Vertical FDI versus Outsourcing: The Role of Technological Complexity and Absorptive Capacity Arti Grover Delhi School of Economics [email protected], [email protected]

Non-Technical Summary Offshoring of white collar jobs began in the software sector where significant activities were transferred to foreign locations, leading to the creation of software expertise and resources in less developed countries like India. However, most often it is expected that the offshoring of white collar jobs takes place within a firm through its subsidiaries vis-à-vis arm’s length agents in a host country. Recently, we have witnessed a rising trend in outsourcing of complex products like medical electronics, designs of digital devices from the likes of Dell, Hewlett-Packard, Motorola, and Philips to third party vendors. How is outsourcing different from offshoring? There are two alternative ways in which a final good producer, also called the sourcing firm, can organize its input production. It can either make it herself or it can buy it from a third party vendor. The former mode is referred to as the intra-firm production transfer or vertical Foreign Direct Investment (VFDI) while the latter mode where the sourcing firm decides to buy the inputs from outside contractors or arm’s length agent, is referred to as outsourcing. In this paper, I differentiate between VFDI and outsourcing and build a model of sourcing firm where these firms must choose the mode of organizing its offshore production. Till date, the internalization literature has explored the make or buy decision of the sourcing firm through the lens of the source country. Even though the sourcing firm in question makes an organizational choice of its input production in the host country, the literature has largely ignored the host country variables in determining this organizational decision. The literature find that the organizational choice of international production sharing depends on the characteristics of the final good – like the degree of standardization of the good, the capital intensity of the final good, the intensity of the offshored input in the final good, the productivity of the sourcing firms, the legal framework and the market thickness in the host country. A common factor across the determinants of the “make or buy” or the internalization decision is that they critically affect the cost of the sourcing firm. The cost of technology transfer is one such variable which is simply missing in the vertical production transfer context. In this paper, we explore how the technology transfer cost affects the internalization decision through its dependence on the complexity of the offshored input and its interaction with the host country factors like the quality of its labor force or its absorptive capacity. We propose that the higher is the technological complexity (or quality) of the offshored input, the higher is the cost of technology transfer but higher is the productivity of the host country labor. Empirical evidence helps us support that the technology transfer costs are borne by the sourcing firm in an internal production transfer while the same costs are borne by the vendor if the production of the input is outsourced. In deciding on its organizational form a sourcing firm faces a tradeoff between high productivity and low contractual costs in the VFDI mode while a low technology transfer costs in the outsourcing mode. The firm’s choice between the two alternative modes must therefore depend on the complexity of the offshored input and the absorptive capacity of the host country. Our model is an extension of the Antràs (2005) model. His model features two inputs, the high-tech and the low-tech, which are combined to produce the final good. By assumption, only the low-tech input is can be offshored. The firm must decide between VFDI and outsourcing once it decides to offshore the input to take advantage of the low wages in the host country. The author finds that if the final good is high-tech, it would be profitable for the firm to offshore the input through the VFDI channel. The reason for their choice

is explained through the property rights theory of the Grossman-Hart-Moore seminal papers. In any relationship, the agent who contributes more to total surplus must get the residual control rights in order to minimize the distortions due to suboptimal effort. In an offshoring relationship, the final good producer supplies the high-tech input while the supplier makes investment for the low-tech input. Therefore for optimality, the agent who contributes more to the relationship should get the residual control rights. For the case of a high-tech good, it is the final good producer who makes investment in high-tech input. Hence intrafirm production transfer emerges which keeps the supplier affiliated with the sourcing firm and the sourcing firm retains greater control over input production. In our model, the introduction of technology transfer costs that is contingent on the technological complexity (or quality) of the input and the absorptive capacity of the host country change the tradeoffs for the sourcing firm. Besides accounting for whether a good is high-tech or low tech, the sourcing firm must also gauge the technological complexity of the input and the technology transfer cost it may impose if the input is offshored to a subsidiary. We find that, a high-tech good is more likely to get offshored through intra-firm transactions only for intermediate range of technological complexity of the input. At low and high levels of technological complexity of the offshored input, the sourcing firm is better off engaging an unaffiliated supplier. The intuition for this result is as follows. By making an intra-firm transfer, the sourcing firm faces lower host country wages and lower contractual costs but higher costs of technology transfer vis-àvis international outsourcing. At low technological complexity, the distortion in technology transfer investment by the supplier is low, while at high technological complexity the savings from technology transfer cost forces the sourcing firm to choose an unaffiliated supplier over an affiliated one even though the final good is a high tech one. The model delivers implications for policy on technology and absorptive capacity in the host country. In real world, we do not observe outsourcing of inputs embodying complex technology on a wide scale. If the host country government encourages (say, by subsidizing) technology investment by the domestic vendors, then the probability of gaining an outsourcing contract vis-à-vis an intra-firm contract for technologically complex input, given the level of absorptive capacity will certainly increase. Moreover, if a host country enlarges its absorptive capacity by heavy investment in education to build comparative advantage in inputs embodying complex technology, then they stand to gain by getting more value-added work through both VFDI and outsourcing.

Acknowledgements: I am extremely grateful to Partha Sen for his guidance and motivation throughout the course of this project. The paper has benefited immensely from discussions with Gene Grossman, Pol Antràs Uday Bhanu Sinha and Abhijit Banerji. I am also thankful to the participants of the CEPR Applied Industrial Organization Workshop 2007, the Royal Economic Society conference 2008 and the International Economic Association conference 2008 for their useful comments.

1

Technical Abstract: Technology transfer costs have a profound influence on a firm’s organizational choice in a production sharing relationship. To explore this nexus, we associate the technological complexity of the off-shored input with the organizational mode of international production sharing by extending the Antràs (2005) model. We modify the Antràs model by proposing that the low-tech input, as qualified within the model, cannot be produced in the low wage south without costly technology transfer. The cost of technology transfer in turn depends on the technological complexity of this input and the absorptive capacity of the host country. Our model refines the results obtained in Antràs (2005). We find that 1. For high-tech goods, intra-firm transfer is preferred vis-à-vis outsourcing only for intermediate range of technological complexity of the off-shored input, 2. On the other hand, for low-tech goods, where the likelihood of outsourcing is higher in Antràs, intra-firm offshore contract is still possible for low to mid range of technological complexity. Our model has policy suggestions for host countries which aspire to maximize their benefits from the expanding global production sharing phenomenon. As the wage gap between the source and the host country falls, cost considerations for offshoring disappear. New sources of comparative advantage should therefore be created in the host country by subsidizing technology investment and higher education to build higher absorptive capacity. Keywords: Outsourcing, Foreign Direct Investment, Technology Transfer, Absorptive Capacity JEL Classification: D23, F12, F23, L22, L23

2

Section 1: Introduction Several studies have found a correspondence between the level of technological complexity of a product (final good or the fragment being offshored) and the sourcing firm’s organizational structure. Historically, high technological complexity of a product has been associated with vertical integration. Based on field research conducted in the US, Singapore, UK and India, Aron and Singh (2002) find that lower end processes like data transformation or customer service which embody less complex skills are outsourced to a third party. On the other hand, complex inputs in the global value chain are offshored to an affiliated supplier. Similar conclusions can also be found in Gereffi et al (2003) and Davidson and McFetridge (1985). Most of these studies rely solely on the transaction cost economies to explain the tendency for vertical integration at a higher level of technological complexity of the input. This, however, leaves a gap in our understanding of some real life examples. There is a rising trend of outsourcing of complex products like medical electronics, designs of digital devices bought from Asian manufacturers by the likes of Dell, HewlettPackard, Motorola, and Philips. As a specific illustration, we note that Dell outsources both the design of its digital devices as well as the assembly of laptop components, which are tasks belonging to the highest and lowest technological complexity category respectively, while at the same time, Dell operates captive centers for customer support, a job of intermediate technological sophistication. What explains the outsourcing of tasks at the extreme ends of technological sophistication while internalizing tasks of intermediate complexity? To understand the internalization decision of a sourcing firm that off-shores technologically differentiated inputs (from the perspective of the host country), we introduce technology transfer costs in the Antràs (2005) internalization model1. The issue of outsourcing versus vertical integration in an international production sharing context is not new. Research on offshoring has revealed that the organizational structure of a sourcing firm is influenced by the degree of standardization of the offshored input, capital intensity of the final good, intensity of the offshored input in the final good, productivity of sourcing firms, legal framework and market thickness in the host country. One common denominator across all these factors is that these variables crucially impact the cost borne by the sourcing firm. Surprisingly, the cost of technology transfer – transmission and assimilation – that has been central to the theory of multinational corporations (MNC) and horizontal production transfer since the last three and half decade has been overlooked with regard to vertical production transfer. In this paper, we incorporate technology transfer cost as a crucial variable that affects a firm’s “make or buy” decision. Technology transfer costs are as crucial in a vertical relationship as in horizontal Foreign Direct Investment (FDI) or licensing. Particularly, in an outsourcing transaction with an unaffiliated supplier, technology acquisition and assimilation costs are significant. A survey of the Business Process Outsourcing (BPO) vendors in India (The Hindu Business Line, 2005) reveals that 25.2% of total wage cost is spent on training of its employees to produce inputs of the quality standards specified by its buyer. Arora et al (2000) in their extensive fieldwork on Indian software outsourcing and BPO industry find that a significant amount of specialized training (an average of 2-3 months) for all employees, including the skilled employees, is undertaken after recruitment. In 2004, Caliber Point Business Solutions Limited, a third party BPO service provider to Hexaware Technologies, made substantial investments in technology infrastructure like fiber optic technology for the backbone of Local area Network, Dell Intel Xeon Servers and Network Security using Stonegate Firewall and IDS and Tata Honeywell CCTV. Our model builds on the Antràs (2005) model where the final goods are produced using two inputs – high-tech and low-tech, of which only the low-tech input can be offshored. However, in our model, we propose that even this low-tech input cannot be produced in the host country without incurring technology transfer costs. Technology transfer cost borne by a sourcing firm depends on the technological complexity of the offshored input and the absorptive capacity of the host country. We model technology transfer cost as another type of relationship specific investment which should borne either by the sourcing firm or the supplier, depending on the mode of organization. When offshored inputs are differentiated in terms of their technological complexity, they necessarily entail different costs of technology transfer. Therefore, the

1

Antràs (2005) model relies on the property rights theory (Grossman-Hart-Moore) to explain a firm’s choice of organizational mode.

