Predictive Validity
If you want your test or other research design to predict future outcomes, you need to ensure that its predictive validity is high. Predictive validity is a subtype of criterion validity together with concurrent validity. It is widely used for checking tests and other measurements in education, human resources management, and psychology.
The expected outcomes may include possible ways of performance, behavior patterns, or even a health disorder that may occur in the future due to some conditions.
What Does Predictive Validity Mean?
Predictive validity has to see how the situation will look in the future according to the test taken and its results. This test has to correlate with a specific variable that is unavailable and the moment but can be evaluated only at a certain point of time in the future, for example, when you have placed your students in different-level groups and two weeks have passed to see whether the level they are at works perfectly for them.
If you want to evaluate predictive validity, you need to check how well the test results predict future performances, behaviors, or situations. For instance, ESL students with higher placement test scores can perform better at a higher level. Here, you will get evidence that the outcome predicted by the test design can be available in the future.
The contexts for checking the predicted outcomes can differ and be related to schooling, medicine, or work. The context has to be narrowly defined for the most accurate prediction of individual performance.
For example, a big IT company often experiences bad complaints from its employees about their poor workplace conditions. There are no specific reasons for such complaints, so the company’s owner asks for conducting a specific type of research to understand what is going on.
To comply with these requests, you create an employee work-life balance satisfaction survey to see what hardships the employees experience. You want to obtain the predictive validity of your test, so you ask those employees who have just been hired to answer the questions of the survey.
After a year, you offer them the same or similar test again and observe how their thoughts and experiences have changed. You can also observe how many new-hired employees have stayed with the company. If the correlation between the preliminary and the following tests is high, you can conclude that your test has high predictive validity because it can highlight the work-life balance issues within the company the employees are not satisfied with.
If the correlation is high between the results of the two tests, the situation is not so bad. If the correlation is poor, your test (as well as workplace conditions in the company) needs improvements. In short, the test conducted after a year shows how high the predictive value of your initial test is.
All in all, all the tests that are meant to check job candidates, students, patients, and services need to have predictive value, or they will be useless.
Comparison of Predictive and Concurrent Validity
The main difference between predictive and concurrent validity, which are both subtypes of criterion validity, is the time when the two measures are applied. Though they both are related to specific validation strategies based on the comparison of test scores with the well-established and widely used criterion or “gold standard,” concurrent validity is measured while the criterion variables and test scores are received and assessed simultaneously. As for predictive validity, the creation variables can be obtained only after you receive the scores of the test or any other research measurement.
Nevertheless, the criterion that all the results are compared to (a gold standard) has been accurately measured for the construct you are interested in long beforehand.
Measuring Predictive Validity
To measure predictive validity, you need to compare the test scores against the scores of a well-established measurement called the criterion or “gold standard.”
The measure you want to validate has to correlate with the criterion variable. You can calculate this correlation with the help of a correlation coefficient or Pearson’s r. This coefficient shows the stability of the relations between the two variables. The values of the coefficient can range between +1 and -1. These values can be properly interpreted in the following way:
- r = +1: You have got an ideal positive correlation.
- r = 0: You haven’t got any correlation at all.
- r = -1: You have obtained an ideal negative correlation.
If you are short of time or want to save effort, use online coefficient calculation tools, such as Excel, SPSS, or R. These pieces of statistical software or any other tools will do all the calculations automatically.
If the correlation is absolutely positive, you can be sure that the predictive validity is high. It means that your test can support your hypothesis and predict the outcomes correctly. Nevertheless, even if the correlation is present, you cannot argue that there is the same strong causation.
So, you have developed the work-life balance survey for the IT company. You measure the correlation between the initial survey results and the obtained outcomes after a year. If this correlation is something like r = 0.72, your survey has high predictive validity. It is obviously higher than for the correlation coefficient of r = 0.25 received for other tests.
If the correlation between the test and the criterion is high, the predictive validity of the test is also high. If you do not obtain the correlation at all or it is entirely negative, the predictive validity of your test is poor, and you will need to make some improvements to it.
Final Thoughts
You can see now what predictive validity as a part of criterion validity is. You understand that having measured this type of validity can influence the overall validity of your research method or design. However, predictive validity only cannot indicate any criterion validity if taken separately. It can only be considered while paired with concurrent validity. If both types give high scores, the criterion validity of your test is also high.
You need to know how to calculate the correlation coefficient. This calculation plays a great role in assessing other types of validity too.