Child Visitation and Performance: The Evidence

• Author: William S. Comanor
• Position: Professor of Economics at the University of California
• Pages: 763-770

Page 763

The author is Professor of Economics at the University of California, Santa Barbara and Professor of Health Services at the University of California, Los Angeles. He served previously as Chief Economist and Director of the Bureau of Economics at the U.S. Federal Trade Commission. In preparing this paper, he thanks Mr. Niklas Hunter for very helpful research assistance.

I Introduction

In her paper on the topic of child custody, Margaret Brinig asks whether joint physical custody is really in the best interests of the children.¹ To answer this question, she explicitly adopts an economic approach and employs an econometric model. Unfortunately, she gets the econometrics wrong, so her conclusions do not follow. Indeed, if her empirical tables indicate anything, they offer weak support for the joint custody regimes that she seeks to disparage.

At the outset of her paper, Brinig observes that two states, Iowa and Maine, have recently moved toward a presumption for joint physical custody.² And while most states do not provide an explicit preference, many encourage some form of shared custody.³She questions whether this trend is really in the best interests of children. While theoretical or psychological arguments can be made on both sides of this question, she acknowledges that the conflicting positions can be resolved only by looking at the evidence. On this point, I agree.

Where my differences arise is how we should interpret her statistical findings. In particular, what does her regression analysis really show? In this comment, I accept her statistical tables but ask what conclusions should be drawn from them.

II Brinig's Empirical Methodology

Brinig's objective is to discover whether children fare better with joint than single custody. For this purpose, she employs a national sample of junior high school and high school-aged Page 764 children,⁴ but limits its scope to those living with their mothers and not their fathers. Based on her analysis, Brining reaches the following conclusion:

[T]he worst situation for children was when they visited their fathers infrequently. Otherwise, however, there was no increase in custodial time that made a statistically significant difference . . . . The only exception to this rule appears . . . where children seemed less likely to engage in alcohol, tobacco, and marijuana use if they stayed with their father several times a month.⁵To support these conclusions, however, Brinig merely refers the reader to her statistical tables and does not even describe her findings. Indeed, there is little discussion in her article of either her empirical methodology or results. An important function of this comment is to fill that void.

A striking feature of Brinig's approach is that it is limited to adolescents and ignores younger children. Yet, a parent's input is likely to be greater at younger ages. For this reason, her analysis cannot indicate the full effect of custodial arrangements.

Furthermore, Brinig's data fails to account for the length of time under which different custodial arrangements have lasted. We do not know if the children whose performance is described have lived with both parents for long periods of time, or never knew their father. This consideration is ignored in her statistics.⁶What we have instead is a "snapshot" of living arrangements, where there is likely as much hidden as revealed.

Even within this limited data set, Brinig restricts her approach still further. Out of the many possibilities included in the data, she deals with only four measures of child well being. These measures are: 1) depression, as measured on a psychological scale, 2) the number of times per month drugs, alcohol and tobacco were used, 3) "juvenile delinquency" as measured on a fifteen-point scale, and 4) morbidity as reflected by the chances of dying or being killed when young.⁷ An obvious question is why she selected these four variables from the large number of possible alternatives. For example, she did not include teenage pregnancy rates or physical activity or scholastic achievement. Page 765

The major difficulty with her analysis, however, is not with the dependent variables she selects but rather with her explanatory variables. Her primary variables reflect the number of times during the prior year that the child stayed overnight with his or her biological father.⁸ As might be expected, nearly half of the valid observations in her sample indicate that they never did so. On the other hand, only fifteen percent report that they stayed overnight at least once a week, while an additional seven percent report staying overnight about once a month.⁹ Thus, more than three-fourths of the respondents in her sample state that they rarely if ever spent long periods of...