Convergent Validity
Table of contents
Research constructs can often be abstract, meaning that you cannot observe or measure them physically. Emotions, attitudes, scientific and psychological concepts, and types of behavior belong to such constructs. It is difficult to establish construct validity in such cases because you cannot use physical tools for its measuring. That is why you need proper operationalization and assessment of convergent and divergent (discriminant) valid.
Convergent validity is a part of construct validity, and it should be measured before discriminant validity to understand the degree of correlation between your test or other measuring instruments with other tests.
Remember!
It is used to ensure that the two tests measure the same or similar constructs. The correlation between them should be from moderate to high. If you obtain a high correlation, it is obvious that the overall construct validity of your research project is high too.
For example, you are planning to use two different research methods to collect data about procrastination in high-school students - observation and a questionnaire. You need to get similar test scores from both of them to identify that they are actually measuring the same construct. If the correlation is high, it means that convergent validity is appropriate.
What Does Convergent Validity Mean?
Convergent validity indicates how much theoretically correlated constructs correlate with each other in practice. It is a subtype of construct validity used to evaluate how perfectly a test or other research tool measures a construct it is meant to measure.
For instance, if you have two tests - for measuring stress and anxiety, these tests are seen as correlated. People who display high scores in stress show the same high scores in anxiety. It means that these two tests have high convergent validity.
The Difference Between Convergent and Discriminant Validity
Discriminant and convergent validity are estimated together to check the construct validity of a research project. If taken separately, the two subtypes of construct validity cannot provide desirable results. However, they are not similar or the same. Convergent validity highlights similarities, while discriminant validity indicates differences.
Therefore, convergent validity shows that two tests that theoretically relate to the same construct are actually related to each other. Discriminant validity demonstrates that two tests that are not meant to relate to the same construct are truly unrelated.
The importance of these types of validity is to show that a new test should correlate with similar tests and constructs but should not correlate with unrelated constructs or tests to avoid bias and unnecessary extraneous variables that could distort the test results.
If you see zero or weak correlation between the results of the tests measuring self-esteem and love for desserts in participants, the test for self-esteem measurement has high discriminant validity. It measures only the construct it was developed for but not other constructs.
When both subtypes of validity - convergent and discriminant - are high, you can ensure that it is evidence of perfect construct validity.
Examples of Two Subtypes of Validity
If you want to check the convergent validity of your new test, you should do the following:
- ✔️ take at least one more test that measures a similar construct and compare the new test scores against the test scores of the similar construct test - you can measure the scores of the procrastination questionnaire against the scores of the same or similar procrastination or laziness test;
- ✔️ choose two different methods for measuring your construct, for example, observation and a survey about self-esteem, and compare the obtained scores.
Suppose you are going to research gaming addiction in senior teenagers and have developed a self-report questionnaire to fill in. You include one more scale or a questionnaire about nightmare experiences in heavy gamers.
Academic research has provided a lot of evidence that people who exhibit gaming addiction often suffer from nightmares. So, you expect that the results of these two scales will converge.
If you obtain the scores of these two questionnaires highly correlated, you can conclude that your gaming addiction test has high convergent validity. On the other hand, if you got the scores from the gaming addiction test that did not correlate with the scores of the nightmare test, you could assume that your gaming addiction test had low convergent validity and should be improved.
You can also pick up two different methods for collecting data on gaming addiction. You can develop a self-report questionnaire and a structured interview. The results obtained from these two designs should correlate because they refer to the same construct of gaming addictions, so the convergent validity should be high. If not, the validity is low, and you will need to look for other research methods.
Remember that you need to check convergent validity before you test discriminant validity. You have to ensure that your test has a high convergent validity and then a low (zero) discriminant validity to evaluate construct validity properly.
Measuring Convergent Validity
The convergent validity of your test should show the positive correlation between different measures or measuring techniques applied to the same construct as well as a strong correlation with the test scores obtained for similar constructs. In short, if you have two similar measurement scales, people who score high on the first scale will score the same high on the other scale.
Important!
The levels of correlation are assessed with the help of a correlation coefficient or Pearson’s r. It ranges between +1 and -1 and shows the direction and strength of the correlation between the two variables.
You can see the following interpretations of these coefficients:
- r = 1: The correlation is absolutely positive.
- r = 0: No correlation is observed.
- r = -1: The correlation is absolutely negative.
Use statistical software online, like Excel, R, or SPSS, to calculate Pearson’s r automatically and save time.
You also need to do relevant literature research to find similar tests that can correlate with your new test for obtaining high convergent validity. You also need to know about other constructs that can relate to the construct of your study.
The r-value above 0.50 can be sufficient for appropriate convergent validity. Simultaneously, correlations between related constructs need to be higher than correlations between unrelated or slightly related constructs.
Suppose you want to research delayed emotional response. It is about the reactions to any stressful or hazardous situation that is not expressed immediately but after some time because of specific blocks in the brain. So you have developed a new test to measure delayed reactions.
Academic literature often describes delayed emotional reactions in connection with severe fear or panic. It means that the construct of delayed response is related to the constructs of fear and panic. So, you can compare the delayed emotional response scale against those of fear and panic predispositions.
You pick out the sample of 55 respondents and offer them three different tests to complete. Then, you calculate the correlation coefficients between the scales related to delayed response, fear, and panic. The results you have obtained show that the delayed emotional response scale correlates with the fear scale as r = 0.38 and with the panic scale as r = 0.66. These results show that the positive correlation between delayed response and panic is stronger than between delayed response and fear. Though, the latter correlation is also positive. These findings correspond to the academic literature, so you have got evidence of how to prove the high convergent validity of your delayed emotional response test.
You can also assume the high construct validity. However, convergent validity is not enough for such an assumption. You also need to check the discriminant validity before making any final statements. In short, you have to show that there is a zero or negative correlation of your delayed response scale with such unrelated constructs as cooking ability or color preference and others.
Final Thoughts
You can see that convergent validity plays an important role in the assessment of the overall construct validity. However, it is essential to check it only in combination with discriminant validity. You should establish convergent validity first before checking divergent validity. You also need to know how to calculate it and compare it to other types of validity. The proper use of various checking techniques and tips will lead to enhancement in the overall construct validity and provide more reliability and consistency to your entire academic work.