Social Networks - Ties, Hubs, Weak and Strong Ties

Groups are bounded: you are in or out (2.2.4). But a great deal of social life runs through connections that are not group memberships at all — the acquaintance who mentions an opening, the cousin who knows someone, the chain along which a rumour, a job or a disease travels.

Those connections form a network, and studying them turns out to explain things that no account of groups can.

Ask ten working people how they got their current job. Do it properly — press past the first answer.

You will find that a large proportion did not get it by answering an advertisement. They got it because someone told them about it, or someone mentioned their name.

Now ask the follow-up question, which is where the interesting finding lives: how close were you to that person?

Most people expect the answer to be "very close" — a best friend, a brother, a father's colleague. In practice you will find a striking number of answers like: a man I studied with and hadn't seen in four years. My neighbour's brother-in-law. Someone I met at a wedding. A former colleague I'm friendly with but never call. My old landlord.

Not strangers. Not intimates. The middle layer — the people you know but do not see often.

That was Mark Granovetter's finding, published in 1973 in a paper called "The Strength of Weak Ties," and it is among the most cited pieces of research sociology has produced.

Why should distant acquaintances be more useful than close friends, who care about you far more and would help you far more willingly?

The answer, once seen, is obvious — and it has nothing to do with willingness.

Your close friends know each other. They move in the same circles you do, attend the same events, and talk to the same people. Which means they know roughly what you know. The information circulating among them has already reached you.

Your acquaintances do not know each other, and do not know your friends. They move in different circles. Which means they are connected to pools of information that your close network has no access to.

So the value of a tie for getting new information depends not on its strength but on its reach — on whether it connects you to a part of the social world you are otherwise cut off from. And weak ties are far more likely to do that.

This is the founding insight of network analysis: structure explains outcomes that motivation cannot. Your friends want to help you more, and can help you less.

The idea, in plain words

The vocabulary is worth having precisely, because it will recur in Parts 5, 7, 9 and 11.

Tie strength. Granovetter defined it by a combination of the amount of time, emotional intensity, intimacy and reciprocal services in a relationship. Strong ties: family, close friends. Weak ties: acquaintances, former colleagues, friends of friends.

Bridge. A tie that is the only connection between two otherwise unconnected parts of a network. Granovetter's structural point: a bridge is almost always a weak tie, because strong ties tend to be embedded in dense clusters where alternative paths exist.

Density. The proportion of possible ties that actually exist in a network. A village neighbourhood is dense (everyone knows everyone); a city professional network is sparse.

Centrality. How well-positioned a node is. Two kinds matter most: degree centrality (how many ties you have) and betweenness centrality (how often you lie on the shortest path between others). These are not the same, and betweenness is often the more consequential — a person with few ties who happens to be the only link between two clusters holds real power.

Structural hole. Burt's concept from 2.2.6: a gap between two unconnected clusters. The person who bridges it — Simmel's tertius — gains information advantage (they hear both sides first) and control advantage (they decide what passes).

Homophily. The strong and near-universal tendency for ties to form between similar people — same class, caste, religion, education, occupation, age, region. "Birds of a feather," and one of the most robust findings in the whole discipline (it is also what 1.1.2 called homogamy, applied beyond marriage).

Small world. The finding that most people in a large population are connected by surprisingly short chains. Milgram's 1967 experiment asked people in the American Midwest to forward a packet towards a target in Boston through personal acquaintances; among the chains that completed, the median length was around five or six intermediaries — hence "six degrees of separation." Use the finding with care: the majority of chains never completed, the sample was small and unrepresentative, and later replications have produced mixed results. The general structural point — that sparse networks with a few long-range ties produce short paths — has survived and is well supported mathematically; the specific number is folklore.

It is tempting to hear "networks" and think of a neutral web. They are not neutral, and this is where the concept becomes sociologically serious.

Homophily plus referral hiring reproduces existing distributions. Return to 1.3.2's interview: eighty per cent of hires came through referrals. If people's networks are homophilous — and they are — then referral hiring reproduces the existing composition of an organisation with great efficiency, without any individual discriminating. Every participant can be entirely fair-minded and the outcome is a workforce that looks like the one that was there before. This is 1.5.2's debunking exactly: a latent function with no author.

Networks are themselves unequally distributed. Whose network contains a doctor, a lawyer, a person who has been to university, a person who knows how a bank works? Network resources are patterned by class, caste, region and gender — which is precisely why Bourdieu counted them as a form of capital (2.1.13).

