PART II THE CENTRAL RESEARCH QUESTIONS BASED ON LITERATURE SURVEY AND INDUSTRY OVERVIEW

Transcription

1 PART II THE CENTRAL RESEARCH QUESTIONS BASED ON LITERATURE SURVEY AND INDUSTRY OVERVIEW The objective of this part is to raise our central research questions pertaining to the technological development of the Indian Pharmaceutical Industry. We begin with a survey of the theoretical and empirical literature on the economics of R&D and technology generation, in chapter 2 and chapter 3 respectively. Although the literature in this area has a distinct bias in favour of developed industrialised economies, which operate at the frontiers of global innovation, there is a considerable amount conceptual and empirical analyses to suggest that minor and adaptive R&D activities in less developed economies have resulted in major economic gains and have positively contributed to their technological capabilities. Next in chapter 4, we present a brief holistic overview of the growth path of the Indian pharmaceutical industry to set the stage for raising our central research questions, to be addressed in this thesis. 9 I

2 Chapter 2 REVIEW OF THEORETICAL LITERATURE 2.1 INTRODUCTION The theoretical foundations of technological progress can be traced back to the writings of the classical economists like Adam Smith and Karl Marx as early as the 18 1 h and 19 1 h centuries. However, the classical perspectives on technology were formalised only in the writings of Schumpeter, who incorporated the concept of technological progress into a more formal structure of economic analysis in the context of a market driven capitalist production system. The Schumpeterian theory constitutes the centrepiece of later theoretical developments on technology. However, one main cornerstone of this theory is its reliance on technological breakthroughs and major innovations as the main source of economic gains through technological progress. The importance of diffusion and minor innovations has been recognised and incorporated in the economic literature much later, particularly after the emerg~nce of Japan as an economic superpower. 1 We divide this chapter in two broad groups, one discussing the theories of major innovations and the other presenting a conceptual and theoretical framework for diffusion and minor innovations. In this context we also discuss the development of an economic literature particularly addressing the problems of technology generation in less developed economies. At the outset, however, we present an overview of the concept of research and development, which lies at the core of the entire process of technology generation in any economy. The chapter is organised in the following way. Section 2.2 gives an introductory note to research and development. Section 2.3 discusses the various strands of economic literature on major innovations. Section 2.4 presents a theoretical framework for diffusion 1 See Rosenberg and Steinmuller ( 1988). 10

3 and minor innovation. Section 2.5 summarises the literature on R&D and technology generation literature in less developed countries. The final section 2.6 is for an overall summary and conclusion. 2.2 RESEARCH & DEVELOPMENT (R&D): An Introduction R&D led technology generation should ideally be viewed as a production process, where R&D effort coupled with other research inputs are transformed into research outputs, like invention, innovation or diffusion. However, it is important to note that the production process of knowledge through R&D is distinct from the process of commodity production because of three unique and inherent characteristics of the former. We would like to highlight on these three unique characteristics of R&D process at the very outset. Appropriability: The outcome of an R&D process can be described as invention, innovation, diffusion or imitation depending on the field of use and the nature of the output' itself. 2 These outputs of R&D may, however, be broadly defined as generation of new knowledge. To understand the problems of appropriability, we make a distinction between two forms of R&D output: (I) output with no physical element (i.e. new information) and (2) output with physically visible element (i.e. new machines). In case of the former, one encounters serious problems of appropriability. Traditionally "knowledge" was treated as a pure public good because of its nonexhaustible nature- knowledge acquired by an individual is not reduced by sharing it with others. But unlike a pure public good, a large amount of private investment is visible in the field of research and development. Thus it must bear some private good 2 These concepts will be elaborated later. 11

4 characteristics.as well. Therefore, the usual appropriability conditions must hold for effective marketing of knowledge. The sale of new knowledge becomes problematic because of asymmetric information. While the buyer of new information is not ready to pay the price without knowing what is he buying, the seller can not completely reveal this information as the buyer then will have no incentive to pay for it. The seller, therefore, has to give some signals about the nature of the new knowledge. IPR Regime and Appropriability: Patents are often regarded as a mechanism through which the ownership of the new knowledge can be procured only by partially disclosing the features of the R&D output thereby overcoming the problems of appropriability. But it has several nontrivial limitations. It is applicable only for a fraction of all possible R&D activities. For disembodied process innovation and for some product innovations, the scope of inventing around is very high (automobile components, for instance). Firms in such cases do not apply for patents as they apprehend that even the slightest disclosure of specification required to be submitted with the patent application, would make the output vulnerable to imitation. There are certain other kinds of innovation where patent application is not feasible (McDonald's burger, for instance). 3 Accordingly the effectiveness of intellectual property right (IPR) regime will vary from industry to industry. Taylor and Silbertson (1973), for instance, found that the impact of the British patent system on inventive activities was negligible except for the chemical and pharmaceutical industries. There are studies, which address the welfare aspects of patents and IPR regime. Helpman (1993) analysed the issue of IPR in a dynamic general equilibrium model with two regions, North (industrialised countries) and South (less developed countries). 12

5 Innovation takes place in North but is imitated in South. If the global IPR is strengthened, imitation possibilities are lowered and South loses out. Glass and Saggi ( 1995) developed a similar model, but assumed that adaptation to the condition of South is costly for an innovating firm in North. In this framework, a weak IPR of South is welfare improving for both countries if adaptation of the technology is easy. Maskus (1997) pointed out that countries adopt stronger patent regime as their technological capability increases in order to facilitate horizontal transfer of technology. Mansfield (1994) found IPR regime of host \ countries to have a significant impact on the kind of technology transfer, especially in chemical industries. Absence of adequate protection may lead to transfer of older vintages of technologies..,..,... The benefit~, of a weak IPR derived from widespread domestic diffusion of technology are unambiguous. But, a weak IPR is not always costless, since it restricts access to the latest technologies, assuming that the latest technology is important for the country concerned. 4 Risk and uncertainty: Though R&D is often treated as investment, it differs from the investment on commodity production because of its riskiness and uncertainty. This uncertainty can be thought ofhaving two dimensions: external and internal. By internal uncertainty, we refer to the firm's inability to predict its next period's technology level, i.e., the riskiness of the R&D process itself. External uncertainty is associated with generating funds for research activities. For obvious reasons, firms can not reveal detailed information about its R&D programme to a lender. The lender may not have prior knowledge of the proficiency level of the firm and therefore refuses to lend. As a result, firms may have to depend on 3 As a consequence, many other forms of appropriation, like diversification, complementary sales and service promotion have been resorted to. These mechanisms put emphasis on the internal use of the new knowledge by the innovator himself. We will discus some of these mechanisms in due course. 13

6 internal liquidity to finance R&D activities or it may divert loans for production purposes to facilitate R&D. In both cases, the total funds available for production purposes would reduce as a result of which production may suffer. If loans available for production purposes are diverted to other channels of riskier activities like R&D then the firm's creditworthiness may also go down, especially if the success rate in R&D is low. Scale of operation: Another distinctive feature of R&D activities pertains to its scale effect. Unlike capital and labour, R&D is likely to display increasing returns to scale. This perhaps reflects the importance of accumulation of knowledge as an efficiency building parameter in carrying out R&D activities. The argument of increasing returns to scale has been refined by Nelson and Winter (1977) by categorizing it into two different sub-issues. In the first case, research output is characterised by indivisibility. Here a minimum input level is necessary to get any output at all. Secondly, in some cases, R&D inputs come in indivisible units e.g. scientists with particular specialization. Thus a large research team has the scope of finer division of labour. 2.3 THE THEORIES OF MAJOR INNOVATION The Schumpeterian Hypothesis Schumpeterian theory starts from the definition of innovation. Whereas an invention may be defined as the first technological breakthrough in the field of basic science, innovation is defined as the first commercial use of invention. 5 In other words, invention may be identified with basic research (aimed at deriving fundamental knowledge), while innovation is closely linked to applied research and development 4 Often the host country may not be interested to bring in the latest technology. The loss in these cases may not be significant. See Ray et a! ( 1999) for details. 14

7 (associated with engineering and commercial application) as commonly distinguished in the literature. 6 Broadly speaking, an innovation is associated with an invention with a lag. 7 According to Rosenberg (1976), Schumpeter also recognised the importance of another form of R&D output, namely diffusion and imitation that are the subsequent stages of the technological cycle. As opposed to invention and innovation, diffusion or imitation do not constitute any major technological breakthrough, but their role is still important in economics for their ability to generate substantial economic value. As Rosenberg (1976) puts it "though it (diffusion) needs little creativity, but it adds considerable economic value to the society". Surprisingly, however, subsequent theories have added little to relate diffusion with R&D efforts. We shall focus on this issue later in details. Schumpeter explained technological progress as a disequilibriating force that causes the economy to deviate from the circular flow model. The circular flow model depicts the economy in a stationary-state equilibrium with perfect competition (similar to the Walrasian state of equilibrium). Every firm is in perfect equilibrium, costs equal income, prices equal average costs and net profits are zero. The circular flow follows from continuous adoption of small external changes that are absorbed in familiar routines of company behaviour. Through innovation, the economy is driven away from this neighborhood of equilibrium. Then gradually, as the effects of innovation "wears off', the economy settles at a new equilibrium until a second innovation takes place. 8 5 See Rosenberg (1974, 1976). 6 See, for instance, Tirole (1988). 7 Note that invention and innovation can also take place simultaneously. Schumpeter (1975(1942)) suggested that in a mature phase of capitalism, firms are able to endogenise its inventive R&D activities. 8 The Schumpeterian approach gives full credit to the brilliance of individual entrepreneurs in bringing about technological progress. But according to the Marxian framework, social conflict and class interests are believed to be the most important determinants of technological progress. To quote Marx, "A critical history of technology would show how little any of the inventions of the eighteenth century are the work of a single individual." (as cited by Rosenberg (1982), page 6). Marx suggests that technological progress acts as an instmment used by the entrepreneurial class to generate greater profits defin;:d as the surplus of value added over wage bill. 15

8 An innovation results in a transient monopoly structure where the first mover gets a temporary edge over its competitors. Before diffusion takes place, during the transient monopoly phase, the innovator enjoys super normal profits. In the Schumpeterian sense, profit is thus a reward for innovativeness, which gets eroded as diffusion takes place. Ultimately, we return to a perfectly competitive market structure with zero profits. 9 Thus the possibility of a long run monopoly, as opposed to a transient one was considered as a 'rare' occasion in the Schumpeterian framework. In fact, Schumpeter did not believe in any static form of market structure. He viewed capitalism as a dynamic evolutionary process where market structure continues to change, largely driven by 'technological breakthroughs'. He termed this phenomenon as 'creative destruction'. 10 Primarily by technological progress of a firm Schumpeter refers to innovations resulting in new input combinations. Hagedoom (1989) has however stated that in some of Schumpeter's writings, concepts similar to a 'shift' in production function are used to characterise technological progress/ innovation. We reproduce a part of his original text to demonstrate the breath of the Schumpeterian concept of innovation or technological progress as: f. The introduction of a new good -- that is one with which consumers are not yet familiar -- or a new quality of good. 2. The introduction of a new method of production, that is one not yet tested by experience in the branch of manufacture concerned, which need by no means be founded upon a discove1y scientifically new, and can also exist in a new way of handling a commodity commercially. 9 It may be of interest to note that in the Marxian framework profits never get squeezed to zero as it includes the returns to capital in contrast with the Schumpeterian structure where profits are a residual (attributable to innovativeness) and the returns to capital and labour are regarded as part of income. 10 Strassmann ( 1959) criticises this viewpoint. He cites example when inspite of "major innovations", industries with two generations of technologies coexisted. 16

9 3. The opening of a new market, that is a market into which the particular branch of manufacture of the count1y in question has not previously entered, whether or not that market has existed before. 4. The conquest of a new source of supply of raw materials or half-manufactured goods. again irrespective of whether this source already exists or whether it has first to be created. 5. The canying out of the new organization of any indust1y, like the creation of a monopoly position (for example through trustification) or the breaking up of a monopoly position. 11 This is sufficiently broad a concept of technological progress capable of representing technical change, diversification and a change in the element of market structure (through creative destruction). Schumpeter's theory, however, is often narrowly construed in the later literature only to explain the link between market structure and R&D, and has thus failed to appreciate the all-pervasive nature of the Schumpeterian argument. Schumpeter's view that monopoly situations and R&D are intimately relat,ed may be traced to two distinct, but apparently contradictory, arguments. First, monopolies are the natural breeding ground for R&D, and secondly, its creation is a necessarily evi1. 12 To understand why monopoly firms may supply more innovation than competitive firms, we quote Schumpeter: 13 "There are superior methods available to a monopolist which either are not available at all to a crowd of competitors or are not available to them readily: for there are advantages which. though not strictly unattainable at the competitive level of enterprise, are as a matter of fact secured only on the monopoly level, for instance, because monopolisation may increase the 11 Schumpeter (1980(1934)), pp 66 and Schumpeter (1939) pp as cited in Hagedoorn (1989). 12 Monopoly promotes appropriation of new knowledge, but at the cost of its wider use. Hence, 1t IS regarded as an evil. There is a clear trade-off between the dynamic gains from new knowledge and the static loss resulting from a lower monopoly output. 13 It is often suggested that it was with the publication of Capitalism, Socialism and Democracy in 1942 that the importance of large firms enjoying monopoly power as engines of technological progress was recognised for the first time. This was a deviation from the spiri~ of Schumpeter's earlier writtings (Schurnpeter 1934, 1939) where new and small firms were posited as more conducive to technological progress. 17

10 sphere of influences of better and reduces the sphere of influences of inferior, brains. or because the monopoly enjoys a disproportionately higher financial standing. " 14 In other words, large firms are better qualified, or perhaps more eager to undertake R&D than smaller firms because of following reasons. First, R&D ts characterised by increasing returns to scale which a large firm can exploit better. Second, since R&D activity involves a high level of risk that is difficult to eliminate with insurance (for reasons of moral hazard), large firms may be more willing to take these risks as it can diversify this risk over a wide range of product lines. Third, the production pattern in a large firm is more systematic and routinised making it easier for them to implement a new innovation. Fourth, a monopolist does not have any ready competitor who can imitate his innovation or can circumvent an existing patent on his innovation. According to Tirole (1988), however, Schumpeter remains vague about whether the above arguments are associated with monopoly power or with bigness of firms. In this context, Dasgupta (1986) argues that "Schumpeter seems to be saying that large ~ize has distinct advantage" whereas "elsewhere in the same page he recognises that monopoly and large firm size are not identical". In most of the empirical work, the Schumpeterian theory has been broadly interpreted as stating that "bigness and fewness encourage technological advancement". 15 According to Dasgupta (1986) the lack of clarity of Schumpeter's writing was partly responsible for most of the empirical research attempting to establish a unidirectional relationship between market structure and research effort, ignoring the process of creative destruction. Fisher and Temin (1973) argued that the true Schumpeterian hypothesis is not sufficient to ensure that bigness or fewness encourages technological advancement. This view is shared by Dasgupta ( 1986). 14 Schumpeter (1986 (1954)) pp Fisher and Temin (1973) 18

11 To analyse the role of firm size in R&D, Fisher and Temin (1973) proposed to examine the responsiveness of R&D output to firm size defined as: 0 = (S/RF)8(RF)/ as, where Sis the number of non-r&d work force, R is the number of workforce engaged in R&D, F(R, N) denotes the value of per worker output from R&D expressed as a function ofr and N. Where N is the total number of workers in a firm, A value of 0 > 1, according to Fisher & Temin, is consistent with the "broad Schumpeterian hypothesis". However, given that 0 = (S/R)8R/8S + (S/F)8F/8S, whether value of the research output will increase more than proportionately with firm size will depend "not only on whether R&D inputs increase but also on whether output per unit input does not decrease with firm size". Moreover, they also argue that interpreting the "true" Schumpeterian hypotheses as 8F/8R >0 and 8F/8N > 0 (i.e. large R&D departments are more productive than small (given firm size), and a given R&D spending yields a greater return per worker in large firms than in the smaller ones), 16 does not guarantee 0 > 1 required for the "broad Schumpetrian hypotheses" to hold. Accordingly Stoneman (1983) suggests that if Fischer and Temins's (1973) interpretation is accepted, then the "broad Schumpetrian hypotheses" is a misleading derivation of Schumpeter's view. 17 According to Nelson and Winter (1977), the Schumpeterian hypotheses involve the demand as well as the supply side of new knowledge generation. While the supply 16 According to Fisher and Temin ( 1973 ), these two relations constitute the so-called "true" Schumpeterian Hypothesis. 17 Fisher & Temin's (1973) interpretation has been challenged and modified by several authors like Rodriguez ( 1979) and Kohn & Scott ( 1982). However, their basic contention remains unaltered. 19

12 side deals with firm size and the returns to scale in research production, 18 the demand side focuses on the incentives to produce new knowledge. Unfortunately, most of the studies have considered only the supply side, "and that too only in a limited manner". 19 To analyse the demand side factor, Nelson and Winter (1977) have qualified the concept of monopoly power by distinguishing monopoly power in product market and monopoly power in R&D output market. Both are important in determining the rewards of innovative activity. In fact, according to Nelson, what might have played a crucial role behind the Schumpeterian proposition that "monopolies are the natural breeding ground for innovation", was perhaps the belief that information monopoly is not secured enough. Therefore monopoly in the product market was sought as an alternative mechanism to appropriate the output of research and recover at least a part of the cost of innovation. However if information monopoly is completely secure, then the above arguments supporting the Schumpeterian Hypotheses become weak and vulnerable The Neoclassical Models: Search for an ideal market structure Indeed along this line, it was Arrow (1962a), who posed the first major theoretical challenge to the Schumpeterian Hypothesis. Arrow showed that it is the perfectly competitive market structure, and not monopoly, which is more conducive to R&D. This result spurred off a parallel line of modelling, which is seen in the literature as an antithesis to what Schumpeter had postulated. Arrow compared the incentives for innovation in three kinds of market set ups, namely socially managed economy, competitive economy, and monopoly. His findings suggest that due to secured position in the product market, a monopolist has generally less 18 Nelson, unlike other, acknowledged the Scumpeterian process of creative destruction and suggests that size distribution of firm can not be viewed as static. Nelson's view on the returns to scale in R&D has already been discussed earlier. 19 Nelson & Winter (1977) 20

13 incentive for innovation. He found socially managed system as most favourable for R&D activities ranking even higher than a competitive firm. This is because a social planner can trace the source of externalities (like rise in labour productivity due to learning by doing) which an individual firm under a competitive structure can not. Arrow (l962a) essentially compared the surplus that an innovating firm can extract following an innovation under different market forms. He showed that this appropriable surplus is least for a monopolistic producer who therefore has the least incentive to innovate. According to Nelson and Winter (1977), Arrow's (1962a) conclusion is based on the assumption of a completely secured property right in the information market. But this can be questioned. It is now perhaps well-established that information monopoly is not secured, especially in disembodied process innovation, which Arrow had considered. Nelson and Winter (1977), therefore, argue that the incentive for innovation might justifiably lie on appropriating some returns before imitation takes place, which may be ~~~1 facilitated by monopoly power. X X. ( L.-)() :); "-&1-4'-f ' \o I Moreover, Arrow did not consider the threat of potential entrants in a monopoly set up. In this case, the monopolist actually replaces himself as a cheaper production source with innovation, an effect known as replacement effect. However, if there is a threat of potential entry, a monopolist may actually be thrown out of the market if the entrant gets hold of an innovation, which is drastic in nature. 20 This would encourage innovation by the monopolist through the efficiency effect. Therefore, the incentive to innovate in monopoly markets would perhaps be higher than what Arrow had postulated. Stiglitz (1986) criticised Arrow's (1962a) construct because of its partial equilibrium framework. He has also argued that for a more meaningful comparison between different market structures one should characterise two economies by two THESIS Si649 Pr TH5006

