Divorce as the Right to Exit

The leading cause of divorce is marriage.

— An ironic aphorism

Divorce is usually discussed as a breakdown: there was a marriage, it was supposed to continue, and something failed. There is truth in that description, but as an institution divorce is built in a more complicated way. It registers a defeat and grants the right to leave a cooperation that has become unprofitable, dangerous or impossible. Any long union creates shared assets — housing, children, loans, habits, family ties and plans — so leaving never amounts to simply disappearing. The legal regime of divorce doesn’t cancel past obligations; it defines how to close them out. Historically divorce was often forbidden, or available to one side only. From the outside that looked like protecting the family: if leaving is too easy, one partner can walk out after the other has invested in children, a home and a dependent position. A difficult exit protects the weaker party from being abandoned, and at the same time protects the stronger one from the consequences of abuse. If leaving is impossible, a bad marriage stops being a union and becomes a locked room.

“After everything I’ve put into this, I can’t just leave.” The sentence sounds like a statement of loyalty. Sometimes it describes an arithmetic error. Sometimes it describes a perfectly real cost of separation. The two cases look identical until past expenses are separated from future ones.

The ticket is already bought

In the classic experiment, people were more likely to agree to continue an unprofitable project if money had already been put into it. Past costs affected the decision even though there was no way to recover them. Economic logic requires comparing future options only. If continuing brings ten units of harm and leaving brings five, the hundred already spent doesn’t change the choice. It has vanished in both scenarios. In a relationship, the unrecoverable items can be years, effort, a wedding, a move and attempts to repair a conflict. Continuing doesn’t bring them back. It adds new years to one of the possible futures.

But the advice “forget everything you’ve invested” is too crude. Investments can change the future.

Investment as a tie, not an argument

Children in common, property, friends and habits create real benefits and costs ahead. These are not sunk costs. They determine what happens after a separation. Ten years of marriage do not, by themselves, make the eleventh sensible. But over ten years a system of mutual support may have appeared that the person will lose. There may be children whose care will need rebuilding. One partner may have lost a career, making the transition expensive. What has to be cleared out of the decision is the wrong way of using history, not history itself. The past matters to the extent that it has changed future trajectories.

The difference shows up in two similar statements. “I’m staying because I’ve already spent fifteen years” — sunk costs. “I’m staying because leaving now would put me out of a home and require a complicated childcare arrangement” — the future cost of the transition. In practice the two reasons often come together.

In Caryl Rusbult’s investment model, commitment to a union grows with satisfaction, accumulated investments and the weakness of the alternatives. A long shared history really can make a union more valuable: children, friends, the skills of living together and mutual help all appear. The same history can hold people in place when current quality is low, because leaving destroys part of those specific investments. The past affects the price of the decision, but doesn’t settle it automatically.

The right to exit changes more than the number of marriages that end. The staggered introduction of unilateral divorce across US states made it possible to evaluate the reforms in state panel data. The authors linked them to declines in several indicators of family distress, including female suicide, domestic violence and spousal homicide. The long-run effect on the overall divorce rate turned out to be relatively small and largely temporary. The design supports an explanation running through a changed outside option and bargaining inside the unions that survived, but it doesn’t observe that bargaining directly.

Sometimes the decision to postpone a separation has value. A couple can go through treatment, wait out a child’s school year, save money, or check whether behaviour changes after a specific agreement. Waiting becomes an option: for a limited price, a person gets extra information and keeps the ability to decide later. An option is useful if the deadline and the criterion are defined. “Let’s see how it goes” with no conditions turns into indefinite renewal. Each new period creates new investments and makes the next exit harder. A good trial period requires observable criteria: the lying stops, duties are redistributed, treatment begins, and conflict can be discussed safely. Change is tested by behaviour, not by the intensity of promises.

The end of the trial period

A limited wait has an unpleasant condition attached: it has to end at some point. If the criterion was set in advance and wasn’t met, another extension no longer collects new information — it postpones the use of old information. The partner didn’t start treatment, the lying didn’t stop, the load wasn’t redistributed, discussing conflict is still unsafe: the trial produced a result, even if not the one hoped for. The ability to push the deadline back again sometimes looks like patience, but economically it resembles an option that only one side is allowed to extend indefinitely for free. Such an instrument stops limiting risk and starts serving the avoidance of a decision.

This doesn’t turn the criterion into an automatic order to leave. Over the course of the test, health, children, housing, available support or the cost of the transition itself may have changed. But then you have to admit honestly that the problem has changed, and re-evaluate both trajectories; you can’t pretend the old test is still running. There is no universal exit date, but there is a checkable boundary: continuing no longer produces information, doesn’t change the conditions, and rests mainly on the hope that next month will, for some unknown reason, be different.

People often compare the pain of separating now with the familiar state of the relationship now. That is an unequal comparison. Leaving contains a transition period; continuing has a cumulative effect. What needs comparing is the two trajectories in one year, three and five. One contains a sharp loss and possible recovery. The other contains familiar stability, which may improve, stay as it is, or get worse. Such a forecast is inevitably wrong, but it beats trying to prove the value of the past by continuing indefinitely. Closing a project doesn’t devalue what has already been lived.

The limit of optimising the exit

There are situations of violence, threat and coercion where the model of a joint project stops being appropriate. The problem changes from maximising the couple’s welfare to one person’s safety. Negotiations and trial periods can increase the risk. In less dangerous cases, economic language helps separate two traps. The first is staying in order to justify an unrecoverable past. The second is ignoring the real future cost of leaving under the banner of rationality. The calculation includes the expected quality of continuing, the probability of change, transition costs, children, housing, health and available support. It contains no automatic answer of “leave” or “stay.”

The number of years lived doesn’t prove the value of the next one. It only reports how much history can no longer be changed. The decision belongs to the future.

Main sources

Arkes, H. R., & Blumer, C. (1985). The psychology of sunk cost. Organizational Behavior and Human Decision Processes, 35(1), 124–140.

Rusbult, C. E. (1983). A longitudinal test of the investment model. Journal of Personality and Social Psychology, 45(1), 101–117.

Stevenson, B., & Wolfers, J. (2006). Bargaining in the shadow of the law: Divorce laws and family distress. Quarterly Journal of Economics, 121(1), 267–288; Wolfers, J. (2006). Did unilateral divorce laws raise divorce rates? A reconciliation and new results. American Economic Review, 96(5), 1802–1820.