Quasi-Experimental Design: When and How to Use It

Quick Answer: What Quasi-Experimental Design Is

The definition.
A quasi-experimental design is a research design that tests the effect of an intervention or treatment on an outcome without random assignment of participants to conditions. It includes manipulation of an independent variable and a comparison group, but groups are formed by natural circumstances rather than randomization.

When to use it.
When random assignment is impossible, unethical, or impractical. Common in education, public health, policy evaluation, and economics, where researchers compare schools, clinics, regions, or cohorts that already exist.

The main types.
Nonequivalent groups design, interrupted time series, regression discontinuity, and difference-in-differences. Each addresses confounding through a different statistical or structural strategy.

A quasi-experimental design is the most widely used research design in real-world evaluation studies, where random assignment is rarely possible. It tests causal hypotheses about the effect of an intervention, policy, or treatment, but it does so without the randomization that defines true experiments. This article explains what quasi-experimental designs are, the main types and how each one works, how they differ from true experiments, the threats to validity that affect them, and how to choose the right quasi-experimental design for your study.

For a broader overview of where quasi-experimental designs fit among research methodologies, see our research methodology guide for graduate students. For the foundational comparison with true experimental designs, see our article on experimental research design.

What Is a Quasi-Experimental Design?

A quasi-experimental design tests the effect of an independent variable on a dependent variable using a comparison group, but without randomly assigning participants to the experimental and comparison conditions. The "quasi" prefix means "almost" or "resembling." A quasi-experiment resembles a true experiment in structure: there's an intervention, a comparison, and a measurable outcome. What's missing is the random assignment that lets researchers rule out pre-existing differences between groups.

In a true experimental design, the researcher randomly assigns participants to conditions, which evenly distributes pre-existing differences across groups. In a quasi-experimental design, the groups already exist. Students attend different schools. Patients use different clinics. States adopt different policies. The researcher compares these naturally occurring groups, knowing that they may differ in ways that have nothing to do with the intervention itself.

This loss of randomization is the central methodological challenge in quasi-experimental research. Strong quasi-experimental designs use statistical or structural techniques to make causal claims defensible despite the absence of random assignment.

When to Use a Quasi-Experimental Design

Researchers turn to quasi-experimental designs in four common situations.

When Random Assignment Is Unethical

Some independent variables can't be ethically manipulated. You can't randomly assign children to experience trauma. You can't randomly assign adults to smoke for thirty years. You can't randomly assign communities to lose access to clean water. When the research question concerns an exposure or condition that researchers can't ethically impose, quasi-experimental comparisons of naturally occurring exposed and unexposed groups are the next-best option.

When Random Assignment Is Impractical

In school-based research, you usually can't reassign students between classrooms or schools. A researcher studying a new curriculum compares classrooms that adopt it to similar classrooms that don't. In public health, you can't randomly assign cities to different smoking laws. A researcher studying smoking bans compares cities that pass them to similar cities that don't. The practical realities of field research often rule out randomization even when it would be theoretically possible.

When You're Studying a Naturally Occurring Event

Sometimes a policy change, natural disaster, or other event creates a "natural experiment." A state raises the minimum wage. A hurricane disrupts schooling in one region but not another. A new screening program rolls out in some clinics first. Researchers can take advantage of these events to study causal effects using quasi-experimental methods, even though no one designed the event as an experiment.

When You're Evaluating an Existing Program

Program evaluation almost always uses quasi-experimental designs. A school district wants to know whether its mentoring program improved graduation rates. A hospital wants to know whether a new triage protocol reduced wait times. These programs are usually already in place, with no opportunity to randomize who received them. Researchers compare participants to non-participants, or compare outcomes before and after the program began.

The Main Types of Quasi-Experimental Design

Four quasi-experimental designs are widely used. Each addresses the absence of randomization through a different structural or statistical strategy.

