Dueling websites on the internet argue over the meaning of marriage in our lives and whether the current generation of young adults is missing out on a life of fulfillment by pursuing casual relationships into their 20s and 30s. The "promarriage" cite is at Ruger's National Marriage Project, at http://marriage.rutgers.edu/. A more skeptical site, The Alternatives to Marriage Project, can be found at http://www.netspace.org/atmp.
One piece in this debate are studies, from both the US and Europe, showing that couples who live together before marriage divorce at higher rates than couples who marry immediately. Setting aside the very important question of whether marriage is a route to personal fulfillment, I will explore in this essay reasons to be cautious about inferring causality from studies which show a correlation between cohabitation before marriage and probability of divorce.
For the past six years I've used the cohabitating statistics to illustrate to my developmental psychology class the concept of spurious correlation. This is also called an illusory correlation, or "third variable problem" because there may be a third variable, other than the variable "did/did not cohabitate", which is the actual causal factor.
I introduce this in class this by asking students to come up with non-causal explanations for the statistically valid (and cross-culturally replicable) finding that people who cohabitate divorce at higher rates than people who don't. I present on a slide from a Cosmopolitan article which blithely assumes causality, in that the writer asks, "Should you live together before you get married?"
Students come up with a wealth of reasons to be skeptical, including:
The "third variable" problem is not immediately obvious to everyone in the lecture hall. Some students initially ask if the inferential problem here is that we the samples weren't followed long enough, or if what we really need is more information about the two groups.
But with a few minutes thought students readily come up with ideas about "third variables" which get at the following point:
People who choose to marry without first cohabitating are already a different group of people than those who decide to cohabitate first.
When we draw conclusions from a study, everything has to be the same between the two groups *except* for the variable of interest (in this case, cohabitating or not). Scientists achieve this by randomly assigning members of a population to treatment conditions (marriage or cohabitating). In psychology, we often don't have "experimental control over the variables of interest." The results is that "correlation is not causation" is one of the mantras psychologists learn to repeat.
So what do researchers do? They try to match two groups on as many of the other potential variables as possible. This would include issues that the callers brought up, such as finances and religious beliefs. But even this won't work, as a thought experiment will make clear.
Scientists often work from experiments of nature. We can imagine an experiment of nature which accidentally ends up effecting random assignment. Imagine a group of couples desiring marriage. After they've decided to marry, a political or natural disaster happens and communities are thrown into relocation camps or disaster shelters. The disaster happens at a time such that some couples have just been married, but other couples' weddings were still pending. In the relocation camps (or shelters) there is considerable privation and confusion, and the couples end up living together regardless of whether they had been able to have a marriage ceremony performed.
What is our prediction? Couples who didn't participate in a marriage ceremony divorce at greater rates? My intuitions don't go this way (or at least not strongly -- let me know if yours do.)
You might say, but in this "natural experiment" everyone *wanted* to be married. They were going to be married and then were kept from it artificially. That is why both groups look the same, and why we don't see an effect of which group happened, by chance, to be married before being thrust together in a common living space.
Ah. Exactly. So the important variable is *desire to be married* or *valuing the state of being married* NOT cohabitating before marriage.
Would Cosmo have grabbed any attention in their article had the headline read, "Couples who value marriage are less likely to get divorced than couples who don't value marriage"?
We would like there to be easy answers to why marriages work, like, "don't live together." One piece of this seems to be "value marriage." The cohabitation studies may well be making this point. The pro-marriage theorists could mention these studies in terms of whether the desire to cohabitate may be a *marker* of discomfort with marriage. But we should refrain from drawing simplistic conclusions about living together before marriage. There is currently no reason to believe that living together in and of itself -- independently of one's values -- is a risk factor for later divorce.