What is Predictive Validity?
Mary McMahon
Mary McMahon
Predictive validity is a measurement of how well a test predicts future performance. It is a form of criterion validity, in which how well the test works is established by measuring it against known criteria. In order for a test to have predictive validity, there must be a statistically significant correlation between test scores and the criterion being used to measure the validity.
One of the classic examples of this is college entrance testing. When students apply to colleges, they are usually required to submit test scores from examinations such as the SAT or the ACT. These scores are used as a basis for comparison, with evaluators looking at the performance of students who have had similar tests in the past. The belief is that the test scores can predict how well a student will perform in college. High test scores tend to be correlated with strong college performance, making students with high scores appealing to admissions departments.
The college test scores example is also an excellent example of the weaknesses of predictive validity. Some students who take such tests do not go to college, which means that no data is generated to correlate their test scores and their college performance. This creates a hole in the data set, which can undermine the validity of such tests. Standardized testing has also been accused of some biases that can work against particular students, especially students in racial minorities. They may perform poorly on the test and well in college, skewing the results.
Statistical significance can be challenging to calculate. Huge numbers of factors can influence test results, especially when they involve data from a test and a criterion measure that are collected at different points. Predictive validity influences everything from health insurance rates to college admissions, with people using statistical data to try and predict the future for people based on information which can be gathered about them from testing.
Predictive validity is most commonly used when exploring data in the field of psychological study and analysis. It is used to collect information about various populations, and to create generalizations which may be useful when assessing individuals. For example, it is often used by big companies that administer a test to prospective employees, comparing test data from current employees to determine whether or not someone will be a good fit with the company.
Mary McMahon
Ever since she began contributing to the site several years ago, Mary has embraced the exciting challenge of being a AllTheScience researcher and writer. Mary has a liberal arts degree from Goddard College and spends her free time reading, cooking, and exploring the great outdoors.
Mary McMahon
Ever since she began contributing to the site several years ago, Mary has embraced the exciting challenge of being a AllTheScience researcher and writer. Mary has a liberal arts degree from Goddard College and spends her free time reading, cooking, and exploring the great outdoors.
AS FEATURED ON:
AS FEATURED ON:
-
![SAT and ACT tests used by colleges and universities are an example of predictive validity.]()
By: David GSAT and ACT tests used by colleges and universities are an example of predictive validity. -
![Insurance rates are based on predictive validity.]()
By: emiliezhangInsurance rates are based on predictive validity. -
![IQ testing, commonly used to determine which students should be placed into a gifted program, assumes predictive validity.]()
By: Lisa F. YoungIQ testing, commonly used to determine which students should be placed into a gifted program, assumes predictive validity. -
![Standardized testing has been accused of some biases that can work against particular students, especially those who are in the racial minority.]()
By: AntonioDiazStandardized testing has been accused of some biases that can work against particular students, especially those who are in the racial minority.
Discussion Comments
I remembered a phrase we learned in statistics class: "Correlation doesn't imply causation." I think that's what predictive validity measures. It measures correlation, not causation.
So exam scores are correlated to how the student will do in college. But ability as a student is not caused by the exam scores, there are many different factors out there. Still, this is the best criteria to use for selecting students. There is no better alternative.
The article pointed out something I hadn't realized. This type of validity measurement needs to compare two scores, one before selection and one later. It doesn't make sense to reach conclusions when you are only measuring for people whom you have selected.
I did not do too well on my GRE and was refused to several top graduate schools. I applied again to one of the same schools which had rejected me, this time to a different program and was accepted. I had a 4.0 GPA during my entire graduate education. If the schools which rejected me looked at how successful I was in graduate school, I would completely fall out of the data they have about GRE scores and GPA scores.
I think my personal experience is a good example how predictive validity doesn't really work for everyone.
I do wish that health insurance companies did not rely on predictive validity when providing insurance to certain people.
There was a report about a woman on TV who had a rare genetic disease and how she was not able to get treatment because the insurance companies refused to give her insurance. They were predicting that her health care needs and expenses would be huge because of her disease which would cost the company more than they were willing to pay.
This is also why our car insurance rates go up if we have an accident. Because predictive validity says that we are more likely than a person who has not had any accidents to cost the insurance company.
Post your comments