Convergence in Communications and Beyond 9780080547022, 9780444502162

Citation preview

EDITED BY

ERIK BOHLIN SCHOOL OF TECHNOLOGY MANAGEMENT 6, ECONOMICS, CHALMERS UNIVERSW OF TECHNOLOGY S-412-96 GOTEBORG, SWEDEN

KAROLINA BRODIN CENTER FOR COMMUNICATIONS AND INFORMATON RESEARCH STOCKHOLM SCHOOL OF ECONOMICS 55-113-83STOCKHOLM SWEDEN

ANDERS LUNDGREN CENTER FOR COMMUNICATIONS AND INFORMATION RESEARCH STOCKHOLM SCHOOL OF ECONOMICS S-I 13-83 STOCKHOLM SWEDEN

BERTIL THORNGREN CENTER FOR COMMUNICATIONS AND INFORMATION RESEARCH STOCKHOLM SCHOOL OF ECONOMICS S-l 13-83 STOCKHOLM SWEDEN

2000

ELSEVIER AMSTERDAM - LONDON - NEW YORK - OXFORD - PARIS - SHANNON

- TCO) and p constant is equivalent to shifting the distribution slightly to the right. The effect of this shift is unambiguously to shift some mass out of the leftmost cell: Prob(y = 1) must decline. Clearly, some probability is shifted into the rightmost cell: Prob(y = 4) must increase. What happens in the middle cells is ambiguous (Kennedy 1992). The partial effects are calculated for both continuous and discrete variables following the separate procedures outlined in Greene (1997), and are reported in Table 6. The marginal effects sum to zero across rows, which follows from the requirement that probabilities add to one. The most interesting result to arise from these calculations is that increasing FLAT RATE will shift demand towards higher use. This means an increase in an ISP's flat-rate price will lead to an increase in usage and exacerbate congestion, should it exist. An increase in usagesensitive prices has the potential to decrease usage.

8The likelihood ratio test of the hypothesis of homoskedasticity is based on h = n In s2 - C, n, In :s , where n ns; s2 is the pooled least-squares residual variance.

= C,

~

~

I15

8. Pricing and Residential Internet Trafic

Table 5 . Estimated Internet Use Duration Equation

Variable

Estimate Standard Error

t ratio

Economic indicators

INCOME1

0.1 137

0.2429

0.4679

INCOME2

0.4506*

0.1719

2.6215

INCOME3

0.2713*

0.1304

2.0799

INCOME4

0.0205

0.1 138

0.1798

ISP BILL

0.0224*

0.0025

8.7985

FLAT RATE

0.591 1*

0.1035

5.7105

TIMED USE

-0.3973*

0.1463

-2.7152

0.1201

0.1279

0.9390

CHAT

0.4623*

0.1111

4.1625

EMAIL

0.0534

0.1747

0.3056

FTP

0.2577*

0.1017

2.5345

GAMES

-0.0399

0.1212

-0.3293

CABLE

-1.0682*

0.3766

-2.8368

LESS288

0.2355

0.1966

1.1983

MONTHS Demographics

0.0089**

0.0046

1.9195

AGE

-0.01 11*

0.0034

-3.2207

MALE

0.1740

0.1272

1.3674

MULTIPLE USERS

0.0353

0.0956

0.3696

Pi

1.2952*

0.0800

16.1899

P2 Sample size 591

2.2434*

0.0958

23.4278

Log likelihood

-68 1.47

Restricted log likelihood

-793.94

Internet use indicators

BROWSE

Home technology

Partition boundaries

59 1

Note: pois normalised to zero. * denotes significant at the 5% level. ** denotes significant at the 5% level.

G. Madden and S. Savage

116

Table 6. Marginal Effects

HOURS=l

HOURS=2

HOURS=3

HOURS=4

INCOME1

-0.0228

-0.0307

0.0190

0.0345

INCOME2

-0.0789

-0.1059

0.0656

0.1192

INCOME3

-0.0478

-0.0642

0.0398

0.0722

INCOME4

-0.0051

-0.0068

0.0042

0.0077

ISP BILL

-0.0039

-0.0053

0.0033

0.0059

FLAT RATE

-0.1019

-0.1368

0.0847

0.1540

TIMED USE

0.0648

0.0870

-0.0539

-0.0979

MONTHS

-0.0016

-0.0022

0.0014

0.0025

BROWSE

-0.0242

-0.0325

0.0201

0.0365

CABLE

0.1765

0.2370

-0.1468

-0.2667

CHAT

-0.0799

-0.1073

0.0665

0.1207

EMAIL

-0.0310

-0.0416

0.0258

0.0468

FTP

-0.0454

-0.0610

0.0378

0.0687

GAMES

0.005 1

0.0069

-0.0043

-0.0077

LESS288

-0.0414

-0.0556

0.0345

0.0626

0.0016

0.0022

-0.0014

-0.0025

MALE

-0.0336

-0.0452

0.0280

0.0508

MULTIPLE USERS

-0.0085

-0.01 14

0.0071

0.0128

Variable

AGE

7.

CONCLUSIONS

The Internet has public good characteristics. Unrestricted access leads to network overuse and creates congestion. The approach adopted so far is a supply-side (or engineering) one. There is no doubt that software programs and innovative network hardware developments have expanded bandwidth exponentially. Despite these capacity improvements, diminished QoS and a general decline in user comfort are regularly reported. Further, expanding the capacity of a public good often only allows latent demand to be realised. Economists advocate demand-side solutions to alleviating congestion through better demand management, the efficient pricing

8. Pricing and Residential Internet Traffic

117

of complements and substitutes, and by setting efficient use prices. Many governments are currently considering reforms to the telecommunications sector that will see a more efficient provision of Internet complementary and substitute services such as international and longdistance telephony and data transfers. The application of traditional economic policy to curb Internet congestion, through efficient pricing by taxing use, will most likely reduce congestion. However, such policies ignore the positive network externalities that are correlated with network size and which offset congestion costs. Other than the standard connectivity effect, positive network externalities include: the stimulation of market innovations such as electronic commerce and the development of new markets; and the provision of access to persons who would otherwise not be able to participate in this technology. Several attempts have been made to develop prices that reflect the social cost of use. Both optimal and edge pricing paradigms propose some form of usage-sensitive pricing. Setting aside the practical difficulties of determining what the appropriate prices should be, little attention has been given to the consumer acceptance of usage-sensitive pricing as opposed to the dominant flat-rate scheme that is currently in operation. The intended contribution of this study is to provide evidence as to acceptability or otherwise of usage-sensitive prices to consumers. To operationalise this notion a web-based survey was developed and offered to Australian residential Internet users. The model estimates clearly indicate that ISP subscribers on flat-rate agreements use the Internet more. Survey data suggest, that ISP subscribers prefer flat-rate pricing (or a combination of flat-rate and usage-sensitive pricing) over usage-sensitive pricing schemes. Ordered logit model estimation confirms this preference. Respondents on flat-rate pricing plans are likely to use the Internet for a greater number of hours per week than respondents on usagesensitive pricing plans. Model results suggest that if Australian ISPs continue down the path initiated by AT&T in 1996 and introduce low-cost flat-rate unlimited Internet access plans, then there will be increased consumption and worsening congestion. The imposition of usagesensitive pricing plans will force consumers to reduce (or shift) use but potentially at the cost of reducing total usage and hindering the development of an important new technology. Thus, usage-sensitive pricing appears a very blunt instrument whose use is not costless. Whilst this research represents the first Australian study of Internet use which examines the impact of flat versus usage-sensitive pricing, many basic questions still need to be addressed. Among these from a policy perspective is the applicability of the results to populations outside Australia. Second, the analysis of business demand, in particular small business demand, is outside the ambit of this study.

REFERENCES ABS (1 998), Use of the Internet by Householders, Australia, February 1998, Catalogue Number 8147.0, AGPS, Canberra. Alston, R. (1998), Building the Information Economy, Media Release, June 11. Ben-Akiva, M. and Lerman, S.R. (1985), Discrete Choice Analysis, MIT Press, Cambridge. Berndt, E., Hall, B., Hall, R. and Hausman, J. (1974), “Estimation and Inferences in Nonlinear Structural Models”, Annals of Economic and Social Measurement,Vol. 3 , No. 4, pp. 653-666. Bohn, R., Braun, H.W., Claffy, K.C., and Wolff, S. (1994), “Mitigating the Coining Internet Crunch: Multiple Service Levels via Precedence”, Technical Report, University of California, San Diego.

118

G. Madden and S. Savage

Claffy, K. (1 996), “Closing the Internet Statistics Gap”, in TeleGeography 1997: Global Telecommunications Traffic Statistics and Commentary, TeleGeography, Washington, D.C. CNN ( 1 996), “MCI, BT Plan to Stretch the Web”, 10 June; ~littp://cn1ifii.co1ii/~iews/960610 linci_bt/index.litinl>. DFAT (1997), Putting Australia on the New Silk Road: The Role of Trade Policy in Advancing Electronic Commerce, AGPS, Canberra. Greene, W.H. (1997), Econometric Analysis, Third Edition, Prentice Hall International, New Jersey. ITU (1 997), Challenges to the Network: Telecoms and the Internet, ITU, Geneva.

Kennedy, P. (1992), A Guide to Econometrics, Third Edition, MIT Press, Cambridge. MacKie-Mason, J.K. and Varian, H.R. (1999, “Pricing the Internet”, in Kahin, B. and J. Keller (eds.), Public Access to the Internet, Harper & Row, New York. MacKie-Mason, J.K. and Varian, H.R. (1 996), “Some Economics of the Internet”, Department of Economics Technical Report, University of Michigan. Metcalf, B. (1 999, “Predicting the Internet’s Collapse”, Info World, Vol. 4 , December. NSF (1 996), “Report on the NSF-sponsored Workshop on Internet Statistics Measurement and Analysis”, February 19-20; .

Network Wizards (1998), Internet Domain Survey; . OECD (1 996), Information Infrastructure Convergence and Pricing: The Internet, Committee for Information, Computer and Communications Policy, OECD, Paris. OECD ( I 997), Information Technology Outlook, OECD, Paris. OECD (1 998), Internet Traffic Exchange: Developments and Policy, Working Party on Telecommunication and Information Services Policies, OECD, Paris. Paxson, V. (1994), “Empirically Delivered Analytical Models of Wide Area TCP Connections”, IEEE/ACM Transactions on Networking, Vol. 2. Sarkar, M. (1997), “Internet Pricing: A Regulatory Imperative”, in McKiiight, L.W. and J.P. Bailey (eds.), Internet Econonzics, MIT Press, Cambridge. Shenker, S., Clark, D., Estrin, D., and Herzog, S. (1996), “Pricing in Computer Networks: Reshaping the Research Agenda”, Teleconzmunications Policy, Vol. 20, pp. 183-201. TeleGeography Inc. (1 996), TeleGeography 1996: Global Telecomniunications Traffic Statistics and Conmentary, TeleGeography, Washington, D.C.

8. Pricing and Residential Internet

Tragic

119

APPENDIX With utility inaxiinised locally, an interiin decision considers whether the current (discrete level of) Internet usage is adequate or a longer duration is required. The process ceases when the host is satiated in Internet use (Ben-Akiva and Lerinaii 1985). Denote as U,," the utility host n receives froin continuing to consume inore Internet tiine given that current consuinption is contained within band i, and as U,,,' the utility obtained by n stopping at band i when current consuinption is contained within band i. Accordingly, the probability that host n is satiated with band i consuinption is:

With utility randoin and the sum of deterministic (V) and stochastic components, the disturbances of the utility differences (U,,"- U,,,") assuined independently logistically distributed, this joint event can be written as the product of binary logit probabilities:

0 2000 Elsevier Science B.V. All rights reserved Convergence in Communications and Beyond E. Bohlin, K. Brodin, A. Lundgren and B. Thorngren (Editors)

CHAPTER 9

Closing in on the Web Consumer - A Study in Internet Shopping Micael DahlCn Center f o r Consumer MarketingKenter f o r Information and Communications Research, Stockholm School of Economics, Stockholm

Abstract. In this paper, a study of Internet users is presented. The purpose of the study is to learn more about who shops and who will shop on the Web. With the help of theories on store choice, distance shopping and adoption of innovations, comparisons are made between shoppers and non-shoppers. Three different shopper categories are identified: shoppers, active non-shoppers and passive non-shoppers. These are found to differ with respect to several factors, such as shopping orientation, socioeconomic status, attitudes towards distance shopping and Internet usage experience. Implications of these findings are discussed.

1. INTRODUCTION Some say the Internet will take over all commerce. Everybody will soon shop on the Web. Others say it’s just a fad, that the Net cannot offer anything of lasting value. Many systems for electronic commerce have failed before (e.g. Prestel, the internationalization of Minitel, and more; cf. Falk, 1983). The World Wide Web has some advantages over prior systems, of which the most prominent are its user-friendliness and the critical mass of Web users. The diffusion of Internet use continues rapidly, and it has been called the fastest-growing medium of all times (Eighmey and McCord, 1995). A solid platform for trade is being established with many active companies and a large global - or local - customer base. However, not every Internet user shops via the computer and there is hardly any reason to believe that everyone will in any foreseeable future. For example, in the latest GVU surveys, the proportion of shoppers seems to stabilize around 18-19 per cent (Pitkow and Kehoe, 1997) of Internet users. The same figure, 19 per cent, is reported in a poll by Businessweek (1997). Internet is one marketing channel in competition with many others, and will have to comPete with supermarkets, department stores, convenience outlets, traditional mail order catalogues and so on for the consumers’ favor. Alba et al. (1997) explain the existence of several different store formats by the fact that consumers are heterogeneous. Different formats satisfy different needs and, hence, different consumers. The question is, who will choose to shop on the Internet and why?

122

M. Dahlin

2. THEORY The Internet as a shopping channel bears a resemblance to retail store formats as well as traditional mail order. The Net offers the interaction of a store (or, at least, it is getting closer and closer), while the transactions are still made over distance like mail order. When trying to learn why people shop - or do not shop - on the Internet, guidance can be obtained from research on these two shopping channels. The Internet offers a direct link between seller and consumer, and is hence the electronic equivalent of the physical retail store. Thereby, the Web competes with all other retail formats for the consumers’ patronage. A way of assessing the new shopping channel’s impact on consumer shopping may thus be to relate it to existing retail formats from a consumer viewpoint. As mentioned earlier, consumers have different needs and therefore there exist several retail formats to cater to these needs. The prospective Internet shoppers could be found among consumers with needs and behaviour that can be fulfilled by this new shopping channel, in relation to other channels. Alba et al. (1997) claim that the Internet offers the same advantages as traditional mail order and is superior to this format. Hence, the features that attract consumers to mail order shopping should also attract them to Web shopping and existing mail order customers ought to be a potential target group for Internet vendors. The Internet can be viewed as the intersection between retailing and mail order. Combining these two formats into a new shopping channel faces the consumer with a decision whether to change hidher patronage. This resembles the process of adopting an innovation. Shopping on the Web is a new way of doing things and entails a process of trying and evaluating something novel, an innovation. Hence, theories on adoption of innovations and innovativeness could shed light on who shops and will shop on the Web. Although the Internet is a rather new phenomenon, consumer shopping behaviour has been a subject of research for a long time and it is believed that existing theory on how and why people shop in different ways can be applied in this new setting, as the people - and their needs - are still the same. Following the logic that different store formats appeal to different consumers, one can speculate that the Internet will, in the short term at least, not appeal to everyone, but rather to a special kind of consumers -Web consumers. Many models of store choice and general buying decision models have been constructed and tested empirically (cf. Hansen, 1972, for an overview). Overall, these models consist of two kinds of influencing factors, These are store characteristics and consumer characteristics. Examples of the former are assortment, prices, service and location (cf. Isaacson, 1966; Schiffman et al., 1977; Malhotra, 1983; Engel et al., 1995). A comparison of characteristics between the Internet and other shopping formats is made by Alba et al. (1997). In this paper, focus will be on consumer characteristics. A prominent feature in store choice models is the consumer’s shopping orientation (cf. Spiggle and Sewall, 1987; Solomon, 1992; Laaksonen, 1993). Stone (1954) categorizes consumers into four different shopping types. The economic consumer is rational, goal-oriented and seeks to maximize the value of his or her money. The personali~edconsumer wants service and personal contact, while the ethical consumer shops with a conscience, e.g. supporting the local store. The apathetic consumer, finally, sees shopping as a necessary but unpleasant chore to be dealt with as painlessly as possible. Bellenger and Korgaonkar (1980), in a later study, identify a fifth shopping type called the recreational shopper. This person enjoys the shopping experience and devotes much time to it. Solomon (1992) puts these together into a

9.Closing in on the Web Consumer - A Study in Internet Shopping

123

sct of five shopping categories. Laaksonen (1993) points to the fact that a consumer can be found in different categories depending on the product (e.g., a person may be an apathetic grocery shopper and a recreational clothes shopper). The consumer’s socioeconomic status is also modeled to have an impact on store choice (cf. Stone, 1992; Engel et al., 1995). Here we find factors such as age, sex, income and education. Berenson (1976) proposes that consumer confidence affects store choice. This is also found in studies on mail order and other kinds of distance shopping customers. Cox and Rich (1964) found that telephone order customers perceived a lower risk in shopping through the channel than did people who did not buy on telephone order. Spence et al. (1970) conclude that risk perception is an important factor when deciding whether to purchase something on mail order. Looking more into the literature on mail order, Cohen (1982), Simon (1984) and Skaug (1994) all mention convenience as an important reason for shopping through the channel. Relatedly, dzaiculty in getting to a store is also a factor which increases the likelihood of purchasing the product on mail order. So far, we have viewed the Internet as a shopping channel, like a retail store or mail order. As a new kind of shopping channel, the Net can also be conceived as an innovation (cf. Kotler, 1991). Purchasing something on the Internet could be considered as adopting an innovation. Adoption models have much in common with buying decision models (for a review of different adoption models, see Robertson, 1971). Models like the classic A D A start with awareness as the first step in the adoption process. Awareness and knowledge are essential, since the consumer must know of and know something about the innovation before s/he will consider adopting it. The next step in the process is interest. Interest in innovations is generally thought to be dependent on the consumer’s degree of iizizovativeness (Paltschik, 1985). People with a higher degree of innovativeness are believed to be more prone to adopt innovations. Consumers can be categorized as adopters, active rejecters or passive rejecters of an innovation (Engel et al., 1993). Adopters embrace the new product, while rejecters do not. Of the latter, active rejecters make an informed decision not to adopt. whereas passive rejecters really have not taken this into consideration. DahlCn (1997), in a study of Internet usage, finds that users mature as they become more experienced. This maturation process means that users become more focused and practically oriented in their use of the Net. As a result, the Internet is viewed more as an instrument to be used for practical purposes. Shopping could be such a purpose. Hoffman and Novak (1997) find that experienced users are more likely to buy things over the Internet.

3. HYPOTHESES To summarize the previous section, we have compared the Internet as a shopping channel to retail stores and mail order. Consumer characteristics affecting the choice of store or shopping channel have been discussed. The Internet has also been looked upon as an innovation in the sense of a new shopping channel, and factors affecting the adoption of innovations have been elaborated upon. Finally, the concept of Internet maturation as users become more experienced, and its possible effect on shopping behaviour, have been considered. This raises the

M . Dahle‘n

124

following hypotheses regarding shopping behaviour (Le. whether a person shops on the Internet or not). H1: Shopping behaviour is stable and predictable, as present Web shoppers are more prone to shop on the Internet in the future than non-shoppers.

H2: Shopping orientation (as discussed above) affects shopping behaviour.

H3: Socioeconomic status affects shopping behaviour. H4: Risk perception regarding traditional distance shopping, and distance shopping behaviour influence Web shopping proneness. Greater perceived risk will decrease the inclination to shop, whereas more frequent distance shopping behaviour will increase the inclination for Web shopping.

H5: Perceived difficulty in getting to retail stores affects Web shopping proneness positively.

H6: Knowledge of where one can shop on the Internet affects the likelihood of Web shopping positively. H7: Innovativeness affects Web shopping proneness positively. H8: The inclination to shop increases with Internet usage experience The first hypothesis concerns the belief that Web shoppers are a special kind of consumers and that not everyone will shop on the Internet. Identifying the present shoppers would therefore give a good idea of future Web consumers, since they are by and large expected to be the same. The second hypothesis states that the consumer’s shopping orientation will affect his or her store choice and thereby Internet shopping behaviour. It is believed that the Net attracts some shopping types more than others. The third hypothesis focuses on socioeconomic factors such as income, education, profession, age and sex. It is hypothesized that these factors will influence shopping behaviour. The next two hypotheses concern attitudes towards traditional distance shopping and distance shopping behaviour. It is hypothesized that Internet shoppers have a more favourable attitude towards distance shopping and use this channel more frequently than Internet nonshoppers (H4). It is also believed that Internet shoppers perceive greater difficulty in getting to retail stores (H5). Hypotheses H6 and H7 concern shopping on the Internet as adoption of an innovation. Since awareness and knowledge normally are the first steps in the decision process, it is hypothesized that Web shoppers are more knowledgeable with respect to shopping sites (H6). Web shoppers are also hypothesized to be more innovative than non-shoppers (H7). Finally, it is believed that Internet usage experience makes users more instrumental and thereby more inclined to shop on the Web. In the eighth hypothesis it is assumed that Web shoppers are more experienced Internet users.

4.METHOD To test the hypotheses generated above, a survey was conducted. This was done in cooperation with Passagen, Sweden’s second most visited Web site at the time. Using Java script,

9. Closing in on the Web Consumer - A Study in Internet Shopping

125

every 200th visitor to the Passagen Web site was approached with a dialogue box upon entering the site. This dialogue box stated the purpose of the study and asked the visitor to participate and fill out a questionnaire. Consenting visitors were transported to a Web site with an on-line questionnaire. With the help of cookie files, all visitors were “tagged” so that once having visited the Passagen site a visitor could not be selected upon a new visit, whether selected or not the first time. Thereby, a random selection without replacement sampling procedure was enabled, in order to obtain as representative a sample as possible of the Passagen visitors (cf. Newbold, 1991). There were no links to the survey site elsewhere, to ensure control of the sample. For more details on the methodology, see DahlCn (1998). During the first week of June 1997, 2642 responses were recorded. All respondents were anonymous, thus making an analysis of non-response bias impossible.

5.

RESULTS

The representativeness of the results must be viewed with the sample and response rate in mind. The method of sampling on-line on a specific Web site yields a sample that is not randomly selected from the entire population. To analyze the results, statistical methods in the form of cross-tabulations and ANOVAs have been used. The cross-tabulations should be read horizontally, comparing the percentages for the different factor groups over the dependent variable. Column percentages are compared row-wise with the total row percentage as basis. ANOVAs with Tukey’s Post Hoc test compare mean values between the three shopper categories (see next paragraph). Following the distinction between adopters, active and passive rejecters, the respondents were divided into three shopping-behaviour, or shopper, categories: shoppers, active nonshoppers and passive non-shoppers. Shoppers were defined as consumers who at some time had purchased something on the Internet. Active non-shoppers had tried or simulated shopping on the Internet but not actually bought anything. Passive non-shoppers, finally, were defined as those who had neither tried or simulated shopping nor bought anything on the Net. In the survey it was found that around 17 per cent had bought something on the Internet. This figure corresponds fairly well with the surveys mentioned in the introduction. The nonshoppers could be divided into approximately 11 per cent active and 7 2 per cent passive nonshoppers. These three categories were used as a basis of comparison in the following analyses.

5.1 Buying Intention and Shopping Orientation To test hypothesis 1, the respondents were asked about their intention to buy anything on the Internet in the near future. There were several response categories, ranging from within a week to more than a year from now. A dichotomous variable was constructed, coding response categories within a year as buying intention and others (“more than a year”, “never”, “do not know”) as no intention. The results are pictured below. Approximately 20 per cent of the respondents report having intentions to shop on the Internet. This is only slightly more than the share of present shoppers of 17 per cent, mentioned above. In Table 1, it is shown that the consumers who intend to shop on the Net are clearly over-represented among the present shoppers, as suspected. Among the passive non-shoppers

M . Duhlitz

I26

there is a clear over-representation of consumers with no intention to shop on the Web. Interestingly, it can be seen that consumers with a shopping intention are over-represented also among active non-shoppers. The results favour the beliefs expressed in hypothesis H1. The ,first hypothesis is supported: present Internet shoppers are more prone to shop on the Internet in the,future than non-shoppers. In order to test the second hypothesis, respondents were asked to categorize themselves as the one of five shopping types who closest resembled their shopping orientation. This was done for each of several product categories. The product categories selected were those of products that are frequently bought on the Internet or through traditional distance shopping channels. For each product category, contingency tables were produced comparing shoppers and non-shoppers for different shopping types. The following results were obtained. Table 1. Buying Intention in Different Shopper Categories

Shoppers Buying Intention

No

Yes

Total

Chi-square = 706.7; p c 0.0001

Count % Within Shopper Categories Count % Within Shopper Categories Count 97Within Shopper Categories

Shopper Categories Active Passive Non-shoppers Non-shoppers

145

182

34.7%

68.95%

1605

92.1 70

Total 1932 79.1%

273

82

138

493

65.3%

31.1%

7.9%

20.3%

418

264

1743

2425

100.0%

100.0%

100.0%

100.0%

9. Closing in on the Web Consumer - A Stud)) in Internet Shopping

127

Even though relatively few purchases had been made in each product category, significant results were obtained in several categories. In Table 2, it is shown that there were two shopping types that had shopped more often on the Internet than other shopping types. These are the economic consumers and apathetic consumers. Hypothesis H2 is thus supported: shopping orientation has an effect on Internet shopping behaviour. Table 2. Shopping Types with Over-representations of Web Shoppers

Product Category

Shopping Type

Groceries

Apathetic Consumer""

Technical Appliances

Economic Consumer" * *

Sports/Health/Leisure

Not Significant

Apparel

Economic Consumer**

Software

Apathetic Consumer:'

Literature

Economic Consumer; Apathetic Consumer'k:k

Music/film

Not Significant

BanWInsurance Services

Apathetic Consumer:l:'k*

Tickets (Travel, Events, etc.)

Apathetic Consumer*'k

5.2 Socioeconomic Influences Testing the hypothesis that socioeconomic status affects shopping behaviour, the following variables were analyzed: income, age, sex, education and occupation. Using ANOVA analysis, mean monthly income before taxes was compared between the three shopper categories. Age, sex, education and occupation were analyzed with the help of contingency tables. As shown in Table 3, shoppers had higher incomes than non-shoppers. On average, shoppers' incomes were 6100 Swedish kronor higher than active non-shoppers' (p = 0.02) and 6500 Swedish kronor higher than passive non-shoppers (p < 0.0001). Table 3. Mean Monthly Income before Taxes Shopper type

Mean Monthly Income

Shoppers

21,300 SEK

Active Non-shoppers

15,200 SEK

Passive Non-shoppers

14,800 SEK

M . Dahle'n

I28

Table 4. Shopper Categories by Age Aae 20-27

27-40

years

years

- 19 years Shopper

Shoppers

categories

Count %Within

50

81

9.0%

14.6%

Count

48

71

8.6% 457

+ years

Total

97

373

19.1%

17.0%

67

59

245

12.8%

11.6%

11.6%

11.2%

401

368

351

1577

72.5%

63.4%

69.2%

71.8%

555

553

580

507

2195

100.0%

100.0%

100.0%

100.0%

100.0%

Age Active

40

145

25.0

Non-shoppers yo Within Age Passive Count Non-shoppers yoWithin

82.3

Age Total

Count %Within Age

Chi-square = 65.3; p < 0.0001

As shown in Table 4, there was an over-representation of Web shoppers among the respondents especially in the ages of 27 through 40. The passive non-shoppers, on the other hand, are over-represented among the younger age categories. Table 5 . Shopper Categories by Gender

Shopper categories

Shoppers

yo Within

Active Non-shoppers

Total

Count Sex Count yo Within

Sex Passive Count Non-shoppers yo Within Sex Count Sex

Male 307

Female 82

19.7%

115%

155

95

250

10.0%

13.3%

1 1.0%

1093

539

1632

75.3'7"

70.3% 1555

100.0%

'

Total 389 17.1YO

71.9%

71 6

2271

100.0%

100.0%

Chi-square = 26.2; p c 0.0001 There was a significant under-representation of shoppers in the category of female respondents, whereas there was an over-representation of non-shoppers, as seen in Table 5. The opposite holds for male respondents.

9. Closing in on the Web Consumer - A Study in Internet Shopping

129

Table 6. Shopper Categories by Education College/

ShODDer ShopDers , . categories

Count %Within Education Active Count Non-shoppers yo Within Education Passive Count Non-shoppers yo Within Education

Total

Count %Within Education

Junior High School High School 45 171 16.6%

University 162

19.8%

Post Graduate Studies 18

Total 396

29.5%

17.2%

33

108

103

11

255

8.3% 319

10.5%

12.6%

18.0%

11.1%

80.4%

72.9%

749

555

397

1028

100.0%

100.0%

67.7% 820 100.0%

32

1655

52.5%

71.8%

61

2306

100.0%

100.0%

Chi-square = 33.3; p < 0.0001

Turning to education, respondents were asked about the highest level of education they had finished or attended at the present time. Table 6 shows a bias in the sense that a significantly larger share of shoppers and active non-shoppers was found in the higher education categories. The under-representation of shoppers in the lower educational levels could partly depend on the fact that these respondents were still in school with lower disposable incomes. Table 7. Shopper Categories by Occupation Occupation WhiteSelfBlue-collar collar Researcher Unemployed Employee Employee employed 25 52 183 10 22

Student 96

Homemaker _ or . . Retired 7

11.4%

20.0%

Active Count 86 Non-shoppers oo/ Within Occupation 10.2%

3

11

29

108

8.6%

11.6%

11.6%

25

59

71.4%

Shopper Shoppers Categories

Count %Within Occupation

Passive Count 661 Non-shoppers Within Occupation 78.4%

26.3%

20.8%

20.8%

33.3%

Total 395

12.6%

17.1%

7

12

256

12.394

23.3%

6.9%

11.1%

169

588

13

140

1655

62.1%

67.6%

66.9%

43.3%

80.5%

71.8%

35

95

250

879

30

174

2306

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

Q/~

Total

Count 843 %Within Occupation 100.0%

Chi-square = 62.7; p < 0.0001

In Table 7, it is shown that shoppers were under-represented in the student category. There was an over-representation of shoppers and active non-shoppers in the categories of blue- and white-collar employees, self-emp1,oyed and researchers. Basically, this means everybody with a job.

M. Dahlin

130

Summarizing the tests, it was found that shoppers and active non-shoppers had higher income on average, whereas respondents in the ages of 27-40 years were especially prone to shop. Web shoppers were over-represented among the male kespondents, while non-shoppers were over-represented among female respondents. Shoppers and active non-shoppers were shown to be over-represented in the higher education categories. As for occupation, employees overall and self-employed respondents showed similar tendencies; shoppers were overrepresented among these. In conclusion, results were mixed. Hypothesis H3 is supported: socioeconomic status affects shopping behaviour.

5.3 Distance Shopping Attitudes and Behaviour For the fourth hypothesis, questions were constructed to measure consumers’ attitudes towards traditional distance shopping, focusing on risk perception. One question concerned respondents’ ability to judge the product beforehand, and the risk that the product would not live up to expectations. Respondents were asked to agree or disagree with the following statement: “the product rarely lives up to expectations”. Agreement was measured on a seven-point Likert scale, where 1 meant “agree completely” and 7 meant “disagree completely”. In the same way, agreement was measured with statements concerning the accuracy of distance shopping deliveries: “the wrong product is often delivered”, “the product is often damaged in the transport”, “the product is often delivered late or not at all”. From the responses to these three questions, a mean index was computed, “delivery reliability”. Responses were compared between the shopper categories, using ANOVA analysis. The results are pictured below. Table 8. Mean Agreement with Statements of Respondents’ Ability to Judge Distance Shopping Shoppers

Active Non-shoppers

Passive Non-shoppers

Product Judgement Reliability

4.5

2.8

2.3

Delivery Reliability

5.1

4.4

3.1

Differences (Tukey’s Post Hoc) between all groups p < 0,0001; Scale: 1 = “agree completely” and 7 = “disagree completely” Passive non-shoppers were less confident than the others that the product would live up to expectations, as indicated in Table 8. Web shoppers were the only ones who were somewhat closer to disagreement than agreement with the statement that the product rarely lives up to expectations. Overall, respondents had more confidence in delivery reliability. Shoppers were fairly positive, while active non-shoppers were close to neutral and passive non-shoppers fairly negative. Distance shopping behaviour was also tested for. The respondents were asked how many times they had bought something on mail order and television home shopping in the last year. Table 9 shows the mean number of times for shoppers and non-shoppers.