14 different market structures rather than dividing total economy into several market structures. He has shown that if labour is taken as the only factor of production and if there are two economies, one characterised by monopoly and the other by perfect competition, then research intensity would depend on the elasticity of labour supply. If we further assume inelastic and identical labour supply (particularly, production labour) in both economies, then pre-innovation output would be the same in both (not less in monopoly). Thus the assumption of cost spreading advantage of the competitive economy (to appropriate the innovation) would not hold any more. 21 Arrow (1962a) deals with one innovation already made. The process leading to that innovation is not analysed. We now discuss some of the later literature, which incorporates this phenomenon. Most of these models assume an oligopolistic market structure, and thereby possibly weakening the seemingly obvious one to one correspondence between market power and monopoly. Dasgupta & Stiglitz (1980) assumed a process innovation with all firms behaving identically. This identical nature of all firms causes diffusion to take place instantaneously. 22 Instantaneous diffusion provides a negative impetus for firms to undertake R&D activities. While Arrow (1962a) considered technological choice to be discrete, Dasgupta and Stiglitz (1980) argue that firms may be interested in marginal (continuous) changes. This could be a marginal decrease in cost or a marginal increase in risk or a marginal increase in the speed of research. They have also criticised the "neo- 20 Drastic innovation are those which result in post innovation monopoly price to be strictly less than the pre-innovation marginal costs, thus driving out all competitors after innovation. See, for instance, Tirole (1988) chapter Stiglitz has also questioned another result of Arrow (1962a), namely the incentive to innovate would be the maximum under a social planner. He argued that to maintain subsidy in R&D, the social planner (effectively government) will have to collect revenue and there could be a distortionary effect of govt revenue collection. In that case the appropriable surplus would be less than what Arrow liad shown. 22 Gomulka ( 1990) argued that if this was not the case, the innovating firm, in the absence of imitation, could have derived a considerable advantage to exploit ~cale economies in R&D. This would have resulted in ti1eir gaining con:>iderable market power which would lead utlimately to a monopolistic market structure. 22

15 Schumpeterian" view, which considers market structure to be exogenous. 23 According to them, market structure and research intensity are simultaneously determined and there are other theoretical dimensions which affect a firm's technology decisions. In particular, demand elasticity, nature of capital market and the cost and nature of R&D have important bearing on firm's technology decision besides concentration and market power. Some ofthe results derived in this paperin the context of R&D in an oligopolistic market structure are discussed below. In an attempt to endogenise market structure, market equilibrium with free entry is assumed. The firms are typical profit maximisers and the number of firms is determined by the zero profit condition at the margin. If the number of firms is large, they show that at the margin (with free entry) industry profit equals total industry R&D expenditure (non zero). This means each firms in equilibrium enjoys a mark up of price over costs and thus holds some amount of monopoly power. This is in line with the Schumpeterian argument that monopoly power is necessary to do R&D. To quote Dasgupta and Stiglitz (1980), "since R&D involves fixed cost we can not expect an industry engaged in it to be characterised by perfect competition." It has also been shown that high concentration is associated with high research intensity, for a given elasticity of demand. In this model, the elasticity of demand does not play any direct role in determining optimum research intensity (Z*) of the industry, but it determines the level of concentration (measured as inverse of the total number of firms). High elasticity would result in smaller number of firms, which therefore indirectly results in a high Z*. Another result of Dasgupta and Stiglitz ( 1980) supports Schmookler' s ( 1962) view that growth in demand serves as a stimulus to R&D activity geared towards cost 23 The neo-schumpeterians are "concerned with a possibh: causal flow... from market structure to technological innovation". (See Scherer (1980) as cited in Dasgupta and Stiglitz (1980) pp 266). 23

16 reduction and quality improvement. They captured the level of demand by the size of the market, and found a positive relationship with firm's R&D expenditure. 24 By assuming entry barrier, Dasgupta and Stiglitz (1980) have also tried to explain whether an increase in the number of firm results in higher cost reduction. Their finding suggests that following an increase in the number of firms total industry research increases but R&D of each firm goes down. This was interpreted as a rise in wasteful, duplicative research activity. In order to explain the relation between market structure and speed of research, Dasgupta and Stiglitz (1980) also constructed a model of patent race in the absence of uncertainty. 25 It is shown that if the first winner is granted patent, 26 and with the possibility of potential entry, active firms in the race would hold Cournot conjecture about themselves but would operate at the entrants reaction curve, signifying Stackelberg conjecture. Ultimately market will be characterised by "at most one firm doing R&D". But it should be noted that this phenomenon (only one firm doing R&D) does not; in any way, reflect that competition is weak. In fact, it is the result of intense competition. The apparent similarity of this result with the Schumpeterian hypothesis is only limited to the creation of transient monopoly that any major innovation brings about. Thus to summarise, this model considers a wide array of theoretical determinants of nature and direction of R&D activity. Although they have endogenised market structure, but the static nature of their model fail.s _to capture the dynamic interaction between R&D and market structure in an effective manner. 27 Though they have acknowledged the role of monopoly power to facilitate R&D, their result is by no means 24 If o: is the elasticity of cost reduction and ~:: is inverse elasticity of demand, then equilibrium solution requires 1:: > o:(l-~::). Inelastic: demand (high~::) is favourable for the equilibrium condition to hold. 25 Models incorporating uncertainty is discussed by Kamien and Schwartz ( 1978). 26 We have already argued that effectiveness of patent protection in process innovation is not unquestionable. 24

