What is Criterion Validity?
Criterion validity (aka criterion related validity) is the degree to which scores from a construct assessment correlate with a manifestation of that construct in the real world (the criterion).
The construct assessment is a test, measurement instrument, or psychological inventory that assesses a latent construct. Because you can’t observe the construct, it’s impossible to look at the assessment scores and know whether they accurately reflect the intended construct.
The criterion is an observable outcome that theory suggests will relate to the construct. It must be an accepted standard of comparison.
Criterion validity evaluates whether the construct and criterion correlate as theory states. Does the assessment instrument produce construct scores that correlate with measurable, real-world outcomes in a manner consistent with theory? For example, does a psychological inventory for a condition correlate with observed manifestations of that condition?
For example, the following cases exhibit criterion validity if the constructs correlate with the criteria.
- Psychological depression inventory scores correlate with behaviors of depressed people.
- Reading assessment scores correspond to demonstrated reading abilities in the classroom.
- Behavior assessment scores correlate with observed behavior.
- SAT scores predict first year of college cumulate GPAs.
Typically, the designers of an inventory or assessment instrument develop it to provide critical diagnostic information and help decision-makers. If the test does not measure the intended construct (i.e., it is invalid), it can lead people astray with incorrect information. Consequently, assessing criterion validity is crucial.
Criterion validity has three sub-types, which depend on when researchers measure the criterion outcome relative to the assessment.
- Predictive validity: Assessment occurs before the outcome.
- Concurrent validity: Assessment occurs at approximately the same time the researchers measure the outcome.
- Retrospective validity: Assessment occurs in the present while someone else measured the outcome in the past.
Learn more about Validity in Research: Types and Examples.
Evaluating Criterion Validity
A crucial aspect of criterion validity is that an accepted standard of comparison exists. You need a recognized, observable measure that theoretically must correlate with the construct in a known manner. If one doesn’t exist, you can’t assess criterion validity.
If construct scores and observable outcomes correlate correctly, it suggests that the assessment instrument truly measures the construct it was designed to measure. That gets to the core of validity’s definition.
Conversely, if there is no correlation or a correlation with the wrong sign, it suggests that the instrument does not measure the intended construct—it is invalid.
Learn more about Interpreting Correlation Coefficients.
Criterion Validity Example
Suppose researchers develop an assessment that measures aggressive tendencies (the construct) in school children. If this assessment exhibits criterion validity, higher test scores should reflect a higher potential for aggression in the real world.
The researchers assess a sample of students using the instrument. Then they use a carefully designed system to quantify observed aggressive behaviors in a classroom that the project monitors.
To evaluate the criterion validity, they determine whether higher assessment scores correlate positively with aggressive behaviors. If a positive correlation exists, it provides evidence that the assessment measures aggressive tendencies as intended.
Reference
Validity in psychological tests (umn.edu)
Comments and Questions