Y,Closing in on the Web Consumer - A Study in Internet Shopping

131

Table 9. Mean Number of Purchases on Mail Order and Television Home Shopping Shoppers

Active Non-shoppers

Passive Non-shoppers

Mail Order

4.0

2.8

1.1

TV Home Shopping

0.7

0.6

0.4

For purchases on mail order, significant differences were found between the groups (all differences p < 0.0001, Tukey's Post Hoc). Shoppers were four times more frequent mail order customers than passive non-shoppers. Although statistically significant, differences in TV home shopping purchases were small, as this channel did not seem to be very popular. In summary, Web shoppers seemed to be more favourable towards traditional home shopping, in attitude and behaviour (except for TV home shopping). Hypothesis H4 is thus supported: Internet shoppers perceive lower risk in traditional distance shopping and are more frequent distance shoppers. Risk perception regarding traditional distance shopping, and distance shopping behaviour, influence Web shopping proneness. The fifth hypothesis was tested with a question about how difficult respondents perceived it to be to get to a retail store. The respondents were asked, for the same product categories as in the previous analysis, to indicate on a 7-point scale how difficult it is to get to a retail store. The answers were compared product-wise between those who had bought the product on the Internet and those who had not. The results are shown in Table 10. Table 10. Perceived Difficulty in Getting to Retail Store Product Category

Shoppers

Non-shoppers

Difference

Groceries

2.94

1.42

1 S2""

Technical Appliances

3.64

2.65

0.99""

Sports/Health/Leisure

2.36

2.14

Not Significant

Apparel

2.92

2.06

0.86""

Software

3.93

2.65

1.28""

Literature

3.20

2.19

1.01""

Music/Film

2.94

2.29

0.65"

BanWInsurance Services

3.03

2.23

0.70"

Tickets (Travel, Events, etc.)

4.1 1

2.65

1.46""

* = -- p < 0.01; ** = p < 0.0001; Scale: 1 = "retail store very easy for me to get to" and 7 = "retail store impossible for me to get to"

M. Dahlin

I32

Statistically significant differences are found in all product categories but one. Judging from the responses overall, the respondents do not perceive any great difficulties in getting to retail stores. Only in the case of tickets did the Web shoppers have an average above the neutral point of 4. Nevertheless, the pattern is consistent. Web shoppers on average perceive greater difficulties in getting to retail stores. Hypothesis H5 is thereby supported: Web shoppers perceive it to be more difficult to get to a store. Perceived difficulty in getting to retail stores affects Web shopping proneness positively.

5.4 Shopping Knowledge, Innovativeness and Internet Experience The sixth hypothesis states that Web shoppers are more knowledgeable than non-shoppers with respect to shopping opportunities on the Net. This was tested by asking the respondents how many shopping sites they knew of. The mean stated numbers of sites were compared between the shopper categories using ANOVA analysis. The following results were obtained. Web shoppers on average knew of 14.1 shopping sites on the Net. This is significantly more (p < 0.0001) than non-shoppers. Also, the active non-shoppers knew of significantly more shopping opportunities on the Web (mean = 9.2) than the passive non-shoppers (mean = 3.0). Hypothesis H6 is supported: Web shoppers have greater knowledge of where one can shop on the Internet. Knowledge of where one can shop on the Internet affects the likelihood of Web shopping positively. Testing the idea from hypothesis H7 that Web shoppers are more innovative, respondents were faced with three statements. They were asked to indicate their agreement with the statements on a 7-point scale, where 1 = “agree completely” and 7 = “disagree completely”. The mean ratings were compared between the shopper categories using ANOVA analysis. Results are pictured below. Table I 1. Innovativeness in the Different Shopper Categories Statement

Shoppers

I often try new things

1.89

Active Non-shoppers 2.1 1

Passive Non-shoppers 2.34

I wait and see others try new things before I try myself I do like everyone else

5.02

4.64

4.41

5.73

5.64

5.53

Scale: 1 = “agree completely” and 7 = “disagree completely” As shown in Table 11, the mean ratings differ between the shopper categories in the hypothesized direction. The differences between shoppers and non-shoppers are statistically significant (p < 0.0001) for the first two statements. However, the differences are small. The respondents exhibit a rather high degree of innovativeness overall. The seventh hypothesis is not supported. The final hypothesis, H8, states that Web shoppers are more experienced Internet users. Following the conceptualization of DahlCn (1997a), experience is operationalized as length of

9. Closing in on the Web Consumer - A Study in Internet Shopping

133

usage for the purpose of this test. Respondents’ reports on how long they had used the Net were compared with ANOVA analysis. As suspected, Web shoppers were found on average to have used the Internet for a longer time than non-shoppers. Web shoppers had had access to the Net for 15.9 months, 5.1 months longer than active non-shoppers (mean 10.8 months) and 6.1 months longer than passive nonshoppers (mean 9.8 months) (both differences p < 0.0001). Approximating experience with length of usage, the hypothesis is supported: Web shoppers are more experienced Internet users. The inclination to shop increases with Internet usage experience.

4. DISCUSSION Summarizing the results in the previous section, it was found that about 17 per cent of the respondents had bought something on the Internet. Adding to this. 11 per cent had shown interest in shopping on the Internet, having tried or simulated Web shopping. This group was called active non-shoppers, indicating an active decision not to shop on the Internet - as opposed to the 72 per cent non-shoppers who had not even tried or simulated shopping on the Web. Testing for buying intention, however, showed that the respondents with buying intentions were clearly over-represented among the active non-shoppers. Thus, the active nonshoppers could be regarded as soon-to-be shoppers. When asked about their intentions of buying something on the Internet, 20 per cent of the respondents stated that they planned on shopping within a year. This is only slightly more than the present share of Web shoppers. It was found that there was a very clear overrepresentation of people with buying intentions among the present shoppers and active nonshoppers. The buying intention of passive non-shoppers was very low. This indicates a fairly stable pattern, where a distinction between shoppers, active non-shoppers and passive nonshoppers is meaningful and useful in an analysis to learn more about who shops and will shop on the Internet and why. The three shopper categories consisting of shoppers, active nonshoppers and passive non-shoppers were used as a basis for analysis in testing the proposed hypotheses. Looking at consumers’ interaction with the retail industry, Stone (1954) and authors after him identified different shopping types. In this study, consumers’ shopping types were examined and compared between shoppers and non-shoppers on the Net. It was found that Web shoppers were clearly over-represented among two shopping types in the different product categories. These were the economic consumers, who look for bargains and are priceconscious, and the apathetic consumers, who strive for convenience and do not like the shopping experience. The hypothesis that shopping orientation affects Web shopping behaviour was supported, in the sense that economic and apathetic consumers are more prone to shop on the Internet. Socioeconomic status was compared between the shopper categories in order to find out whether this affects shopping behaviour. It was found that age seemed to matter. Income was higher among shoppers. Also, gender had an effect in the sense that Web shoppers were overrepresented among men while especially active non-shoppers were over-represented among women, indicating a more cautious approach to Web shopping by the latter. Higher-educated consumers were found to have an over-representation of Web shoppers. Overall, socioeconomic status does seem to matter. Demographically, the traits that have been ascribed to early

134

M . Dahlkn

Web users in previous consumer surveys such as GVU (1995) and others (the classic picture of the Internet user), namely male, 30-ish, well educated, wiih fairly good income, were overrepresented among Web shoppers in this study. As the Web bears a resemblance to traditional distance shopping channels such as mail order, it was hypothesized that Web shoppers would have more favourable attitudes towards distance shopping and be more frequent distance shoppers than non-Web shoppers. The suspicions were confirmed, as Web shoppers perceived less risk in distance shopping, thought they were better able to judge products beforehand, and shopped more on mail order than nonshoppers. It was also found that Web shoppers perceive it as more difficult to get to retail stores than do non-shoppers, as hypothesized. This confirms the resemblance of the Web to traditional distance shopping, from the consumers’ point of view. Using the Web as a shopping channel can be viewed as adopting an innovation. In order to adopt something new, consumers must be aware of it and know something about it. Therefore, it was hypothesized that Web shoppers would be more knowledgeable with respect to shopping opportunities on the Net. The hypothesis was confirmed. Web shoppers should also be more innovative than the other categories, since they were the first to adopt the new shopping channel. However, the differences in innovativeness between the shopper categories were small. This could perhaps be explained partly by the fact that Web users in general can still be described as innovative, being a minority among consumers. This may hold especially true for Internet users that participate in electronic surveys. Finally, the concept of Internet maturation and the idea that Web users become more instrumental with experience were considered. The hypothesis that Web shoppers are more experienced than non-shoppers was supported. As a final comment, the tests of the hypotheses have focused on Web shoppers compared to non-shoppers, a dichotomous division. However, in the analyses a threefold categorization has been made, dividing non-shoppers into active non-shoppers and passive non-shoppers. It was found that the active non-shoppers were somewhere in between shoppers and nonshoppers, as they showed more interest in shopping than passive non-shoppers and could be viewed as prospective shoppers. In the analyses it is also found that active non-shoppers are closer to shoppers than the passive non-shoppers are. The three shopper categories can be viewed as three parts of a spectrum.

5. CONCLUSIONS This study confirms the GVU (Pitkow and Kehoe, 1997) results indicating a rather stable share of Web consumers. Among those with intentions to buy something on the Internet in the foreseeable future, there was a clear bias towards present shoppers and those who had tried or simulated shopping. A conclusion from this could be that the increase in Web shopping, in the near future at least, will come from increased expenditures by these two groups. With the subsidization of Internet connections from telecom companies and other actors, it will probably not be long before the Net is as common as the TV. The Internet user may soon be anybody. However, the Web shopper is not and will probably not be so for a while. Two groups of consumers have been found to be especially prone to shop on the Web. These are the economic and the apathetic consumers. Hence, the price and convenience as-

.

9. Closing in on the Web Consumer - A Study in Internet Shopping

135

pects of the Internet seem to be important. Price could be reduced as redundant intermediaries between buyer and seller can be eliminated. Also, the search costs are lowered as consumers can scan a more vast array of alternatives easily. However, the Internet is still difficult to find its way around, and in order to facilitate consumer search so that the price aspect of the Net can be exploited, measures need to be taken (Dahlin, 1997a). There are several alternatives, such as better search engines and intelligent agents who help consumers in their search. In the end, though, it all depends on the companies, who must facilitate the search by giving out the necessary information. Of course, there is always a resistance between companies to engage in price competition as previous experience has shown (cf. Bailey, 1996). The convenience aspect of the Web puts demands on the organization behind the virtual store front. In order to compete with convenience outlets, Web stores must be available when the consumer feels like making a visit, and deliveries must be tailored to the customer’s needs. This calls for an efficient organization that can meet these demands without too much increase in overall operating costs. Along the same line as the discussion of consumer price search above, companies must be visible to consumers in order to make them aware of the shopping abilities that exist. This study showed that consumer knowledge of Web shopping opportunities was fairly low, and that knowledge of shopping sites had a positive relation to shopping behaviour. Just as companies reason that there must be a critical mass of shoppers before they invest in the shopping channel, consumers need to know of a fairly large supply before they adopt the channel. As the Internet users do not find or come to the shopping sites, the shopping sites must “come to them”. This calls for marketing communication efforts, in order to inform about the opportunities that exist. The Web shopping opportunities ought to be communicated on the Net to help consumers find their way. Companies with traditional marketing channels as well should use these to make customers aware of “virtual alternatives”. In the study it was found that Web shoppers were biased towards traits that describe the stereotypical Internet user. A possible explanation for this is the finding that Web shoppers are more experienced Internet users, as they have used the Net for a longer time. The long-time users are also the ones that are pictured in the early demographic studies. Hence there is an intersection. The question then is: which of the factors is most important to Web shopping behaviour, socioeconomic status or usage experience? Time will tell. From the findings in this study, one can conclude that there is a big market for Web shopping among the consumers in the distance shopping industry. Web shoppers were found to be more favourably disposed towards distance shopping, and were also more frequent mail order customers. The Internet faces the same problems as traditional distance shopping for many products, i.e. physical delivery, but has advantages in the user interface. Information is essential. Web shoppers were found to have better confidence in their ability to judge products beforehand. This factor is important. In order to appeal to consumers having less confidence, companies must help them with information in their pre-purchase evaluation of products. The interactive part of the Net facilitates an instant dialogue between buyer and seller, which has yet not been used to its full potential. Also, the unlimited space on the Web facilitates storage of vast amounts of information, databases etc. which can be turned to the customer’s use. Product quality and features must be easy to assess, either through availability of thorough information and help in information processing, or through product standardization (of which well-known brands are a form).

M. Dahle‘n

1-36

REFERENCES Alba, J., J. Lynch, B. Weitz, C. Janiszewski, R. Lutz, A. Sawyer and S. Wood (1997), “Interactive Home Shopping: Consumer, Retailer, and Manufacturer Incentives to Participate in Electronic Marketplaces”, Journal of Marketing, Vol. 61, No. 3, pp. 38-53. Bailey, P. ( 1996), “The Emergence of Electronic Intermediaries”, Proceedings of the Seventeenth International Conference on Information Systems, pp. 39 1-399. Bellenger, D. N. and P. K. Korgaonkar (1980), “Profiling the Recreational Shopper”, Journal of Retailing, Vol. 56, No. 3, pp. 77-92. Businessweek (1997), “Special Report. BWIHarris Poll: A Lot of Looking, Not Much Buying - Yet”, Business Week, October 6; . Cohen, J. (1982), Building a Mail Order Business, McGraw-Hill. Cox, D. F. and S . Rich (1964), “Perceived Risk and Consumer Decision Making”, Journal of Marketing Research, Vol. l , November. DahlCn, M. (1997), “Navigating the Never-Ending Unknown. - Cruising the Web”, COTIM-97 Conference, Brussels, 23-26 November 1997. DahlCn, M. (1998), “Controlling the Uncontrollable. Towards the Perfect Web Sample”, ESOMAR Worldwide Internet Seminar and Exhibition, Paris, 28-30 January 1998. Eighmey, B. and L. McCord (1995), “Adding Value in the Information Age: Uses and Gratifications of the World Wide Web”, COTIM -95 Proceedings, pp. 174-182. Engel, J. F., R. D. Blackwell and P. W. Miniard (1993), Consumer Behavior, The Dryden Press. Falk, T. (1983), Kopa via datorer [Buying via computers], Foundation for Distribution Research. GVU (1995), GVU’s 3‘* User Survey; . Hansen, F. (1972), Consumer Choice Behavior, The Free Press. Hoffman. D. L. and T. P. Novak (1997), “Measuring the Flow Experience Among Web Users”, Preliminary draft; . Isaacson, H. L. (1966), “Store Choice. A Case Study of Consumer Decision Making”, University Studies in Retail Research, Vol. 8. I

Kotler, P. (1 99 I), Marketing Management. Analysis, Planning, Implementation, and Control, Prentice-Hall International. Laaksonen, M. (1 993), “Retail Patronage Dynamics. Learning about Daily Shopping Behavior in Contexts of Changing Retail Structures”, Journal of Business Research, Vol. 28, No. 112. Malhotra, N. K. (1983), “A Threshold Model of Store Choice”, Journal ofRetailing, Vol. 59, No. 2. Newbold, P. (1 99 l), Statisticsfor Business and Economics, Prentice-Hall International. Paltschik, M. (1985), Adoption av innovationer. En studie i syfte att utveckla och empiriskt prova en referensram for individens adoption av innovationer [Adoption of innovations. An empirical study], Publications of the Swedish School of Economics and Business Administration, No. 35. Pitkow, J. and C. Kehoe (1997), GVU’s 7’h User Survey; . Robertson, T. S . (1971), Innovative Behavior and Communication, Holt, Rinehart and Winston Inc.

9. Closing in on the Web Consumer - A Study in Intrrnet Shopping

137

Schiffman, L. G., J. F. Dash and W. R. Dillon (1977), “The Contribution of Store-image Characteristics to Store-type Choice”, Journal of Retailing, Vol. 53, No. 2. Simon, J. L. (1984), Getting into the Mail-order Business, McGraw-Hill. Skaug, J. E. (1994), Kundattityder till postorder. En studie om postorderkundernas attityder till postorderhandeln [Customer attitudes towards mail order], Business School at the University of Goteborg, Business Administration, Department of Marketing. Solomon, M. R. (1992), Consumer Behavior, Allyn and Bacon. Spence, H. E., J. F. Engel and R. D. Blackwell (1970), “Perceived Risk in Mail-order and Retail Store Buying”, Journal of Marketing Research, Vol. 7, August. Spiggle, S. and M. A. Sewall (1987), “A Choice Sets Model of Retail Selection”, Journal of Marketing, Vol. 5 1, No. 2. Stone, G. P. (1954), “City Shoppers and Urban Identification: Observations on the Social Psychology of City Life”, American Journal of Sociology, Vol. 60, No. 1.

PART IV

CONVERGENCE AND BUSINESS NETWORKING

0 2000 Elsevier Science B.V. All rights reserved Convergence in Communications and Beyond E. Bohlin, K. Brodin, A. Lundgren and B. Thorngren (Editors)

CHAPTER 10

Governance and Technological Change: Transaction Costs in Telco-equipment Supplier Networks Denis Phan’ Ecole Nationale Supe‘rieure des Te‘le‘communicationsde Bretagne, Departement Economie, Brest Thierry Sommer France Telecom, Branche Grand Pub W D i v i s i o n MultimedidDSP, Paris

Abstract. This chapter aims to explain governance forms in the history of the public switching equipment industry by embedding transaction cost theory (TCT) in a technology variable. On the basis of the smallest building block in TCT, “technologically separable interfaces”, three periods in the switching history are delineated: electromechanical, digital and intelligent architectures. For each period, there is a movement towards more standardized interfaces and components, making the technologically separable interface increasingly small. The period of the future - Internet Protocol - will further accentuate this condition. A case study of AT&T’s governance structure suggests a trend towards disintegration, related to the trend of increasing technologically separable interfaces in switching technology.

1. INTRODUCTION To what degree is Transaction Cost Theory (TCT) able to highlight organizational choice analysis in a technologically intensive sector like the public switching industry? A previous study by Guertman and Quelin (1993) used TCT in order to analyze the coordination practices in this sector in an international comparative perspective. However, it did not mention the latest developments in technology and organization, like intelligent network architectures (INA), increasing competition in the telecommunications area (Phan, 1996). This chapter uses Transaction Cost Theory to explain the organizational forms observed in the telecommunications industry’s production of public switching equipment in the recent past. Its second goal is to emphasize some limits of this theory in this specific case and to help extend its scope. ’ The authors acknowledge F. Lotter and S. Saussier (ATOMIParis I), E. Dognin, J. P. Simon, M. Trtheux (France Telecom) for their helpful comments on an early French draft version of this paper, and E. Bohlin, G. Dang Nguyen (ENT-BretagnelICI UBO) and J. Ormrod for their contributions to the English version for the ITS 1998 biennial conference. All remaining errors are our own.

142

D.Phan and T. Sommer

Thefirst section gives a brief discussion of the principles of TCT and particular attention is paid first to the role of technology. This is not well defined in Williamson’s work, where a “semi-weak” form of technological determinism is put forward. The way “technologically separable entities”, bounded by “technologically separable interfaces”, are defined is, according to Williamson, beyond TCT’s domain of analysis, because there is no way to identify a transactional problem in their definition. Therefore technologically closed entities would be the new “black boxes” of transactional analysis. But this first limitation makes “standard” TCT unable to take into account the process of technological evolution itself, when technological change is not an “exogenous” factor. A second theoretical limitation of TCT is the relationship of governance structures and the institutional environment. Williamson (1993a) recognizes both the influence of the institutional environment on governance structures and a feedback effect of governance structures upon the institutional environment. But he claims that these latter effects are “of the second order” and beyond the scope of TCT theory. In this paper we discuss the logical and empirical consequences of this principle, in particular when the actors’ strategies are designed to influence the institutional arrangements according to their self-interest. The consequences of both theoretical limitations are that it is difficult - logically, and empirically in some cases - for the actors to measure ex ante the transaction cost for different kinds of governance structure, and TCT cannot by itself explain the transition between different regulatory modes. The second section analyzes the characteristics of transactions between telecommunications operators and equipment manufacturers in the domain of public switching. On the basis of “technologically separable interfaces”, TCT analysis enables us to distinguish between three periods. In the first one, the switching system as a whole is a unique technological entity (designed using electromechanical principles), which gives rise to a single transaction between telcos and equipment manufacturers, with a high degree of asset specificity. In the second period, the switching system is split into two different technological entities: hardware and software, and two different transaction characteristics. Exogenous technologically separable interfaces appeared early with space-division electronic switches, as a consequence of the diffusion of computer characteristics into the telecommunications industry. But a separate transaction for software and hardware, respectively, appeared only with the fully digital switch (time-division). Transactions concerning software become the support and the main challenge of asset specificity, first reinforcing it, but then globally reducing it in the long run, for both institutional and technical considerations. With this separation, new transactions occurred through an upgrade of the software for technical reasons or through marketing necessity. The introduction of each new software release provided a competitive advantage for fast movers. Thus, growing competition in the telecommunications services gave strong endogenous incentives to the separation of customer services from basic network administration. According to these incentives, the third period, with the emergence of the Intelligent Network Architecture at Bellcore, led to more modular systems with new opportunities for standardization (decrease in asset specificity) and for new technological separation in a third period. The definition of technologically separable entities is consequently mainly endogenous, unlike the previous stage. However, interactions with the computer industry remained important, in the context of the gradual convergence between the computer world and the telecommunications world, through growing networking activity in data communications (ATM, Internet). The third section raises the question of the appropriateness of governance structures to their technological and institutional environment in the long run, taking as an example the

IO. Governance and Technological Change

143

United States, from the creation of the Bell Company to the present day. Despite theoretical limitations, TCT enables us to compare ex post the relative efficiency of governance structures in a given institutional and technological context. TCT also allows us to highlight when governance structures became inadequate, faced with technological and institutional evolution, as is shown in this paper in the case of telco-equipment suppliers.

2. A FRAMEWORK FOR THE “CLASSICAL” APPROACH OF TRANSACTION COST THEORY (TCT) This section briefly presents useful TCT tools for assessing the power and the limits of TCT as a suitable theoretical framework for analyzing the relationships between public switching equipment manufacturers and Telecom operators. TCT was developed early by R. H. Coase in his famous article “The nature of the firm”, published in 1937. This paper underlines that the market is not the only means of coordination for economic transactions; the firm can also be considered as an alternative form of coordination. Coase explains the complementarity between the firm and the market through the existence of transaction costs. Transaction costs are the resulting burden of the resources mobilized in order to achieve a transaction through the market. The firm exists because the cost of achieving specific transactions is lower within an organization than on the market. In the seventies, O.E. Williamson proposed a new framework for analyzing transaction costs. He acknowledged his indebtedness to Coase’s seminal work. Williamson attributed to Coase the merit of having presented the problem of economic organization though a comparison between organizational forms.’ Hence, the goal of TCT is to assess the different forms of governance structures in relationship to the corresponding transaction cost. Since the end of the eighties, the TCT research program has been enhanced, by taking into account more explicitly the influence of the institutional environment on organizational choices. Moreover, debates on the role of technology and organizational learning have influenced the recent development of TCT.

2 . 1 The Main Concepts of Transaction Cost Theory (TCT) The Sources of Transaction Costs

The transaction is the analytical unit of TCT. A transaction can take place in several organizational frameworks, such as the market or the firm. Every form of coordination is costly. What is the source of these costs? To answer this question, it is convenient to underline two fundamental assumptions concerning the behavior of economic agents and three characteristics. These elements are used to described a transaction. The first assumption concerns the cognitive nature of the rationality of economic agents. Agents are assumed to search for the less costly form of coordination for organizing their transactions. But their cognitive capabilities are assumed to be “limited” (Williamson, 1985, pp. 45-46).’ Under this assumption, it is impossible for agents to write “complete” contracts.

’

Apart from Coase, Williamson recognizes his intellectual debt to K. Arrow, A. Chandler and H. Simon. “The Institution of Capitalism” is dedicated to these four authors. In particular, Herbert Simon supervised Williamson’s Ph.D.

* This approach is derived from Herbert Simon. Williamson (1985) uses the term “bounded rationality” and not “procedural rationality”, also introduced by Simon (1976, 1978). On this question, Williamson refers frequently

144

D. Phan and T. Sommer

A “complete” contract is assumed to protect agents against all possible events that might occur in a trade. The incompleteness of contracts might lead to unpleasant effects in relation to opportunism, the second behavioral assumption. Economic agents, driven by their own selfinterest, are tempted to manipulate information deliberately, in order to gain advantage for themselves. Three attributes can characterize the nature of a transaction: the level of frequency, the level of uncertainty and the level of asset specificity.3 That is, the combination of these three attributes would serve to justify the choice between possible forms of coordination, generally in order to protect against the risk of opportunism. Taking into account the behavioral assumptions and the characteristics of these attributes, the exchange process is therefore intrinsically costly. The corresponding costs are specifically the transaction C O S ~ S The .~ aim of TCT is to make an inventory of and assess the institutional arrangements in order to reduce these transaction costs.

The Structures of “Governance” “Governance structure” denotes the organizational form in which a transaction takes place. The two extreme forms of the set of governance structures are the market and the “hierarchy” (coordination inside a firm). Between these “pure” forms exists a large set of “hybrid forms”. Some of them can be qualified as “quasi-market”, some as “quasi-integration”. One can also consider another structure, like public authorities organizing transactions and other forms of coordination beetween several firms. Some specific costs are attached to each governance structure. It is possible to assess these costs using two main indicators: the degree of control and the degree of motivation. The degree of control refers to the capacity of the regulation structure to contain opportunistic behavior, and to enable the compatibility of actions between the agents concerned by the transaction. A weak degree of control by the market may lead to an inefficient allocation of resources (market failure). The degree of motivation refers to the agents’ incentive to search for efficiency in production. There is a hierarchy failure when the hierarchy is unable to produce enough incentives in comparison with the competitive pressure.s

to Simon (1957): “behaviour is intendedly rational, but only limited so”. Like the latter, he justifies bounded rationality by the limitation in the cognitive competence of the agents: “mind is scarce resources” (Simon, 1978, p. 15, quoted by Williamson, 1985, p. 46) but does not have a great deal of interest in the cognitive dimension of rationality itself. Only the economic consequences of bounded rationality matter, such as the incompleteness of contracts. In TCT, agents remain calculating and lead mainly through their self-interest. From this point of view, one can interpret this form of bounded rationality as a weak form of instrumental rationality (Walliser, 1989). Asset specificity may take different forms. Williamson (1993b) mentions, for instance: site specificity, physical and human asset specificity, dedicated assets. Cf. Williamson (1985, p. 55). The ex ante transaction costs include, for instance, the costs of searching for information, negotiation costs, and the costs related to the safety of the relationship. The ex post transaction costs include control costs, adaptation costs, correction costs and conflict costs. Lack of incentives is, in TCT, the main limitation in the size of firms: “the transfer of a transaction out of the market into the firm is regularly attended by an impairment of incentives” (Williamson, 1985, p.161). The bureaucratic phenomenon (Leibenstein, 1987) is taken into account at a lower level. It is not possible for the management of large companies to limit themselves only to the problem of coordination between departments, given the numerous sources of incentive disabilities, enforced by auditing limits.

’

10. Governance and Technological Change

145

The Transactional Choices

The institutional comparison is the fundamental principle which enables the choice between different governance structures to ensure a given transaction. “The Transaction Costs are economized by assigning transactions (which differ in their attributes) to governance structures (the adaptive capacities and associated costs of which differ) in a discriminating way” (Williamson, 1985, p. 18). From the TCT point of view, the choice of an organizational form results from the intentional search for an appropriate structure to minimize the set of governance costs and the related production costs, for a given set of technological choices. In its positive version, the aim of TCT is to understand how a large variety of characteristics associated with transactions within the capitalist economy lead to a plurality of institutional arrangements: the governance structures. In its normative version, TCT advocates a regulation structure with a strong degree of control when the transaction attributes are selected by the agents, or are characterized by a strong degree of asset specificity. This is a major prediction of the theory.‘

2 . 2 The Place of Technology: A “Semi-weak” Determinism TCT is involved in the debate on the role of technology for the determination of the firm’s boundary. According to Williamson, the degree of influence of a technology over the choice of organizational structures corresponds, in general, to a so-called “semi-weak” form of technological determinism:’ “technology is neither fully determinative of nor irrelevant to economic organization” (Willliamson, 1988, p. 357). Known technological options would delimit the set of feasible governance structures. The notions of technologically separable entities and technologically separable interfaces are fundamental for understanding the nature of organizational choice allowed through TCT. It makes it possible to go beyond the notion of “standard” theory, in which the firm looks like a “black box”, a pure “production function” (Willliamson, 1988, pp. 356, 358). Technologically Separable Entities and Technologically Separable Interfaces

The use of these notions implies two consequences. First, under the “semi-weak” form of technological determinism assumption, all processes of production can be stylized by one or several inseparable entities, given the technological knowledge. Between these entities, “technologically separable interfaces” appear. At these points, transactions are possible: “A transaction may thus be said to occur when goods or a service are transferred across a technologically separable interface. One stage of processing or assembly activity terminates and another begins” (Williamson, 198 1, p. 1544). The number of technologically separable interfaces can be considered as fixed when it is no longer possible to “break down” a particular production process to identify new interfaces. The choice for governance structure occurs as many times as there are technologically separable interfaces.

‘This result is illustrated by numerous theoretical studies. Cf. Klein and Shelansky (1995). ’ Willliamson (1989) proposes to distinguish four degrees in technological determinism. There is a “strong” technological determinism when the technology is assumed to be the unique determinant of economic organization. There is a “semi-strong” form when the technology is not the main determinant, but matters. There is a “semi-weak” form when the technology defines a set of feasible organizations within TCT which alone allows the choice. Finally, there is a “weak’ form when the technology plays an insignificant role, having as a consequence “technology does not matter”, and organizational forms are completely determined by transactional factors.

146

D.Phan and T. Sommer

Second, there can exist a set of already known technological options in order to organize the production behind every interface. These different options are to be found in the form of different attributes at the transaction level (notably in the scope of asset specificity). Within this set of possible situations, agents have to choose an organizational form. They select the transaction attributes, the governance structures, and the technological configuration of production in the best way according to their interest. In that way, they can group some subsets of entities within firms or within other “hybrid” forms of governance (alliances, joint ventures, and other forms of inter-company agreements). The transactional arbitration of TCT is therefore equivalent to stating the problem of the organization of productive activities in these terms: how can the segment of activity be spread out among economic organizations in order to economize transaction costs, for a given technological background? Are Technologically Closed Entities the “Quarks” of TCT ?

One problem remains unsolved. The area within an entity considered as “inseparable” is outside TCT’s domain of analysis, because it is no longer possible to identify a transactional problem. Therefore technologically closed entities would be the new “black boxes” of transactional analysis. Williamson (1987) recognizes the limits of TCT for taking account of the creation and the evolution of firms.8 In fact, TCT does not manage to explain how the set of technological options appears and what factors contribute to their emergence. However, taking into account the possible evolution within the set of technologically closed entities plays a non-negligible role in the definition of the entities’ boundaries. Some authors argue that the question of endogenous technical change may be solved by the introduction of another dimension, like competence, or capabilities.’ But for TCT “fundamentalists”, it is by nature a radically different approach, because transactions are no longer the only unit concerned in the analysis (Lotter, 1997).1° All this advocates considering the “standard” TCT framework essentially as a “static”, normative approach. Learning and technologically strategic choices are not integrated within this “standard” analysis. In recent papers, Beije (1996) and Noteboom (1996) underline the difficulties raised by the introduction of dynamics together with a TCT analytical framework. Beije (1 996) introduces key concepts for understanding the technological dimension. That is, technological learning implies taking into account competence - as mentioned above - but also strategies and path-dependency phenomena. In this way, the relative advantage of a governance structure no longer depends only on its capacity to minimize transaction costs, but also on its ability to make technological learning easier, and to capture private rent from innovation as well.” Finally, from the point of view of technology analysis, the analytical framework of “standard” TCT alone is therefore essentially “static”; in this sense it is not possible

’

This restriction is also underlined by Chandler (1992): “I find the earlier growth of the industrial firm difficult to explain fully in terms of transactions, agency and other information costs, so I find it hard to explain the recent process of expansion and contraction with these same concepts” (Chandler, 1992, pp. 98-99). The question of capabilities refers to the debate on the complementarity between TCT and evolutionary theory (cf. note 17 for more references.). IO This interpretation must be brought closer to Williamson’s position on the institutional environment and close to North’s research program. ‘ I Noteboom (1996) further develops the issue of bounded rationality in the cognitive sense. Cognitive capacities depend on subjects, but also on environments. As a result, preference formation and innovative capacities must be endogenous and the way in which agents seek to learn matters.

‘)

10. Governance and Technological Change

147

to take into account the process of technological evolution itself, and the technology remains an “exogenous” factor. This kind of argument is still valid when we move on to the institutional environment.

E l

+

Competence

11

Technologically Black Box?

11

. .. ~

9

.

b

Learning Process

Transaction

E l Competence

1

Technologically

1

Black Box?

Figure 1. .4Two-dimensional “Non-classical” Analysis to Integrate Learning and Transactions.

2 . 3 Taking into Account the Institutional Environment The institutional environment is defined by Davis and North (1971) as “a set of fundamental political, social and legal ground rules that govern economic and political activity (rules governing elections, property rights and the rights of contracts are examples of the ground rules)”.’*More specifically, Schotter (1981, p. 11) defines a social institution as “a regularity in social behavior that is agreed to by all members of society, specifies behavior in specific recurrent situations, and [is] either self-policed or policed by some external authorities”. For North (199 l ) , institutional elements consist of “constraints” which structure political, economic and social interactions. Such interactions may be either formal rules (constitutions, law, property right) or informal constraints (taboos, custom, traditions and codes of conduct). The relative features of the various forms of coordination for transactions (i.e. costs of governance) are dependent on the characteristics of the institutional macro-environment (Williamson, 1993a, 1996). Thus, modifications in the institutional environment may aiter the relative performances of governance structures which organize these transactions. The Three Levels (at least) of the Global System: Institutional Environment, Governance Structures and Individual Agents

The influence of the institutional environment on governance structures may be illustrated by Williamson’s diagram (Figure 2: Strata Diagram - Williamson, 1993a). The relationship between the individual (agents) stratum and that of the governance structure denotes the influence of individual behavioral characteristics (limited rationality, opportunism) over the choice of governance structures. Williamson recognizes the significance of specific evolutionary

’* Op. cit. p. 133. Davis and North define “an institutional arrangement as an arrangement between economic units that govern the ways in which these units can cooperate or compete”. The governance structure is typically such an “ institutional arrangement ”.

D. Phan and T. Sommer

148

dynamics for the institutional environment, on the one hand, and for governance structures, on the other hand.” But for Williamson, these dynamics must relate to two different levels of analysis. On the one hand, the institutional environment concerns the global macro-relationships between agents and activities. On the other hand, governance structures concern the micro-economic level of transactions. As a consequence, governance structures and the institutional environment would emerge and evolve differently with respect to intentionality (Williamson, 1996, p. 5). Moreover, he considers the feedback effect of governance structures upon the institutional environment as of the second order (Williamson, 1993a). Given all these arguments, Williamson claims that these two levels must be studied separately. The two higher strata may influence individual choices as well, for instance through certain forms of “social influence”, “social customs” or “conformist” behavior (Jones, 1984; Akerlof, 1990; Bernheim, 1994). This effect of “endogenous preferences” (Bowles, 1998) denotes the influence of the institutional environment and governance structures upon an individual stratum. These effects, recognized by Williamson, of lower importance, are indicated as dotted lines in Figure 2 (endogenous preferences and strategies). The latter are frequently neglected by TCT, which focuses on the two main relationships (institutional and individual effects upon governance structures, indicated as a solid line). However, TCT refers frequently to secondary effects, but Williamson suggests that others’ analytical frameworks are more appropriate for studying these relationships (Williamson, 1993a, p. 115).

e

Institutional Environment Properties

I

Strategies

I

I

1 Governance Structures

Behavioral Characteristics

I + Endogenous Preferences

I

I

I I I I I

I I

+I

V1

Agents ”Second-Order Effects”

----

I

Figure 2. Strata Diagram. Source: Williamson [1993a]

’’

“Both the institutional environment and the institutions of governance have evolutionary origins, [but] the ramifications of each are different”: Williamson (1996, p. 5, underlined by us).

10. Governance and Technological Change

149

Williamson’s argumentation on “second-order effects” is not fully persuasive, especially when the time perspective is long enough for significant backward effects not to be avoided. For instance, the accumulation of small events may create an effect of tension by accumulation which may suddenly be relaxed by a dramatic change beyond a certain threshold (Lordon, 1996). In the case of the governance structure, the history of business highlights that the feedback of organizations upon the institutional environment is sometimes negligible but may become more significant in other cases, with possible institutional break-up. For instance, Fliegstein (1990) shows how large American firms’ structures can be viewed as a consequence of American institutional competitive rules (such as antitrust laws), on the one hand, but the behavior of these firms in order to find new way to bypass this institutional framework must be viewed as the main cause of the evolution of this institutional framework, on the other hand. The AT&T case, as studied by Mueller (1993), is a good example in the telecommunications field of such behavior. A governance structure which may generate such a tension would be stable in the short term (locally), and strongly unstable in the long term.I4 Elsewhere, one governance structure may be inefficient in a given institutional environment, and efficient in another. Hence, institutional reform may sometimes be suggested in order to intentionally modify - in a “strategic” way - the institutional environment in order to make a given governance structure more efficient.Is In a recent paper, Magnusson and Ottosson raise the question of the underdevelopment of the feedback effect of governance structures on the environment, and the difficulty of “standard” TCT in explaining the real (historical) process of institutional change. On the one hand, the action of actors on the institutions matters. “Clearly the actors will try to influence institutional arrangements according to their self-interest” (Magnusson and Ottosson, 1996, p. 361). On the other hand, because the actors are also bounded by the existing institutions, pathdependency phenomena may explain why evolution can also lead to a lock-in in inefficient institutions as well as governance structures. Because a real choice depends not only on transaction costs, but also on technological, market and institutional conditions, these authors argue that actors cannot measure ex ante the transaction cost for different kinds of governance structure, and TCT cannot by itself explain the transition between different regulatory modes. As a result, these authors claim that TCT “has been most valuable in developing economic history, but its use as an explanatory factor in itself is highly questionable”(id. p. 361). That is, a complete assessment of the scope and limits of the TCT analytical framework for relationships between the equipment manufacturers and operators in the public switching system market must take into account the complete set of factors (institutional, technological, organizational) which affect a transaction. Despite the complexity of the question, we shall try to propose a synthetic framework in order to summarize the (omitted) effects of technological and institutional factors within Williamson’s original diagram.

l 4 In another way, similar results can arise when the accumulation of small events drives a dynamic system into a “critical state” where there is no proportionality between incremental stimulus and the resulting effect, as in the case of avalanches. IS Transformations in the institutional environment may be committed intentionally for strategic reasons or instrumental ones. The position of R.H. Coase on the interest of selling Hertzian frequency by auction is a good example of such an approach (Coase, 1994). Coase proposes a set of arguments to demonstrate the superiority of the market over the administrative decisions of the FCC to organize this transaction. However, to make the sale by auction possible, it is previously necessary to create property rights over the Hertzian frequency spectrum. Thus, the intrinsic qualities of the market (as a particular governance structure) justify an institutional reform to make possible the use of the auction-market as an efficient solution for a transactional problem.

D. Phan and T. Sommer

150

Technological and Institutional Factors: A Synthesis

In this way, Figure 3 amends Figure 2 by proposing complementary relationships. (2’) represents the impact of individual choice on the set of technological options (innovations), on the one hand, and (2) represents the contribution of technology to the modification of the arbitration space of TCT, on the other hand. Both questions are usually seen as exogenous by the theory. To summarize, TCT recognizes that evolution in technology and evolution in the institutional environment are important features for the problem of organizational choices. However, they are generally treated as exogenous by the “standard” approach of governance structures. These factors define a given transactional space where agents would arbitrate between the different institutions, the governance structures. “Standard” TCT will be applied in the second part of this paper to public switching transactions for a given technological and institutional environment. With this constraint, TCT will be used as a static allocative theory of organization, and a history of organizational forms will be reduced to an assessment in a “static comparative” way.“

I I t3 Institutional Environment

1

.....................

...............\

i Technologically i Closed i Segment(s) i....................................... (BlackBox)

: :

:

111

4

Interface

1 L

1’

.................. ................., Technologically i Closed i Segment(s) i....................................... (BlackBox) i

1 Individual Choices

1

Figure 3. Synthesis

16

Dynamic evolution is often referred to in recent papers, but little is proposed in practice. However, one may conceive extensions for the TCT standard research program. Lotter (1997), for instance, suggests a more “dynamic” approach for the evolution of transaction characteristics. Such an extension is suggested by Williamson himself, in a recent work. The problem is that it is not certain that this extension will be sufficient to make endogenous the technological choices and the dynamics of innovation. Another way is to suggest a complementary approach for TCT, like the evolutionary theory of a firm based on capabilities (Winter, 1991; Langlois, 1992; Foss, 1993, 1996; Langlois Robertson, 1995; Brousseau, 1996).

10. Governance and Technological Change

151

3. THE TRANSACTIONS BETWEEN EQUIPMENT MANUFACTURERS AND OPERATORS IN THE PUBLIC SWITCHING SYSTEMS INDUSTRY: CHARACTERISTICS AND ATTRIBUTES On the basis of “technologically separable interfaces”, TCT analysis enables us to distinguish between three periods. In the first one, the switching electromechanical system is an unique technological entity, which gives rise to a single transaction between telcos and equipment manufacturers. In the second period, the switching system is split into two different technological entities: hardware and software, and two different transaction characteristics. Exogenous technologically separable interfaces appeared early with space-division electronic switches, as a consequence of the diffusion of computer industry characteristics into the telecommunication industry. But a separate transaction for software and hardware, respectively, appeared only with the fully digital switch (time-division). A third period began with the emergence of the Intelligent Network Architecture (INA), which led to more modular systems, opportunities for new technological separation and more standardized interfaces. By opposition to the previous stage, the definition of new technologically separable entities became mainly endogenous. In this section, we study the attributes of transaction for the two earlier periods, and provide hints for the transition towards the third period, to be completed in the last section.

3.1 From the First Switch to Electromechanical Generation In the 1910s, the first automatic switch, invented by Alvin Strowger, simply imitated the actions of previous manual operators. The basic principles of this system were not to change for decades. Two main components had to be distinguished within the machine, one designed for action and the other designed for control: A selector within the switching machine allowed a circuit to be temporarily created between an incoming line and an outgoing line.I7 In order to increase the capacity of this machine, these components were connected one to the other via a mesh and thus constituted a “connection network” with the function of switching the line. This “connection network” was driven by instructions from the command subsystem of the switching device. The latter accomplished the “intelligent” function: to establish the communication and to administrate the system. In more recent electromechanical switching systems, the command subsystem drove a relay which opened or closed a circuit. This role was to perform logical operations at a low level of complexity. These components are not naturally programmable. In this context, it is practically impossible to easily upgrade the command subsystem, which integrates the general specifications of the network.

Network Specificity Unlike a milling machine or a lathe, of which certain models may be used indiscriminately in numerous contexts, a public switching device may not be deployed outside the public telecommunications network without lack of efficiency. The latter must be considered as a specific system. For a telecommunications operator, the control of the whole switching system generates an important market power over competitors wishing to be interconnected to this network. 17 The path established between two lines is a physical one (spatial): metal contacts make the connection. With the transistor, they were replaced by electronic gates.

152

D. Phan and T. Sommer

However, a secondhand market exists for switching devices. The obsolete equipment may be deployed again in the less developed countries under the following conditions: the host network must have the same characteristics as the original network, which implies technical lags. It is also necessary to mention the case of regional American companies which h a w common technical characteristics, given their common past within the Bell System (Western Electric equipment). Uncertainty Over the Telco-Equipment Supplier Relationship

During the era of electromechanical switching, operators were often in a local monopoly for telephony services. The uncertainty about transactions resulted principally from considerations about the continuity of supply (uncertainty about the long-term survival of the supplier) and about problems technical by nature. Table 1. Asset Specificity for Electro-Mechanical Switches

Nature of the Specificity

Switching Equipment

Site Specificity

No

Dedicated Assets

Yes, network a s “specific system”

Specific Human Resources

Specialized teams for switching system

3.2 Digital Technology: A New Technological Separation, Coming from Computer Industry Design The emergence of the so-called “digital paradigm”” in telecommunications has gradually transformed the production framework and especially the structure of this industry, in the field of public switching. In particular, the introduction of electronics has gradually modified the attributes and the number of transactions. Since the end of the sixties, switching systems have been extended to become a specialized type of computer.” The “intelligence” of telecommuSome authors, including Abernathy and Utterback (1978) and Dosi (1982, 1984) have extended to technological practices the use of the concept of “paradigm”, used initially in fundamental sociology research by Kuhn (1970). Dosi (1982) thus defines a technological paradigm as: ‘‘ a ‘pattern’ of solutions of selected technoeconomic problems based on highly selected principles derived from the natural sciences, jointly with specific rules aimed at acquiring new knowledge and safeguarding it, whenever possible, against rapid diffusion to the competitors ” (Dosi, 1988, p. 1127). According to this analytical proposal, an innovation is considered as “radical” when the potentialities that it conceals eventually lead to the fundamental transformation of the dominant design. Radical innovation constitutes the core of a new “knowledge base”, intended to increase with new solution rules, associated with new products and new methods of production. In telecommunications, digitization may be analyzed as a radical innovation (Brousseau, Petit and Phan (eds.), 1996; Dang Nguyen and Phan, 1998). In this paper, we do not develop the role of governance structures in the emergence of digital technology (timedivision switching) which have been previously analyzed by Quelin (1992). ”) Telecommunications engineers rapidly understood the advantages of replacing cabled automata guaranteeing the automatic switch control functions via programmed software activating a calculator. In 1949, the first development project of such a machine, Eco, was launched at Bell Labs. In 1957, at the first ISS (International Switching Symposium), Bell Labs announced the SPC (stored program control), a switch controlled by means of electronics. It was only in 1970 that CNET, in collaboration with CIT, developed the first prototype of a totally

10. Governance and Technological Change

1.53

nications networks has become more flexible and they have a higher capacity, as a result of the generalization of electronic computation and the implementation of software inside the switching equipment. Since the early generations of electronic switching,’’ the “software” component has assumed an increasing role within the core of telecommunications systems. This increasing importance of software led to a proportion of 50% of the total cost of the switch by 1988, and between 70% and 75% by 1994. From the TCT point of view, the possibility of separating hardware and software may be understood as the opening of a previously technologically closed entity. It was not previously possible to develop the control subsystem and the connection network in a switch separately. A new technological interface in the TCT sense will appear as a result of this innovation. Therefore, the potentiality of a new transaction level between network operators and suppliers of switching systems also appears. The importance of this transaction is enhanced by the possibility of upgrading the software part of the switch.” As a consequence, transactions over public switching systems may first be located at two levels: the hardware level and the software level (software platform), with attributes different from the TCT sense. This technological architecture was not “thinking strategically” in the telco-equipment area, but it is clearly transposed from the computer industry, as an exogenous feature. Without interface standardization, the asset specificity dedicated to these two transactions results mainly from the specific nature of the network itself. This specificity results from the technological path historically followed by manufacturers in the related sector. Market fragmentation and the existence of local monopolies do not contribute to a costly standardization process for the equipment produced by the manufacturer. Nevertheless, the importance of transaction on software will be increasingly enhanced by the possibility of upgrading the software part of the switch, which allows a different frequency of transaction between hardware and software, and an endogenous incentive toward more standardization and more modularity. The Path Dependence of Specificity

After the “digitization” of the control subsystem, then of the switching device (time-division), the switching device has remained specifically dedicated to a given public telecommunications network. These have great capacities, which distinguish this kind of equipment from private equipment (PABX). In the latter, administrative functions are less sophisticated and do not have, or have a very simple, invoicing function. Finally, they establish circuits (physical or virtual), that distinguish the public switching system from the material used in the data network (such as routers and packet-switches) which appeared with the development of telecomputing). With digitization, the technical characteristics of public networks have become more diversified. This results from past choices in switching systems: the division into national protected markets, the differentiated technological paths resulting from past strategies and the industrial

electronic switch, integrating a logical circuit connection function. This time-division switching system, called the E10, was then the largest computing system of its time (Libois, 1983; Chapuis and Joel, 1990). Space-division, stored program control (SPC): cf. Chapuis and Joel (1990, Chs.. VI & VIII) By analogy with micro-computing, this operation resembles the upgrade of software to a new version on a personal computer.

2o

21

I54

D. Phan and T. Sommer

policy of the “national champion” in some European countries and Japan. Such a division has contributed to the enforcement of the specific dimension of public network systems.22 Frequency of Transaction

The frequency of transactions relating to public switching equipment is determined by the growth of the network, by the duration of material, or more recently by its obsolescence. More specifically the latter enforces the need for building a long-term relationship between the operator and his suppliers. During the electromechanical period, switches had a very long duration, sometimes longer than forty years. This implies a low frequency for the renewal of the whole stock of switching equipment.23With the digitization of controls, the duration of the switching equipment has been considerably reduced. However, the possibility of upgrading the software components makes it possible to maintain duration long enough for the hardware. With the absence of standardization, the initial choice for a specific system and a particular supplier therefore tends to limit the scope of possible future opportunities - that is, in particular, the opportunity for an operator to choose another supplier. This type of constraint contributes to the creation of exclusive relationship conditions with a supplier of switching

Rotary Crossbar

E10 N3 AXE 11 F E10 N1

MT25 AXE E10 B3 ATM

Figure 4. The Generations of Subscriber Switches in France (Duration by System).

’’

Cf. Fransman (1992) for an example of an evolutionary approach to R&D at AT&T, BT and NTT. On public policy and industrial strategies, cf. Dang Nguyen (1983); Quelin (1992); Sally (1993). On national systems of R&D in telecommunication, cf. Grupp and Schnoring (1992). 23 24

In 1980, there was still a Strowger switch in BT’s network which had been working since the 1920s

TCT interprets this phenomenon as a “fundamental transformation” (Williamson, 1985, p.61). Certain operators tried to get around this problem by opening up the market to a supplier, but such a policy was contested for the complexity that it created for the administration of the network.

10. Governance and Technological Change

155

Uncertainty Related to the Functional Dimension of Hardware and Systems

With digitization, technical uncertainty increased (i.e. uncertainty about the effective performance of the equipment). Uncertainty increased about the survival of the supplier in the long term, given the large investments in R&D required and the necessity to maintain and to periodically upgrade software. Moreover, with liberalization, telecommunications operators were to be faced with increasing (real or expected) competition on the market. Digitization has not radically changed the nature of the transaction concerning the hardware part of the switching system, even if some attributes had to evolve. In particular, asset specificity related to the systemic nature of the network remains the crucial factor. The main innovation results from the emergence of a new transaction concerning the software component. The latter allows a greater frequency in the renewal of functionalities related to the switching system, by upgrading software. This innovation has led to more significant important changes in the nature and the evolution of asset specificity at the software level. Specificity

A range of switches have the characteristics of a specific physical asset from the viewpoint of transactions linked with software. Until very recent times, switching software was “tailormade” to satisfy the needs of an operator. Moreover, this software was developed by manufacturers with the help of proprietary languages, which reinforced their “capture” of the operator by the manufacturer. This logic can be found in the early development of software for intelligent network platforms : the slowness of the processes of standardizing programming interfaces proved the reluctance of equipment manufacturers to abandon proprietary systems for open, modular structures.*’ The technological characteristics particular to the network strengthen this effect even more. The specificity of a system can be conveyed in the writing of the software. In the absence of standardized interfaces, switching software cannot be deployed again towards a different system. Collaboration between the specialized services of the operator and the equipment manufacturers concerning the specifications of this proprietary software also constitutes a specific human asset which it is difficult to deploy without avoiding important learning costs. Frequency as a Competitive Advantage for Fast Movers

A new transaction may occur when purchasing new equipment equipped with more efficient software and new functions. It is then necessary to upgrade the software of the older equipment. Another type of upgrading occurs at regular intervals. This is partly linked to the commercial policy of the operator. It means introducing into the network the intelligence necessary to carry out new functions and to supply new services on the market.26The introduction of a new 2’

This problem is classical in the computer industry, where the eighties were marked by a massive decrease in closed proprietary systems to the benefit of modular systems. But the latter remained largely proprietary within each module. With TCT’s “Dynamics” proposed by Langlois, Robertson ( 1995), taking these “modular systems” into account allows other factors to be integrated into the analysis, such as the demand for modularity and “upgradability” which this suggests, although the analysis framework of “standard” TCT only covers one of the aspects of the problem.

‘‘In France, the first upgrade (or “Digital Version” = DV) dates from 1987, when the introduction of the Nume‘vis service (ISDN) made it necessary to update the switching network. Three other operations of this type were then planned in order to improve the services entrusted to ISDN (DV2, DV3, DV4). In the innovation “technology push’ model which characterized telecommunications operators in the pre-competitive period, the

156

D. Phun and T. Sommer

release for a whole network is a long and complex operation - software in a modern switch represents over a million instructions - and above all, the operator is tied to the supplier(s) already present in the network, who alone is(are) able to upgrade their software. Today, the time necessary to implement a new version of the software throughout the set of switches in a classic telephone network is still very long, although it is tending to decrease. Table 2. Comparison of Switching Software Development Times (1993) Operator

Design and Specification

Industrial Development

Testing

Introduction in the Network

Total

Version Frequency

FT (VN3)

18 months

30 months

12 months

12/18 months

72 months

12 months

DT (S12)

8 months

14/16 months

1 months

3 months

26 months

6 months

AT&T

12 months

11 months

1 months

6 months

30 months

6 months

MCI

6 months

18 months

3 months

3 months

30 months

1 2 months

6 months

18 months

3 months

12 months

39 months

12 months

Bell Atlantic

(RBOC) Sources: ENST-Bretagne; L’Echo des Recherches: Kubiak and Baptiste (1995) Depending on the operators and the manufacturers, the process is split into different phases of variable length which reflect various factors difficult to isolate: efficiency of the process, capacity of the equipment to evolve, the number of functions and of services per version, etc. In the above table, these times are given for AT&T, who manage a long-distance network, and for Bell Atlantic, a local, general operator. Uncertaintyfrom the New Institutional Framework

Uncertainty is profoundly linked with the liberalization of the operator’s market. A competitive risk is associated with times which are too long to develop switching software, in particular at services level. It is the “time to market” which allows the more reactive operators to benefit from a temporary rent which can be transformed into a permanent advantage when “club effects” are decisive in attracting customers (for instance for new mobile networks). With the growing liberalization of the market, upgrading software has become a test with which to measure the capacity of operators to. withstand the competition of new service providers, who might act either from other public infrastructures or from private equipment.

implementation of a “DV” was limited to collaboration between the specialized services of the CNET (France Telecom’s research center) and the manufacturer. The commercial services were gradually involved in defining the new versions of switching software. The requirements and specifications of the DVs were established by the operator, but the development was ensured by the manufacturer. In this distribution can be found the providercustomer roles close to specialized sub-contracting, Given the precursor role of Alcatel in Digital (time-division) switching, the complete DV cycle had lasted 6 years in France by 1992. This long delay has decreased dramatically with new generations (Kubiak and Baptiste, 1995).

10. Governance and Technological Change

157

This risk casts doubt on the very existence of this kind of transaction in the future. This question brings us back to that of the place of intelligence in telecommunications networks. The question is raised both from a technical point of view (the competition of Internet-type networks, where the intelligence is located at the users’ end) and from an economic point of view (opening up of networks and request for the interconnection of competitors). For the moment, however, the intelligent network architecture enables a partial solution to be found to the problem of the length of switching software development cycles.

A Need for Integration Table 3 summarizes the characteristics of specific assets relative to transactions linked with public switching. This degree of specificity is high, but a trend towards a reduction has begun since the beginning of electronic switching. Table 3. Asset Specificity for Digital Switches

Nature of the Specificity Site Specificity Dedicated Assets

Specific Human Resources

Switching Equipment (Hardware)

Software

No

No

Yes, network as “Specific System”

Yes, proprietary systems enforced by the effect of the network as a specific system.

Specialized teams to maintain and upgrade the available function (Decrease with harmonization effort of the supplier)

In conclusion, the specificity of assets appears as the most decisive characteristic for analyzing transactions in public switching. Empirical observations seem to confirm the results of the theory. In industrialized countries, the privileged relationships between telecommunications operators and the equipment manufacturers are the most frequent governance structures. The United States have even seen more than a century’s effective integration in the Bell System. However, digitization and liberalization have also opened up new opportunities which are expressed as falling incentives for this specificity, and therefore a potential questioning of integrated governance structures.

3.3 Liberalization and Falling Incentives for Asset Specificity: Towards the End of Integrated Governance Structures? The liberalization of networks has deeply modified the institutional environment of public TCT shows that the old governance structures are tending to become more and more inadequate with this environment, which is more oriented towards the market. However, the emergence of new structures largely depends on strategic, technological and institutional choices, on which “standard” TCT can only cast a little light, because it partly considers them 21

Cf. Brousseau, Petit and Phan (1996), for a review of these changes and their consequences.

I58

D. Phan and T. Sommer

as “exogenous”. First of all, the emergence of digitization was favored by governance structures with a high degree of control, as underlined by Quelin (1992). Asset specificity was first reinforced by the characteristics of transactions associated with the software. But a trend has since been observed towards a drop in specificity. This largely comes from explicit arbitration where asset specificity plays a strategic role. However, in the absence of interface standardization, interfaces are sufficiently specific to dissuade operators from radically changing the nature of their relations with equipment manufacturers. In the USA, organizational adjustments are nonetheless being outlined, turning towards governance structures closer to the market. Intelligent Network Architectures

Inside an analog/electromechanical telecommunications network, the signaling (numbering, demand for circuits, signal for free or busy circuits, taxation) conveys a low volume of information and performs together with transmission on the same channel. In this period, it was not technically pertinent to separate transmission and signaling. This technical constraint is today no longer valid. The signaling circuit (so-called “CCITT Number 7”) is different from the communication circuit. The opening of this technological entity, formerly closed, has numerous consequences. Intelligent Network Architectures (WAS)allow a new stage to be reached in the modularity of telecommunications systems. These architectures extend the separation between hardware and software by dedicating a specific system to the process of the information and another to the supplying of services. Contrary to the solution that consists of making some specific switches more specialized for network services, INAs first separate the control function and the services facilities into distinct levels, driven by a separate signaling network (CClTT Number 7). That is, an intelligent network is made up of configurable components in order to satisfy the customer’s demands more easily. This is possible through a modular architecture which separates the different network functions (signaling, administration of the network, services provision) from the transportation of information itself, and as a result makes the supply of service independent of the underlying infrastructure. In a second step (IN2) modularity is increased by providing three different interfaces, with the possible opening up to customer services facilities, as an intelligent peripheral through a “Services Switching Point”. For some cases, service provision may be performed by an ordinary work-station located beyond the “Services Control Point”, that is, constituting a specific case of modularization within the software component itself. See also Box 1. In all interfaces, a standardization question is raised for the different levels of transaction considered. According to Man-Sze and Eirwen (1993), the motivation for the standardization of such interfaces varies among players’ interests and the different areas of standardization. Everybody has an interest in standardization for resource management interfaces. Telcos are looking for a decrease in equipment cost, and equipment suppliers are looking for an increase in the market area. If telcos and software providers have an interest in standardizing programming interfaces, hardware providers have a low incentive to do this, in order to keep the telcos as captive customers. At the higher level, only customers and software suppliers have an incentive for standardization. The motivation of the equipment supplier to refuse standardization is the same as previously (to keep the telcos as captive customers). For the telcos, standardization at the service level is an incentive for both competitors and customers to request interconnection to the network; the greater the level of diversity, the greater the local monopoly barriers. Finally, the attribute of transaction greatly depends on the standardization

IO. Governance and Technological Change

159

policy, so the design of new separable interfaces in the INA process clearly appears as an endogenous strategic choice, and not as a given exogenous, purely technical fact.

Box 1: Intelligent Network: Increasing Modularity IN 2

IN 1

P

Service Creation and Execution Service Management System

...........

..................................................................................

...............................

Transaction Processing and Programming Interface : Service Control Point

(SIB)

Intelligent Network Infrastructures ...............................................................

CCITT N"7

((,

......

Resource Management Interface : Service Switching Point

Telco Network Resources Management

Intelligent

In 1984 Ameritech began work on a concept of total separation between service features and switching process. Bellcore further developed the concept of INA by 1985-1986 which gave the IN1 concept, a centralized architecture based on the existing telephone network. Connection with service information was to be located at Service Control Points, with access from switches provided by CCITT No 7 Signaling System. This IN1 was deployed by the RBOC during 1986-1992. During this period, Bellcore studied a new conceptual framework, IN2, and carried out an interim solution named IN1 +, by 1988. IN2 introduced more modularity, with SCP's Service Independent Building Block (SIB) and Elementary Function at the level of the service switching point. In addition, more interfaces were designed for the services and a new system, called Intelligent Peripheral, provided a platform for deploying services assistance capabilities. There exist three interfaces between the elements of the system: first, the resource management interface, located between network resources and IN infrastructures; second, a transactional interface located between the intelligent network and service facilities; third, another interface between the service management system and the services design functionality (creation and execution). The Drop in Asset Specificity

The digital paradigm is expressed via the great increase in fixed costs in the telecommunications equipment industry. R&D expenses have risen considerably with the abandonment of

D. Phan and T. Sommer

I60

purely electromechanical technologies, overtaken by digitally controlled (SPC) space-division technology, then by time-division switching. Equipment manufacturers had to face the problem of increasing fixed costs. They diversified their production in order to benefit from economies of scope, and looked for economies of scale in production. In the framework of the exclusive relationship between operators and manufacturers, systems were set up to get part of this constraint borne by the operator (support via market studies or in the procedures for signing public markets which could be seen through an increase in dedicated specific assets). With the liberalization of the services market and the loss of monopoly income, the operators became more reluctant in giving their support to manufacturers. In order to extend series, and to deliver to increasing numbers of public customers, the manufacturers then tried to overcome the effect of specific systems associated with each national network. Harmonizing ranges between switching systems then became necessary (Quelin,, 1992). The development of competition at an international level was thus accompanied by a strong movement of concentration in the industry. Equipment manufacturers, of whom there were 36 in the world in 1975, were only 12 in number in 1990. Furthermore, under competitive pressure, manufacturers were gradually compelled to abandon the logic of proprietary systems, in the same way that large computer manufacturers had to do in the not so distant past.28 With the development of INA the specific nature of networks was to decline. However, specificity in INA is still important today, and manufacturers are trying to keep this specificity as high as possible. Only the next step in technological change, with the Internet, could significantly reduce this specificity. Degree of network specificity

harmonization of the scope of supplied equipment

Strong

Weak

1

Digital (Time Division) Switching

time

Figure 5. Evolution in the Degree of Specificity in Transactions in Equipment. It should also be added that public switched networks are going to be faced with increasing competition from “IP” databases (based on the Internet Protocol) for providing telecommunications services. At the end of the 1990s, real-time services appeared on the Internet (IP teSpecifically, the high development costs associated with proprietary languages led them to migrate towards standard markets (Unix, C++).

2R

10. Governance and Technological Change

161

lephony, videophony). This world-wide network could not be ignored. Designed to interconnect heterogeneous networks, it is built from standardized equipment, routers, produced by a fast-expanding competitive industry, for which asset specificity is not a significant variable. The choices made by those involved in the industry with the diffusion of the digital paradigm are in favor of less specific assets to support public switching transactions. This could be the case particularly if the regulator decided to impose the openness and the standardization of interfaces in the framework of the opening of XNAs, or if an equipment manufacturer decided unilaterally (but successfully) to abandon proprietary systems and join an open system strategy, playing both on club - or “network” - externalities and on technological development^.^^ Each of these innovations could result in a profound change in characteristics linked to transactions. According to TCT, these changes in characteristics should then be expressed by a search for regulation structures more adapted to operating transactions. In many cases, this could be closer to the market. But the standardization of “technological separation interfaces” is the condition necessary for the market to become an efficient means of coordination. Without this, specific assets remain high and the temptation toward vertical integration remains strong, which is contradictory to the will to promote competition, even more so as there are not very many equipment manufacturers on the market. The forms of such an evolution themselves remain open. This evolution may be stimulated “upwards” (regulation”) or “downwards” (deviant innovative strategy), associated with “defacto” standardization via the market.

4.

THE EVOLUTION OF GOVERNANCE STRUCTURES: THE AMERICAN CASE

The “Ancien Rkgime ’”I of telecommunications favored the development of close relations between operators and manufacturers in the sector, which could even go as far as vertical integration. That was the specific case of the United States where AT&T remained an operator integrated in the production of equipment from its origins until 1996. Similarly, GTE, the principal independent network operator on a local level in the USA, kept an integrated activity producing public switching equipment until the end of the eighties. For several decades, the diffusion of the digital paradigm and the process of liberalization, whose own logic did not exclude a strong degree of interdependence, progressively changed the technological and institutional environment of governance structures. The only American example makes it possible to fruitfully confront what can be learned from TCT with the facts, and therefore to assess the scope and limits from a historical perspective.

This strategy was successfully used by Sun Microsystems in the eighties, when it newly entered the workstation market, and it rapidly became the leader. This success is largely due to the exploitation of network (club) externalities linked with “open” systems, with the consumers’ demand for diversity and “upgradability” and their preference for a modular system, and, finally, to Sun’s capacity to remain the leader technologically after much innovation (Garud and Kumaraswamy 1993). 30 Implementation may then be affected in either a cooperative (normalization) or a decentralized way (standardization through the market - or self-organized). 31 Cf. Phan (1996) for an economic analysis of this “Ancien Rigirne” of telecommunications (the old-fashioned plain telephone service organization: integrated, monopolistic, electromechanical).

29

D. Phan and T. Sommer

162

4.1 The Integrated Model From its creation in 1878, American Bell was a manufacturing company. At that time, the telephone service was indissociable from the telephone itself, and Graham Bell held the patent rights. After having ousted the monopoly of the Western Union Telegraph from the market, the new company was to prosper from 1879 to 1894 under the legal protection which Bell’s patents conferred on it. In 188 1, the Bell system reinforced its production potential by acquiring the electrical materials subsidiary of its ex-rival (Western Electric). After a period of competition, between 1894 and 1934, AT&T was to transform its advantage as a precursor into a local monopoly situation regulated by the public authorities (Mueller, 1994). “The company will establish (...) its domination over the whole of the sector (‘local’ services, trunks, equipment) by controlling strategic patents and by acquiring associated patents, by consolidating its control over the local operating companies and especially by becoming owners of the long-distance transmission network” (Simon, 199 1). The special nature of assets thus results from the initial conditions of the emergence of telephony, where the service (voice communication) was strongly linked to its technical support (the telephone). But it also very largely results from technical and strategic decisions taken later with the aim of strengthening AT&T’s position as a monopoly. The legal environment in the United States was hostile to concentration and to vertical integration (the Sherman Act). The integrated governance structure between the operator and the manufacturer was attacked in court as early as 1910. T.N. Vail, the head of AT&T, was to set up as a system the principles of unity, integrity and universality of the network, and to plead for the creation of a monopoly regime controlled by the public authorities (Simon, 1991; Mueller, 1994; Phan, 1996). ~

United States Environment by 1934 Law : Antitrust Law /Communication Act 1934 Regulation : FCC / PUC Vertical Integration :Bell System

(AT&T + Regional Companies)

~

Telephone Market :Bell in the Monopoly Marginal Competition at the Local Level (20%)

Figure 6. Vertical Integration within the Bell System.

IO. Governance and Technological Change

163

From 1934, the institutional framework in which AT&T operated stabilized. The company, which controlled the whole of the long-distance service (AT&T Long Lines) and 80% of local telephony, underwent two levels of control: that of the Federal Communication Commission, created by the Communications Act, and that of the Public Utilities Commission at State level. During the half century which separated 1934 and 1984, AT&T was to pay the price of defending its monopoly and the integrity of its network in lawyers’ fees and propaganda. The company managed to maintain this skzatus quo by invoking “universal service”, with the backup of regulators and of certain large administrations (Simon, 1991), sometimes having to renounce certain important matters, as shown in the Consent decree of 1956.

4.2 The Imbalance between the Institutional Framework, Technological Evolution and Governance Structures In 1982, after the third antitrust lawsuit against AT&T, the American Department of Justice and the operator came to a compromise about separating the long-distance activity from the local operations of the network. This decision aimed to enable a competitive situation to be established in the long-distance market. Seven regional companies (the RBOCs) in a severely regulated local monopoly were to be created by this split-up. This became effective in 1984. At the end of this agreement, AT&T kept the long-distance service, remained integrated in the equipment sector with its subsidiary Western Electric, and obtained the authorization to enter the computing markets.3* United States 1984 Law : Antitrust Law /Communication Act 1934 Regulation : FCC / PUC Divestiture / Breaking Up : Consent Decree (1982)

I

Vertical Integration : AT&T

Market

.

Local Monopolies

I

Network and Trunk Services

Competition on Trunk Network

U Other Sources of Provision (SuPPlYind Figure 7 . Vertical Integration after 1984.

32

Controlling the “long lines” and verticd integration have been the key to AT&T’s success since the beginning of this century. AT&T’s managers then found themselves in what Dang Nguyen and Phan (1997) have called the “backward-looking dynamics” of expectations.

164

D.Phan and T. Sommer

This compromise in 1984 created an unstable situation from a transactional analysis point of view. Thus, the features of switching transactions for local networks are always characterized by a network specificity, which was strengthened over time in the Bell system. Nevertheless, the market became the governance structure imposed during these transactions. To satisfy the requirement of these particular transactions, Bellcore, the laboratory common to the RBOCs created in 1984, had an interfacing role in the new regulation structure, to avoid the risk of market failure, and especially to coordinate the technological choices and technical norms between RBOCs. From then on, the RBOCs were threatened with being locked in the relationship with AT&T, as the latter supplied them with network equipment (Filoche, 1997). As TCT claims, when objectives differ between firms, the risk of opportunistic behavior increases, and with this risk, so do the transaction costs. Because the RBOCs ceased to belong to the Bell System, their marketing policies and/or investment choices no longer depended on a sole decision maker. To coordinate the objectives of the parties involved would thus enable the situation to stabilize, but such a complete alignment of objectives was hardly conceivable in the new competitive framework. Everybody could equally demand particular evolution in material or specific functionalities for switching software. With the emergence of Intelligent Network Architectures, a growth in the modularity of software functionalities was seen in the USA. In particular, certain operators designed servicecreation platforms on workstations connected to “Service Control Points” with the aim of developing services autonomously, in relation to equipment manufacturers. So, MCI selected vertical integration after the in-house development of this service function; Bell Canada and Bell Atlantic experimented with mixed methods which combined in-house development, associations with software engineering companies and sometimes the outside purchase of offthe-peg software solutions.33This evolution placed the acquisition, control and development of competence, rather than the features of the transaction itself, at the center of the operators’ strategy. For previously mentioned methodological reasons, this type of problem has not given rise to significant developments in the TCT framework. Faced with this new trend towards modularization and with the explosion of the software market, AT&T decided in 1995 to re-focus its activities on services and abandon the production of public equipment. This second break-up of AT&T became effective in 1996 and the manufacturing activities of the operator were brought together in a new, legally independent company, Lucent T e c h n ~ l o g i e s . ~ ~

Derian (1991); for example, at the beginning of the eighties, Bell Canada bought a turnkey billing platform from GTE data systems, and a services platform from MCI.

13

34

AT&T broke up into four different companies according to their basic competence and without any significant financial connections: one company has kept the name AT&T (for long-distance and mobile telephony, phone cards, Communications Services Group, AT&T Universal Card Service Corp., AT&T Solutions, AT&T Wireless Services); one company, Lucent Technologies, oversees all manufacturing activities (Network Systems Group, Global Communications Systems, Consumer Products, Paradyne and Microelectronics); there is one company for computing equipment (AT&T GIS (ex NCR)); and there is one finance company, Capital Corp.

IO. Governance and Technological Change

165

United States Environment by 1998 Law : Antitrust law /Communication Act 1996 Regulation : FCC / PUC

Vertical Disintegration

- -0theT. Suppliers I I

of Telecom Services

1-

I

-1

I I

I

,

--I

-Suppliers Other Of

I Equipments

I I I

, - , - - I

Figure 8. AT&T’s Second Break-up. AT&T’s second break-up may be justified from a company point of view through its desire not to penalize its manufacturing activities in a situation where it was in competition with its customers (Rey and Tirole, 1996). It seems therefore that after 118 years of vertical integration, controlling equipment is no longer a key element of AT&T’s strategy, and they are especially seeking to develop in the multimedia market or as a “global” operator. This decision also seems to be coherent with the new American institutional context. In 1996, the Telecommunications Act brought into being the ‘‘ decompartmentalization of all the telecommunications markets anticipated by AT&T. The local monopolies of the regional companies are destined to disappear, like the restrictions made on the RBOCs in 1982 concerning their operating in services and manufacturing activities. ”

5.

CONCLUSIONS

5.1 Lessons from a Practical Point of View Transactional analysis enables three periods to be distinguished. At the time of electromechanicallspace-division switching, the switching system can be seen as a unique technological entity, which gives rise to a single transaction between telecommunications operators and equipment manufacturers. The high degree of asset specificity has sometimes been reinforced by national strategies, protected by a local monopoly. In the American case, the strategy of AT&T has led to its vertical integration within the Bell System. As soon as space-division electronics appears, the digitization of switches makes it possible to distinguish the hardware from the software component in the equipment, corresponding to a new technologically separable interface in the TCT sense. Transactions concerning software are to become the support and the main stake of asset specificity. Originating in highly

166

D. Phan and T. Sommer

controlled governance structures, time-division switching is first expressed by a reinforcing of this specificity. However, the spread of the digital paradigm, into which this technology is integrated, and the process of liberalization of telecommunications, direct the choices of those involved towards a decrease in asset specificity. In fact, the growing costs of R&D associated with the taking up of digital systems by equipment manufacturers induce them to multiply the number of clients and to harmonize their range of equipment. As a consequence, asset specificity has to be reduced. At the end of this period, public switching has become a global oligopolistic market, controlled by a few major companies and/or networks of alliances (Quelin, 1996). Vertically disintegrated operators, like RBOC, make the equipment supplier market more competitive, and favor the emergence of Intelligent Network Architecture at Bellcore. The latter begins an additional step in the process of more modular switches with new opportunities for technological separation. (However, the adjustment process of the governance structures seems largely disconnected from such a process. TCT has some difficulty taking into account these dynamic aspects, even though this theory makes it possible to analyze the final consequences of such a process.) “Vertical disintegration” is observed, as in the case of AT&T or of GTE, when the attributes of transactions are still characterized by high levels of specificity. In fact, equipment manufacturers and operators have to arbitrate between short-term and long-term considerations. In the short term, it is in the interest of equipment manufacturers to maintain the specific nature of the assets associated with their proprietary systems, in order to keep operators dependent on them, while it is in the interest of the operators to take a flexible technological stance, in order both to make their existing investments pay and to avoid committing themselves to making costly or dead-end choices, while technologies of the future (like ATM switching or multimedia data networks) are far from being stabilized. In the longer term, equipment manufacturers will certainly have to decrease their costs and commit themselves to more modular and more standardized systems. But then they will be in competition with the manufacturers of computer networks and systems integrators. The operators will themselves have to make their networks evolve, but in the meantime, competition leads them to worry first about adapting their services to the needs of their clients.

5.2 Theoretical Lessons The analysis of transactions in the public switching telecommunications industry confirms the limitations of TCT highlighted at a more general level by authors mentioned in the first part of this paper. According to Groenewegen (1996), TCT is adequate for comparative static analysis between governance structures. Moreover, the analysis confirms that TCT is able to identify the instability of the current situation. But, according to the theoretical arguments of Beije (1996), Noteboom (1996), Magnusson and Ottosson (1996), among others, this theory is experiencing some difficulties in describing the evolutionary process. That is no doubt why most research about these questions uses an evolutionary approach.” In fact, TCT provides a framework to compare governance structures across time and/or across space, for a given in35

All these works underline the “historical” dimension of the process, caused by irreversibility factors. For instance, Fransman (1992) and Quelin (1992, 1996) underline the role of competence in the resulting technological trajectories. More specifically, Wolff (1996) analyzes the nature of the learning process related to the forming of competence, in the case of strategic alliances. For the latter author, the transaction cost approach, at least in its standard version, suffers from a certain number of limitations when we try to analyze the dynamics of telecommunications agreements. “They can all be attributed to the essentially static dimension of this approach” (Wolff, 1996, p. 458).

10. Governance and Technological Change

167

stitutional environment, along comparative static principles. W e can thus put forward elements to justify the existence of different governance structures in different contexts. On the other hand, the explanation for the evolution of one structure towards another goes beyond the field of “standard” TCT. Wishing to use TCT in a morphogenetic way, if institutional evolution is considered as given, would lead to a functionalist interpretation of governance structures. To integrate in the long run, it seems necessary to describe more specifically the feedback effect of governance structures on the institutional framework (in the short run, such an effect is considered to be of the “second order” by Williamson). It would also be necessary to better integrate the trade-off between technology and strategic adaptation behavior, i.e. to consider a field wider than the one which “standard” TCT refers to. Such a research program remains to be constructed. Several directions are possible. Lotter (1997) thinks that we can develop evolutionist dynamics by keeping transaction as a sole support for the analysis. Other authors claim that greater integration between TCT and the evolutionist research program, judged to be complementary, should append significant enhancements (Foss, 1996; Brousseau, 1996). For Chandler (1992) the facts observed by the historian of company organization refer to this approach more than any other.36 The relative independencehnterdependence that has been highlighted between the institutional-organizational level and the technological level could thus be treated by using the notion of “co-evolution” borrowed from biology (Nelson, 1994). However, these developments remain merely exploratory and at the program stage. Therefore, this study suggests that understanding the evolution of the relationships between equipment manufacturers and telecommunications operators would certainly be enriched by such contributions.

5.3 Further Research Finally, the public switching manufacturer’s ability to innovate in the 1970s and 1980s is known to be linked to vertical integration or vertical quasi-integration with telcos (Dang Nguyen, 1983; Quelin, 1992; Harris, 1993). In such “hybrid forms of governance”, the distribution of competence between partnership matters. For instance, one may raise the question of the opposing cases of French and German hybrid forms of governance in relation to competence in digital switching technologies. In France, competence is mainly located at the CNET.” In Germany, the main location of competence is at Siemens. Is this asymmetric distribution of competence between partners more significant than transaction costs when explaining the role of governance structures in the emergence of digital technologies (i.e. Alcatel’s early leadership)? Recently, Praest (1998) has used patents in the telecommunications equipment industry to analyze the effect of competence accumulation and strategic behavior at the level of the firm throughout industrial dynamics. No doubt further research about the telecommunications industry will make significant progress by using a pluralist approach. Another issue that might open new avenues for research is the convergence between IP (Internet Protocol) data routed networks and switched networks. Given the congestion of the Internet and the necessity to price-discriminate among the users, it is likely that the routers 36

c‘ ... the evolutionary theory of the firm, which emphasizes continuous learning that makes a firm’s assets dynamic, provides an understanding of why in the past and how new firms grew by expanding into new markets through the process of integrating production and distribution. I find the earlier growth of the industrial firm difficult to explain fully in terms of transactions, agency and other information costs, so I find it hard to explain the recent process of expansion and contraction with these same concepts” (Chandler, 1992, pp. 98-99).

37

The research center of the former DGT -Direction Gdne‘rale des Te‘ltcommunications - now France Telecom.

168

D. Phan and T. Sommer

will become more and more complex and integrate some of the features that one finds in telephone switches. From an equipment design point of view, this will lead to a greater similarity between the two types of products. This, on the other hand, will have some consequences for the industry structure. Hence the focus will be shifted to horizontal - or “lateral” - integration among public telecommunications equipment and private data network manufacturers, attested by the recent takeover of Bay Networks by Nortel, instead of the vertical quasiintegration. In addition, it is important to remark that IP is gaining ground in relationship with the building of a “post-telephony’’ economy (Fleurot and Sommer, 1999). This process shifts the frontier between telecommunications and the computer industry. In this context, vertical relationship has to be investigated in a complex and fast-moving environment.

REFERENCES Abernathy, W.J. and Utterback, J.M. (1978), “Patterns of Industrial Innovation”, Technology Review, VOI.80, NO.7, pp. 40-47. Akerlof, G.A. (1980), “A Theory of Social Custom, of Which Unemployment May be One Consequence”, Quarterly Journal of Economics, Vol. 94, No. 4, June, pp. 75 1-775. Beije, P.R. (1996), “Transaction Costs and Technological Learning”, in Groenewegen (ed.) (1996), pp. 309-326. Bernheim, B.D. (1994), “A Theory of Conformity”, Journal of Political Economy, Vol. 102, No. 5, pp. 841-877. Bowles, S. (1998), “Endogenous Preferences: The Cultural Consequences of Markets and other Economic Institutions”, Journal of Economic Literature, Vol. 36, March, pp. 75-1 11. Brousseau, E. (1996), “New Institutional Economics and Evolutionary Economics: What Convergence?”, 8th EAEPE Conference Antwerp, Belgium, 7-9 November. Brousseau, E., Petit, P. and Phan, D. (eds.), (1996), Mutation des Te‘le‘cornmunications,des Industries et des Marche‘s, ENSPTT-Economica Paris. Chandler, A.D. (1992), “Organizational Capabilities and the Economic History of the Industrial Enterprise”, Journal of Economic Perspectives, Vol. 6, No. 3, pp. 79-100. Chapuis, R.J. and Joel, A.E. (1990), 100 Years of Telephone Switching, North Holland. Coase, R.H., (1937), “The Nature of the Firm”, Economica 4, pp. 386-405; reprint in Williamson and Winter (eds.) (1991). Coase, R.H. (1994), “Faut-il vendre les frdquences ?’, Re‘seauxN”64, mars-avril. Coriat, B. and Weinstein, 0. (1995), Les nouvelles the‘ories de l’entreprise, Livre de Poche Refdrences Dang Nguyen, G. (1983), Les te‘le‘communicationsen Europe durant la de‘cennie 1970-1980: Thesis, Institut Universitaire EuropCen, Florence, 376 pp. Dang Nguyen, G. and Phan, Denis (1998), “Learning and the Diffusion of the Digital Paradigm in Information and Communication Technology”, in Macdonald, S . and Madden, G. (eds.), Telecomrnunications and Socio-Economic Development, Elsevier Science Publishing, pp. 275-293. Davis, L. and North, D.C. (197 l), Institutional Change in American Economic Growth, Cambridge University Press. Delattre, P. (197 l), Systdrne, structure,fonction, e‘volution,Maloine, Paris.

10. Governance and Technological Change

169

Derian, J.C. (1991), Les nouveaux rapports entre les exploitants de les fournisseurs d’e‘quipements de te‘le‘communication,Etude rCalisC pour France TClCcom DPS, November. Dosi, G. (1982), “Technological Paradigms and Technological Trajectories”, Research Policy, Vol. 2, NO.3, pp. 147-162. Dosi, G. (1988), “Sources, Procedures, and Microeconomic Effects of Innovation”, Journal of Economic Litterature, Vol. 26, pp. 1120-1171. Elster, J. (1983), Sour Grapes, Cambridge University Press. Englander, E.J. (1988), “Technology and Oliver Williamson’s Transaction Cost Economics”, Journal of Economic Behavior and Organisation, Vol. 10, pp. 339-353. Filoche, C. (1997), “Technological Innovation and the Dynamics of Organizational Forms: The Case of Software in The Telecommunications Industry.”, Actes du 3” Colloque Economie Publique Applique‘e Brest, 12-13 juin. Fleigstein, N. (1990), The Transformation of the Corporate Control, Harvard University Press, Cambridge. Fleurot, I. and Sommer, T. (1999), “Web as a New Mass Medium Charging for Internet Services”, Second Berlin Internet Economics Workshop, May. Foss, J. (1994), “Why Transaction Cost Economics Needs Evolutionary Economics”, Revue d’Economie Industrielle, No. 68, 2” trimestre. Foss, J. (1996), “Capabilities and the Theory of the Firm”, Revue d’Economie Industrielle, No. 77, 3” trimestre pp.7-28. Fransman, M. (1992), “AT&T, BT, NTT, Vision, Strategy, Corporate Competence, Path Dependency and the R81e of R&D’, paper presented at the Ninth International Conference at the International Telecommunications Society, Sophia Antipolis, June 1994. Gaffard, J. L. (1990), Economie industrielle et de l’innovation, PrCcis Dalloz, Paris. Garud, R. and Kumaraswany, A. (1993), “Changing Competitive Dynamics in Network Industries: An Exploration of Sun Microsystems’ Open Systems Strategy”, Strategic Management Journal, Vol. 14, pp. 351-369. Groenewegen, J. (ed.) (1996), Transaction Cost Economics and Beyond, Kluwer Academic Publishers, Boston. Gmpp, H. and Schnoring, T. (1992), “Research and Development in Telecommunications, National Systems under Pressure”, Telecommunication Policy, JanuaryIFebruary, pp. 46-66. Guertman, M. and Quelin, B. (1993), “An Empirical Study of Hybrid Forms of Governance Structure: The Case of the Telecommunication Equipment Industry”, Groupe HEC, Working Paper, 1993. Harris, J.E. Oxley (1993), “Competitive Implications of Vertical Relations Between Telecommunications Equipment and Services”, paper presented at the ITS Regional Conference, Goteborg, June. Jones, S.R.G. (1984), The Economics of Conformism, Basil Blackwell, Oxford. Klein, P. and Shelansky, H. (1996), “Empirical Research in Transaction Cost Economics: a Survey and Assessment”, WP BBP-60 Center for Research in Management, University of California at Berkeley; forthcoming in Journal of Law,Economics and Organization. Kubiak, G. and Baptiste, G. (1993, “Raccourcissement des temps de cycle dans les paliers logiciels”, L’e‘chodes Recherches, Vol. 160, pp. 15-26. Kuhn, T. H. (1970), The Structure of Scientific Revolutions, Chicago University. Langlois, R.N. (1992), “Transaction Cost Economics in Real Time”, Zndustrial and Corporate Change, June, pp. 99-127.

I70

D. Phan and T. Sommer

Langlois, R.N. and Robertson, P. L. (1995), Firms, Markets and Economic Change A Dynamic Theory for Business Institutions, Routledge, London and New York. Lazonick, W. (1991), Business Organization and the Myth of the Market Economy, Cambridge University Press. Leibenstein, H. (1987), Inside the Firm, the Inefficiencies of Hierarchy, Harvard University Press, Cambridge, Mass. Libois, J-L. (1993), Gbnndse et croissance des tkle‘communications, CNET-Masson, Paris. Lordon, F., (1996), “Formaliser la dynamique et les crises rCgulationnistes”, in Boyer, R. and Saillard, Y. (eds.), The‘orie de la regulation, L’etat des savoirs, La DCcouverte, Paris, pp. 264-272. Lotter, F. (1 994), “CoQts de transactions et fondements de l’intervention publique”, W.P.94-16, Centre ATOM-UniversitC de Paris I. Lotter, F. (1997), “ElCments de dynamique transactionnelle gouvernance et organisation industrielle”, WP 97-01, Centre ATOM UniversitC Paris I. Magnusson, L. and Ottosson, J. (1996), “Transaction Costs and Institutional Change”, in Groenewegen (ed.) (1996), pp. 35 1-364. Man-Sze, Li and Eirwen, Nichols (1993), Intelligent Networks: Strategies for Customized Global Services, Ovum, London. Menard, C. (1995), L’e‘conomie des organisations, La Dkouverte, Paris. Mueller, M. (1993), “Universal Service in Telephone History”, Telecommunication Policy, July; trad. franqaise: “Le service universe1 dans l’histoire du tC1Cphone: une reconstruction ”, Reseaux, no 66, 1994. Nelson, R. R. (1994), “The Co-evolution of Technology, Industrial Structure and Supporting Institutions”, Industrial and Corporate Change, Vol. 3, No. 1, pp. 47-64. North, D. C. (1990), Institutions, Institutional Change and Economic Perjormance, Cambridge University Press. Noteboom, B. (1996), “Toward a Learning Based Model of Transactions”, in Groenewegen (ed.) (1996), pp. 35 1-364. Phan, D. (1996), “L’ouverture des rCseaux de tClCcommunications”, in Brousseau, Petit and Phan (eds.) (1996). Piaget, J. (ed.) (1967), Logique et connaissance scientifque, NRF EncyclopCdie de la PlCiade, Paris. Piaget, J. (1968), Le structuralisme, PUF, Paris. Praest, M. (1998), “A Dynamic Approach to the Analysis of Technological Competence, Accumulation and Strategic Behaviour in Telecommunications”, Communications & Strategies, Vol. 29-1, pp. 79-1 14. Quelin, B. (1992), “Trajectoires technologiques et diffusion de l’innovation: l’exemple des Cquipements de tClCcommunications”, Revue d’Economie Industrielle, Vol. 62, No. 4, pp. 83-105. Quelin, B, (1996), “Dynamique concurrentielle et globalisation: la concentration de l’industrie des Cquipements de tClCcomunications” in Brousseau, Petit and Phan (eds.) (1996), pp. 431-483. Rey, J. and Tirole, P. (1996), “A Primer on Foreclosure”, Working Paper, IDEI, Toulouse. Riordan, M. and Williamson, O.E. (1983, “Asset Specificity and Economic Organisation”, International Journal of Industrial Organisation,Vol. 3. Sally, R. (1993), “Alcatel’s Relations with the French State: the Political Economy of a Multinational Enterprise”, Communications & Strate‘gies,N”9, 1” trimestre.

10. Governance and

Technological Change

171

Saussier, S. (1996), “ThCorie des coQts de transaction et durCe des contrats: une analyse empirique”, Journe‘es AFSE, Caen. Schotter, A. (1981), The Economic Theory of Social Institutions, Cambridge University Press. Simon, H.A. (1976) “Rational”, in Latsis (ed.), Method and Appraisal in Economics, pp. 129-148; trad. fraqaise, “De la RationalitC Substantive a la RationalitC ProcCdurale”, Pistes, 3-octobre 1992, pp. 25-43. Simon, H.A. (1984), “On the Behavioral and Rational Foundation of Economic Dynamics”, Journal of Economic Behavior and Organization,Vol. 5 , pp. 35-55. Simon, J.P. (1991), L’esprit des rkgles, re‘seaux et re‘glementation aux Etats-Unis, Collection logique juridiques, L’Harmattan, Paris. Wallis, J.J. and North, D.C. (1994), “Integrating Institutional Change in Economic History. A Transaction Cost Approach”, Journal of Institutional and Theoretical Economics, Vol. 150, No. 4, December, pp. 609-624. Walliser, B. (1989), “Instrumental Rationality and Cognitive Rationality”, Theory and Decision, Vol. 27, July/September, pp. 7-36. Williamson, O.E. (1979, The Markets and Hierarchies: Analysis and Antitrust Implications, The Free Press, New York. Williamson, O.E. (1981), “The Modem Corporation: Origins, Evolution, Attributes”, Journal of Economic Literature, Vol. 19, December, pp. 1537-1568. Williamson, O.E. (1985), The Economic Institutions of Capitalism, The Free Press. Williamson, O.E. (1987), “Transaction Cost Economics: The Comparative Contracting Perspective”, Journal of Economic Behavior and Organization, Vol. 8, pp. 617-625. Williamson, O.E. (1988), “Technology and Transaction Cost Economics: A Reply”, Journal of Economic Behavior and Organization, Vol. 10, pp. 355-363. Williamson, O.E. (1992), “Markets, Hierarchies, and the Modern Corporation: An Unfolding Perspective”, Journal of Economic Behavior and Organization, Vol. 17, pp. 335-352. Williamson, O.E. (1993a), “Transaction cost Economics and Organization Theory”, Industrial and Corporate Change, Vol. 2. Williamson, O.E. (1993b), “The Economic Analysis of Institutions and Organizations. In General and with Respect to Country Studies”, Working Paper No. 133, OCDE, Paris. Williamson, O.E. (1995), “Hierarchies, Market and Power in the Economy: An Economic Perspective”, Industrial and Corporate Change, Vol. 4, No. 1, pp. 21-49. Williamson, O.E. and Ouchi W. G. (1981), “The Markets and Hierarchies and Visible Hand Perspectives”, in Van & Joyce (eds.), Perspectives on Organizing Design and Behavior, New York, pp. 347-370. Williamson, O.E. and Winter, S. G. (eds.) (1991), The Nature of the Firm, Oxford University Press, 1991. Winter, S. G. (1991), “On Coase, Competence and the Corporation”, in Williamson and Winter (eds.) (1991), pp. 179-195. Wolff, S. (1996), “Dynamique des accords inter-entreprises dans le secteur des tClCcommunications: analyse de deux cas contrastCs”, in Brousseau, Petit and Phan (eds.) (1996), pp. 457-484.

0 2000 Elsevier Science B.V. All rights reserved Convergence in Communications and Beyond E. Bohlin, K. Brodin, A. Lundgren and B. Thorngren (Editors)

CHAPTER 11

The Impact of Information Technology on the Boundaries of the Firm Marcus Garbe" Argonauts, Miinchen

Abstract. This paper uses the property rights and transaction cost approaches to analyze the impact of IT on the boundaries of the firm. The impact of IT on the specificity of investments and on the determinants of integration is discussed. The hypothesis that market compared to hierarchy will gain efficiency with IT can only be weakly supported by theory. A trade-off between the flexibility effect and the complementarity effect of IT is identified, and depending on the overall effect of these two, IT might support a trend towards integration as well as towards deintegration. A case study of the production process of printed matter (pre-press and print) in Germany delivers empirical evidence.

1. INTRODUCTION Information Technology (IT) has enhanced efficient information flows with a view to coordinating economic activities in markets and hierarchies. The efficiency and reliability of exchanging information has increased, making information about supply and demand of a given good available on a world-wide basis. However, questions remain: What is the impact of IT on the efficiency of the coordination in the firm? Will there be a shift of the efficient boundaries of the firm? The aim of this paper is to discuss the hypothesis that the market will gain efficiency compared to hierarchy through the application of IT.

1.1 Different Conceptual Frameworks to Explain the Impact of IT The impact of IT on the efficiency of coordination of economic activities has been analyzed in the framework of transaction cost economics (TCE) (Williamson, 1975; 1979; 1991). The analysis shows that the impact of IT on the efficiency of coordination is to a high degree * This paper has been mainly written while I was visiting the Massachusetts Institute of Technology working as a guest researcher in Spring 1998. I would like to thank Erik Brynjolfsson, Oliver Hart, Susan Helper and Jose Tribolet for helpful comments. My adviser Carl-Christian von Weizsacker, Walter Elberfeld and Birgitta Wolff provided valuable input on an earlier version of this paper.

I 74

M . Garbe

determined by a reduction of the specificity of investments (Garbe, 1998). However, the arguments to give an answer to the shifting boundaries of the firm are vague; TCE is not able to provide a detailed analysis of the benefits and costs of integration since the costs of integration are not much discussed. Williamson uses exogenous variables such as rising costs of bureaucracy or a lack of incentives as one possible source. Due to the limitations of TCE to determine the efficient boundaries of the firm, the results are not robust. However, the property rights theory (PRT) is used as a new theoretical lens to examine the given hypothesis (Hart, 1995). PRT is a framework which is potentially able to give a better understanding of the boundaries of the firm and how they are affected by the use of IT. The key variables used in this paper which will help to understand the mechanisms behind the changing boundaries of the firm are the specificity of investments and complementarity of assets. The former represents the explanatory variable for the shifting boundaries of the firm mainly used in TCE, the latter in PRT. One obstacle to gaining clear-cut results is constituted by the nature of IT: IT might support flexibility and yield a trend towards de-integration in an environment where common ownership has been the efficient ownership structure without the use of IT. On the other hand, if the complementarity effect dominates, a trend towards integration might be supported in an environment where separate ownership of each asset has been the efficient structure without the use of IT. These results will be obtained if IT is considered to be a catalytic medium for an efficient information flow and is treated as a given, but not separately modeled, variable.'

1.2 Organization of the Paper In the following section the specificity effects of IT are discussed in brief and a classification of different types of specificity is introduced. In the third section, the impact of IT on the determinants of contracting (complementary assets and independent assets) is analyzed. In the last section the paper delivers empirical evidence, a case study of the production of printed matter (pre-press and print) in Germany. Conclusions follow.

2. DE-SPECIFICATION OF IT INVESTMENTS: TCE PERSPECTIVES The literature contains evidence that IT might reduce specificity of investments. One observed shortcoming of the literature is the generalization of a de-specification effect of lT.Different types of specificity are not distinguished, although the impact of IT on these various types differs. The purpose of Table 1 is to overcome this inadequacy and to present a synopsis of the impact of IT on specificity. It demonstrates that the impact of IT on asset specificity depends on the type of specificity. The different types of asset specificity are based on the classification developed by Williamson (1985) and Benjamin, Malone and Yates (1986), and have been extended by the author.

'

The impact of IT will be treated as (1) decreasing average unit cost of IT, (2) increasing information availability and processing capacity, (3) increasing standardization and interconnection (Clemons, Reddi and Row, 1993). See Jacobides (1996; 1997).

I I . The Impact of Information Technology on the Boundaries of the Firm

I 75

Table 1 suggests the following interdependence of IT and asset specificity: Specificity might be reduced if the factor “information” has a higher relevance as an input factor compared to other factors. The mobility of specific investments is of great importance. If a specific investment has a highly immobile character (e.g. exploration of mineral resources) or specific presence is required to develop its highest value in cooperation with another asset, IT cannot reduce specificity. IT enables markets to serve as an efficient coordination tool for a higher level of specific investments as compared to a “non-IT world”. If specificity can be reduced, the market gains efficiency, since the need to safeguard specific investments in a firm has been reduced. Thus, IT might support a trend towards disintegration. However, since the costs of integration are not well developed in TCE, the results of the impact of IT on firm boundaries are not robust.

3.

IMPACT OF IT ON CONTRACTING: PRT PERSPECTIVES

The property rights theory as applied to integration has mainly been developed by Grossmann and Hart (1986). It analyzes the incentive effects of different ownership arrangements and compares the value (marginal return on investments) of alternative asset coalitions. The main benefit of this approach is that it is based on a formal model, which allows the precise quantitative determination of the costs and benefits of integration. This is an important advance towards the identification of the efficient boundaries of the firm compared with TCE. On the other hand, PRT focuses on the very formal structure of a firm: the efficient distribution of ownership rights, but is not able to provide a picture of the informal structure of the firm. TCE on the other hand, can analyze more complex relationships, such as relational contracts.’ The consideration of formal and informal structures of a firm would be valuable to overcome the current limitations of the theory of the firm.’ In this paper I use a simple model of Hart (1995) to illustrate the impact of IT on the boundaries of the firm. It is assumed by Hart (1995) that in the case of complementary assets, integration represents the efficient ownership structure. Under conditions of integration, incontractible residual rights will not occur, because the exchange between the upstream and the downstream parties is coordinated by a single party through common ownership. In the case of independent assets, where de-integration represents the efficient ownership structure, there might be parts of the exchange relation which are not fully contractible (level of quality, delivery dates, after-sales services). In this case, the allocation of the residual rights will have an important impact on the bargaining position of the parties after they have made specific investments to support the given exchange relation. However, IT will not directly increase the share of contractible elements of a given exchange relation since the contracting problem does not depend on information asymmetries, but on ill-defined property rights. IT might have an impact on the efficient ownership structure by changing the level of complementarity or independence of assets, and this in turn will affect the level of incontractible elements of a given exchange relation. The key question is then how IT affects the characteristics of complementary and independent assets.

M. Garbe

176

'able 1: rype of

In- iact of IT on Different Typ IT application

ipecificity nvestments in T

of Specificity mpact on specificity

teduction of specificity h e to increasing standardization and a ransition to more flexible software and iardware, the redeployability of IT investnents will increase. (Clemons, Reddi and IT,~,v~;. j ,j = 1,...,J ,

0 otherwise os:

where I is the indicator function (i.e. a function taking value 1 if the argument in brackets is true and 0 otherwise). This gives an “observational rule” to link observed (strategies) and unobserved (profits and capabilities) characteristics of the firm. Lastly, at this point, the firm is described by the system of equations (l), (2) and (3):

W e could interpret the system (4) as a structural model; unfortunately, it cannot be used for econometric purposes because the capabilities C,, are not observed. Therefore we are compelled to resort to a reduced form of model, in which the expression for the capabilities is substituted in the profits:

yielding:

n;: = (r, xr:,x;, qi,mi; ) >

9

,

1

= 1,...,N , t = 1,...,T , j =1, ...,J .

14. Firms’ Heterogeneity and Dynamics of Ently in a New Sector

243

Now, we have to specify a functional form for the expected profits and a distribution for the error terms. W e choose a linear functional form:I4

where: 9

p,’ is the vector of coefficients of the variables entering the “capabilities creation function”;

pi 0

is the vector of coefficients of market variables;

p: is the vector of coefficients of the variables capturing the behaviour of other firms; p i is the vector of coefficients of the unobservable variables entering the capabilities creation function.

Now, we have only to choose a distribution for the errors r\,, and o,‘, . In the following we will suppose that p i = p, ; in this case, the random term p, .r\,, is the same in all the J profit equations and can therefore be eliminated from our analysis, since it disappears when considering the difference between two profits: this hypothesis will be justified by the Appendix. Moreover we suppose that the o:~s are independently and identically distributed Weibull random variables.” Under these assumptions we obtain, for the strategies d!: , a multinomial logit model, in which:

k=l

The Appendix will test the correct specification of the multinomial logit model (8). In this form, the model is not identified since we have to set to 0 the coefficients of an alternative. While it is clear that any other choices would have been equally acceptable, in the estimation we set to 0 the parameters relative to the multimedia service provision, since this provides a useful way of testing, via direct Wald tests, the difference between the other two alternatives and this one. Moreover, note that the variables capturing the behaviour of the competitors (that is, CLUSMSP and WORLDMSP) are present both in x: and in y,, , and therefore the esNote that under quite broad conditions, no loss of generality is involved in the choice of a polynomial, since every continuous function of uniformly limited arguments can be approximated to any desired degree of precision by a (Chebyshev) polynomial as the specified one; the only problem is the degree of the chosen polynomial, but we believe that in this case a linear form should give good results. I s However, the random unobserved factors entering the capabilities production and decision processes are SO many and so erratic that a central limit theorem could be invoked to justify the choice of a normal distribution for p i .qil . Our choice has been motivated mainly by the performance of the tests of Appendix A. and

M.G. Colombo, P. Gurrone, R.G. Seri

244

timated coefficients will be the sum of the two: in this case the coefficients of the structural model are not identified. Concluding, we have estimated the model by maximum likelihood using the following loglikelihood function: ''

i, j , i

Given the model (S), the research hypotheses may be empirically specified, by formulating a set of hypotheses upon the coefficients pNoTHING and pmo. A positive estimated value for a parameter of the pNoTHING vector implies that an increase in the related explanatory variable lowers the profit expected from MSP with respect to the profit expected from NOTHING; a positive estimated value for a parameter of the pNWo vector implies that an increase in the related variable lowers the profit expected from MSP with respect to the profit expected from NWO. Hypothesis I : Historical capabilities. In accordance with the discussion carried out in Section 2 on the partly competence-destroying nature of the MIS paradigm, technological knowledge specific to the carriers, as proxied by DIG, is likely to be obsolete and unable to support MSP conduct (i.e. the corresponding coefficient in PINOTHING is null)." A more radical view is that this kind of expertise may even act as a liability with respect to adaptation of internal capabilities to MSP (resulting in strongly positive coefficients). CELL proxies both technological and marketing capabilities: a mixed effect is thus expected and no hypothesis is explicitly formulated. As to the management and marketing expertise, if it has been developed to cope with the requirements of more advanced telecommunications markets (proxied by T R , CATV and ZNT), a firm is more likely to embrace MIS strategy. Further, other things being equal, our proxies for historical capabilities in management and marketing activities are likely to be less relevant to NWO conduct. An increase in TR, CATV and ZNT should favour MSP rather than NOTHING or NWO; accordingly, our hypotheses can be specified as follows:

Consistency, asymptotic normality and efficiency are guaranteed reasoning along the lines of Kaufmann (1987) and Fokianos and Kedem (1998). "

Rather, they might help in incrementally changing a firm's routines, if some opportunities arise in segments of

multimedia other than MIS. It follows that coefficient p i w o

may now turn out to be positive.

14. Firms’ Heterogeneity and Dynamics of Entry in a New Sector

245

Hypothesis 2: Learning devices.Neither R&D nor learning through experience are likely to be relevant for entry into MIS. Accordingly, the effect of RDTOT and INITMSP should be null in the MSP equation; on the contrary, RDTOT could favour the incremental learning process underlying a NWO strategy:

As regards learning by imitation, unfortunately, given the reduced form of (8), we are not able to discriminate between imitation as “follow-the-leader’’ and imitation as a learning process. However, it is reasonable to assume that the former affects a firm’s strategies much more rapidly than the latter; therefore, at least in the current development phase of the MSP industry, p 3 will capture bandwagon effect rather than learning through imitation. Previous sections have argued that interaction with other firms is the central learning process activated by telecommunications firms that are looking for a set of capabilities suitable for supporting their entry into MIS. The PINOTHING coefficient for COOP is thus likely to be negative; even the corresponding coefficient in p l m o should be negative, since a stronger experience in co-operative relations with other firms and organisations, other things being equal, may put NWO behaviour at a relative disadvantage.

Furthermore, as was said earlier, coefficients P3NOTHING and p 3 N W O will especially describe a “bandwagon” strategy. Our expectations as to this coefficient are as follows (see the considerations developed in Section 3). Some of the nearest competitors (those belonging to the same cluster of the firm) can be described as world-leaders in MSP if CLUSMSP is large and WORLDMSP is relatively small; by contrast, they are likely to be backward in MSP if CLUSMSP is small and WORLDMSP is relatively large. Accordingly, a “bandwagon” effect should be accepted if the coefficient of CLUSMSP is negative and the coefficient of WORLDMSP is positive for the NOTHING alternative; as to NWO, a similar line of reasoning might be applied. In conclusion, P3NOTHING and p 3 N W O are expected to assume the following signs:

P3,CLUSMSP

( o, P3,WORLDMSPNOTHING ) 0,

5. THE RESULTS The maximum likelihood estimates of (8) are presented in Table 3, together with the number of observations, the degrees of freedom and the log-likelihood value. It is to be recalled that the coefficients measure the impact of the variables on the incremental profit of NOTHING and NWO alternatives with respect to the baseline alternative, MSP. Let us briefly comment on the firm-specific control variables. Size ( E M P ) has a strong positive effect, according to the classical arguments; moreover there seems to be a statistically

246

M.G. Colombo, P. Gurrone, R.G. Seri

non-significant but nevertheless non-negligible effect of saturation when approaching large sizes (EMP”2);the absence of a significant second-order effect is probably due to the fact that the firms of the sample, coming from local monopolies, are medium- or large-sized. The firm’s profitability and labour productivity as measured by ROS and PROD have no effect. As to control variables capturing the effects of traditional and newer markets, the attractiveness of the traditional market (VLINES) has negative effects on a firm’s entry likelihood into the MSP business at a 95% confidence level: investments in the new interactive service market are likely to take the firm far from its core market. Moreover, high values of VLINES also seem to favour NWO entries, pointing to similarities between the traditional telecommunications activities and the multimedia hardware and video services. As to WINT and WCATV, we are aware that our proxies for the attractiveness of the multimedia markets are weak. Given such a caveat, they show the expected effects in the NWO alternative, even though with a low confidence level. P ~ , W I N ? O T H I ~ G is positive, yet statistically not significant. As was said earlier, the estimation of our model in a reduced form inhibits us from distinguishing bandwagon effects from learning by imitation. However, due to the different dynamics of learning process and “follow-the-leader” conduct, test (13) is likely to illustrate the latter especially. A “bandwagon” effect is partly accepted. As to the NOTHING alternative, with a 95% confidence level, the coefficient of CLUSMSP is negative and the coefficient of WORLDMSP is positive. As to the NWO alternative, the CLUSMSP and WORLDMSP coefficients are estimated not to be significantly different from 0. The two variables may be weak proxies for the conducts of interest, or may be correlated to omitted variables proxying for a bandwagon effect in the NWO business; alternatively, the firm may interpret its rivals’ early entries in MIS as a signal of the profitability of such a move. Coefficients of the y , , and x> variables allow us to test our research hypotheses empirically, as specified by Section 4. Estimates reported by Table 3 yield the following results. Hypothesis 1. Historical capabilities. According to (13), DIG, proxying for technological knowledge in the traditional industry, should impact positively on NWO and NOTHING as compared to MIS. A positive yet not significant effect has been estimated in the NWO alternative, while the NOTHING coefficient is even lower: while leaders in the old technological order seem to be neither advantaged nor disadvantaged in MIS, the view of technological capabilities as strongly inertial has not been given empirical evidence. In addition, we argued that the management and marketing expertise developed in view of the more sophisticated telecommunications markets could easily be enlarged to support the MSP activities. The estimated coefficients for TR are negative at a 90% confidence level in the NOTHING equation and not significantly different from 0 in the NWO equation; the effects of CATV show an unexpected positive sign, yet not significantly different from 0 (at the standard confidence levels); only one of the two estimated coefficients for INT shows the “right” sign, but, again, neither is significantly different from 0.

In sum, evidence provided to the hypothesis that the whole set of marketing and management capabilities developed for serving a more advanced telecommunications market may adapt to support entries into MIS is mixed. Moreover, while a competence-enhancing nature of the MSP business with respect to the technological knowledge generated inside the telecommunications industry can be rejected, more evidence is needed for concluding about the competence-destroying aspects of the new paradigm. Such results on historical capabilities might be due to the fact that our variables only partly proxy the capabilities of

14. Firms' Heterogeneity and Dynamics of Entry in a New Sector

247

interest and much caution is needed in interpreting them; more appropriate variables are needed to support a more conclusive discussion on the relationship between the MSP paradigm and historical capabilities of the telecommunications firms. Table 3. Estimates of Model (8)

Usable Observations 385 Degrees of Freedom Function Value -239.98 Variable Coeff Std Error T-Stat Incremental profits of NOTHING with respect to MSP CONST 5.946795 2.449286 2.42797 -2.52525 0.0130 17 EMP -0.03287 1.21903 5.24E-05 EMP"2 6.38E-05 -1.26213 2.18565 ROS -2.75857 -0.40987 0.007646 PROD -0.00313 0.10433 1.007036 0.105066 DIG Y,f 0.4305 1 0.000907 0.00039 CELL 1.977023 1.1084 CATV 2.19133 1.20755 97.31966 117.5188 INT 0.001956 -0.00348 - 1.78004 TR 0.30132 2.839265 RDTOT 0.855525 0.21655 -0.00322 -0.01487 INITMSP -4.4 1914 0.10244 -0.4527 COOP 0.008849 -0.01839 -2.07867 Y and E. CLUSMSP 0.004704 WORLDMSP 0.009581 2.03697 6.994 153 1.96854 13.76824 xi: VLINES 37.86813 -39.8276 - 1.05174 WCATV WINT 0.199049 0.923601 0.21551 Incremental profits of NWO with respect to MSP 2.535804 -1.60438 CONST -4.06839 0.01 1803 0.84501 EMP 0.009974 EMP"2 -3.2E-05 4.15E-05 -0.76667 2.332241 ROS 1.0251 0.43953 PROD -0.00233 0.007841 -0.29674 0.963835 1.15539 DIG 1.113604 0.000365 -0.37898 CELL -0.00014 CATV 2.135222 2.174547 0.98192 -1.0342 -85.945 83.10276 INT 0.001543 -0.29606 TR -0.00046 2.823998 -0.06083 RDTOT -0.17178 INITMSP 0.027455 0.18809 0.14597 0.092589 -2.32206 COOP -0.215 CLUSMSP -0.00011 0.008308 -0.01369 Y and XE WORLDMSP 0.005098 0.004553 11 1966 VLINES 16.27131 6.804949 2.391 1 xi: WCATV 55.35995 41.54132 1.33265 WINT -1.49028 1.053236 -1.41496 if

if

349

Signif 0.015 184 0.011561 0.222832 0.206903 0.6819 0.9 16906 0.666825 0.26769 0.227219 0.07507 0.76317 1 0.988135 0.000001

0.037648 0.041653 0.049006 0.2929 17 0.829366 0.10863 0.398106 0.443279 0.660274 0.766667 0.24793 1 0.704706 0.326141 0.30 I042 0.767185 0.95 1494 0.883946 0.02023 0.989081 0.262858 0.016798 0.182647 0.157082

M.G. Colombo, P. Garrone, R.G. Seri

248

Hypothesis 2 . Learning devices: Learning by search and learning by doing. While hypothesis (1 1) posited a positive effect of RDTOT as to NWO, the estimated coefficient P l , R D T o p W O is statistically not different from 0. Our hypothesis is not empirically accepted: a firm’s involvement in R&D seems to imply a higher propensity towards neither MSP nor NWO. Hypothesis 2. Learning devices: Learning by interacting. Our hypotheses (1 2) are accepted on empirical grounds, since both the P1,coopi coefficients are negative (99% and 95% confidence level, respectively). The experience in co-operation as measured by COOP has a strong positive effect on the probability of entering the multimedia sector as a multimedia service provider: having interacted with other subjects allows the firm to act as a system integrator, by co-operating with service providers, content providers, software firms, and so on. Moreover, P1,CoopNWOis negative (Le. cooperation seems to give the firm a peculiar channel to access competencies suitable for the interactive service provision), and smaller than P , , C ~ ~ ~ N O T H in I Nabsolute G value (Le. low-medium values for COOP may result in a NWO strategy). The managerial capabilities built in by the firm through a history of cooperative relations are shown to favour the definition and exploitation of a set of external access channels to the relevant knowledge. In sum, a network of alliances seems to be an effective learning device in supporting the telecommunications firms’ early entry trials into the MSP business.

6. CONCLUSIONS This paper has empirically addressed the evolution of a firm’s capabilities and strategies in a rapidly changing technological, competitive and institutional environment. The entry strategies of telecommunications operators in the multimedia service business have been analysed to gain early evidence both of the direct impact of environmental changes, as market signals and bandwagon effects, on a firm’s conduct, and of the relative effectiveness of the different learning devices in updating the internal routines and coping with the opportunities provided by a new breakthrough. The paper has argued that the multimedia service business is still living in an era of technological and institutional variation, and that the nature of the multimedia breakthrough is unlikely to be “competence-enhancing’’ with respect to most of the distinctive capabilities of the telecommunications carriers (i.e. technological ones). Our central hypothesis is that, under such circumstances, interactive learning through alliances is a very effective capabilitybuilding mechanism for telecommunications carriers willing to enter multimedia service provision, while entirely internal learning devices are likely to perform rather poorly. As to the econometric model, after having related the firm’s capabilities to both the observed firm-specific historical variables and internal learning activities, strategies, profits and capabilities have been represented in a discrete-choice model: the telecommunications carriers can either offer multimedia interactive services, or resort to a multimedia network operation strategy, or stay in the traditional telecommunications industry. A panel of data has allowed us to estimate a reduced form of multinomial logit model. The empirical evidence supports the view that co-operation is an effective learning device in supporting the telecommunications firms’ entry trials into the multimedia service provision

14. Firms’ Heterogeneity and Dynumics of Entry in a New Sector

249

business. On the contrary, in the current phase of the new business, imitation of the competitors, learning by doing, and research and development do not allow carriers to suitably adapt their capabilities. As to the direct effect of the context on the strategy-making process, enduring perspectives in the traditional telecommunications industries are likely to keep telecommunications operators near to their core market; moreover, a “bandwagon” effect is partly accepted. As to the role of historical capabilities, it should be pointed out that leaders in the old technological order are not advantaged in service provision; the competence-enhancing nature of the multimedia service provision with respect to technological knowledge can be rejected. However, more evidence is needed for concluding about the competence-destroying aspects of the new paradigm.

REFERENCES Anderson, P. and Tushman, M.L. (1990), “Technological Discontinuities and Dominant Design: A Cyclical Model of Technological Change”, Administrative Science Quarterly, Vol. 35, pp. 604633. Arthur, W.B. (1989), “Competing Technologies, Increasing Returns and Lock-in by Historical Events”, Economic Journal, Vol. 99, pp. 161-131. Barnett, W.P. and Burgelman, R.A. (1996), “Evolutionary Perspectives on Strategy”, Strategic Management Journal, Vol. 17, pp. 5-21. Brooks, R.D., Fry, T.R.L. and Harris, M.N. (1998), “The Size and Power Properties of Combining Choice set Partition Tests for the IIA Property in the Logit Model”, Working Paper, Royal Melbourne Institute of Technology. Colombo, M.G. (1998a), “Some Introductory Reflections”, in Colombo, M.G. (ed.), The Changing Boundaries of the Firm, Routledge, London and New York. Colombo, M.G. (1998b), “The Organisational Form of Strategic Alliances. Transaction Costs Economics and Beyond”, DRUID Summer Conference, Bornholm, Denmark. Colombo, M.G. and Garrone, P. (1997), “The Multimedia Paradigm: An Evolutionary Approach”, Strategies and Communications, Vol. 28, pp. 217-242. Colombo, M.G. and Garrone, P. (1998), “Common Carriers’ Entry into Multimedia Services”, Information Economics and Policy, Vol. 10, pp. 79-98. Dosi, G. (1988), “Sources, Procedures, and Microeconomic Effects of Innovation”, Journal of Economic Literature, Vol. 26, pp. 1120-1171. Fokianos, K. and Kedem, B. (1998), “Prediction and Classification of Non-stationary Categorical Time Series”, Working Paper, Ohio State University. Garrone, P. and Rossini, A. (1998), “The Role of Technology and Product Capabilities in a New High-Tech Business: The Case of Cellular Services”, Journal of High-Technology Management Research, Vol. 9, pp. 285-310. Hannan, M.T. and Carrol, G.R. (1993, “An Introduction to Organizational Ecology”, in Carrol, G.R. Hannan, M.T. (eds.), Organizations in Industry, Oxford University Press, New York and Oxford.

250

M.G. Colombo, P. Gurrone, R.G. Seri

Haveman, H.A. (1993), “Follow the Leader: Mimetic Isomorphism and Entry into New Markets”, Administrative Science Quarterly, Vol. 38, pp. 593-627. Henderson, R.M. and Clark, K.B. ( I 990), “Architectural Innovation: The Reconfiguration of Existing Product Technologies and the Failure of Established Firms”, Administrative Science Quarterly, Vol. 35, pp. 9-30. Henderson, R.M. and Mitchell, W. (1997), “The Interaction of Organizational and Competitive Influences on Strategy and Performance”, Strategic Management Journal, Vol. 18, pp. 5-16. Kaufmann, H. ( 1 987), “Regression Models for Nonstationary Categorical Time Series: Asymptotic Estimation theory”, The Annals of Statistics, Vol. 15, pp. 79-98. Lundvall, B.A. (1988), “Innovation as an Interactive Process: From User-Producer Interaction to the National System of Innovation”, in Dosi, G. et al., (eds.) Technical Change and Economic Theory, Pinter Publishers, London and New York. McFadden, D.L. (1981), “Econometric Models of Probabilistic Choice”, in Manski, C. and McFadden, D.L. (eds.), Structural Analysis of Discrete Data with Econometric Applications, MIT Press, Cambridge. McFadden, D.L., Train K.E. and Tye, W. (1981), “An Application of Diagnostic Tests for the Independence of Irrelevant Alternatives Property of the Multinomial Logit Model”, Transportation Research Record, Vol. 637, pp. 39-46. MethC, D., Swaminathan, A. and Mitchell, W. (1996), “The Underemphasized Role of Established Firms as the Sources of Major Innovations”, Industrial and Corporate Change, Vol. 5, pp. 11811203. Mitchell, W. ( 1 989), “Whether and When? Probability and Timing of Incumbents’ Entry into Emerging Subfields”, Administrative Science Quarterly, Vol. 34, pp. 208-230. Nooteboom, B. (1992), “Towards a Dynamic Theory of Transactions”, Journal of Evolutionary Economics, Vol. 2, pp. 281-299. Prahalad, C.K. and Hamel, G. (1990), “The Core Competence of the Corporation”, Haward Business Review, May-June, pp. 79-91. Teece, D.J., Rumelt, R., Dosi, G. and Winter, S. (1994), “Understanding Corporate Coherence. Theory and Evidence”, Journal of Economic Behavior and Organization, Vol. 23, pp. 1-30. Thurstone, L.L. (1927), “A Law of Comparative Judgement”, Psychological Review, Vol. 34, pp. 272286.

14. Firms’ Heterogeneity and Dynamics of Entr)]in a New Sector

251

APPENDIX A To test the correct specification of the multinomial logit model (S), the McFadden, Train and Tye (1981) test has been performed. Even if the test is undersized, i.e. badly biased towards accepting the null hypothesis (it accepts the hypothesis of correct specification more than desired), Monte Carlo studies (Brooks, Fry, Harris (1998)) show that this test “has very high power despite having an effective size of zero” and recommend the use of this test. It appears to be, therefore, the best test, if large deviations from the null hypothesis of the multinomial logit model should be detected. As all of the choice partition tests, it is based on the following property of the multinomial logit model, the so-called Independence of Irrelevant Alternatives. In a J alternatives multinomial logit, the probability of an alternative is given by: P(Y, = o)=

1

1 + k e x p b k .x)’ k=2

k=2

therefore, we have:

It is straightforward to note that this expression does not depend on the presence of alternatives different from j , and j , ; as a consequence, the so-called Independence of Irrelevant Alternatives (IIA) holds (under IIA, the addition or the subtraction of alternatives to the choice set does not change the quotient of the probabilities of two choices). Therefore the estimated parameters should not be very different if we eliminate an alternative (or a group of alternatives) from the sample, together with the individuals that have chosen these alternatives. This test is essentially a modification of a likelihood ratio test and consists of the following procedure:

I . Calculate the parameters

p

of the complete model;

pR

2 . Calculate the parameters of a binary model obtained by eliminating all the alternatives but two (model with restriction of the choice set) and obtain the value of the log-likelihood functionlb,).

3. Substitute the correct values of the log-likelihood function

p,

IbF).

in the model of the previous point and obtain the value of

M.G. Colombo, P. Garrone, R.G. Seri

252

The value 2 .

kbR)-

is asymptotically distributed as a

x;, where p is the number of pa-

rameters of the model with restriction of the choice set. Table A. 1. Results of the IIA Test.

OBEs

Full model

309

-88.88463

M S P and NWO

202

-87.16390

NOTHING and

259

NOTHING and

Restricted 2 [ @ R ) - l @ r ) ] model -86.94396 3.8813

DoF

Lambda

Prob

I8

28.86932

0.99980922

MSP -86.55755

1.2127

18

28.86932

0.99999998

-109.05160 -107.23928

3.6246

18

28.86932

0.99988458

NWO

From Table A. 1, it is clear that, according to this test, the multinomial logit specification appears to be highly suitable for our purpose since it is accepted by all of the choice partition sets at a critical value of 99.98%.

0 2000 Elsevier Science B.V. All rights reserved Convergence in Communications and Beyond E. Bohlin, K. Brodin, A. Lundgren and B. Thorngren (Editors)

CHAPTER 15

Industrial Organizational Implications of Electronic Commerce Steven Globermanr Centerfor International Business, Western Washington University, Bellingham, Washington

Abstract. This paper considers the potential impacts of the emergence and growth of electronic commerce on industrial organizational attributes such as ease-of-entry, seller concentration and average firm size. While much of the relevant literature argues that electronic commerce will profoundly change the industrial organizational attributes of a range of markets, primarily in the direction of making those markets more contestable, this paper offers a more guarded and qualified appraisal. Specifically, electronic commerce will generally reduce barriers to entry, seller concentration and average firm size; however, these effects are likely to be modest for many products, especially for those that are either “experience” goods or “credence” goods. Anti-trust laws will not be rendered obsolete by electronic commerce; however, antitrust authorities may need to reorder their priorities.

“History shows that every time a new universal communication path has opened it has led to profound change in the way commerce is done. Clipper ships, railroads, telephones, broadcasting - and now the Internet.”‘

1. INTRODUCTION The emergence and growth of electronic commerce has received a substantial amount of attention in the popular press, as well as among policy-makers. Among other potential consequences that have been discussed is the potential for electronic commerce to substantially change traditional organizational arrangements in many industries. Among other things, numerous experts have discussed the potential for a substantial contraction of intermediation services in many industries, as well as dramatic changes in the optimal size and scope of firms. Such changes, if they emerged, could significantly alter traditional modes of competi-

* The author thanks Professor Charles Steinfield for helpful comments on an earlier draft. He also thanks an anonymous reviewer for useful suggestions. A quote given by Irving Wladawsky-Berger, general manager, Internet division, IBM as cited in Business Week, 23 March 1998, Special Advertising Section, p. 1.

S.Globerman

254

ticn in the industries affected, as well as the net benefits of alternative private sector business strategies and public sector industrial policies including anti-trust policies. The purpose of this paper is to review and assess the potential linkages between electronic commerce and industrial organizational characteristics. The analysis blends both theoretical and empirical insights with a particular focus on the relevance of electronic commerce to government competition policies. The paper proceeds as follows. Section 2 provides a brief description of electronic commerce, its current status and predictions of future growth. Section 3 identifies and discusses hypotheses regarding the potential linkages between electronic commerce and industrial organizational characteristics. Section 4 evaluates the hypotheses identified. Section 5 discusses some policy implications of our analysis.

2.

DESCRIPTION OF ELECTRONIC COMMERCE

2.1 Definition of Electronic Commerce The term electronic commerce is used in association with many events in modern telecommunications and information technology. The most encompassing definition of electronic commerce would include any form of business activity conducted on the electronic medium (Wigand, 1997). This would include electronic data interchange (EDI), electronic mail and related types of communication. While ED1 has been explicitly equated with electronic commerce in the past, most experts consider it more properly a subset of electronic commerce. ED1 is the communication of business information via computers, usually for the purpose of ordering, invoicing and sending contracts or other pertinent commercial information. To this extent, ED1 can be viewed as a set of value-added services comprising a subset of electronic commercial transactions.* Electronic mail might be viewed similarly. Non-commercial transactions conducted electronically would not qualify as electronic commerce by most standards, although, as a practical matter, the margin between commercial and non-commercial activity is vague. Typically, a set of commercial transactions must be carried out over the Internet/World Wide Web to be considered electronic commerce. While many commercial transactions are carried out on private electronic networks, the main hypotheses linking electronic commerce to industrial organizational changes focus on “public access” networks, of which the Internetmorld Wide Web is the dominant model.

2.2 The Magnitude of Electronic Commerce Systematic data on the use of electronic commerce are unavailable. What one finds instead, and then primarily for the United States, is ad hoc observations and estimates about utilization, often for specific applications. In order to summarize the relatively disparate estimates of the volume of electronic commerce Table 1 identifies the source of the estimate, as well as a quantitative or qualitative evaluation of the level of activity, either in a specific sector or more generally. What is clear from the limited information summarized in Table 1 is that electronic commerce as defined above has only barely begun to be implemented, but it is growing quite rapidly. In particular, retail purchases using the Internet are growing quite rapidly. By contrast,

’ This is not to say that “business-to-business” electronic commerce has the same commercial implications as “business-to-retail consumer” electronic commerce.

15 lndustrtal Organizational Implications of Electronic Commerce

2 55

EDI, which has a much earlier initial date of introduction than do other electronic commerce applications, seems to have diffused at a relatively slow pace. For example, it is estimated that around 3,000 US firms were using ED1 in 1986. This number had gone up to only around 100,000 in 1994 (Pfeiffer, 1995). In 1988, the number of European (including UK) firms using ED1 was 2,350. The number increased to 20,000 in 1994. Pfeiffer (1995, p. 112) notes that with millions of business organizations as potential ED1 users, these statistics indicate the still rather embryonic stage of EDI utilization in practice, and he does not expect this situation to change significantly anytime soon. The embryonic stage of electronic commerce illustrated by the information summarized in Table 1 is further underscored by Auger and Gallaugher’s ( 1997) survey of a sample of small and medium-sized businesses in October 1995. Of the approximately 140 firms that responded to the questionnaire, around 90% were based in the United States. Most of the business Web sites identified in the survey had been in operation for one year or less. The majority of products sold were consumer tangible goods. Average sales were only around $1.2 million, and less than 12% of the Web sites had encrypted credit card capabilities. The two most popular ordering methods were telephone (83%) and postal mail (78%). Most of the items sold on the Internet were relatively low-price (less than $50), and more than half of the respondents (54%) had less than 5% of their total sales coming from the Internet. Table 1. Estimates of Current Electronic Commerce Activity Author

Comments

1. Auger and Gallaugher (1997)

Over half of the Fortune 500 companies have some sort of Web presence. Over half of these firms are “just playing or dabbling”.

2. Phelan (1996)

Sales originating from the Internet often account for less than 1 % of total company sales.

3. Schecter (1997)

On-line sales of pre-recorded music comprise a very small fraction of the market.

4. The Economist (1997)

In 1996, all consumer transactions were in the range of $500-600 million. Of the Fortune 500, only 5% were con-

ducting transactions on the Internet. 5 . Wigand (1997)

Over 45,000 firms in the US exchange data electronically.

6. Dah1 and Lesnick (1996)

The general public has virtually no idea as to the use of the electronic payment systems.

7. Rowan (1 997)

Around 8% of surveyed Canadians used the Internet for shopping.

8. Anders (1998)

Direct sales to consumers on the Internet will total $7.2 billion in 1998.

Source: Author’s review of literature.

S. Globerman

256

More recent estimates of the magnitude of electronic commerce activities continue to underscore their relatively small size and focused scope, as well as their rapid rate of growth. In this regard, Table 2 reports estimates of Internet shopping dollar values in different product categories for 1997 along with projections for the year 2000. The estimated dollar value for 1997 is around $2.4 billion.’ The value is expected to jump to around $12 billion at the turn of the m i l l e n n i ~ mWhile .~ the latter estimate might be considered relatively optimistic, it can be put into perspective by considering that total retail sales in the United States in 1996 were approximately $204 billion. It is also clear from Table 2 that on-line computer hardware and software sales comprise a disproportionate share of on-line purchases, with travel expected to comprise a disproportionate share in the future. Computer electronics is also a relatively large sector for business-to-business commerce on the Internet. Specifically, it has been estimated that computer electronics accounted for almost half of total business-to-business Internet commerce in 1997. Motor vehicles and petrochemicals are industry sectors in which such commerce is expected to grow rapidly in the first decade of the new millennium.’ Table 2. Estimates of Internet Shopping (millions of dollars) 1997

2000

PC Hardware and Software

863

2,901

Travel Entertainment

654

4,741

Books and Music

298 156

1,921 76 1

Gifts, Flowers and Greetings Apparel and Fashion

149 92

591 361

Food and Beverage

90

354

Jewelry

38

107

Sporting Goods

20 19

63 93

Consumer Electronics Other

65 197 TOTAL $2.444. $12,090. Source: “The Virtual Mall Gets Real,” Business W e e k , 26 January, 1998, pp. 90-1.

’ More recently, it has been estimated that goods and services sold on-line to US and European consumers will top $5.1 billion in 1998. See “The Click Here Economy,” Business Week, 22 June, 1998, p. 122. The same source estimates that US businesses will exchange $17 billion in goods and services in 1998. See ibid., p. 124. A relevant qualification is that on-line payments are still relatively rare. See “Internet Commerce Booms”, The Bellingham Herald, 22 November, 1998, E2. Forrester Research Inc. predicts global electronic commerce revenue will reach $3.2 trillion by 2003. See “Why They’re Nuts About The Net”, Business Week, 23 November, 1998, p. 52. Given the dynamic nature of commercial transactions on the Internet, a forecast through 2003 is highly speculative.

’ See Anders (1998).

IS.Industrial Organizational Implications of Electronic Commerce

257

Table 3. Some Projections of Future Electronic Commerce Activity Author

Projections

1. Auger and Gallaugher (1997)

Forrester Research found that $240 million in sales were conducted through interactive media in 1997. On-line sales are expected to surpass $5 billion by the end of this century.

2. Sims (1995)

32% of Web users have already bought products over the Internet and 91% plan to do so in the future.

3. Wigand (1997)

McKinsey and other consulting firms have predicted that there will be a $4-5 billion home shopping market in the US in 2003.

4. Cronin (1996)

Chrysler, which puts its Internet sales in 1996 at 1.5% of total sales, projects that by the end of the century this figure will be 25%.

5 . Schecter (1997)

Internet research group Jupiter Communications estimates that music sales over the Internet will hit $1.6 billion by 2002, representing 7.5% of the US market.

6. Coy (1998)

Electronic commerce will reach about $350 billion by 2002 from an estimated $22 billion in 1998, including direct sales to consumers and transactions between businesses.

Source: Author’s review of literature.

A summary of some additional forecasts of Internet sales provided in Table 3 suggests that the $10 billion estimate by the year 2000 is at the high end of estimates. Nevertheless, the various estimates of the future growth of electronic commerce suggest that it might well openup non-trivial markets, at the margin, for a number of commercial activities. The issue the paper now turns to is how the industrial organization of these electronic markets is likely to differ from the industrial organization of more conventional markets.

3.

ELECTRONIC COMMERCE AND INDUSTRIAL ORGANIZATION: SOME HYPOTHESES

Phelan (1996) among others notes that a major characteristic of electronic commerce is the reduction of transaction costs. The latter, in turn, are now well recognized to exert a strong influence on the governance structures and the nature of competition in a range of economic activities.

258

S. Globerman

3.1 The Nature of Transaction Costs The costs of transacting are essentially comprised of the following: (1) search costs: the costs of physically searching for products, sellers and buyers:

(2) contracting costs: the costs of setting up and carrying out contracts; (3) monitoring costs: the costs of ensuring that the terms of the contracts are met; (4) adaptation costs: the costs incurred in making changes to contracts over time.6

Most of the focus of the discussion of the impact of electronic commerce has been on the linkage to search costs. Specifically, the Internet has been highlighted for its ability to offer rich detail and specificity regarding the availability and prices of different goods and services. Moreover, since consumers have greater control over the Internet search process than they do over other media, such as television and radio, the cost of gathering this information should decline, and, as a result, markets for the affected products should become more efficient (Hoffman and Novack, 1997). In this context, one would expect that the products most significantly impacted by the emergence of electronic commerce would be so-called search goods. Economists identify a product as having “search qualities” if the consumer can establish a product’s quality prior to the purchase by inspection. Within this broad definition, inspection includes visual and tactile inspection. Computer equipment is an obvious example of a search good, as technical specifications are quite meaningful and easily communicated to potential buyers. The large share of electronic commerce accounted for by this product category is, therefore, unsurprising. Securities traded on the Internet are also search goods, as the companies involved are usually listed on major stock exchanges. If the consumer must utilize the product to determine its quality, the good is said to have “experience” q ~ a l i t i e sAn . ~ example of an experience good is the automobile. The subjective characteristics of the product, such as “road feel”, steering responsiveness and so forth, can best be established by actually driving the car. Economists also talk about “credence” goods whose qualities may not be readily determined even after consumption. An example of the latter is medical care. Many products can be downloaded to consumers, thereby allowing them to try electronic “free samples”. Obvious examples include software, including software games, music, books, magazines and financial advice and related services. Hence, the Internet in its present form might emerge as a viable medium for selling these types of experience goods.S Specifically, the Internet may make it more convenient and less costly for consumers to identify and evaluate a range of experience goods. For experience goods that cannot be readily “sampled” elec-

‘This categorization of transaction costs is discussed in Wigand, op.cit., p. 8. A component of search activity is the verification of the claimed attributes of products. Where it is difficult for producers to “validate” their product claims, markets may be characterized by a “lemons” problem, and reliable producers may be driven from the market. For a discussion of this phenomenon on the Internet, see Lu (1998).

’ A brief discussion of these distinctions can be found in Carlton and Perloff (1990, pp. 596-98). ’ Indeed, the startling growth of on-line bookseller Amazon.Com attests to the relative ease of purchasing books through the Internet.

15. Industrial Organizafional Implications of Electronic Commerce

259

tronically, producers can attempt to ensure quality in other ways, such as by investing in brand names, by offering product satisfaction warranties and so forth. However, it is unclear how the Internet, p e r se, will affect the costs that producers need to incur in order to create the “trust capital” needed to make their quality claims credible to potential consumers. Traditionally, large accumulated sunk costs in brand names and trademarks have been a major “hostage” that firms have made available to potential consumers in order to create trust (Klein and Leffler, 1981). To the extent that the Internet allows firms to lower (or avoid) the sunk cost investments that have been traditionally required to market and promote experience goods, consumers may become even more concerned about buying ‘‘lemons’’ through the Internet, and the marketing of experience goods may not be strongly affected by the emergence of this new distribution medium.’ It is also technically possible for credence goods to be sold over the Internet. For example, there are psychologists who are selling their services on-line to clients, primarily through electronic mail, and the emergence of relatively low-cost video conferencing is allowing an increasing number of psychologists and other health care workers to have “face-to-face” consultations with patients. However, where the long-term effectiveness of such treatments is uncertain to the patient, even after some accumulated experience, the individual’s purchase of credence goods is likely to be strongly guided by recommendations from contacts including family, close friends and other professionals. In this context, it is unclear that the Internet will substantially reduce the costs associated with search for credence goods, since information on the Internet about the availability of suppliers may be strongly discounted by prospective buyers. In effect, for many individuals, reliable information about credence goods will remain the recommendations of members of valued reference groups who are trusted because they have a history of satisfactory relational exchanges with those individuals.” Thus, it is likely to be “traditional” search goods whose transaction costs will be most significantly impacted by electronic commerce. Table 4 generally confirms that search goods and relatively inexpensive experience goods that can be electronically sampled constitute the majority of products that have been purchased on the Internet, to date. It also suggests that, as consumers become more confident about the security of Internet payment systems and the ability of Internet software providers to limit the establishment of fraudulent web sites, more expensive search goods, such as real estate and luxury consumer items, will become more prominently traded, or at least identified for purchase on the Internet. There has been less discussion and much less consensus surrounding the impact of electronic commerce on other types of transaction costs. For example, it can be argued that the widespread adoption of standardized ED1 (electronic) contracts would lower the average costs of setting up contracts, since a repetitive activity with relatively high variable costs would be replaced by a “once-and-for-all’’ effort with relatively high fixed (and sunk) costs but relatively low variable costs. However, besides the unresolved legal issues surrounding electronic contracts, it is unclear that transactions between parties, even those who regularly do business with each other, are sufficiently standardized that the need for contract modifications and extensions on an ongoing, and perhaps unpredictable basis, will be obviated.

’ This point is also made in Lu (1998). In the next section of the paper, we will discuss differences in costs between Internet marketing and marketing through more traditional channels. lo As an illustration, the vast amount of on-line medical data appears to be used by patients primarily to bring ideas and questions to their existing physicians rather than as a source of new physician services. See Hafner (1998).

S. Globerman

260

Table 4. Current and Future Internet Best-Selling Categories (By Value) 1995

Future

Real estate

Computer-related hardware, software

Computers and accessories

High-tech equipment

Software

Travel services

Travel

Investments

Audio and consumer electronics

Training

Financial services

Boilerplate legal services Prestige real estate Male-oriented luxury goods Specialty hobbies and craft Technical employment services Audio recordings and consumer electronics Industrial equipment and parts

Source: Phelan (1996) This caveat is particularly relevant to the extent that electronic commerce results in the expansion of relevant geographical markets, since differences in legal regimes, contractual customs and so forth may oblige firms to enter into multiple bilateral agreements with a resulting loss of standardization. To the extent that many contracts remain activity-specific or customer-specific, there may be no compelling cost advantages to contracting electronically. Virtually nothing has been written about the potential linkages between monitoring and adaptation costs and the implementation of electronic commerce. Again, to the extent that exit costs are lower in electronic businesses than in conventional businesses, the risks of opportunism will be perceived to be higher, all other things being the same, and firms may see a need to monitor their electronic transaction partners more closely than they do their transaction partners who use conventional technology. On the other hand, the Internet is a relatively low-cost vehicle for spreading notoriety about opportunistic behavior. One can imagine, for example, industry trade associations maintaining electronic bulletin boards in which members can register opinions about the quality of the services they have received from specific suppliers. To the extent that perceived risks of opportunistic behavior are higher for electronic commercial activities, underlying contracts may have to be more complete and complex which, in turn, makes it more likely that they will have to be adapted more frequently over time. The associated costs of maintaining contracts in force, including associated costs such as litigation, may therefore also be higher for electronic commercial transactions. On the other hand, if the use of electronic commerce expands the geographical scope of relevant product markets, as will be argued below, buyers and sellers should have increased options to deal with other trading partners, thereby reducing the costs associated with switching trading partners. Lower

15. Industrial Organizational Implications of Electronic Commerce

261

switching costs, in turn, should reduce incentives to act opportunistically, all other things being constant, which would reduce the need for comprehensive contracting.” In summary, the main and relatively unequivocal impact of electronic commerce would seem to be on search costs. Specifically, the costs of searching for prices and other market attributes are allegedly lowered substantially through use of the Internet. The implications of this development are discussed next.

3.2 Potential Implications of Electronic Commerce As it becomes less costly to search for information about prices and other attributes of potential transactions, including potential transactors, electronic commerce should lead to increased integration of relevant markets. This is likely to be particularly true for geographical markets, since it is relatively costly for (especially) retail shoppers to compare and contrast product prices (and other search attributes) across physically separated geographical markets. Since local markets are usually more concentrated than national (or international) markets for most products, the adoption and spread of electronic commerce activities could lead to significant reductions in structural measures of industrial concentration. ’* The reduction in search costs may also lead to increased vertical integration in affected industries. The term that tends to be used in the relevant literature is “disintermediation”. Specifically, as it presumably becomes easier for buyers to identify preferred sellers and vice versa, the economic incentives for intermediation with attendant economies of scale are reduced.” Participants in markets will be less willing to pay agents to create and expand markets when information about potential supply and demand is more widely disseminated and less subject to cost indivisibilities associated with its collection and dis~emination.’~ Electronic commerce is also suggested to contribute to increased competition in relevant markets by lowering barriers to entry, particularly for small firms. In particular, electronic commerce allegedly reduces the necessary sunk cost investments associated with more traditional distribution channels. As a case in point, Solomon (1995) suggests that it costs as little as US $1,000 a year to open an “electronic storefront” on the Internet which is currently accessible by as many as 20 million people; however, a problem is gaining visibility on the Internet. Information web sites are fragmented and numerous, so that access to a web site in a technical sense is not equivalent to access in an economic sense. In order to gain easier access to individuals browsing the Internet, many on-line merchants have begun using the heavily trafficked net search engine sites (the most frequently visited sites) as a springboard to their sites. While several search engines have recently introduced shopping tools designed to search according to price or product, they still include only a few hundred merchants. Technology that can scour the entire World Wide Web and the thousands of on-line merchants is still not

” An early and seminal discussion of the environmental determinants of opportunistic behavior is provided in Williamson (1975).

’* For an enthusiastic statement of how electronic commerce spells the end of geography and borders as industrial organizational constructs, see Kobrin (1995). 13 For a discussion of electronic commerce and intermediation, see Picot, Bortenlanger and Rohrl (1997). Lu (1998) highlights the importance of the nature of.the relevant intermediation activity. 14

For example, the rise of electronic stock trading has been implicated in the recent decline in the price of seats on major financial exchanges. See Barboza (1998).

262

S. Globerman

abailable.’’ As well, shoppers remain reluctant to authorize on-line payments. Hence, many transactions are still consummated through traditional distribution channels such as mail catalogues, even though consumers may have identified the product through the Internet. In summary, electronic commerce is seen by some observers to be a boon to competition and, especially, to the entry and growth of small firms. Markets are expected to become more atomistic with dominant firm power eroded over time. Information impactedness problems, which create a demand for market intermediaries, will be mitigated, and there will be more direct dealing between buyers and sellers. Business-to-business reach should also be broadened, thereby enabling increased competition in input bidding processes. In short, electronic commerce can be seen as contributing to a re-emergence of the attributes of “classical” competitive markets on (potentially) a global scale. To the extent that this vision proves accurate, it will have profound effects on private and public sector business strategies. In particular, much of the re-emerging activist philosophy on antitrust policy might have to be rethought.

4.

EVALUATING THE IMPLICATIONS

The hypotheses outlined in the preceding section suggest that the adoption and spread of electronic commerce applications will move product markets closer to the classical ideal of perfect competition with resulting and substantial gains in consumer welfare. The thinking in this regard is that electronic commerce will erode the market power of dominant firms by facilitating entry on the part of small, emerging enterprises, primarily because the Internet’s relatively low-cost service and its non-hierarchical architecture lower barriers to entry associated with high marketing, promotion and selling costs using “conventional” channels.

4.1 Market Contestability Some observers caution that the industrial organizational changes hypothesized by many are exaggerated in magnitude, if not entirely unlikely to materialize in the foreseeable future. For example, Garcia (1997) argues that visibility rather than access is at a premium on the Internet. Given the Internet’s loose management structure and its exponential growth, the major problem facing small firms on the Internet is making themselves known and differentiating themselves from others. As noted above, in seeking to promote their availability and offerings on the World Wide Web, merchants are paying millions of dollars for top billing on popular sites such as Yahoo! In a similar vein, the main on-line sports web sites are attempting to increase viewers by cross-promoting with “main events” featured on other media, such as sister cable stations. This observation suggests that substantial “up-front” costs for entrants and smaller firms seeking to expand may still be substantial in the world of electronic commerce. Notwithstanding the preceding caveat, smaller firms apparently do feel that the emergence of electronic commerce has helped more than hurt their abilities to compete. For example, the relatively rapid and recent proliferation of small, high-growth companies in Europe has been credited, in part, to the spread of the Internet which helps puts these companies in touch with new customers.16 In a related vein, early surveys of companies involved in Internet-facilitated l5 16

See “The Virtual Mall Gets Real,” Business Week, 26 January, 1998, pp. 90-1.

Other factors cited include a growing trend toward deregulation on the Continent and the emergence of secondary stock markets. See “Startups To The Rescue,” Business Week, 2 March 3 , 1998, p. 50.

15. Industrial Organizational Implications of Electronic Commerce

2 63

electronic commcrce revealed that company size had little or no bearing upon success on the Internet, implying that success on-line does not necessarily result from benefits such as scale economies and a large staff (Auger and Gallaugher, 1997).” In this regard, the rash of mergers in sectors likely to be affected by electronic commerce, for example financial services, seems curiously at odds with the technological trend ostensibly mitigating the traditional advantages of large firm size. A possible rationalization of this seeming paradox is that many consumers prefer to deal with established brand-name products on the Internet, especially when buying from foreign suppliers.” Hence, while it may be cheaper for smaller firms to communicate information to new consumers about prices and product characteristics through electronic commerce, the need to make claims (even about the likelihood of receiving shipments paid for) may predispose many risk-averse consumers to favor brand-name suppliers on the Internet even more than in conventional marketing channels. Auger and Gallaugher (1997, p. 61) embellish upon the barriers to competition raised by the asymmetric ownership of “reputational capital” with their observation that anecdotal evidence suggests that Internet shoppers are more interested in gathering information on-line than in obtaining the lowest price.” In turn, they interpret this evidence as a suggestion that Internet activity may favor those products which do not face competition, but which have the purchase decision impeded by product complexity, lack of knowledge and other factors that can affect buyer uncertainty. While it might be argued that intermediaries such as “electronic malls” will arise to essentially provide the reputational capital that smaller retailers do not have, and would find prohibitive to acquire by investment on their own, there is no strong evidence that early electronic malls have provided any significant advantages to the small retailers affiliated with them.20 Finally, entry conditions may be affected by an enhanced ability of sellers to offer new products to consumers. To the extent that new products are substitutes for existing products, the impact will be pro-competitive. The reverse may be true if new products are complements to existing products. An example is American Airlines’ program to offer frequent fliers oneto-one marketing software. With this software, preferred customers can streamline their booking process by creating a profile of their home airport, seating and meal preferences and so forth.*’ In sum, the linkages between electronic commerce, barriers to entry, and industrial concentration are potentially complex and activity-specific. In particular, for “pure” search goods, the widely held view that electronic commerce will reduce barriers to entry for smaller sellers

” The remarkable success of Internet start-up companies such as Amazon.com is additional anecdotal testimony to the ability of small entrants to exploit the capabilities of Internet marketing.

‘*

Auger and Gallaugher (1997) interprete their survey finding that larger firms obtained a significantly greater share of their Internet sales from outside of their home countries as evidence that foreign customers prefer to buy from larger organizations. l 9 An emphasis on “price-shopping” appears particularly marked for on-line buyers of products such as autos and airline tickets. See “The Click Here Economy,” Business Week, 22 June, 1998, p. 124. However, to date, except for a few exceptions, on-line prices are not radically lower than prices in conventional stores. See “So Where Are All the Bargains?” Business Week, 22 June, 1998, p. 162. 2o

See “Point, Click - and Spend,” Business ‘Week, 15 September, 1997, pp. 74-6.

21

See “Now it’s Your Web”, Business Week, 5 October, 1998, pp.164-178.

264

S. Globerman

and lead to greater integration of geographical markets, with resulting decreases in industrial concentration and market power of incumbents, may wel! be valid, on balance: however, in the case of goods which consumers need to experience, to some extent, to be sufficiently confident of producers’ claims, the Internet may fail to be a significantly more robust marketing channel for small firms and early entrants than conventional marketing channels. Indeed, large firms with well-known international brand names, and that also distribute their products through conventional marketing channels, may be the relatively large beneficiaries of the technological properties of electronic commerce. As such, while some markets may become more contestable, others may become less so.22

4.2 Disintermediation The popular view that electronic commerce will lead to the gradual disappearance of intermediaries (between consumers and sellers) across a range of markets may also be too simplistic. The basic notion here is that the demand for specialized knowledge is not necessarily reduced by the growth of electronic commerce. Rather, the relative values of different types of specialized knowledge change. For example, knowledge about how to actually execute transactions where the parties “know” each other is being rapidly devalued by electronic commerce. An example in this regard is the stockbrokerage function. Simple executions of buy-and-sell orders are increasingly being performed on-line without the intermediation of traditional brokerage personnel.23 On the other hand, knowledge about the reliability of on-line sellers, the creditworthiness of on-line buyers, the relative quality of competing experience goods being sold on-line, and so forth, may become even more valuable with the growth of electronic commerce applications. Established intermediaries participating in conventional markets may be more or less successful in transferring their dominance to electronic markets. For example, traditional market research companies may have no competitive advantage in processing sales and marketing data on Internet transactions to develop profiles of segmented groups of Internet buyers. Indeed, success in gathering and processing Internet “traffic” data may require a different set of skills than those possessed by traditional media rating agencies such as Neilson’s. On the other hand, where successful intermediation largely draws upon embodied and idiosyncratic knowledge of specific markets, incumbent intermediaries could enjoy strong competitive advantages when acting as “electronic intermediaries”. An example here is financial intermediation such as bond rating services. While it is relatively easy to imagine Netscape or Microsoft opening up their own electronic brokerage services, presuming they could get government approval to do so, it is harder to imagine displacing either Dun and Bradstreet or Moody’s as rating agencies for corporate and government securities. As another example, the discount brokerage company, Charles Schwab, is planning to create a new niche as a source of financial information for investors who trade over the Internet. In short, the impact of electronic commerce on the extent and nature of intermediation, as well as the identities of the intermediaries likely to be more or less successful is, again, likely to be activity-specific. Entirely new intermediary functions may also emerge and grow, such 22

Early experience confirms the pro-competitive effects of the Internet for such products as discount brokerage services, wholesaling and retailing of computer equipment and retailing of books and recordings. It was recently estimated that 30% of all US retail stock transactions are done via the Internet. See “Citigroup,” Business Week, 20 April, 1998, p. 36. 21

15. Industrial Organizational Implications of Electronic Commerce

265

as the rating and “guaranteeing” of electronic money. Indeed, the emergence of new intermediary functions is almost assured by the innovative nature of electronic commerce. Specifically, one might expect the sale of new products on the Internet that, in turn, will create a demand for entirely new intermediary services.

4.3 Abuse of Dominance It has been argued that the more widespread and transparent information about prices and other conditions of sales facilitated by the Internet will reduce manifestations of anticompetitive pricing such as price discrimination and informal price collusion, by enabling buyers to more easily identify and arbitrage price differences between sellers and between geographical markets. Cross-price elasticities of demand among sellers may also increase under electronic commerce, as geographical distances between sellers and potential buyers lose their importance as means of segmenting markets. As a consequence, the temptation to cheat on informal price agreements should increase with the growth of electronic commerce. Conversely, it has been argued by Picot, Bortenlager and Rohrl (1997) that electronic mediation of market transactions will not automatically lead to reduced prices compared to conventional market organizations, since sellers can be expected to implement various market strategies to reduce market transparency in order to preserve previous price and income levels. For example, they may quote prices on the Internet as the basis for further negotiation, rather than as firm offers which will be filled if buyers meet the quoted price. In this way, some price discrimination remains possible based upon the buyer’s urgency for the product, his or her opportunity costs of the time spent haggling over the product and so forth. The use of bundled pricing and complicated charging schedules can also obscure price differences among sellers. To the extent that individual web sites become well-known for vending certain products or for “guaranteeing” quality and reliability of the merchandise sold at the sites, areas of bottleneck market power may emerge on the electronic network. Traditional competition policy issues, of whether and to what extent vertical integration into retailing by web site operators (or other suppliers of critical inputs to the Internet) enhances dominant firm market power, may arise in these cases.24 Less traditional concerns about potential anti-competitive effects of strategic alliances may also become more prominent, given indications that electronic commerce is encouraging the growth of business partnerships.

5.

CONCLUSIONS

It seems reasonable to believe that a major technological event such as the emergence and growth of electronic commerce will change the industrial organization of a wide range of markets; however, the precise nature, magnitude and timing of the changes are difficult to predict. The predominant thinking as expressed in the relevant literature is that the changes will be widespread and substantial, and that the major change will be a substantial increase in the contestability of markets, as entry barriers decrease and industrial concentration also decreases.

24

This concern is raised by Garcia (1995).

266

S. Globerman

This paper expresses a more skeptical view on the various dimensions of the issue. In particular, it suggests that the rate of diffusion of electronic commerce may be relatively slow, beyond the few applications that currently account for the bulk of retail electronic transactions. Moreover, whole categories of goods that can be classified as experience or credence goods may be particularly slow to be adopted for electronic commerce applications. This view is obviously qualified by the potential for innovations (both technological and managerial) to emerge which make the Internet a much more robust distribution channel for services and related products. The paper also cautions that increased contestability is not a certain, nor even the most likely, outcome of the growth of electronic commerce in a range of market applications. Specifically, in some cases, large incumbent firms may enjoy even greater competitive advantages in an electronic commerce environment than they do in conventional markets. In particular, the enhanced need for investments in reputational capital may impart relatively high barriers to entry into electronic commerce.2s For these and related reasons, it seems doubtful that even a relatively fast and widespread adoption of electronic commerce would significantly mitigate the potential role for competition policy in developed economies; however, such adoption could alter the priorities of competition policy authorities, as well as the degree to which monitoring and prosecution of specific competitive abuses are emphasized relative to others. One obvious consideration in this regard is the heightened potential need for greater coordination across national (or regional) competition policy authorities. Pricing and other conspiracies are likely to become more “international” with the growth of electronic commerce. Fraud, misleading advertising, baitand-switch and other “garden variety” abuses against consumers are also likely to increase to the extent that “hit-and-run” entry of dishonest merchants is cheaper in electronic markets than in conventional markets. On the other hand, efficiency arguments in support of tied selling, below-cost pricing and related practices may be more compelling when associated with electronic commerce. For example, tied selling may become a widespread way for innovative sellers on the Internet to internalize the economic rents associated with their innovations, e.g. by tying the sale of a new (and easily imitated) Internet service to the sale of some other good or service which is protected by trade secrets or by patents, trademarks or copyrights.26Below-cost pricing could become more ubiquitous as a means to encourage consumers to try out experience goods, in effect serving as an “up-front” warranty payment. A readier acceptance of potentially abusive behavior in favor of stricter monitoring of access to critical (if not bottleneck) nodes on the Internet, may be a beneficial tradeoff for competition policy authorities when policing the electronic marketplace.

25

Early evidence suggests that it is difficult for new firms to establish a widely accepted brand name through Internet-related selling and promotion alone. 26

Examples of this are discussed in Globerman (1995).

