What is External Validity?
External validity asks: are the results generalizable beyond the scope of the study? External validity is about the degree to which conclusions about a group of people in a study would be the same for other persons in the population who are in other places at other times.
Difference Between Internal Validity and External Validity
In research, there are main types of validity: internal validity and external validity. Internal validity is the measure of how well a study is structured and conducted, and how the results reflect the group that is being studied. Internal validity controls extraneous variables, eliminates alternative explanations, focuses on accuracy, strong research methods, and conclusions that are warranted.
When there is external validity in research, the following principles apply:
- the outcomes apply to practical situations,
- findings can be generalized,
- results apply to the world at large, and
- results can be translated into another context.
Thus, the study’s results are applicable to different contexts, persons, situations, periods, or events.
The results provide an estimate of the exact truth about the population. For example, if you conducted research to obtain the voting preferences of Americans, you could collect data from about 500-600 Americans who were randomly selected. According to the results, you would be able to generalize and infer that the participants’ voting preferences are similar to those of Americans in general.
Types of External Validity
The three broad types of external validity are population, temporal, and ecological.
Population Validity
Population validity is how well the experimental sample in your research represents and reflects a target population. If you do not choose a representative population, you will not be able to generalize the study’s population to the target population because they will not be similar.
Temporal Validity
Temporal validity relates to how well your research findings are generalizable to different periods of time other than the specific time that your research was conducted. The research that has temporal validity are those that are recent or those that involve something that has not changed since the study’s completion.
Ecological Validity
Ecological validity is the similarity between your research’s setting and the target real-world setting to which you want to generalize. To have ecological validity, factors that can influence the research’s outcome variable, such as the materials, methods, and environment, need to approximate the target setting.
Threats to External Validity
Threats to external validity are any factors between experimental conditions and the real-world, which reduce your results’ generality.
- Sampling Bias: Sampling bias occurs when there are units from the target population that were not selected appropriately so that the sample is not representative and generalizable to the population.
- Hawthorne Effect: Hawthorne effect occurs when participants are inclined to work harder and perform better because they are part of an experimental study. When they do this, it changes their behavior, which causes the results of the study to be less generalizable to real-world settings.
- Aptitude-Treatment Interaction: Aptitude-treatment interaction is the concept that the study’s treatments or more or less effective for different people because of their specific abilities. In experimental studies, this means that the optimal learning results can be achieved only when the instruction chosen for the experiment matches the aptitude of each learner.
- Situation Effect: The situation effect relates to the varied situational factors that can threaten the external validity, such as the researcher’s characteristics, the setting, the location, or the time of day.
How to Address Threats and Increase External Validity
The following are some ways to address threats and increase external validity. They do not all pertain to one type of research but may be dependent on the type of research that is being conducted and its parameters and design.
Avoid Sampling Bias
If the participants volunteer or self-select or volunteer to participate in the research, a result that does not represent the target population may occur. To control this threat, you can obtain a representative sample from the population you are studying. One way, based on the sampling model, suggests that you do a good job of drawing a random sample from a target population. For example, in an experiment, if you have groups of randomly selected participants to receive a treatment or receive a placebo, the two groups will be similar at the beginning of the experiment. Therefore, if the results indicate that one group had a better outcome, you can conclude that one intervention or treatment was better than the other.
Create Standards for Participation
Create detailed, specific inclusion and exclusion standards for participation in the research. For example, your results would not be generalizable if you were comparing the salaries of men and women who worked as architects, and 90% of the participants were men; the results would not reflect the general population of male and female architects. Having inclusion and exclusion standards, therefore, helps you identify those participants who best represent the target population.
Replicate Your Study
Replicate your study with different people, at different times, in different conditions, or in various settings to increase the results’ generalizability. For example, if you are conducting a clinical experiment, you could conduct a field experiment in a natural setting outside of the laboratory. Also, regarding time, you may have done a study on the amount of time school children spent doing outdoor sports, and the study happened to taken place during a rainy season. You would have obtained different results if you had done the research at a time when the weather was better.
Ensure Participants Have Real Experience
To address the Hawthorne effect, ensure that the participants have a real experience during the study by informing them about the aim of the study, the goal, and what to expect so they do not behave in a different way than they would in real life.
Ensure Appropriate Sample Size
You will need to have an appropriate sample size to achieve reliable, generalizable results. You are likely to have false negatives and inconsistent results if the sample size is too small. However, it is not good to have a very large sample size that may be difficult to manage.
Note: You can refer to the following webpages for more information:
