How can you create the network effect

Network Effects: Theoretical Basics

Transcript

1 Network Effects and Network Externalities WS 14/15 Basic Seminar Network Effects: Theoretical Basics Dr. Jürgen E. Blank Submitted by: Pirmin Dewald and Sebastian Robers Pirmin Dewald Sebastian Robers Im Horst 4 Erlenstrasse Kaiserslautern Kaiserslautern 7th semester WI-Mb 7th semester WI-Mb Matriculation no .: Matriculation no .: 381953

2 I Table of Contents 1. 33

3 II List of Figures Figure 1: The direct network effect ... 5 Figure 2: Vertical network ... 8 Figure 3: Two-sided markets Figure 4: The different levels of compatibility Figure 5: Central standardization problem Figure 6: Welfare with perfect competition Figure 7: Welfare with Price regulation Figure 8: Welfare in a monopoly Figure 9: Network properties and the spread of shocks ... 30

4 1 1. Introduction Network effects, which are sometimes also referred to as network externalities, express that the behavior of one person influences at least the well-being of another person. Network effects therefore occur when the demand for a good depends on whether another person also consumes this good. 1 The broadest definition of network externalities is provided by Michael Katz and Carl Shapiro. The benefit that a consumer derives from the use of a good often is an increasing function of the number of other consumers purchasing compatible items [...] we call these positive consumption benefits network externalities. 2 This results in the following consequence: the more consumers use a compatible good, the greater the benefit for each individual consumer. 3 In some networks it happens that the owner of all network nodes is identical. Since this does not correspond to the definition of externalities, the action of an economic subject to bring about a welfare change outside the market, the term network effects was used in later works. 4 In the following we will refer to the term network effects in a uniform manner. A good example of network effects can be shown using telephones. If a consumer is the sole user of a phone, he does not benefit from it. Only when a certain number of people have a phone, i.e. a network has formed, do users benefit. In general, network effects can be divided into direct and indirect network effects. 1 cf. (Corsten & Gössinger, 2008a, p. 559) 2 (Katz & Shapiro, 1986, p. 146) 3 cf. (Gröhn, network effects and competition policy: An economic analysis of the software market, 1999, p. 25) 4 cf. . (Gröhn, Network Effects and Competition Policy: An Economic Analysis of the Software Market, 1999, p. 28f.)

5 2 2. Basics 2.1 Introduction to network effects In the literature, one often speaks of network externalities 5, network effect 6, network effect 7, increasing return of adoption or 8 of DEOS 9 (demand side economies of scale). Furthermore, a distinction is made between horizontal, i.e. direct, and vertical, i.e. indirect, network effects. 10 In the following we use the most common terms, direct and indirect network effects. In the case of direct network effects, the individual network participants can draw a direct, i.e. immediate, advantage for themselves. Indirect network effects only result in an indirect connection between network size and compatible technologies for the network subscriber. Direct network effects Basics of direct network effects The most common definition of direct network effects comes from Katz and Shapiro, who still spoke of network externalities. The benefit that a consumer derives from a good depends on the number of other consumers of the good. 12 A consumer who decides on a product does not make this decision about the basic benefit (T), i.e. the personal requirements for a product, but also refers to the network size (N) in 5 cf. (Katz & Shapiro, 1986, p .146) 6 cf. (Thum, 1995, p.5) 7 cf. (Wiese, 1990) 8 cf. (Arthur, 1989); (Arthur BW, 1994) 9 cf. (Farrel & Saloner, 1986, p.940) 10 cf. (Gröhn, 1999, p.25f.) 11 cf. (Farrel & Saloner, 1985, p.70) 12 cf. (Katz & Shapiro, 1986, p.146)

6 3 incorporates its decision in order to maximize its utility (U). This results in: 13 U! = U! N, T with U! N, T

