Wrong About What Will Make You Happy

Ask someone how they would feel two years after becoming paralysed from the waist down. Then ask how they would feel two years after winning a large lottery prize.

The predictions are confident, and they are wildly apart, and in 1978 Philip Brickman and colleagues went and asked people in both situations.

Both groups had moved substantially back toward their prior baseline. Lottery winners were not much happier than controls and reported taking less pleasure in ordinary activities. Accident victims were less happy than controls and by nothing like the margin anybody predicts.

The study is small and has real methodological limits, and it has been criticised properly. What has survived, in much larger longitudinal datasets following the same people through marriage, divorce, disability, promotion and bereavement, is the central observation: people adapt to major life changes far more and far faster than they expect to.

Adaptation is not complete, and the exceptions matter. Unemployment appears to leave a lasting mark even after re-employment. Chronic pain and noise do not fade. The death of a spouse takes years. But the systematic gap between the forecast and the outcome is enormous, and it runs in the same direction almost every time.

Dan Gilbert and Tim Wilson named the error the impact bias: we overestimate both the intensity and the duration of our future emotional reactions. The duration component is the larger one. Asked how long a breakup will hurt, people say months and mean it; the data say weeks.

This would be a curiosity except that almost every consequential decision anybody makes is a prediction about how they will feel afterwards. Should I take this job, move to this city, marry this person, have this operation, buy this house? The estimate you are worst at is the one the decision turns on.

So why is the error so consistent? Three mechanisms, and each suggests a partial fix.

The first is focalism. Asked to predict your happiness a month after a bad outcome, you picture the outcome. You do not picture the month. The work, the meals, the weather, the things that will actually occupy your attention. The forecast is about the event, and the experience will be about the life it happens inside.

The fix follows directly and works: asking people to sketch a diary of a typical future day before making the prediction reduces the error measurably, because it forces the rest of life into the estimate.

The second is the psychological immune system, which is Gilbert's term for the rationalising machinery that makes bad outcomes bearable. Rejected from the job, you notice that the commute was awful and the manager seemed difficult. Left by a partner, you gradually recall the things that did not work. This machinery is reliable and it operates largely outside awareness, which is exactly why people leave it out of their forecasts. You cannot anticipate a consolation you have not yet constructed.

An interesting consequence is that it works better on large blows than small ones. A major setback triggers the full rationalising apparatus; a minor irritation is not serious enough to activate it and simply stays irritating. People are frequently more durably bothered by a small slight than by a large disaster, which nobody predicts about themselves.

The third is that adaptation itself is invisible from the outside of it. You cannot simulate the process by which something stops being noticeable, because noticing is what your simulation consists of.

That last point explains the most practically useful finding in this literature, which concerns what does not fade.

Adaptation is fast for a change in level and slow for a repeated irritation. A larger house becomes the new normal within months and stops registering. A ninety-minute commute does not, because each day's journey is a fresh instance rather than a new baseline, and there is nothing to adapt to except the fact that it keeps happening.

Studies consistently find long commutes associated with lower wellbeing that does not attenuate over years, while the house at the end of it stops contributing. People who trade a shorter commute for a bigger house are making a forecasting error with a known shape, and it is one of the few cases where the research gives specific advice.

Finally, the intervention Gilbert and Wilson found works better than anything people are willing to do.

They call it surrogation. Rather than imagining how you would feel doing something, ask someone who is currently doing it how they actually feel. In their experiments, a stranger's report of their current experience predicted participants' future feelings better than the participants' own detailed imaginings did.

People find this deeply implausible. Asked to choose between a description of a stranger's experience and their own simulation, they take the simulation, on the grounds that they are unusual and the stranger is not them.

They are unusual. The individual differences are real. They are also, reliably, smaller than the forecasting error, which means the stranger's actual data beats your own imagination about yourself, and the reason this advice is so rarely taken is that it does not feel like it could be true.