17 "Schumpeterian" as the latter explicitly considered a monopolistic market structure as opposed to an increase or decrease in market concentration in an oligopolistic market considered in this model. In fact Dasgupta and Stiglitz (1980) suggest that the "pace of innovative activity must be traced neither to the degree of concentration nor to the degree of monopoly in the industry in question, but to more basic ingredients like demand conditions,... nature of the capital market etc". 28 According to Stoneman & Leech (1980), the Dasgupta-Stiglitz model suffers from certain limitations; it considers only one innovation to be innovated at a time; it also considers only process innovation. They argued that the static nature of the model also inhibits the accumulation of knowledge. Accordingly, Stoneman and Leech (1980) developed a model where.a firm's R&D expenditure can either be directed towards cost reducing process innovation or to a demand-stimulating product innovation. Their model suggests that: 1. If innovation is of cost reducing type then, keeping other things unchanged, R&D to sales ratio would be negatively related with market share. 2. R&D to sales ratio varies negatively with the expected retaliation. (i.e. spill over reduces research effort). 3. The more productive the R&D process, the greater is R&D to sales ratio. 4. A higher R&D expenditure to sales ratio is associated with higher economies of scale and lower discount rate. 29 The first and second results seem to contradict each other. Larger market share should imply less scope of spill over and therefore a higher R&D to sales ratio according 27 As a proxy for time dimension they resorted to cross section comparisons among markets with varying degrees of monopoly and research intensity. 28 They subscribe to the view that if firms have to fund R&D activity internally in the face of capital market imperfections, then the size of the firm does matter. 29 The discount rate shows how one values the future vis-a-vis the present; a low rate necessarily implies a higher value attached to the future and therefore encourages a firm to undertake R&D expenditure. 25

18 to second result. But first result suggests the opposite. Cowling (1972), made an attempt to reconcile these two arguments by saying that ideally retaliation increases with concentration up to duopoly level, causing R&D intensity to fall up to this point. But then R&D intensity should be very high at the monopoly level in the absence ofretaliation. 30 There exist another strand in the literature, which attempts to relate R&D expenditure with the date of innovation. Higher the rivalry, shorter would be the development period requiring higher R&D expenditure. Kamien & Schwartz (1980) suggest that as rivalry increases, innovations would be introduced more rapidly. But there is an inflexion point of rivalry, up to which this trend continues and beyond which firms start postponing the innovation date Internalisation vis-a-vis Networkin~ Internalisation of R&D Both the schools of thought, the Schumpeterian and the neo-classical, emphasise upon relationships between size/ market structure and R&D. However, one basic point needs to be addressed at this juncture: why is it necessary to intemalise the R&D effort of firms? The neoclassical paradigm has sought to find the answer in terms of transaction costs and principal-agent problem. Transaction cost arises out of bounded rationality. If R&D is contracted out, an agreement must be reached involving a transaction (of research outcome) to be carried out at a future date. Due to bounded rationality, it is difficult for the contracting parties to specify all contingencies that can arise in this transaction. This affects transaction costs in two ways. First, it increases the ex-ante cost of drafting a suitable contract. Secondly, it 30 It is often postulated that extent of retaliation varies with firm size where smaller firms may expect greater retaliation from their larger counterparts as "larger firms often undertake protective R&D or build excess capacity". (Freeman (1965), as cited in Stoneman (1983) pp 37). Thus any defmite result is difficult to p:edict. 26

19 raises the ex-post cost of administration to amend the contracts in case of unforeseen contingencies. In this context, Kreps ( 1990) distinguishes between self-interest and opportunism. An opportunistic behaviour is different from self-interest in the sense that the former arises out of the dishonesty on the part of contracting parties unlike the latter (which is simply a reflection of the assumption of "rational individual") and results in a typical principal-agent problem. While transaction cost arises out of the incapability of the contracting parties to take care of all possible contingencies (for spot as well as forward transaction), the principal agent problem arises because of the possibility to violate the laid out norms by the contractee due to their opportunistic behaviour. 31 It is well known that appropriation, tacitness of knowledge and uncertainty are central issues in any technology transaction. Hence problems of transaction cost and moral hazard arise in research networking. DeBresson and Am esse ( 1991) highlight that costs arising out of the threat of leakage (of knowledge) is particularly true if both the parties are potential competitors in the final product market, since network is often a very "loose and decomposable system". 32 However, research networking and collaborations takes place, all these problems notwithstanding. In a comprehensive survey of the literature on networking, Debressen and Am esse (1991) point out that synergetic creation of knowledge through interaction, dynamic technological accumulation, and social learning are the three major thrusts of networking, especially when uncertainty is very high. 33 In this context the transaction cost theory has been criticised as "insufficient" because of its lack of focus on these prospective dynamic gains of networking. In fact, durable networks are believed to reduce opportunistic behaviour to a great extent and thereby minimise the principal-agent 31 Note that the contracting parties may be completely rational in this case. See Kreps (1990). 32 Other main costs are the costs of coordination. See Hakansson (I 989), Smith et a! (I 99 I), and Szarka (1990) for some explorations of hierarchy in networking resulting in take-over or break down of networking. 27

20 problem. 34 If indeed these dynamic gains are realised, then one can argue in a nutshell that dynamic gains from flexibility in the size of investment (and sunk cost) and a broader scope of technological exchange may outweigh the static transaction costs of networking. This may raise the relative benefits of networking vis-a-vis internalisation ofr&d. Intemalisation of Production vis-a-vis Subcontracting The threat of leakage of knowledge generated through R&D also comes from production networking. According to Williamson ( 1985) problems of transaction cost play an important role in shaping the structure of production organisation as well. Firms may want to integrate vertically in order to minimise transaction costs. However, there are other explanations of vertical integration as well. Foreclosing rival firms from accessing scarce raw materials, creating other forms of entry barriers and attempting to increase monopoly power are some of them. Stigler (1951) presented a life cycle explanation to this issue. In his opinion, urge for vertical integration increases both at the initial phase and the late phase of an industry's development. At the initial phase, a product is untried and untested, and many technical problems may have to be overcome. This is perhaps best tackled when the firm has control over its entire production organisation. At the late (declining) stage, integration is preferred since supporting and complementary industries also decline along with the industry in question. Williamson (1985) however, predicts a reduction in the vertical integration at the matured phase of development of an industry. Blair and Kaserman (1983) believe that uncertainty can cause vertical integration as upstream firms can eliminate demand variability due to the behavioural uncertainty of. the downstream firms. 35 There may be additional 33 See Lund vall ( 1990), White ( 1988), Pavitt et a! ( 1989) for details. 34 Sax en ian ( 1991) also points out that trust is the single most important factor behind persistent networks. 35 Kathuria ( 1996) argues that a part of this behavioural uncertainty can also be explained in terms of transaction costs. 28

21 considerations with regard to vertical integration in less developed economies, which we discuss later in section 2.5. All the above arguments notwithstanding, contract manufacturing continues to remain as an important form of production organisation, perhaps due to a natural limit to vertical integration. Indeed, contract manufacturing involving two given parties can prove to be a stable form of production networking if we consider the learning costs associated with making a contract. As Lundvall (1988) argues, building up of a contract involves a learning process and hence breaking an existing contract proves to be costly as it nullifies the efforts already made to build up a relationship by two parties. Citing Arrow ( 1974 ), he compares establishment of codified information flow with investment on physical capital, where development of common code is time consuming and involves learning. This creates an inertia to switch contracting parties, if both contractor and the contractee are risk averse in nature. 2.4 THEORIES OF DIFFUSION AND INCREMENTAL INNOVATION Models incorporatine Diffusion The neoclassical models in the Arrovian tradition (following Arrow (1962a)) gave less than necessary importance to diffusion as an R&D output. Neither the time cost nor the research cost involved in the diffusion process received adequate attention in these models. A possible reason could be found in the underlying concept of 'drastic innovation', where the monopoly price charged by the innovator is lower than the previous marginal cost threatening to drive the competitors out of the market. The means the competitors will have a very high incentive for instantaneous imitation at whatever cost. This is perhaps why these models assumed instantaneous diffusion without 29

22 explicitly taking into consideration the costs involved. 36 Hagedoorn (1989) notes that technology is often subject to considerable modifications and changes during its diffusion process. This continuous change of a given technology may actually be the outcome of in-house R&D at firm level geared towards 'minor' innovation (which we discuss later at length). Stoneman (1987) identified three prevailing approaches to analyse the phenomenon of diffusion. In the first (information based) approach, diffusion essentially reflects a process of information acquisition and learning. A new technology is not adopted instantaneously because all relevant information about the new technology is not readily available to potential users. It is assumed that the number of potential users is fixed. The time path of diffusion reflects learning by each of the potential buyers, which can either be an epidemic process or a bayesian process. The former process assumes the expected future profit and firm size to remain fixed during the diffusion process, only the degree of uncertainty regarding the expected future profits declines with diffusion over time. The latter process assumes continuous revision of firm-size and expected profits in a Bayesian manner, thus making the number of end users endogenously determined. In the second (difference-based) approach, on the other hand, firms are assumed to have the necessary information regarding new technology. But inter-firm difference exists in terms of either their financial capability or their valuation of the new technology. The cost of acquiring the technology is also assumed to fall over time. In this approach, adoption decision (and the time of adoption) is determined by the equilibrium condition of marginal cost of adoption being equal to the marginal benefits derived. In both of these approaches it was assumed that benefits from adoption is independent of the number of the users. 36 Tirole (1988) shows that there can be a case where a non-drastic innovation will never be imitated in a Bertrand competition framework with identical costs. 30