7 4 Number of all possible connections (F) between the nodes of the network. The sum of all possible connections follows from: 18! F = !!! (n 1) =! (!!!)! with n = 1 ,, N Directional direct network effects The communication or the connections in a network can go in one direction, ie unidirectional, as well as in both directions, ie bidirectional. 19 In the case of unidirectional networks, in Figure 1 (The direct network effect) this would correspond to node A, node B, for example, the exchange possibilities cannot be calculated in generalized terms. The predictability of the network depends mainly on the structure of the network. An example of unidirectional networks is software compatibility. Newer software is usually downwardly compatible with previous versions so that the files of the old software can still be read and edited. However, updates for upward compatibility are extremely rare with the older software. In the case of bidirectional networks, in Figure 1 (the direct network effect), this would correspond to node A, node B, for example, the exchange possibilities can be calculated differently than with unidirectional networks. The number of connections results from N (N 1) exchange possibilities. This results in (N 1) new connections for the new node. The nodes already integrated in the network are each given a new exchange option. The difference, i.e. the number of new connections, results in the bidirectional network from: 20 ΔF = 2 (N 1) 18 cf. (Gröhn, 1999, p.25f.) 19 cf. (Economides, 1996, p.675) 20 see (Gröhn, 1999, p.26)

8 Metcalfe s Law The higher the probability of being able to connect with another user of the same or compatible technology, the greater the benefit that will be drawn from the network. The most common example is the telephone network. A telephone alone is of no use. Users A and B each have a phone, so A can contact B and vice versa. Two connections have already been made. If user C is now added, the number of possible connections increases to six, if another user D is added, 12 connections are already possible. 21 The following figure illustrates the situation. Figure 1: The direct network effect (Source: Saloner, Shepard, & Podolny, 2001, p.309) This growth can be represented for small networks with a simple formula for the total benefit: nn 1 with n = number of users For large networks the overall benefit is represented by the rule of squares. The total benefit here corresponds to: cf. (Erhardt, 2001, p.25f.) 22 cf. (Shapiro & Varian, 1999, p.184ff.); (Saloner, Shepard, & Podolny, 2001, p.308ff.)

9 6 n! with n = number of users Empirical studies Between 1986 and 1991, Neil Gandal tried to empirically prove direct network effects. To this end, he has empirically examined spreadsheet programs. 23 At that time, the market was dominated by Lotus and thus seen as the standard. Lotus’s competitors designed their products and the associated file formats to be compatible with the industry standard. Neil Gandal compared the direct network effects through hedonic price indices, which make the goods comparable through a large variation of various influencing factors that are significant for the products. 24 The result of Gandal's empirical studies was that programs that were compatible with the Lotus standard were on average more expensive than the programs on the market that were not compatible with the Lotus standard. On average, the compatible programs were priced at $ 365 and the incompatible programs were priced at $ 85. In order to anticipate possible criticism that Gandal only examined expensive products that cover other market segments and only showed compatibility as an additional function, Gandal also examined cheaper products at a price below $ 200 separately. The result of this second empirical study was that the Lotus compatible programs had an average price of $ 151. The incompatible programs averaged $ 80. In these studies, Gandal was able to show the extent to which network effects unfold their effect and how consumers' willingness to pay increases or changes. 25 However, it should be noted that with compatibility standards, in addition to direct network effects, indirect network effects can also play a role. 23 see (Gandal, 1994); (Gandal, 1995) 24 cf. (Triplett, 1989, p. 127ff.) 25 cf. (Erhardt, 2001, p. 26f.)

10 7 2.3 Indirect network effects Basics of indirect network effects In addition to direct network effects, a distinction is also made between indirect network effects and vertical networks. Basically, indirect network effects say that as the user base increases, a certain good becomes more attractive as soon as there is an increase in utility without there being a direct relationship. 26 A vertical network consists of different complementary goods (Fig. 2). A single good or a component in the vertical network is of no use on its own. The benefit only arises through the entire system. The example of a computer illustrates this network very well. The computer cannot be used without an operating system, hardware and applications. A fully functional network is only created when all components are present. It is essential that there can of course be several providers for each component. However, the components also influence each other on the different levels. 27 In the following we will look at the two main causes of indirect network effects. 26 cf. (Erhardt, 2001, p. 27) 27 cf. (Gröhn, 1999 p. 27)

