Blogs And The Economics Of Reciprocal (in)attention

April 21st 2009, Conference of the Royal Economic Society

Blogs and the economics of reciprocal (in)attention Alexia Gaudeul, Chiara Peroni and Laurence Mathieu.

Outline of the presentation 1. Blogs, context and motivations. 2. Objectives of the research 3. The norm of reciprocity 4. The dataset: Livejournal friendship patterns. 5. Empirical Analysis. 6. Conclusion and further research.

Context of the paper  Growth in participation in Internet-mediated nonmonetary participative processes Social networking, open source software, wikis, social news websites such as digg or reddit, etc...

 Particularly important for companies that manage communities of self-motivated individuals. eBay, Amazon, Google, Microsoft, Facebook.

 Also important for media, culture, journalism...

Motivations for blogging Different reasons for blogging. Make friends. Interacting with other bloggers and in communities. Become popular and well-known. Expressing oneself and publicizing own work. Update far located and dispersed real life friends (web presence…).

Results  Study of the behaviour of bloggers. We show that attention and content are the currencies of trade within bloggers’ networks. We evidence the importance of reciprocity norms in blogging networks.

A model of content production (1) 1 e 1 n31

n21 n12 n23

n13

2 3 e3 e2 n32  ei effort by agents i.  Agent i devotes attention nij to the content produced by j and nik to the content produced by k.

A model of content production (2)  A simple additive form for the total utility of agent is:

Ui(n,e) = d sumj≠i nji ei + sumj≠i nij ej - C(ei) Utility from being read

Utility from reading others

Cost of content production

 Content producers choose how much attention to devote to each other and how much content to produce.

A model of content production (3)  We suppose there is Free entry and exit in the blogging market Perfect information on the activity of each bloggers. ‘Attention budget’ T, the same for all agents. Attention shared equally between all friends. Number of friends’ readers independent of one’s number of friends

Hypothesis  H1: Network size: Bloggers who display higher levels of content production and general blogging activity will have a higher number of friends.  H2: Network imbalances: Bloggers with more readers (‘friend of’) than readees (‘friends’) will produce more content than others.

Norm of reciprocity: Caveats!  Other possible forms of reciprocity A friend for a friend. A comment for a comment. Reciprocate only if content of high enough quality.  Other possible motivations: reader only, writer only.  External effects, e.g. ‘real world’ celebrities.  Limitations in the data available

The dataset: LiveJournal  Livejournal is essentially an aggregation tool with lockin effect Write and read public and private entries Participate in communities and exchange comments

 Use of proxies for attention exchanged and expended from a number of bloggers’ activity. Content production (posting per day) Interaction (comments, community participation) Commitment (length of time active)

The dataset: LiveJournal  2767 users.  Main characteristics (median): Three years’ activity (and counting!). One post every two days. Each receive 1.5 comments. Balance between friends / readers (26 / 25). Balance between comments received / made. Membership in 14 communities.

Friends vs. friends of

Reciprocity equation ln(friends ofi)=a+b*ln(friends)+ui ln(friends)

ln(friends of) 0.985***

constant

0.103

obs

2496

R2 F (1, 2494)

se

t-value

P>|t|

0.005

208.98

0.000

0.017

6.07

0.000

0.942 43673***

BP

9.39

(0.002)

RESET

3.41

(0.017)

***: <1%;**: <5%;*: <10%

Regressions (1)  Activity equation 1 to test H1: ln(friend ofi)=cXi+vi with Xi a matrix of independent variables measuring the activity of the blogger i  Activity equation 2 to test H2: ln(friend ofi/friendsi)=dXi+wi with Xi a matrix of independent variables measuring the activity of the blogger i.

Activity equation 1 and 2 Dependent variable: ln(communities) ln(entries)

0.152*** 0.579***

0.170*** 0.576***

-0.030*** -0.007

ln(comments received per post)

0.776***

0.742***

0.070***

ln(comments made per friends) ln(duration)

-0.463***

-0.525***

0.138***

0.601***

0.603***

-0.025***

ln(friends of)

0.474***

ln(friends of/friends)

constant obs adj R2 F stat ***: <1%;**: <5%;*: <10%

ln(friends of/friends)

0.781*** 1334 0.880 1550***

0.950*** 1334 0.902 1646***

0.240*** 1357 0.275 83***

Additional regressions  ln(readersi/friendsi) instrumented by Xi. Shows that deviating from the norm of reciprocity has a negative effect on number of readers and friends.

 Use of a measure of network asymmetry Shows that part of the effect is due to asymmetry, not imbalance, in a blogger’s network of relations.

Conclusion  Activity significant predictor of size of friend list Which comes first? Topic for further study based on panel data set.

 Interpretation of ‘reciprocity norm’: does a ‘balanced’ friends list lead to more or less friends? Those who care about reciprocity may be less willing to friend others (risk of friendship not returned). Those who care about reciprocity may reciprocate more, thus accumulating more friends over time. Net effect is ambiguous, as seen in contradiction in results of OLS and IV regression.

Future research  Is there saturation in reciprocation, i.e. limit to the size of a blogger’s network?  Further research in the effect of systematically reciprocating or to not reciprocating. --> Collection and analysis of panel data Do you befriend (link) because of content produced, or do you produce content because of friends?

 Further analysis of reciprocation at the individual level.