Are we Bowling at all? An Analysis of Social Capital in Online Networks

Javier Sajuria*, Jennifer vanHeerde-Hudson*, David Hudson*, Niheer Dasandi* & Yannis Theocharis**

* University College London ** University of Mannheim

Overall Research Question

Is there evidence of social capital in online networks? In which way?

Putnam claims that ICTs are not able to create social capital for 4 reasons:

Face-to-face interactions carry much more contextual information than technology-based interactions
Online interactions foster "cyber-balkanisation"
The presence of a digital divide
The Internet is more about entertainment than communication

Neo-capitalist approaches understand social capital as the relative advantage of a member in a network.

They (Burt 2005) claim the distinction between brokerage and closure:

Brokerage: bridging over a structural hole in a network

Closure: the level of connectedness of a person in a group

Communitarian approaches are more concerned about the aggregate benefits of social capital.

They understand social capital as networks able to mobilise resources and information, based on norms of trust and reciprocity.

Putnam argues that social capital can be separated between bridging and bonding.

Bridging: Inter-group ties

Bonding: Intra-group ties

However, he claims the difficulty of measuring them. In most cases, researchers rely on self-reported data.

Provide a measurement of the presence of social capital in online networks.
Bring together no-capitalist and communitarian approaches by using brokerage and closure as indicators of bonding and bridging social capital

We test observed Twitter networks against network simulations using three widely used theoretical models:

Erdos-Renyi (random graphs): Random networks according to given number of nodes and ties.
Barabasi-Albert: incorporates the notion of preferential attachment (i.e. nodes are more likely to connect to heavily linked nodes)
Watts-Strogatz: similar to BA, but incorporates topological features such as node neighbourhood and triadic closure.

The levels of bridging and bonding social capital formed through online interactions are significantly different than random
The networks formed through online interactions are, on average, less dense and weaker than the theoretical expectations.
In online networks, bonding and bridging social capital operate in coordination, strengthening each other.

Three "Twitter events":

Each dataset was cleaned leaving the sender, the message and all the recipients of the tweets. Then the networks were constructed using R (igraph)

Two measures were used as metrics for brokerage and closure.

Brokerage -> Average Network Constraint Index

Constraint refers to the ability of a node of concentrating connections and being able to move information across the network

Closure -> Average Local Clustering Coefficient

This measure accounts for how close is a node's neighbourhood of becoming a clique.

Almost all t-tests show a significant difference between the observed values and the models, supporting H1.
Correlation coefficients between brokerage and closure are positive (or negative, given the way in which Constraint is measured). This supports H3.

The difference between the OWS and the other two cases in terms of brokerage brings an interesting question about the role of professional brokers
The theoretical models we used are not able to account for the structural features of the online networks under study
Online networks seem efficient to create bonding social capital, which in turn fosters trust and reciprocity
Bridging social capital, on the other hand, does not seem to appear organically in online networks
The interplay between bridging and bonding social capital goes in the right direction.