Quasi-Experimental Design | Definition, Types & Examples
A quasi-experimental design is used to establish a cause-and-effect relationship between independent and dependent variables. However, unlike in a true experiment, participants are not randomly assigned to experiment conditions.
Quasi-experimental designs still involve the manipulation of an independent variable but may lack control over extraneous variables that could impact a study’s outcomes.
| Characteristic | Experimental design | Quasi-experimental design |
|---|---|---|
| Random assignment of participants | ||
| Manipulation of an independent variable | ||
| Control of extraneous variables | ? |
What is quasi-experimental design?
The term “quasi” means “resembling,” and quasi-experiments indeed resemble true experiments. To better understand quasi-experimental design, a brief review of true experimental design may be helpful.
An experimental design involves the manipulation and measurement of independent and dependent variables. By systematically varying an independent variable and measuring the resulting changes in an independent variable, a researcher can establish a cause-and-effect relationship.
A key element of true experimental design is random assignment—participants are randomly assigned to different study conditions, such as treatment and control groups. However, in many cases, this random assignment is not possible. For example, it would be unethical to withhold an effective medical treatment from a group of patients. Researchers may instead choose to use a quasi-experimental design.
Quasi-experimental designs still involve the manipulation of an independent variable, but because they lack random assignment, it becomes more difficult to control for potential confounds (things that could influence the study’s outcomes).
Example of quasi-experimental design
The following example illustrates how quasi-experimental and true experimental designs differ.
To explore this question using a true experimental design, researchers could randomly select students to receive free lunches. Students who do not receive free lunches could be compared as a control group. If the test scores of the students receiving free lunches improve more than the control group scores, the school board could conclude that free lunches improve academic performance.
There are obvious ethical issues with randomly withholding free lunches from students. A quasi-experimental design may be a better option in this scenario. The school board might look at historical data on academic performance before and after other school districts implemented free lunch programs. If the introduction of free lunches coincided with improvements to standardized testing scores at these schools, the school board might conclude that free lunches improve academic performance.
What are the types of quasi-experimental design?
There are many different types of quasi-experimental designs. They differ in their inclusion of a control group and when they collect data (before and/or after an intervention). Some common designs are summarized in the table below.
Nonequivalent groups design
In this design, a control and experimental group are chosen that appear similar to each other. Nonequivalent groups design is a form of between-subjects design: participants experience only one condition, and you compare outcomes between conditions.
However, because these conditions (or groups) are not randomly assigned, they must be considered “nonequivalent.” By making the nonequivalent groups as alike as possible, researchers can try to minimize the impact of confounds.
Pretest-posttest design
In this design, a dependent variable is measured once before and once after a treatment is implemented. A pretest-posttest design is similar to a within-subjects design: each participant experiences all conditions (in this case a pretreatment condition and a posttreatment condition), and differences between these timepoints are compared.
A disadvantage of a pretest-posttest design is that it can be difficult to isolate the treatment as the factor that caused changes between pretest and posttest scores.
Interrupted time series design
An interrupted time series design is similar to a pretest-posttest design, but multiple measurements are taken before and after a treatment. The treatment “interrupts” the time series of measurements. If measurements from before a treatment are consistently different from measurements taken after, researchers may conclude that their treatment is effective.
Regression discontinuity design
Often, people are separated into groups based on an arbitrary cutoff level (e.g., students with a score of 49% may be assigned to a “failing” group, while students with a score of 51% are assigned to a “passing” group). These groups may be used to determine treatments (e.g., students in the “failing” group may receive extra tutoring).
Because the threshold used to define groups is arbitrary, people whose scores fall immediately above and immediately below this threshold are likely quite similar (returning to our example, students who score 49% and 51% on a test have virtually identical performance). If these highly similar people are compared as treatment and control groups, differences observed between them are likely due to the treatment rather than any inherent differences between the groups themselves.
To evaluate the impact of this scholarship, the college compares academic outcomes for students with an incoming average of 79%, who did not receive the scholarship, and students with an average of 80%, who did. If there is a significant difference in the first-year GPAs of these two groups, this suggests that the scholarship had a causal effect on academic success in college.
Natural experiment design
Sometimes, an external event may naturally separate participants into treatment and control groups. By studying differences between these groups, researchers can identify the causal effects of this external event. Although they involve random assignment, natural experiment designs are considered quasi-experimental because they are observational.
