13.3.07

HOW OTHERS SEE US. The Badger Blog Alliance's Lance Burri tackles an abstract of an article in Child Development.

The abstract begins with a description of the sample.
Using a representative sample of low-income, primarily minority adolescents (N=647, aged 10–14 years at Wave 1), this study examined bidirectional longitudinal relations between nonresident father involvement, defined as contact and responsibility for children's care and behavior, and adolescent engagement in delinquent activities.
As passes through panel data go, this is a relatively small sample. Economists who work with the Panel Study on Income Dynamics and other such longitudinal databases ("longitudinal" refers to observing a number of cases at different times; if the cases are the same individuals or businesses at each time, it's a "balanced" panel) get nervous if their sample size drops below 2,000. I've managed to make do with 52 power companies or 34 steel mills.

Mr Burri quotes the second sentence, with emphasis added.
Autoregressive and fixed effects models found that higher nonresident father involvement predicted subsequent decreases in adolescent delinquency, particularly for youth with initial engagement in delinquent activities.
His response: "duh." What the abstract is really telling us, however, is that the seemingly innocuous decision of a father to get more involved in his kid's life once the kid has had a scrape with the law introduces all sorts of measurement difficulties that the "autoregressive" (meaning the error you committed last year influences the error you commit this year) and "fixed effects" (meaning we can generalize about people with similar characteristics) models attempt to cope with. There's also a phenomenon called "self-selection" that can arise in problems such as this.

The abstract concludes,
As adolescent delinquency increased, so too did father involvement, suggesting that nonresident fathers may increase their involvement in the face of adolescent problem behavior, with this pattern driven primarily by African American families.
Mr Burri notes,

That sounds like good news, and at odds with the usual wisdom. I wonder how true it is? And under what circumstances?

Sorry, I didn’t read the whole study. I know. Bad blogger. Bad. But come on, that was the abstract, and it was barely comprehensible.

Thus one difficulty in making use of academic research in formulating public policy. The conclusion is not that out of line with "the usual wisdom." Put simply, it means "Absentee dads get more involved after their kids have a brush with the law. The propensity of African-American dads to get involved in that way is somewhat greater." In a "low-income, primarily-minority" sample the frame of reference is "compared to other low-income dads belonging to other minorities." Left unsaid is whether greater involvement by those fathers before the kids run afoul of the authorities would keep those kids out of the juvenile justice system, but that's even harder to tease out of the data.

And the "barely comprehensible" nature of the abstract follows from the nature of the research. The authors have to emphasize the use of state-of-the-art methods of statistical inference in order to pass muster with the referees, who, if they are conscientious, will be more interested in the methods the authors used to cope with the measurement errors than they will be with the political reliability of the conclusions.

Academic abstracts are sometimes like that. Here, a relatively turgid abstract introduces a simple message: absentee dads get more involved after their kids get into trouble. That's called "putting out fires." Sometimes a relatively simple abstract introduces a complicated message. Consider "All elliptic curves are modular, thus xn + yn = zn has no nontrivial solutions in integers for n>2." The proof runs some 300 pages.

No comments: