7. APS Data Review

Although the National Teams and Survey Vendors in each country are among the best available, all submitted data is reviewed and tested before being approved for inclusion in the master GEM database. Some of the aspects that are examined for all submitted data include:

Missing or refused questions

Respondents are occasionally not asked all the necessary questions in the questionnaire—often because of a misunderstanding of the survey skip pattern.  Other times, a respondent may be asked extra questions, which they should not be asked according to skip pattern instructions, causing the respondent confusion and survey fatigue. This may result in refusal to answer specific questions or to complete the survey altogether. All responses are examined for skip logic errors and excessive missing values.

Performing the process of random assignment incorrectly

In past years, but no longer included as part of the GEM Adult Population Survey, was a set of questions that were to be randomly assigned to one-half of all respondents. For these years, all national data sets were examined to see if this random assignment process was properly implemented. The current version of the survey no longer includes any random assignment of questions, so this test is no longer needed.

Incomplete interviews

Not all individuals contacted are eligible for the GEM APS and not all those eligible who start the survey complete it. Most GEM countries had a very low percentage of incomplete surveys.

A high refusal rate

A high refusal rate increases the respondent bias, and therefore reduces the likelihood that the survey reflects the true experience of the population. The overall refusal rate for each national data set is examined and compared to prior years, other nations, and other surveys being conducted using similar techniques.

A gender and or age ratio imbalance

The overall age and gender distribution for each national data set is compared with that that would have been expected from a completely unbiased sample. Any deviations are noted, and, if necessary, the National Team and/or their survey vendor are asked to respond.

Poor translations of open-ended responses

Responses to questions about business type must be recorded verbatim and, if not in English, translated in-full in the APS dataset. Some teams may provide shortened, unclear records of these responses, making them more difficult to code accurately and precisely.

Teams are encouraged to liaise with their Survey Vendor if they expect to, or have encountered any of the above problems in the past. The GEM APS Data Team is available to answer any queries teams may have in this respect, and can often also put teams in contact with others who have also experienced but overcome these problems. The above problems are common however, and are often picked up during the harmonization process.


Loading...