3

production cost of an offshored input borne by a final good producer differs across different products (or varieties of a given product) and hence their internalization decision. We differentiate between vertical foreign direct investment (VFDI) or vertical integration and international outsourcing (IO) in the usual Grossman-Hart-Moore way of contractual bargaining power, wherein the sourcing firm has greater bargaining power in an intra-firm production transfer vis-à-vis a market transaction. Additionally, using evidence from existing studies on horizontal production transfer relationships and current offshoring surveys, we distinguish between intra-firm and arm’s length production contract with respect to the technology transfer cost borne by the sourcing firm in the two alternative modes. Specifically, we argue that the technology transmission cost incurred by a sourcing firm in an internal production transfer is substantial and forms a part of its relationship specific investment (RSI), while the subsidiary manager has little incentive to invest in technology assimilation. Au contraire, if the offshore production is contracted to an outside supplier, then the supplier has to incur a significant proportion of the technology transfer costs while the sourcing firm has little motivation to bear the costs of technology transmission2. Since the technology transfer costs borne by the sourcing firm varies with its organizational mode, so naturally internalization decision becomes a function of this cost which in turn depends on the technological complexity of the offshored input along with the host country’s absorptive capacity. In the Antràs (2005) model, the decision to internalize depends on the intensity of the low-tech input (the offshored input) in the final good. The sourcing firm makes RSI in the high tech input while the supplier makes RSI in low-tech input. Grossman-Hart-Moore property rights theory suggests an intra-firm production transfer for a high-tech good (to minimize contractual distortions in the RSI). This is because the supplier’s contribution in total surplus is low relative to the sourcing firm and distortions in RSI is lower if the residual rights of control lies with the sourcing firm. On the other hand, for a low-tech good the supplier makes a greater contribution in total surplus vis-à-vis the sourcing firm. In such a case, optimization, that minimizes contractual distortions in RSI, should lead to an outsourcing contract. We extend the Antràs (2005) model by intertwining the intensity of the offshored input with the technological complexity of this input. Given the absorptive capacity of the host country, under certain parameter restrictions, our model reveals that even a low-tech good (high intensity of the offshored input) is likely to be offshored through intra-firm transactions at intermediate range of technological complexity. At extreme ends of technological complexity, the sourcing firm may be better off engaging an unaffiliated supplier, just as in the case of Antràs (2005) model. For high-tech good, VFDI is preferred at mid range of technological complexity as in Antràs (2005), but there exists international outsourcing at higher end. The intuition for this result can be explained by the distortion in the offshored input of a low tech good and technology transfer cost at varying levels of technological complexity. A sourcing faces a tradeoff between technology transfer cost and the RSI in the low-tech and high-tech input in the two alternative organizational modes. If the sourcing firm’s share in total surplus is much higher in vertical integration vis-à-vis international outsourcing, then, by choosing to integrate at low levels of technological complexity, the sourcing firm has to bear technology transfer costs but has low RSI in the low-tech input and high RSI in the high-tech input. If the final good is intensive in the low-tech input, then the sourcing firm would rather choose international outsourcing because the RSI in low-tech input is extremely important in this case. On the other hand, if the final good is intensive in high-tech input, then the sourcing firm should integrate because the RSI in the hightech input is relatively more important. At mid range of technological complexity, integration always yields higher RSI in the low-tech and high-tech input because high technology transfer cost distort an unaffiliated supplier’s RSI in the low-tech input. This implies that the sourcing firm should choose to integrate for all goods at mid range of technological complexity. At higher levels of technological complexity, even though the RSI in the inputs is low in the arm’s length mode, the burden of technology transfer cost is too high to warrant VFDI. Therefore, irrespective of whether a good is a low-tech or high-tech, there exists an arm’s length relationship at high levels of technological complexity. Our result is empirically testified in Borga and Zeile (2004) where the volume of intra-firm trade falls with increasing R&D intensity of the affiliate3 (the R&D intensity of the affiliate could be a good proxy for the technological complexity of the offshored input). Competition among suppliers ensures that the sourcing firm has to incur minimal technology costs. The evidence from Borga and Zeile (2004) is relevant to this paper only to the extent that the intra-firm trade they are talking about is the import of inputs from the parent firm to the affiliate for further processing.

2 3

4

Our model highlights the possibility of different trends that can emerge in the organizational structure of fragmented production. If we consider a dynamic version of this model, where technological complexity of inputs increases overtime with technological progress in the north, then, we can explain the observed organizational changes in the offshoring industry. The BPO industry has witnessed relationship in the form of “Build-Operate-Transfer” (BOT) whereby a sourcing firm initially establishes an outsourcing relationship with an unaffiliated supplier and then after achieving a certain level of maturity takes over the production unit to make it a captive one. Our model predicts that relationship of this form will most likely transpire for a low-tech good where the supplier has minimal share in surplus. This kind of relationship has emerged in case of Aviva Plc insurance in India. 24/7 Customer, a BPO firm, was hired to implement this BOT Model. Our model, on the other hand, also explains the transformation of a “captive”, or an affiliated supplier unit to a third party vendor or an unaffiliated supplier, as exemplified by General Electric India operations. In December 2004, GE Capital International Services’ Indian operation was sold off to a third party and GENPACT was born. The model delivers implications for policy on technology and absorptive capacity in the host country. In real world, we do not observe outsourcing of inputs embodying complex technology on a wide scale. If the host country government encourages (say, by subsidizing) technology investment by domestic vendors, then the probability of an outsourcing contract for technologically complex input, given the level of absorptive capacity will certainly increase. Moreover, if a host country enlarges its absorptive capacity by heavy investment in education to build comparative advantage in inputs embodying complex technology, then they stand to gain by getting more value-added work through both VFDI and outsourcing. The paper beyond this point is organized in the following manner. Section 2 discusses the literature associated with this research area. In section 3 we develop the model and discuss the consequences of introducing technology transfer costs in the Antràs (2005) model. Section 4 discusses the results of the model and section 5 makes a conclusion. Section 2: Related Literature In this section, we intend to assimilate the internalization literature with the literature on technology transfer costs and contract theory. The importance of technology transfer costs was first emphasized by Teece (1977) three decades ago. Based on the information obtained for twenty-six projects of U.S. firms in chemicals and petroleum refining and machinery, he found that the cost due to technology transmission can range from about 20%-80% of a project cost. Later several studies examined the importance of technology transfer costs in affecting the horizontal mode of entry by a multinational firm. For example, Mattoo et al (2001) build a theoretical model where establishing a subsidiary is preferred to acquisition of a domestic firm if the cost of technology transfer is high. This is because the sum of the technology transfer costs and the acquisition price outweighs the benefits of lower marginal costs of the domestic (acquired) firm. Hence, acquisition may not be a profitable strategy of entering a foreign market if technology transfer costs are high. In our paper, we propose that the burden of technology transfer cost can be shifted by transferring ownership share and therefore impact the internalization decision of a firm. A major stumbling block in relating the technology transfer literature to internalization decision in a vertical production transfer context is that there is little research on vertical transfer of technology and the related costs. Therefore, we have to rely on studies relating to horizontal technology transfer (HTT). However, even the HTT models do not offer much evidence on the cost sharing patterns between the transferor and the transferee or the resource cost of technology absorption borne by recipient firms. Recently, several instances from BPO firms and sourcing firms, as cited in the introduction, confirm our belief that technology absorption is a substantial proportion of costs and therefore can be expected to impact the internalization decision of the sourcing firm. The other strand of literature which we incorporate in our model is contract theory which has been known to influence the sourcing firm’s decision to internalize since Grossman and Hart (1986). Incompleteness of contracts is an inevitable feature that sets in when transaction happens between two independent entities (See for example Antràs and Helpman, 2004 and Antràs, 2003, 2005). With incomplete contracts, these models argue that the bargaining power of the sourcing firm is higher in VFDI mode vis-àvis outsourcing. 5

In the Antràs (2003) model, a final good requires both capital and labor. To simplify the analysis, Antràs assumes that the sourcing firm always makes complete investment in capital while the supplier contributes to investment in labor. This implies that for a capital-intensive good, the sourcing firm contributes more to aggregate surplus relative to the supplier. In such a case, minimizing contractual costs or optimization based on property rights theory of Grossman-Hart-Moore requires integration with the supplier. Antràs (2005) is also based on the same principle as Antràs (2003) where the two inputs are labeled as hightech and low-tech instead of capital and labor. The sourcing firm is assumed to make RSI in the high-tech input and therefore we expect an intra-firm production transfer for a high-tech good. The model is extended to include dynamics of a production cycle where the final good standardizes with time implying that the intensity of high-tech input falls with time and therefore the same input may be outsourced at a later stage of production cycle. This highlights that the degree of standardization is also higher for a product that is outsourced relative to a product that is produced by a MNC subsidiary. Besides contractual differences between VFDI and outsourcing, Antràs and Helpman (2004) rank the fixed organizational cost of integration higher relative to outsourcing. The interaction of differential fixed and variable costs with firm productivity of Melitz (2003) framework produces interesting results on sorting pattern of heterogeneous firms. Firms in their model sort themselves into four groups – firms that either integrate or outsource in the north or the south. In their model, the most productive firms always offshores to the south to an unaffiliated supplier. To understand these two strands of research together in one framework, we split technology transfer costs into transmission and assimilation costs. Transmission cost is the cost to shift codified knowledge like blueprint, formulas, management techniques customer list, or tacit knowledge like know-how, information gained from experience which is usually borne by the transferor. Assimilation cost is the expenditure on R&D by the supplier, cost of training workers to adapt to new technology, or acquiring new technology from the technology market. These costs are typically borne by the suppliers unless the host government makes it mandatory for the investing foreign firm to make investment on technology absorption or acquisition. For the VFDI (vertical integration) mode, we would expect high technology transmission costs in proportion to technology assimilation cost. This is justified by the high share in total surplus appropriated by the parent firm in an intra-firm transaction which induces it to invest in costly technology transmission. At the same time, a low share in total surplus accruing to the affiliated supplier reduces the incentive of the subsidiary manager to invest in assimilation of technology. Our insight is spelled out in a survey by Chuang and Chang (1993) on foreign affiliates and domestic licensee firms (and joint ventures) in Taiwanese pharmaceutical industry. They find that foreign subsidiaries in the host country do not give much importance to the cost of technology transfer in their profitability analysis, while the licensee firm cares a lot about the cost of technology for profit maximization. They explain their results by emphasizing that domestic firms rely on external market to channel the ingestion of technology and thus have to bear pecuniary expenditure and adaptation costs. Per contra, a subsidiary obtains technology from its parent firm which precludes any transaction in the technology market. Therefore technology transfer costs do not appear to be an important consideration for its profitability. Teece (1977) survey found that technologies closer to the frontier are transferred to a subsidiary vis-à-vis an arm’s length agent. Since the cost of transferring technology is positively related to its age, a parent firm spends more resources in transmitting technology to a subsidiary vis-à-vis an arm’s length unit. UNCTC (1987) also finds empirical evidence supporting the fact that intra-firm technology transmission cost is far more significant vis-à-vis the cost for transferring technology to independent parties in cases of US and German firms. This is a reflection of the MNC preference towards fully controlling the assets transferred to overseas establishments. Since full control of technology transferred is not granted in case of an arm’s length contract, the MNCs may not prefer to bear the cost of technology transfer in an arm’s length relationship. Further, technology transfer by a sourcing firm to an unaffiliated supplier is also limited by the fact that a third party vendor (TPV) usually provides services to more than one client. Therefore, if a client transfers its technology to the supplier, it undertakes a risk that its technology maybe used by the supplier to serve its competitors as well. For example, a credit card services firm ‘A’ has a proprietary “technology” to segment customers based on their risk profiles for default-risk analysis. If this firm contracts externally with a risk analytics service provider firm, then, firm ‘A’ will not transfer its profiling technique to the vendor as the vendor may use this technique to serve a competing credit card service firm ‘B’. In general, any rational 6

sourcing firm will not transfer its technology to a TPV to the extent possible. Hence, the TPV has to invest on its own training and technology acquisition contrary to a captive (subsidiary) unit which can depend on its parent firm for technology. Examples can be found in the Indian third party BPO companies like VisualSoft Technologies Ltd, Zensar Technologies, iGate Global Solutions etc. who have to spend a considerable proportion of their revenues on technology acquisition and absorption. Since a sourcing firm has little incentive to invest in technology transmission in an arm’s length relationship, therefore the technology assimilation and acquisition costs gain more importance in an outsourcing contract. It makes sense for an unaffiliated supplier to make investment in costly technology acquisition because she appropriates a higher share in total surplus (vis-à-vis the affiliated one). Chudnovsky (1991) report on north-south technology transfer finds that technical assistance to local suppliers is crucial for meeting their performance metric, however, this is precisely the area where assistance from final good producers are missing. Egan and Mody (1992) find that in a shoe manufacturing subcontracting relationship the buyer is willing to transmit only the minimum information required to get the product out of the production cycle. If the product must adhere to stringent quality specifications before being accepted, then it is entirely left to the supplier’s discretion to take up the contract, get involved in the manufacturing process and produce the good of requisite quality at lowest possible cost. Thus, the supplier has to incur a large part of the technology transfer or adaptation cost. We now re-state the assumption implied by the existing literature on technology transfer cost and organizational mode:

Assumption 1: In case of VFDI, the parent firm incurs a significant share of technology transfer costs, while in case of outsourcing, it is the unaffiliated input supplier who bears a large proportion of this cost.