11 8 Figure 2: Vertical network (source: Gröhn, 1999, p.27) Causes for indirect network effects Complementary products and services An important cause for indirect network effects can be traced back to complementary goods. Complementary goods are goods that are offered together. 28 Although the main component has its own benefit, it still needs complementary goods or components in order to provide extensive benefits for the consumer. This is shown again well in a computer, which offers a relatively small benefit without a screen, printer or storage media. Even a record player is not very useful with just a few records. Only a certain number and selection of plates leads to a relatively high benefit, which satisfies the consumer. Accordingly, a good gains in usefulness if it has the widest possible selection of complementary 28 cf. (Corsten & Gössinger, 2008, p. 389)

12 9 products and services. If demand increases, the number of complementary goods can increase because the product field becomes more attractive for other suppliers and manufacturers. This also has the consequence that the initially high costs for a good, which may be high, decrease as the number of users increases due to increased production. The network thus continues to gain users, which in turn leads to higher benefits and lower prices for all goods and therefore becomes more attractive for potential users. At this point it is worth mentioning that the greater the benefit, the greater the willingness of customers to pay. Increasing the attractiveness and usefulness of a good can ultimately justify the demand for higher prices. Learning effects Learning effects are another cause of indirect network effects. These are also known as information spillovers. They are not only to be found in complementary products, as in the previous section, but generally occur in innovative technologies. The effects that innovative technologies have on consumers are fraught with a certain degree of uncertainty. It is not clear in advance which technology is better with competing standards and, above all, which applications are preferred by users. The Sony company can be mentioned as an example at this point. Sony had made VCRs. Its original intended function was to serve as a recording medium for private use. At that time it was not yet clear that video recorders would be used to play cassettes that had already been written on, such as feature films. The application behavior of the consumers consequently plays a central role in the development of application possibilities and properties that are placed on the technology. Consumers learn to deal with the new technology and may discover new possible uses for the product through use. This learning effect helps the 29 cf. (Erhardt, 2001, p. 27f.)

13 10 manufacturers to optimize their product and opens up new possibilities for complementary goods. 30 Another important aspect are very strong innovations that confront the user with unknown technologies in the form of innovations. This in turn has the consequence that new, as yet unknown application possibilities arise and the radical innovation is associated with a high level of technical uncertainty. 31 If potential users cannot yet guess the extent of the effectiveness of a new technology, learning effects are of decisive importance. If new information is provided through the learning effect of a consumer who has dealt with the new technology, this leads to a self-reinforcing process. Other potential users get a deeper insight into the areas of application of the new technology so that they can choose the technology with the least amount of uncertainty. 32 Robin Cowan explains this as follows: Early adoptions provide information with which beliefs are updatet, and as the process proceeds opinionons gain strength. Eventually beliefs that one technology is superior are strong enough that it is exclusively adopted. 33 Furthermore, it can be seen that, with regard to empirical studies of indirect network effects, only little information resulting therefrom is available. An empirical analysis was carried out, for example, with regard to the connection between software use and the complementary goods available for it. A significant connection between the complementary goods could be determined. As part of this analysis, training courses in learning the software were also observed very frequently. 34 The Lotus program was also empirically investigated with regard to indirect network effects. Among other things, the use of complementary goods and training courses were measured here. Both, complementary goods as well as training, showed a higher usage at Lotus 30 cf. (Erhardt, 2001, p. 28) 31 cf. (Kuhn, 1976) 32 cf. (Cowan, 1992, p. 285f.) 33 ( Cowan, 1992, p.286) 34 cf. (Shurmer, 1993)