23 This last assumption is discarded in the game theoretic approach, where the benefit from the new technology is believed to vary with the number of users. If X is the number of users at any point of time, then g(x) is the valuation function where 8g(X)/8X is generally assumed to be negative. Thus there is an incentive to be the first user. If, however, 8g(X)/8X >0, then the diffusion process will not be automatic and some form of governmental intervention would be required to start the process. Stoneman (1987) also introduced the supply side of the diffusion process along with its demand side to endogenise the price of the new technology (to be determined by the interaction between demand and supply), hitherto considered as parametrically given. The introduction of the supply side makes it possible to examine the welfare gains/losses of diffusion defined in terms of consumer and producer surplus. Using this demand-supply framework, Stoneman (1987) fommlated a two-period model of diffusion. In this model, diffusion can occur only through the sale of capital goods, assuming that information about the new technology is not saleable separately. Since new technology (assumed to be a process technology) reduces cost of production, supplier industry can hope to extract the surplus generated by the use of new technology (for simplicity we may assume it to be the difference between constant marginal costs before and after adoption). Thus profits of the supplier industry is generated through R&D resulting in new technology. Stoneman (1987) also derived the optimum diffusion path under different market forms using the difference-based approach. One may note that if the producer of a ~ew technology is also a producer of the commodity produced with that technology, in violation of Stoneman's assumption, then the incentives for technology diffusion through sale of technology/ capital goods may disappear. Cohen & Levinthal (1989) explicitly considers the role of in-house R&D in order to enhance absorptive capacity and better assimilate knowledge available in the public 31

24 domain. They argued that "R&D not only generates new information, it also enhances the firm's ability to assimilate and exploit existing information." This role, according to them, "would change the traditional (negative) relationship between spill-over and R&D incentive into a positive one". 37 Then, besides the information generating aspect of R&D, they consider absorptive capacity building dimension of R&D as well. The absorptive capacity (or learning) is, therefore, endogenised in this model. One should note that learning through R&D, as conceptulised here, is different from learning by doing. 38 The latter is related to gains in efficiency for a given technology, whereas the former refers to a change in the absorptive capacity implying a change in the technology level of a firm. In-house R&D is expected to enhance the firm's capability to extract and exploit outside knowledge as well as its ability to access intra-industry spillover. Outside knowledge refers to technology embodied in suppliers' machines or generated through research at' universities and governmental organisations. The model defines Zi, the i-th firm's stock of scientific knowledge as:... Z; = M; + Y;(BLMj + T) j# where e = the extent of spillover. 0 < e <1. Mj = jth firm's R&D investment, T =extra-industry (outside) knowledge, Mi = ith firm's (concerned firn1) investment in R&D Yi = fraction of knowledge in the public domain that a firm IS capable of assimilating and exploiting, i.e:, the firm's absorptive capacity. The model assumes gross return to Zi to increase at a diminishing rate (oll/ozi >0, and 37 Cohen & Levinthal (1989), italics own. 32

25 Again Yi = Yi (M, p) (with YM >0 and YMM< 0 ). P is the characteristics of extra industry knowledge like the complexity of knowledge. It is assumed that. When outside knowledge is less targeted (generally the research at university level or that in the basic science area) to a firms requirement, R&D should be enhanced to maintain a given level of absorptive capacity. Higher values of p are therefore associated with a greater dependence on own R&D to acquire absorptive capacity. This is because with higher p, first of all, the marginal impact of R&D is higher, and secondly, firms would tend to increase R&D as less absorptive capacity can be acquired from a given R&D if outside knowledge is more complex. One important aspect of the model is that it acknowledges firms' need to rely on own in-house R&D effort even to access existing intra-industry knowledge. In this regard, it is a development over the information-based approach, which considered only the time cost to explain diffusion. 39 In this model, a firm is a typical profit maximiser in a static framework. The model obtains three results using comparative static analysis. The results pertain to the direct effects on R&D of (a) ease of learning (b) intra-industry spill over and (c) teclmological opportunity. Result (a) describes that there is positive relationship between a firm's incentive to conduct R&D and the complexity of knowledge (p): its own R&D becomes more critical \ to build up absporptive capacity as p increases. Also a rise in p reduces the competitors' levels of absorptive capacity and therefore their ability to exploit the knowledge base of the concerned firm. This reduces the threat of leakage and provides additional incentives for this firm to conduct R&D. 38 Arrow (1962b) made a pioneering attempt to introduce and formally incorporate the concept of 'learning by doing' into mainstream economic theory. 33