15. Industrial OrganizationalImplications of Electronic Commerce

267

REFERENCES Anders, G. (1998), “Click and Buy”, Wall Street Journal, December 7, R4. Auger, P., and Gallaugher, J. (1997), “Factors Affecting the Adoption of an Internet-Based Sales Presence for Small Business”, The Information Society, Vol. 13, No. 1, pp. 55-74. Barboza, D. ( 1998), “Elcctronic Trading Makes Big Inroads”, International Herald Tribune, June 1314,p. 11. Carleton, D.W. and Perloff, J.M. (1990), Modern Industrial Organization,Harper Collins Publishers. Coy, P. (1998), “YOUAin’t Seen Nothing Yet”, Business Week, June 22, pp. 130-1. Cronin, M.J. (1996), Global Advantage on the Internet, Van Nostrand Reinhold, New York. Dahl, A. and Lesnick, L. (1996), Internet Commerce, New Riders Publishing, Indianapolis. Garcia, L. (1995), “Networking and the Rise of Electronic Commerce: The Challenge for Public Policy”, Business Economics, Vol. 30, No. 4, pp. 7-14. Globerman, S. (1995), “The Economics of the Information Highway”, in Courchene, T. (ed.), Technology and Public Policy, John Deutsch Institute, Kingston, pp. 243-80. Hafner, K. (1998), “Can the Internet Cure the Common Cold,” New York Times, July 9, D1. Hoffman, D.L. and Novak, T.P. (1997), “A New Marketing Paradigm for Electronic Commerce”, The Ifforination Society, Vol. 13, pp, 43-54. Klein, B., and Leffler, K. (1981), “The Role of Market Forces in Assuring Contractual Performance”, Journal of Political Economy, Vol. 89, pp. 615-41. Kobrin, S.J. (1999, “Regional Integration in a Globally Networked Economy”, Transnational Corporations, Vol. 4, No. 2 . Lu, J. (1998), “Lemons in Cyberspace: A Call for Middlemen”, in Bohlin, E. and S.L. Levin (eds.), Telecoinmunications Transformation: Technology, Strategy and Policy, 10s Press,, Amsterdam, pp. 235-253. Pfeiffer Hagen, K.C. (1993, The Diffusion of Electronic Data Interchange, Springer-Verlag, Heidelberg. Phelan, S.E. (1996), “Internet Marketing: Is the Emphasis Misplaced?” paper presented at the Annual Meeting of the Australian and New Zealand Academy of Management, Wallongong, NSW, Australia, December 4-7. Picot, A,, Bortenlanger, C., and Rohrl, H. (1997), “Organization of Electronic Markets: Contributions from the New Institutional Economics”, The Information Society, Vol. 13, pp. 107-23. Rowan, G. (1997), “Internet Home Access Almost Doubles: Statscan”, The Globe and Mail, November 28, B6. Schecter, B. (1997), “The Internet is not Music to the CD Retailers”, The Financial Post, October 22, p. 16. Sims, D. (1995), “Who Buys on the Web”, Web Review, October 13; . Solomon, S. (1995), “Staking a Claim on the Internet”, Inc Technology, Vol. 16, No. 13, pp. 87-91. Wigand, R. (1997), “Electronic Commerce: Definitions, Theory, and Context”, The Information Society, Vol. 13, pp. 1-16. Williamson, O.E. (1973, Markets and Hierarchies: Analysis and Antitrust Implications, The Free Press, New York.

0 2000 Elsevier Science B.V. All rights reserved Convergence in Communications and Beyond E. Bohlin, K. Brodin, A. Lundgren and B. Thorngren (Editors)

CHAPTER 16

Local Versus Global Markets in Electronic Commerce: Towards a Conceptualization of Local Electronic Commerce Strategies Charles Steinfield Michigan State University, East Lansing Alwin Mahler' MCI WorldCom Germany, Frankfurt Johannes Bauer Michigan State University, East Lansing

Abstract. This chapter introduces the case for locally focused electronic commerce and offers strategies that local merchants might use to capitalize on their combined physical and visual storefronts. The analysis suggests that developments in electronic commerce should not always be considered free from the constraints of geography. This holds in particular for those local retailers who offer products readily available elsewhere. They may need a web strategy that emphasizes the synergy between their local presence and their web storefront. On the other hand, a global web strategy makes sense for local players with niche products that are not readily available.

1. INTRODUCTION According to the popular press, 1998 was the year that Internet-based electronic commerce truly arrived (New York Times, December 29, 1998). It is certainly becoming a more prominent force in the consumer retail markets (ibid.), and projections of total retail sales on the Internet range from $7 billion by the year 2000 to more than $1 15 billion within a 5 to 8 year period (U.S. Department of Commerce, 1998). Although still a relatively small proportion of total retail activity, Web-based commerce appears to be growing more quickly than other retail formats.' The U.S. Commerce Department (1998) estimates that PC hardware and soft* Any statements contained in this article are statements of the author only and do not necessarily reflect the

opinion or viewpoint of MCI WorldCom. I A 1995 retail industry report estimated total retail sales for in-store formats to be $1.7 trillion, while sales on the Web were $1 billion (Chain Store Age, 1995). Web sales growth was estimated at loo%, while in-store sales growth

2 70

C. Steinfield, A. Muhler and J . Buuer

ware sales exceeded $850 million in 1997, with some firms such as Cisco and Dell Computer now using the Web for the majority of their sales activity.’ The few widely cited success stories such as Amazon.com engender a great deal of interest among researchers and policy-makers alike in the future of Web-based retailing. The primary emphasis in much of the current debate about electronic commerce is on the global nature of electronic markets, and the lower costs of reaching global markets (e.g. Amazon.com, a firm that did not even exist a couple of years ago, now sells books in more than 150 countries). A small startup company with a good Web design team can appear to be just as large and sophisticated as an existing large firm, without having to make the investments in “bricks and mortar” and without an expensive private global telecommunications network. The Web helps to reduce transaction, search and coordination costs, facilitating globalization of electronic commerce. Even companies operating in highly specialized niche markets can find a sufficiently large audience on the Web to gain economies of scale once their products and services are made available globally. The rhetoric suggesting that electronic commerce fosters global markets is convincing, and exerts a powerful influence on the business strategies of companies going on-line (Steinfield & Klein, 1999). Indeed, a recent article in The Economist proclaimed, “The Internet is affecting all businesses in similar ways. Every industry, for example, has suddenly become part of a global network where all companies are equally easy to reach. As a result of these changes, many businesses that survived mainly because they were conveniently placed, or because they provided information that was hard to find, will soon have to find some other raison d’Ctre” (The Economist, 1997). In contrast to the global-market focus of much electronic commerce literature, Steinfield and Klein (1999) argue that much electronic commerce activity is regionally or locally focused. They point out that many regions, states and cities provide malls, directories, or other guides to business and government services in their respective geographical communities, using electronic commerce to foster economic development. Steinfield, Mahler and Bauer (1999) further argue that local retailers may not be properly equipped to compete in a global market. Rather, they may be better served by using their local knowledge and physical presence in a market to compete more effectively with distant Web-based businesses for customers in the community. Clearly the Internet and the World Wide Web can support both global and local commerce. Our goal in this paper is to introduce the case for locally-focused electronic commerce, and particularly electronic retailing, while offering strategies that local merchants might use to capitalize on their combined physical and virtual storefronts. We begin by reviewing the principal theoretical frameworks that can explain both global and local approaches, and follow with a brief overview of empirical evidence that illustrates local merchant electronic commerce strategies. Our chapter concludes with implications for policy-makers concerned about the impacts of electronic commerce on local economies.

was only 2-3%. Given the increase in commercial activity on the Web since 1995, these are likely to be conservative figures for Web-based sales.

’ Dell reportedly sells more than $3 million worth of computers on-line every day (US.Department of Commerce, 1998).

16.Local Versus Global Markets in Electronic Commerce

2.

2 71

LOCAL VERSUS GLOBAL ELECTRONIC MARKETS

2.1 The Case for Globalization The factors favoring growth of global electronic commerce have been clearly articulated in the burgeoning electronic commerce literature.3 The Web lowers transaction costs that formerly served as a barrier to entry in local markets, enabling consumers to become aware of and transact with electronic retailers who may be located anywhere. Newcomers to any local market formerly incurred sunk costs to establish a presence. There is, of course, a spatial dimension for products that influences their location of production and consumption. Some services such as haircuts or lawn care must be both produced and consumed locally, while others such as large appliances may be more costly to transport. However, many other types of goods and services can be produced anywhere and delivered either electronically or physically to consumers. For these latter products, the Web reduces search and signaling costs that normally would be a barrier to reaching a distant market. Earlier analyses by Malone, Yates and Benjamin (1987) spelled out the network effects, particularly in the area of consumer search, where electronic brokerage capabilities of networks enable consumers to search for and locate products that match desired features and prices. The relatively low costs of creating a Web presence, which is then accessible at the same cost by consumers connected to the Internet worldwide, enable firms to use the electronic storefront as a surrogate to establishing a physical presence in a local market. This surrogate local presence affords some advantages over physical retailers. Web-based retailers do not require a building or rented space in each market, and they may operate with less or no inventory. Larger, distant retailers may have significant economies of scale, through such factors as volume discounts on input^.^ In addition, a competing Web-based firm may enjoy a cost advantage due to its location near important factors of production, which would not be available to local physical retailers. Others have also emphasized the ability of Web retailers to bypass many of the intermediaries in the retail distribution value chain, affording further potential cost savings (Wigand and Benjamin, 1995; Wigand, 1997).’ Well-designed Web businesses are also more likely to have a relatively high degree of transaction automation, improving efficiency and further reducing costs relative to physical retailers. These economies can potentially enable Web-based retailers to easily undercut the prices of local retailers who formerly faced little or no competition. Despite some empirical evidence to the contrary (Bailey and Brynjolfsson, 1997; Palmer, 1997)6, there is a general expectation that For an economic analysis, see Choi, Stahl and Whinston (1997). Wigand and Benjamin (1995) and Wigand (1997) make a case for growing bypass of retailers by producers using direct sales over the Web. Recent analyses by the U S . Department of Commerce (1998), the Institute for Technology Assessment (Coates, 1998; Garcia, 1998), and the Aspen Institute (Bollier, 1995) all emphasize the reduced transaction costs afforded by electronic commerce leading to rapid growth. Although some evidence suggests that size of firm has less relevance in Web retailing, given the relatively low costs to establish a Web presence (Kennedy and Dietsch, 1995). Nonetheless, there is clear potential for large retailers who then go on the Web to use their buying clout to acquire products at a lower per unit cost than both small Web retailers or small physical retailers.

’ However, Sarkar, Butler and Steinfield (1995;

1998) present arguments for increasing, rather than decreasing, numbers of intermediaries in the electronic commerce environment.

‘ Bailey and Brynjolfsson (1997) report, for example, that in their empirical analyses the prices of books, CDs and software were higher on average on the Web than in Boston area stores. Palmer (1997) found no significant differ-

2 72

C. Steinfield, A . Mahler and J. Bauer

prices will be lower on the Web. Finally, wcll-known retailers and manufacturers can benefit from their brand name recognition in accessing new local markets. Using transaction cost theory, we can conclude that electronic commerce implies new competition for local retailers, particularly those offering products that are readily obtainable from other sources, and that are easily transported’. Of course, as Steinfield et al. (1999) suggest, local merchants might respond by establishing their own Web presence, making up any business lost to Web-based competitors by expanding into new geographic markets themselves. However, they are skeptical about the likely success of this strategy for many smaller local retailers. They note that Web startup firms face significant barriers in their attempts to attract customers. The sheer number of new Web businesses reduces the likelihood that people will chance upon a Web store, necessitating large marketing and advertising expenditures to get noticed. Smaller local retailers may also not have the business systems in place to adequately serve distant customers, even if they do attract them. The ability to process electronic orders, verify payments, ship to distant customers, properly apply sales tax regimes, handle returns, and many other skills need to be acquired. Moreover, despite the increasing sales activity on the Web, lack of trust remains a strong inhibitor. Unknown virtual Web stores are more likely to experience problems due to lack of trust than established brand names. In fact, recent surveys of Web users show that lack of a local physical presence inhibits purchases (GVU Center, 1997).8Other surveys suggest that problems with unreliable products and difficulties in returning goods have dissuaded Internet shoppers from being repeat buyers (Coates, 1998). In summary, electronic retailing trajectories that favor the success of relatively few established Web brands with easy access to global markets raise important questions about the role of electronic commerce for smaller, local merchants. Is electronic commerce another threat, just like the arrival of large chains with better economies of scale, that serves to take away market share from small, local retailers? Or can local retailers make up the losses in sales to global, virtual Web stores by “going global” themselves, and using the Web to expand their reach into new markets? Despite the attention that electronic commerce has received nationally and internationally, there has to date been little emphasis on its significance for local, physically based, community businesses. Auger and Gallaugher (1997) do compare small and large Web-based businesses through an empirical study of motivations to establish an on-line presence and barriers to success. They note, however, that smaller and medium-sized firms go on-line to increase sales outside their geographic region, rather than to improve business in their local market.

2.2 Towards a Local Orientation in Electronic Commerce Rather than using the Web to attract distant clients, an alternative strategy for local merchants is to use it to better defend their home territory, and offer better service to the local commuence between in-store prices and prices in Web stores, catalogs and cable TV shopping channels across a variety of products. One plausible explanation is that Web-merchants were simply using a price discrimination approach that allowed them to find buyers willing to pay extra for the added convenience of Web shopping.

’ Note that this is even more true for information products that can be delivered electronically via the Internet. ’ In one survey of online shoppers (GVU Center, 1997), more than half of all respondents required or preferred an off-line presence before shopping on the Internet. Results can be found on the Web at: .

16. Local Versus Global Markets in Electronic Commerce

2 73

nity. For a variety of reasons, articulated below, local merchants may enjoy a competitive advantage over their Web-only rivals when using a hybrid Web-plus-physical-presence strategy.' These reasons can be broadly grouped into four categories: ( 1 ) trust and embeddedness , (2) consumer needs and behavior, (3) services and applications that capitalize on complementarities between the Web and their physical presence, (4) local knowledge, and (5) local initiatives for economic development.

Trust and Embeddedness A commonly cited impediment to on-line shopping is consumers' lack of trust in the legitimacy of the Web-based store (Bollier, 1995; Coates, 1998). Indeed, the emerging digital certificate infrastructure appears to be motivated by a desire to demonstrate to potential consumers that merchants are not merely criminals masquerading as a business in order to obtain credit card numbers. Consumers who recognize the Web store as an extension of an existing business may perceive it to be more legitimate, and have more trust in the store. This is not only likely to be a local phenomenon, but certainly influences perceptions of national brands as well. Nonetheless, we expect that consumers in any particular local market will have more confidence in a Web business if they can associate it with a particular physical store that they have visited or seen in their community. It may also be the case that distant consumers will perceive a Web business to be more legitimate if they are made aware that it has been a successful physical business. Embeddedness is a term that has been used by sociologists when describing the extent to which economic exchanges are predicated upon patterns of social relations in any community (Granovetter, 1985). According to Granovetter, embeddedness is often considered a problem by economists, who argue that when economic exchange is determined by social relations, inefficient allocation of resources can result." However, he also notes that social relations often facilitate trust, permitting exchanges without expensive contracts or legal fees and thereby reducing costs. Recent empirical work by Kraut et al. (1998) suggests not only that electronic exchanges between buyers and sellers are associated with interpersonal linkages, but that the quality of electronic exchanges complemented by interpersonal relationships is higher than that of electronic-only exchanges. Embeddedness that enhances trust can occur in more than familial relations. Local merchants may belong to chambers of commerce, volunteer organizations, churches, and other types of community groups. Through these efforts, they become recognized and trusted members of the community. These associations may extend to the Web, for example when other local businesses provide reciprocal linkages or endorsements. Consumer Needs and Behavior

Consumers have a variety of needs and preferences that influence their shopping behavior. Here we mention only a few possible consumer needs and shopping preferences that may offer a competitive advantage to a local merchant who also establishes a locally oriented Web pres-

' Recent data seem to support the existence of strong synergies between bricks-and-mortar and Internet retailing (see Economist, 1999). l o Consider, for example, the inefficiencies resulting from a producer buying needed inputs from a family member rather than a lower-cost competing supplier.

2 74

C. Steinfield, A. Mahler and J. Bauer

ence. Our goal is to illustrate those cases where the combination of Web plus physical presence would have a greater chance of capturing business than a Web-only presence. With a locally available physical store, consumers’ perceived risks in engaging in on-line shopping are potentially reduced. As noted above, problems associated with the return of faulty merchandise are a major reason for lack of repeat on-line purchases (Coates, 1998). A local physical presence reduces such risks by providing a low-cost means for returning flawed goods, or seeking technical assistance with a particular product. People may also learn about products at the Web site, but confirm product quality by visiting the store to physically inspect the goods prior to purchase. Such pre-purchase and after-sales services not only reduce risk, and thereby lower consumer transaction costs; they can further build trust in the Web store. Shoppers are not homogeneous, and may have different preferences that influence their use of both regular and Web-based stores. Various market segmentation approaches, such as SRI’s Values and Lifestyles (VALs) and Internet VALs (iVALs) are examples of marketers’ attempts to understand diverse consumer preferences that affect purchasing behavior.” It has been suggested, for example, that the Web excels in “search” goods, for which there are particular product features that enable evaluation by consumers prior to use. The Web facilitates search allowing consumers to locate desired products more easily (Klein, 1998). Klein (1998) argues that the Web even turns “experience” goods, which are thought of as products that can only be evaluated after trying them, into search goods, by using multimedia capabilities of the Web to permit on-line experience.” Coupled with a purchasing style characterized by extensive research prior to actual purchase, the Web can be a powerful complement saving consumers on the high search costs associated with such information gathering. Others suggest that many consumers prefer the social and personal experience of shopping, engaging in interaction with others in the marketplace as they shop (Sarkar et al., 1995). With both a Web and physical store such diverse preferences can be addressed. There are likely to be shoppers who exhibit hybrid patterns, such as preference for gathering information on-line, but making actual purchases in a store. Some may prefer the convenience afforded by the Web, and use it to locate desired products, but due to the lack of trust in the security of the Web, prefer to make actual transactions in person. Perhaps most important are the consumers who desire immediate gratification, and do not want to wait for products to be shipped. They may identify products on-line, but cater to their desires for immediate gratification by picking up the product at the local store. These are all examples where the physical store would capture purchases that otherwise would not have been made with only a Web-based store. Complementarities Between Web and Physical Presence

The Web store and the physical store can support each other in many different ways, capitalizing on natural complementarities (Steinfield et al., 1999). These include cross-promotions, joint-service provisions, and value-added services. Cross-promotions are perhaps the most straightforward example of a natural complementarity. Web stores may provide the bargains that people may expect when shopping on the Internet, but also offer coupons for in-store purchases. Marketers are accustomed to the use of “loss-leaders” as an approach to increasing ”

Information on VALs and iVALs can be obtained from their Web site at: .

” For example, music is normally thought to be an experience good. Web sites such as CDNow, however, allow consumers to search by artist or genre, and hear a short preview prior to purchasing music.

16. Local Versus Global Markets in Electronic Commerce

2 75

store traffic, and realize that once in the store, other purchases for non-discounted merchandise are more likely. A physical store can use its Web site to highlight local events, such as a reading or performance at the store, helping to bring in traffic. The Web store may also provide information about additional services that are available at the store and add value to products purchased on-line. A company might use electronic mail for direct marketing, not only to advertise the Web site, but also to provide information about in-store products and services. Some services might be provided jointly, levering the investment in physical and Web presence. A good example would be when a computer store offers a product on-line, but provides installation and repair services at its premises for customers purchasing from its Web site. Finally, the Web store may function as a source of value-added services for customers who have purchased or plan to purchase goods at the physical store. A purchaser of a musical instrument, for example, may find additional sheet music at the store’s Web site. Also, Heikilla, Kallio, Saarinen and Tuunainen (1998) describe the emergence of Web-based grocery sites, where customers gain convenience by shopping on-line. A local grocery store then delivers the chosen items. These various complementary approaches represent a sampling of strategies a retailer might use to leverage its investment in physical and Web distribution channels. In many cases, the natural complementarities offer a distinct advantage over Web-only stores that might require shipping a product back for additional service or installation work. Local Knowledge Local merchants who target a local market should be able to capitalize on their local knowledge to compete effectively against distant virtual stores. Bouwman (1999), for example, points out that in many parts of the world, there are local or regional language differences that can be captured in the Web content to increase the appeal of the site. Local customs, tastes and product preferences will be known by local merchants, but not necessarily by distant virtual sellers. Prominent local citizens who offer endorsements, reference to local landmarks, awareness of important local events that may influence purchase patterns, and many other local content strategies should help make a merchant’s Web site more meaningful and appealing to the local target market. It would be difficult for globally oriented Web businesses to capture the same degree of local relevance. RegionalKity Economic Development

In many cities, regions, or states, electronic commerce is becoming a foundation for economic development policy (European Commission, 1996; Steinfield & Klein, 1999). Many communities are actively organizing electronic commerce sites into business directories or malls to boost the local economy (Steinfield & Klein, 1999). Zimmermann (1997), for example, describes the role of the Electronic Mall Bodensee in stimulating commerce in the Bodensee region. Local communities may also be active in developing infrastructure to help businesses in the area participate in electronic commerce. In this regard, Fesenmaier & Es (1 999) describe how a rural region is pursuing Internet infrastructure to help compete with urban areas, while Damsgaard and Farhoomand (1999) illustrate the emerging e-commerce infrastructure strategy for the city of Hong Kong. Local merchants can therefore benefit by joining such local or regional initiatives, receiving assistance in the infrastructure aspects as well as the usage and marketing aspects.

2 76

C. Steinfield, A. Mahler and J . Bauer

3. EMPIRICAL RESEARCH ON LOCALLY ORIENTED ELECTRONIC COMMERCE Research examining the electronic commerce strategies of local merchants is difficult to find. Steinfield et al. (1999) examined the Web sites of 125 local merchants in the Mid-Michigan area of the United States, in order to explore how they linked their physical and Web presences. The Web sites were randomly selected from various business directories, and examined specifically for references to the physical location as well as for the use of strategies that reflected synergies between physical and Web presences. In-depth interviews were conducted with a small set of businesses that were offering complete Web stores, including the ability to order and pay for merchandise on-line. A key interest of the interviews was to learn more about the local versus global motivations for establishing the Web site. In particular, they asked if businesses went on-line to better serve the local market and defend against new on-line competitors, or to extend their reach to new distant markets. The findings suggest that currently, few local merchants actually use their Web presence to complement their physical stores, and locally oriented strategies were not common (Steinfield et al., 1999). Among the results reported, was that, although nearly all of the Web sites examined at least identified the city in which the business is located (98%), there were few other attempts to provide synergies between the on-line and physical presences. The primary strategy was to make some additional reference to the physical location of the business, or the community in which it is found, with more than half (57%) providing this type of information. Most also provided a local telephone number (86%), with only a small number (6%) offering only an “800” number and no local number. However, few offered any incentives for Web site visitors to come to the physical location through such devices as coupons or notices of events or sales. None offered any options for in-store pickup of merchandise ordered on-line. Only 2% mentioned any return or refund policy, and none of these noted whether this could be handled at the physical store. Most of the Web sites for the local businesses in this study were relatively unsophisticated, and only a small percentage (1 8%) even allowed on-line ordering of merchandise or services. An even smaller amount allowed for on-line payment (14%) with a few sites relying on preexisting accounts with customers who ordered online. These results indicate that while most local businesses do attempt to use their physical presence to establish their legitimacy as an existing company, few go beyond this simple strategy to seek competitive advantage over distant Web businesses.

3.1 Factors Influencing Success or Failure of Web Strategies Steinfield et al. (1999) conducted interviews with several store managers to discuss the rationale for establishing the Web site, its role in their business operations, and the results of their activities on the Web. Interviews also explored how the Web site began, who manages it, and the features of the products they make available from their Web site. In two of the five cases, the introduction of a Web site was considered an outright failure, with little or no new business originating from the Web site. Two cases were marginally successful, in that the local merchants were able to recover the cost of designing and maintaining the Web site, but had not yet experienced significant increases in revenues. Only in one case, a supplier of stained glass products, did the merchant report significant revenue growth as a result of establishing a Web presence. The case studies illustrate a few important patterns that

16. Local Versus Global Markets in Electronic Commerce

2 77

highlight factors influencing the success or failure of the Web sites of‘ local businesses. Success factors identified by Steinfield et al. (1999) include initiating the Web presence with a business strategy in mind, offering niche products that are not readily obtainable elsewhere, having previous experience with mail order or other ways of catering to distant customers, and, of the greatest significance for our conceptual arguments here, offering content and services that not only directed customers to the physical store, but added value to in-store offerings. First, rather than resulting from any well-thought-out business strategy, four of the five cases illustrate that local merchants also are “jumping on the bandwagon” when they set up a Web site. In most of the unsuccessful cases, merchants had little or no expertise, and were totally reliant on a local contact to help them get started. In these bandwagon cases, it appears that issues of long-term site management were not addressed, and the Web sites themselves remained relatively unsophisticated and unchanging. The site owners were unaware of any new competitors they might face due to the Web. The most successful case was one in which there was a clear business strategy (improving customer service) that guided the Web site development. This company, the stained glass supplier, was well aware of its competitors, on and off the Web, and in local and distant markets. A second success factor relates to specific product characteristics. Researchers have often noted that one effect of electronic markets is to reduce asset specificity by expanding access to a larger number of potential suppliers (Malone, Yates and Benjamin, 1987). Seen from the perspective of the seller, the electronic market enables a seller of niche products to aggregate buyers from distant markets. Stained glass certainly can be considered a specialized product where there are few suppliers, and relatively little demand in any single locality. In this situation, the Web represents a perfect tool for accessing distant customers, who do not necessarily have local options to satisfy their demands. The other products were less niche-like, and these merchants were more likely to face competing suppliers in other markets. It is therefore not surprising that the stained glass supplier experienced the most success. The five case-study businesses also differed greatly in their degree of previous experience with serving distant markets. This appeared to be something of a “readiness” factor, with firms having a pre-existing mail-order capability the most likely to gain from their Web-based activities. Finally, in the one truly successful case, the emphasis was not entirely based upon becoming a global, virtual enterprise. Rather, the stained glass site provided a detailed history highlighting the physical location of the store and factory. It provided an on-line map to the factory outlet, and information about and enrollment in classes and other activities that would take place there. The company noted that many customers drove from quite some distance away, and often arrived with the printed-out map from the Web site in hand. The other sites noted their physical location, but did not otherwise exploit any potential synergies between their physical location and their Web site. In general, the majority of the cases emphasized the strategy of expansion to new markets over the defense of the local market. For local merchants with no real niche product or service, and relatively little price advantage over other larger players, or all virtual players, this appears to be a misguided competitive strategy.

C. Steinfield, A . Mahler and J . Bauer

2 78

4.

CONCLUSIONS

Clearly there is a lack of good empirical data on the relative success of electronic commerce in local markets. There may be sophisticated local users of the Web in some communities, particularly larger cities that have deeper pools of Web-literate personnel, greater local competition, or a better-developed technical infrastructure. Nonetheless, our analysis in this chapter suggests that developments in electronic commerce should not always be considered free from the constraints of geography. Every business is not a global business, although many can certainly extend their reach into distant markets through the Web. Yet this is a double-edged sword, since just as local players attempt to hide their locality and become more global, large global firms can use the Web to enter small local markets. In one sense, many local merchants have already experienced similar threats from large chains that physically entered their local market. For example, Wal-Mart used their backoffice efficiencies and economies of scale to threaten inefficient local merchants. To survive, many local merchants learned how to cater to the specific needs of the local community. The problem now is that the Web helps larger firms who do have back-office efficiencies, to reduce also the costs associated with front-office services. Mass customization represents a coming trend that may give Web-based stores the same ability to cater to individual customer wants and desires. Local merchants who establish a Web presence may imitate the Web strategies of their larger competitors, but without the same back-office efficiencies they will be less effective. Hence, local merchants must look elsewhere for their competitive advantage in electronic commerce, leveraging their embeddedness in a community, local knowledge, and existing physical presence in ways that would be difficult for distant Web businesses to duplicate. Our analysis suggests that for local players who have niche products that are not readily available, and who have developed a capability to service distant customers, a global Web strategy makes sense. However, for those who are offering products that are readily obtainable elsewhere, they may need a Web strategy that emphasizes the synergy between their physical presence in a local market and their Web storefront. W e think this characterizes many local retailers. Yet in our research, we could not find many good examples of complementarity between local physical presence and a merchant’s Web site. Few mentioned their Web site in their traditional store, in their advertisements, or in the telephone directory, and few featured special services for local customers, or highlighted local information as a way of building trust on their Web site. Rather, we were more likely to find mismatches between Web site strategy and the nature of the local business. Clearly a strong conclusion is that local merchants need to be deliberate about their target audience - local or global - and offer the appropriate complementary services on their Web site. The rub is that most local merchants do not possess the technological sophistication to implement Web strategies on their own, as clearly indicated in our empirical review of local Web sites. This suggests that electronic commerce raises an important policy question for local communities. Unlike upstart companies that can take advantage of the Internet in designing their management, marketing, and sales processes, established firms must modify their existing organization. Especially small and medium-sized enterprises (SMEs) seem to have difficulties in adjusting their processes to the opportunities and constraints of the Internet. Thus, there seems to be a role for local and regional communities as well as trade associations in providing educational and consulting services to local merchants. Such organizations could

16. Local

Versus Global Markets in Electronic Commerce

2 79

also assist in providing some of the required infrastructure on a shared basis. Examples of successful peers could help overcome the knowledge gap for SMEs. From the perspective of local communities, providing such information and assistance may well have the character of a public good securing continued local investment and employment by the business community.

REFERENCES Auger, P. and Gallaugher, J. (1997), “Factors Affecting the Adoption of an Internet-based Sales Presence for Small Businesses”, The Information Society, Vol. 13, pp. 55-74. Bailey, J. and Brynjolfsson, E. (1997), “In Search of “Friction-free Markets”: An Exploratory Analysis of Prices for Books, CDs, and Software Sold on the Internet”, paper presented to the TwentyFifth Annual Telecommunications Policy Research Conference, October, Washington, D.C. Bollier, D. (1995), The Future of Electronic Commerce: A Report of the Fourth Annual Aspen Roundtable on Information Technology, The Aspen Institute, Washington, D.C. Bouwman, H. (1 999), “E-commerce: Cyber and Physical Environments”, Electronic Markets, Vol. 9, No. 1, April, pp. 58-64. Chain Store Age (1993, State of the Industry Report, August. Choi, S.Y., Stahl, D. and Whinston, A.( 1997), The Economics Technical Publishing, Indianapolis.

of

Electronic Coinmerce, Macmillan

Coates, V. (1998), Buying and Selling on the Internet: Retail Electronic Commerce, Institute for Technology Assessment, Washington, D.C. Damsgaard, J. and Farhoomand, A. (1999), “Electronic Commerce in Hong Kong: The Impact of Regional Dimensions on the Diffusion of Electronic Commerce”, Electronic Markets, Vol. 9, No. 1, April, pp. 73-80. Economist (1997), “In Search of the Perfect Market”, The Economist, September 14; ihttp://www.economist.com/editorial/freeforall/14-9-97/ec 1.html>. Economist (1999), “The Real Internet Revolution”, The Economisr, August 21, pp. 53-54. European Commission (1996), Building the European Informution Society for Us All - First Rejlections of the High Level Group of Experts, January 1996, Brussels; . Fesenmaier, J. and van Es, J. (1999), “Rural Development: Communications and Computing Technologies Create a Rapidly Changing Environment”, Electronic Markets, Vol. 9,No. 1, April, pp. 81-86. Garcia, L. (1998), Global Electronic Commerce and Transaction Costs, Institute for Technology Assessment, Washington, D.C. Granovetter, M. (19 8 3 , “Economic Action and Social Structure: The Problem of Embeddedness”, American Journal of Sociology, Vol. 91, No. 3, pp. 481-510. GVU Center (1997), GVUS Eighth WWW User Survey, October; .