The assumption that we state so explicitly in our paper is usually implicit in models like Bartel et al (2005). They find that an increase in the rate of technological change increases outsourcing because it allows the sourcing firm to use the services of the supplier based on leading edge technologies without incurring the sunk costs of adopting new technologies. The assumption implicit in their analysis is that it is always the supplier of the input who subsumes all the cost of technology in an outsourcing relationship. Section 3: The Model: Consider a world with two countries – the developed north and the low wage south Consumer preferences are such that a unique producer of variety j, of good y faces the following isoelastic demand function: −1 (1) y ( j ) = λ p( j ) 1 − α Where p(j) is the price of good y(j) and λ is a given parameter known to the producer. In the north, any variety of the final good y is produced using two inputs, high-tech, x h , and low-tech, x l , with intensity (1-z) and z respectively. ⎛ x ⎞ y = ⎜⎜ h ⎟⎟ ⎝1− z ⎠

1−z

⎛ xl ⎜⎜ ⎝ z

⎞ ⎟⎟ ⎠

z

If however, the low-tech input is offshored, then the production of final good depends on the technological complexity, denoted by i, of the low-tech input. ⎛ x ⎞ y = ⎜⎜ h ⎟⎟ ⎝1−z ⎠

1−z

⎛ x l (i ) ⎞ ⎟⎟ ⎜⎜ ⎝ z ⎠

z

(2)

The low-tech input is differentiated (from the host country’s perspective) with respect to its technological complexity (or quality4), i. The more complex the technology employed in producing a low-tech input, the

4

I am grateful to Pol Antràs for suggesting this interpretation.

7

higher is the quality of the input5. From the perspective of the north, we assume that are no technology transfer costs for producing the low-tech input in the north, and also that the home produces one unit of the low tech input using one unit of labor at all levels of technological complexity. Thus, technological complexity matters only when the low-tech input is offshored and therefore the low-tech input is differentiated only from the perspective of the host country. In this paper, we model the internalization decision across firms in an industry that produces the final good y using technologically differentiated low-tech inputs. It is also possible to re-interpret this model as an organizational choice model of offshoring several inputs of a single firm. Suppose a firm has 3 low-tech inputs which are combined with a high tech input to produce the final good. The three low tech inputs vary in their technological complexity and therefore in the technology transfer costs. Our model can generate a firm’s organizational decision across these three low tech inputs. We will explain the two alternative interpretation of our model through an example at a later stage in this section. By assumption, the South lacks the capability to produce a high-tech input like R&D. Thus, it is only the low-tech input that can be offshored. In the Antràs (2005) model, one unit of labor is required to produce one unit of the low-tech input, irrespective of the location of production. Grossman and Rossi-Hansberg (2006) introduced the possibility of a divergence in labor requirements to produce the offshored input contingent on the location of production as well as the complexity of the task. A unit of a task, in their model, requires β t (υ ) > 1 units of host country labor vis-à-vis one unit of domestic labor, where t (υ ) is an increasing function of the complexity of task υ and β reflects the overall feasibility of offshoring at a point in time. Their formulation suggests that as the complexity of tasks increases, it becomes more costly to offshore to the host country. However, one feature missing in their model is that it does not allow for cost savings (or productivity advantage) from offshoring a more complex task to the host country. It has been observed that the cost savings per employee for a call center agent who performs simple, low skill tasks is much smaller relative to a skilled employee who performs complex offshored jobs like risk analytics. Figure 5 shows the graph of the cost of doing various types of tasks by a sourcing firm in the home location, near shore location (say, as Canada is for firms in the US) and offshore locations (for example, as India is for sourcing firms in the US). BPO task can be interpreted as a relatively simpler task vis-à-vis a KPO (knowledge process outsourcing) task. We find that the cost savings derived from offshoring (to near shore or offshore location) a complex task (KPO task) is higher. In our model, we incorporate the cost as well as the benefits of offhsoring a complex input.

Figure 1: Costs for BPO and KPO services in USD per hour at On Shore, Near-Shore and Off-Shore locations Source: E-valueserve (2005)

5

Higher is the quality of the input, higher is its productivity and technology transfer cost.

8

In the south, the production of the low-tech input depends not only on the employment of labor, also on the labor productivity in the host country, T s (i ) .

L

s xl

, but

⎡ ⎤ (3) x l (i ) = x l ⎢ L s , T s (i ) ⎥ x ⎣ l ⎦ The low tech input of each good i embodies firm specific or product specific technological complexity i. A high i implies the use of more advanced technology in production6. ∂ 2 T S (i ) ∂T S (i ) > 0, <0 ∂i ∂i 2 0 ≤ i ≤ 1, T S (0 ) = 0, T S (1) = 1 The effect of technology on productivity: A more complex and advanced technology is believed to bring in higher productivity. For example, Acemoglu, Antràs and Helpman (2006), build a model where a more advanced technology is implicitly more productive. We assume that the productivity increases at a decreasing rate due to a rise in technological complexity of the input. To ease the interpretation of the technological complexity of the low tech input, i, and how it is different from z,, we can consider the example of a final good, say, a consulting project. A consulting project of a firm can be treated as a final good y(i), produced using two inputs – the high tech and the low tech. A consultant’s strategic analysis of the client’s problem is a high-tech input, an input available in the north only, while data analysis of client information is an example of a low tech input which can be offshored. If the final good producing firm requires relatively fewer inputs from data analysis vis-à-vis a consultant’s strategic analysis, then the firm provides a service/project which is intensive in the high-tech input and the parameter z is low for such a service/project. The complexity of the low-tech input, that is, the technique used for data analysis, differs across projects. For example, the project may involve sophisticated data analysis tools, say using complex software like - SAS or it may simply carry out data analysis using excel. Technological complexity, i, of SAS is higher relative to Excel and accordingly, the value and efficiency for data analysis is higher in SAS. Even if the contribution of data analysis is the same in the final project (same z) of two final good producers, the internalization decision of the two firms will depend on what technique is used to make data analysis, that is, the technological complexity of the low-tech offshored input. We now take an example from manufacturing to explain the difference between z and i and also to reinterpret the model’s implications on the internalization decision of various inputs of a single firm. A laptop production requires, say, three different low tech inputs. The R&D input required for developing and improving various features of a laptop for the first time is an example of a high-tech input. R&D is always done in the north and its intensity is represented by (1-z) in our model. Within the low-tech input production stage, there are many possible operations varying by their technological sophistication. Consider a firm that has three different operations comprising its low-tech input – design for laptops, customer support services and the assembly of laptops. Designing laptops is presumably a job of highest technological complexity, followed by customer support and then followed by assembly process. Since the cost of fragmented production varies by the technological complexity of the input, there is bound to be different organizational modes of offshoring these three low-tech inputs. Dell7, for example, outsources design and assembly to Asian manufacturers, the jobs involving highest and lowest technological complexity, but maintains four captive centers in India for its customer support, an operation of intermediate technological sophistication.

Our model attempts to explain how a difference in technology of a low-tech input affects its organizational mode.

Assumption 2: The production of the low -tech input is linearly proportional to productivity and labor. The technology that we refer to in case of offshoring is typically available in the technology market and the supplier does not have problems finding a technology solution. 7 Dell had significant outsourcing relations with third parties like Wipro BPO and Sitel, but it still maintained four captive contact centers in India at Bangalore, Hyderabad, Mohali and Gurgaon. See De (2006). 6

9

x l (i ) = T S (i ) L Sx

l

x l = x l (L ,0 ) = 0, S xl

x l = x l (L ,1) = x l (L S xl

S xl

)= L

( 3′ ) S xl

i = 0 means that the technological complexity of the offshored input is too low to justify production in the South, T S (0 ) = 0 , and i = 1, implies that leading edge technology is used to produce the low tech input which makes productivity of southern labor high enough to match the productivity of the northern labor. If technological sophistication adds to productivity, it cannot come without cost. A higher level of technological complexity has to be matched by a corresponding rise in efforts to transmit and assimilate the technology. In the example of the consulting project discussed above, there are costs of running data analysis in SAS – license costs and training costs. Excel, which has lower technological complexity, has a lower technology transfer cost vis-à-vis SAS. The assumption that technology transfer cost is a positive function of technological complexity of the low tech input is straightforward and follows directly from Teece (1977) observation that technologies embodying more complex mechanisms require more resources to be transferred. In addition to this cost, Chuang and Chang (1993) model suggests that technology transfer cost depends on many factors, namely, the mode of entry of the sourcing firm, absorptive capacity of the host country and the level of technological development in the host country. To endogenize the technology transfer cost with respect to the mode of organizing production fragmentation, we use assumption 1. The sourcing firm (supplier) understands that there is little incentive for the supplier (sourcing firm) to invest in technology assimilation (transmission) in an intra-firm (external) production contract and hence she decides to take a small fixed payment, TT fS (TToN ) from the supplier (sourcing firm) in lieu of its unverifiable and insignificant investment in technology transfer. To simplify the algebra, and without loss of generality, we assume that these fixed payments are insignificant and close to 0. Assumption 3: TToN ≈ 0, TT fS ≈ 0 Baranson (1970), Mattoo et al (2001), and Pack and Saggi (1997) have stressed on the importance of the host country’s absorptive capacity in affecting the technology transfer costs. Teece (1977) study found a negative relationship between cost of transferring technology and the host country’s absorptive capacity. Further, Eicher and Kalaitzidakis (1997), in their theoretical model, emphasized the importance of local human capital necessary to absorb FDI technology. Assumption 4: The functional form for technology transfer cost is the same irrespective of the mode of organization of fragmented production. C = C (i , ξ ) Where the cost of technology transfer is a function of the absorptive capacity of the host country, ξ and the technological complexity or the quality of low tech input, i. ∂C ∂C ∂ 2C > 0, >0 < 0, ∂i ∂i 2 ∂ξ

Technology transfer cost increases with i at an increasing rate and decreases with ξ . We assume that the technology transfer cost can be separated in i and ξ : ~ (4) C = Ω (i ) C (ξ ) As in Antràs (2005), we consider three possible organizational forms: (1) Vertical integration in the North/ Domestic outsourcing (DO) (2) Unaffiliated Supplier in the South: International Outsourcing (IO) and (3) Affiliated Supplier in the South: VFDI. Section 3.1: Domestic Outsourcing or Vertical Integration in the North Antràs (2005) assumes that vertical integration and domestic outsourcing in the north are equivalent because of complete contract enforcement in the north. To maintain this equivalence in our model we need to additionally assume that all firms in the north are capable of adopting a technology without incurring any technology transfer cost. Demand and production function is given by (1) and (2) respectively. Assuming that

