In an earlier post of mine, I took Slate's Emily Yoffe to task for her ill-informed and intellectually dishonest characterization of the research about the impact of single motherhood on children. Over at the Freakonomics blog, economist Justin Wolfers looks at two new studies that concern a similar subject: the economic impact of marriage and divorce. Viewing these studies side by side is highly edifying. Call the the pair of them the Goofus and Gallant of divorce research. The Do-Bee and the Don't-Bee. Or what you will.

One of the studies is a scholarly paper in a peer-reviewed journal which seeks to disentangle the question of whether, in Wolfers' words, "divorce causes lower income, or whether lower income (or its correlates) cause divorce." The authors, Elizabeth Ananat and Guy Michaels, attack the problem of determining the direction of the causality by using what is called an instrumental variables approach. An instrumental variable (IV) is a variable which is correlated with something that is believed to be causing the outcome, but is not believed to be correlated with other unknown and unmeasured factors which may also be causing the outcome. In other words, it needs to be fairly random in nature.

The paper Wolfers looks at wants to measure the impact of divorce on income. To do an IV analysis, the researchers needed to find a factor that is correlated with divorce, but is not correlated with other unmeasured factors that may also have a causal impact on income (such as, for example, a person's health status). The IV they settle on is the sex of the first-born child. Research strongly suggests that if the first-born child is female, a couple is more likely to divorce, all other factors held constant. Therefore, the sex of the first -born seems to be a good IV, because while it does impact the likelihood of divorce it does not seem to be related to unmeasured factors that may have an impact on income.