And in India, jati functions as a network as well as a status. The classic work on trading and business communities — Marwari, Gujarati Bania, Chettiar, Nattukottai and others — shows caste-based networks supplying credit without formal contracts, information about markets and counterparties, enforcement through reputation, and apprenticeship. The same mechanisms operate in migration: chain migration, in which one migrant's network draws successive arrivals from the same village into the same city, occupation and neighbourhood, is one of the best-documented patterns in migration research (11.3).

Notice the double edge, and it is the honest conclusion. These networks are genuine resources for those inside them — they made economic activity possible where formal institutions were weak, and they support migrants who would otherwise arrive with nothing. And they are exclusion machinery for everyone outside, because the same closure that makes them work is what keeps others out.

Trust and exclusion are the same mechanism viewed from two sides. That sentence is worth carrying into Part 8.

A necessary clarification, because "social capital" is used in the literature for three genuinely different concepts and confusing them is a common error.

Bourdieu's version (class). Social capital is the resources accessible through one's network, and it is a form of capital — unequally distributed, convertible into economic and cultural capital, and part of how privilege is reproduced (2.1.13). The emphasis is on inequality.

Coleman's version (closure). Social capital inheres in relations that facilitate action — especially in dense, closed networks where everyone knows everyone, because closure enables norms to be enforced and trust to be extended. His classic application: communities where parents know each other's children can sustain expectations that isolated parents cannot. The emphasis is on function.

Putnam's version (civic). Social capital is the stock of associational life, trust and civic engagement in a community, which he argued makes democratic institutions work better and which he documented declining in late-twentieth-century America (Bowling Alone). The emphasis is on collective goods. Putnam's distinction between bonding social capital (within a homogeneous group) and bridging social capital (across groups) is genuinely useful — and note that it is Varshney's finding in 2.2.5 in another vocabulary.

These are not compatible. Bourdieu's version is about how networks produce inequality; Putnam's is about how associational density produces public goods. A community can be extremely high in Putnam's social capital and be tightly closed, exclusionary and hostile to outsiders — dense bonding capital with no bridges is exactly what a well-organised riot requires.

Always specify which sense you mean.

Go deeper

Diffusion through networks

Networks are how things travel: information, rumours, technology adoption (2.1.15), job openings, political mobilisation, disease, behaviour.

Three findings worth having:

Structure shapes speed and reach. A dense cluster spreads something quickly within itself and poorly beyond; a few long-range weak ties can carry it across the whole population. This is the mechanism behind small-world effects.

Simple and complex contagion differ. Information can spread through a single contact — one person tells you, and you know. Behaviour that carries risk or cost usually requires multiple independent sources before someone adopts it: you need several people you trust to have done it. Which means that for behaviour change, dense clusters with reinforcing ties can outperform long weak-tie bridges — the opposite of the job-information case. This distinction, developed by Damon Centola among others, corrects a common over-generalisation of Granovetter.

And network position predicts influence better than personality does. The person who is listened to in an organisation is often the one occupying a betweenness position, not the one with the most impressive characteristics. This is the network version of 2.2.1's point about positions.

What network analysis actually does

You will meet this as a method in 5.5 and 7.6. In outline: researchers collect relational data (who talks to, trusts, lends to, works with, marries whom), represent it as a matrix or graph, and compute structural measures — centrality, density, clustering, path length, the presence of structural holes.

Its power is that it makes structure measurable without asking anyone about attitudes. Its limits are real: relational data is hard to collect, boundaries have to be drawn somewhere arbitrary, and the meaning of a tie ("friend") varies enormously between respondents.

Harrison White and colleagues built structural sociology on the premise that positions in networks explain more than attributes of individuals. It is one of the strongest programmes in the discipline for demonstrating that structure is not a metaphor.

Networks are not a metaphor — and not everything

Two cautions in one heading.

Against the sceptics: a network is not a figure of speech. Ties are observable, countable relations, and their patterns have measurable consequences for wages, health, mobilisation and mortality.

Against the enthusiasts: network position is not the only thing that matters. A perfectly placed broker with no money, no credential and a stigmatised master status will not convert their position into much (2.2.2). Network analysis is strongest when combined with an account of what flows through the ties and what resources the nodes hold — which is why the best work in this tradition (Granovetter's own embeddedness argument, 1.4.2) always keeps institutions and interests in the picture.

"A network is a group." Networks have no boundary, no membership and no requirement of shared identity. That is what makes them a different tool.

"Weak ties are better than strong ties." Better for novel information. For support, care, money in a crisis, and behaviour change requiring reinforcement, strong ties are what work. Ask what is flowing.

"Six degrees of separation is an established fact." Milgram's completion rate was low and the sample small. The structural principle is sound; the number is folklore.

"Social capital is a good thing." It depends whose and which. The mechanism that produces trust inside a group is the mechanism that excludes those outside — and dense bonding capital without bridges is a resource for exclusion and violence as readily as for cooperation.

"Networks are neutral." Homophily plus referral means networks reproduce existing distributions without anyone discriminating.

"Well-connected people are more sociable." Position and personality are different things. Betweenness can be a product of biography, geography or accident.

Why it matters

It explains outcomes that intentions cannot. Nobody in the referral chain discriminated; the outcome was reproduction of the existing composition anyway. Once you can see structure, a whole class of unjust outcomes becomes explicable without villains — and therefore addressable by changing procedures rather than by accusing people (1.5.2).

It gives you a realistic account of opportunity. "It's not what you know, it's who you know" is a cliché that turns out to be an under-description: it is who they know, and whether your ties reach beyond the circle that already knows what you know.

It makes the meso level concrete. 1.3.2 said most of social life happens in the middle, between individual and society. Networks are the clearest available way to study that middle, and they are measurable.

And it explains the double edge of community. The trust that makes a community work and the closure that keeps outsiders out are the same structural property. That is one of the more important things sociology has to say, and it applies to caste networks, migrant communities, professional associations and elite schools alike.

A social network is a set of nodes and ties. Unlike groups, networks have no boundary, membership or shared identity.

Granovetter's strength of weak ties: acquaintances are more useful than intimates for novel information, because close friends know what you know while weak ties reach other clusters. A bridge — the only link between two parts of a network — is almost always weak.

Key vocabulary: density, degree and betweenness centrality, structural holes (Burt), homophily, and the small world finding (Milgram's six degrees — structurally sound, numerically folklore).

Networks are inequality machinery: homophily plus referral hiring reproduces existing distributions without anyone discriminating; networks are themselves unequally distributed; and jati-based business and chain migration networks show that trust and exclusion are the same mechanism seen from two sides.

Social capital names three different concepts: Bourdieu's (network resources as a form of capital reproducing inequality), Coleman's (closure enabling norm enforcement), and Putnam's (civic association producing public goods, with bonding versus bridging capital). Always specify which.

Diffusion: structure shapes speed and reach; simple contagion (information) spreads through weak ties, while complex contagion (costly behaviour) needs multiple reinforcing sources.

Social network — a set of actors and the ties between them.

Node / tie — an actor in a network; a relationship between actors.

Tie strength — the combination of time, emotional intensity, intimacy and reciprocity in a relationship.

Strong / weak tie — close relationships versus acquaintanceships.

Bridge — a tie providing the only connection between two parts of a network.

Strength of weak ties — Granovetter's finding that weak ties supply novel information because they reach other clusters.

Density — the proportion of possible ties that exist.

Degree centrality / betweenness centrality — number of ties; frequency of lying on shortest paths between others.

Structural hole — a gap between unconnected clusters, advantageous to whoever bridges it (Burt).

Homophily — the tendency of ties to form between similar people.

Small world — the property of sparse networks with some long-range ties producing short path lengths.

Social capital — network-based resources; used in Bourdieu's (class), Coleman's (closure) and Putnam's (civic) senses.

Bonding / bridging social capital — ties within a homogeneous group versus ties across groups (Putnam).

Chain migration — migration in which existing migrants' networks draw successive arrivals from the same origin.

Simple / complex contagion — spread requiring one contact versus spread requiring multiple reinforcing contacts.

Embeddedness — the location of economic action within concrete social relations (Granovetter, 1.4.2).

One — the job question. Ask five working people how they found their current job, and how close they were to whoever helped. Record the answers. You are replicating Granovetter with a sample of five.

Two — map your bridges. List people you know who know almost nobody else you know. Those are your bridges to other clusters, and they are where unexpected information will come from.

Three — test homophily. Take the ten people you speak to most. Note their class background, education, religion, caste or ethnicity, and region. Count how many differ from you on each. Most people are surprised by how low the count is.

Four — the double edge. Identify one community network you know that provides real support to its members. Then list who is outside it and what that exclusion costs them. Both halves are true at once, and holding both is the lesson.