280

C. Steinfield, A. Mahler and J . Bauer

Heikilla, J., Kallio, J., Saarinen, T. and Tuunainen, V.K. (1998), “Grocery Shopping for the Elderly and Disabled: Finnish EC Experiments”, Electronic Markets, Vol. 8, No. 2, pp. 17-19. Kennedy, W. and Dietsch, J. (1995), Making Money Online, Sunsite. Klein, L. (1998), “Evaluating the Potential of Interactive Media through a New Lens: Search Versus Experience Goods”, Journal of Business Research, Vol. 41, No. 3, pp. 195-203. Kraut, R., Steinfield, C., Chan, A,, Butler, B. and Hoag, A. (1998), “Coordination and Virtualization: The Role of Electronic Networks and Personal Relationships”, Journal of Computer Mediated Communication, Vol. 3, No. 4; chttp://www.ascusc.org/jcmc/vol3/issue4/~aut.ht~~, Malone, T., Yates, J. and Benjamin, R. (1987), “Electronic Markets and Electronic Hierarchies: Effects of Information Technology on Market Structure and Corporate Strategies”, Communications of the ACM, Vol. 30, No. 6, pp. 484497. New York Times (1998), ’99: A Year to Make Good on Electronic Commerce”, e-commerce report, December 29; . “

Palmer, J. ( 1 997), “Electronic Commerce in Retailing: Difference Across Retail Formats”, The Information Society, Vol. 13, pp. 75-91. Sarkar, M., Butler, B. and Steinfield, C. (1999, “Intermediaries and Cybermediaries: A Continuing Role for Mediating Players in the Electronic Marketplace”, Journal of Computer Mediated Communication, Vol. 1, No. 3; Sarkar, M., Butler, B. and Steinfield, C. (1998), “Cybermediaries in Electronic Marketspace: Towards Theory Building”, Journal ofBusiness Research, Vol. 41, No. 3, pp. 215-221. Steinfield, C. and Klein, S. (1999), “Local versus Global Issues in Electronic Commerce”, Electronic Markets, Vol. 9, No. 1, April, pp. 45-50. Steinfield, C., Mahler, A. and Bauer, J. (1999), “Electronic Commerce and the Local Merchant: Opportunities for Synergy between Physical and Web Presence”, Electronic Markets, Vol. 9, No. 1, April, pp. 51-57. U S . Department of Commerce (l998), The Emerging Digital Economy, Department of Commerce, Washington, D.C.; .

Wigand, R. and Benjamin, R. (1993, “Electronic Commerce: Effects on Electronic Markets”, Journal of Computer Mediated Communication, Vol. 1, No. 3 ; ~http://www.ascusc.org/jcmc/voll/issue3 /wigand.html>. Wigand, R. (1997), “Electronic Commerce: Definition, Theory, and Context”, The Information Society, Vol. 13, pp. 1-16. Zimmermann, H.-D. (1997), “The Electronic Mall Bodensee (EMB): An Introduction to the EMB and its Architectural Concepts”, Electronic Markets, Vol. 7, No. 1, pp. 13-17.

PART VI

CONVERGENCE AND REGULATORY TRANSFORMATION

0 2000 Elsevier Science B.V. All rights reserved Convergence in Communications and Beyond E. Bohlin, K. Brodin, A. Lundgren and E. Thorngren (Editors)

CHAPTER 17

Toward a Theory of the Global Liberalization of Telecommunications: Implications for Convergence Regulation Gene Mesher College of Business Administration, California State University, Sacramento Edward E. Zajac University of Arizona, Tucson

Abstract. Although the movement to liberalize telecommunications is now worldwide, few attempts have been made to build a political economy model of this process, especially with regards to international differences in liberalization rates. Using Noam’s pioneering work on this topic as a starting point, this paper extends Noam’s work by integrating extant public choice/political economy models into a three-stage political economy model of liberalization. In our theory, the Iron Law of Political Economy plays a fundamental role. The Iron Law is the use of the state’s coercive powers by those in power to the benefit of favored groups at the expense of introducing inefficiencies into the economy. This law will play an important role in the regulatory transformation related to convergence.

1. INTRODUCTION Although the movement to “liberalize” the telecommunications industry is sweeping the world, and country after country is privatizing its government-owned telecom business and passing laws to guarantee new entrants into the business’ various facets, the rate of liberalization has varied widely from country to country. This paper addresses two main questions related to the liberalization process: (1) Why did the liberalization movement occur?

and (2) Why is it occurring at different rates in different places?

The paper will also address the price of regulatory change related to convergence.

284

G. Mesher and E. Zujac

In a series of seminal papers, Noam (1987, 1989, 1992, 1994) advanced a theory of why liberalization was inevitable, thus offering an answer to the first question, but not addressing the second. This paper tries to build on Noam’s work. First, we try to flush out further aspects of answers to Question 1. Second, we make a start on addressing the much more difficult Question 2. In particular, we develop a three-stage model of the liberalization process. Stage I of this model is an equilibrium dominated by an interest group Noam calls the “postalindustrial complex” which is a coalition of the PTT monopoly, its suppliers, residential and rural users, trade unions, the political left and the newspaper industry (whose postal and telegraph rates were heavily subsidized). We envision the postal-industrial complex as having captured the coercive power of the state. The result is an equilibrium in which operated the above-mentioned Iron Law of Political Economy. We model this equilibrium by applying Becker’s (1976) modification of the Stigler-Peltzman theory of regulation, a model that we call the “SPB” model of regulation. We also point out that, based on Wilson’s (1980) categorization of types of politics, this equilibrium is a case of “client politics”. Stage 11 is a complex process of forces working to unravel the entrenched postal-industrial complex, together with interests within the complex resisting change. We devote little discussion to the unraveling forces because these are dealt with by Noam in his “network tipping” model (Noam, 1992, 1994). Likewise, we only briefly consider the forces resisting change. We feel that these are aptly captured by Charlotte Twight’s theory of institutional change (Twight, 1994, 1998). Her theory is based on constitutional-level political transaction costs and highlights the manifold means available to entrenched interests to resist change and stay i n power. We envision that the Stage 11process will evolve to another equilibrium, which in our view is best captured by Becker’s (1983) oligopoly theory of political competition. In Wilson’s schema this form of equilibrium fits into Wilson’s (1980) “interest group” politics, consisting of two groups of more or less equal political power battling to a standoff. One possibility for modeling the two groups is survivors of the postal-industrial complex as one group, and Noam’s ( 1987) “service-information coalition,” consisting of a “second electronics sector” and large telecom users, as the other. There are other possibilities. For example, in the USA at present, the two coalitions could be modeled as the local exchange carriers, together with state commissions and numerous allies, as one coalition, and the interexchange carriers, the FCC, and their allies as the other. Finally, we posit that Stage III will again consist of a new, client politics equilibrium, but one that is much more efficient than that of Stage I.

2.

THE DRIVERS OF LIBERALIZATION

Noam (1989) traces the origins of the postal-industrial complex to the sixteenth century, when European rulers realized that a government-owned postal monopoly could be a cash cow. When the telegraph and telephone were invented, most governments moved to bring these new forms of infrastructure under the cash cow umbrella, resulting in the creation of the now familiar government-owned Postal, Telegraph and Telephone (PTT) monopolies. Unlike mail delivery, the monopoly was profitably fed by a domestic equipment industry, all overseen by government bureaucrats. The result was the formation of the “postal-industrial complex.” In

17. Toward a Global Telecom Liberalization Theory

285

the USA, except for a brief period before World War I, the postal and telegraph/telephone monopolies were separate, but Noam’s theory applies nonetheless. Noam’s (1987) paper lists a dozen causes for the erosion of the postal-industrial complex: (1) the growth in the service economy, resulting in an explosive growth in telecom demand and the creation of telecom expertise at the level of the firm, ( 2 ) reduction of the centralized network’s economies of scale and scope, (3) users’ demand for a wide range of differentiated products rather than the PTT’s limited menu of standardized products, (4) new niche users with specialized needs such as travel agents and airlines, financial institutions, etc., not speedily met by the PTT, (5) increasing control of the network by the user at his premises, (6) new technologies such as cellular and fiber (Noam would undoubtedly now add the Internet and PCS), (7) internationalization of transactions and the reduction of transmission costs leading to alternative transmission paths that can circumvent PTT regulation, (8) merging of technologies, particularly computing and telecom technologies, (9) government programs that restrict the PTT’s monopolizing practices, such as government support of the electronics industry and the regional and supranational collaborations like the NAFTA, the EU, the WTO, etc., (10) the emergence of value-added services, (1 1) saturation of basic service weakening the PTT’s argument that competition would divert resources from the primary social goal of universal service, and (12) an emerging wedge between postal and telecom services, and a breakdown in governmentllabor support of PTTs. In his 1989 paper, Noam further argues that the above powerful centrifugal forces drove the independent computer and component industry to form a “second electronics sector” that could directly compete with the postal-industrial complex and could surpass it technologically. For example, Noam points out that integrated circuit technology was largely developed by firms that were not part of the postal-industrial complex. Further, the second electronics sector formed an alliance with the large service users to form a politically potent “servicesinformation coalition.” As a US example, Noam cites American Express, IBM, Time, TWA, Silicon Valley firms and Citicorp as being members of the “service-information coalition.”

3. THE ROLE OF IDEOLOGY’ Noam’s papers at the end of the 80s and the beginning of the 90s were remarkably comprehensive and prescient. Nonetheless, it seems to us that Noam ignored some important factors. Perhaps foremost amongst these has been the role of ideology in the formation of public polIcy. In the last few decades a number of economic experiments seem to have shown the superiority of market economies over centrally controlled ones. Thatcherism stoked the British economy, and a number of South American countries, such as Argentina, Brazil and Chile, have come under the control of neo-classically oriented economists who have instituted market-oriented reforms, apparently with great success. And, we also have seen the collapse of the Soviet Union and the descent of dependency theory-inspired state capitalism in Latin America. These and other experiences have command-and-control ideology on the run, to be replaced

’

This section is based on several sources including: Brooks (1976), Danielian (1939), Lowry (1973), Mueller (1997), Paine (1921), and Phillips (1988).

G. Mesher and E. Zajac

286

bq a general acceptance of market ideology. This ideological sea change has undoubtedly contributed to the worldwide movement to liberalize telecommunications. Perhaps equally important is the demise of the theory that telecom is a natural monopoly, a theory that until recently was uncritically accepted. Part of the acceptance rested on what seemed an obvious dictum: good social planning requires the “elimination of wasteful duplication.” Thus, John Stuart Mill’s 1848 idea that efficiency required that there should only be one water company and one gas company in London struck a responsive chord, not only with the public but with academic economists. By the late 1890s, they had fully espoused the doctrine that there existed “natural monopoly” industries2, and that telecom was one of them. According to natural monopoly theory, left to competition, a single firm in the telecom industry would end up taking over the entire market and would be capable of charging exorbitant prices that reaped the firm exorbitant profits and in the process restricted output from the welfare-maximizing level. The remedy for this unacceptable situation was to have one firm, thus “eliminating wasteful duplication,” but to have the firm be either government-owned or government-regulated. That way the public could have its cake and eat it. Government ownership or regulation would keep prices low and quality high and, at the same time, would ensure that the full efficiencies of economies of scale were captured. In the USA, natural monopoly theory fit well with the progressive movement that flourished at the beginning of this century. This argued that most government failings, such as fraud, corruption and incompetence, could be cured by the appointment of civil servants on the basis of technical expertise rather than on the basis of political connections. It was an easy step to argue that natural monopoly firms, including telecom, should be overseen by a meritocracy of expert technocrats. Thus was born the movement to create state regulatory commissions, a movement that began in 1907 with the creation of the New York and Wisconsin commissions, and then quickly spread to most states of the union. The confluence of the birth of state regulatory commissions and natural monopoly ideology presented AT&T’s president, Theodore Vail, with an opportunity that he seized. In 1894, AT&T’ s basic telephone patents expired, thus destroying its legal telephone monopoly. Competition evolved quickly. By the early 1900s there were over 4000 phone companies. Price wars caused demand to explode, and AT&T was essentially driven into bankruptcy. In 1907, its bankers, led by J. P. Morgan, took over and appointed Vail president. Vail faced two political problems. First, he had to avoid the telephone industry’s being overly controlled by the federal government. In 1913, two years after the Supreme Court’s breakup of Standard Oil and American Tobacco, AT&T signed what is now known as the Kingsbury consent agreement with the Department of Justice in order to avoid that fate. Next, Vail had to prevent AT&T and its competitors from coming under the regulatory control of thousands of municipalities in the USA. His solution for this was to embrace the natural monopoly and progressive arguments. He proposed that AT&T be subject to regulation at the state level by the newly forming state regulatory commissions, and, in return, that “wasteful duplication” be eliminated by the commissions’ mandating that telephone service be provided by a single firm. Thus, Vail formed a political coalition with state regulators that undercut any movement to nationalize either the telephone industry or fragment it, eliminated “home rule” (regulation at the municipal level), and gave AT&T a telephone monopoly in most major urban areas. All this was justified by John Stuart Mill’s theory of natural monopoly. ~

’For the history of natural monopoly theory and its academic acceptance, see Lowry (1973).

17. Toward a Global Telecom Liberalization Theory

287

Vail’s strategy was brilliantly successful. It brought AT&T out of bankruptcy; AT&T’s regulated monopoly became profitable and dominated US telephony from the end of World War I until AT&T’s breakup in 1984. The monopoly only began to seriously unravel in the early 70s when the FCC allowed new entrants into both the long distance transmission and customer premises equipment markets. The success of these experiments showed the falsity of natural monopoly theory. Interconnection of competitors and AT&T was not trivial, but doable. Many competing firms and suppliers of customer premises equipment could fruitfully coexist on the same network. Natural monopoly theory for telecom was abandoned, to be replaced by “essential facility” or “bottleneck” theory3. This held that, although telecommunications was not a natural monopoly, single, unregulated ownership of certain facilities would allow the owning firm to exclude rivals and give the firm an economic advantage. For economic welfare, bottleneck theory holds that such a bottleneck should be made available to all rivals. However, to prevent the owning firm from reaping monopoly profits, there should be regulation of the rates the firm charges rivals for using the bottleneck4. The two theories have crucially different consequences. Natural monopoly theory says that there should be a single firm and all rivals should be kept out. Bottleneck theory says that all rivals who wish to enter should be allowed in, and thus allows the possibility of many firms. The important point is that the success of the US experiment with liberalization thoroughly discredited the theory that telecommunications is a natural monopoly. This undermined the main theoretical argument available to entrenched interests who would maintain the postalindustrial complex status quo. At the same time, the experiment got started R&D on how to connect competing networks and highlighted the importance of getting right interconnection ground rules.

4.

THEORY

4.1 The Postal-Industrial Complex Equilibrium a.

Illustration of the Theoretical Model - the “Postal-IndustrialComplex” Equilibrium in the USA Before the AT&T Breakup

As already mentioned, Vail’s efforts led to a period of stability that lasted roughly from the end of the first World War until the early 70s. What evolved during this period was a coalition of AT&T and its affiliates - its operating companies and over a thousand (mostly rural) telephone companies, and its regulators - the Federal communications Commission and state Some authors distinguish between a “bottleneck” and an “essential facility,” with a ’bottleneck” becoming an essential facility to a rival only when the “bottleneck’s” owner refuses to lease it to the rival. We, however, follow common usage and use the terms “bottleneck” and “essential facilities” interchangeably. For example, Baumol and Sidak (1994, p. 7) state, “...the LECs’ [Local Exchange Carriers’] monopoly services constitute inputs for the activities of the rivals of these firms in other arenas - inputs without which the rivals cannot hope to operate. These monopoly services, consequently, are deemed ‘bottlenecks’ or ‘essential facilities’ - meaning that a LEC, if it were to operate completely without regulatory constraint, could use the services or facilities in question to force rivals to bend to its will or to destroy those rivals altogether.” In the USA, the landmark antitrust case that lays out the essential facilities doctrine is USA vs. Terminal Association of St. Louis, 1912.

288

G. Mesher and E. Zajac

regulatory commissions. Elsewhere one of us has called this coalition, Old Regulated Bell, or ORB. It reached a mature equilibrium in the 1960s’. The primary beneficiaries of ORB were residential telephone consumers. They were by far the most numerous of Bell’s customers and had enormous political clout. This was direct in states (roughly 113) that elected regulators, but was also very strong in states where they were appointed. Zajac has argued that the satisfaction of residential customers by the provision of basic telephone service that continually decreased in price and increased in quality drove ORB’s political and technological strategy (Zajac, 1990a, 1990b). Because it generated inefficiencies, however, the strategy was not without cost. Briefly, some of the main inefficiencies were the following6. (i)

Products and services that the public wanted were not offered; ORB considered them to be “frills” that might interfere with the provision of basic telephone service. Until the late 1950s AT&T’s “attitude as to color of equipment was much the same as Henry Ford’s had been as to the Model T - that is, the customer could have any color desired, so long as the color was black” (Brooks, 1976, p 265). Newly appointed Chairman Kappel’s 1950s introduction of color telephones was considered “almost a smack of depravity to some traditionalists within the company” (Brooks, 1976, p. 266). Starting at the turn of the century, AT&T made the “French” telephone, with microphone and hearing piece in a single handset, for foreign export, but made only the candlestick phone available in the USA until 1927. The reason for this? The French phone’s acoustics made it less suitable for long distance than the candlestick phone. Never mind the fact that customers might have been delighted to buy French telephones in spite of their inferior acoustics (Fagen, 1975). The denial of products and services the public wanted became apparent when the FCC deregulated the customer premises equipment market in the early 70s. Bell rivals promptly brought to market all kinds of “frill” customer premises equipment.

(ii) As measured by the competitive market standard, Bell’s products were over-engineered (Bell System saying: an elephant is a mouse designed by Bell Labs). Bell’s standard product design-life was forty years, and products were designed to be able to take all sorts of abuse without failing. Great stress was placed on reliable, prompt service (another Bell System saying: We provide only one grade of service - superb). (iii) Prices were distorted from the values they would have had in a competitive marketplace. The main effect was that local residential and rural rates were artificially low, while long distance and business rates were artificially high. A famous example is that a single business line was priced at roughly double that of a single residential line, even though both were providing essentially the same service. Generally speaking, ORB was quick to sub-

’ See Zajac (19904. Wenders makes a similar observation. What Zajac calls “old regulated Bell,” Wenders refers to as ”the cartel.” See Wenders (1987).

‘

We use the benchmark of a competitive industry to gauge inefficiencies. Other benchmarks are possible. For example, ORB always justified its actions by Bell Labs’s tremendous scientific and engineering achievements, argued to be the result of the guaranteed funding that ORB’s postal-industrial complex provided. ORB thus advocated a more Schumpeterian, dynamic view of efficiency. However, ORB’s claims can be neither proved nor refuted and are problematic. We avoid this debate in this paper and simply adopt the competitive benchmark. For a further discussion, see Zajac (1990a) and (1998).

17. Toward a Global Telecom Liberalization Theory

289

sidize basic residential service by charging high rates for business, long distance, and “frill” services. (ivl Under regulation, price discrimination had to be justified on the basis of cost differentials, with the burden of proof on the utility, and little price discrimination was practiced. At the same time, demand continually increased. In these circumstances, engineering planning models dictated that high levels of spare capacity be built in order to meet demand with the high quality service expected of Bell. ORB could not act as the airlines have acted after deregulation wherein flights are routinely filled by creative price discrimination. The result was the massive deployment of idle resources. (VI

AT&T suffered from Parkinsonian bloat. Large bureaucracies worked hard at jobs that contributed little to the bottom line and would not have existed in a competitive industry. Evidence for this is that from 1984 to 1990 AT&T reduced its workforce by 25% - from 364,000 to 274,000 employees (Zajac, 1990b). At the same time, AT&T’s output increased.

To our knowledge, the monopoly rents captured by Bell’s employees and the deadweight losses because of its lack of product offerings, over-engineered products, and pricing distortions have never been estimated, but they undoubtedly were enormous. We have focussed on the US experience because we are most familiar with it. But the same kinds of stories can be told of most of the telephone administrations of the world as they reached equilibrium during the mature years of their own “postal-industrial complex.” In each case there was a basic tradeoff. The conferring of benefits on some favored sector of the economy came at the cost of inefficiencies. Those in power had the choice of making the telephone system more efficient or benefiting some favored sector of the population, and thereby favoring themselves by generating more votes in their favor. At equilibrium, these contending forces were in balance. This is the basis of the formal model that we construct in the next section.

b.

The Formal Model

We start by applying Becker’s variant of the Stigler-Peltzman theory of regulation to characterize the equilibrium reached by the postal-industrial complex’, which we will refer to as the SPB model of equilibrium. To build it, we start by assuming that the aim of “politicians” is to maximize “votes.” Here, we use the terms “politician” and “votes” for want of better terms to capture more general ideas. In the US ORB example, the “politicians” were the leadership of the Bell System and the its regulators, while “votes” were either direct votes, as in the election of state regulatory commissioners, or public support for ORB’S policies and actions. In general, rather than the politicians in a western democracy, we can imagine any group of persons that has less than totally despotic control of the telecom industry. In an authoritarian society these could be government officials or bureaucrats. Likewise, rather than wishing to maximize “votes,” the controllers could be trying to maximize wealth or power. As in the US story, our model assumes that “politicians” will obtain “votes” by conferring benefits on some members of society, and that the benefits come at the cost of policies that generate inefficiencies. These can be in the form of simple waste, deadweight losses, bloated bureaucracies, etc. Whatever their form, we assume that they will become apparent to “voters” ’See Stigler (1971), Peltzman (1976) and Becker (1976).

G. Mesher and E. Zajac

290

who will then be motivated to vote against the “politicians.“ In a non-democratic society, resistance to the policy of those in power may take forms other than votes such as a diminution in the size of bribes or a reluctance to pay them, schemes to displace government officials by others less sympathetic to prevailing policies, etc. In symbols, the “politician’s” problem is to maximize V = V(N,I) where V = net “votes” for the “politician” N = net benefits, that is, beneficiaries’ gains less losers’ losses due to inefficiencies I = aggregate inefficiencies (deadweight losses from price distortions, waste, bribes, etc.) W e further assume that the greater the net benefits, the more beneficiaries will be motivated to work to generate “votes” in favor of the “politicians.” Likewise, we assume that the greater the inefficiencies, the more motivated will be those opposed to the “politicians” to work to generate “votes” against them. Mathematically, these assumptions are expressed as: av/aN>o,

V increases with an increase in net benefits, N;

av/aI 0,

Also, the fact that aV/aN > 0 and aV/aI < 0 implies that the gradient of the V function points in the northwest direction. Finally, we note that we can expect the level curves of the V function to be convex. This follows from the fact that along a V level curve we expect that for a given change, AI, we expect that a larger AN will be required to stay on the level curve as I increases. These considerations imply that we have a picture of the V level curves such as that shown in Figure 1.

Inefficiencies (I)

I

I

Figure 1. Level Curves of Net “Votes” for a Hypothetical “Politician”, V(N,I)

17. Toward a Global Telecom Liberalization Theory

291

At the same time, net benefits, N, are related to the inefficiencies, I. At one extreme, if competition prevails, we expect neither benefits to a favored group nor inefficiencies. Both N and I are zero. At the other extreme, if inefficiencies are sufficiently large, losers’ losses will overcome beneficiaries’ benefits, and net benefits, N, will again be zero. This gives us an N(1) function that is shaped like an inverted “U”. W e call this the ‘“(1) hill,” as depicted in Figure 2.

h

k

Inefficiencies (I)

Figure 2. The N(1) “Hill” of Beneficiaries Gains Less Losers’ Losses The combination of Figs. 1 and 2 yields an equilibrium in Figure 3, where the N=N(I) hill is tangent to a level curve of the V function’.

I

~$” G

bi l!qiwunnpt Inefficiencies (I)

Figure 3. Combined Graph of V(N,I) and the N(1) Hill Showing a Tangential Equilibrium Point

Readers familiar with Peltzman’s (1976),paper will note the similarity between our Figure 3 and Peltzman’s Figure 11. Also, readers will note that we have extended Peltzman’s model to take account of Becker’s critique of it (Becker, 1983).

292

c.

G. Mesher and E. Zajac

Implications of the Model

The model is very general and, thus, very crude. Nonetheless, it has some important implications: Implication 1

In equilibrium, “politicians” have no incentive to increase efficiency (decrease inefficiency). An efficiency increase means a left-ward move along the N(1) hill, which in turn means a move to a V level curve of fewer votes.

Implication 2

The equilibrium is locally stable. From Figure 3 it is clear that an increase in inefficiency will also result in fewer votes. So, either an increase or a decrease will lose the “politician” votes.

Implication 3

Any N(1) hill will generate an inefficient equilibrium.

d.

Interpretation of the Model

The mathematics used to derive the model may obscure the intuition underlying it. That intuition, however, is straightforward and reflects our discussion of the US case. The model tries to capture what might be called The Iron Law of Political Economy: Politicians benefit favored groups by creating economic inefficiencies’. The model brings out the fact that there are two basic processes at work: (1) the economic process that allows net benefits to be generated at the sacrifice of efficiencies, and (2) the political process that allows the same number of “votes” (amount of power, amount of wealth) to be generated by various net benefit and inefficiency combinations. In both processes there are tradeoffs between net benefits and efficiency. Locally (in the economist’s jargon, “at the margin”), these tradeoffs or “rates of substitution” are measured by the ratio, AN/AI. Graphically, this ratio is the slope of the N(1) hill or the slope of a level curve of constant number of votes. The model simply says that, in equilibrium, these two slopes (tradeoffs or rates of substitution) must be equal. If they are not, “politicians” will have the possibility of generating more “votes” by trading off efficiencies for “votes.” For example, at a point of the N(1) hill to the left of the equilibrium point in Figure 3, the slope of the V = constant curve is less than the slope of the N(1) hill. By decreasing efficiencies (increasing inefficiencies) by a small amount and thereby buying more “votes”, the “politicians” can hoist themselves to a higher V = constant curve. Likewise, at a point to the right of the equilibrium point, the “politicians” can move down the N(1) hill and to higher V by increasing efficiencies (decreasing inefficiencies).

4.2 Destruction of the Postal-Industrial Complex Equilibrium and Movement Toward a New Equilibrium a.

Location of the Potential New Equilibrium

What happens to the V(N,I) and N(1) functions as a result of the eroding forces described by Noam and augmented by us? The SPB model is general and fundamental. In most societies, we can expect to see the tradeoffs between net benefits and efficiencies at work, and, eventu~

’For a striking, non-telecom list of examples of the Iron Law of Political Economy at work, see Schleifer (1998, pp. 142-143).

17. Toward a Global Telecom Liberalization Theo y

2 93

ally, we expect that a new equilibrium will ensue. In the new equilibrium we expect that competitive forces will prevent inefficiencies and will make whatever inefficiencies that exist more transparent - “politicians” will have a tougher time hiding bribes, incompetence, bloated bureaucracies, etc. This in turn means that, for a given level of inefficiencies, I, net benefits, N will be harder to generate. The result then will be a lowering and flattening of the N(1) curve. Greater transparency also means that, for a given level of inefficiency, I, larger net benefits, N, will have to be generated for the “politicians” to harvest the same number of votes. Further, we expect that “votes”, V, will more sharply fall off with increasing I, that is, for a given (NJ), ldV/dII will be greater than it was in the postal-industrial complex equilibrium. The net result will be that every level curve such as V = V I will move upward and rotate counterclockwise. Putting these arguments together, we conclude that the various eroding forces should result in pressures to move the equilibrium point in Figure 3, toward the origin. This potentially new SPB equilibrium should be a point at lower levels of inefficiency, I, and lower levels of net benefits, N. This, of course, is the touted societal benefit of “liberalization.” It is by no means automatic that the new SPB equilibrium will be attained, and, if attained, how long it will take to attain it. For one, as a preliminary to the attainment of a new SPB equilibrium, we can expect an unraveling of the postal-industrial complex, a process that Noam addressed with his “network tipping” model (Noam, 1992, 1994). For another, pressures to move toward a new SPB equilibrium can be expected to be resisted by those who will lose by movement from the old equilibrium. In turn, we can expect these pressures to be countered by the interests that will gain from the attainment of a new equilibrium. This means that in addition to the various forces identified by Noam and by us moving from the old equilibrium, we have the forces of resistance and counter forces. W e can expect entrenched interests to use these forces in efforts to keep the benefits they have enjoyed. The result is a complicated story of movement away from the postal-industrial complex equilibrium toward a new SPB equilibrium, a story that varies from country to country and depends on the country’s political history and climate. Thus, Vail’s alliance with State regulators and “independent” phone companies almost guaranteed that residential users of local phone service would be a favored group in the USA. Not surprisingly, ORB put great stress on “pricing for development,” with the goal of every home having a telephone. But this was not the case in every country. As late as 1975, France’s “penetration rate” was only about a third of that in the USA, 37 main lines per 100 population in the USA versus 13 in France (Waverman and Sirel, 1997). Likewise, until 1957 Malaysia and Singapore were both part of British Malaya. But, as we have pointed out, political forces have propelled the telecommunications market structures of these two countries along quite different paths. Malaysia has a very competitive structure while Singapore’s main firms are either government-owned or government-linked (Mesher and Zajac, 1997). A further discussion of movement from equilibrium in specific countries takes us beyond the scope of this paper. However, it is useful to consider briefly Charlotte Twight’s theory of institutional change and Gary Becker’s oligopoly model of political competition. b. Resisting Movement Toward a New Equilibrium

Charlotte Twight has written extensively on her theory of institutional change, based on her notion of constitutional-level political transaction costs (Twight, 1994, 1998). These are the

294

G. Mesher and E. Zajac

“costs to individuals of negotiating and enforcing collective political agreements that influence the scope of government authority - the line dividing what is handled by the government from what is regarded as outside its purview” (Twight, 199’8). Her point is that an incumbent government and/or its officials have both the incentive and the ability to heap political transaction costs on those who oppose the implementation of its policies. Twight (1998) states: ...officeholders may mitigate resistance by misrepresenting a bill’s contents, using incremental strategies, tying controversial measures to popular ones, using tax strategies that obscure a program’s cost, and the like. ... Artificially increased political transaction costs thus drive a wedge between voter preferences and political action that reflects those preferences. The theory identifies various determinants influencing an individual officeholder’s decision to favor a transaction-cost-increasing measure: executive and party support for the measure, impact on officeholder job security and perquisites, third-party payoffs, officeholder ideology, the measure’s complexity and perceived importance to constituents, publicity, time, and the existence of an appealing rationale for the measure.

Twight applies her theory to an analysis of the passage of the 1965 act that created Medicare and of the passage of the Health Care Portability and Accountability Act of 1996. In both cases, all of the forces mentioned above were used by the Act’s proponents to win passage. Although Twight has applied her theory to instances where entrenched interests effected change, it obviously also applies to cases where entrenched interests work to block change.

e.

Modeling the Rivalry Between an Entrenched Postal-Industrial Complex and a Coalition of Rivals

The postal-industrial complex in equilibrium is an example of what Wilson (1980) calls “client politics.” The government (in our terminology, the “politicians”) uses its coercive powers to levy a small tax on one group in society in order to favor another. These costs tend to be widely diffused and small so that they generate little or no opposition to their implementation, while the benefits are concentrated on a group that is politically powerful. In the case of telecommunications, the beneficiaries were typically residential users of local services and rural customers, with costs levied on business and long distance customers. This circumstance differs somewhat from client politics in that some residential consumers could also be heavy users of long distance services. Nonetheless, Wilson’s main notion of client politics carries through; the tax was relatively diffused and the favored group was politically powerfUl’0. As we have argued above, however, the rise of rival interests to the postal-industrial complex and the exposure of the inefficiencies inherent in the postal-industrial complex equilib-

I o Wenders (1987, p.154, p. 156) points out that what is important is the cost of becoming informed and taking political action. Unlike the railroad, trucking, or airline industries, telecommunications customers got detailed monthly bills and knew that a politically accessible state regulatory agency determined rates. The costs of residential customers becoming informed and takmg action was thus low. Wilson’s four-way categorization might profitably be modified. What is really important is not whether beneficiaries and victims of a policy change are concentrated or diffused but whether their costs of becoming informed and taking political action are high or low.

17. Toward a Global Telecom Liberalization Theory

2 95

rium has brought about a new situation, one which we feel is more appropriately modeled by Becker’s (1983) oligopoly model of political competition. This model envisions two strong interest groups in political rivalry. Wilson calls this situation “interest group politics”. As is usually assumed in oligopoly models in industrial economics, each group measures the actions of the other and adjusts its actions accordingly. In the case of political rivalry, Becker assumes the actions will consist of the application of “political pressure.” This can consist of letter-writing campaigns to influence legislators’ votes, contributions to political campaigns for a legislator’s election or re-election, promises to bestow government jobs on a legislator’s district, etc. The result is that each group will have an “influence function,” as shown in Figure 4, where for simplicity, the influence function is depicted as a straight “influence line.” As shown, the influence line of Group 2 has a slope of less than unity (dp2/dpl