10

one unit of labor produces one unit of the each input, x h and x l , the profit of the northern firm is respectively given by:

Π

N

= λ1−α ζ zα x h( 1−z )α x lzα − w N x h − w N x l

(z ) Where ζ zα = (1 − z ) h and w , h ∈ (N , S ) represents wages in the north and south respectively. This case is exactly the same as in Antràs (2005). Profit maximizing price and equilibrium profit is given by: − (1− z )

p N ( z) =

−z

wN

(5a)

α

⎛ wN Π =λ (1 − α ) ⎜⎜ ⎝ α N

⎞ ⎟⎟ ⎠

−

α 1−α

(5b)

Section 3.2: International Outsourcing- Unaffiliated supplier in south Assumption 1 and 3 together imply that the technology transfer cost in an outsourcing relationship is borne by the supplier. In an outsourcing contract, the RSI for the sourcing firm comprises of its commitment for producing the high tech input only, while the supplier makes RSI in technology transfer as well as the low tech input. Assumption 5: As in Antràs (2005), competition among southern suppliers of low-tech input drives their profit to zero. A transfer payment, T, from the supplier to the sourcing firm has to be allowed for such that the profit of the outsourcing partner is driven to zero. The profit function for the sourcing firm that outsources a low tech input to a TPV in the south is given by: Π oN = φ R − w N x h + T

(

)

= φ λ1−α ζ zα x h( 1−z )α (x l (i )) − w N x h + T Where R denotes the total revenue from the relationship and φ is the share of the sourcing firm in the total revenue. φ is also a measure of the bargaining power of the sourcing firm. In the Antràs (2003) model, the supplier (sourcing firm) makes relationship specific investment in labor (capital) while in Antràs (2005) RSI of the supplier (sourcing firm) is its commitment to produce the low tech (high-tech) input. Au contraire, in our model, we have two components of RSI for the supplier in an outsourcing relationship– investment to produce the low-tech and investment on technology transfer. The unaffiliated supplier chooses the amount of labor to employ and maximizes his profits. The objective function of an unaffiliated supplier is given by: Using ( 3′ ) and (4) we get: Π oS = (1 − φ ) R − w S L Sx − C L Sx − T zα

(

l

)− w

l

⇒ Π oS = (1 − φ ) λ1−α ζ zα x h( 1−z )α (x l (i ))

zα

(

⇒ Π oS = (1 − φ ) λ1−α ζ zα x h( 1−z )α (x l (i ))

zα

S

~ L Sx l − Ω (i ) C (ξ ) L Sx l − T

)− ⎛⎜⎜ w

S

⎝

~ + Ω (i ) C (ξ ) T S (i )

⎞ ⎟ x l (i ) − T ⎟ ⎠

~ The term ⎛⎜ w + Ω (i ) C (ξ ) ⎞⎟ is the Average Efficiency cost (AEC), that is the per unit cost of producing the S ⎟ ⎜ T (i ) ⎠ ⎝ S

low-tech input adjusted for productivity. First order condition for profit maximization for the sourcing firm and supplier and setting T to make the supplier break even leads to the following expression for the sourcing firm’s ex-ante profits and profit maximizing price in IO equilibrium:

( )

⎡ w N 1− z ( AEC )z ⎤ Π oN = λ [(1 − zα ) + ϕα (2 z − 1)] ⎢ 1− z ⎥ z ⎢⎣ φ (1 − φ ) α ⎥⎦

−

α 1−α

(6a)

11

⎡ (w N )1− z ( AEC )z ⎤ Po = ⎢ 1− z ⎥ z ⎢⎣ φ (1 − φ ) α ⎥⎦

(6b)

The profit maximizing price in Antràs (2005) when outsourcing is chosen is given by:

(w ) (w ) N 1− z

S z

(6c) α φ 1− z (1 − φ ) z Our price equation, (6b), is analogous to the profit maximizing price in Antràs (2005), that is, equation (6c), except that the southern production cost is augmented to include the technology transfer costs and adjusted for the productivity effect. N Profit of the Sourcing firm with Domestic Production Let Θ 1 = Π N = Profit of the Sourcing firm with Internatio nal Outsourcing ΠO International outsourcing is preferred to domestic production in the north, that is, Θ1 < 1 if, p =

wN AEC = ω > L 1 (φ , z , α ) . wS wS

(7a) 1−α

⎤ zα 1−α Where, L (φ , z , α ) = ⎛⎜ φ ⎞⎟ ⎡ 1 ⎜ 1 − φ ⎟ ⎢ (1 − zα ) + φα (2z − 1) ⎥ ⎝ ⎠⎣ ⎦

1 1 z

φ Since our parameter of interest is i, and ω and L 1 (φ , z , α ) are given for given i, so we examine the behavior of AEC with respect to i. Now, ~ ∂ ( AEC ) ∂Ω (i ) C (ξ ) ∂T S (i ) 1 (7b) [w + Ω (i ) C~(ξ )] − = 2 S S ∂i

∂i

T

(i )

∂i

(T (i ))

S

⎡ For optimization, we require that ∂ ( AEC ) = 0 ⇒ η Ω =ηT S ⎢1 + ∂i

⎣

⎤ wS ~ ⎥ Ω (i )C (ξ ) ⎦

(8)

S Where η Ω = ∂Ω (i ) i and η = ∂T (i ) i T ∂i Ω (i ) ∂i T S (i ) Condition (8) implies that AEC reaches an optimum when the elasticity of the cost of technology transfer with respect to technological sophistication, i, is equal to the weighted elasticity of productivity of southern labor with respect to the technological complexity. Let us define the technological complexity at which (8) holds true as: i o* . S

Proposition 1: The function AEC reaches minima at i o* . Mathematically, the cost function is convex with respect to the technological complexity, i, while the productivity function is concave, thus, AEC is a convex function. Intuitively, proposition 1 implies that at lower levels of technological complexity, a marginal increase in technological complexity of the low-tech input adds more to the productivity of southern labor vis-à-vis technology transfer costs. However, at higher levels of technological complexity, a marginal increase in technological complexity of the low tech input adds more to technology transfer costs relative to its contribution in raising the southern productivity. We would expect this because technology improvements are more costly at higher end of technological complexity. Hence, AEC falls for low levels of technological complexity, i < i o* , while for i > i o* it rises. The RHS of equation (7a) can be depicted by the bold curve in figure 2a8. The LHS of equation (7a) is simply the relative wage, which is exogenously given in our model and therefore represented by a horizontal line. The intersection of the relative wage schedule with AEC schedule defines the range of technological complexity for DO relative to IO. 8

The position of the curve also depends on parameters like z , α , φ and ξ

12

A sourcing firm stands to gain from international outsourcing vis-à-vis domestic outsourcing for two reasons: One, it gains from lower wages in the host country and two, it does not have to bother about the technology transfer costs. On the other hand, the sourcing firm may lose from IO vis-à-vis DO due to contractual distortions in IO which lead to suboptimal RSI in both the low-tech input and technology transfer. Any technological complexity below i o implies low southern labor productivity and therefore a lower output of

the offshored input. Thus, for i < i o low productivity of southern labor (due to low i) as well as the distortion from suboptimal RSI in the offshored input and technology transfer outweighs the savings from cheap southern labor. Thus DO is preferred relative to IO. Similarly, for i > i o , the cost of technology adaptation is higher relative to its contribution to labor productivity implying a higher distortion in RSI in the low tech input and therefore also in technology transfer. Thus, the sourcing firm does not prefer to outsource internationally for i > i o . Table 1 gives the effect of increase in technological complexity on the profitability of the sourcing firm and the resulting organizational decision. We depict the range of i for which IO is preferred relative to DO in figure 2a. Range of i

i < io

o o o

io < i < io

o

i > io

o

Effect on the Profit of a Sourcing Firm Extremely low efficiency of southern labor Contractual distortion and under investment in x l Above effects make production in north more profitable despite higher wages Efficiency effect of southern labor and contractual distortion outweigh the high northern wages High technology transfer cost for the unaffiliated supplier creates large distortion in RSI in x l which cannot outweigh the higher northern wages

Organizational Form DO

IO DO

Table 1: The tradeoff faced by the sourcing firm in choosing between IO and DO AEC (i , ξ ) wS

Relative Wages

ω

AEC (i , ξ ' ) wS

ξ < ξ'

ω

Domestic International Outsourcing Outsourcing

0

io

i o*

Domestic Outsourcing

io

1 Technological Complexity, i

Figure 2a: Tradeoff between domestic production and International Outsourcing

13

Proposition 2: Only at intermediate levels of technological complexity is international outsourcing preferred relative to domestic outsourcing. At low and high levels of technological sophistication, domestic outsourcing dominates international outsourcing.

Proposition 3: The range of technological complexity where IO is preferred relative to DO depends on the absorptive capacity of the host country. Further, this range is an increasing function of the host country’s absorptive capacity.

From equation (7a), we find that ∂i > 0 for i > i o* and ∂i < 0 for i < i o* . The dotted line in figure 2a ∂ξ ∂ξ indicates the shift in AEC schedule due to a rise in absorptive capacity. Thus outsourcing expands at both ends of technological complexity with a rise in absorptive capacity. Section 3.3: Vertically Integrated supplier in the South or VFDI We retain the Hart-Moore (1990) assumption that the sourcing firm has a higher share in surplus (bargaining power) in an intra-firm transaction vis-à-vis a market transaction because if the supplier misbehaves in an intra-firm relationship, the sourcing firm can fire the affiliated supplier and appropriate a share of the lowtech input. The share of the sourcing firm in the VFDI mode is given by: (9) φ = δ α + φ (1 − δ α ) > φ Where δ is the proportion of output expropriated by the sourcing firm if the manager of the subsidiary is fired. In our model, the multinational firm assumes RSI in technology transfer (to train the host country labor) in a VFDI contract. We set T ′ such that competition among suppliers drive their profit down to zero. The profit function of a multinational is given by: ~ Π fN = φ R − w N x h − Ω (i ) C (ξ ) L Sx l + T ′

(

)− w

x (i ) ~ x h − Ω (i ) C (ξ ) lS + T ′ T (i ) RSI on the part of the integrated supplier comprises of its commitment to produce the low-tech input only. The subsidiary manager maximizes: x (i ) Π Sf = (1 − φ )R − w S lS − T ′ T (i ) First order conditions for maximization of the Multinational firm’s and the supplier’s profits under the VFDI mode yields price as: = φ λ1−α ζ zα x h( 1−z )α (x l (i ))

z S ⎡ ⎞ N 1− z ⎛ w ⎜ ⎟ ( ) ⎢w ⎜ T S (i ) ⎟ ⎢ ⎝ ⎠ Pf = ⎢ z 1− z (1 − φ ) α ⎢ φ ⎢ ⎣

zα

N

⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

(10a)

