Skip to content
Back to search results

Telephone-assisted Survey — CATI Survey

Introduced by
Groves & Kahn (foundational comparative study); CATI systems developed by Charles Cannell and colleagues at University of Michigan
Year
1970s (widespread from mid-1970s; Groves & Kahn 1979 seminal text)
Type
Quantitative / mixed-mode data collection
Family
Data collection
Font size

What does it do?

A telephone-assisted survey is a structured data-collection method in which a trained interviewer administers a standardised questionnaire to respondents over the telephone, often supported by Computer-Assisted Telephone Interviewing (CATI) software.

It combines the efficiency of remote administration with the response-quality advantages of live interviewer guidance, making it widely used in social, public-health, market-research, and political polling contexts.

The intuition behind the method

Think of a telephone-assisted survey as a structured questionnaire interview conducted at a distance: a trained interviewer reads questions from a screen, records answers in real time, and can clarify misunderstandings — without needing to travel to the respondent. CATI software routes the interviewer through skip logic automatically, reducing error. Compared with a self-administered postal or online survey, the live interviewer raises response rates and allows follow-up probing; compared with face-to-face interviews, the telephone mode is far cheaper and reaches geographically dispersed samples quickly.

History

Telephone surveys emerged as a serious academic method in the 1970s when Robert Groves and Robert Kahn published their landmark comparison of telephone and face-to-face interview methods (1979), demonstrating that telephone data quality was comparable for many variables at substantially lower cost. CATI systems were developed concurrently at the University of Michigan Survey Research Center and were commercially widespread by the early 1980s. The method dominated probability-sample survey research through the 1990s. Declining landline penetration and rising non-response rates from the 2000s onward prompted the shift to dual-frame (landline plus mobile) designs and, eventually, the integration of telephone modes into mixed-mode and online panels.

When to use it

Use a telephone-assisted survey when you need a cost-effective way to reach a geographically dispersed probability sample, when your questionnaire is short-to-medium in length (typically under 20 minutes), and when a live interviewer is needed to handle skip logic, clarify questions, or improve response rates among populations unlikely to complete self-administered surveys. It is particularly appropriate for population-health surveillance, political polling, customer satisfaction tracking, and rapid-turnaround policy research. Avoid it when the questionnaire requires visual stimuli, complex scales, or documents the respondent must consult; when the target population has very low telephone ownership or high caller-ID screening rates; or when interview length exceeds 25–30 minutes, as cooperation rates decline sharply for longer calls.

Limitations and considerations

  • Coverage bias: households without a landline or mobile phone, or with unlisted numbers, are systematically excluded from random-digit-dialling frames.
  • Declining response rates in many countries due to increased caller-ID screening, call blocking, and general survey fatigue.
  • Not suitable for surveys requiring visual aids, cards, photos, or physical materials.
  • Interviewer effects — tone, accent, gender — can influence responses, requiring careful training and monitoring.
  • Limited to populations literate enough to process spoken questions and willing to engage by phone.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Reaches geographically dispersed or mobile populations without the travel costs of face-to-face fieldwork.
  • Live interviewer presence raises response rates and allows real-time clarification, reducing item non-response.
  • CATI software automates skip logic, timestamps responses, and flags errors instantly, improving data quality.
  • Faster fieldwork turnaround than postal surveys; results can be available within days for short instruments.
  • Suitable for sensitive topics where anonymity over the phone is perceived as higher than in face-to-face settings.

Applications

  • National and regional health-behaviour surveillance (e.g., Behavioral Risk Factor Surveillance System, BRFSS).
  • Pre-election political polling and public-opinion tracking studies.
  • Customer satisfaction and service-quality measurement in large service organisations.
  • Rapid-turnaround policy evaluations where face-to-face fieldwork is not feasible within the project timeline.
  • Epidemiological follow-up interviews with patient cohorts recruited through clinical records.
  • Labour-market and employment surveys collecting workforce participation data from household samples.

How it works

How it works

1

The researcher designs a standardised questionnaire with fixed response categories and, where applicable, CATI skip-logic programmed into interviewing software. Questions are worded neutrally and pre-tested in a pilot round to check clarity, timing, and interviewer instructions.

