ASIS Midyear '98 Proceedings

Collaboration Across Boundaries:
Theories, Strategies, and Technology

Collaboration Among Competitors - A First Economic Analysis

Eberhard Stickel
Viadrina University Frankfurt (Oder), Germany


Collaboration between firms that are located in different phases of the value chain is a well-studied problem. Industrial economics provides a rich theory to study reasons for vertical integration or disintegration. Transaction cost theory is the framework usually used. Collaboration between competitors, however, is a much less studied problem. Transaction cost theory only considers costs of contractual arrangements and production costs. Costs and benefits due to competition are not considered. In this paper it is shown how transaction cost theory may be extended, in order to treat effects due to competition between (collaborating) firms. Two possible forms of collaboration are studied, namely information sharing and joint development of information systems. The notion of problem specific knowledge is introduced. Depending on the degree of distribution of problem specific knowledge and the risk attitude of competing firms different predictions about benefits of collaboration are possible.

  1. Introduction
  2. Transaction Cost Theory
  3. Microeconomic Model and Results
  4. Collaboration Incentives
  5. Conclusion


Increased competition together with decreasing margins and lower budgets forces companies to rethink their way of doing business. The efficient collection of relevant information and the efficient use of modern information technology are key success factors in today's world.

The institutional structure of supplier-purchaser-relationships has changed significantly over the last few years (Bauer & Stickel, 1996). More often coordination through markets and collaboration (network organizations, strategic alliances) instead of pure hierarchical coordination may be observed.

The role of IT in this process was investigated in the literature. Malone et al. (1987) propose the so-called 'Move-to-the-Market'-hypothesis. Basically it is stated that complex products are more likely to be obtained through hierarchical than through market coordination due to communication costs. Since IT may decrease communication costs Malone et al. argue that a shift towards market coordination should occur. Also, product specifity should decrease. Unfortunately, the authors do not investigate the specificity of IT-investments. Highly specific IT-investments may allow opportunistic behavior and may incur additional costs (agency costs). This increase may very well exceed the savings with respect to communication costs.

More work on the subject was done by Clemons & Reddi (1994). They proposed the so-called 'Move-to-the-Middle'-hypothesis. It is argued that the use of IT involves a learning curve. Hence, long-term collaboration will be preferred to short-term market arrangements. Long-term collaboration will allow companies to longer enjoy the benefits of their IT-investments. Empirical research basically supports this hypothesis (Bauer & Stickel, 1996, p. 49).

The work of Malone et al. and Clemons & Reddi is, like most other work in this area, concerned with the problem of vertical integration. In this paper issues related to horizontal integration will be studied. It is investigated by means of a microeconomic model, whether there are incentives for competing firms to collaborate. The areas of information sharing and collaborative IT-development are studied in more detail. It should be pointed out, however, that no complete theory is offered. Instead, some observations and directions that may influence emerging theories are sketched.

The paper is organized as follows. Section 2 presents an outline of transaction cost theory. The work on vertical integration heavily relies on the concept of transaction costs. Section 3 presents our microeconomic model and basic results. In section 4 we discuss implications for information sharing and collaborative IT-development. Finally, we conclude in section 5.



Transaction cost theory investigates the way transactions between partners are organized. According to Commons (1931), p. 652 a transaction may be seen as an exchange of property rights on resources. Usually, a transaction is governed by contracts. Transaction costs are costs caused by organizing a transaction (Williamson, 1989, p. 139). The main assumptions underlying transaction cost theory are as follows:

Typical transaction costs are costs of searching optimal partners, costs of negotiating an agreement, costs of safeguarding against opportunistic behavior of partners (e.g. monitoring costs), costs of adapting a contract to a changing environment and costs for closing a contract.

Rational (and bounded rational) companies will choose the organizational structure of their transactions in such a way that the sum of production and transaction costs is minimized (Williamson, 1990, p. 25). Incomplete contracts allow opportunistic behavior of one of the partners. This negative impact increases with the specificity of the investments made. The same holds for increasing uncertainty.

Consequently, markets are more attractive than hierarchies in case of low specificity and uncertainty. In such a case, opportunistic behavior may not provide sufficient benefits. Also, production costs should be lower if non-specific goods are bought in the market place (e.g. scale economies). Finally there is no need for control mechanism as in the case of a pure hierarchy. Collaboration lies exactly between the extreme coordination forms of market and hierarchy. Fig.1 summarizes the main hypotheses of transaction cost theory with respect to vertical integration. Fig.1 suggests that there is an optimal institutional structure for each potential type of transaction.

Horizontal integration is concerned with the design of institutional arrangements on the same level of the value chain. Within this paper we restrict our attention to the case of information sharing and the discussion of collaborative IT-development.

Information collection usually involves more or less high search costs and possibly the design of suitable information systems to evaluate information collected and to facilitate decisions based on it. Search and development costs may be shared by competing firms. Basically, this forms an incentive to collaborate.

In terms of transaction cost theory uncertainty, specificity and opportunism determine the institutional arrangement chosen. Specificity, however, does not seem to be of importance in the cases considered here. First of all, information is not an asset in the usual way. It may be duplicated arbitrarily and it is not destroyed or changed during its use as a resource. Moreover, it does not have any (market) value after it is known. Once information is available to all transaction partners, there is no room for opportunistic behavior like in the case of physical assets.

On the other hand, information may be seen as a strategic resource. In that case a company intends to get a competitive advantage by collecting and evaluating information competitors do not have or cannot have. Hence, opportunity costs of information sharing should be considered. These costs are incurred if otherwise exclusive information is shared among competitors.

The same basically holds for strategic IT-development projects. Very often substantial know-how is necessary in order to build information systems. In case of collaboration there certainly is some sort of transfer of know-how between the companies. This in turn allows opportunistic behavior of one of the participating partners. Consequently, collaboration may not be desirable or not possible at all in such cases. In that case the distribution of knowledge should be analyzed carefully. If relevant knowledge for IT-development is equally distributed over the competing firms the risk of opportunistic behavior is greatly reduced.

Information systems built may be seen as (physical) assets. Hence, the problem of asset specificity has to be addressed. It should be expected that all transaction partners have access to the system developed. Consequently, there is no risk of opportunistic behavior arising from asset specificity in that case. Also, note, that today the development of portable systems is greatly simplified through the use of programming languages like JAVA (Seitz & Stickel, 1997).

To summarize, the desirable degree of collaboration is basically influenced by the distribution and the degree of problem specific knowledge:

If problem specific knowledge is not required, collaboration should be favorable since search and/or development costs may be shared. It cannot be ruled out, however, that collaboration makes sense even if problem specific knowledge is required and unequally distributed. In that case transfer payments may need to be negotiated.

The discussion has shown that basically it should be possible to adapt transaction cost theory to the problem of collaboration between competitors. The microeconomic model of section 3, however, will show, that extensions are necessary. It will be shown that opportunistic behavior in the sense of maximizing returns (utility) at the expense of competitors influences collaboration decisions. This is a dimension that is not yet captured by transaction cost theory (in the case of vertical integration contracts are formed between potential partners and not between competitors).



To simplify the analysis we will restrict ourselves to the case of duopolistic competition. Hence, we have two competing companies F1 and F2. Suppose that both firms are price takers (case of homogeneous duopoly, no product differentiation is considered) and that quantities are set simultaneously. To be specific, assume that both firms face an aggregate demand function given by

p = a - b (x1 + x2 )

where xi represents the quantity offered by firm Fi. Without loss of generality we may set b = 1. The parameter a is assumed to be stochastic. This assumption introduces uncertainty. For the sake of simplicity it is assumed that a is equally distributed on the interval [A,B]. By gathering suitable information both firms are able to predict the parameter a before production decisions for the next period need to be made. Consequently, four scenarios may be considered:

Furthermore, we assume that both firms have the same variable production costs c. Without loss of generality we may then set c=0. Search costs for information collection may be considered fixed during one period (they are independent of the level of production).

We assume that both firms try to behave in a rational manner. One way of rational decision making for firm Fi is to maximize the expected utility Eui ( Ri ) with Ri given by

Ri = pi xi = (a - x1 + x2 ) xi

Hence, utility functions need to be introduced. We assume that both firms have constant risk aversion with degree ai such that a1 < a2 holds. Thus, firm F1 is less risk averse than firm F2. These assumptions lead to the utility functions (Copeland & Weston, 1982, pp 103)

ui ( r ) = 1 - exp( -ai r ).

In Stickel (1997) the model was analyzed for the case of F1 being risk neutral, while F2 was a maxi-miner. In that case closed form solutions of the corresponding optimization problems may be derived. In case of the utility functions used here, exact solutions may be computed numerically only. It is possible, however, to generally prove the statements that are derived from the numerical example presented. Since the focus of this presentation is not on formal derivations the solution path is only briefly sketched. In the first scenario both firms solve the optimization problem

maxxi Eui (xi (a - x1 - x2)).

This yields two reaction functions x1(x2) and x2(x1), respectively. The point of intersection of these reaction functions determines a Cournot-equilibrium (Pindyck & Rubinfeld, 1992, p. 434). After the computation of this equilibrium point expected utilities and expected returns for both firms may be calculated.

In the second scenario firm F1 predicts the stochastic parameter a. Firm F2 does not but anticipates the fact that F1 does so (of course the result of the prediction is not known to F2). Now, F1 maximizes

maxx1 u1 ( x1(ã - x1 - x2 ));

while F2 maximizes

maxx2 Eu2 ( x2 (a - x1(a) - x2 )).

Here ã is the parameter predicted by firm F1. Again the results are two reaction functions from which a Cournot equilibrium may be computed. Note, that the result for firm F1 will depend on the value of the parameter predicted. Hence, expectations need to be taken in order to get general results. Again, we get expected utilities and returns for both firms. The third scenario may be analyzed in the same way as the second one. In the fourth scenario firm Fi solves the optimization problem

maxxi ui ( xi (ã - x1 - x2 )).

This yields two reaction functions x1(x2) and x2(x1(a)). Again, expectations need to be taken to get expected utilities and returns.

In order to motivate our conclusion we consider the following example. Note, however, that the results derived by means of this example are true in general.


Let A = 4, B = 8, a1 = 0.1and a2 = 0.7 be given. We then get the results contained in table 1 and table 2, respectively.

Firm 1/Firm 2
No Prediction
Firm 1/Firm 2
No Prediction
No Prediction
5,78 3,96
No Prediction
3,02 4,42
4,85 4,15
3,72 4,15

Table 1: Expected profits for firm F1 (left) and firm F2 (right); actions of F1 (F2) are shown in rows (columns)

Firm 1/Firm 2
No Prediction
Firm 1/Firm 2
No Prediction
No Prediction
0,420 0,322
No Prediction
0 ,817 0,880
0,366 0,332
0,906 0,910

Table 2: Expected utility for firm F1 (left) and firm F2 (right); actions of F1 (F2) are shown in rows (columns)

Based on these results the following conclusions may be drawn:



So far, we did neither consider search costs nor costs for developing suitable information systems. We also did not take into account the distribution of problem specific knowledge in the two firms. In the following various scenarios will be analyzed.

The results obtained are not surprising in case of equally distributed problem specific knowledge. Both firms have sufficient incentives to collaborate. In that case we may see information sharing strategies, as well as joint application development. Joint data and process modeling projects may be an example of such efforts (see e.g. the case of Germany where insurance companies and banks work together on such modeling tasks). Moreover, it is possible to 'outsource' development efforts to an independent third party. In that case the competitors only provide the problem specific knowledge needed by third party's developers. Without collaboration the involvement of a third party could be dangerous since it may sell results or know-how gained to potential competitors.

The results obtained in case of unequally distributed problem specific knowledge are surprising at least in the case that firm F2 has such knowledge. Why should it be beneficial for this firm to transfer know-how to its competitor? Expected profits of firm F2 decline but its expected utility increases due to reduction of risk (variance of profits). This decrease in uncertainty could be enough incentive for the firm to collaborate. Moreover, search and development costs may be halved (at least).



Transaction cost theory provides the basis for studying collaboration incentives between competitors. The theory needs to be extended, however, in order to capture the effects of competition between rival firms.

In this paper two possible areas of collaboration, namely information sharing and joint development of information systems have been studied. The distribution of problem specific knowledge turns out to be of major importance. If this knowledge is equally distributed, both firms have enough incentives to collaborate.

In case of unequally distributed problem specific knowledge the risk attitudes of the competing firms need to be investigated more closely. In the model presented a less risk averse firm is rewarded for its willingness to take higher risks through higher expected profits. If this firm has more problem specific knowledge it has no incentive to reduce uncertainty which basically is the goal of collaboration in the two cases studied in this paper. Things are different if the more risk averse firm has more problem specific knowledge. It has an incentive to transfer know-how to the competitor with lower knowledge. By doing so it may further reduce uncertainty without losing too much of its expected profits.

