When the fieldwork is completed, researchers have a great deal of data but little or no information. Researchers may have 200 completed questionnaires; but until they have been tabulated and analyzed, they represent only raw data. What is needed to transform these data into information is a procedure for organizing and compiling the bits of data contained in each of the 200 questionnaires.
Much of the data tabulation activity consists of counting the number of responses to a specific category of a specific question – for example, how many households are heavy consumers of yogurt? On the surface this may seem to be a relatively simple task. Yet, when researchers go through 200 questionnaires they may encounter one in which the recorded responses are not perfectly clear. As a result, tabulations made by researchers A could be different from those made by researchers B. Such “counting” would not be accurate or reliable. Tabulating is more than just counting.
If accurate and reliable information is to be obtained, it is necessary to establish a set of procedures to use when preparing raw data for tabulation. These include making a preliminary check of all the completed questionnaires, editing individual questionnaires establishing categories into which different responses ca be classified and coding responses.
Making a Preliminary check improves Data quality:
After receiving all questionnaires from the field, a preliminary check is made before they are subjected to the detailed editing and coding work. Since the questionnaires should have been checked as part of the procedures for controlling the field force, few problems should be encountered. Nevertheless, a preliminary check for five possible problems should be made to ensure that the data on the completed questionnaires are of high quality.
Adherence to sampling instructions: If the interview was not made with the proper respondent, it should be rejected. For example if the sampling universe consisted of home owners, interviews with renters would not be acceptable. To the extent that they can, editors should make sure that the sampling requirements have been met. In random sampling where interviewers identify the households to be sampled, it is impossible to determine if the “rights” households were selected. But if households were selected from prepared lists editors can verify if the proper households were interviewed.
Legibility: Editors should review the responses to all questions on each questionnaire to assure the legibility of the responses.
Understandability: Answers to open ended questions are often difficult to interpret. The interviewer may have abbreviated the answer to such an extent that it is not clear what the respondent meant. For example, often it is impossible to know to what such words “this” or “it” refers.
Completeness: All questions are expected be answered since “blanks” can mean different things; (1) no answer or refusal; (2) the question was not applicable and therefore, was not asked or (3) the interviewer failed to record the answer. The only acceptable reason for a question’s being left blank is (2).
Consistency: Each questionnaire is examined to determine if it is internally consistent. An example of inconsistency would be on a travel questionnaire, where the respondent reports not using a car and later, in answer to another question mentions driving to a particular site.
When the preliminary check uncovers one of these problems, editors must choose one of four possible courses of action: (1) If time permits the questionnaire may be returned to the interviewer for “translation” or “interpret” what is recoded; (3) The applicable question can be eliminated from the tabulation; or (4) The entire questionnaire can be rejected. Because of options (1) and (2) a preliminary check often makes it possible to obtain missing data or to clear up other difficulties while the field force is still intact and the survey still fresh in the interviewers minds.–