2

A probability or purposive sampling frame is constructed from telephone number registries, random-digit dialling (RDD) lists, or existing panels. For RDD, numbers are generated or drawn to cover landline and mobile subscribers in the target population. Coverage bias — households without phones or who refuse to be listed — must be assessed and documented.

3

Selected numbers are called and potential respondents are screened for eligibility according to study criteria (e.g., age, location, experience). Informed consent is obtained verbally before the interview begins. Multiple call-back attempts are planned in advance to reach initially unavailable numbers and reduce non-response bias.

4

The interviewer reads each question exactly as scripted, following CATI-prompted skip patterns. Neutral probing ('Can you tell me more?') is used to clarify ambiguous answers without leading the respondent. Interviewer tone, pace, and adherence to protocol are monitored by supervisors, often in real time through silent call monitoring.

5

Responses are entered directly into the CATI system during the call, eliminating a separate data-entry stage and reducing transcription error. The system timestamps events and flags out-of-range entries immediately, enabling on-the-spot correction.

6

Completed records are exported to statistical software. Non-response weights, post-stratification adjustments, and design-effect corrections are applied as needed before descriptive and inferential analyses are conducted.

Common pitfalls

  • Ignoring mobile-phone coverage: samples drawn only from landline registries systematically miss younger and lower-income respondents; dual-frame or RDD designs covering both landline and mobile numbers are now considered essential.
  • Questionnaire length underestimation: researchers accustomed to self-administered instruments often design calls that run 30–45 minutes, causing high mid-interview drop-out; pilot test for actual call duration and cut ruthlessly.
  • Skipping call-back protocols: contacting each number only once inflates non-response rates and introduces availability bias; AAPOR guidelines recommend at least five to eight contact attempts before coding a number as non-contact.
  • Failing to apply non-response weights: completing calls does not guarantee a representative sample; post-stratification weighting by known population parameters (age, sex, region) is almost always necessary before generalising results.

Frequently asked questions

What is random-digit dialling (RDD) and why is it used?

RDD generates telephone numbers randomly within valid area-code and exchange combinations rather than drawing from a published registry. This ensures that unlisted numbers and newly assigned numbers have a chance of inclusion, producing a more representative frame of all telephone subscribers. The trade-off is a higher proportion of non-residential and disconnected numbers that must be screened out before the interview stage.

How does CATI differ from an ordinary phone interview with a paper questionnaire?

In a CATI system the questionnaire is displayed on the interviewer's screen and the software automatically routes through skip patterns, randomises response-option order where required, enforces range checks, and records answers directly to a database in real time. A paper-based telephone interview requires manual page turning, manual skip logic, and a separate data-entry stage — each introducing additional error. CATI substantially improves data quality and reduces the time from fieldwork to analysis.

What response rate is acceptable for a telephone survey?

There is no single threshold, but AAPOR (American Association for Public Opinion Research) guidelines require transparent reporting of response, cooperation, refusal, and contact rates using standardised formulas. In contemporary conditions, completion rates of 10–30% from RDD frames are common; what matters most is documenting and analysing non-response patterns and applying appropriate weighting, rather than achieving an arbitrary threshold.

When should I choose a telephone survey over an online survey?

Prefer a telephone survey when your target population has low internet access or digital literacy, when you need a probability sample without an existing panel, when the topic benefits from interviewer clarification, or when you need rapid turnaround without the time to recruit an online panel. Choose an online survey when questionnaire length exceeds 20 minutes, when visual stimuli or complex instruments are required, or when cost constraints make interviewer labour prohibitive.

How do I handle language barriers when administering telephone surveys to multilingual populations?

Best practice is to translate and back-translate the instrument, hire bilingual interviewers matched to the language preferences of sampled respondents, and offer a language-selection step at the start of the call. Multilingual CATI scripts can be programmed so the same skip logic and coding apply across language versions, allowing pooled analysis with a language-of-interview control variable.

Related methods

Sources

  1. Groves, R. M., & Kahn, R. L. (1979). Surveys by telephone: A national comparison with personal interviews. Academic Press.

    Open publication reference
    Copied.
  2. Lavrakas, P. J. (1993). Telephone survey methods: Sampling, selection, and supervision (2nd ed.). Sage.

    ISBN: 978-0803950795
    Copied.