I would be tempted first to modify PGStats approach to use similar coding to add a flag variable or two and then look at results for other questions and see what the responses of these questionable participants look like compared with the rest. Depending on collection method you might be seeing something like "I don't want to answer this particular question but didn't give me that as a choice". There may also be a technological reason, depending on how the survey was administered, where the first response was propogated to other questions when the respondent skipped the remainder. Also a third thing if you didn't have control on the data entry/collection is a possible change in the collection insturment. ... View more