Validity Testing Explained

When you come up with a survey question, you need to make sure that it is getting you the results that you expect. Writing up a good question is only the first step. Another method of finding out if the survey is getting you the results that you want is by testing for convergent validity and divergent validity.

These can be defined as: Convergent Validity – When two similar questions reveal the same result. Divergent Validity – When two opposite questions reveal opposite results. Testing for this type of validity requires that you essentially ask your sample similar questions that are designed to provide you with expected answers.

If they don’t, the questions might not be valid.

Imagine you are under the assumption that those that would buy your product again are satisfied, as that would rightfully be what is expected. Testing for convergent validity in a survey may look something like this: Question 1: Would you buy product X again if given the chance? Question 2: How satisfied are you with product X? If they say yes to the first question, but they do not score the product very highly in the second question, the question may have failed the validity test.

Divergent validity is designed to see if you get the expected opposite result, because that should also help imply that the question is answering in the way you wanted it to answer. For example: Question 1: Do you wish you did not own product X? Question 2: Would you buy product X again if given the chance? If they answered yes for the first question, and yes for the second question, it would imply that the question was too confusing, because you did not receive the opposite response you expected.

This would be divergent validity.

Most people, when they can, employ both types of validity measures. It does cost money to run the survey enough times to test each question depending on the length of your survey, but validity measures including both convergent and divergent validity are both useful ways to make sure that your questions are asking what you expect them to ask – something that is not always found when pretesting a survey. It is in the best interest of survey researchers to employ various validity measures to make sure the survey is going to gather data effectively.