Share

Theory Development

15. Summary

A recent medical study found an association among teenagers between having no breakfast and sexual activity. Did they discover a new causal fact? Or is this a case where the fact predicts but does not explain? How do you make sense of it? In this chapter we have described two ways.

One is to think of a more abstract connection between things associated with the two variables sex and breakfast that could explain it: it might be that teenagers with less parental supervision are more likely to have sex and that having breakfast is a proxy for such things as parents getting up early and sending the kids off to school, and this is a proxy for parental interest and concern. Parental interest and concern might influence teenagers choices, including those involving sex. We can test this more abstract relationship, which makes more sense than the no breakfast/sex connection, but may not in fact be true, by seeing if it applies to other choices and other measures of parental interest and concern. If it worked in these other cases, we would have some reassurance that this is what is going on in this case. If we had a lot of connected relationships that worked in a way that fit this basic idea, we would be even more reassured. 

The second approach would be to break down this relationship between no breakfast and sex, using data and our background knowledge. We more or less know that breakfast doesn’t have any direct causal link to sexual activity. The relation doesn’t exist for older people.  So what is going on here? Just by breaking the relation down to smaller subsets, using the kinds of categories that allow us to apply our background knowledge (such as income level of family, ethnicity, etc.) we are probably going to see changes in the relationship. We may be able to get data on things that our background knowledge would suggest might have to do with the relationship– such as the answers the teenager gives to the question “is your mother annoying?” It might be that the data show that facts about family relations more or less correspond with the relationship, and that breakfast is just a confounder. Or it might be that the practice of family breakfast is associated with ethnicity, and facts about the culture of a particular high-risk group explain both skipping breakfast and sexual activity. The data will enable you to choose– and you can keep breaking down the categories until you have results that fit with your background knowledge and make sense.

Is this an airtight, mechanical process? No. The results of statistical research are often puzzling. It is often hard to get the kind of data that clearly distinguish between alternative hypotheses. Confounders and hidden causes are everywhere. Welcome to the world of social research!