and Contract Enforcement 1

Transcription

1 Foreign Direct Investment and Contract Enforcement 1 Zhigang TAO School of Business The University of Hong Kong Susheng WANG Department of Economics Hong Kong University of Science and Technology July, 1998 Suggested running head: Foreign Direct Investment Correspondence to: Susheng WANG, Department of Economics, Hong Kong University of Science and Technology, Clear Water Bay, HONG KONG. Tel: (852) We would like to thank sincerely Editor John Bonin, three anonymous referees and Danyang Xie. Their invaluable comments and suggestions have contributed to substantial improvement of the paper. We would also like to acknowledge nancial support from the Research Grants Council of Hong Kong. 1

2 Abstract A long-standing deterrent to foreign direct investment in developing countries is weak enforcement of binding contracts. A local rm may learn business skills from a cooperating multinational rm and subsequently do business on its own based on the acquired skills. In a two-period doublemoral-hazard model, non-binding contracts are shown to be preferred by all parties, implying that contract enforcement is unnecessary. Our results shed light on the puzzling phenomenon that substantial FDI has been carried out under contractual arrangements in developing countries in which contract enforcement is problematic. They can also explain some interesting stylized facts of contractual joint ventures between multinationals and local rms in the early stage of an economic transition. Keywords: foreign direct investment, contract enforcement, contractual joint ventures, double moral hazard, learning by doing. Journal of Economic Literature Classi cation Numbers: D2, F2, L2. 2

3 1. Introduction In the past three decades, foreign direct investment (FDI) has increased substantially. In particular, during the eighties, FDI worldwide grew faster than GDP and trade by a factor of four and three respectively (Neven and Siotis, 1993). Not surprisingly, almost all outward FDI comes from developed countries. Developing countries are attracting an increasingly larger share of inward FDI. 2 Multinational corporations from developed countries are looking for markets as well as sources of low-cost production in developing countries through either ownership arrangements, e.g., wholly-owned subsidiaries and equity joint ventures (EJVs), or contractual arrangements, e.g., contractual joint ventures (CJVs). In recent years, FDI has played an important role in the economic transition of East Asian countries, East European countries and the former Soviet Union. A longstanding deterrent to FDI 3 in many developing countries is weak contract enforcement. 4 Companies in developing countries try to master trade secrets from their partners through cooperation and conduct business on their own based on the acquired knowledge. Indeed, some governments in developing countries have declared openly that the absorption of superior technologies is a key condition for the approval of a FDI project. Weak contract enforcement, especially regarding the penalty for violation of binding contracts, is thought to deter FDI in developing countries. Evidence reveals that, in developing countries even with poor contract enforcement, a substantial amount of FDI has been carried out under contractual arrangements. For example, when China began to attract FDI in 1979, the law for EJVs was enacted in anticipation that the form of EJV would be widely used by foreign investors. 5 Interestingly, the form of CJV was instead more widely used even though there was no law for CJVs until In fact, from 1980 to 1987, CJVs outnumbered EJVs, see Wang (1992). 2 Developed countries supplied 97 to 99% of the total FDI throughout the period (Hummels and Stern, 1994). On the other hand, developing countries have obtained an increasingly larger share, 21% in 1993 and 40% in 1994, of total inward FDI (Banks, 1995). 3 Other deterrents to FDI include restrictions on foreign ownership and requirements for local content, both of which are beyond the scope of this paper. 4 In this paper, short-term contracts and long-term non-binding contracts on veri able revenues are always enforced. What we mean by weak contract enforcement is that of long-term binding contracts (see (Grout, 1984) for contract enforcement problem of short-term contracts). For ease of exposition, contracts in this paper, both binding and non-binding, are meant to be long-term unless otherwise speci ed. 5 In an EJV, participating rms have equity shares. The venture s pro ts are split among the participating rms according to equity shares that are constant throughout the cooperation unless they are voluntarily exchanged. 6 The de ning feature of CJVs is the absence of equity shares so that the pro ts are split according to the revenue shares speci ed in the contracts. 3

4 Table 1: Contractual Joint Ventures a Company Duration Revenue Share b Technology Stone cutting 10 year 50/50 Equipment from West from 1986 Germany, some from Italy. Dried duck 5 years 55/45 Basically traditional processing from 1985 technology. Confectionery 6 years 50/50 for rst 3 years Packaging machinery, manufacturing from /45 for last 3 years imported from Japan. Footwear 10 years 50/50 Equipment from Japan. manufacturing from 1985 Restaurant unspeci ed 10/90 Hong Kong style from 1984 restaurant supervision. Kitchen equipment 8 years 0/100 for rst 3 years NA for restaurant from /60 for next 2 years manufacturing 50/50 for next 1 year 60/40 for last 2 years. Jewellery 10 years 56/44 Hong Kong designs. manufacturing from 1985 over and above repayment capital to HK side Camera 10 years 30/70 of net pro t NA manufacturing from 1985 a This table is adopted from Thoburn et al. (1990). b Chinese revenue share compared to foreign revenue share. Substantial FDI using contractual arrangements, despite weak contract enforcement, presents a puzzle to economists as well as to legal scholars. Hall (1992) highlights the puzzle in his discussion of economic transition in Asia and Eastern Europe. In an attempt to solve this puzzle, we investigate contractual arrangements for FDI in settings in which the enforcement of binding contracts is problematic. In particular, we consider a foreign rm that provides technology or know-how and a local rm that provides complementary inputs, both involved in a joint two-period project. The 4

5 success of the project in the rst period depends on unveri able e ort by each rm. Furthermore, we consider the possibility that, through cooperation in the rst period, the local rm may learn the necessary skills from the foreign rm, in which case the local rm may conduct business on its own in the second period. 7 If the good technology is provided by the foreign rm, the possibility of learning is determined by the local rm s e ort in the rst period, which is a characterization of learning by doing. The two rms can sign either a two-period binding contract under which the local rm is not allowed to do business on its own upon learning the technology or a twoperiod non-binding contract under which the local rm is allowed to leave the contract arrangement upon learning the technology. Note that the former requires contract enforcement whereas the latter does not. We show that the equilibrium non-binding contract is preferred by both partners to the equilibrium binding contract. Thus, contract enforcement is unnecessary. We also nd that, under the equilibrium nonbinding contract, the foreign rm s revenue share is decreasing or constant over the duration of the cooperation. Interestingly, this theoretical result is consistent with the puzzling empirical fact of decreasing foreign revenue share for CJVs in China, see Table 1 for examples. Learning is a key incentive device in our model of FDI because it depends on endogenously determined e ort in the rst period. Under the non-binding contract, the local rm works harder in the rst period in order to learn the technology and the foreign rm extracts a larger revenue share in the rst period to compensate for its potential second-period loss from learning. On balance, both rms are better o under the non-binding contract. A surprising implication of our results is that contract enforcement is not a prerequisite for FDI in developing countries, especially in the early stage of economic development. The nding sheds light on the puzzle that there has been substantial FDI in developing countries in which contract enforcement is problematic. It may also explain why FDI under CJVs was popular in China despite its weak contract enforcement, at least in the early stage of the country s economic reform. However, we stress that our conclusions depend crucially on some important stylized facts. We consider a scenario in which the foreign rm s know-how does not have any immediate application without some complementary inputs or skills from the local rm. This captures a characteristic feature of FDI in developing countries in which only the local rm has access to certain inputs and resources so that the foreign rm must cooperate with the local rm to obtain the resources. During the cooperation, 7 A survey of CJVs in China reveals that multinational rms generally provide technologies that are not too complex or sophisticated for Chinese rms to learn. See Table 1 for some examples. 5

6 however, it is di cult for the foreign rm to protect its know-how, e.g., management skills and style, market knowledge and experience, from being acquired by the local rm. Indeed, a substantial portion of FDI by foreign rms is for sources of lowcost production; the technologies and know-how involved are conventional and easy to learn, see Table 1 for some examples. Moreover, since the foreign rm s know-how is embedded or di cult to transfer, the local rm needs to work with the foreign rm for a certain time period in order to learn the technology. Such learning-by-doing implies a lag and creates an incentive for the foreign rm to cooperate with the local rm, which is in turn willing to give up a large share of revenue to the foreign rm in the initial period of their cooperation. Our model and its implications are most suitable for the early stage of economic development. In Section 2, FDI is modeled as a two-period moral-hazard problem with learning and various contractual arrangements are discussed. Sections 3 and 4 investigate equilibrium non-binding and binding contracts, respectively. Section 5 presents the main result. Section 6 highlights the special characteristics of the equilibrium contracts. Finally, the paper concludes with some general remarks. 2. The Model We model FDI as a way of pooling complementary inputs between rms from di erent countries. A foreign rm has technology or business know-how and a local rm has some complementary inputs in production and marketing. The two rms undertake jointly a project through a contractual arrangement. The project lasts for two periods. The success of the project in the rst period depends on unveri able e orts from both the foreign rm and the local rm: E 1 from the foreign rm and e 1 from the local rm. For simplicity, the success probability of the project is assumed to be p 1 = E 1 e 1, where E 1 is 0 or 1 and e 1 2 [0, 1]. 8 The foreign rm s e ort is interpreted as providing either good technology, E 1 = 1 with cost C 1 > 0, or bad technology, E 1 = 0 with zero cost. The local rm s e ort is interpreted as developing the foreign rm s technology or know-how into a product with cost c(e 1 ) = β 1+β e β 1+β 1, where β > 0. Since c 0 (e 1 ) = e 1/β 1 and e 1 2 [0, 1], 1/β represents the local rm s productivity. 9 Once successful, the project generates observable and veri able revenue R 1 > 0; otherwise it has zero revenue. 8 The simplifying assumption of discrete E 1 re ects the fact that the multinational rm provides technology or know-how rather than labor e ort. 9 This speci c form is chosen for the simplicity of derivations. Cost can be written as c(e) = Ae γ for a closed-form solution, where A and γ 0. 6

7 Through cooperation in the rst period, the local rm may learn the foreign rm s technology. The probability of doing so is given by φ = ke 1 E 1, where k 2 [0, 1]. 10 If the foreign rm provides good technology, the harder the local rm works in the rst period, the more likely it is to succeed in learning the technology. The probability of learning depends on an exogenous parameter k that may be interpreted as the speed of learning. As k increases from 0 to 1, learning becomes more likely. On the other hand, e 1 and E 1 are endogenously determined in equilibrium. Hence, learning has an endogenous component, i.e., learning by doing, that is based on the e orts expended by the local and foreign rms in the rst period. Our model focuses on the possibility that the local rm will learn the foreign rm s know-how through cooperation and do business on its own afterwards. This situation applies to the early stage of economic development, e.g., the early years of China s economic transition, when foreign rms technologies and know-how are conventional and easy to learn and local rms are given exclusive access to certain inputs and resources by their governments. If the local rm has learned the technology in the rst period, the success probability of the project in the second period, p 2, depends solely on the local rm s e ort in the second period e 2, i.e., p 2 = e 2, where e 2 2 [0, 1]. If the local rm does not learn the technology, the success probability depends again on the foreign rm s e ort, E 2, and the local rm s e ort, e 2, i.e., p 2 = e 2 E 2, where e 2 2 [0, 1] and E 2 = 0 or 1. In the second period, the foreign rm provides either good technology, E 2 = 1 with cost C 2 > 0, or bad technology, E 2 = 0 with zero cost. The local rm incurs cost c(e 2 ). Once successful, the project generates observable and veri able revenue R 2 > 0 in the second period; otherwise it has zero revenue. Without losing generality, 11 assume that the foreign rm designs the contract. In a standard principal-agent model, the principal does not provide any input; if the agent is risk-neutral, the optimal contract has the agent paying a lump-sum fee or franchise fee upfront and then getting all the revenue, see Hart and Holmström (1987). In our model, both the local and the foreign rms must provide costly e orts. In particular, the foreign rm can provide good technology at positive cost or bad technology at zero cost. If the quality of the foreign rm s technology can be veri ed by a third party such as a court, our problem would be subsumed in the standard principal-agent model and the optimal solution would be the franchise fee. The franchise fee constitutes the rst-best solution as the local rm has the maximum incentive to supply e ort while the foreign rm is induced to provide the good technology. 10 This learning structure can be generalized to the Cobb-Douglas form, see Remark 4 in Section See Remark 2 in Section 5 for the case in which the local rm designs the contract. 7

8 Our model is concerned with a situation in which the quality of the foreign rm s technology cannot be veri ed when the technology is delivered. This is characteristic of FDI in developing countries (Marin and Schnitzer, 1995). While the quality of the foreign rm s technology cannot be veri ed, it could be inferred by the local rm. However, the crucial point is that the local rm cannot verify its inference to a third party because the success of the joint project depends on both the quality of the foreign rm s technology and the local rm s e ort. It is impossible for a third party to determine whether a failure of the joint project is due to the foreign rm s bad technology or the local rm s low e ort. In the above situation, the franchise fee solution will not work. The foreign rm would accept the franchise fee and supply bad technology at zero cost, because the foreign rm knows that the local rm cannot prove to a third party that the technology is of low quality. Given the inferiority of the franchise fee solution, share contracts emerge as optimal contracts in our model. 12 Under such contracts, the foreign rm obtains payo s only when the project is successful; thus, it provides good technology if it has a signi cant share. We consider two possible share contracts. First, the foreign rm can write a nonbinding contract that allows the local rm to do business on its own upon learning the technology. Speci cally, if the project is successful in the rst period, the foreign and local rms get X 1 and R 1 X 1, respectively. In the second period, the local rm will work on its own if it has learned the technology; otherwise it will be in both rms interests to continue their cooperation. In the former case, the local rm gets the entire amount R 2 if the project is successful; in the latter case, the foreign rm and the local rm get X 2 and R 2 X 2, respectively, if the project is successful. Second, the foreign rm can write a binding contract that prevents the local rm from doing business on its own upon learning the technology. Speci cally, if the project is successful in the rst period, the foreign and local rms get X 1 and R 1 X 1, respectively. In the second period, regardless of whether or not the local rm has learned the technology, i.e., regardless of whether or not the foreign rm is needed to provide good technology, the foreign and the local rms get X 2 and R 2 X 2, respectively, if the project is successful. 12 We model foreign direct investment as a team moral hazard problem; we would like to thank an anonymous referee for bringing the existing literature to our attention. Holmström (1982) shows that the rst-best outcome can be achieved if contingent violation of the team s budget constraint is allowed and credibly implemented by an outsider who receives the di erence between the total surplus and payo s to the team members. The incentive schemes for achieving rst-best outcomes require the team members to post bonds. Such schemes may not be feasible for foreign direct investment in developing countries in which the local rms face initial capital constraints. The literature also establishes the solution to the team moral hazard problem if the game is repeated inde nitely. Our paper studies scenarios in which weak contract enforcement and poor protection of intellectual property rights make it practically impossible for the foreign and local rms to cooperate inde nitely. 8

9 The di erence between binding and non-binding contracts lies in the rms secondperiod payo s. Under the binding contracts, the foreign rm always gets X 2, even if the local rm has learned the technology in the rst period and the foreign rm s e ort is no longer needed in the second period. In contrast, under the non-binding contracts, the foreign rm gets X 2 only when the local rm has not learned the technology in the rst period and the foreign rm s e ort is still needed in the second period. Clearly, binding contracts require contract enforcement whereas non-binding contracts do not. 3. Non-Binding Contracts 3.1. E ort Choices To derive the equilibrium non-binding contract, we analyze rst the choice of e orts by the two rms for a given contract (X 1, R 1 X 1 ; X 2, R 2 X 2 ). In the second period, if the local rm has learned the technology, the success of the project depends only on the local rm s e ort. Under non-binding contracts, the local rm will choose to work on its own to capture all the revenue. The local rm chooses e 2 to maximize its second-period pro t: 13 π 2,F max e 2 0 e 2R 2 c(e 2 ), where c(e 2 ) = β 1+β 1+β e β 2. The rst-order condition is R 2 = c 0 (e 2 ) = e 1 β 2, which yields e 2,F = Rβ 2, π 2,F = β (e 2,F ) 1+β β. This solution will be shown to be the rst-best solution, so it is denoted by the subscript F. If the local rm has not learned the technology, it may be bene cial for both rms to continue the cooperation in the second period. Speci cally, given the foreign rm s e ort E 2, the local rm chooses an e ort to maximize its second-period pro t: ^π 2 max e 2 0 e 2E 2 (R 2 X 2 ) c(e 2 ), (3.1) 13 Throughout this paper, we use π i to denote the local rm s pro t in period i and i to denote the foreign rm s pro t in period i. 9

10 which yields 8 < [E 2 (R 2 X 2 )] β, if X 2 R 2, ^e 2 = : 0, otherwise, ^π 2 = 1 1+β 1 + β ^e β 2. On the other hand, given the local rm s e ort e 2, the foreign rm s expected pro t is 2 = E 2 (e 2 X 2 C 2 ). The foreign rm would provide good technology, ^E2 = 1, if and only if e 2 X 2 C 2 0, otherwise ^E 2 = 0. Thus, there are two possible Nash equilibria, depending on parameter values (R 2, β, C 2 ) and X 2. If X 2 R 2, and (3.2a) (R 2 X 2 ) β X 2 C 2 0, (3.2b) the Nash equilibrium is ^E 2 = 1, ^e 2 = (R 2 X 2 ) β (3.3) with ^ 2 = ^e 2 X 2 C 2, ^π 2 = 1 1+β 1 + β ^e β 2 ; otherwise the Nash equilibrium is ^E 2 = 0 and ^e 2 = 0, with ^ 2 = 0 and ^π 2 = 0. The local and foreign rms choose their e orts noncooperatively and simultaneously, and their e ort levels depend critically on the above conditions. When the conditions are satis ed, the foreign rm has the incentive to provide the good technology and the local rm puts in positive e ort. When the condition is not satis ed, the foreign rm does not provide the good technology and the local rm abandons the project, i.e., there is no cooperation in the second period. In the rst period, the foreign and local rms again choose their e orts simultaneously. Speci cally, given the foreign rm s e ort E 1, the local rm chooses e ort to maximize total pro t: ^π N max e 1 0 e 1E 1 (R 1 X 1 ) c(e 1 ) + δ ke 1 E 1 π 2,F + (1 ke 1E 1 )^π 2, (3.4) where ke 1 E 1 is the probability that the local rm has learned the technology in the rst period, and δ is the discount factor. Here, the subscript N is used to denote the non-binding contract. The solution of (3.4) is: 8 < E1 R1 X 1 + kδ(π 2,F ^e 1 = ^π 2) ª β, if X1 R 1 + kδ(π 2,F ^π 2), : 0, otherwise, 10

11 and ^π N = ^π 1 + δ^π 2, where ^π β 1 + β ^e β 1. On the other hand, given the local rm s e ort e 1, the foreign rm has expected total pro t N = E 1 (e 1 X 1 C 1 ) + δ(1 ke 1 E 1 )^ 2, where E 1 (e 1 X 1 C 1 ) is the foreign rm s expected pro t in the rst period and ^ 2 is the foreign rm s pro t in the second period if the local rm has not learned the technology (with probability 1 ke 1 E 1 ). The foreign rm will provide the good technology, i.e., choose ^E 1 = 1 as opposed to ^E 1 = 0, if and only if or e 1 X 1 C 1 + δ(1 ke 1 )^ 2 δ ^ 2, e 1 (X 1 δk ^ 2 ) C 1 0. Again, there are two possible Nash equilibria, depending on parameter values (R 1, R 2, β, C 1, C 2 ) and X 1. If X 1 R 1 + kδ(π 2,F ^π 2), and (3.5a) R1 X 1 + kδ(π 2,F ^π 2 ) β (X1 δk ^ 2 ) C 1 0, (3.5b) then the Nash equilibrium is ^E 1 = 1, ^e 1 = R 1 X 1 + kδ(π 2,F ^π 2) β, (3.6) with ^ N = ^e 1 X 1 C 1 + δ(1 k^e 1 )^ 2, ^π N = ^π 1 + δ^π 2 ; otherwise the Nash equilibrium is ^E 1 = 0 and ^e 1 = 0, with ^ N = δ ^ 2 and ^π N = δ^π 2. When conditions (3.5a) and (3.5b) do not hold, neither rm has the incentive to cooperate in the rst period. However, the two rms may still cooperate in the second period, as speci ed in (3.1) through (3.3). Thus, the rms total pro ts are not necessarily zero Contract Design We now analyze the design of the contract (X 1, R 1 X 1 ; X 2, R 2 X 2 ) to maximize the foreign rm s pro t. Under conditions (3.2a), (3.2b), (3.5a) and (3.5b), which imply Nash equilibrium e orts of (3.3) in the second period and Nash equilibrium e orts of (3.6) in the rst period, the foreign rm solves: N max X 1, X 2 2( 1, 1) ^e 1X 1 C 1 + δ(1 k^e 1 )^ 2 11

12 which yields: where X 1,N = 1 R1 + kδ(π 2,F 1 + β π 2 + β 2 ), X 2 = R β, µ β µ β e 1,N = β R1 + kδ(π 2,F 1 + β π 2 2 ) β, e βr2 2 =, 1 + β E 1 = 1, E 2 = 1, π N = π 1,N + δπ 2, N = 1,N + δ 2, π 1,N β (e 1,N ) 1+β β, π β (e 2 ) 1+β β, 1,N 1 β (e 1,N ) 1+β β C1, 2 1 β (e 2 ) 1+β β C2. Under the optimal choice of X 1 and X 2, constraints (3.2a) and (3.5a) are automatically satis ed. Constraints (3.5b) and (3.2b) are equivalent to 1,N 0 and 2 0, respectively, which de ne a feasible set of (R 1, R 2, β, C 1, C 2 ) for the equilibrium nonbinding contract. If either 1,N < 0 or 2 < 0, there will be no cooperation in either the rst period or the second period. These cases will be dealt with in Section 5.1. In summary, Lemma 1. Non-binding Contract If 1,N 0 and 2 0, there exists an equilibrium non-binding share contract (X 1,N, R 1 X 1,N ; X 2, R 2 X 2 ) with e orts e 1,N, e 2, E 1, E 2 and pro ts π N and N. The foreign rm s expected rst-period pro t is e 1,N X 1,N C 1 = 1,N + δke 1,N 2, and its expected discounted second-period pro t is δ 2 δke 1,N 2. As δke 1,N 2 is the foreign rm s expected second-period loss from learning, 1,N and 2 can be interpreted as the learning-free pro ts. Thus, under the equilibrium non-binding contract, the foreign rm extracts a larger revenue share in the rst period to compensate for its expected second-period loss from learning. Notice that, in the rst period, the local rm s payment to the foreign rm may exceed the total revenue (X 1,N > R 1) when the local rm learns fast (with a large k). Hence, the local rm may need liquidity to pay the foreign rm in the rst period. Compared with borrowing money to pay a franchise fee for a technology of uncertain quality, it should be easier if the revenue is realized in the rst period and more revenue is expected in the second period. 12

13 4. Binding Contracts 4.1. E ort Choices To derive the equilibrium binding contract, we analyze rst the choice of e orts by the two rms for a given contract (X 1, R 1 X 1 ; X 2, R 2 X 2 ). In the second period, if the local rm has learned the technology, the success of the project depends only on its own e ort. However, as the contract is binding, the local rm only receives a share of the revenue R 2 X 2. The local rm thus solves: ^π 0 2 = max e 2 0 e 2(R 2 X 2 ) c(e 2 ), which gives 8 < (R ^e 0 2 X 2 ) β, if X 2 R 2, 2 = : 0, otherwise, ^π 0 2 = β (^e0 2) 1+β β. As the foreign rm does not need to provide the technology, 14 its second-period pro t is ^ 0 2 = ^e0 2 X 2. If the local rm has not learned the technology, the success of the project depends on both rms e orts. The analysis is the same as Section 3.1, with the outcomes given by ^e 2, ^E2, ^π 2 and ^ In the rst period, given the foreign rm s e ort E 1, the local rm chooses an e ort to maximize its total pro t: which yields ^π B max e 1 0 e 1E 1 (R 1 X 1 ) c(e 1 ) + δ[ke 1 E 1^π (1 ke 1E 1 )^π 2 ], 8 < [E 1 (R 1 X 1 ) + δke 1 (^π 0 2 ^e 1 = ^π 2)] β, if X 1 R 1 + δk(^π 0 2 ^π 2), : 0, otherwise, and ^π B = ^π 1 + δ^π 2, where ^π β 1+β β ^e Under our speci cation of the local rm s payo, the local rm has no interest in requesting unnecessary e ort from the foreign rm. 15 Note that ^π 0 2 and ^π 2 have same mathematical formula. However, the former is always obtainable while the latter depends on the conditions (3.2a) and (3.2b). 13

14 Here, the subscript B is used to denote the binding contract. On the other hand, given the local rm s e ort e 1, the foreign rm has expected total pro t given by: i B = E 1 (e 1 X 1 C 1 ) + δ hke 1 E 1 ^ 02 + (1 ke 1 E 1 )^ 2, where E 1 (e 1 X 1 C 1 ) is the foreign rm s expected pro t in the rst period, ^ 0 2 = ^e 0 2X 2 is the foreign rm s pro t in the second period if the local rm has learned the technology (with probability ke 1 E 1 ), and ^ 2 = ^E 2 (^e 2 X 2 C 2 ) is the foreign rm s pro t in the second period if the local rm has not learned the technology (with probability 1 ke 1 E 1 ). The foreign rm will provide the good technology, i.e., choose ^E 1 = 1 as opposed to ^E 1 = 0, if and only if i e 1 X 1 C 1 + δ hke 1 ^ 02 + (1 ke 1 )^ 2 δ ^ 2, or e 1 [X 1 + δk(^ 0 2 ^ 2 )] C 1 0. Again, there are two possible Nash equilibria, depending on parameter values (R 1, R 2, β, C 1, C 2 ) and X 1 and X 2. If X 1 R 1 + δk(^π 0 2 ^π 2 ), and (4.1a) (R 1 X 1 ) β [X 1 + δk(^ 0 2 ^ 2 )] C 1 0, (4.1b) the Nash equilibrium is ^E 1 = 1, ^e 1 = (R 1 X 1 ) β, (4.2) with i ^ B = ^e 1 X 1 C 1 + δ hk^e 1 ^ 02 + (1 k^e 1 )^ 2, ^π B = ^π 1 + δ^π 2 ; otherwise the Nash equilibrium is ^E 1 = 0 and ^e 1 = 0, i.e., there is no cooperation in the rst period Contract Design We now analyze the design of the contract (X 1, R 1 X 1 ; X 2, R 2 X 2 ) to maximize the foreign rm s total pro t. Under conditions (3.2a), (3.2b), (4.1a) and (4.1b), which imply Nash equilibrium e orts of (3.3) in the second period and Nash equilibrium e orts of (4.2) in the rst period, the foreign rm solves: i B max ^e 1X 1 C 1 + δ hk^e 1 ^ 02 + (1 k^e 1 )^ 2 X 1, X 2 2( 1, 1) 14

15 which yields: X 1,B = R 1 kδβc 2, 1 + β X 2,B = X 2, µ β β e 1,B = (R 1 + kδc 2 ) β, 1 + β e 2,B = e 2, E 1 = 1, E 2 = 1, π B = π 1,B + δπ 2, B = 1,B + δ 2, where π 2, e 2, X 2 and 2 are de ned in Lemma 1, and π 1,B β (e 1,B) 1+β β, 1,B 1 β (e 1,B) 1+β β C 1. Under the optimal choice of X 1 and X 2, constraints (3.2a) and (4.1a) are automatically satis ed. Constraints (4.1b) and (3.2b) are equivalent to 1,B 0 and 2 0, respectively, which de ne a feasible set of (R 1, R 2, β, C 1, C 2 ) for the equilibrium binding contract. If either 1,B < 0 or 2 < 0, there will be no cooperation in either the rst period or the second period. These cases will be dealt with in Section 5.1. In summary, Lemma 2. Binding Contract If 1,B 0 and 2 0, there exists an equilibrium binding contract (X 1,B, R 1 X 1,B ; X 2, R 2 X 2) with e orts e 1,B, e 2, E 1, E 2 and pro ts π B and B. Here, the foreign rm s expected rst-period pro t is e 1,B X 1,B C 1 = 1,B δke 1,B C 2, and its expected second-period pro t is δ 2 + δke 1,B C 2. As δke 1,B C 2 is the foreign rm s expected second-period gain from learning, 1,B and 2 can be interpreted as the learning-free pro ts. That is, under the equilibrium binding contract, the foreign rm reduces its rst-period revenue share by the amount of its expected second-period gain from learning. 5. The Dominant Contract This section investigates the rms contractual preferences, given any combination of parameter values. To do that, we rst discuss the possibility of one-period contracts and the characteristics of the rst-best solution. 15

16 5.1. One-Period Contracts As shown in Lemmas 1 and 2, there exist equilibrium two-period binding and nonbinding contracts if 2 0, 1,N 0 and 1,B 0. However, under those conditions, it is also feasible for the rms to cooperate in one period but not in the other. We call such a contractual relation a one-period contract. Moreover, when the above conditions fail, one-period contracts prevail except in a special case where no contract exists. Lemma 3. One-Period Contracts (i) If 1st 0, there exists an equilibrium one-period contract (X 1,1st, R 1 X 1,1st) in the rst period with e orts e 1,1st and E 1,1st and pro ts π 1st and 1st, where X 1,1st = R β + kδ 1 + β π 2,F, e 1,1st = µ βr1 1 + β + kδβ 1 + β π 2,F E 1,1st = 1, π 1st = β (e 1,1st ) 1+β β, 1st = 1 β (e 1,1st ) 1+β β C 1, β, and π 2,F is de ned in Section 3.1. (ii) If 2 0, there exists an equilibrium one-period contract (X 2, R 2 X 2) in the second period with e orts e 2 and E 2 and pro ts π 2 and 2, where X 2, e 2, E 2, π 2 and 2 are de ned in Lemma 1. The derivations of these two contracts, together with the parameter conditions, are similar to those in Section 3 or 4. Note that, in our model, the local rm needs to work with the foreign rm for one period in order to learn the technology, which means that contract enforcement is unnecessary for implementing one-period contracts First-Best Solution The rst-best solution is considered here as the benchmark. The foreign rm is assumed to provide good technology when necessary and the local rm is assumed to 16

17 put forth rst-best e ort, thereby maximizing joint pro t. How the rms divide their total joint pro t is of no concern. Denote e 1,F (R 1 + δkc 2 ) β, e 1 (R 1 + δkπ 2,F )β, e 2,F = Rβ 2, and J 1,F β (e 1,F ) 1+β β C 1, J β (e 1) 1+β β C 1, J 2,F β (e 2,F ) 1+β β C 2. As will be shown, J 2,F is the joint pro t in the second period; and J 1,F pro t in the rst period if J 2,F 0, while J 1 is the joint is the joint pro t in the rst period if J 2,F < 0. Similarly, e 2,F is the local rm s second-period e ort; and e 1,F rm s rst-period e ort if J 2,F 0, while e 1 J 2,F < 0. is the local is the local rm s rst-period e ort if Lemma 4. First-Best Solution (a) If J 1,F 0 and J 2,F 0, the rst-best solution is a two-period cooperation with e orts e 1,F and e 2,F and total joint pro t J 1,F + δj 2,F. (b) If J 1,F < 0 and J 2,F 0, the rst-best solution is a one-period cooperation in the second period with e ort e 2,F and total joint pro t J 2,F. (c) If J 1 0 and J 2,F < 0, the rst-best solution is a one-period cooperation in the rst period with e ort e 1 and total joint pro t J 1. (d) If J 1 < 0 and J 2,F < 0, the rst-best solution is no cooperation. The second-period e orts of the share contracts are generally lower than in the rst-best solution because the local rm has to share revenue with the foreign rm. The exception is that, under the non-binding contract, the local rm can work on its own and apply the rst-best e ort if it learns the technology. The rst-period e orts of the share contracts are, however, not generally lower than in the rst-best solution. In expectation of a future pro t, the local rm has the incentive to learn the technology (under the non-binding contract) or it is given the incentive to work hard by the foreign rm (under the binding contract). In essence, e orts across periods under the share contracts are distorted by moral hazard. 17

18 5.3. Main Result Since e 1 and e 2 are part of the success probabilities, we need e 1 and e 2 to each be no greater than one. To guarantee this, we suggest two simple su cient, but not necessary, conditions in the following lemma. Lemma 5. If R 1 + C 2 1 and R 2 1, then e 1,N, e 1,B, e 1,1st, e 1, e 1,F, e 2,F, e 2 each no greater than one. are Let us now investigate the rms preferences over equilibrium binding, non-binding, and one-period contracts. We call a contract the dominant contract if it is preferred by both rms to all other contracts. Proposition 1. Dominant Contracts (1) If 1,N 0 and 2 0, the dominant contract is the non-binding contract in Lemma 1. (2) If 1,N < 0 but 2 0, the dominant contract is the one-period contract in the second period in Lemma 3. (3) If 1st 0 but 2 < 0, the dominant contract is the one-period contract in the rst period in Lemma 3. (4) If 1st < 0 and 2 < 0, there is no feasible contract. Proposition 1 establishes the rms preferences for all possible parameter values. That is, any combination (R 1, R 2, β, C 1, C 2 ) of parameter values will be in one of the four cases in Proposition 1. Proposition 1 shows that the dominant contract must be either the non-binding contract, case (1), or the one-period contracts, cases (2) and (3). Recall that neither non-binding contracts nor one-period contracts need contract enforcement. Therefore, Proposition 1 implies that contract enforcement is unnecessary for all parameter values. This may explain why FDI has not been deterred by weak or non-existent contract enforcement in some developing countries. It may also explain why a great deal of FDI, especially in the early stage of an economic transition, has been carried out under contractual forms in those countries. Example. By choosing three sets of parameter values corresponding to the rst three cases in Proposition 1, we calculate the dominant contracts: 18

