Intended for healthcare professionals


Dynamic spread of happiness in a large social network: longitudinal analysis over 20 years in the Framingham Heart Study

BMJ 2008; 337 doi: (Published 05 December 2008) Cite this as: BMJ 2008;337:a2338

The Contagion of Happiness

Fowler & Christakis report that people given a questionnaire (see
footnote) to assess their ‘happiness’ tend to have similar scores to their
family, neighbours and friends. As the authors freely acknowledge, this is
not very surprising, since relations and friends share similar
circumstances. However, Fowler & Christakis carry out further analysis
to support their conjecture that there is an actual causal relation
involved in the association, not merely a correlation. They are sure
enough about their conclusion to conclude that their findings have
‘relevance for public health’, a serious and important claim. Have they
made their case?

There seem to me to be three main arguments. The first is that the
survey was taken at several points in time, allowing a longitudinal
analysis. This meant that the changes in ‘happiness’ of a target
individual (the ‘ego’) could be analysed, in an attempt to see if changes
could be related to the present and past state of ‘happiness’ in the
related individual (the ‘alter’). The key data presented in their Fig 4
show that if a ‘alter’ changes in state from unhappy to happy, this
increases the probability that the ‘ego’ will be happier in the second
survey. The probability increase is 0.25 if the ego and alter are friends
living nearby, but close to zero if they are living far way. This is an
impressive figure, but I could not see from the arguments advanced why
this could not mean that nearby friends tended to share changes in
fortune, particularly since far-away friends included those that had left
the town.

The authors try to address this point by pointing out that the
effect is much stronger between strong, mutual, friends than between
asymmetrical pairs in which one names the other as a friend, but the
object of their affection does not reciprocate. The authors state that
‘..if some third factor were explaining both ego and happiness, it should
not respect the directionality of the tie.’ This reasoning is hard to
follow. If the ‘third factor’ or hidden variable were more highly
correlated between mutual friends than between more distant ones, as one
would expect, the findings are exactly what one would see. It needs to be
demonstrated that mutual friends do not have closer socio-economic ties
than those in which the friendship is presumably much weaker.

The third argument is related to the last. The effects of an alter’s
happiness are found to be much increased if they live next door rather
than in the same block (presumably in the US sense of that term). The
authors argue that this demonstrates a direct contagion, rather than the
influence of a hidden variable. They state that ‘…socioeconomic status
probably cannot explain the clustering of happy people as next door
neighbours'. The key word here is ‘probably’. This is a simple question
of spatial scale, which should be addressed by evidence, not conjecture.
One has to assume, of course, that next door neighbours did not get
together to answer the questionnaire together over a few beers.

A key issue is assortative mating
(, which the authors dub
‘homophily’. Happy people may choose happy friends, or as the adage has
it ‘misery seeks company’. The authors state that ‘Alter’s happiness in
the previous exam helps to control for homophily.’ This is a promising
line, but it is not pursued: the word does not appear again in the
article, and in the supplementary notes only to say that the differential
analysis controls for baseline states, and thus for homophily, a rather
opaque statement. A simple version of the assortative mating hypothesis
would that ego and alter happiness just after they form their friendship
should be more related than a few years later, supposing the friendship to
have survived. The contagion theory says that the relation should be
equal. It would be interesting to know the facts.

Aldous Huxley said wittily that ‘several excuses are always less
convincing than one’. The same is true in Science. An accumulation of
individually weak arguments makes only another weak argument. I did not
find a single one of Fowler & Christakis’ arguments individually
compelling. The fact that there are several might mean that they share a
common defect. This would not matter if they had presented their argument
in the traditional scientific way as a conjecture, consistent with the
evidence but not tested by it. Instead, their conclusions state that
‘People’s happiness depends on the happiness of others with who they are
connected’. They go further. Since, they say, happiness has a causal
role to play in health (another unproven conjecture), it follows that
their findings have relevance for public health. Presumably, the idea is
that if we wish to increase someone’s happiness, and thus their health, we
should be trying to make their friends happier.

They go even further, into the realms of abstract philosophy. ‘Human
happiness’ they state, ‘ is not merely the province of isolated
individuals.’ We seem here to be entering Durkheim territory, where
suicidal impulses, rather than coming from individual brains, hover like a
shroud over a society. Durkheim’s hypothesis has been thoroughly
discredited in the case of suicide, so why should we now want to
rescusistate it for ‘happiness?’ Even if the author’s causal hypothesis
were true, it would still not rule out happiness being an individual brain

The real purpose of these conjectures becomes clear in the final
paragraph, when the authors say that their conclusion (the causal effect
of inter-personal happiness on health) ‘provides a conceptual
justification for the speciality of public health’. I take it that this
is a plea for resources to be diverted. The plea has my sympathy, but it
does not belong in a scientific article. As President Obama has recently
and eloquently said:

‘Promoting science is about ensuring that facts and evidence are
never twisted or obscured by politics and ideology’. (Quoted in FT
Editorial, Jan 17 2009).


Happiness is never really defined in the paper, and is certainly not
measured by anything other than self-report, with all the weakness that
such a measure implies. The questionnaire is described as being validated,
but it would be useful to know if the validation extends beyond its
predicting depression.

Competing interests:
None declared

Competing interests: No competing interests

24 January 2009
Michael J Morgan
Professor of Visual Neuroscience
Applied Vision Research Centre, City University London