Abstract
Network externalities spur the growth of networks and the adoption of network goods in two ways. First, they make it more attractive to join a network the larger its installed base. Second, they create incentives for network members to actively recruit new members. Despite indications that the latter “peer effect” can be more important for network growth than the installed-base effect, it has so far been largely ignored in the literature. We address this gap using game-theoretical models. When all early adopters can band together to exert peer influence—an assumption that fits, e.g., the case of firms supporting a technical standard—we find that the peer effect induces additional growth of the network by a factor. When, in contrast, individuals exert peer influence in small groups of size n, the increase in network size is by an additive constant—which, for small networks, can amount to a large relative increase. The difference between small, local, personal networks and large, global, anonymous networks arises endogenously from our analysis. Fundamentally, the first type of networks is “tie-reinforcing,” the other, “tie-creating”. We use survey data from users of the Internet services, Skype and eBay, to illustrate the main logic of our theoretical results. As predicted by the model, we find that the peer effect matters strongly for the network of Skype users—which effectively consists of numerous small sub-networks—but not for that of eBay users. Since many network goods give rise to small, local networks, our findings bear relevance to the economics of network goods and related social networks in general.
Similar content being viewed by others
Notes
With some simplification, Skype works in the following way. After registering with the Skype service, a user builds her personal contact list by sending contact requests to other users, or receiving and accepting requests herself. She can then call each person on her personal contact list via Skype. Calls are transmitted over the internet, and are free of charge for both parties. The Skype software also include other features such as conference calls, calls to fixed-line phones (SkypeOut), sending SMS or Instant Messaging. In September 2005, Skype was acquired by the Internet auction provider eBay.
See http://about.skype.com/ (accessed June 9, 2011).
See http://www.sec.gov/Archives/edgar/data/1498209/000119312511056174/ds1a.htm#rom83085_9 (accessed June 9, 2011).
See http://www.skype.com/intl/en-us/tell-a-friend/ (accessed June 9, 2011).
Word-of-mouth communication is a much researched topic in marketing and communication research. For an early work in the field of communication research see Lazarsfeld et al. (1944), who introduce the distinction between word-of-mouth communication and mass media influence in the context of voter behavior. For an early contribution in the field of marketing see Arndt (1967), who analyzes the effect of word-of-mouth communication on the diffusion of a new product in an experimental setting.
To model this situation, one would have to determine how exactly peer influence is exerted on a person who would take the same action out of her own account. For the purpose of the current analysis, however, this question can remain unanswered. What matters is that up to a certain network size both effects can work concurrently, while beyond this size only the peer effect is relevant.
In real life, it is a plausible situation that a network starts out as an insider tip among friends with growth largely through peer influence, while later it becomes popular and further adopters join because of the installed-base effect. However, this scenario is characterized by various overlaying effects, not all of which are captured in our models: the peer effect, the installed-base effect, and the effect of popularity. Popularity of a network good implies that information about it becomes more readily available and may even be impossible to ignore, which increases the perceived benefits and reduces the cost of adoption for all not-yet-adopters. This “popularity” effect is not captured in our models, though.
We use the term “consumer” to keep the presentation simple. However, the potential adopters of the focal good may also be firms or other institutions.
In fact, monotonicity of u(x) need not be assumed. Since each consumer is characterized only by the standalone utility she derives from the good, we are free to assign values of x in such a way to the consumers that their standalone utility decreases with x. Thus, we obtain \(u'(x) \le 0\) by definition rather than by assumption.
A third possibility would be that the function equals zero at one or more points and is negative otherwise. These roots would constitute equilibria which are unstable against deviations to lower values of y. We refrain from pursuing this case further since we focus on stable equilibria.
A solution y to Eq. 1 at which the sign changes from negative to positive would mean that, after a small deviation to \(y - \epsilon < y\), the marginal consumer would experience a negative net utility and would hence not join the network, reducing its size further. In contrast, a positive deviation of the network size to \(y + \epsilon \) would imply that the marginal non-consumer would derive a positive utility from joining the network. She would consequently do so and increase the network size further.
That is, they have realistic expectations concerning the adoption that will take place in stage one, but are myopic with regard to peer effects coming into play in stage two. Without this assumption, some agents who do adopt in the base case would refrain from doing so, because they do better by deferring their adoption until the second stage, thus not being an influencer. We will relax the assumption later on that each early adopter becomes an influencer in stage two.
