At the end of the lesson, the student will be able to:

1. describe ex post facto research and compare it

to experimental research;
2. state conditions needed to infer a causal
relationship;
3. describe alternative explanations in ex post
facto research and identify cases in which
these are or are not plausible;
4. describe methods of partial control and
identify cases in which they would be useful;
5. distinguish between proactive and retroactive
ex post facto designs;
6. identify questions for which ex post facto
research would be the method of choice; and
7. state the major weakness of ex post factor
research.
As researchers probe such
educational questions as “Why
are some children better readers
than others?”, “What is the
effect of single-parent homes on
achievement?” and “Why do
some youths become delinquent
while others do not?” they find
that only some questions can be
investigated through
experimental research.
If you want to investigate the
influence of such variables as home
environment, motivation,
intelligence, parental reading
habits, age, ethnicity, gender,
disabilities, self-concept, and so
forth, you cannot randomly assign
students to different categories of
these variables.
Independent variables such as
these are called attribute
independent variables. An
attribute variable is a
characteristic that a subject has
before a study begins.
Ex Post Facto Research

o from Latin for “after the fact”

o conducted after variation in the variable of interest has already
been determined in the natural course of events
o sometimes called causal comparative because its purpose is to
investigate cause-and-effect relationships between independent
and dependent variables
Ex Post Facto Research vs. Experimental Research
Consider the question of the effect of students’ anxiety in an
achievement testing situation on their examination
performance.

v The ex post facto approach would involve measuring the

already existing anxiety level at the time of the examination and
then comparing the performance of “high anxious” and “low
anxious” students.
v The experimental approach could randomly assign subjects to
two exam conditions that are identical in every respect except that
one is anxiety arousing and the other is neutral.
If you wish to reach a conclusion that one variable (X) is the cause of
another variable (Y), three kinds of evidence are necessary:
1. A statistical relationship between X and Y has been
established.
2. X preceded Y in time.
3. Other factors did not determine Y.
Think about this!

1. Why does the current

administration of the Department
of Education prefer randomized
experimental research to ex post
facto research?

2. Why do researchers conduct ex

post facto research?
PLANNING AN EX POST FACTO
RESEARCH STUDY
1. State the research problem, usually in the form
of a question.
Examples:
What is the relationship between variable A and variable B?
What is the effect of variable A on variable B?

The researcher then states a hypothesis about

the expected relationship and defines the
variables in operational terms.
2. Select two or more groups to be compared.
These two groups should differ on the variable of interest,
but they should be similar on any relevant extraneous
variables. They are selected because they already possess
the variable of interest, for example, smoker/nonsmoker
and retained/not retained.
3. Determine whether your question requires a
proactive or a retroactive design.
a. The proactive ex post facto research design begins
with subjects grouped on the basis of an independent
variable (such as father present/ father not present or
retained/promoted). The researcher then compares
these preexisting groups on measurers of dependent
variables (such as self-confidence, mental health, and
academic performance).
b. The retroactive ex post facto research seeks possible
antecedent causes (independent variables) for a
preexisting dependent variable. Whether students
graduate from high school or drop out is a variable
that cannot be manipulated. Therefore, a researcher
would use retroactive ex post facto research to
investigate hypotheses about possible causes (such
as truancy, attitude toward school, ambition.
ALTERNATIVE EXPLANATIONS IN EX POST
FACTO RESEARCH
When investigators can control the treatment (X)
and then observe the dependent variable (Y) as in
experimental research, they have reasonable
evidence that X influences Y.
Ex post facto research, on the other hand, lacks
control of the independent variable and thus has
lower internal validity. If researchers cannot
control (X), they may be led to inappropriate
conclusions.
When interpreting ex post facto research, one
should consider alternative explanations, such as
common cause, reverse causality, and the
presence of other independent variables.
1. Common Cause

In an ex post facto investigation, you must consider the possibility that

both the independent variable and the dependent variable of the study
are merely two separate results of a third variable—that they have a
common cause.
1. Common Cause

If you use a school’s total budget as an independent

variable and cases of diagnosed learning disability as
a dependent variable, you might find a positive
correlation between the two variables.
Does this mean that an increase in total school budget
leads to an increase in cases of learning disability?
A more plausible explanation is that the relationship is
spurious. An increase in school size/number of
children attending could account for both the budget
and the cases of diagnosed learning disability
because funding is tied to the number of students.
1. Common Cause

It is well established that the average income of

private high school graduates is much higher than the
average income of public and parochial high school
graduates.
Does this mean that private schools better prepare
students for financial success?
Or is the difference due to the fact that those families
with enough money to send their children to private
schools are also able to finance their children’s
professional training and set them up in business?
2. Reverse Causality

In interpreting an observed relationship in an ex post facto study, the

researcher must consider the possibility of reverse causality—that
the reverse of the suggested hypothesis could also account for the
finding. Instead of saying that X causes Y, perhaps it is the case that
Y causes X.
2. Reverse Causality

If you find that college students who drink have a

lower grade point average (GPA) than nondrinkers,
you cannot automatically conclude that alcohol
consumption depresses academic performance.
Perhaps bad grades drive students to drink (or any
number of common causes could lead to both
drinking and poor grades.)
2. Reverse Causality

Investigations on the effects of child-rearing practices

have revealed that children who are frequently
punished show more aggressive behavior. Can you
conclude that parental punishment leads to
aggressive children? Or are aggressive children more
likely to be punished?
3. Other Possible Independent Variables

Independent variables other than the one considered in the ex post

facto study may bring about the observed effect on the Y variable; that
is, in addition to X1, other variables, X2 and X3, may also be
antecedent factors for the variation in the dependent variable.
3. Other Possible Independent Variables

The recorded suicide rate in Sweden is among the

highest in the world. Does this mean the Swedish
environment causes more people to commit suicide?
Does it mean that the Swedish people are more suicide
prone than others?