The model presented allows to capture only a small portion of possible environmental factors. It should be noted, however, that it may be generalized in various directions without altering the main results. To be specific, the results hold if another distribution of the stochastic parameter a is chosen (e.g. a beta distribution). It is also possible to use a more sophisticated demand function. Basically it is only necessary that prices decline if the quantities offered increase, which is quite a plausible assumption. Also, the assumptions underlying the Cournot equilibrium may be weakened. The results remain the same if there is a firm that announces its quantities set prematurely. The fact that only two competitors have been analyzed is not essential. The model may be easily extended to study an arbitrary oligopoly. The essence of the results presented remains the same, although more cases need to be distinguished. Finally, the results hold in oligopolistic markets where firms set prices (product differentiation is possible). In that case the demand function has prices as inputs and quantities as outputs (Martin, 1988, pp. 131).

It has to be noted, however, that only two areas of collaboration have been studied in this paper. Of course there are other possibilities and areas and it remains to be explored if this requires further extensions of transaction cost theory.



Bauer, S. & Stickel, E (1996). "Contributions of IT to the Genesis of Network Organizations." In Khosrowpour, M. (Ed.), Information Technology Management and Organizational Innovations. Proceedings of the 1996 Information Resources Management Association Conference (pp. 46-52). Harrisburg: Idea Group Publishing.

Clemons, E. & Reddi, S. (1994). "The Impact of I. T. on the Degree of Outsourcing, the Number of Suppliers, and the Duration of Contracts." In Nunamaker, J. & Sprague, R. (Eds.), Proceedings of the 27th Annual HICSS, Vol. IV (pp. 855-864). Hawaii: IEEE.

Commons, J. (1931). Institutional Economics. American Economic Review, 21, 648-657.

Copeland, T. & Weston, J. (1983). Financial Theory and Corporate Policy. 2nd Ed., Reading: Addison Wesley.

Malone, T. & Yates, J. & Benjamin, R. (1987). Electronic Markets and Electronic Hierarchies. CACM, 30, 484-497.

Martin, S. (1988). Industrial Economics. New York: Macmillan.

Pindyck, R. & Rubinfeld, D. (1992). Microeconomics. 2nd Ed., New York: Macmillan.

Seitz, J. & Stickel, E. (1997). "Cooperative Software Development Supporting Financial Product Representation and Distribution in the World-wide Web." In Wojtkowski, G. & Wojtkowski, W. & Wrycza, S. & Zupancic, J. (Eds.), Systems Development Methods for the Next Century, (pp. 345-355) New York: Plenum Press.

Stickel, E. (1997). "Problems Associated with Cost/Benefit Analysis for Strategic Information Systems." In Wojtkowski, G. & Wojtkowski, W. & Wrycza, S. & Zupancic, J. (Eds), Systems Development Methods for the Next Century, (pp. 101-114) New York: Plenum Press.

Williamson, O (1989). "Transaction Cost Economics." In Schmalensee, R. & Willig, R. (Eds.), Handbook of Industrial Organization, Vol. 1, (pp. 135-182). New York: North Holland.

Williamson, O. (1990). Die ökonomischen Institutionen des Kapitalismus - Unternehmen, Märkte, Kooperationen. Tübingen: UTB, in German.

Williamson, O. (1991). "Comparative Economic Organization - Vergleichende ökonomische Organisationstheorie: Die Analyse diskreter Strukturalternativen." In Ordelheide, D. & Rudolph, B. & Büsselmann, E. (Eds.), Betriebswirtschaftslehre und Ökonomische Theorie, (pp. 13-49) Stuttgart: Poeschel, in German.


Paper presented at the 1998 midyear meeting of the Association for Information Science, May 17-20, 1998, Orlando, Florida.

Return to ASIS MY98 Proceedings Table of Contents

Copyright © 1998, Association for Information Science. All rights reserved. No part of this document may be reproduced in any form whatsoever without written permission from the publisher.
The opinions expressed by contributors to this publication do not necessarily reflect the position or official policy of the Association for Information Science.

Last updated 5/14/98

Proceedings edited by Barbara M. Wildemuth.
Conference web pages maintained by Jan White.