19 Table 2. Dominant Contracts a Parameters R 1 = 0.7, R 2 = 0.9 b R 1 = 0.1, R 2 = 1 R 1 = 1, R 2 = 0.1 Non-Binding Contract One-Period Contract One-Period Contract e 1,N = 0.57 e 1,1st = 0.58 Dominant X 1,N = 0.67 X 1,1st = 0.68 Contract e 2 = 0.55 e 2 = 0.58 X 2 = 0.6 X 2 = 0.67 π N = 0.22 π 2 = 0.13 π 1st = 0.13 N = = st = 0.19 Conditions 1,N = 0.17, 2 = ,N = 0.11, 2 = st = 0.19, 2 = 0.29 a C 1 = 0.2, C 2 = 0.3, k = 0.7, δ = 0.9 and β = 0.5. b Under the same parameter values, the binding contract yields: e 1,B = 0.54, X 1,B = 0.40, e 2 = 0.55, X 2 = 0.6, π B = 0.21, B = Clearly, both rms get larger pro ts under the non-binding contract than under the binding contract. Remark 1. Simulations reveal that case (2) is likely to happen when R 2 /R 1 large, and case (3) is likely to happen when R 1 /R 2 is very is very large. The intermediate case is covered by case (1). The case with high relative costs C 1 /R 1 and C 2 /R 2 is covered by case (4). Remark 2. If the local rm is to write the contract, it will cooperate with the foreign rm only when it has not learned the technology and will o er the foreign rm revenue shares such that the expected revenue just covers the costs C 1 and C 2, under which it is in the foreign rm s own interest to provide good technology. The foreign rm will earn zero rent under both the equilibrium binding and non-binding contracts, whereas the local rm prefers the non-binding contract under the same conditions, 1,N 0 and 2 0, as in Proposition 1. Remark 3. We have also considered contingent contracts where the second-period revenue shares depend on the learning outcomes. Implementation of such contracts may involve less transaction costs than that of binding contracts. Even assuming same transaction costs for implementing all types of contracts, we nd that the equilibrium contingent contract is either the non-binding or one-period contract and it dominates the binding contract. Remark 4. A more general model has also been considered, in which p i = e α i Ea i, c(e i) = e γ i, φ = eβ 1 Eb 1, where α, β, γ, a, b > 0. All the results still hold under minor additional conditions. Hence, our conclusions are robust to di erent model speci cations. 19

20 6. Characteristics of CJVs To understand the dominance of the equilibrium non-binding contract, we now study its key characteristics in comparison with the equilibrium binding contract. Among others, we nd that the non-binding contract induces higher e ort incentive from the local rm than the binding contract does. Stylized facts of CJVs in China are presented along with the theoretical results and the implications for contract enforcement are discussed. We rst establish an equivalence result between binding and non-binding contracts when the possibility of learning is ruled out. Characteristic 1. No Learning Without learning, i.e., k = 0, the binding contract is identical to the non-binding contract, with X 1,N R 1 = X 1,B R 1 = X 2 R 2 = β ; e 1,N = e 1,B ; π N = π B, N = B. In this case, the local rm has to cooperate with the foreign rm in each period to carry out the project, and the foreign rm has to provide the technology in each period. Consequently, our two-period model degenerates into two consecutive, oneperiod models and the two rms are indi erent between the binding and non-binding contracts. Therefore, the presence of learning is crucial to contractual preferences in our model. The revenue shares in the presence of learning are characterized in the following. Characteristic 2. Patterns of Revenue Shares With Learning 1. The foreign rm s revenue shares in the second period are always the same under both the binding and non-binding contracts: X 2 R 2 = β. (6.1) 2. Across periods, under the non-binding contract, the foreign rm demands a larger revenue share in the rst period: X 1,N R 1 > X 2 R 2. (6.2) 20

21 3. However, under the binding contract, the foreign rm demands a smaller revenue share in the rst period: X 1,B R 1 < X 2 R 2. (6.3) In the second (or last) period, there are no future gains or losses from learning for either rm, no matter whether the cooperation is governed by the binding or the nonbinding contract. Thus, the second-period revenue shares need not be ne-tuned for the redistribution of pro ts so that they are the same for both contracts. However, in the rst period, there are potential future gains or losses from learning for both rms, depending on whether the cooperation is governed by the binding or the non-binding contract. While the foreign rm does not have direct control over its second-period gains or losses from learning, it can design contracts with optimal revenue shares for the redistribution of pro ts. Under the non-binding contract, the foreign rm may receive nothing in the second period. Thus, in the rst period, the foreign rm demands a larger revenue share. In particular, we can show that X 1,N = 1 1 β e β 1,N + kδ 2. This implies that, in the rst period, the foreign rm receives its expected second-period loss from learning, i.e., kδ 2, in addition to the learning-free share, i.e., 1 1 β e β 1,N In contrast, under the. binding contract, the foreign rm expects to gain from learning in the second period. Thus, in the rst period, the foreign rm accepts a smaller revenue share. In particular, we can show that X 1,B = 1 1 β e β 1,B kδc 2. This implies that, in the rst period, the foreign rm accepts less than the learning-free share. We would like to highlight that, under the non-binding contract, the revenue share for the foreign rm declines over the two periods of cooperation. 16 pattern does not depend on the cost ratio C 1 /C 2, revenue ratio R 1 /R 2 Notice that such a and the discount factor δ. 17 Interestingly, this theoretical result is consistent with the experience with CJVs in China (see Table 1). What is the underlying factor that determines the rms preferences over the two possible contracts? By comparing N with B and π N with π B, we have: 16 In the current model, learning has only two possible outcomes: learn or not learn. More realistically, learning is a continuous process that could be approximated by a sequence of small steps, e.g., learn 0% or 10% in the rst year, learn 10% or 20% in the second year, learn 20% or 30% in the third year, and so on. Our basic result holds for the more general case. 17 One might think the decreasing revenue share for the foreign rm is a direct consequence of its decreasing costs, and dismiss the result as trivial. Our analysis indicates that more fundamental reasons are responsible for the phenomenon. Indeed, even increasing costs will still imply decreasing revenue shares. 21

