SOCB05 – Lecture 2 – May 16 2013
Research Design and the Logic of Causation
The Logic of Causation
Correlation does not equal causation
Not really interested in finding out if a specific thing will happen, but the
factors that will make it more likely for it to happen.
Scientific explanations rest on the idea that events and conditions have causes.
Leaves room for outliers, or things that go against the general pattern.
Probabilistic in nature.
o X is not always followed by Y, but the presence of X makes Y more
o A change in the independent variable, will lead to a change in the
o The cause
o Impacts other variables
o The variable that is depended on
o The effect/ explained variable
o Being impacted upon
o Depends on another variable
Variables vs Attributes
Attributes: Describe the intensity, magnitude, and nature of the variable.
o The values that a variable takes.
o How it is measures
A variable must have at least two attributes
If attributes do not vary, we cannot talk about a variable but a constant
o “People with lower levels of education are more likely to be more
Warning: Do not confuse attributes with variables
Low level of education is not a variable
More prejudice is not a variable
o “In the U.S., Republicans are more likely to be in favour of the death
penalty than Democrats”
I.V. : Political Affiliation
D.V. : Opinion towards death penalty.
NOTE: Variables are not inherently dependent or independent
Conditions for Causation
Idiographic: Complete, in-depth understanding of a single case. Nomothetic: An attempt to find independent variables that account for the
variations in a given class/category of phenomenon.
o More common in sociological research because sociologists aren’t
really looking for individual cases.
What do we need for a relationship to be causal?
o When 2 variables are observed to be related.
o It is never perfect. Outliers always exist.
o Has two forms
Change in one variable is associated with change in a second
When an attribute of one variable is associated with an attribute
of another variable.
o Cause has to be before the effect
Example: Buying a fancy car does not cause one to make
enough money to afford one.
o Can be problematic because sometimes we cannot tell what happens
o Observed correlation between two variables that cannot be explained
by a third variable.
The incorrect inference of a causal relationship between two
variables where the relationship is in reality only accidental
Despite two variables being correlated and one preceding the
other. It may be possible that neither is a cause of the other.
Length of hair and grades in SOCB05
Length of hair is correlated with academic success in the class
o Temporal Order
Starting the class with long hair precedes getting a good grade
in the class.
o BUT there is a spurious relationship because there is a third variable,
which would be gender. Which may explain why people with longer
hair do better.
Different from spuriousness
The original relation between the two variables is explained by a third variable
that acts in between the two.
Example: Drinking doesn’t lead to pregnancy. But drinking does lead to
higher chances of unprotected sex. Which leads to pregnancy.
o Drinking doesn’t exactly cause pregnancy, but it can help it..
Social Science vs. Lay Science
Complete Causation? o SS: More likely to say that something is “one of the causes” rather
than saying that something is the “only cause”.
o SS: Likes exceptional cases, but it doesn’t change the pattern.
Majority of Cases?
o SS: Talk about how people are more likely to do something.
Units of Analysis