What Is a Quasi-Experiment in Psychology?

A quasi-experiment is a study that tries to show one thing causes another, the same goal as a true experiment, but it skips one step: letting chance decide who ends up in which group.1Scribbr. Quasi-Experimental Design | Definition, Types & Examples Researchers work with groups that already exist, which lets them study questions where shuffling people at random would be impossible or unfair.2UPI Study. What Are Quasi-Experimental Designs in Psychology? That missing step is also what makes the conclusions a little less certain, so quasi-experiments sit one rung below true experiments in how much weight their evidence carries.

What Random Assignment Does and Why Quasi-Experiments Lack It

In a true experiment, people are assigned to conditions at random, for example by a coin flip or a computer draw.3Quillbot Blog. What is the difference between a true experiment and a quasi-experiment? Because chance spreads differences like age, mood, or motivation evenly across groups, any gap in results can be pinned on the treatment. In a quasi-experiment, people are placed in groups by non-random criteria, such as which school they attend or which clinic they visit.1Scribbr. Quasi-Experimental Design | Definition, Types & Examples The groups may therefore differ before the study even starts, and those differences could explain the results instead of the treatment.

Take a study-skills program. In a true experiment, students from one school are randomly split into a program group and a no-program group. In a quasi-experiment, one school gets the program and a similar nearby school serves as the comparison. The second school’s students might already study more, have better teachers, or come from wealthier homes, so a difference at the end could come from the schools, not the program.

Researchers can try to choose comparison groups that look alike on age, grade level, or income. But they can only match on things they thought to check. Chance balances traits nobody measured or even thought of, such as stress at home or how motivated people are, and that is the unique strength of random assignment. A quasi-experiment can only argue that its groups were similar enough, while a true experiment can rely on chance to do the balancing.

Everything else can look the same. A quasi-experiment still has a treatment or change being tested and an outcome being measured.1Scribbr. Quasi-Experimental Design | Definition, Types & Examples It usually has a control group too, meaning a comparison group that does not get the treatment. On paper the two designs can look nearly identical, so the assignment method is the detail to check.

The word “quasi” means “almost” or “resembling,” not “fake” or “sloppy.” A quasi-experiment can be carefully designed and very large. The label describes how groups were formed, not the quality of the work.

Quasi-experiments also differ from simple observational studies. In a quasi-experiment, a specific treatment or change is introduced or happens at a known time. In a purely observational or correlational study, researchers just measure things as they are, with no planned change. So quasi-experiments aim at cause and effect, while correlational studies mostly show that two things go together.

The Main Types of Quasi-Experiments

The most widely used type is the nonequivalent groups design.4Conjointly. Nonequivalent Groups Design It relies on intact groups, like two classrooms or two clinics, that researchers believe are similar.4Conjointly. Nonequivalent Groups Design Both groups are measured before, one gets the treatment, and both are measured again afterward.5Research Methods in Psychology. Quasi-Experimental Research The real question is not whether the treated group improved, but whether it improved more than the group without treatment.5Research Methods in Psychology. Quasi-Experimental Research

“Nonequivalent” means the groups were not made equal by chance, so they may differ in ways that matter even if they look alike. Measuring both groups first shows how different they were at the start, so researchers can focus on change rather than final scores. That partly protects against one group simply starting ahead.

A second type watches one group over time. A time series is a long run of measurements of the same thing taken one after another.6American Psychological Association. Experimental and Quasi-Experimental Designs for Generalized Causal Inference An interrupted time series checks whether that run shifts at the exact moment a change happens, such as the date a seat belt law took effect.6American Psychological Association. Experimental and Quasi-Experimental Designs for Generalized Causal Inference If the change worked, the numbers afterward should jump to a new level or start rising or falling at a different rate.6American Psychological Association. Experimental and Quasi-Experimental Designs for Generalized Causal Inference This before-and-after over time design can be used with a single person, a single group, or even a whole economy.7Judgment and Decision Making. Causal learning with interrupted time series data

One measurement before and one after can’t show this, because a single change between two points could be random ups and downs. Many measurements before the change establish the normal pattern, including ordinary wobbles, so a real break stands out.