After they have gone through the adoption process they might, in a third stage, act as influencers themselves. However, we restrict our analysis here to two stages. An extension to three or more stages would of course be feasible. However, if one aims at making the temporal structure more realistic, then a more suitable choice would be to introduce continuous time instead of three or more stages. We refrain from pursuing this approach in order to keep the model tractable.
Of course, our model is a simplification of this scenario by allowing only two levels of communication intensity: either two individuals belong to the same network or they do not. We also abstract from the fact that some individuals will belong to more than one sub-network.
The high degree of consistency of the three intersection points for each of the two services can be quantified using vector algebra. Consider, for the case of eBay, the line ending at the upper corner, denoted line P for “peer influence.” Normalizing the length of line P to unity, its intersection with line L is at 0.241 while that with line G is at 0.190. The difference thus equals 0.051, or 5.1 % of the length of line P. In an analogous way, the normalized differences between the two intersections obtains as \(0.266 - 0.211 = 0.055\) for line L, and as \(0.585 - 0.510 = 0.075\) for line G. For the case of Skype, the normalized differences are 0.086 (P), 0.082 (L), and 0.033 (G).
This variable is operationalized as follows. 1: None of my friends or colleagues had told me about Skype (Ebay) and its features; 2: One or more of my friends or colleagues had told me about Skype (eBay) and its features; 3: One or more of my friends or colleagues had invited me to register with Skype (eBay); 4: One or more of my friends or colleagues had tried to persuade me to register with Skype (eBay); and 5: One or more of my friends or colleagues had pressed me to register with Skype (eBay).
This variable is operationalized as follows. 1: I have not told any of my friends or colleagues about Skype (eBay) and its features; 2: I have told one or more of my friends or colleagues about Skype (eBay) and its features; 3: I have invited one or more of my friends or colleagues to register with Skype (eBay); 4: I have tried to persuade one or more of my friends or colleagues to register with Skype (eBay); 5: I have pressed one or more of my friends or colleagues to register with Skype (eBay).
We caution the reader that a platform such as Facebook is far more complex than the relatively simple network services of Skype and eBay that we used as illustrations. Thus, one has to be careful in applying our result to this case. Still, despite many differences the three services share fundamental characteristics, and our theoretical analysis is kept in general terms. We thus think that the underlying mechanisms at work are robust.
References
Abrahamson E, Rosenkopf L (1997) Social network effects on the extent of innovation diffusion: a computer simulation. Organ Sci 8(3):289–309
An MY, Kiefer NM (1995) Local externalities and societal adoption of technologies. J Evol Econ 5:103–117
Arndt J (1967) Role of product-related conversations in the diffusion of a new product. JMark Res IV:291–295
Baumgarten SA (1975) The innovative communicator in the diffusion process. J Mark Res 12(1):12–18
Bettencourt LMA, Cintrón-Arias A, Kaiser DI, Castillo-Chávez C (2006) The power of a good idea: quantitative modeling of the spread of ideas from epidemiological models. Physica, A 364(15):513–536
Birke D, Swann GMP (2006) Network effects and the choice of the mobile phone operator. J Evol Econ 16:65–84
Block JH, Köllinger PD (2007) Peer influence in network markets: an empirical investigation. Schmalenbach Bus Rev 59:364–386
Brynjolfsson E, Kemerer CF (1996) Network externalities in microcomputer software: an econometric analysis of the spreadsheet market. Manage Sci 42(12):1627–1647
Cowan R, Miller JH (1998) Technological standards with local externalities and decentralized behaviour. J Evol Econ 8(3):285–296
Church J, King I (1993) Bilingualism and network externalities. Can J Econ 26(2):337–345
Deroïan F (2002) Formation of social networks and diffusion of innovations. Res Policy 31:835–846
Domingos P, Richardson M (2001) Mining the network value of customers. In: Proceedings of the seventh ACM SIGKDD international conference on knowledge discovery and data mining, San Francisco, California, pp 57–66
Economides N, Himmelberg C (1995) Critical mass and network size with application to the US FAX market. Stern School of Business Working Paper 95–11
Farrell J, Saloner G (1985) Standardization, compatibility, and innovation. Rand J Econ 16(1):70–83
Farrell J, Saloner G (1986) Installed base and compatibility: innovation, product preannouncements, and predation. Am Econ Rev 76(5):940–955