It should be noted that in case of VFDI, the presence of technology transfer cost does not directly distort prices because the technology transfer cost is determined by the supplier’s decision on x l while it actually borne by the multinational firm. Thus, technology transfer cost is like a fixed cost to the multinational. Equilibrium profit of the MNC is given by: ⎡ N 1−z ⎛ w S ⎞ z ⎢ (w ) ⎜⎜ S ⎟⎟ ⎡ Ω (i ) C~(ξ ) ⎤ ⎢ ⎝ T (i ) ⎠ N Π f = λ ⎢(1 − zα ) + φ α (2z − 1) − (1 − φ )zα ⎥ ⎢ 1−z S z w ⎦ ⎢ φ (1 − φ ) α ⎣ ⎢⎣ Assumption 6: To simplify algebra, let us take φ = ½ as in Antràs (2005)

⎤ ⎥ ⎥ ⎥ ⎥ ⎥⎦

−

α 1−α

(10b)

14

Using the above assumption and simplifying further, we conclude that for VFDI to yield a positive profit9: ⎡ 2 − α (1 + δ α ) + 2zαδ α S ⎤ i < Ω −1 ⎢ w ⎥ =b α ~ ⎣ α z (1 − δ )C (ξ ) ⎦

(11)

VFDI case is in contrast to the case of IO, where Π oN > 0 for all range of i. This result comes by because in case of international outsourcing it is the unaffiliated supplier who makes RSI in technology transfer, while the resulting gain in productivity due to technological complexity is shared by both the supplier and the sourcing firm. In case of VFDI, the MNC makes RSI in technology transfer, while both parties enjoy surplus from productivity gain. Proposition 4: The sourcing firm stands to lose from VFDI (in absolute terms) if the technological complexity of the low-tech input is higher than a critical level, b defined in (11). To evaluate the relative prevalence of VFDI vis-à-vis DO we compare (10b) with (5b), the respective profit functions of the sourcing firm in the two alternative modes of organization. N Let Θ 2 = Π = Profit of the Sourcing firm with Domestic Production Profit of the Sourcing firm with VFDI Π fN VFDI is preferred to domestic outsourcing in the north if Θ2 < 1 ,that is, ⇒ ω > L 2 (φ , z , α , i , ξ ).

Where

1 T

S

⎛ φ L 2 (φ , z , α , i , ξ , w S ) = ⎜⎜ ⎝ 1−φ

Let Ψ = L 2 (φ , z , α , i , ξ , w S ).

(12a)

(i ) ⎞ ⎟⎟ ⎠

⎡ ⎤ ⎢ ⎥ ⎢ ⎥ 1−α ⎢ ⎥ ~ ⎢ ⎡⎛ 1 − 1 α (1 + δ α (1 − 2z ))⎞ − 1 α z (1 − δ α ) ⎛⎜ Ω (i ) C (ξ ) ⎞⎟ ⎤ ⎥ ⎟ ⎥ ⎜ ⎢ ⎢⎢⎜⎝ 2 w S ⎟⎠ ⎥⎦ ⎥⎦ ⎠ 2 ⎝ ⎣⎣

1−α zα

1

φ

1 z

1 T S (i )

L (φ , z , α , i , ξ , w S ) ∂T S (i ) ∂Ψ ∂L 2 (φ , z , α , i , ξ , w S ) 1 = . S − 2 . 2 ∂i ∂i 4443 T (i ) ∂i 3 T S (i ) 14442 12

[

+

]

(12b)

+

Where ⎤ ⎡ ⎥ ⎢ ~ ∂L 2 (φ , z , α , i , ξ , w ) 1 ⎥ ⎛ 1 − α ⎞(1 − δ α ) ∂Ω (i ) C (ξ ) = L 2 (φ , z , α , i , ξ , w S ) ⎢⎢ ⎜ ⎟ ~ ∂i ∂i wS ⎛ C (ξ ) ⎞ ⎥ ⎝ 2 ⎠ ⎥ ⎢ ⎛⎜ 1 − 1 α (1 + δ α (1 − 2z ))⎞⎟ − 1 α z (1 − δ α ) ⎜ Ω (i ) ⎟ ⎜ w S ⎟⎠ ⎥⎦ ⎢⎣ ⎝ 2 ⎠ 2 ⎝ * Let the level of technological complexity at which ∂Ψ = 0 be i f . ∂i S

The first term in the above derivative denotes the effect of an increase in technological complexity on technology transfer costs, which is positive, while the second term is its effect on labor productivity. Since Ω (i ) is a convex function in technological complexity, i, it implies that L 2 (φ , z , α , i , ξ ) is also convex. Further,

1

T (i ) S

is also a convex function in i. Therefore, Ψ is a convex function. Thus, for i < i *f , an increase

in technological complexity increases the profitability of the sourcing firm by increasing productivity of the host country’s labor relative to an increase in technology transfer cost.

Proposition 5: For reasons corresponding to proposition 1, for i < i *f , ∂Ψ < 0 while for i > i *f , ∂Ψ > 0 ∂i

9

∂i

For z>½ or z<½,the numerator is always positive.

15

* It reaches minimum at i f and then it rises. We graph the RHS of equation (12a), that is the function Ψ in

bold in figure 2b10. The LHS of equation (12a) is the relative wage of northern labor, which is exogenous in our model and is shown by a horizontal line. The intersection of the two functions at i f and i f defines the

range of technological complexity for VFDI and DO. The prevalence of DO below i f and above i f can be explained as in case of international outsourcing. In table 2 we describe the trade-off faced by a multinational in choosing between VFDI and DO.

Range of i

i
o o o

if
o

i >i

o

f

Effect on the Profit of a Multinational Firm Extremely low efficiency of southern labor Contractual distortion and under investment in x l Above effects make production in north more profitable despite higher wages Efficiency effect of southern labor and contractual distortion outweigh the high northern wages High technology transfer cost for the multinational firm cannot outweigh the higher northern wages

Organizational Form DO

VFDI DO

Table 2: The tradeoff faced by a Multinational firm in choosing between VFDI and DO

Proposition 6: Only at intermediate levels of technological complexity is VFDI preferred relative to domestic outsourcing. At low and high levels of technological sophistication, domestic outsourcing dominates VFDI.

Proposition 7: The range of VFDI (vis-à-vis DO) is increasing in the absorptive capacity of the host country.

From equation (12a), it is clear that ∂i > 0 for i > i *f and ∂i < 0 for i < i *f . Thus the possibility of hosting ∂ξ ∂ξ VFDI vis-à-vis DO expands at both ends with a rise in absorptive capacity of the host country. The dotted line in figure 2b indicates the shift in Ψ due to rise in absorptive capacity. Thus VFDI expands at both ends of technological complexity with a rise in absorptive capacity.

ω

Relative Wages

L 2 (φ , z , α , i , ξ , w S ).

1 T S (i ) L 2 (φ , z , α , i , ξ ' , w S ).

1 T S (i )

ξ < ξ' ω

Domestic Outsourcing

0

i

Vertical FDI

f

i *f

Domestic Outsourcing

i

b

f

b’

1

Technological Complexity, i

Figure 2b: Tradeoff between northern production and Vertical FDI 10

The position of the curve also depends on parameters like z , α , φ , ξ and δ

16

Section 3.4: Comparing International Outsourcing with VFDI To compare the profit functions of the sourcing firm in the two alternative regimes of offshoring, that is VFDI and IO, we need to compare equation (6a) with (10b). N Let Θ = Π o = Profit of the Sourcing firm with International Outsourcing Profit of the Sourcing firm with VFDI Π fN

Using equation (9) and φ = ½, we get: 1− α

⎤ αz ⎡ ⎛ 1 ⎞ ⎥ ⎢ ⎜1 − α ⎟ ⎥ ⎢ 2 ⎠ ⎝ Θ =⎢ ~ ⎞ ⎤⎥ ⎡⎛ 1 ⎛ ( ) ξ C 1 ⎢ ⎢⎜ 1 − α (1 + δ α (1 − 2z ))⎞⎟ − α z (1 − δ α ) ⎜ Ω (i ) ⎟ ⎥ S ⎟⎥ ⎜ ⎢ ⎢⎣⎝ 2 w 2 ⎠ ⎝ ⎠ ⎥⎦ ⎥⎦ ⎣

⎤ ⎡ ⎡ ⎤ wS 1 ⎥ ⎢ ⎢ ⎥ ~ 1− z S ⎢ w + Ω (i ) C (ξ ) ⎦ α α ⎥ ⎣ (1 + δ ) z (1 − δ )⎦ ⎣

(

)

To have meaningful comparison between VFDI and international outsourcing, we need to hold i < b , because we know that for i > b either IO or DO exists, but not VFDI. Let us now look at the partial derivative of Θ with respect to i . ⎤ ⎡ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ α S (1 − α )(1 − δ ) w ∂Θ ∂Ω (i ) ~ ⎢ ⎥ C (ξ ) − S =Θ ~ ⎥ ⎢ i ( ) ( ) ∂i ∂ w Ω i C ξ + ⎛ ⎞ 14243 ⎢ ⎜ ~ 443 ⎥ C (ξ ) ⎟ 1442 + ⎞ 1 ⎛ 1 α α + ⎥ ⎢ 2 ⎜ ⎜ 1 − α (1 + δ (1 − 2z ))⎟ − α z (1 − δ )Ω (i ) S ⎟ w4 2 444 ⎥ ⎢ ⎜⎜ ⎝14244424444 42444 3 ⎟⎟ 3⎠ 1 + + ⎠ ⎦⎥ ⎣⎢ ⎝

∂Θ < 0 , that is, as technological complexity increases, the profitability from VFDI increases if: ∂i ⎡ 1 − δ α − 2(1 − z )αδ α ⎤ wS ⎥ = a i < Ω −1 ⎢ α ~ ⎣ ((1 − α ) + α z ) (1 − δ )C (ξ ) ⎦

(13)

Clearly, a < b , else production will never be offshored via intra-firm contract. ∂Θ > 0 , that is, as technological complexity increases, the profitability from outsourcing increases if: ∂i

a
Proposition 8: If in equilibrium, international outsourcing occurs for i < a , then a small increase in technological complexity, still less than ‘a’, will induce a regime switch from international outsourcing to VFDI.

Proposition 9: If in equilibrium, VFDI occurs in the range a < i < b , then a small increase in technological complexity in this range will switch the organizational form of fragmented production to international outsourcing. Thus, at higher level of technological complexity, the organizational form of fragmented production is likely to be an external one. Proposition 8 and 9 helps us narrow down five possible ways in which a sourcing firm may choose its organizational form with changes in technological complexity of the offshored input. We depict the possible configuration of organizational forms in figure 311.