The third type uses a cut-off score. A regression discontinuity design is used when treatment is given to everyone above or below a set score.8PubMed. The Goldilocks Regression Discontinuity Design: A Dual-Threshold Framework for Evaluating Interventions Outside an Optimal Range People just above and just below the line are very similar, so a sudden jump in results right at the cut-off points to the treatment.9PubMed Central. Introduction to Regression Discontinuity Design The catch is that the result tells you about people near the cut-off, not necessarily those far from it.9PubMed Central. Introduction to Regression Discontinuity Design

Here is how each might look in psychology. These are illustrations, not specific published studies.

  • Existing groups: one school runs an anti-bullying program, a similar school doesn’t, and both are surveyed before and after.
  • Before-and-after: monthly counts of anxiety-clinic visits are tracked for years and checked for a shift after a new campus counseling policy.
  • Cut-off: students scoring below a set reading score get tutoring, and those just under and just over the line are compared later.

A close cousin is the natural experiment, where an event outside anyone’s control creates the comparison, like a disaster hitting one region. Researchers can then compare affected and unaffected people, as a study did with mental health service use after an earthquake. Nobody assigned the groups, so these share the same strengths and weaknesses as other quasi-experiments.

Why Psychologists Use Quasi-Experiments

Quasi-experiments are used where fully randomized trials are not feasible or not ethical.10PubMed Central. An Introduction to the Quasi-Experimental Design (Nonrandomized Design) A hospital can’t randomly delay treatment for patients just to create a comparison group.2UPI Study. What Are Quasi-Experimental Designs in Psychology? And many things psychologists care about, like divorce, earthquakes, or job loss, can’t be assigned to anyone.11PubMed Central. Mental health service use after the Van Earthquake: evidence from a quasi-experimental difference-in-differences study For these questions, a quasi-experiment is often the most practical way to test cause and effect.

Schools often adopt a program as a whole, so splitting a classroom randomly would be impractical. Laws and policies apply to everyone at once, which is why before-and-after designs fit them well.

Public health psychology offers another example. Researchers compared re-employed and still-unemployed people over time, and finding a job after three years out of work improved how healthy people said they felt. No one could ethically assign people to stay unemployed, so a quasi-experiment was the realistic choice.

These studies are also valued for giving practical answers from real settings rather than a lab.2UPI Study. What Are Quasi-Experimental Designs in Psychology? Real schools and clinics have mixed groups, busy staff, and interruptions that a controlled lab removes. A program that works only under tightly controlled conditions may not work once ordinary staff run it, so testing it where it will actually be used tells decision-makers whether it is likely to work in practice.

In one study of about 1,400 primary school children aged 7 to 11, a school mental health program improved self-esteem compared with similar children who didn’t get it.12Humanities and Social Sciences Communications. A quasi-experimental controlled study of a school-based mental health programme to improve the self-esteem of primary school children The same program showed no effect on emotional well-being, family, friends, or school life.12Humanities and Social Sciences Communications. A quasi-experimental controlled study of a school-based mental health programme to improve the self-esteem of primary school children

What Can Go Wrong in a Quasi-Experiment

The largest challenge is selection bias: comparison groups that differ from the treated group in ways that affect the outcome.13PubMed Central. Selecting and Improving Quasi-Experimental Designs in Effectiveness and Implementation Research The second is history bias, when outside events during the study change the outcome on their own.13PubMed Central. Selecting and Improving Quasi-Experimental Designs in Effectiveness and Implementation Research A program might look effective because of a new school principal, a holiday, or a pandemic, not the program itself.

Groups that differed from the start are easy to miss. Volunteers for a therapy program may be more motivated than people who don’t sign up, and motivated people often improve faster no matter what treatment they get. Comparing volunteers with non-volunteers can then overstate how well the therapy works. Asking how people got into each group is the quickest check.

The bias can also run the other way: it can hide a real benefit. If a program is given to the neediest students, they may do worse afterward even though the program helped them. So the question is not only whether groups differ, but which way the difference would bend the results.

That is one reason researchers measure the starting point. Measuring both groups before the treatment shows whether they started out similar on the outcome. If they start far apart, the final gap can’t be credited to the program without adjusting for it. The check helps, but it cannot reveal differences in things nobody measured.

