Short communicationCost range and the stable network structures
Introduction
Since Jackson and Wolinsky (1996) started with a value function that is defined on networks directly, the study concerning network formation has gained focus. A large of literature has sprung from this study based on the work of JW and here have been many significant results. Among them, the simulations of Hummon (2000) found that there were equilibrium structures that were not anticipated in the formal analyses of JW. To resolve the discrepancy, Doreian (2006) examines the cost-benefit conditions under which the different network structures emerge. By the potential evolutionary sequence with line addition and line deletion, Doreian gets three main conclusions concerning the equilibrium structures in different cost range when the number of vertexes is 3, or 4, or 5, respectively. We examine these interesting results and find out some different results for Theorem 4.
His Theorem 4 reads: Theorem 4 For graphs with n = 5 vertices, the equilibrium structures are: For γ > δ, the null graph (G1) results; For δ > γ > δ − δ3, the star (G11), the cycle (G20), G12 and G19 result; For δ − δ3 − γ > δ−δ2, the star (G11), the cycle (G20), G17 and G26 result; and If γ < δ − δ2, the complete graph results.
We present three problems with this statement.
Section snippets
G17 is not the equilibrium structure for δ − δ3 > γ > δ − δ2
The cost range δ − δ3 > γ > δ − δ2 implies the connecting condition δ3 < δ − γ < δ2. From the utility perspective suggested by Jackson and Wolinsky (1996), it means that an indirect link with two steps is better than direct link (sometimes referred to as adjacent) and direct link is better than indirect link with three steps. Then, for vertex e and d in G17 (all relevant structures are presented in Fig. 1), the formation of line ed means that the indirect link with three steps is replaced in the calculation
The structure G19 is not the equilibrium structure for δ > γ > δ − δ3
The cost range δ > γ > δ − δ3 implies the connecting condition 0 < δ − γ < δ3 and this implies that indirect link with three steps is better than direct link. Then, for vertex b, or c, or d in G19, the severance of link bc or cd means that the direct link (bc or cd) is replaced in the calculation of the utilities. The severance of bc or cd can increase the utility δ3 − (δ − γ) for b, or c, or d so that the line bc or cd will be severed.
For example, for vertex d, the severance of link dc means that it can get
The conclusion about “1. For γ > δ, the null graph (G1) results” is incomplete
Consider the circle (G20). The utility of each vertex in G20 is 2δ − 2γ + 2δ2. For γ > δ, the direct link will not provide any positive utility to vertex. So, when γ > δ, no vertex builds a new link. Then, under what conditions will a vertex in the circle sever some links when γ > δ? We analyses the problem as follows:
No matter which vertex severs a link, the structure G20 will be replaced by a line formed by vertexes a, b, c, d, and e.
Given that vertex a or c severs the link ac, it implies that the
Conclusion
By analysis, we make certain that some results of Theorem 4 in “Actor network utilities and network evolution, social networks” are inaccurate. And we think the correct conclusion is as following:
For graphs with n = 5 vertices, the equilibrium structures are:
- 1.
For γ > δ + δ2 − δ3 − δ4, the null graph (G1) results;
- 2.
For δ + δ2 − δ3 − δ4 > γ > δ, the null graph (G1), the cycle (G20) results;
- 3.
For δ > γ > δ − δ3, the star (G11), the cycle (G20), G12 result;
- 4.
For δ − δ3 > γ > δ − δ2, the star (G11), the cycle (G20), and G26 result; and
- 5.
If γ <
Acknowledgements
We gratefully acknowledge the support of the National Science Foundation of China (No. 70571062 and No. 70121001).
References (3)
Actor network utilities and network evolution
Social Networks
(2006)
Cited by (8)
Types of stable network structures based on exhaustive search
2013, Social NetworksCitation Excerpt :For graphs with n = 5 vertices, the equilibrium structures are: Since Jackson and Wolinsky (1996) raised the puzzle that the stability and efficiency of social networks are not always compatible, the stability problem of social networks is researched by many scholars using approaches of simulation analysis and exhaustive search (Hummon, 2000; Doreian, 2006, 2008; Xie and Cui, 2008a,b). This paper considers the exhaustive search in order to find all the possible transitions among isomeric social networks with five vertices, and identifies the seven types of stable network structures: complete, star, chain, cycle, near-chain, near-cycle and null.
Topologies of strategically formed social networks based on a generic value function-Allocation rule model
2011, Social NetworksCitation Excerpt :Doreian (2006) explores the same issue in a systematic manner and establishes the conditions under which different pairwise structures are generated. Some gaps in the analysis of Doreian (2006) are addressed by Xie and Cui (2008a,b). Jackson (2003) reviews several models of network formation in the literature with an emphasis on the tradeoffs between efficiency with stability.
Actor utilities, strategic action and network evolution
2008, Advances in Strategic ManagementFormation of Stable Strategic Networks with Desired Topologies
2015, Studies in Microeconomics