Information Technology Outsourcing: Asset Transfer and the Role of Contract

Transcription

1 Information Technology Outsourcing: Asset Transfer and the Role of Contract Young Bong Chang Business School Sungkyunkwan University 25-2, Sungkyunkwan-ro, Jongno-Gu, Seoul,Korea, Seoul, Republic of Korea Vijay Gurbaxani Center for Digital Transformation The Paul Merage School of Business University of California Irvine, CA Kiron Ravindran IE Business School Calle María de Molina, Madrid, Spain June 2015

2 Information Technology Outsourcing: Asset Transfer and the Role of Contract Abstract Information Technology Outsourcing (ITO) has become the predominant mode of acquiring information systems services, providing clear evidence that the economics of service delivery favor external service providers over in-house information systems departments. An interesting feature of many large ITO arrangements is that assets necessary for service delivery are transferred to the vendor. The argument in favor of such asset transfers, based in Property Rights Theory, is that they are necessary to incentivize vendors to continue to invest in the transaction-specific assets to improve service. On the other hand, Transaction Cost Economics predicts that transferring such assets increases bilateral dependence and will elevate the risk of post-contractual opportunistic behavior. The contracting challenge is to specify the terms of exchange to achieve the client s objectives for outsourcing while managing the transaction risks. Given the role of asset transfer in ITO engagements, we develop a theoretical framework to derive propositions on contract design in the presence of asset transfer. In particular, we recognize the complementary role of compensation mechanisms, specifically the pricing scheme and the use of performance incentives. We have compiled a unique dataset that provides an opportunity to examine sensitive information on contract structure. We test our propositions by comparing large ITO contracts that include asset transfer to those that do not. We find that asset transfer does significantly affect contract design, manifested in the inclusion of clauses that protect both clients and vendors. Outsourcing objectives are more likely to be met when contracts include compensation mechanisms that complement asset transfer. Keywords: IT Outsourcing, Asset Transfer, Contract Structure, Transaction Cost Economics, Property Rights

3 1 Information Technology Outsourcing: Asset Transfer and the Role of Contract 1. Introduction External provision has become the predominant mode of acquiring information technology services. The worldwide market for IT outsourcing (ITO) was estimated at $288 billion in 2013 and predicted to grow at over 5.4% annually through 2015 (Overby 2013). Over 60% of the respondents in a large global survey indicated that outsourcing was a standard practice within their company and reported high levels of satisfaction (Deloitte 2012). While these data indicate the dominance of ITO as the preferred delivery option, examples of contractual disputes and unraveled outsourcing relationships are also not uncommon (DiamondCluster 2006; Deloitte 2012). These observations suggest that companies see economic benefit in sourcing IT services from external providers, but that these relationships are also fraught with risk. The academic literature has recognized this tradeoff. Prior research has noted the advantages that specialized providers possess, generated by economies of scale and specialization. It has also recognized the significant transaction costs inherent in ITO arrangements, stemming from the deployment of relationship-specific assets and the considerable technological and business uncertainty in multi-year deals. While some analysts have argued against external service provision (Wholey et al. 2001) because of the difficulty in writing efficient contracts in these settings, the widespread adoption of IT outsourcing indicates that its economic advantages seem to outweigh the contractual concerns. Economic theory suggests that outsourcing contracts will seek to maximize the gains from the arrangement and economize on transaction costs. This paper studies how contracts can be used to manage the tradeoffs between the risks inherent in outsourcing arrangements with the available gains. Before proceeding further, we describe the scope of IT outsourcing arrangements that we are interested in given the many different interpretations in the literature. We study ITO arrangements that