Nonequivalent Groups Design

The most common quasi-experimental design. The researcher compares an intervention group to a comparison group that wasn't randomly assigned, usually measuring both groups before and after the intervention. The pretest lets the researcher check whether the groups were similar at baseline and measure change within each group over time.

The classic example is a school-based curriculum study. Classrooms that adopt the new curriculum form the intervention group. Similar classrooms that keep the existing curriculum form the comparison group. Reading scores are measured at the start and end of the school year. If the intervention group improves more than the comparison group, that's evidence the curriculum worked. The weakness is that the two groups may differ at baseline in ways that affect how much they improve, regardless of the intervention.

Interrupted Time Series Design

In an interrupted time series design, the researcher measures the outcome at multiple points before and after an intervention, looking for a discontinuity at the point of intervention. The pre-intervention measurements establish a trend. If the trend changes sharply at the intervention point, that's evidence the intervention had an effect.

A public health example: a city passes a smoking ban in 2015. Researchers collect monthly emergency room admissions for heart attacks from 2010 through 2020. If admissions decline sharply starting in 2015 and the new trend differs from what the pre-2015 pattern would predict, that's evidence the ban affected cardiovascular health. Adding a comparison city without a ban (a controlled interrupted time series) strengthens the design considerably.

Regression Discontinuity Design

A regression discontinuity design takes advantage of a cutoff rule. Eligibility for a program, scholarship, or treatment depends on whether a score crosses a threshold. Students just above and just below the cutoff are similar on most characteristics but differ in whether they receive the intervention. The researcher compares outcomes for those who barely qualified to outcomes for those who barely missed, treating the cutoff as a near-randomization mechanism.

An education example: a scholarship is awarded to students who score above 1200 on an entrance exam. A student who scores 1201 is similar in ability and motivation to a student who scores 1199, but one receives the scholarship and the other doesn't. Comparing the long-term outcomes of the two groups isolates the effect of the scholarship itself rather than the effect of being a high-scoring student. Regression discontinuity designs are widely used in economics and education policy research.

Difference-in-Differences Design

A difference-in-differences design compares the change in an outcome over time between a group that received an intervention and a group that didn't. The key assumption is that, in the absence of the intervention, both groups would have followed parallel trends. The intervention's effect is the difference between the actual post-intervention change in the treatment group and the change that would have been expected based on the comparison group's trend.

An economics example: one state raises its minimum wage in 2018. A neighboring state doesn't. Researchers compare employment rates in both states from 2015 through 2022. The difference-in-differences estimate is the change in employment in the treatment state from before to after 2018, minus the corresponding change in the comparison state. This design has become standard in policy evaluation and is widely used in quantitative research in economics, public health, and labor studies.

Submitting a quasi-experimental study for journal review or graduate committee approval?

Editor World's native English editors specialize in research methodology across disciplines including education, public health, economics, and policy evaluation. Choose your own editor by subject expertise, receive your manuscript back with Track Changes, and get an optional certificate of editing for journals that require one. BBB A+ accredited since 2010. 100% human editing, no AI at any stage.

Get Your Dissertation Edited

Quasi-Experimental vs True Experimental Design

The two designs share several features. Both manipulate an independent variable. Both use a comparison group. Both aim to support causal inference. The difference is in how participants end up in each condition and what that means for the strength of the causal claim.

In a true experiment, random assignment ensures that the experimental and control groups are statistically equivalent at baseline. Any difference in outcomes can be attributed to the intervention itself. In a quasi-experiment, the groups exist for reasons unrelated to the research, and they may differ at baseline in ways that affect the outcome. The researcher has to account for those differences using statistical or structural methods.

This is why true experiments produce stronger causal evidence and why regulatory agencies require them for high-stakes decisions like drug approvals. But true experiments aren't always possible, and for many of the most important research questions in education, public health, and policy, quasi-experimental designs are the only feasible approach. A well-executed quasi-experiment can support credible causal claims, especially when the design uses pretest measures, comparison groups, and statistical adjustments to address the most plausible confounders.