Perhaps there is truth in one or both of these

hypotheses. It is equally possible, however, that the
actual independent variable is the honesty of coroners
in Sweden compared with the honesty of coroners in
other countries.
Think about this!
1. The chief of police finds that the
proportion of Filipino Muslims who
are arrested for traffic violations is
much higher than the proportion of
Filipino Christians in his city. State a
possible explanation the chief
should consider before concluding
that Filipino Muslims are more
reckless drivers than others.
2. Why is reverse causality not a
credible explanation in this case?
A supervisor of student teachers notices that those with pierced
tongues get lower ratings from their mentor teachers than do
other student teachers. The supervisor might hypothesize the
following:
a. The same lifestyle that leads to pierced tongues leads to
poor student teaching performance.
b. Pierced tongues lead to poor health, which leads to poor
student teaching performance.
c. Student teachers who are not doing well decide to have
their tongues pierced.
d. Mentor teachers imagine that pierced tongues indicate
antisocial attitudes and give student teachers with
pierced tongues lower ratings than they deserve.
Which of the previous hypotheses represent the following
possibilities?
1. Reverse causality
2. Cause/effect
3. Common cause
Think about this!
4. Alternate independent variable
PARTIAL CONTROL IN EX POST
FACTO RESEARCH
There are strategies for improving
the credibility of ex post facto
research, although none can
adequately compensate for the
inherent weakness of such research
— lack of control over the
independent variable.
These strategies provide partial
control of the internal validity
problems of common cause and
other possible independent
variables. Among these strategies
are matching, homogeneous
groups, building extraneous
variables into the design, analysis
of covariance, and partial
correlation.
1. Matching

Matching the subjects in the experimental and

control groups on as many extraneous
variables as possible. This matching is usually
done on a subject-to-subject basis to form
matched pairs.

If you are interested in the relationship between scouting

experiences and delinquency, you could locate two
groups of boys classified as delinquent and
nondelinquent according to specified criteria.
In such a study, it would be wise to select pairs from
these groups matched on the basis of socioeconomic
status, family structure, and other variables known to be
related to both choosing the scouting experience and
delinquency.
2. Homogeneous Groups

Instead of taking a heterogeneous sample

and comparing matched subgroups within it,
an investigator may control a variable by
including in the sample only subjects who are
homogeneous on that variable.

If academic aptitude is a relevant extraneous variable,

the investigator could control its effect by using subjects
from only one academic aptitude level.

If gender is a variable to be controlled, you can use only

males or only females in the research.
3. Building Extraneous Variables
Into the Design
It may be possible to build relevant
extraneous independent variables into the ex
post facto design and investigate their effect
through the use of two-way and higher order
analyses of variance.

Suppose an investigator wants to compare the number of

truancies among students who have been through an
attendance promoting program and among students who
have not been in such a program.
If the investigator thinks that ethnic membership and
gender influence the number of truancies, the subjects
could be classified as Christian, Muslim or Lumad and
also as male or female.
4. Analysis of Covariance

Analysis of covariance (ANCOVA) is

sometimes used to partially adjust for
preexisting differences between groups in
an ex post facto design.
Specifically, it adjusts scores on the
dependent variable for any initial differences
on the extraneous variable.
However, because the adjustment is only
partial, ANCOVA does not “solve” the
problem of initial differences between
groups but only reduces it.
A classic example of the problems inherent in matching and
ANCOVA is the Cicirelli (1969) ex post facto investigation of the
effects of the Head Start program.
This study compared the academic achievement of students
who had been in the Head Start program with the achievement
of those who had not been in the Head Start program.
Children who had been in the program were matched with non-
Head Start children from the same neighborhoods on gender,
racial/ethnic groups, and kindergarten attendance. ANCOVA
was used to adjust for differences in income per capita,
educational level of father, and occupational level of father.
The results appeared to indicate that the achievement of the
non-Head Start group was greater than that of the Head Start
group even when scores were adjusted for initial differences.
The authors concluded that the Head Start program was
harmful.
Think about this!
Read the following abstract from a report of
a study using ex post facto research that was
published in 2000 and answer the following
questions:
1. What is the independent variable?
2. What are the dependent variables?
3. How are the dependent variables
operationally defined?
4. Is this study proactive or retroactive?
5. What is the population of interest?
6. How was the sample selected?
7. Is reverse causality credible?
8. Is common cause credible?
9. Which partial control strategy was used?
 • Ex post facto research is used to
investigate relationships when the
Summary researcher cannot randomly assign
subjects to different conditions or
directly manipulate the independent
variable.
• Ex post facto research begins with
subjects who differ on an observed
dependent variable and tries to
determine the antecedents (cause) of
the difference; or the researcher
begins with subjects who differ on an
independent variable and tries to
determine the consequences of the
difference.
 • Although there are many disadvantages of ex
post facto design, it nevertheless is frequently
Summary the only method by which educational
researchers can obtain necessary information
about characteristics of defined groups of
students or information needed for the
intelligent formulation of programs in the
school. It permits researchers to investigate
situations in which controlled variation is
impossible to introduce.
• Attributes such as academic aptitude,
creativity, self-esteem, socioeconomic status,
and teacher personality cannot be
manipulated and hence must be investigated
through ex post facto research rather than
through the more rigorous experimental
approach.
 • The possibility of spurious
relationships is always present in ex
Summary post facto research. Considering the
possibilities of common cause,
reversed causality, and possible
alternate independent variables can
help educators evaluate such research
more realistically.
• Several partial control strategies can
help researchers avoid gross errors in
ex post facto designs, but none can
entirely solve the problems inherent in
those designs.
• Always exercise caution when
interpreting ex post facto results.