Group Comparisons: Differences in Composition Versus Differences in Models and Effects—Page 1
Differences in Composition Versus Differences in Models and Effects
This is the first of a series of handouts that will deal with techniques for comparing
groups. This initial handout notes that, when comparing groups, it is important to realize that
groups can differ in two ways:
There can be compositional differences between groups. Specifically, means on the IVs may
differ between groups.
The effects of IVs can differ between groups. A variable might have a stronger effect on one
group than it does on the other. Indeed, the direction of an effect may even differ between
groups. The model that describes one group may be very different from the model that
For example, blacks may have lower levels of education and less job experience than do whites.
As a result, they may tend to have lower levels of income, even if the effects of education and
job experience are the same for both groups. Simple T-tests or ANOVA tests can determine
whether there are significant compositional differences between groups.
Or, blacks may have similar levels of education and job experience, but the effects of these
variables may be less for them, e.g. a year of education is worth less to a black than it is to a
white. As a result, blacks may tend to have lower incomes than comparable whites.
Compositional, or mean, differences between groups on the IVs may suggest that differences on
the DVs are “justified”, e.g. blacks earn less than whites because they are less educated; women
earn less than men because they are concentrated in lower-paying occupations, or have less
continuous service with the same company. Of course, one must then ask what produced the
compositional differences. It may be, for example, that race is a cause of education and job
experience; that is, race may be an
cause of income, because race affects education and
job experience which in turn affect income.
Differences in effects raise questions about why those differences exist. If blacks benefit less
from education than whites, is this perhaps because of discrimination? Or do other factors need
to be considered in the model?
It is important to keep compositional differences and differences in effects separate. Researchers
will sometimes confuse the two, muddling the discussion of why group differences exist. In
particular, researchers sometimes focus a lot on their models, and overlook how important
compositional factors can be in explaining group differences.
Differences in Composition.
Returning again to our hypothetical data from 400 whites and 100
blacks: the following t-tests and descriptive statistics reveal that blacks have lower levels of
education and job experience than do whites. These differences are all highly significant. These
lower levels of education and job experience probably are part of the reason that black income is
also lower than white income.