Social Networks and their Economics: Influencing Consumer Choice by Daniel Birke

Author:Daniel Birke [Birke, Daniel]
Language: eng
Format: mobi
ISBN: 9781118699645
Publisher: Wiley
Published: 2013-07-23T14:00:00+00:00

The most important variables used to test whether churning decisions are interdependent or not are the first four. As can be seen from Table 4.3, churn is strongly influenced by the decision of neighbours to churn (number of churned neighbours). In other words, churn decisions of consumers are interdependent. One additional neighbour decreases the time to churn by about 904 days. On average each subscriber is estimated to stay with the current network for an average of 7.5 years, which shows that each churner has a considerable impact on their neighbours' decisions to churn.

As can be seen in the last column, the average value of this variable is just 0.59, meaning that most subscribers do not have contact with churners (see also Table 4.1). This partly reflects the relatively short period for which data are available (at the beginning of the observation period, this variable is zero for all subscribers). However, there are a few subscribers who have a lot of churned neighbours and are strongly influenced by neighbourhood churn. The estimates of the squared term imply that the total impact of the variable is increasingly negative until eight churn neighbours, but that each additional churn neighbour does not add linearly to the churn probability. Note also that two friends of a subscriber often also have a relationship and a neighbour's influence on the focal node might be multiplied by indirect links.

Apart from the number of churners with whom a subscriber is in contact, the frequency of contact is also important (weights of churned neighbours). The higher the interaction frequency between a subscriber and a churner, the more likely that she will also churn. This makes sense, as a high interaction frequency is an indicator of a closer relationship and influence should be higher between socially close subscribers. However, the marginal effect of this variable is lower than (number of churned neighbours) and 10 additional communications would only decrease the time to churn by 25 days. Again, the relationship is non-linear, but increasing in magnitude and negative over the whole support.

The centrality measures degree centrality (unweighted) and degree centrality (weighted) show the expected signs as well. Especially subscribers who have more neighbours are estimated to churn earlier than neighbours with fewer calling partners, which is consistent with the economic argument that subscribers who use their mobile phone more often have higher incentives to overcome switching costs and switch to new carriers.

A high percentage of neighbours who use a different network (% of off-net neighbours/100) is an indicator of a higher churn probability. As prices for off-net calls are higher than prices for on-net calls, it is more beneficial to churn if a high percentage of neighbours are off-net. Again, I included a squared term to take non-linearities in the functional form into account. The turning point of % of off-net neighbours/100 is at 46% off-net members and the impact of this variable would only get positive for values over 93%, which is the case for only a very small minority in the sample.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.