I just finished reading a bunch of research on social networks and social capital, and would like to share my notes with you. The research explains social capital very clearly from a science point of view. I know everyone is looking for the Holy Grail, so I thought this might help the search.

Social networks are collections of social entities (mostly human in this research) and the collection of connections between these nodes. This collection of connections is the basis for social capital, and the basic properties are important to know and understand. Read about them in this blog, the next blog will use them to explain social capital based on these properties.

Directionality is simply the way the tie moves. 2-way of 1-way and if yes, what direction?

Reciprocity is an important factor in social capital. Does the advice flow one-way? Is it symmetric? Reciprocity is crucial for positive affect, cooperation and trust between ties. Generally, when tie A helps tie B, tie B will help tie A later, or trust of affection weakens.

Multiplexity defines whether the relation is multi-layered. This is the case when good friends offer each other advise, the layers are friendship and advise, which influence each other but should be valued apart from each other.

Tie strength is extremely complex and very important to social capital. Nobody has been able to take into account all the complex variables, probably because they change constantly, should be measured very precise and are mainly intuitive. Researchers define many variables that are often overlapping. I wrote this for anyone interested in playing around with them, so i’ll just show em all.

• Amount of time (intensity),
• interval,
• duration of interval and total duration.
• Emotional intensity
• Intimacy (mutual confiding)
• Reciprocal services
• Activity
• Frequency of activity
• Depth
• Closeness
• Valence (affective, supportive or cooperative character of tie)

Homophily is the tendency for actors to connect to other actors of their own sort. This can be good and bad. Similarity breeds connection, so that’s the easy part. But the similarity might have a negative effect on learning about new ideas and products and changing one’s opinion. The upside is multilayeredness of ties, so similarity might exist on one level, but on other levels the nodes may teach each other and influence each other’s opinions.

Now that the ties are defined, ill share some network properties with you.

Transitivity is the driving force in how networks evolve. It defines that a connection between node 1 and 2 cant exist without a connection between node 2 and 3 (while 1 and 3 don’t have to have a tie). For example if I work at a coffee shop and sell bagels from the bakery next door, I am clearly tied to that baker. I’m also tied to the supplier of the baker’s flower supplier. The more affection and valency exists between me and the baker, the more likely it becomes that ill use that supplier when the baker can’t supply something directly and I have to bake it myself.

Structural balance. This property lies close to transitivity, but looks at the positive or negative affect or valence. Negative valence can effect triades, for instance when you befriend 2 people who are enemies. This may bring awkward situations where both parties are not eager to share any knowledge or be open to connect.

Density
How many ties are possible in a network, and how many are actually there. In a network of 5 consumers, 10 non-directional ties can exist. With 6 connected ties, the density is 60%

Closure and local clustering
Closure is the density among those in a network with whom an actor has a tie. If all your friends know each other, closure is 100%.

Centrality
The importance of an actor within a certain network. Has some sub-properties.
Degree centrality
Basic measurement, in-degree shows the people that connect to you, out-degree shows the people that you connect to. Especially in-degree centrality is great in defining popularity.

Closeness centrality
How many steps does it take for actor A to reach actor B? The less steps to each node in a network, the more closeness can be attributed to the measures person.

Betweenness centrality
Measures how ‘in the middle’ someone is. Very important since the person in the middle has the option of keeping networks separated, thus controlling flow of information and value.

Social cohesion
This identifies closely connected subgroups of networks. When many actors of a subgroup are connected, information, attitudes and values are more easily diffused.

Structural equivalence
This focuses on 2 nodes, and sees how many overlapping contacts, values, similar information sources etc. the nodes both have.

Structural isomorphism
This is similar to structural equivalence, but does not necessarily look at the exact persons. If 2 people have completely different networks but the portfolio of their networks is similar, isomorphism is important because equivalence would give a 0% result but actually the 2 people have very similar backgrounds. Just coming from different persons.
Interesting
The research defines some interesting notes using the above properties.

While many networks enjoy high homophily and transitivity, many people think that’s the reason why closed groups are not good. When people are attracted to eachother by the same properties like race and gender, the nodes in the group are similar and closed of from outside knowledge and information. The following example explains how homophily is actually a good thing for network expansion. Imagine a firm with coworkers with 3 properties, nationality (French versus german), sports (baseball of hockey) and department (finance versus marketing). 2 french people may easily connect, but their interest in sport or department may differ. So on that level they can exchange knowledge and information.This may actually aid in briding the finance and marketing department based on the similar language.

6-degrees of separation was the research everybody still talks about. ‘I know the american president in 6 steps’. The research was actually started with 296 individuals who had to send a postcard to 1 random pre-selected person. The longest path was 11, the shortest 1. The 6-degrees is therefore based on the average. Also, only 64 postcards arrived, so 71% gave up because they didn’t know where to go to or because they weren’t motivated to cary out the task (or anyone in the network in between the start and end wasn’t motivated). Important is that in later studies researchers found that high local clustering within subnetworks that are connected by only a few nodes was enough to find a really short path. The longest paths where in non-dense networks.

In the upcoming blog, I’ll share my notes explaining social capital based on these network properties.

For anyone interested in tools to measure the properties, you should get your hands on ‘social networks and marketing’ by Christophe van den Bulte and Stefan Wuyts (Marketing Science Institute).