11

In figure 3, we have assumed that i f < a and b < i o

17

DO

IO

VFDI

a

io

0

IO

DO

b

DO

1

io

Technological Complexity, i

IO

VFDI

0

DO

b

DO

1

io

0

Technological Complexity, i

Case 3

DO 0

a

1

io

VFDI if

a

IO

b

DO io

1

Technological Complexity, i

Case 4

IO

io

b

Case 2

IO

a

if

DO

Technological Complexity, i

Case 1

DO

IO

a

io

0

VFDI

DO

b

1

io

Technological Complexity, i

Case 5 Figure 3: Possible configuration of organizational forms It should be noted that case 1 and 2 are observationally equivalent because the order of the chosen organizational forms with respect to the technological complexity of the input is the same in the two cases. Our task is now is to chart out the characteristics (in terms is z) of the final good for which we would observe these cases. We would observe case 1 (or 2) if the cost schedule for the VFDI mode cuts the cost schedule for the IO mode from above at low i and cuts it from below at high i. This implies: At high i: ∂Ψ > ∂AEC ∂i ∂i ∂ ∂ AEC Ψ At low i: < ∂i ∂i We compare the AEC and the Ψ function to chart out the preferred organizational mode at various ranges of technological complexity. To compare the slopes of the cost curves in the VFDI and the IO mode, consider equation (7b) and (12b). ∂Ψ ∂AEC if: ~ (14a) (C (ξ ) Ω (i ) + w S ) (1 − α ) (1 − δ α ) < 2 Γ w S < ∂i ∂i ~ Where Γ = ⎛⎜ 1 − 1 α (1 + δ α (1 − 2z ))⎞⎟ − 1 α z (1 − δ α ) ⎛⎜ Ω (i ) C (ξ ) ⎞⎟ S ⎟ ⎜ ⎝

⎠

2

~ ⎛ ∂Γ C (ξ ) ⎞ 1 = α − α (1 − δ α ) ⎜⎜ Ω (i ) S ⎟⎟ ∂z w ⎠ 2 ⎝

2

⎝

w

⎠

Clearly, it is more likely for ∂Γ > 0 if the technological complexity of the offshored input is low. Hence, the ∂z

inequality (14a) is more likely satisfied for a high z final good , that is a low tech good, when i is low. 18

Now, treating the case where we need ∂Ψ > ∂AEC . This holds if: ∂i ∂i (C~(ξ ) Ω (i ) + w S ) (1 − α ) (1 − δ α ) > 2 Γ w S

(14b)

Clearly, the LHS is higher for higher i and RHS is lower for higher i. Also, it is more likely for ∂Γ < 0 if the ∂z

technological complexity of the offshored input is high. Hence, inequality (14b) is more likely satisfied for a low tech good, when i is high. In sum, organizational forms outlined in case 1 (or 2) are typically observed for a low-tech good. We draw the cost schedules for the VFDI and IO modes for these cases in figure 4a. Relative Wages

DO

IO

VFDI

IO

DO

VFDI : L 2 (φ , z , α , i , ξ , w S ).

1 T S (i )

IO :

ω

0

io

P

a A

b

io

L 1 (φ , z , α ) ~ . w S + Ω (i )C (ξ ) w S T S (i )

(

)

1

Technological Complexity, i

Figure 4a: Possibility of multiple switches between regimes for a low tech good Now consider case 3. Case 3 requires that the VFDI cost schedule cuts the IO cost schedule from below at

L any given technological complexity, high or low. This implies that ∂Ψ > ∂AEC when Ψ = s1 AEC . w ∂i ∂i This happens if: (15a) i >a Where a is given by (13). Now: α ∂a 1 (15b) > 0 if δ > 3 ∂z

and

∂a < 0 if ∂z

δα <

1

(15c)

3

Condition (15a) is more likely satisfied for a high tech good (low z) if δ α > tech good if δ

α

<

1

3

1

3

(condition 15b) and for a low-

(condition 15c). Thus, case 3 is valid configuration for a high tech as well as a low tech

good depending on the value of δ α . We depict (with one possible configuration of AEC and Ψ ) this case in figure 4b below. As we will see below, a high value of δ simply increases the incentive for the sourcing 19

firm to make higher RSI in the high-tech input in the VFDI mode because she can appropriate a greater share of output if the supplier misbehaves. Since RSI in x h is relatively more important for a high-tech good (a low z), this implies that VFDI is more likely and occurs for a larger range of i when for a high-tech good when δ is high. On the other hand, when δ is low, then it motivates the supplier to make higher RSI in x l in the VFDI mode. Since x l is relatively more important for a low-tech good, we are more likely to observe VFDI

Relative Wages

even for a low-tech good in low to mid ranges of i when δ is low.

DO

VFDI

DO

IO

VFDI : L 2 (φ , z , α , i , ξ , w S ).

1 T S (i )

IO :

L 1 (φ , z , α ) ~ . w S + Ω (i )C (ξ ) w S T S (i )

(

)

ω

0

i

a

f

B

b

io

1

Technological Complexity, i

Figure 4b: Possible organizational regimes for a high-tech good Cases 4 and 5 correspond directly with the Antràs (2005) results. In case 4, the VFDI cost schedule would completely envelop the IO cost schedule (in this case for all technological complexity below b), while the reverse holds for case 5. As in Antràs (2005), case 4 corresponds to a high tech good, while case 5 corresponds to a low tech good. We now discuss the intuition for why case 1 (or 2) is typical for a low-tech good only while case 3 could be for a low-tech as well as a high tech good. The intuition for cases 4 and 5 is clearly laid in the Antràs (2005) model which we have briefly discussed in section 2 of the paper. For a low-tech good, the supplier’s contribution is relatively more important for the relationship. Therefore, to develop the intuition for organizational regimes across varying levels of technological complexity, we consider the RSI in the low-tech input in the two alternative modes. The first order conditions for maximizing the sourcing firm’s (choosing x h ) and the supplier’s (choosing x l ) profit gives rise to the following relationship between x h and x l in an outsourcing relationship: φ 1 − z AEC o (16a) xl x ho = 1−φ z wN Where x lo and x ho is the RSI in the high tech and low tech input respectively in an outsourcing relationship. Solving the model using equation (1), (2) and (6b) yields the following expression for x lo : x lo = λ φ

α (1−z ) 1−α

αz

1

(1 − φ )1+ 1−α z α 1−α

α (1−z ) N − 1−α

(w )

⎛

αz ⎞

( AEC )−⎜⎝ 1+ 1−α ⎟⎠

(17a)

Since AEC is a convex function of the technological complexity, the above expression implies that x lo would be a concave function of i. 20

Now consider the case for VFDI. The first order necessary conditions for maximizing the sourcing firm’s and the supplier’s profit gives rise to the following relationship between x h and x l :

φ

1−z x ho = 1−φ z

wS

T S (i ) f xl wN

(16b)

Where x l f and x hf is the RSI in the high tech and low tech input respectively in an intra-firm relationship. Solving the model using equation (1), (2) and (10a) yields the following expression for x l f : x lf = λ φ

α (1−z ) 1−α

(1 − φ )

1+

αz 1−α

zα

1 1−α

(w ) N

−

α (1−z ) 1−α

⎛ wS ⎞ ⎜⎜ S ⎟⎟ ⎝ T (i ) ⎠

⎛ αz ⎞ − ⎜ 1+ ⎟ ⎝ 1−α ⎠

(17b)

Since T S (i ) is an increasing function (concave) of the technological complexity, the above expression implies that x l f would also be an increasing (concave) function of i. Note that for φ =

1

αz α (1−z ) x lf 1+ 2: = (1 + δ α ) 1−α (1 − δ α ) 1−α o xl

⎛ ⎞ wS ⎜ S ⎟ ⎜ w + Ω (i ) C~(ξ ) ⎟ ⎝ ⎠

⎛ αz ⎞ −⎜ 1+ ⎟ ⎝ 1−α ⎠

(17c)

Since Ω (i ) is an increasing (convex) function of the technological complexity, the above expression implies that the RSI in the low tech input in the VFDI mode increases with a rise in technological complexity and at an increasing rate vis-à-vis the IO mode. This implies that a rise in technological complexity increases distortions in the investment of the low tech input in the IO mode, while in the VFDI mode it only raises the investment in the low tech input. The rationale for this result is embedded in the assumption that the technology transfer cost is borne by the unaffiliated supplier in the IO mode while it is undertaken by the multinational firm in the VFDI mode. When technological complexity rises, the unaffiliated supplier must weigh the ensuing benefit of rising southern labor productivity against the rising technology transfer cost. As equation (17a) suggests, her RSI in the low tech input rises at a decreasing rate, reaches a maxima and then falls. Per contra, the affiliated supplier does not need to worry about the technology transfer cost when technological complexity rises. For a subsidiary manager, a rise in technological complexity brings gains in terms of rising southern labor productivity and no costs of technology transfer. Therefore, her investment in the low tech input rises with a rise in technological complexity. As equation (17b) suggests, an affiliated f supplier’s investment in the low tech input increases at a decreasing rate. Therefore, x lo must rise at an xl increasing rate. Is it possible for x lo > x l f ? Yes, but only for low levels of technological complexity, i and if (necessary but not sufficient condition) α (1−z ) 1−α

(1 + δ ) α

αz α 1+ 1−α

(1 − δ )

(18a) <1 Note that the above function is decreasing in z, implying that this condition is more likely satisfied for a higher value of z, that is, for a low tech good. When z is low, then condition (18a) is not likely satisfied and x lo < x l f even for low i. Note that the LHS of condition (18a) is decreasing in δ . Therefore, condition (18a) imposes an lower limit on δ . The existence of IO (and not VFDI) for a high value of δ in a low tech good can be easily explained12 as in case 3 depicted in figure 4b. However, as i increases, the RSI in the low-tech input falls in the IO mode vis-à-vis the VFDI mode as the unaffiliated supplier needs to incur the technology transfer costs. Thus, at mid range of technological complexity, the sourcing firm prefers to have an intra-firm production transfer even for a low-tech good. Further increases in i reduces the profitability of the sourcing firm in the VFDI mode vis-à-vis the IO mode even though the distortion in x l in the IO mode is high. This 12 If δ is high the RSI in the low-tech input in the VFDI mode is low. Thus, if the final good is low-tech, then the sourcing firm prefers to have an unaffiliated supplier.

21

happens because the cost of technology transfer in the VFDI mode is borne by the sourcing firm. Therefore, at higher levels of technological complexity the sourcing firm prefers to engage an unaffiliated supplier. There are two possible ways in which the functions x lo and x l f could be mapped together for a low tech good. In figure 5a we depict the case with high z (low-tech good) where condition (18a) is satisfied such that x lo > x l f for low i while in figure 5b we depict the case with high z, where condition (18a) is never satisfied. The possible configuration of organizational modes for a low tech good outlined in figures 5a and 5b are based only on the RSI in the low-tech input. They ignore the technology transfer cost that the sourcing firm needs to bear in an intra-firm production transfer. Therefore, in each case, the IO regime may set earlier than technological complexity b. Figure 5a corresponds to case 1 (or 2) of figure 3, while figure 5b corresponds to case 3. We do not depict a graph for equilibrium RSI in the low-tech input for a high tech good (which would be similar to figure 5b) because for a high-tech good, the RSI in the low-tech input is not as significant as the RSI in the high-tech input. A sourcing firm’s organizational choice for a high tech good would depend on the RSI in the alternative modes in the high tech input rather than the low-tech input. DO

IO

VFDI

DO

IO

RSI in low tech input

xl

VFDI

DO

IO 0

io

P a

A

b

io

1

Technological Complexity, i

Figure 5a: Equilibrium organizational forms for a low-tech good when x lo > x l f for low i For a low tech good, the RSI in the low-tech input is extremely important and therefore the sourcing firm decides on its organizational form based on which mode generates larger RSI in the low tech input as well as the technology transfer cost. As is clear from the above figure, if condition (18a) is satisfied such that IO leads to higher RSI in the low-tech input at low i, then, the sourcing firm decides on DO for i < i o . For the range of technological complexity i o < i < P , the sourcing firm prefers IO, for P < i < A it prefers VFDI and for A < i < i o it decides on IO. If condition (18a) is not satisfied, then, the RSI in low tech input is shown in figure 5b. In such a case, the sourcing firm prefers to have VFDI for i f < i < B and IO only for

B
22

xl

VFDI

DO

IO

RSI in low tech input

DO

VFDI

DO

IO 0

i

f

a

B

b

io

1

Technological Complexity, i

Figure 5b: Equilibrium organizational forms for a low-tech good when x lo < x l f for all i For a high tech good, the RSI in the low-tech input is always higher in the VFDI mode vis-à-vis the IO mode for all i. However, when the final good is high-tech, the high tech input assumes greater significance and the final good producer cannot decide on its organizational form based on the RSI in the low tech input. It needs to evaluate the investment in the high-tech input under the alternative organizational forms. Using equation (16a), (16b), (17a), (17b) and φ = 1 2 , we get that: αz α (1−z ) x hf α 1− α + 1 α 1−α ( ) ( ) = + − δ δ 1 1 x ho

⎛ ⎞ wS ⎜ S ⎟ ⎜ w + Ω (i ) C~(ξ ) ⎟ ⎝ ⎠

⎛ αz ⎞ − ⎜⎜ 1+ ⎟⎟ ⎝ 1−α ⎠

f f By examining the above condition, as in case of x lo , we can say that x h increases at an increasing rate with i xl x ho because Ω (i ) is an increasing (convex) function of the technological complexity. For x ho > x hf , we need that at low i:

α (1−z ) +1 1−α

(1 + δ ) α

αz α 1−α

(1 − δ )

(18b) <1 (Necessary but not sufficient condition13) This is a much stronger condition vis-à-vis condition (18a). This condition is also more likely satisfied for a high z, that is, for a low-tech good. For a high tech good, we can state with certainty that condition (18b) is not satisfied and x ho < x hf . The behavior of RSI in the high tech input for a high-tech final good is represented in figure 5c. In such a case the organizational choice is clear. The sourcing firm prefers to have VFDI for technological complexity lying in the range i f < i < B (where B < b) and IO for B < i < i o because the RSI in the high-tech input is always higher in the VFDI mode and VFDI is not feasible beyond technological complexity b. Deciding the organizational form solely on the basis of RSI in the high tech input would be inappropriate because the sourcing firm needs to incur the technology transfer cost in an intra-firm relationship. Therefore, in figure 5c the IO regime may emerge at i = B, well before the technological complexity of the offshored input hits the level b. This case corresponds to case 3 in figure 3.

13

Note that the LHS of condition (18b) is increasing in δ . Therefore, condition (18b) imposes an upper limit on δ .

23

xh RSI in high tech input

DO

VFDI

DO

IO

VFDI

DO

IO 0

i

f

a

B

b

io

1

Technological Complexity, i

Figure 5c: Equilibrium organizational forms for a high-tech good ( x ho < x hf for all i) Section 3.5: Application of the Model Section 3.5.1: Effects of a fall in Northern Wages due to Offshoring It is observed that the rate of growth of wages in countries which host offshoring contracts is about 20% per annum, which is much higher than global average. At the same time, there are concerns in the sourcing countries about job losses and fall in relative wages. Glass and Saggi (2001) document the fall in relative wages in the north through their theoretical model of offshoring intermediate inputs to the south. However, they do not deal with the issue of how this change in relative wages would impact the relative prevalence of VFDI and IO mode. We use our model to analyze the relative prevalence of the two organizational forms. To simplify our analysis, we consider the case of a fall in home country wages without any change in the host country wages. In figure 6a, we show the impact of a fall in northern wages which leads to a fall in relative wages from ω to ω ′ . The bold lines define the new range of technological complexity for international production sharing. In figure 6a, we observe a fall in off-shoring at the two ends of technological complexity. With a fall in north-south wage differential, one moves from multiple regime switch situation to a unique regime switch situation. The result is intuitive because a fall in relative northern wages decreases the profitability of the sourcing firm through offshoring to the host country at all levels of technological complexity. Offshoring falls at both the low as well as high end of technological complexity because at the two ends, the wage savings from offshoring are lower vis-à-vis contractual distortions. Since outsourcing is dominant at high end (and low end) of technological complexity, therefore, for low-tech goods, we would expect outsourcing vis-à-vis VFDI to decline with a fall in north-south wage differential. The effect of a fall in northern wages due to offshoring on a high tech good or a low tech good (when δ is low) is similar. We show this in figure 6b. In both cases, outsourcing falls relative to VFDI. The reason for this can again be explained in terms of the RSI in the low-tech and high tech input in the two modes in the presence of technology transfer costs. At low i, IO occurs in a low-tech good case only when δ is high. If however, relative wages in the north fall, then the range of technological complexity for offshoring falls. At this level of technological complexity, VFDI is more likely than IO. Similar is the case with high-tech good. Thus, outsourcing vis-à-vis VFDI falls with a rise in northern wages.

24

Relative Wages

DO

IO

IO

VFDI

DO

VFDI : L 2 (φ , z , α , i , ξ , w S ).

1 T S (i )

IO :

ω

L 1 (φ , z , α ) ~ . w S + Ω (i )C (ξ ) w S T S (i )

(

)

ω′

DO

0

io

VFDI

P

IO

DO

a A

b

io

1

Technological Complexity, i

Relative Wages

Figure 6a: Impact of a fall in relative wage for a low-tech good (when δ is high)

DO

VFDI

DO

IO

VFDI : L 2 (φ , z , α , i , ξ , w S ).

1 T S (i )

IO :

L 1 (φ , z , α ) ~ . w S + Ω (i )C (ξ ) w S T S (i )

(

)

ω

ω′ DO

0

i

f

VFDI

a B

IO

DO

b

io

1

Technological Complexity, i

Figure 6b: Impact of a fall in relative wage for a high-tech good (or a low-tech good when δ is low) If relative wage differential between the north and south falls due to rising wages in the host country, the qualitative nature of our results would not change14. Our model therefore suggests that a fall in wage differential between the source and the host country (due to falling wages in the home country or rising wages in offshoring destinations) represents a loss in its comparative advantage. This phenomenon is likely to slow 14 Besides the downward shift of the relative wage schedule, the cost curves would also shift, which makes the exact analysis difficult to follow without specific functional forms or numerical simulations.

25

down the process of international production fragmentation as well as the growth in income and employment in the host country emanating from the offshoring industry unless the host country works on to build new sources of comparative advantage. One such source of comparative advantage has been discussed in Acemoglu, Antràs and Helpman (2006), where better contracting institutions can influence the level of production sharing by impacting relative productivity of the final good sector. Our model suggests yet another source of comparative advantage – the host country’s absorptive capacity and technology expertise. A high level of absorptive capacity and proficiency in technology can sustain a more technologically sophisticated good by lowering the cost of technology transfer and hence widen the range of offshoring. Section 3.5.2: An Exogenous rise in the Host Country’s Absorptive Capacity For a low-tech good, a rise in absorptive capacity of the host country lowers the cost of technology transfer and therefore increases the range of technological complexity for offshoring. Is the increase in offshoring more in the form of intra-firm contracts or is it more of an external transaction? In this sub-section, we attempt to evaluate the relative increase in one mode of organization vis-à-vis the other due to a rise in absorptive capacity of the host country. By partially differentiating Θ wrt ξ , we get that for i < a , ∂Θ < 0 and for i > a , ∂Θ > 0 , implying ∂ξ

∂ξ

Relative Wages

that IO falls relative to VFDI for low i while it increases for high technological complexity. We depict the effect of a rise in ξ in a low-tech good with high δ in figure 7a. DO

IO

VFDI

IO

DO

VFDI : L 2 (φ , z , α , i , ξ , w S ).

1 T S (i )

IO :

ω

DO IO

0

io

VFDI

P

a A

IO

L 1 (φ , z , α ) ~ . w S + Ω (i )C (ξ ) w S T S (i )

(

)

DO

b

io

1

Technological Complexity, i

Figure 7a: The effect of rise in the absorptive capacity on a low-tech good In the case of a low-tech good with high δ , IO may increase more than VFDI because IO has the opportunity to expand at both ends of technological complexity vis-à-vis domestic outsourcing. For a hightech good case with high δ , (or a low-tech good when δ is low), we need that ∂Ψ > ∂AEC such that ∂Θ > 0 ∂i

∂i

∂ξ

for i > a . Therefore, IO expands vis-à-vis VFDI. However, VFDI also expands relative to DO at lower end of technological complexity and IO expands vis-à-vis DO at higher end of i. The model cannot determine whether VFDI increases more vis-à-vis IO without choosing specific functional forms for Ω (i ) , T s (i ) and

~ C (ξ ) . We depict the effect of a rise in absorptive capacity on a high-tech good in figure 7b.

26

Relative Wages

DO

VFDI

DO

IO

VFDI : L 2 (φ , z , α , i , ξ , w S ).

ω

1 T S (i ) IO :

DO 0

i

IO

VFDI f

a

B

b

L 1 (φ , z , α ) ~ . w S + Ω (i )C (ξ ) w S T S (i )

(

)

DO

io

1

Technological Complexity, i

Figure 7b: The effect of rise in the absorptive capacity on a high-tech good Section 4: Discussion Perhaps a widely held notion is that firms do not outsource the production of technologically complex inputs. The trend to buy technologically complex inputs from unaffiliated suppliers is not completely absent though. For instance, Dell contracts out the design for notebooks, Personal Computers, digital televisions. Hewlett-Packard seeks external assistance to develop servers and printers. Motorola purchases designs for its cheapest phones from unaffiliated suppliers. These firms acquire complete designs of digital devices from Asian developers, modify them to suit their own specifications and finally stamp the products with their own brand name. The trend is fast spreading from electronics sector to navigation systems, pharmaceutical and even consumer goods. For example, Boeing is working with HCL Technologies, an Indian third party service provider, to co-develop software ranging from navigation systems and landing gear to the cockpit controls. Similarly, 20% of Procter & Gamble’s new product ideas come from external source. Besides reasons relating to labor cost arbitrage, outsourcing technologically complex products (within the basic stage of production) can also be rationalized by the fact that these products require specialized skills and knowledge which can only be offered by a broad network of specialists around the world. This helps in explaining why many pharmaceutical companies have begun to outsource basic research15. Our analysis in the previous section shows that whether a final good is low tech or high tech, it is always offshored externally if the technological complexity of the offshored input is high. This is because at higher levels of technological complexity, the cost of technology transfer is high which creates a strong disincentive for the sourcing firm to undertake an intra-firm production transfer. In case of VFDI the MNC makes RSI in technology transfer while the ensuing productivity gain is shared by the supplier as well. As technological complexity of the low-tech input crosses a threshold, technology transfer cost rise and the MNC is no longer willing to bear the cost of technology transfer. At this point, the MNC is better off sharing a larger part of the surplus in return for the unaffiliated party’s RSI in technology transfer. Our result shares a 15 In contrast to Antràs (2005), Acemoglu, Aghion and Zilibotti (2005) show that firms closer to technology frontier (intensive in high tech input) are more likely to outsource high tech inputs to focus on R&D. Inputs of a high-tech good are more technologically complex than the inputs of a low tech good. Thus, their model is also capable of generating a result similar to ours that more technologically complex inputs are outsourced.