Comparing outcomes for eligible residents who did and did not receive coverage allowed researchers to evaluate the impact of Medicaid.
Quasi-experiments may also involve a combination of several different designs.
Pros and cons of quasi-experimental design
Like any research design, quasi-experimental designs have certain advantages and disadvantages.
Advantages of quasi-experimental design:
-
- Higher external validity than true experiments: Quasi-experiments are generally conducted in real-world settings rather than controlled laboratory settings, so they may better reflect reality.
- Higher internal validity than non-experimental research: Quasi-experimental studies allow for some control of external factors, so their results may be less influenced by confounds than non-experimental designs.
- Feasibility: In certain situations, quasi-experimental research may be more feasible than experimental studies due to ethical or practical concerns.
Disadvantages of quasi-experimental design:
- Lower internal validity than true experiments: Because they lack the random assignment of true experiments, quasi-experimental designs are more susceptible to the impact of confounds.
- Lack of control over treatment or intervention: Often, researchers conducting quasi-experimental research do not implement or oversee the intervention they are studying, which could further reduce internal validity.
- Reliance on existing data: Quasi-experimental designs often use data that are already collected, and these data may be incomplete or difficult to access.
Frequently asked questions about quasi-experimental design
- What is an interrupted time series design?
-
An interrupted time series design is a quasi-experimental research method. It is similar to a pretest-posttest design, but multiple data points, called a time series, are collected for a participant before and after an intervention is administered. The intervention “interrupts” the time series of observations.
If scores taken after the intervention are consistently different from scores taken before the intervention, a researcher can conclude that the intervention was successful. Considering multiple measurements helps reduce the impact of external factors
- Why is randomization important in an experimental design?
-
Randomization is a crucial component of experimental design, and it’s important for several reasons:
- Prevents bias: Randomization ensures that each participant has an equal chance of being assigned to any condition, minimizing the potential for bias in the assignment process.
- Controls for confounding variables: Randomization helps to distribute confounding variables evenly across conditions, reducing the risk of spurious correlations between the independent variable and the outcome.
- Increases internal validity: By randomly assigning participants to conditions, you can increase the confidence that any observed differences between conditions are due to the independent variable and not some other factor.
- What is regression discontinuity design?
-
Regression discontinuity design is a quasi-experimental approach that compares two groups of participants that are separated based on an arbitrary threshold. This method assumes that people immediately above and immediately below this threshold are quite similar. Any subsequent differences between these groups can therefore be attributed to interventions that one group does or does not receive.
For example, imagine you’re testing the efficacy of a cholesterol medication. You administer this medication only to patients whose cholesterol levels exceed 200 mg/dl. You then compare heart health indicators of patients with cholesterol levels slightly over 200 mg/dl, who do receive the medication, to patients with cholesterol levels slightly below 200 mg/dl, who do not receive the medication. If the heart health of the former group improves relative to the latter group, you may conclude that the treatment worked.
- What is pretest-posttest design?
-
A pretest-posttest design is a quasi-experimental research design. Two data points are collected for a participant: one from before an intervention is introduced and one from after an intervention. A difference in these scores may indicate that the intervention was effective.
For example, imagine you complete a depression inventory before and after a 6-week therapy program. An improvement in your score may indicate that the program worked.
- What is the difference between a true experiment and a quasi-experiment?
-
In a true experiment, participants are randomly assigned to different study conditions. A quasi-experiment lacks this random assignment.
True experiments are also usually conducted in controlled laboratory settings, which facilitates control of confounding variables that may impact study results. Quasi-experimental designs often collect data in real-world settings, which increases external validity but reduces control of confounds.
Finally, both true experiments and quasi-experiments generally involve the manipulation of an independent variable to determine its causal effect on a dependent variable. However, in a quasi-experimental study, researchers may have less control over this manipulation (e.g., they may be studying the impact of an intervention or treatment that has already happened).
- When should I use quasi-experimental design?
-
Practical or ethical concerns may prevent researchers from using a true experimental design:
Practical concerns that prevent researchers from conducting a true experiment may include the cost of a study or the time required to design the experiment and collect and analyze data.
Ethical concerns may also limit the feasibility of true experimental research. It would be unethical to intentionally prevent study participants from accessing medication or other treatments that the researcher knows would benefit them.
In these cases, a quasi-experimental design may be more appropriate.