Outside events can fool a before-and-after study. If a law starts the same month a recession hits, any change could come from either. Results that slowly improve anyway may look like a treatment effect if the earlier trend is ignored. Long runs of measurements before the change help separate a real shift from an existing trend.

Being watched can change behavior too. The Hawthorne effect is when people change their behavior because they know they are being observed.14Wikipedia. Hawthorne effect A group that knows it’s getting special attention may improve partly from that attention.

How Much You Can Trust the Results

Common evidence rankings place quasi-experiments at Level II, below randomized trials and above non-experimental studies at Level III.15Medical College of Wisconsin Libraries. LibGuides: Evidence Based Practice: Study Designs & Evidence Levels Quality varies within the category, and stronger quasi-experimental designs give more convincing evidence for cause and effect.16PubMed Central. The Use and Interpretation of Quasi-Experimental Studies in Medical Informatics A single-group before-and-after study is weaker than one with a comparison group and many measurements.

Without random assignment, “convincing” comes from several kinds of evidence pointing the same way. A sharp change exactly when a treatment starts is harder to explain away. Results that match across different groups and places are more trustworthy.

Evidence levels describe the typical strength of a design, not the quality of any single study. A well-run quasi-experiment can be more convincing than a small, poorly run randomized trial.

How Researchers Make Quasi-Experiments Stronger

One tool is matching people who look alike on paper. Propensity score matching tries to balance two nonequivalent groups on characteristics researchers can measure.17PubMed. Propensity scores: an introduction and experimental test Each person gets a score for how likely they were to receive the treatment, based on things like age, gender, and past health care visits.18PubMed Central. Web-based Mindfulness-Based Cognitive Therapy for reducing residual depressive symptoms: An open trial and quasi-experimental comparison to propensity score matched controls Each treated person is then paired with an untreated person who has a similar score.18PubMed Central. Web-based Mindfulness-Based Cognitive Therapy for reducing residual depressive symptoms: An open trial and quasi-experimental comparison to propensity score matched controls In one depression study, an online mindfulness therapy group was compared with matched patients in usual care, and the therapy group improved more.18PubMed Central. Web-based Mindfulness-Based Cognitive Therapy for reducing residual depressive symptoms: An open trial and quasi-experimental comparison to propensity score matched controls Matching only balances what was measured, so unmeasured differences such as motivation can still remain.

Treated people with no similar untreated match may be left out, since there is no fair comparison for them. The result is a smaller but more comparable pair of groups.

Another tool is comparing change in both groups. Difference-in-differences compares how much the treated group changed with how much the comparison group changed over the same time.11PubMed Central. Mental health service use after the Van Earthquake: evidence from a quasi-experimental difference-in-differences study If both groups improve because of an outside trend, that shared improvement is subtracted out. Imagine a treated school’s scores rise 10 points and a comparison school’s rise 6. The program gets credit for only the extra 4 points, where without the comparison you’d wrongly credit it with all 10. This works only if both schools would have changed alike without the program. Researchers can also use instrumental variables, a method meant to get closer to cause and effect when groups weren’t randomly formed.19Utrecht University. Instrumental variables in psychology

Finally, there is openness about methods. A reporting checklist called TREND asks authors of non-randomized studies to describe their design, comparison groups, and how they adjusted for bias.20PubMed Central. Improving the reporting quality of nonrandomized evaluations of behavioral and public health interventions: the TREND statement

How to Recognize a Quasi-Experiment in a Study or Assignment

Some practice cases trip students up. Comparing men and women on memory is not a true experiment, because gender can’t be assigned; with no treatment being tested, it is a correlational comparison rather than a quasi-experiment. Testing a drug where a computer randomly picks who gets it is a true experiment. Comparing two hospital wards after one adopts a new routine is a quasi-experiment. Studies often say “quasi-experimental” or “nonrandomized” right in the title.

Some studies fit more than one description, with both randomized and non-randomized parts. Ask how the key comparison was formed. If the main treatment comparison was not randomized, the study is quasi-experimental even if other parts were randomized. Judge by how people got into the groups being compared, not by how technical the study sounds.