about the whole population from results obtained from an experiment with the
sample. In this regard the sample is representative of the population but is not
perfectly so. This is another reason that results from an experiment on humans
are tentative.
Correlation Method
Science gives the most reliable results when controlled experiments are
possible. In many cases, though, controlled experiments are not possible. In
these circumstances a useful method of investigating causation is to look for
correlation. Two variables are correlated when they move in concert. They
might be positively correlated in that they rise and fall together, or they may be
negatively correlated in that as one rises the other falls, and vice versa.
In the simple case the two factors will move in perfect concert. Realistically,
though, the concert will not be perfect. This happens because outcomes are
often a result of the operation of several factors whereas simple correlation
measures how one factor changes compared to one other.
To illustrate correlation, assume that an economist wants to know if there is a
relationship between the productivity of a factory in a town and the success on
the pitch of the town’s football team. First, they would need a definition of
productivity and a definition of success for the football team. Then they would
make a graph. On one axis they would put the productivity of the factor, while
on the other axis they would put the success of the football team.
This graph will indicate if there is correlation and also how close it is. If there
is correlation it suggests that there is some connection between the
two
factors. The question then becomes what sort of a connection is it. While
correlation suggests a connection, it has nothing to say on what the connection
is. This conclusion is captured in the maxim that correlation (between A and
B) does not equal causation (between A and B) .
In fact there are several possible connections. To explain these we can label
the correlated items as A and B. These are the possibilities. (i) A causes B. (ii)
B causes A. (iii) In a less simple case, A and B mutually cause each other. (iv)
In a second less simple case, the causal relationship does not lie between the
two correlated factors, A and B, but comes from a third factor. For example,
when tar on the roads softens and melts people buy more ice creams. It is not
a case of the softening tar causing the purchase of ice creams nor the purchase
of ice creams causing the tar to soften. Instead a third item, an increase in air
temperature, causes both the tar to soften and melt and people to purchase
more ice creams.
 |
 |
 |