22 Characteristic 3. Contractual Preference Suppose k > 0. In the rst period, the local rm always chooses more e ort under the non-binding contract than under the binding contract, i.e., e 1,N > e 1,B. Characteristic 3 highlights the importance of learning as an incentive device. In our model of endogenous learning-by-doing, under the non-binding contract, the prospect of doing business on its own and capturing all future revenue gives the local rm an extra incentive to work hard in the rst period. On the other hand, the foreign rm can extract its expected second-period loss from learning in the rst period as illustrated by Characteristic 2. Because of the increased incentive for the local rm to provide e ort, both rms are better o under the non-binding contract. This rms contractual preference could also be understood in terms of a positive externality. Under the non-binding contract, the prospect of learning and implementing the foreign rm s technology motivates the local rm to work hard during the initial cooperative phase and increase total pro t, in general, and the foreign rm s pro t, in particular. In other words, a positive externality results from the non-binding contract in that the local rm s hard work to learn the technology also has the side e ect of increasing pro ts. However, such a positive externality does not arise under the equilibrium binding contract. The presence of an externality casts serious doubt on the popular policy view attributing a negative incentive for incoming FDI to an unsettled legal environment. An interesting feature of our results is that the preference for the non-binding contract is independent of the revenue ratio R 1 /R 2. This distinguishes our paper from the work of Anton and Yao (1994), who study how an innovator sells his (readily pro table) innovation in the absence of property rights. When dealing with a buyer, the innovator can always threaten to sell the innovation to another buyer and thus create a competitor in the output market. This threat is shown to guarantee some return from the innovation to the innovator when certain conditions concerning revenue are satis ed. In contrast, our results hold even if the foreign rm could contract with another local rm in the second period as represented by a smaller R 2 /R 1. Finally, it is interesting to probe the robustness of the rms contractual preference when it is easier for the local rm to learn the foreign rm s technology as represented by a larger k. Faster learning depends on the nature of the foreign rm s technology or know-how. Local rms may nd it easier to catch up in certain industries. Moreover, k could be a ected by the degree of contract enforcement. Interestingly, the following result shows that, with faster learning, the rms preference for the non-binding contract is strengthened further and the revenue share for the foreign rm declines faster across periods. 22

23 Characteristic 4. Contract Preference Under Faster Learning The pro t differences π N π B and N B di erences X 1,N R 1 X 2 R 2 and X 2 R 2 X 1,B R 1. increase as k increases as do the revenue share Intuitively, with a higher k, the local rm has even more incentive to work hard and learn the technology under the non-binding contract. The foreign rm, on the other hand, can demand an even greater revenue share in the rst period to compensate for its larger expected second-period loss from learning. Thus, both rms will prefer the non-binding contract because the induced positive externality is greater with faster learning. k. The last issue of interest is whether the rms will prefer faster learning, i.e., a larger Characteristic 5. Payo s Under Faster Learning non-binding contract, π N and N, increase as k increases. Both rms pro ts under the Our characterization of k addresses an important issue: the protection of intellectual property rights (IPRs). The impact of IPR protection on FDI has not been examined theoretically until recently. 18 While poor protection of IPRs in developing countries is thought to deter inward FDI, empirical evidence has been inconclusive. In fact, countries in Latin America and East Asia that o er only limited protection of IPRs were the most active places for FDI in the 1980s. 19 In our model, the incentive to learn may be a ected adversely by the protection of IPRs. Characteristic 5 suggests that even the foreign rm prefers less protection of IPRs. Notice that this conclusion relies heavily on a key feature in our model, i.e., a time lag for learning the technology. In our model, the acquisition of foreign technology requires cooperation and sharing of pro ts with the foreign rm in the rst period. The time lag is enough for the foreign rm to compensate for its potential loss of IPRs. Therefore, on balance, the foreign rm prefers compensation by larger initial revenue shares, which have less adverse effect on the local rm s incentive to work, than compensation from future revenues by legal protection. 18 See, for examples, Chin and Grossman (1990) and Diwan and Rodrik (1991), who show that developed and developing countries have di erent preferences concerning IPR protection. 19 The economic environment in a developing country, including cost conditions, market size, quality of work force, infrastructure and macroeconomic conditions, is found to be an important factor determining inward FDI, whereas regulations on FDI are not found to be important factors (United Nations Publication, 1993). 23

24 7. Concluding Remarks Our model addresses a realistic and timely issue of FDI: by cooperating with a foreign rm, a local rm can master new technology and then do business on its own. The non-binding contract is shown to be preferred by both the foreign and the local rm, which implies that neither enforcement of binding contracts nor protection of IPRs is a prerequisite for FDI. Thus, our model can explain the puzzle that signi cant FDI has been carried out under contractual forms in developing countries in which protection of IPRs is weak and contract enforcement is problematic. It also con rms some stylized facts about CJVs including their widespread use in the early stage of an economic transition as is the case in contemporary China. It is interesting to point out that there is little variation among the Visegrad countries (i.e., Poland, Hungary, the Czech Republic and Slovakia) in contract enforcement but much variation in FDI (Peitsch, 1995). Di erences in FDI ow may be due to political and economic stability as well as market size. Those factors can be, to some extent, re ected in the relative sizes of R 1 and R 2 in our model; the prediction of our model for various R 1 /R 2 seems consistent with the above facts. However, the policy implications of our theoretical results for the Visegrad countries depend on careful empirical study of FDI in those countries. Finally, we expect that our model is suitable only for the early stage of an economic transition during which foreign rms know-how and technologies are easy to learn on the one hand, and local rms are given exclusive access to certain inputs and resources by their governments on the other hand. At that stage, learning by doing is an important factor, perhaps the dominant factor, a ecting local rm s incentives. 20 As an economy develops, more sophisticated know-how and technology will be involved, and many other factors will become important. Strict contract enforcement may thus be necessary for the later stage of an economic transition. References [1] Anton, James J., and Yao, Dennis A., Expropriation and Inventions: Appropriable Rents in the Absence of Property Rights. American Economic Review 84, 1: , March For example, the Chinese government recently chose General Motors over Ford in a large joint venture project, simply because General Motors agreed to transfer more technology and thus provide more chances for the local company to learn the technology. 24