong>of</strong> mode may depend on the particular combination <strong>of</strong> methodsused• Designs to test effects increasingly difficult to design <strong>and</strong> interpret• Hard to generalize about mode comparisons
Choosing method <strong>of</strong> data collection:Considerations• Survey population: Characteristics, abilities, resources, <strong>and</strong>interests• Sample frame: Information available in sample frame (area,name, address, telephone number)• Sample design: Clustering, stratification, selection <strong>of</strong> respondent• Study topic: Sensitivity, task difficulty, need to use visual aids• Analytic goals: Comparisons needed with studies using othermodes• Interviewer corps: Ability <strong>and</strong> experience• Response rate: Need for high response rate• Administrative infrastructure: Availability <strong>of</strong> technology <strong>and</strong>experience with technology• Cost: Desired length <strong>of</strong> field period• Cost: Other considerationsChoosing method <strong>of</strong> data collection (continued)• Impact <strong>of</strong> mode on error• Sampling error• Coverage error• Nonresponse error• Measurement error• Comments:• Decisions about error require consideration <strong>of</strong> time <strong>and</strong> budgetconstraints <strong>and</strong> trade-<strong>of</strong>fs• There is no one best mode for all situationsChoice <strong>of</strong> mode: Frame• Three basic types <strong>of</strong> frames: area, telephone, <strong>and</strong> list• Area frame: no list <strong>of</strong> addresses or persons in U.S., so usegeographic/political units as frame• Telephone: Most designs now are list-assisted RDD, <strong>and</strong> increasinglyhave dual-frame designs with cell phones• Mail: Addressed-based sampling (ABS) provide frame for mail surveys<strong>of</strong> household• List: lists vary in the amount <strong>and</strong> quality <strong>of</strong> information in the frame<strong>and</strong> what modes they can support• Information in frame <strong>and</strong> survey design• Frames vary in coverage <strong>of</strong> target population• Information in frame determines how sample unit can be contacted<strong>and</strong> how survey instruments can be distributed• If frame includes address, design can include advance letters,prepaid incentives, etc., regardless <strong>of</strong> mode• Some frames provide ability to subsample within sample clusters onthe frame (e.g., within households)• Frames vary in coverage <strong>of</strong> target populationChoice <strong>of</strong> mode: Coverage <strong>and</strong> other problems insampling frames• Typical frames differ in how well they cover target population• Frames may provide information for assessing or determining eligibility<strong>of</strong> units on the frame (e.g., administrative record in frame based onobservation <strong>of</strong> or self-report from sample unit)• All designs that involve selecting respondents share coverage errors dueto measurement errors made by respondent (e.g., for number <strong>of</strong>persons in household, last birthday, etc.)