4 2 are multi-year, annuity-based arrangements in which a firm provides services on a continuous basis for the duration of the contract. 1 In this paper, we define ITO as, A long-term contractual arrangement in which one or more service providers are assigned the responsibility of managing all or part of a client s information systems infrastructure and operations. This definition is consistent with academic usage (see Dibbern et al. 2004) and with that of industry analysts (International Data Corp. 2006; Gartner 2008). More importantly, our definition sets a specific context for the relationship, which helps identify the nature of the risk and the potential gains in the ITO relationship. Consider a typical situation where a client firm has historically sourced IT services internally and owns the assets underlying service delivery data center and networking equipment, personal computers and mobile devices, proprietary software, facilities, human capital and so on. The firm then decides to source some or all of these IT services from an external provider with the goal of improving performance on dimensions such as cost and quality. In many, but not all cases, the vendor firm, as part of the outsourcing arrangement, acquires assets that were used in service delivery - hardware, software, people and facilities - from the client and uses them to deliver the contracted services. While there can be other motivations for asset transfer, one particularly important reason is that achievement of service improvements requires upfront and ongoing investments in the production assets. This requirement for investment in the underlying production assets sets up the essential tension in this paper. The argument in favor of asset ownership by the vendor derives from property rights theory, which recognizes that firms will likely not make investments in assets they do not own. In this view, conferring the associated property rights to the vendor provides it with the incentive to continue to invest in the relationship-specific assets (RSA) enabling the provision of lower cost or higher quality services over the life of the contract (Grossman and Hart 1986). The argument against transferring asset ownership to the vendor, derived from transaction cost economics, is that the transfer of RSA intensifies bilateral 1 We exclude arrangements that are project or task based and defined in terms of specific deliverables such as bespoke software development. Also note that our dataset predates the emergence of cloud-based delivery.

5 3 dependence in the relationship and leads to increased transaction risk (Williamson 1975; Klein et al. 1978). Williamson (1985) observed that the existence of a durable, transaction-specific asset is often the source of large market transaction costs since special arrangements, typically long-term contracts, are necessary to prevent both parties from acting opportunistically after one firm makes an investment in a specific asset. The client s switching costs increase when the vendor owns the assets by making it more difficult to terminate the vendor and bring services back in-house or transfer their delivery to another vendor. This increases the likelihood for a vendor to engage in post-contractual opportunistic behavior. Simultaneously, the vendor s risk also increases as a result of its significant upfront and largely sunk investment in the acquisition of these RSA. Given their transaction-specific nature, which we discuss more fully later, the market value of these depreciating assets is lower than their value in their current use. This leads to an improvement in the client firm s bargaining position. It is important to note that in annuity-based outsourcing arrangements at the time, the delivery assets purchased by the vendor were used to provide service to the client from which they were purchased. That is, vendors did not typically commingle technology assets between their various clients. In situations where market exchange occurs in spite of high transaction costs, contracts are the primary mechanism by which these risks can be mitigated (Joskow 1985; Kern and Willcocks 2001). While all ITO arrangements are characterized by some level of transaction-specific investments and switching costs, the transfer of assets underlying service delivery has the potential to substantially increase the risk of these arrangements to both parties. Therefore, contracts with asset transfer should include additional protections in the form of clauses that safeguard the interests of both parties. This implies that there should be observable differences in contracts with and without asset transfer. Of course, the goal of an ITO arrangement is to achieve the client firm s objectives, which most often are related to decreasing cost, improving quality, and increasing the business impact of technology.

6 4 It is well recognized that asset ownership, performance incentives and job design 2 are complementary dimensions of a management system (Holmstrom and Milgrom 1991; 1994). Therefore, a well-designed management system that takes these complementarities into account should lead to better performance outcomes. In our case, if assets are transferred to the vendor, the set of outsourced tasks and performance incentives incorporated in the agreement should complement asset ownership and lead to better outcomes. The simultaneous goals of achieving the client s strategic objectives for outsourcing and mitigating the risks of asset transfer creates the central challenge that motivates our research question: How are contracts used to improve IT outsourcing outcomes both in terms of the achieving the goals for outsourcing and mitigating the risk of ex-post opportunism associated with the presence of asset transfer? We develop a conceptual framework based in property rights theory and transaction cost economics to study the use of various contractual elements to improve outcomes and to manage risk. Using this framework, we develop hypotheses about the use of specific contractual clauses to protect the interests of the client and the vendor in case of asset transfer, and whether contractual features lead to better outcomes. Utilizing our focused definition of IT Outsourcing, we implemented a survey to build a reasonably homogenous dataset of ITO engagements in which the source and nature of hold up risk is relatively similar across deals. We conduct a preliminary analysis on this data to verify that there are, in fact, observable differences between contracts with and without asset transfer. Next, we formally test the hypotheses by comparing contracts where assets are transferred (CA) to contracts where assets are not transferred (CNA). We establish that when assets are transferred, contract structure does in fact differ considerably relative to the case where the client continues to own the assets. Specifically, we find that these contracts are more extensive with several clauses dedicated to specifying cooperation from both client and vendor. These contracts are also of longer duration. Moreover, we show that asset transfer to the vendor is 2 Job design refers to the delegation of specific tasks and associated decisions as to how they will be conducted to the agent or vendor.

