Accurate and good quality data were set from the start as a primary objective of the survey. This goal could only be achieved if the collected data could efficiently and effectively be checked. For this purpose, a multilevel checking system of questionnaires was set up immediately after data collection commenced.
1. During the first week of data collection, area supervisors were instructed to collect and check all the questionnaires completed by every interviewer on a daily basis. This action was considered particularly important in the sense that any errors made would be identified at the earliest stage and brought to the attention of the interviewers so that the degree of repetition of such errors would be minimised considerably during the rest of data collection period.
The checks made by area supervisors were of twofold nature:
- First, they were obligated to check the information contained in each questionnaire for errors or inaccuracies. These checks included completeness, validations in relation to information available in the register and between related variables, acceptable ranges, summations, ratios and rational checks.
- Second, they were instructed to make coverage checks. They were asked to come in contact with all of the respondents for whom a questionnaire was completed (either through telephone or personal visit) and to ensure that the questionnaire was filled out according to instructions, that is, during the personal visit of the interviewer to the respondent and that all questions were asked. This checking method was implemented from the very first week of data collection and this proved useful because enumerators knew that all their work was thoroughly checked and that there was no room for shallow approaches.
From the second week onwards, area supervisors were obligated to collect the completed questionnaires from every interviewer and were instructed to carry out numerical completeness and validation checks to all questionnaires, whereas they were instructed to carry out analytical, weekly coverage checks on a sample of at least 10% of the collected questionnaires. They should also ensure that this sample contained at least one questionnaire for each interviewer so that the work of everyone was checked on a weekly basis. Taking into account that every area supervisor was responsible for about 7 interviewers, each of whom completed 20 questionnaires per week, this checking process implied that area supervisors had to check analytically at least 20 questionnaires per week. In cases where mistakes were found repeatedly on the questionnaires of any specific enumerator, the checking unit was asked to raise the percentage of questionnaires checked with respect to the work of that enumerator. Similarly, the percentage of questionnaires checked was raised in the case that mistakes were frequently found in the questionnaires of any particular area supervisor.
2. District officers were requested to set up checking units comprising one or two casual employees working under their guidance and supervision. These units had the obligation to come in contact with those people who reported not having an agricultural holding in order to ensure that the interviewer had actually visited these people and the declaration of not having an agricultural holding is correct. It is noted that interviewers were instructed to report those cases where they visited people who reported that they were not holders of any agricultural holding, providing the name and address of these people.
3. The central office of CYSTAT gave clear instructions to permanent employees to check at least 5% of the questionnaires submitted by each area supervisor. They were also instructed to ensure that the sample of questionnaires chosen for checking purposes should include at least some cases of those questionnaires that were analytically checked at an earlier stage. This process ensured that not only the work of interviewers was being checked but also the work of area supervisors. This level of checking was also done on a weekly basis in order to identify errors and weaknesses as early as possible and to take immediate action for correction purposes and for avoiding repetition and accumulation of mistakes.
4. Finally, checks were carried out during data entry by the software program itself. The program carried out several checks such as consistency checks, valid value and range checks, arithmetic checks etc.
Checks on the accuracy of the data and on coverage were concurrently done.
Tools used for data validation
The software program used during data entry (see point 4 above)
Level of data validation
The first step of this checking process was in the hands of area supervisors (see point 1 above).
The second step of the checking process was carried out by district officers (see point 2 above).
A third level of checking was carried out at the central office of CYSTAT (see point 3 above).