If association and causation are different, as indeed they are, then it is not immediate that their strengths should be measured by the same quantities, as current practice does. This paper gives a few reasons why even effects need to be measured not one, but in several different ways. It is argued, what seems obvious but not reflected upon in current practice, that the research or policy question and the mode of data collection should determine the choice of the measure used. In particular, a measure of effect which avoids Simpson’s paradox, the cross sum ratio, is discussed and compared to the cross product ratio. The results illustrate the need for further research to understand how effects should be measured in different situations.