7 5 complemented by a specific set of contract features - flexible pricing mechanisms and incentives for achieving improved performance outcomes - that jointly lead to improved ITO performance. We add to the empirical literature on the role of contract by conducting an empirical examination of contract structure in ITO arrangements that represent a comprehensive, and often strategic, sourcing relationship. Our study makes several contributions to the understanding of contracts as a means of governing ITO relationships both by mitigating transaction risk and by helping assure better ITO performance outcomes. First, our results complement earlier research by providing a new understanding of how the transaction risks associated with asset transfer can be mitigated using contract. Prior studies have relied on indirectly inferring asset specificity from the nature of the outsourced tasks usually due to data limitations. We take a fundamentally different approach to specific assets. That is, we examine how contracts can be used to mitigate the transaction risks associated with the transfer of assets essential to service delivery from the client to the vendor. Second, our context is very different from, but complementary to prior research, in that we focus on mitigating the risks of opportunism associated with vendors ownership of durable transaction-specific assets in large multi-year IT outsourcing arrangements, which in the traditional Williamsonian (1985) sense is the contracting challenge. Importantly, asset transfer is both observable and contractible. A property of our unique dataset is that the level of risk is very high given the scale and scope of the transaction. Our primary dataset allows us to incorporate and account for factors that are generally unobservable in secondary data, yet can critically influence contract design, such as the strategic objectives for the arrangement. In contrast, most prior studies have used secondary datasets, requiring the grouping of different kinds of IT outsourcing arrangements with substantial differences in the set of outsourced activities and hence in their individual risk profiles. These prior studies, in the main, are focused on smaller deals with lower potential economic consequences. Finally, our study examines the role of contract in the achievement of the client s goals for ITO, which of course, is the primary purpose of external service provision. Specifically, we draw on

8 6 Grossman and Hart (1996) and Holmstrom and Milgrom (1991,1994) to examine the complementarity between vendor ownership of assets, performance incentives and job design. This paper is structured as follows. The next section presents our theoretical approach and a review of the relevant literature. Section 3 derives relevant hypotheses on the role of contract. In section 4, we present the data and the results of a preliminary analysis, which we conduct to verify that there are, in fact, observable differences between contracts with and without asset transfer. This is followed by the formal analysis and presentation of our results in section 5. In section 6, we discuss the results and conclude with the implications of our findings and the limitations of our study. 2. Theory and Related Literature Our research adopts the perspective of incomplete contracting. The central premise of Incomplete Contract Theory is that contracting for all future states of the world is either impossible due to bounded rationality (Williamson 1975) or inefficient due to the ex-ante cost of contract design (Segal 1999). Within this framework we examine two closely related streams of literature: Property Rights Theory (PRT) which deals with ex-ante incentive alignment (Grossman and Hart 1986; Hart and Moore 1990) and Transaction Cost Economics (TCE) (Williamson 1985) which deals with contracting and ex-post economizing of transaction costs. PRT proposes the transfer of residual rights 3 to the relation-specific assets to the vendor as an incentive for vendor investment. However, TCE cautions against the risks associated with the potential opportunism that may arise from transferring these residual rights to a third party. The two theories adopt complementary perspectives in that one view focuses on the benefits of asset transfer ex-ante while the other highlights its costs ex-post. Property Rights Theory PRT applies in situations where an agent requires an asset to deliver a service to the principal. It is assumed that the effort invested by agents is not verifiable even if it is observable. It can then be shown 3 Residual rights, in contrast to specific rights, are those that have not been explicitly specified in the contract.

9 7 that both ownership of the RSA either by the principal or jointly by both parties can lead to underinvestment in these assets by the agent. Agents might choose to under-invest in the asset anticipating that the principal might force the agent to renegotiate terms after the investment has been made. On the other hand, owning the residual rights gives the agent an incentive to invest. Thus, Grossman and Hart (1996) argue that allocation of ownership rights is not a function of coordination cost or contracting ink cost but rather an incentive mechanism to enhance efficient investments, ex-ante. The underlying assumption here is that aligning incentives ex-ante will prevent opportunistic behavior expost. Optimal arrangements for production must keep asset ownership, job design, and explicit incentives in balance. In situations where effort is not verifiable as is the case in our context, and if the principal owns the assets required for service delivery, optimal contracts tend to have restricted incentives for production, especially in multi-task settings. However, when asset ownership shifts to the vendor, and the vendor must allocate effort to multiple tasks, the optimal incentive contract will provide stronger incentives to engage in production to prevent the vendor from being too cautious (Holmstrom and Milgrom 1991). Hence, no discussion of services outsourcing is complete without simultaneous discussion of the appropriate asset ownership structure and the incentive mechanisms. Various streams of literature, ranging from sociology to law, have adopted the PRT approach to examine the allocation of ownership and decision rights. From our perspective, PRT offers an appropriate approach to examine asset transfer in IT outsourcing and suggests that, The relevant comparison is not between the non-integrated outcome and the complete contract outcome but instead between a contract that allocates residual rights to one party and a contract that allocates them to another (Grossman and Hart 1986).

