This action might not be possible to undo. Are you sure you want to continue?
Volume 30, Issue 4
Intellectual property and antitrust limitations on contract: comment
Matthew John Holian San Jose State University
Neil N. H. Nguyen San Jose State University
In their chapter in Dynamic Competition and Public Policy (2001, Cambridge University Press), Burtis and Kobayashi never defined their model's discount rate, making replicating their simulation results difficult. Through our own simulations, we were able to verify their results when using a discount rate of 0.10. We also identified two new types of equilibria that the authors overlooked, doubling the number of distinct equilibria in the model.
We thank Alliance Business School for providing research support during Holian's visit to Bangalore, India, in July, 2010, Lydia Ortega for useful comments on our earlier draft, and an anonymous referee for excellent suggestions. The usual disclaimer applies. Citation: Matthew John Holian and Neil N. H. Nguyen, (2010) ''Intellectual property and antitrust limitations on contract: comment'', Economics Bulletin, Vol. 30 no.4 pp. 2680-2684. Submitted: Jul 18 2010. Published: October 12, 2010.
In order to generate the simulation results reported in Table 1. We begin by briefly describing the model.05 z Our discussion applies equally to the 2001 published version. 2 . The good news is that we were able to verify their results when using a discount rate of 0. Burtis and Kobayashi (2001) made specific assumptions about how z . We present our own simulation results. it is straightforward to take the first-order conditions for profit maximization. and then to solve the resulting system of equations for the Cournot quantities for each type of firm. solve for reaction functions assuming symmetry. Assuming marginal cost for originator and imitators is aO and a I respectively. where qO and qi are the quantities produced by originators and imitators.com/abstract=210088. Let N O be the number of firms producing original software and N I be the number of firms producing imitation versions of the source code. the profit functions for each type of firm j (where j = O . in the Appendix. and then go through our corrections point by point. I ) equal: Π j = [ P (Q ) − a j ]q j − F j where FO and FI are the fixed costs of original development and imitation. also merit clarification. originators and imitators. affect the cost of producing original and imitation versions of the source code. Two other minor issues. making replicating their simulation results difficult.2 However. the level of copyright protection.4 + 0. At the time the decision to invest in producing an original or imitation version of the source code. Specifically. what is most interesting about this model is that it incorporates aspects of both Cournot and Schumpeterian competition. we have discovered some problems with their presentation. Therefore we undertook our own simulations in an attempt to verify their results. taking the number of distinct equilibria in the model from two to four.1 Introduction Michelle Burtis and Bruce Kobayashi (2001) present an interesting model of the software industry. In the model there are two types of firms. as well as variables that correspond with the strength of copyright and contractual protection for intellectual property. they assumed that FO ( z ) = 0. respectively. the level of contractual protection.10. containing our newly identified equilibria. and k . which seem to be essentially identical. Originators spend resources creating software programs and imitators spend resources copying these programs. Most seriously. and to the working paper version available at http://ssrn. But we also discovered two new types of equilibria that the authors overlooked. In our view. As the discussion surrounding the model demonstrates. discussed below. and Q is the total quantity of software units produced and sold by both types of firms. Market demand is given by P(Q) = 1 − N O qO − N I q I . the discount rate was never defined. the model is rich and relates to many issues in both antitrust and licensing.
further entry by either an original or imitation firm cannot be profitable.01k .10.5 Fourth. there was a minor typo in their equation (25). We will now go through our (five) corrections point by point. Third. The number of firms in the model is endogenously determined. The cells in which these new equilibria exist are also shaded in Table 1. Given the at least one version of the original source code exists. Total welfare must include each originator’s and imitator’s profit in addition to the sum of all generated consumer surpluses. These new equilibria are indicated by an “ * ” in Table 1 in the Appendix. we were able to verify their results when using a discount rate equal to 0. As mentioned above. First.1 + 0. the marginal firm can choose to enter either with its own original version of the source code or with an imitation version of the source code. the discount rate was never defined. a distribution of firms is an equilibrium if the following set of conditions are met. Finally. with the correct notation.10 z aO (k ) = 0. we also identified two new equilibria. the correct equations are: (5) q O = [1 − ( N O − 1) qO − N I q I − aO ] / 2 (6) q I = [1 − N O qO − ( N I − 1) q I − a I ] / 2. . Through our own simulations. 5 4 3 The equation. there were errors in their equations (5) and (6).4 Third.01 + 0. no existing firm of one type can make higher profits by changing its decision on whether to produce an original or imitation version. These equations did not include the marginal costs aO and a I . If we assume that the static per-period profits are collected in perpetuity.01k . we have identified several problems with Burtis and Kobayashi’s (2001) presentation. Second. First.001k a I (k ) = 0. making replicating their simulation results difficult. while we verified that all of their proposed equilibria in fact met the equilibrium conditions. The correct equation for total (or gross) welfare is: (14) TW * = (1 − P * (Q*))( N O qO * + N I q I *) / 2r + N O Π O * + N I Π I * . These are the first-order conditions.01 + 0. the present value of profits net of the costs of authorship or imitation turn out to be Π * = [(1 − a O (1 + N I )+ N I a I ) /(1 + N I + N O )] 2 / r − FO O Π * = [(1 − a I (1 + N O )+ N O a O ) /(1 + N I + N O )] 2 / r − FI I where r is the discount rate.FI ( z ) = 0. all existing firms must expect that the present value of their investment in producing an original or imitation version of the software program is positive. is as we reported above: (25) a I (k ) = 0. Thus. imitation versions cannot exist in the absence of at least one original version of the source code. there were errors in their equation (14). Finally.3 Second.01 − 0.