14 than with other comparable products. This can be explained by the dominant position that Lotus took on the market. Positive Feedback Loop Positive feedback is the process that the indirect network effect triggers.Shapiro says: Positive feedback makes the strong grow stronger and the weak get weaker 36 This means that the consumer takes possible network effects into account when making a purchase decision, and that the economies of scale are particularly effective in making this decision. Take the example of a computer purchase. When purchasing, the user will commit to a specific software and hardware combination. This definition creates costs, whereby the consumer tries to invest his money in the best possible way. He must therefore take into account the current and especially the future developments of the technologies, because these will greatly influence the software offer in the future. When it comes to hardware, consumers will choose the one that they believe software developers will make the same decision on. This creates a kind of cycle. The greater the number of current and future consumers for a particular good, the more complementary goods will be offered in the future, and this results in a growing value of the product for the consumer. Finally, it should be noted that not only the current benefit, but also the benefit expected in the future is decisive for the current purchase decision, i.e. the demand, for the good cf. (Swann & Gill, 1993, p. 165) 36 cf. (Shapiro & Varian, 1999, p. 175) 37 cf. (Evans, Nichols, & Schmalensee, 2002, p. 38ff.); (Pohlmeier, 2004, p. 32)

15 Economies of scale on the demand side Network effects can give rise to economies of scale. 38 These economies of scale can arise on the supply side as well as on the demand side. 39 If, for example, the production costs for a quantity produced twice are less than twice as high, this is referred to as economies of scale on the supply side. Since the high fixed costs are now distributed over a relatively large production volume, the production costs do not grow so quickly. This effect is also called fixed cost degression. 40 It is important to mention that with extremely large production quantities, there are also high costs for initial investments, such as expensive machines. Thus one can assume that the size effect on the supply side has a limit. 41 In cases of overloaded production capacities and raw materials that are difficult to obtain, the company can face further cost problems. As a result, the cost of production can be greater than the economies of scale on the supply side. In reality it happens that the entire market cannot be supplied by one company because the corresponding limits have already been reached. The economies of scale on the supply side contrast with economies of scale on the demand side. These are usually achieved through network effects, since an increase in the number of users also increases the value of a product. As described in the previous chapters, the value of a main product increases with the number of users. The increased number of consumers also leads to a higher supply of complementary goods. It can thus be shown that an initially present economies of scale on the demand side will result in a future economies of scale on the supply side. The future size advantage on the supply side relates to the 38 cf. (Shapiro & Varian, 1999, p. 179) 39 cf. (Baldwin, 1987, p. 164ff.) 40 cf. (Pohlmeier, 2004, p. 33) 41 cf. (Shapiro & Varian, 1999, p. 179)

16 13 complementary goods that bring lower prices and higher quality with them. 42 In information technology today, it is assumed that there is a scale advantage in the form of a combination of supply and demand. 43 The costs on the supply side therefore decrease due to growth on the demand side. This in turn leads to a higher attractiveness, which brings with it an increasing demand. Leveraging Leveraging is understood to mean the possibility of expanding one's monopoly power from one market segment to another market segment when several products are combined into one good. This could lead to a displacement of possibly superior products. 45 Consumers strive to combine various components into a system, which can lead to a combination of indirect network effects. At this point one can cite the example of the OPEC countries (Organization of the Petroleum Exporting Countries). The OPEC countries generally have market power with regard to oil deposits. Hence the fear arose that OPEC could expand its power from the oil business to indirectly related industries. There are strong complementary relationships between the levels of indirect networks and indirectly integrated industries. 47 Another view of bundling is that it can promote price differentiation. 48 Furthermore, the literature speaks of a reduction in transaction costs when bundling products into goods. 49 The extent to which leveraging promises success depends 42 cf. (Kolasky, 1999, p. 577ff.) 43 cf. (Shapiro & Varian, 1999, p. 182) 44 cf. (Pohlmeier, 2004, p. 34) 45 cf. ( Reback, Creighton, Killam, & Nathanson, 1994); (Reback, Creighton, Killam, Nathanson, & Gross, 1995) 46 cf. (Siebert & Rauscher, 1985, pp. 211ff.) 47 cf. (Economides & White, 1994, p. 654); (Economides N., 1996, p. 676) 48 see (Posner, 1976); (Blair & Kasermann, 1978); (Schmalensee, 1982) 49 see (Williamson, 1979)