Gandal N (1994) Hedonic price indexes for spreadsheats and an empirical test for network externalities. Rand J Econ 25(1):160–170
Hallinan MT, Williams RA (1990) Students’ characteristics and the peer-influence process. Sociol Educ 63(2):122–132
Hotelling H (1929) Stability in competition. Econ J 39:41–57
Jonard N, Yildizoglu M (1998) Technological diversity in an evolutionary model with localized learning and network externalities. Struct Chang Econ Dyn 9(1):35–51
Katz ML, Shapiro C (1985) Network externalities, competition, and compatibility. Am Econ Rev 75(3):424–440
Katz ML, Shapiro C (1986) Technology adoption in the presence of network externalities. J Polit Econ 94(4):822–841
Katz ML, Shapiro C (1992) Product introduction with network externalities. J Ind Econ 40(1):55–83
Koski H (1999) The installed base effect: some empirical evidence from the microcomputer market. Econ Innov New Technol 8:273–310
Koski H, Kretschmer T (2004) Survey on competing in network industries: firm strategies, market outcomes, and policy implications. J Ind Comp Trade 4(1):5–31
Lazarsfeld PF, Berelson B, Gaudet H (1944) The people’s choice: how the voter makes up his mind in a presidential election. Columbia University Press, New York
Lee E, Lee J, Lee J (2006) Reconsideration of the winner-take-all hypothesis: complex networks and local bias. Manage Sci 52:1838–1848
Majumdar SK, Venkataraman S (1998) Network effects and the adoption of new technology: evidence from the U.S. telecommunications industry. Strateg Manage J 19(11):1045–1062
Rysman M (2004) Competition between networks: a study of the market for yellow pages. Rev Econ Stud 71(2):483–512
Saloner G, Shepard A (1995) Adoption of technologies with network effects: an empirical examination of the adoption of automated teller machines. Rand J Econ 26(3):479–501
Shankar V, Bayus BL (2003) Network effects and competition: an empirical analysis of the home video game industry. StrategManage J 24(4):375–384
Sinan A, Walker D (2012) Identifying influential and susceptible members of social networks. Science 337:337–341
Solomon S, Weisbuch G, de Arcangelis L, Jan N, Stauffer D (2000) Social percolation models. Physica, A 277(12):239–247
Subramani MR, Rajagopalan B (2003) Knowledge-sharing and influence in online social networks via viral marketing. Commun ACM 46(12):300–307
Tomochi M, Murata H, Kono M (2005) A consumer-based model of competitive diffusion: the multiplicative effects of global and local network externalities. J Evol Econ 15:273–295
Tucker C (2008) Identifying formal and informal influence in technology adoption with network externalities.Manage Sci 54(12):2024–2038
Turnbull PW, Meenaghan A (1980) Diffusion of innovation and opinion leadership. Eur J Mark 14(1):3–33
Venkatesh V, Morris M (2000) Why don’t men ever stop to ask for directions? Gender, social influence, and their role in technology acceptance and usage behavior. MIS Q 24(1):115–139
Acknowledgments
We are grateful to many people who shared their insights or commented on earlier drafts of the paper. Special thanks go to Nicoletta Corrocher, Oliver Fabel, Paul A. Kattuman, Tobias Kretschmer, Thomas Rønde, and Catherine Tucker. We also thank participants at seminars at the Academy of Management Meeting, Bocconi University, Deutsches Institut für Wirtschaftsforschung (DIW), the DRUID conference, the European Academy of Management, the TUM/LMU TIME colloquium, and the VfS Industrial Organization Divsion. Errors and oversights are ours alone.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
1.1 Symbols
- c :
-
Variable cost of providing the network good
- n :
-
Size of small sub-network which coordinates to recruit a new adopter
- r :
-
reward paid by seller to a network member recruiting a new adopter
- u 0 :
-
Constant in the linear model of u(x): \(u(x) = u_0 - \lambda x\)
- u(x):
-
Stand-alone utility (i.e., excluding network externalities) which consumer x derives from the good
- u(y):
-
Utility that each adopter derives due to the network effect when the network size equals y
- \(y_i^*\) :
-
Marginal adopter in equilibrium, in model i
- α :
-
Slope parameter in the linear model of v(y): \(v(y) = \alpha y\)
- κ :
-
Cost leveraging factor: When a user exerts peer influence on some not-yet-adopter, the resulting benefit for the wooed individual equals κ times the cost that the influencer incurs
- λ :
-
Slope parameter in the linear model of u(x): \(u(x) = u_0 - \lambda x\)
- ω :
-
\(\omega \equiv \alpha / \lambda \) measures the relative strength of the network externality
Rights and permissions
About this article
Cite this article
Henkel, J., Block, J. Peer influence in network markets: a theoretical and empirical analysis. J Evol Econ 23, 925–953 (2013). https://doi.org/10.1007/s00191-012-0302-4
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00191-012-0302-4