Computer-Assisted Personal Interviewing (CAPI)

Background

To understand the evolution of CAPI it is necessary to understand the history that led to its development and widespread implementation. In the late 1980s, many surveys used early versions of computer-assisted telephone interviewing (CATI). The early CATI systems ran as terminal applications on a mainframe or minicomputer. Computer applications typically used compilers; the central computer had to handle many simultaneous processes to service a CATI research facility. The cost of mainframes and more capable minicomputer systems was so high that the economic case that CATI should replace paper-and-pencil interviewing (PAPI) was tenuous. In addition, CATI facilities tended to use interviewers quite intensively and with close supervision, so interviewers tended to make fewer errors of the sort that computerized systems suppress, at least relative to face-to-face interviewers. With computing costs high, CATI was not a strong value proposition.

As personal computers (PCs) started to penetrate the market, they offered only modest processing power—but CATI interviews did not require much power. An intensively used PC could be cost-effective, and its capabilities matched the CATI task better than a mainframe or minicomputer did. There was no strong need to have a networked solution for PC computing, since CATI facilities could use low-tech case management and scheduling systems and still get the work done.

The PC software solutions for computer-assisted interviewing were adaptations of software first used on minicomputers or mainframes. A boundary constraint was that the compiler needed to have a variant that ran on DOS—the disk operating system for PCs that soon outstripped the use of Apple computers' proprietary operating system. This limited the software options.

By the late 1980s all major survey organizations doing face-to-face interviewing looked to establish a CAPI capability. With limited computing power for laptop computers and the limitations of DOS (which limited executable size because of its address space), these organizations faced a daunting systems challenge. Designers had two major strategic software alternatives. One choice was to follow the existing strand of software development with CATI and program the instrument to run on a laptop, accepting the reductions in memory and processing speed imposed by the technology of the times. The second strategic strand was to represent the instrument not as program code to execute but as a series of data records to be processed one by one. Internal machine instructions became records to be processed in exactly the same way, except that there was no output to the screen. The first application of this second strategy was done by Willem Saris of the Netherlands for smaller, less complex market research and public opinion surveys. In 1989, the Center for Human Resource Research at Ohio State University used a CAPI system based on representing the instrument as data to administer Round 11 of the National Longitudinal Survey of Youth 1979 (NLSY79), a large, complex event history interview that collected socioeconomic data in a one-hour face-to-face interview.

...