Saturday, August 6, 2011

Selection and Science: Cohabitation Research as the Example

I’m going to build on the last two posts about science by focusing in on a related topic that is of great importance in my own field of study—that is, the study of romantic relationships. I want to start by giving you the definition of a selection effect in social science. I’ll start with an example. While it’s less clear than it used to be, decades of studies show that those who live together prior to marriage are less likely, not more, to do well in marriage. This finding keeps making headlines because the historical pattern has been very counterintuitive. After all, most people believe that one of the best ways to improve your odds in marriage is to live together before marriage. I used to say most “young” people believe this, but now it’s really most people of all ages, including the parents of people who are currently young. (More recently, by the way, it is the presumed absence of this finding that has been making headlines.)

There is a complicated debate going on among social scientists how as to whether or not this cohabitation effect is going away as cohabitation before marriage becomes the norm. I think the reports of its demise are premature. Nevertheless, if you are interested in the current findings, our work consistently shows that people who only cohabit after they are engaged or wait until marriage are at lower risk than those who cohabit before there are clear, mutual plans for marriage. That word “mutual” is pretty crucial, here. Sometime, I’ll come back to why that latter finding is very important and what theory it was predicted by before we started to test for it. Suffice for the moment that we have published many studies now with different data sets showing that cohabiting prior to engagement is associated with greater risks in marriage. Personally, I think it’s best to wait until marriage, but if one does not plan to do that, I think it’s wisest to at least be engaged, first.

Okay, that’s not really what I wanted to write about. It just sets the stage. I want to write about selection effects and here I go. The classic explanation for why cohabiting before marriage has been associated with poorer outcomes in marriage, not better, is selection effects. Selection means that there are characteristics associated with certain people that are both related to the risk you are interested in (in this case, doing less well in marriage) but also more associated with the likelihood of doing something like cohabiting prior to marriage that appears to cause risk. Are you still with me? Historically, one could have looked at all the research showing cohabiting before marriage may not be the panacea it’s cracked up to be and suggest that cohabiting prior to marriage actually harm relationships. The counter view is that the cohabiting itself has nothing to do with the eventual outcomes but that it’s the people who were more likely to cohabit—they were more risky to begin with, and they ‘d have been more risky whether or not they had cohabited prior to marriage.

There is a lot of evidence of selection effects in various types of analyses of risk in social sciences. In the case of cohabitation, those who are most likely to cohabit truly are less religious, less traditional, more likely to come from homes where their parents divorced (or never married), more likely to have a lack of confidence in marriage, more likely to have children from prior partners, have had more sexual partners in the past, and are less educated (just name a few things associated with selection). These factors are associated with more problems in marriage and they are associated with greater likelihood of cohabiting prior to marriage (and a much greater likelihood of cohabiting prior to engagement). The idea is that these folks are already “select” for risk in marriage, and they are also more likely to “select” cohabitation prior to marriage or prior to engagement. That’s a selection effect. When a social scientist uses the term, they mean to suggest that what you might have thought was the element causing risk is not causing the risk at all. Selection variables are actually causally linked to risks and outcomes. What I mean here is that selection effects are not the part of the risk that is causal as in related to the behavior in focus—cohabiting before marriage or before engagement.[1] It was already baked in and it just looks like that variable is causal.

Here’s the problem with selection and here’s where this post links to the past two posts about science and materialism (the view that everything that is, is material; and if enough of what’s around is was measured accurately and analyzed correctly, you could explain anything). I believe that selection has the same tie as much thought in other areas of science to the logical extension that people have no free will. You may think I exaggerate, but keep thinking it through. It is where the assumptions lead. In this case of cohabitation, you might think that you decide whether or not to cohabit before marriage or engagement, but in reality, selection says that such things were pretty determined long ago by the behavior of your parents and the setting of your birth and upbringing, and the opportunities these things provided or failed to provide.

My colleague Galena Rhoades and I (along with our colleague Howard Markman) have shown in numerous studies that, even if you “control” for scads of selection variables (scads is scientific speak for a large number), you still can show a clear risk for cohabiting prior to engagement. With selection massively controlled for, such findings provide some evidence suggesting that there is something about the experience of cohabiting that is causing risk on top of the selection into the risky pathway. Ah, but hold the fort (I suppose that means, don’t abandon your position too quickly). If you are a modern day scientist, and you deeply believe that better measurement ultimately causes more thing to be explainable, you would not take studies controlling for known selection variables as evidence of the behavior being risky but, instead, as evidence that you have not yet figured out all the other selection variables to measure in order to wipe any sense that there could be something causal to cohabiting prior to engagement or marriage. This idea is analogous to a scene in one of the Star Wars movies. Remember when the good guys are “flying” through the sea in a submersible, and a giant fish of a sort starts snapping after their little boat, and then a larger, even more giantish fishlike sea creature comes out of a hole to bite down on the large fish pursuing the good guys? Crunch. Qui-Gon Jinn says: “There's always a bigger fish.” A scientist often believes the same (often, rightly so, by the way): There is always another variable or set of variables that, if properly measured, explain what is not yet explained or incorrectly explained.

I will stop here for now but stay tuned. Next, I’m going to argue for how science might be able to show that a person could get a sexually transmitted disease without every touching another human being. And I don’t mean from a toilet seat. I mean to use that point to highlight a problem science has with measuring and analyzing free will. For now, I'll keep a lid on it.

[1] I added the clarity here about selection playing a causal role but not in the way people wonder about the effects of cohabitation, in February 2016.

If you are interested in some of our publications related to cohabitation, here are a few citations.

Kline (Rhoades), G. H., Stanley, S. M., Markman, H. J., Olmos-Gallo, P. A., St. Peters, M., Whitton, S. W., & Prado, L. (2004). Timing is everything: Pre-engagement cohabitation and increased risk for poor marital outcomes. Journal of Family Psychology, 18, 311-318.

Rhoades, G. K., Stanley, S. M., & Markman, H. J. (2009). The pre-engagement cohabitation effect: A replication and extension of previous findings. Journal of Family Psychology, 23, 107-111.

Stanley, S. M., Rhoades, G. K., & Markman, H. J. (2006). Sliding vs. Deciding: Inertia and the premarital cohabitation effect. Family Relations, 55, 499 - 509.

Stanley, S. M., Rhoades, G. K., Amato, P. R., Markman, H. J., & Johnson, C. A. (2010). The timing of cohabitation and engagement: Impact on first and second marriages. Journal of Marriage and Family, 72, 906-918.