27

similarity with Bartel et al (2005). In their closed economy model, an increase in the speed of technological progress encourages outsourcing vis-à-vis intra-firm production transfer. Our model proposes that a final good firm with a more technologically complex input will choose to outsource it (provided the host country has a threshold level of absorptive capacity). The forces driving similar results in the two models are not very different. In Bartel et al (2005) model, acceleration in the pace of technological change raises technology adoption costs for the final good firm and hence increases the per-period unit cost of producing in-house. This shifts the demand for outsourcing outwards irrespective of its service price because it allows sourcing firms to use services based on leading edge technologies without incurring the large and recurrent fixed costs of adopting these new technologies. In our model, we assume that the cost of technology transfer is borne by the supplier only in the outsourcing mode. Therefore, when technological complexity is high, the sourcing firm prefers to outsource the production of the low-tech input. To ensure that the possibility of outsourcing at higher technological complexity does not only remain a theoretical opportunity, an active participation of the host country’s government in globalization process is mandatory. Our model is suggestive of a strong technology subsidy policy in the host country. Since the sourcing firm is not likely to make an intra-firm production contract at high levels of technological complexity, the host country government should subsidize the domestic vendors’ technology investment so as to enhance its overall participation in the global production. The role of the host nation’s government is also critical in raising the level of education and the absorptive capacity of the country. A rise in absorptive capacity decreases the cost of technology transfer and directly boosts international sourcing. Some leading companies have simultaneously adopted a mix of captive and outsourced services wherein more complex and core processes are being handled by the captive unit, for instance in Credit card companies, complex processes which analyze customer behavior are usually offshored within the firm. If a country has low absorptive capacity, its third party outsourcing service providers may get trapped in low value-add work, that is provide service with i o < i < P as is depicted in figure 4a. Given the possibility of multiple switches in regime, a TPV can potentially jump to high value and technologically complex work if the supplier can manage to lower its production costs. When offshoring began in India, it was limited to low end jobs like call centers. Third party outsourcing firms like Progeon and Wipro Spectramind16 have proved that the ability to handle complex processes can be acquired overtime. The cost of an unaffiliated supplier can be lowered if the host country enhances its absorptive capacity through investment in human capital. An increase in absorptive capacity decreases the cost of technology transfer, reduces the distortion in RSI and hence raises the profitability of the sourcing firm. Using equation (15a), we can say that, when z is low, VFDI is more sensitive to technological complexity and therefore an increase in the host country’s absorptive capacity reduces technology transfer costs and raises the relative profitability from VFDI17. Thus, it is likely that offshoring in the form of VFDI increases by a greater proportion as a result of a rise in absorptive capacity when z is low. On the other hand, when z is high, equation 15(b) gives us the result that the relative prevalence of outsourcing increases vis-à-vis VFDI as a result of rising absorptive capacity. A dynamic interpretation of our result is also possible if we view that the technological complexity of inputs rise with time due to technological improvements in the north. Consider figure 4a, where the horizontal axis now represents not only i, the technological complexity, but also time. A recent trend in offshoring business is the method of “Build-Operate-Transfer” (BOT) whereby a TPV uses its skills and knowledge of the local market to create an offshore production unit on behalf of a multinational firm. When this unit reaches a critical mass and a certain level of maturity, the unit is transferred to the MNC by the vendor. The offshore unit is then controlled by the MNC. For example, Aviva Plc, a United Kingdomheadquartered insurer, testing the waters in the Indian market, decided to opt for the BOT model. This model can explain the switch from IO to VFDI with an increase in technological complexity of the input of a low-tech good.

16 Progeon, a subsidiary of Infosys, has concrete plans to enter into more complex BPO activities as part of its expansion strategy. Progeon has formed a partnership with Aceva Technologies Inc to offer finance and accounting solutions. Wipro Spectramind is the second largest third-party offshore BPO providing services in insurance processing, telemarketing, mortgage processing, and technical support services apart from customer services. 17 This is because the fall in technology transfer cost due to a rise in absorptive capacity is higher for VFDI when z is low.

28

Now consider figure 4a and 4b, where again the horizontal axis represents both time and i. A switch from VFDI to outsourcing mode is likely with time as a captive supplier’s maturity evolves. This is exemplified by the transition of a captive of General Electric (GECIS) to GENPACT18 in December 2004. GECIS was a subsidiary of GE in India and in the year 2004, it transformed to a TPV after eight years of operations in India. Convergys, an international BPO firm was also a spin-off from Cincinnati Bell and similarly, Vertex was the internal services arm of United Utilities. There are other firms in this league, for example, Xchanging, Fidelity, T Systems, Siemens Business Services, and so on. Europe, especially Germany, is dominated by IT and business services companies that started out as captives. World Network Services (WNS), one of the biggest third party outsourcing vendor in India, started out as a captive offshore revenue accounting center for British Airways. It has done a very credible job of transforming itself to a dynamic outsourcing services company. Such a transition is predicted in our model for both the low-tech as well as high-tech goods. Empirical evidence relating to our model is found in Borga and Zeile (2004). They regress the volume of intra-firm trade on a number of parent firm related factors, host country characteristics and affiliate related variables. Affiliate R&D intensity is found to be negatively related to the volume of intra-firm trade. We can interpret R&D intensity of affiliate to be a measure of technological complexity the offshored input. The result would then imply that as the technological complexity of the input rises, the probability of VFDI falls. The second important result of Borga and Zeile (2004) crucial for our paper relates the volume of intra-firm trade to the education standards of the host country and its income. Their results suggest that the volume of intra-firm trade falls if the host country has higher levels of education or income. This matches with our model’s intuition for low tech goods. International outsourcing is more sensitive to technology transfer cost for low tech goods. Therefore, with a rise in absorptive capacity which decreases the technology transfer cost, our model predicts a rise international outsourcing or a fall in intra-firm trade for low-tech products. Section 5: Conclusion This paper builds on the framework provided by Antràs (2005). We emphasize the importance of contractual differences between the VFDI and outsourcing and propose that in case of an intra-firm production transfer, a significant proportion of the technology transfer cost is borne by the sourcing firm. Per contra, in a relationship with an outside contractor, the cost of technology acquisition or assimilation assumes more importance which is undertaken by the supplier. The result obtained in the Antràs (2005) model is a special case of our model where the host country and the home country are equally productive. Specifically, in the Antràs model VFDI is preferred to outsourcing if the intensity of high-tech input is high. However, in our model, the technological complexity of the offshored input and the absorptive capacity of the host country is a critical variable that affects the internalization decision of a sourcing firm. In case of a high-tech good, the sourcing firm prefers to an internal transfer only for intermediate range of technological complexity. This is because high technological complexity increases the cost of technology transfer for the MNC and therefore reduces profitability from VFDI vis-à-vis IO at higher ends of technological complexity. On the other hand, even for a low-tech good, the sourcing firm may prefer an intra-firm transfer if the complexity of offshored input is low or mid range. This is explained by the RSI in the low-tech input and the incidence of technology transfer costs on the sourcing firm in the alternative modes of organization. A dynamic interpretation of our model may be used to explain a BOT relationship as well as recent transitions from captive units like GE Capital to GENPACT or British Airways to WNS. In future, it may be valuable to broaden this research by looking at the relative growth and welfare effects of VFDI and outsourcing. Another useful extension can make the current model dynamic, where technology becomes more complex with each instant and improves productivity along with the evolution of absorptive capacity. This may help in highlighting new sources of comparative advantage for the host country. In such a model we would certainly expect absorptive capacity to form an important basis for increasing greater participation in global production. 18 More and more captive spin-offs like that of British Airways-WNS, SwissAir-TCS, Conseco-EXL and GECIS-Genpact are expected to take place in the Indian scenario as the absorptive capacity in India rises.

29

References Acemoglu, D., P. Antràs and E. Helpman, “Contracts and Technology Adoption," mimeo, Harvard University, (2006) Acemoglu D., P. Aghion and F. Zilibotti, “Distance to Frontier, Selection, and Economic Growth, Journal of the European Economic Association 4, (2006): 37-74. Antràs, P., “Firms, Contracts, and Trade Structure,” Quarterly Journal of Economics, (2003): 1375-1418 Antràs, P., “Incomplete Contracts and the Product Cycle,” American Economic Review, (2005): 1077-1091 Antràs, P., and E. Helpman, “Global Sourcing,” Journal of Political Economy, (2004): 552-580 Aron, R and J. Singh, “IT Enabled Strategic Outsourcing: Knowledge Intensive Firms, Information Work and the Extended Organizational Form” Knowledge@Wharton, October 08, (2002) Arora, A., V.S. Arunachalam, J. Asundi and R. Fernandes, “The Globalization of Software: The Case of the Indian Software Industry”, A report submitted to Sloan Foundation, (2000) Baranson, J. “Technology transfer through the international firm”, American Economic Review, Papers and Proceedings, (1970): 435-440. Bartel A.P, S. Lach, and N. Sicherman, “Outsourcing and Technological Change”, NBER working Paper No. 11158, (2005) Borga, M. and W. Zeile, “International Fragmentation of production and the intra-firm trade of U.S. Multinational Companies.” Bureau of Economic Analysis Working Paper 2004-02, (2004) Chudnovsky, D, “North South Technology Transfer Revisited: research Issues for the 1990’s”, Documento de Trabajo nº 2 . CENIT, Buenos Aires, (1991) Davidson, W. H. and D. G. McFetridge, “Key Characteristics in the Choice of International Technology Transfer Mode.” Journal of Industrial Economics, 16(2), (1985): 5–21. De, R., “Captives: The Day of the Captive?”, DqIndia, August 28, 2006. Dunning, J. H., Multinational Enterprises and the Global Economy, Cambridge, UK: Addison Wesley Longman, Inc., (1993) Egan, M.L. and A. Mody, “Buyer-Seller Links in Export Development,” World Development, (1992): 321-334. Eicher, T. S. and P. Kalaitzidakis, “The Human Capital Dimension to Direct Foreign Investment”, in Wong, K. Y. and B. S. Jensen (eds.) Dynamics Trade and Growth, Ann Arbor, MI, Michigan University Press, (1997) E-valueserve, “KPO –Knowledge Process Outsourcing” April 2005a Gereffi, G., J. Humphrey and T. Sturgeon, “The Governance of Global Value Chains”, Review of International Political Economy, 12(1), (2005): 78-104 Glass, A. and K. Saggi, “Innovation and wage effects of international outsourcing”, European Economic Review, 45(1), (2001):67-86 Grossman, S. and O. Hart, “The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration,” Journal of Political Economy, (1986): 691-719. Grossman, Gene M. and Esteban Rossi-Hansberg, “Trading Tasks: A Simple Theory of Offshoring,” unpublished paper, Princeton University, (2006). Hart, O. and J. Moore, “Property Rights and the Nature of the Firm,” Journal of Political Economy, (1990): 1119-1158. Pack, H. and K. Saggi, “Inflows of Foreign Technology and Indigenous Technological Development”, Review of Development Economics, (1997): 81-98, Mattoo, A., M. Olarreagaz and K. Saggi, “Mode of Foreign Entry, Technology Transfer, and Foreign Direct Investment Policy”, Policy Research Working Paper 2737. World Bank, Washington, D.C., (2001) Melitz, Marc J., “The Impact of Trade on Intra-industry Reallocations and Aggregate Industry Productivity.” Econometrica 71 (2003):1695–1725. Teece D. “Technology transfer by multinational firms: the resource cost of transferring technological knowhow”, Economic Journal, (1977): 242-61 The Hindu Business Line, “Attrition costs weigh down BPOs: Study”, Dec. 13, 2005 UNCTC, “Transnational Corporations and Technology Transfer: Effects and Policy Issues,” NewYork, (1987) 30