Predictive Validity

Analogy

Imagine you’re going on a road trip to a destination you’ve never visited before. To estimate the travel time, you use your GPS for an estimated time of arrival (ETA) based on current traffic, historical traffic data, and the route you plan to take. The accuracy of the GPS’ ETA in predicting your actual travel time demonstrates its predictive validity.

Diagram

Note that the two diagrams below have different y-axes when comparing. This diagram shows the actual and estimated travel time for 50 trips using a well-reviewed GPS app on a smartphone –

Here we see that the actual and estimated travel time is comparable to each other – suggesting good predictive validity. The next diagram shows the actual and estimated travel time for 50 trips using a 2009 GPS navigation system that hasn’t had an update in a considerable amount of time –

Here we see that the actual and estimated time are not comparable at all. For example, look at the data point with the lowest value on the x-axis – we see a trip that was estimated to take 3 hours but actually took over 4 hours.

Example

Let’s consider a college admissions test. This test can be expected to predict the students’ performance in college (which can be measured by GPA for this example). To establish the predictive validity of the college admissions test, students’ admissions test scores can be compared to (or correlated with) their first year GPA. A strong positive correlation would provide evidence of predictive validity.

Plain Language Definition

Predictive validity is a way of checking how well a measure can predict future results.

Technical Definition

Predictive validity is the ability of a test, measure, or survey to predict a future outcome, behavior, or performance.