One of the core beliefs of positive psychology, also known as the psychology of happiness, has been seriously challenged. A major reason for the popularity of positive psychology is their claim that happiness leads to improved health and greater longevity. Not so, according to a new study by a research team headed by Dr. Bette Liu of Oxford University published in the medical journal The Lancet.

The study shows the difficulty of drawing causal inferences from correlational data. The majority of previous studies of the happiness-health hypothesis are correlational. The researchers measured both the participants’ happiness and their health at the same time and found a positive relationahip. However, a correlation between A and B could mean that A causes B, B causes A, or both A and B are jointly caused by some third variable, C. In other words, the previous studies have at least two problems.

1. Directionality. Rather than happiness causing good health, it could be that good health is the reason people are happy. Is happiness a cause of good health, or an effect?

2. Third variables. There an infinite number of other variables which might be correlated with happiness and health and might be causing both. An obvious possibility is social class. Poverty could be making people unhappy and also making it difficult for them to lead healthy lives or obtain adequate health care.

The Liu data come from the Million Women Study conducted in the United Kingdom. Participants were recruited between 1996 and 2001 and were tested three years after their recruitment. At this baseline measurement session, they were then asked whether they suffered from a list of common health problems, and to rate their health as “excellent,” “good,” “fair” or “poor.” Then they were asked, “How often do you feel happy?” The alternatives were “most of the time,” “usually,” “sometimes,” and “rarely/never.” Measures were also taken of how often they felt stressed, relaxed, and in control.

Data were also collected for 13 demographic and lifestyle variables: age, region, area deprivation (a measure of the wealth of their census area), education, whether living with a partner, number of children, body mass index, exercise, smoking, alcohol consumption, hours of sleep, religiosity, and participation in other community groups. In 2012, it was determined whether each woman had died and, if so, the cause. The average duration of the study, from testing to outcome, was 9.6 years. Not all women completed the baseline measurements, and those who suffered from serious health problems at that time were eliminated, leaving a total of about 720,000 participants.

Participants were combined into three groups: happy most of the time, usually happy, and unhappy. About 4% of the women had died by 2012. Controlling only for age, the researchers found a strong relationship between happiness and all-cause mortality. This replicates previous studies. However, poor health at baseline was strongly related to unhappiness. When self-rated health was statistically controlled, the relationship between happiness and mortality was no longer statistically reliable. When all 13 demographic and lifestyle variables—some of which were significantly related to mortality—were controlled, the relationship almost completely disappeared. Happiness was also unrelated to heart disease mortality and cancer mortality once baseline health was controlled.

When they controlled for baseline health, the same results were obtained substituting the measures of feeling stressed, relaxed and in control for the happiness measure. These four analyses are illustrated by these graphs. Notice the almost flat lines hovering around RR = 1. A rate ratio (RR) of 1 indicates that a person in this group was no more or less likely to die than anyone else in the sample.

In summary, the results are consistent with the reverse causality hypothesis: Good health causes happiness, rather than the reverse. As one of the authors, Dr. Richard Peto, said, “The claim that [unhappiness] is an important cause of mortality is just nonsense. . . . Many still believe that stress or unhappiness can directly cause disease, but they are confusing cause and effect.”

Negative results are usually not considered a sufficient reason to reject a hypothesis, because many things can go wrong that can cause a study to fail even when the hypothesis is true. However, the Million Women Study must be taken seriously due to its large sample size and long duration. Certainly it would be better if the study had included men and citizens of other countries. However, there is no obvious theoretical reason to think that the happiness-health relationship holds only for men and not women, and previous studies of  the effect size for men and women are inconsistent.

Although the authors statistically controlled 13 demographic and lifestyle variables, it is impossible to control all possible confounding variables. With these negative results, a critic would have to argue that some third variable is masking the relationship between happiness and health. That is, there would have to be a third variable that is positively related to happiness but that increases mortality and therefore counteracts the expected positive effect of happiness on health.

A comment published along with the study criticized their simple, one-item measure of happiness. However, many previous studies have used the same or a similar measure. More importantly, when baseline health was not controlled, the authors replicated the results of previous studies, which suggests that their measure is adequate. Nevertheless, I anticipate that some positive psychologists will speak philosophically about some deeper meaning of happiness, insisting that whatever they mean by happiness is still a cause of good health, despite these negative results.

There are probably hundreds of thousands of professionals—not just clinical psychologists, but pop psychology practitioners from A (art therapists) to Z (well, yoga instructors)—who promise their clients they will be happier as a result of their treatment, and who implicitly or explicitly promise better health as an indirect result. They will either ignore this study or scrutinize it carefully for flaws. It should be fun to watch.

You may also be interested in reading:

On Obama’s Speech