A Good Example of  Bad “Science”

I recently read about a study, carried out by researchers at Yale and Oxford and published in The Lancet, that discovered that “exercise is more important to your mental health than economic status.”

It took me a moment to comprehend that. Is it a question that people are asking? Is it a question that can provide useful answers? Is it a question that you can even study in a scientific way?

As it turns out, the study was based on an electronic survey. The researchers asked 1.2 million Americans: “How many times have you felt mentally unwell in the past 30 days, for example, due to stress, depression, or emotional problems?”

The participants were also asked about their income and physical activities.

The results?

Those that exercised regularly felt bad, on average, 35 days a year. Non-active participants felt bad 53 days a year.

But also: Physically active people felt just as good as inactive people who earn around $25,000 more a year.

There are many problems here.

First, surveys about behavior and feelings are virtually worthless because people aren’t honest in answering such questions.

Second, even if they try to be honest, they cannot accurately remember their behavior and feelings over long stretches of time.

Third, in concluding that exercise creates happiness, The Lancet(which is becoming famous for publishing bad science) is conflating correlation with causality. This is probably the most common error made in understanding studies.

Let’s look at that study again…

It found that people that exercised regularly reported that they felt happy on more days than sedentary people. But that doesn’t mean – at all – that exercise is the cause of that additional perception of happiness.

You can use the same data to come to the opposite conclusion: that happier people are more likely to exercise – i.e., that feeling happy causes exercise.

There are actually dozens if not hundreds of variables that affect happiness. Exercise is just one of them. And since this study was not controlled (did not consider all the other variables, since doing so would have been impossible), what you have here is correlation, not causation.

With a correlation, you can come to whatever conclusion you want. And I believe that the researchers in this case wanted to come to a positive conclusion about exercise.

But it’s much more likely that the study proved what is common sense: that mentally healthy and happy people tend to take better care of themselves in every way.

You can also come to the opposite conclusion regarding income. That for the same cohort of people (those that don’t exercise), making more money does indeed make you happier.