17 14 depends on the market structure. If a monopolist succeeds in bundling his products into goods, he can further expand his monopoly power in a new market. Two-sided markets A set goal or result is achieved in one-sided markets through appropriate measures or competition policy agreements. In two-sided markets, however, these measures cannot be used and they can lead to a negative result. It is therefore important to take a differentiated look at bilateral markets. 50 The concept of bilateral markets is based on the indirect network effects. In the case of indirect network effects, the benefit to the network participants increases when the other connected network grows. The newspaper market is a good example. This initially only consists of advertisers and the readers of the newspaper. On the one hand, advertisers benefit if the readership of the newspaper continues to grow. On the other hand, if the newspaper runs a lot of advertisements, readers will benefit, provided that the advertisements are informative. This form of advertisement creates a positive benefit for the readers. However, ads can also have negative benefits. These are known as negative externalities. In another case, the newspaper reader may be indifferent to the advertisement. The result would be a one-sided indirect network effect, whereby one can no longer speak of a two-sided market here. The newspaper publisher or media company acts as an intermediary between advertisers and readership, i.e. the two networks, which is referred to as an intermediary in the concept of two-sided markets (Figure 3: two-sided markets). The intermediary now sets the price and the circulation size for the newspaper with knowledge of the network effects. In addition to the price elasticity of demand and marginal costs, the strength of the 50 cf. (Evans, 2003b); (Evans, 2003a); (Evans & Schmalensee, 2005); (Dewenter, 2006a); (Dewenter & Kaiser, 2006); (Peitz, 2005); (Wright, 2004)

18 15 network effects to be taken into account. Price increases lead to a reduction in readership and at the same time to a shrinking number of advertising customers. This is followed by feedback to the readership. 51 From this example it can be concluded that prices are relatively low where the two markets react strongly to price changes. Consequently, an increase in prices is only attractive if the benefits outweigh the harm. Figure 3: Two-sided markets (source: Dewenter, 2006a, p.3) Empirical studies So far, only a few empirical studies have been carried out on indirect network effects. This can lead to inadequate results and assessments of these network relationships. 52 Mark Shurmer, for example, conducted surveys of computer users to investigate how the use of software and the goods used in a complementary manner correlate. So 51 cf. (Cordon, 1952); (Bucklin, Caves, & Lo, 1989); (Blair & Romano, 1993); (Chaudhri, 1998); (Dewenter & Kaiser, 2006) 52 cf. (Erhardt, 2001, p. 29)

In 1916, Shurmer was able to demonstrate a strong correlation between the software installed on the respondents' computers and the use of print media to obtain information and participation in training courses for the respective software. 53 Furthermore, Peter Swann and Jas Gill used the spreadsheet program Lotus 1-2-3, the same software that was also used by Neil Gandal for the empirical investigation of direct network effects (see Section 2.2.4), to investigate indirect network effects. To measure the network effect, Swann and Gill examined complementary goods, such as additional programs and additional products, as well as the availability of training opportunities. They received the result that the market leader, Lotus, had a higher level of utility than the products of its competitors. 54 As a result, the empirical studies to date have shown, even if there are not many, that there are positive correlations between complementary goods and that the indirect network effects can thus be empirically proven. 2.4 Real and virtual networks In addition to the direct and indirect networks already discussed, we also speak of real networks and virtual networks. Although these are very similar to the networks already discussed, they are important to mention in order to obtain a comprehensive overall picture of the network theoretical fundamentals. American economists differentiate between real and virtual networks. Here one refers genuinely and virtually to the function of a good that produces network effects. 55 One speaks of a real network when the sole purpose of one good is to establish a connection to another 53 cf. (Shurmer, 1993, p. 231ff.) 54 cf. (Swann & Gill, 1993, p. 165) 55 cf. (Pohlmeier, 2004, p. 37)