Same-sex parenting studies seriously flawed

Saturday, June 15, 2002

Advocates for same-sex couples who have or want to have children claim research shows that the children of such couples are in no way different from the children of traditional couples.

Not so, say researchers Robert Lerner and Althea Nagai, who examined 49 studies in the field in their book, No Basis: What the Studies Don't Tell Us about Same-sex Parenting. Instead, the studies fail to show that any difference exists, which is not the same thing at all.

Is it hard to see the difference? Try a different, and possibly less emotionally charged, example. Stories this week reported astronomers had found a solar system with a Jupiter-sized planet in a Jupiter-like orbit, one of 90 or more extra-solar planets identified since 1995.

Every such study to date has failed to show the existence of an Earth-like planet -- of course, because the instruments available to us aren't powerful enough to find one even if it's there. Wait 10 years.

But, while we're waiting, no one is going about saying, "All the research shows that no Earth-like planets exist." As yet, it shows nothing at all -- and the same is true of the research on same-sex parenting.

We don't know yet whether there are statistically significant differences between these two groups of children. And, as Lerner observes, it's dangerous to be making public policy based on the assumption that we do know.

He points to the example of divorce and single-parent families. It was much the fashion for a while to defend single motherhood as just another lifestyle choice. Remember the scorn heaped on Dan Quayle when he criticized the producers of the show Murphy Brown for putting its lead character in that situation with no hint of the problems nonfictional single mothers struggle to cope with.

Now almost everyone accepts that divorce and single parenthood, as a rule, are bad for children, even if most of them turn out all right. Most children of gay or lesbian parents will probably turn out all right, too. Even if some don't, that is not grounds for denying individuals the right to have and raise children. After all, we don't take kids away from mothers just because they aren't married, nor should we.

But saying there's no difference? No basis.

Lerner faults many of the studies not only for overblown claims, but for serious methodological flaws.

Social-science research intended to demonstrate a causal relationship or the lack of one -- which is the result claimed for many of these 49 studies -- needs to start with a clear formulation of the cause, the effect and the relationship. Only two of the 49 do this, while 18 explicitly start out to prove a negative.

"If your goal is to prove no differences," Lerner observes, "you're bound to reach it, and the poorer research you do, the more successful you will be."

The second step in research is to control for unrelated effects. Researchers need a comparison group; 21 of the 49 studies had none. If they find another variable on which the comparison groups differ, such as education or income, that variable has to be included in the analysis; only one of the 12 studies that found a third variable did so. They have to control for what are called "suppressor" variables, which would tend to mask a difference if one existed. Only one study did this, and even that one failed to report which variables were dropped or added.

Next comes the choice of what to measure and how, and after that sampling. This is a rare population, and hard to reach through random samples. So some researchers used what's called "snowball sampling" -- they found one person, asked that person to identify others, and so on until they got enough cases. Such a group is in no way representative, and nothing you find out about them can be generalized to the rest of the population.

Finally comes the actual statistical analysis -- and here the problem is that all but one of the studies are too small to validate the results they get. They have too little power, like the planet-searchers looking for another Earth. Only two of the studies that analyze the difference between comparison groups have more than 100 subjects. The rest, consequently, have a very high probability of getting the wrong result -- finding no difference -- strictly by chance. In seven cases, that probability is more than 90 percent.

Even if none of the studies singly proves much of anything, surely their sheer number adds up to something? No; the sum of two unreliable numbers is just as unreliable as they are.

If same-sex parenting is problematic, people need to know that, and nobody needs to know that more urgently than the people hoping to embark on it. They deserve better research than this.