Strengths of Quasi-Experimental Design

Quasi-experimental designs have distinct advantages over both true experiments and purely observational research.

Practical feasibility. They work in real-world settings where random assignment is impossible, expensive, or unethical. This makes them the dominant approach in program evaluation, education research, and policy analysis.
External validity. Because they're conducted in natural settings with naturally occurring groups, the results often generalize better to real-world conditions than tightly controlled laboratory experiments.
Ability to study large-scale interventions. Policies, curricula, and public health programs affect populations, not individuals. Quasi-experimental designs can evaluate these at scale in ways no laboratory experiment could.
Use of existing data. Many quasi-experimental designs use administrative data, census data, or routinely collected records, which reduces cost and lets researchers study long time periods.
Causal inference under realistic conditions. With careful design and analysis, quasi-experiments can support credible causal claims even without randomization, especially in regression discontinuity and difference-in-differences designs.

Limitations and Threats to Validity

Quasi-experimental designs face threats to validity that true experiments largely avoid. Researchers need to anticipate these and address them in the design and analysis.

Selection bias. The biggest threat. Groups formed without random assignment may differ in ways that influence the outcome, leading to results that reflect those differences rather than the intervention. For a deeper discussion, see our research bias guide.
History effects. Events that occur between pretest and posttest, unrelated to the intervention, can affect the outcome. A long study period increases this risk.
Maturation. Participants change naturally over time. In studies of children, students, or recovering patients, some of the observed improvement may simply reflect normal development.
Regression to the mean. When groups are selected because they have extreme scores, their scores tend to move toward the average on retest, regardless of any intervention. This can create the appearance of an effect when none exists.
Testing effects. Taking a pretest can affect how participants perform on the posttest, especially if the same measure is used. This is also a concern in true experiments but is harder to detect without random assignment.
Instrumentation. Changes in how the outcome is measured during the study can produce apparent effects that aren't real.
Attrition. Differential dropout between groups can bias the results, especially if participants who leave the intervention group differ systematically from those who leave the comparison group.

Strategies to Strengthen Causal Claims in Quasi-Experimental Research

Strong quasi-experimental research uses several techniques to address the absence of randomization.

Matching

Matching pairs each participant in the intervention group with a similar participant in the comparison group based on observable characteristics. Propensity score matching is a more sophisticated variant that combines multiple characteristics into a single score and matches on that. Matching reduces but doesn't eliminate selection bias, because it can only adjust for differences that are measured.

Statistical Control

Including baseline characteristics as covariates in the analysis adjusts for measured pre-existing differences between groups. Analysis of covariance (ANCOVA) and multiple regression are the standard methods. Like matching, this approach addresses observed differences but can't account for unobserved ones.

Multiple Comparison Groups

When possible, including more than one comparison group strengthens causal inference. If the intervention group differs from both comparison groups in the same way, alternative explanations involving specific characteristics of either comparison group become less plausible.

Pre-Intervention Trend Analysis

In interrupted time series and difference-in-differences designs, examining the pre-intervention trend in both groups tests the parallel trends assumption. If the groups followed similar trajectories before the intervention, the assumption that they would have continued to do so without it becomes more credible.

Sensitivity Analysis

Sensitivity analyses test how robust the results are to alternative specifications, alternative comparison groups, or hypothetical confounders. If the results hold up across reasonable variations, the causal claim becomes more credible. Reviewers at top journals increasingly expect sensitivity analyses in quasi-experimental research.

Quasi-Experimental Design Examples

Three examples show how quasi-experimental designs are applied across disciplines.

Example 1: A Public Health Policy Evaluation

A researcher wants to know whether a state's expansion of Medicaid in 2014 reduced uninsured rates and improved preventive care use. The researcher uses a difference-in-differences design, comparing the expansion state to a similar neighboring state that didn't expand. Uninsured rates and preventive care use are measured annually from 2010 through 2020. The pre-2014 trends are checked for parallelism. The difference-in-differences estimate captures the effect attributable to the expansion. This design is standard in health policy research and produces findings that inform legislative debates and program design.

