Predictive Validity

There are several different types of validity, all of which are used to establish a test's “truthfulness,” or whether the test does what it is supposed to do (the main claim of validity). Among the many different kinds of criterion validity, a subset is concurrent and predictive validity.

Predictive validity allows the measurement specialist to judge how well a test predicts future performance. The interest is in whether the instrument can predict what it theoretically should be able to predict.

For example, the following outcomes would characterize test and measurement tools that have predictive validity.

• A math test predicts future performance in an occupation where sophisticated mathematical skills are essential, such as engineering or computer science.
• A test of spatial skills predicts future performance in an occupation where the ability to visualize in more than two dimensions is essential, such as mechanically oriented tasks and professions.
• A test of personality predicts the configuration of personal skills that might be needed to succeed in a particular profession, such as hotel management.

What these three hypothetical statements have in common is that each of the tests that is developed is predictive of a later outcome (which is the criterion) and hence, if they do their job, they have predictive validity.

Predictive validity is most often established through the simple Pearson product-moment correlation [Page 782] computed between the test itself and the criterion. To then establish the predictive validity of the mathematics test mentioned above, test scores would be collected and then correlated with some criterion used to evaluate engineering skills a certain number of years later. That simple zero-order correlation indicates the strength of the association between the test and the criterion and, hence, the predictive validity of the test.