and Antitrust Issues by J. NI=1. consider first the equilibrium at z=1. B. at z=1. and Kobayashi. 6 To illustrate why these are in fact equilibria. Ellig. This is an equilibrium. where NO. because the originator is making profit equal to 0. that Burtis and Kobayashi (2001) underreported both the number and variety of equilibia. 2.087 (i. . the two additional equilibria that we discovered both have multiple imitator firms. but we also show that it is more interesting—the model supports a wider variety of equilibria than the original authors realized. and in the other case there are three imitators and one originator. k=8. NI=2.001). In particular. our findings indicate not only that their model is correct. further entry by an originator drives originator profit down (to -0. where NO. On the whole. his profit would go down to 0.041). The proof of the second equilibrium we discovered.6 In one case there are two imitators and two originators. Innovation.14. 229-263. (2001) “Intellectual Property and Antitrust Limitations on Contract” in Dynamic Competition and Public Policy: Technology. 2 References Burtis. while further entry by an imitator drives imitator profit down (to -0.Of all of our corrections our fifth. and only one of these had imitator firm entry. follows in a similar manner. M. is in our view our most valuable contribution. both are making positive profits). k=9. if an imitator switches to become an originator.07. his profit would fall to 0. they identified only two types of equilibria. Cambridge University Press: New York. and imitators are making profit equal to 0.139. This is strange for a model of the software industry where imitation is a rather prevalent phenomenon. while if the originator switches to become an imitator.e. However. 3. editor.
Q = .62 NW = 2.255.3* P = 0. Q = .61 NW = 2.256. Q = .75 GW = 4.0 P = . NI=3.74 GW = 4.63 NW = 2. NI=3.020 aO = .0 P = .1 P = .27 NW = 2.74 GW = 4. In this process.10 FO = .0090 IV.75 GW = 4. (z = 0) FI = .83 No Equilibrium At the suggestion of an anonymous referee.255.63 NW = 3.0 P = .040 aO = . Q = .256.070 aO = .0070 VIII. NI=1.75 GW = 4. (k = 4) aI = .62 NW = 3.60 No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium No Equilibrium NO. NI=3.42 NW = 3. (z = 1) FI = .0075 VII.40 I.0 P = . (k = 5) aI = .010 II.0 P = . Q = .0 P = .256. (k = 3) aI = .75 GW = 4.0 P = .61 NW = 2. Q = .30 FO = .83 NO.50 FO = . we attempted to replicate the gross welfare (GW) and net welfare (NW) calculations.81 NO.75 GW = 4. NI=3.2* P = . (z = 3) FI = .40 FO = .0060 No Equilibrium B.75 GW = 4. This can be quickly verified as GW = NW + NOFO + NIFI. NI=3. Q = . NI=3.61 NW = 2.12 NO.82 NO. Q = . Q = . Q = . Q = . Q = . Q = .20 FO = .62 NW = 2.255.0 P = . NI=2. NI=2.0 P = . NI=3.13 NO. (k = 7) aI = .0 P = . Updated simulation results7 A.3 Appendix Table I.97 NO. Q = .249. (k = 6) aI = .0065 IX. Q = .0 P = .080 aO = .63 NW = 3.255.0080 VI. suggesting they forgot to include the discounted present value of consumer surplus.255.97 NO.62 NW = 2.96 NO.255.76 GW = 4. NI=3.090 aO = . we found that Burtis and Kobayashi’s report of GW and NW are off by a magnitude of 10. NI=3.55 No Equilibrium E. Q= .74 GW = 4. 7 .63 NW = 2.87 NO.0 P = . NI=3.12 NO.98 NO.63 NW = 2.74 GW = 4.75 GW = 4.0 P = .50 No Equilibrium D.82 NO.98 NO.74 GW = 4.45 No Equilibrium C (z = 2) FI = .238. (k = 1) aI = . (k = 0) aI = .255.28 NO.13 NO.0095 III. In addition.255. (k = 9) aI = .81 NO.62 NW = 2.256.63 NW = 2.73 GW = 4.060 aO = . (k = 8) aI = .0 P = . (z = 4) FI = .050 aO = .63 NW = 3.255. NI=3.74 GW = 4.75 GW = 4.010 aO = . Q = .267.030 aO = .0085 V. NI=3.74 GW = 4. we discovered the errors in their equations (14) as mentioned in footnote 4 above. NI=3. NI=3.255.
This action might not be possible to undo. Are you sure you want to continue?