Example 2: An Educational Curriculum Study

A school district adopts a new math curriculum in half of its elementary schools, based on which schools requested it. A researcher uses a nonequivalent groups design with pretest and posttest measures of math achievement. Because the schools weren't randomly assigned, the researcher matches participating and non-participating schools on prior achievement, school size, and demographic composition. Pre-intervention trends in math scores are checked for similarity between groups. The analysis includes baseline scores and school characteristics as covariates. The design isn't as strong as a true experiment, but with these adjustments, the results can credibly inform decisions about whether to expand the curriculum to the remaining schools.

Example 3: A Graduate Student's Regression Discontinuity Study

A doctoral student studies whether a university honors program affects long-term career outcomes. Admission to the honors program requires a GPA above 3.5 in the first year. The student compares graduates who entered with a GPA of 3.4 to 3.5 to those who entered with a GPA of 3.5 to 3.6. The two groups are very similar academically but differ in whether they joined the honors program. The student measures career outcomes ten years after graduation. Because the two groups are nearly equivalent except for honors program participation, the regression discontinuity design supports a credible causal claim about the program's effect. This is the kind of design that benefits from careful dissertation editing attention to ensure the methodology section addresses every assumption.

How to Choose a Quasi-Experimental Design

The right quasi-experimental design depends on the research question, the data available, and the timing of the intervention.

Identify whether a sharp cutoff exists. If eligibility for the intervention depends on a threshold (a test score, an income level, an age limit), a regression discontinuity design is often the strongest option.
Check whether you have data over time for both groups. If you have multiple measurements before and after the intervention for a treatment group and a comparison group, a difference-in-differences design is feasible. If you have measurements only for the treatment group over time, an interrupted time series may work.
Assess the comparison group. If you can identify a comparison group that's similar to the intervention group on observable characteristics, a nonequivalent groups design with matching or statistical adjustment is appropriate.
Plan for the threats to validity. Identify the most plausible confounders before data collection. Design the study to measure them, and plan the analysis to address them.
Consider sample size requirements. Different quasi-experimental designs have different statistical power profiles. For more on sample size and population considerations, see our article on population vs sample in research.

Reporting Quasi-Experimental Research in Your Manuscript

Reviewers and committee members evaluate quasi-experimental studies against a specific set of expectations. The methodology section should be unusually transparent because the absence of randomization makes the design's credibility depend on how carefully alternative explanations are addressed.

A clear statement of why random assignment wasn't used, and why the quasi-experimental design chosen is the most credible alternative
A description of the intervention group and comparison group, including how they were defined and how they came to be in each condition
Baseline measures for both groups, with statistical tests of equivalence on relevant characteristics
The specific quasi-experimental design used (nonequivalent groups, interrupted time series, regression discontinuity, difference-in-differences), with citations to methodological sources
The assumptions the design depends on (such as parallel trends in difference-in-differences) and evidence that they hold
The statistical methods used to address selection bias and other confounders
Sensitivity analyses showing whether the results are robust to alternative specifications
An honest discussion of limitations and the strength of the causal claim the design can support

Reviewers at top journals increasingly expect this level of methodological rigor in quasi-experimental research. Studies that present quasi-experimental results as if they were from true experiments, or that fail to address obvious threats to validity, face rejection or major revision requests.

Self-Audit Checklist for Your Quasi-Experimental Study

Before submitting a quasi-experimental study, work through this checklist. If you can answer yes to each question, your methodology section is on solid ground.

