What we found from studying 6 billion friendships

Dogs, social capital, and how we can become more upwardly mobile

In the mid 1980s my mum decided to build a house in our back garden. Apparently it's what she would have done if she had still been living in India. (A conceit used to justify all manner of things throughout my childhood).

Rather than pay for a builder to do it, she decided to manage the project herself. Like most British infrastructure projects, it was dogged by legal challenges, unreliable builders, poor weather and cost overruns.

So after selling our home, we ended up temporarily ‘homeless’ for a year, living on a council estate on the outskirts of Burnley. It wasn’t the most welcoming of environments, and if the shift workers next door didn’t start off as racists, my sister’s violin practice soon radicalised them.

It was here that I developed a novel sociological theory that I have subsequently tried to promote at various attempts but to no avail. And now, thanks to the joy of an unfiltered newsletter, I am delighted to share my theory on dogs and social capital.

Social capital, a concept developed by many, but brought to prominence by Robert Putnam - author of Bowling Alone - refers to the social networks and civic activity that builds trust and cohesion, and connects people to opportunity. While we tend to focus on human, financial, or physical capital when analysing economic growth or social mobility, social capital is less well understood or studied.

Dogs, it turns out, help strangers become neighbours. Every day, I would take my black labrador Boris - named incongruously after the blonde, tax-avoiding tennis player - for a walk around the estate. He was so large that when he went missing once and was taken in by the police they classed him as a rottweiler. Feared and respected from afar, lovable up close, he was the perfect dog. Thanks to him, I got to know people, old and young, started playing football and cricket with them and felt pretty at home. Meanwhile, my sister was stuck playing her violin, bow-ing alone.

The reason I’m ruminating on Boris’ magnetism is thanks to some fascinating research, published today, by my colleagues at BIT (the Behavioural Insights Team), in conjunction with Meta, the RSA, Neighbourly Lab and Stripe Partners. Sadly, it does not shed light on my pet theory, but I’m sure you’ll find the first tranche of results intriguing.

Economic connectedness

When we analyse why some people earn more than others, we often assume the answer can be found in their parents’ income, occupation or genes, what school they went to, or structural factors, such as geography and discrimination. Of course, these factors all matter and often a great deal. But what about their networks - the people who act as role models, raising aspirations, shaping their norms and expectations, or providing information and job contacts.

In 2022, Raj Chetty, an economics professor at Harvard, produced two fascinating papers trying to bring more empirical evidence to the debate on social capital. Based on a novel dataset derived from Meta, containing over 20 billion facebook friends, he analysed what types of social capital might be most associated with people being upwardly mobile.

Chetty distinguished between three main forms of social capital. First, what you might call ‘connectedness’ - the degree to which you are friends with people of a different age, or socioeconomic status. Second, ‘cohesiveness’ - the degree to which your friends are not just connected to you, but to each other - a measure of whether communities have tightly knit friendship networks. Third, civic engagement, for example, through volunteering.

Chetty found that only one form of social capital mattered when it came to helping people rise out of poverty: connectedness across social classes. While we are more likely to befriend people from a similar social class, people living in places with a high degree of ‘economic connectedness’ - the degree to which people below median earnings had friends above median earnings - are much more likely to rise out of poverty.

In his follow-up paper, Chetty and his co-authors looked at what leads people to have higher or lower levels of economic connectedness. About half of the difference is due to segregation - we are more likely to live in neighbourhoods or go to schools with others of a similar social class. The other half is due to a specific form of homophily that he terms ‘friending bias’ where we tend to seek out friendships with those of a similar socioeconomic background.

The UK story

So what happens when you replicate Chetty’s study in the UK? Researchers at BIT and Meta analysed the Facebook data of 20 million UK residents, and then tracked their future education and employment outcomes through administrative data - the Longitudinal Education Outcomes (LEO) dataset.

The findings have many similarities to the US picture but some differences.

Unsurprisingly, as in the US, higher income people tend to have more friends from similarly affluent backgrounds. However, as the graph below shows, this pattern is not as marked as I’d have predicted, or as stark as the US. The poorest 50% of people in the UK still have about half (47%) of their friendships with high-income people, compared to about 39% in the US.

The research points to the same story as the US on economic connectedness. We should not lump all forms of social capital together, as only cross-class connections are strongly associated with upward mobility. Low-income children who grew up in the top 10% most economically connected local authorities in England earn 38% more per year on average (£5,100) as adults relative to low-income children in the bottom 10% local authorities. This association holds when you control for other factors, such as the level of income, education or health within the area.

While the primary focus of the research is on economic mobility, an additional survey of 4,000 adult Facebook users looked at whether higher social capital is associated with wellbeing. Those who have more friends - particularly from high-income backgrounds - report higher levels of happiness and life satisfaction, lower levels of feeling isolated or left out, and higher levels of trust. These patterns persist even when we control for an individual’s income, age, and gender.

Policy implications

While the research does not establish causality, it does suggest that if we can foster cross-class friendships, it will increase social mobility, trust and wellbeing. This form of social capital therefore deserves far more attention than the occasional ministerial speech on the importance of building communities.

The policy challenge is easy to frame: how do we create shared physical and online spaces - children’s centres, primary and secondary schools, universities, workplaces - that expose people from different classes to each other? And once people are within a shared space or institution, how do we nurture shared activities or other mechanisms that might reduce friending bias?

The research suggests three categories are worth focusing on.

First, neighbourhoods and secondary schools are where most people make their friends so arguably deserve the greatest focus, whether by shifting their composition, or increasing the likelihood that people from different backgrounds will interact with each other. Now that we’ve started to be systematic at measuring social capital in these contexts, we should consciously experiment in reducing ‘friending bias’.

Second, universities, unlike most contexts, are where people from low incomes are more likely to be exposed to people from high incomes. They can be engines of mobility, particularly if their ‘friending bias’ can be reduced. The research ranks universities by both ‘exposure’ and ‘friending bias’ so we should be examining the outliers, like Cambridge that score highly on both.

Third, hobby groups are ‘low-bias’ settings, which means you are disproportionately likely to befriend someone from a different social class. While these are not typically where people make their friends, they could be made more salient. It is rather ironic that a study using Meta data doesn't capture insights about how people can connect across class lines online via gaming communities, Discord, or apps that connect particular communities of interest. The fragmentation of culture into niche, online interests is usually lamented, but there may be an upside in how this acts as a social leveller.

Back to Boris

You might think my final policy suggestion is entirely flippant, but, according to a study in Nature, there are around 13 million dogs in the UK, with 31% of households owning one.

Anecdata suggests that repeated interactions with people walking dogs (particularly among other dog walkers) is a powerful social glue, and one that public policy should take more seriously.

Of course, some dogs are much better at lubricating social contact than others. Around £130m is spent in a typical year on buying dogs, so why don't we tilt this procurement towards more social capital enhancing puppies. Like vehicle excise duty, perhaps a variable rate for different puppies?

Maybe Rachel Reeves can smuggle it into the Spring Statement. Or maybe I’m just barking up the wrong tree.

What I’m reading/watching/thinking

Toffler in China by Howard W French.

I was always intrigued by the American futurologist Alvin Toffler. He’s not referenced much these days, but apparently he’s big in China!

Open Socrates: the case for the Philosophical Life. I’m belatedly discovering what was meant by ‘an unexamined life is not worth living’.

I’m still avoiding Adolescence. Enjoying the escapism of White Lotus instead.