Koryu Sato: Between Correlation and Causation

I specialize in "social epidemiology," which clarifies the impact society has on people's health. The purpose of my research is to demonstrate whether a causal relationship exists between socioeconomic status as a "cause" and health status as an "effect." The interesting part of this field is that proving causality is extremely difficult. If one were investigating the effectiveness of a new drug, the causal relationship could be correctly measured by conducting a trial where one randomly assigned group is given the new drug and the other is given a placebo, followed by a comparison of outcomes.

On the other hand, when it comes to investigating the causal effects of society on health, it is extremely rare to be able to conduct randomized controlled trials. Therefore, we must rely on "observational data" recorded in its natural state without artificial manipulation. However, in research using observational data, there is a risk of confusing cause and effect. For example, the observation that "people who exercise are healthy" might be a causal effect of exercise, but it could also simply be that people who are already healthy are the ones exercising.

Recently, I published a paper showing that the risk of elderly people requiring long-term care is lower in cities with a high number of library books, which was featured in several media outlets. Looking at the reactions on social media, many people were convinced, but I also saw some comments suggesting that "cities with well-equipped libraries are simply wealthy and have other well-equipped facilities as well." I myself do not believe that strict causality can be claimed from these research results, and I had asked that a note be clearly included in the articles stating that there might only be a correlation, but I felt that this point was not fully conveyed.

Furthermore, I am making efforts to move from mere correlation toward causation. In my analysis, I account for differences such as individual years of education and income, as well as municipal financial strength indices and population density, to remove those influences. I have addressed most of the criticisms I saw on social media. Among those who have studied the difference between correlation and causation a little, there are a certain number of people who immediately conclude that "observational data equals mere correlation." However, the relationship between correlation and causation is not clearly divided like black and white. It would be the pinnacle of scientific literacy if individuals could read original papers and examine the strength of evidence themselves, but perhaps that is asking too much.