Causal Analysis

Whether causal relationship can be established between two phenomena is highly controversial as a philosophical issue. According to David Hume, causal connections among real events cannot be perceived. Bertrand Russell even suggested that the term causality be banned in scientific discourse. Academic researchers, however, simply cannot help asking why and searching for causal explanations. Then Herbert Simon proposed that discussions of causality be restricted to our model of the reality rather than to the reality per se.

The desire to find causality is motivated by at least three potential benefits. First, causal explanation is believed to transcend time and space and therefore has a much wider scope of applications. Second, causal connections constitute the foundation of good control of the interested phenomena, which is especially important for policy making. Third, most causal statements are subject to the laws of logic and therefore more rigorous than they would be if no logic were involved.

It is unjustified, however, to think that causal analysis is always desirable and superior to other types of analysis. Sometimes, it is perhaps more important to know what has happened than why something has happened, because it is either less important or extremely difficult to ascertain the causal mechanism. Carefully produced and interpreted, descriptive statistics are sufficient in many situations. Furthermore, causality is not a prerequisite of precise prediction. For example, in demography and econometrics, making an accurate prediction is more important than identifying a causal chain or measuring a causal effect.

Finally, to discover that A causes B is beyond the capacity of statistical analysis. Causal mechanisms are either found through heuristic methods, such as experiment and observation, or simply derived from current theories and knowledge. It is only after the causal relationship is proposed that statistical techniques can be applied to measure the size of the causal effect.