Section 5.4: Quasi-experimental Designs | Research


Fundamentals of Social Research

Adam J. McKee, Ph.D.


DRAFT - Do Not Distribute

This content is released as a draft version for comment by the scholarly community.  Please do not distribute.


We’ve looked at the good (experiments), we’ve looked at the bad (pre-experiments), now let’s consider the ugly.  I say ugly because quasi-experiments don’t have the elegant simplicity that we find in true experiments.  What a true experiment does through the elegant simplicity of randomization, the researcher using a quasi-experimental design doesn’t do at all, or does through controls built into the design or through statistical methods.

In a quasi-experiment, the hypothesized cause of observed differences has already occurred.  For this reason, random assignment to groups is impossible; the groups are preexisting. Studies that examine a preexisting characteristic that cannot be manipulated, such as gender, age, or eye color must be undertaken using a quasi-experimental design.

In addition, many phenomena of interest to social scientists would be unethical, immoral, or illegal to create artificially.  Can we really study things like child abuse, infant malnutrition, the effect of alcohol on a fetus, and so forth using true experimental methods?  Absolutely not. Hitler may have sanctioned these methods, but modern research ethics dictate that participants cannot be harmed by participation in research.  Thus, we use the best method available to us—quasi-experimental designs.

In addition to these types of circumstances, post hoc (after the fact) designs like quasi-experimental studies give the researcher the ability to look into the past and look for clues as to why something happened.  An evolutionary psychologist studying the development of human intelligence certainly cannot use experiments; the only methods that work are those that let us look into the past.

The Nonequivalent Control Group Design

This design is very similar to the pretest-posttest control group design we discussed under the heading of true experiments.  The difference here is that there is no random selection or random assignment. The researcher is forced to use intact groups.  In other words, groups are defined by some preexisting criterion, not random assignment.  The lack of randomization in the design means that we are not very confident that our groups are equivalent at the beginning of the study.  That is why we refer to the groups as “nonequivalent.”

The Static Group Comparison

In situations where the researcher can neither randomize nor administer a pretest, the static group comparison design can be used.  This design is almost identical to the nonequivalent control group design except there is no pretest.  This design is one that should be used with great caution. Because there is no pretest, the control group is the only thing that the experimental outcomes can be compared with.  This is fine when randomization assures us that the experimental and control groups are equivalent, but without randomization, the design is full of threats to validity.

Studying Change across Time

Many social scientists are interested in phenomena that occur in people over the course of time.  It is well known that people behave differently at different ages or stages across the lifespan. Two major methods have been developed to consider such changes.

The longitudinal method examines changes in behavior in a group of subjects at more than one point in time.  Let’s say that you are interested in studying the “aging out” of crime by following a group of boys from age 12 to age 45, collecting data every year during the research period.  This is an example of a longitudinal design.

There are several advantages to longitudinal studies.  First, it allows the researcher to answer questions about phenomena that occur over a long period of time.  Second, since data are recorded for the same individuals, those individuals serve as their own controls. That is, most extraneous variables are controlled because each participant brings the same characteristics and experiences to the data collection process every time.  There are problems as well. Longitudinal studies tend to be expensive and take lots of time. Mortality is a big problem; the more time that passes, the less likely it will be that participants will be available to provide new data.

The cross-sectional method has the same aims as a longitudinal study, but goes about it in a different way.  With the cross-sectional study, all data are collected at the same time, and subjects are grouped according to age or stage in the human life cycle.  The primary advantage this method has over the longitudinal method is cost—it can be done a lot quicker. The drawback is that since different individuals are used to represent each stage or age of interest, there is not good control of extraneous variables.

Modification History

File Created:  07/25/2018

Last Modified:  07/25/2018

[ Back | Content | Next]


This work is licensed under an Open Educational Resource-Quality Master Source (OER-QMS) License.

Open Education Resource--Quality Master Source License


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.