Have you stated clearly why random assignment wasn't possible and why a quasi-experimental design is appropriate?
Have you specified which quasi-experimental design you used and cited methodological references for it?
Have you described how the intervention and comparison groups were formed, including any selection process?
Have you tested whether the groups were equivalent at baseline on relevant characteristics?
Have you identified the most plausible confounders and explained how the design or analysis addresses them?
If your design assumes parallel trends, have you shown evidence that the pre-intervention trends were similar?
Have you reported sensitivity analyses or robustness checks?
Have you discussed the limitations honestly, including the threats to validity that your design can't fully address?

If you found gaps in your manuscript working through this checklist, professional editing can help you address them before submission. Editor World's dissertation editing service connects you with native English editors who specialize in research methodology across disciplines, including the statistical and design-focused fields where quasi-experimental research is most common.

Frequently Asked Questions

What is a quasi-experimental design?

A quasi-experimental design is a research design that tests the effect of an intervention on an outcome using a comparison group, but without randomly assigning participants to conditions. It includes manipulation of an independent variable and a comparison group, but the groups are formed by natural circumstances rather than randomization. Quasi-experimental designs are widely used in education, public health, economics, and policy evaluation, where random assignment is often impossible, unethical, or impractical.

What is the difference between quasi-experimental and true experimental design?

The critical difference is random assignment. In a true experimental design, participants are randomly assigned to conditions, which controls for pre-existing differences between groups and supports strong causal inference. In a quasi-experimental design, the groups already exist and weren't randomly assigned, so they may differ at baseline in ways that affect the outcome. True experiments produce stronger causal evidence, but quasi-experimental designs are often the only feasible option for evaluating real-world programs, policies, and interventions. For more on the comparison, see our article on experimental research design.

What are the main types of quasi-experimental design?

The four main types are nonequivalent groups design, interrupted time series, regression discontinuity, and difference-in-differences. Nonequivalent groups designs compare an intervention group to a similar comparison group, often with pretest and posttest measures. Interrupted time series designs look for a discontinuity in an outcome trend at the point of intervention. Regression discontinuity designs exploit a cutoff rule for eligibility. Difference-in-differences designs compare changes over time between a treatment group and a comparison group that wasn't exposed to the intervention.

When should you use a quasi-experimental design instead of a true experiment?

Use a quasi-experimental design when random assignment is unethical, impractical, or impossible because the intervention is a naturally occurring event such as a policy change or natural disaster. Quasi-experimental designs are also the standard approach for program evaluation, where the program is already in place and randomization wasn't built into its implementation. The choice should follow from the research question and the practical realities of the setting, not from a preference for either approach.

What are the main threats to validity in quasi-experimental research?

The main threats include selection bias, history effects, maturation, regression to the mean, testing effects, instrumentation changes, and differential attrition. Selection bias is the biggest concern because groups formed without random assignment may differ in ways that affect the outcome. Strong quasi-experimental research uses matching, statistical control, multiple comparison groups, pre-intervention trend analysis, and sensitivity analyses to address these threats. For more on the cognitive and procedural biases that affect research, see our research bias guide.

Can quasi-experimental designs establish causal relationships?

Yes, when properly designed and analyzed. Quasi-experimental designs can support credible causal claims even without random assignment, especially when they use techniques like matching, statistical control, regression discontinuity, or difference-in-differences. The causal evidence is generally weaker than from a true experiment, because pre-existing differences between groups can't be fully ruled out. But for many research questions in education, public health, economics, and policy, where random assignment is impossible, well-designed quasi-experiments produce the strongest causal evidence available.

How do you report a quasi-experimental design in a manuscript?

The methodology section should explain why random assignment wasn't used, describe the intervention and comparison groups including how they were formed, report baseline equivalence tests, specify the type of quasi-experimental design used with methodological citations, identify the assumptions the design depends on and provide evidence they hold, describe the statistical methods used to address selection bias, report sensitivity analyses, and discuss limitations honestly. Reviewers at top journals expect a higher level of methodological transparency in quasi-experimental research than in true experiments, because the design's credibility depends on how carefully alternative explanations are addressed.