10 8 Transaction Cost Economics TCE, developed largely by Williamson (1975; 1985), has focused on understanding the relationship between governance structure and the characteristics of exchange. It holds that firms would choose to adopt a market-based governance mechanism over a hierarchical governance mechanism if production cost advantages outweigh transaction costs. This viewpoint of economic governance that balances the cost of producing in-house versus that of acquiring the product or service from outside the firm is mainly driven by the extent to which market sourcing invites transaction costs. These transaction costs are, among other things, a result of the asset specificity of the underlying production assets. When durable transaction specific assets are involved in the production of the goods or service, it gives rise to the possibility of ex-post bargaining to appropriate the surpluses accrued from the investment in these assets. Such ex-post bargaining is a central source of the transaction costs. TCE devotes significant attention to the role of durable transaction-specific assets because they elevate ex-post transaction risk and thus pose an additional challenge in securing the production cost advantage that markets can offer. Specifically, investment in relationship-specific assets by either party creates mutual dependence. Combined with uncertainty in supply or demand, it becomes costly to account for future contingencies leading to a heightened risk of ex-post opportunism. It is to minimize this cost of ex-post opportunism that Williamson says, support institutions of contract do matter (Williamson 1985 p. 29). TCE therefore provides us with another lens with which to examine the role of ex-post mechanisms of governance in the contract. 3. Literature Review Given the difficulty of obtaining contractual data, there are only a few studies in the IS literature that have conducted an empirical analysis of contract structure in managing ITO relationships. These papers have generally focused on mitigating specific transaction risks. None of these papers examine the role of contract in helping achieve the performance objectives of the ITO arrangement. In this section, we

11 9 review the few complementary papers that examine how an outsourcing contract is designed to address transaction risks. While empirical analysis of the effect of asset specificity on contract features is generally scarce, there are a few influential papers have done so in focused industry settings. Before we review the IS literature, we briefly highlight the research on managing supplier relationships by Klein et al. (1978) and Joskow (1985, 1987). Klein et al. (1978) examine the potential for appropriating the quasi-rents associated with RSA in various industries. They argue that of the two possible solutions of vertical integration and contracting, the former is likely to be the dominant solution when the cost of contracting is high. In a series of seminal papers, Joskow (1985; 1987) presents the results of an exhaustive case study on the role of contract in minimizing transaction costs in the context of coal procurement by electric power plants. In particular, he notes that even though transaction costs are seemingly high, market exchange is often the observed solution. In light of this observation, he posits that contracts are used to minimize transaction costs and conducts a comprehensive analysis of contracts to understand how contractual features are related to transaction risks. He examines four kinds of sourcing arrangements: from the spot market, from preferred suppliers, from strategic partners, and joint ventures. He shows how differences in asset specificity lead to differences in contractual features. For example, he finds that when the electric utility company and a coal mine jointly invest in sizeable RSA like railroads to carry the coal, the contracts are of longer duration, includes specific contractual guarantees including delivery commitments and dispute resolution, specifies expectations from the vendor such as product quality, and provides flexibility in pricing even though the pricing structure is specified at the outset of the relationship. He concludes that these measures protect both parties from behaving opportunistically and that contract is a crucial mechanism to mitigate risk and reduce transaction costs. Gurbaxani (2007) is among the first studies to demonstrate the important role of contract clauses in mitigating transaction risk for both clients and vendors in ITO arrangements. This qualitative study examines contractual terms used in ten large arrangements in which assets were transferred to the vendor.

