Codebooks in the 1990s; or, Aren't You Embarrassed to Be Running a Multimedia-Capable, Graphical Environment like Windows and Still Be Limited to 40-Byte Variable Labels?

Abstract

Codebooks are designed to reproduce the context in which data were collected. For survey data, they contain the complete text of questions and responses, along with frequencies of each response. Standard PC statistical packages like SAS, SPSS, and Systat are unable to produce a good codebook; problems include: severe limits on the length of variable and value labels, no ability to store extended text describing variables (such as the text of questions), and no ability to handle multiple response variables, among others. As a result most survey research firms use their own custom-written software to produce a codebook. This paper describes important aspects of a project undertaken for the National Opinion Research Center (NORC) to produce the NORC Automated Codebook. It concludes by considering how the characteristics of an ideal codebook are changing as computing and survey research change in the 1990s. Codebook-style output would be useful for more than just survey research; it is an effective way to describe a variety of scientific, marketing, business, and government data. Keywords statistical software, codebook, frequency tables, variable labels, value labels, SAS, SPSS, Systat.