Computer-Assisted Telephone Interviewing (CATI)

Advantages of Computer-Assisted Telephone Interviewing

CATI provides the following advantages:

• More efficient data collection, because the interviewer enters answers directly into the computer rather than sending a paper questionnaire for a separate data capture step.
• More efficient and more accurate questionnaire administration, because the computer delivers the questions to the interviewer in the correct programmed sequence, including any required rotations, randomizations, or insertions of information from a separate data file or from earlier in the interview.
[Page 123]
• More accurate data collection, because the computer can apply various range and logic edits as the answers are entered. These edits can range from hard edits (in which the system will not accept an answer outside certain parameters—for example, age at first marriage being less than 14 years of age) to “query edits” that require the interviewer to confirm that, while unusual, the answer is indeed that intended by the respondent (e.g. to confirm that age at first marriage was indeed only 14 years of age).

While this has been the basic model for CATI systems since they were first introduced in the 1970s, and some CATI systems still have only this questionnaire administration component, technological developments during the past 30 years have provided many more ways in which the computer can assist the telephone interviewing process.

Quality Assurance Monitoring

For quality assurance, most telephone surveys have a sample of interviews monitored by a supervisor, so the researcher can be confident that the questions have been administered by the interviewer as instructed (correct wording, probing) and the answers given by the respondent faithfully recorded or correctly categorized. Computers allow this to be done in an unobtrusive and effective manner, usually by the supervisor listening in on the interview on a separate audio channel while watching an image of the interviewer's screen.

Further assistance by the computer for this process occurs with the automatic recording of the interviewer the supervisor is monitoring and for what time period. A data entry tool for the supervisor then records the results of the monitoring session and a database in which these results are stored. The use of better allocation of monitoring resources, typically by an algorithm, queries the database, so that more experienced interviewers who rarely have errors are monitored less than those who are newer or who have been identified as needing more assistance.

Sample Management and Call Scheduling

Most CATI programs now have at least two modules, one being the questionnaire administration tool already described, the other providing sample management and call scheduling functions, such as the following:

• Holding the list of all the telephone numbers to be called, along with any other relevant frame information, for example, geographic region if the sample is to be stratified by region
• Recording information about the call history, that is, each call made to each number, such as time and date the call was placed, the interviewer who placed the call, and the call outcome (completed interview, refusal, busy signal, etc.)
• Executing calling rules that determine when the next call (if any) should be placed to a number, which could include delays from the previous call, or certain times of day or parts of week
• Prioritizing among numbers competing for delivery at the same time, for example, by queuing numbers that have appointments first, calls to households where previous contact has occurred next, and fresh sample last
• Delivering phone numbers to the next available interviewer appropriate for that number (e.g. previous refusals to refusal converter interviewers)
• Producing sample progress information, such as number of interviews so far completed by strata, number of interviews refused, and amount of sample yet to be worked

The sample management module often has a separate supervisor interface, which enables the supervisor to execute additional sample management functions, such as stopping particular numbers from being delivered to increasing for a limited period of time the priority of numbers in strata where the survey is lagging.

...