12 10 It examines the frequency with which clauses that protect the interests of the client and the vendor are present in the contracts. These clauses include specified purchase and supply obligations, service level agreements, scheduled renegotiation, termination, arbitration, force majeure and competitive restrictions. It finds that contractual clauses are indeed matched to transaction risks. It also observes that total outsourcing contracts were longer than data center contracts resulting from the inclusion of application services, which exhibit high human asset specificity. Andersen and Dekker (2005) studied a large set of purchasing contracts for IT products and associated services with an average contract value of $1,500. Although the context of their study is very different from traditional definitions of ITO, we include it here because they provide a valuable contribution in their measure of the extent to which a contract addresses contingencies. They refer to this as contract extensiveness and operationalize it as a count of the number of contract clauses. Barthélemy and Quélin (2006) study 82 outsourcing transactions including IT and Business Process services and observe that contracts are more complex when outsourced activities have high switching costs or are central to the firm. They define complexity as a weighted average of the ranks that they assign to the mechanism specified in the clause for a set of clauses. They also incorporate measures for how human assets within the client firm adapt to the vendor s business requirements and a binary variable for asset transfer but find no association between these measures and contract complexity. Note that this measure of contract complexity is very different from the measure of contract extensiveness, since the first imposes an ordering on a set of clauses while the latter is a count of the number of clauses. Two more recent and particularly noteworthy studies, Susarla et al. (2010) and Chen and Bharadwaj (2009), analyze datasets of material contracts for IT services publicly disclosed to the U.S. Securities and Exchange Commission. Their datasets include over 100 contracts with average contract values of $13 million and $10 million respectively and mean duration of 40 and 38 months respectively. IDC (2006), using a definition of ITO consistent with ours, reports that the average size of the top 100 IT outsourcing deals in 2005 was $700 million, which suggests that the data in these studies are drawn from a sample

13 11 that reflects a very different class of outsourcing arrangements such as custom software development projects and from smaller client firms. Susarla et al. (2010) study three features of a contract extensiveness, duration, and extension clauses and examine their association with task complexity and scope. Their central premise is that complexity and scope can raise the need for relationship-specific investments, and in turn, the risk of hold-up. They find that complex tasks tend to be governed by extensive contracts but not necessarily long-term contracts. They argue that this is likely explained by the fact that a long-term contract for a complex task may create the risk of inefficient bargaining with the vendor in the future. Provision for contract extension is negatively associated with vendor ownership of residual assets, defined as assets that are the outcome of service delivery. That is, when vendors own the results of service delivery, the existence of a provision for contract extension is less likely. It is worth noting that these residual assets, as contracted deliverables, are significantly different from the ones we study, which are the production assets necessary for service delivery and were once owned by the client. In our context, in which vendors operate IS functions previously run in-house, IT deliverables are rarely, if ever, owned by a vendor. Chen and Bharadwaj (2009) link elements of risk in IT outsourcing to contract design. Based on transaction cost economics, agency theory and relational exchange theory, they categorize contract clauses into three classes: monitoring, property rights protection and contingency planning. They posit that task characteristics -- asset specificity, process interdependence, complexity and uncertainty -- can influence contract extensiveness, as defined by the inclusion of clauses in the three classes. Asset specificity is coded as a binary variable depending on whether service delivery involves the use of customized technologies specific to the client firm. They find that while asset specificity is associated with property rights safeguards and dispute resolution mechanisms (in fixed price contracts), it has no significant association with the overall measure of extensiveness. Summarizing this literature, we conclude that theory suggests that vendors can be incentivized to make ongoing investments in relation-specific assets (RSA) essential to service delivery if the residual

14 12 rights to these assets are transferred to them. However, ex-ante incentives provided by transferring these residual rights will not eliminate ex-post opportunism and likely increases it. Contracts do serve to manage such ex-post opportunism through their use of specific clauses and features that mitigate anticipated forms of transaction risk. While we have made considerable progress in understanding how contracts are designed to mitigate the risks associated with asset specificity, there is much that is not well understood. For instance, there appears to be no consensus on the relationship between contract extensiveness and asset specificity in part due to the different interpretations of asset specificity. Given the difficulty in measuring asset specificity, the studies above have inferred the degree of asset specificity by examining some of its drivers, though each has focused on a different set of assets. In almost all cases, the asset of interest is a deliverable rather than a production asset. There are no statistical studies of the role of contract in ITO when the assets of interest are the primary production assets. And, as we stated above, there are no studies of how contract design helps achieve the goals for outsourcing. 4. Hypothesis Development Fundamentally, IT outsourcing arrangements are subject to a range of business and technological uncertainties making it impossible for contracts to be fully specified. As a result, contracts focus on a limited set of parameters which are likely to be the most relevant or most verifiable (Salanié 1997). By design, ITO contracts specify significant decision rights and payment mechanisms for the service provider, which determine the quantity and quality of effort expended on service delivery. While the vendor is expected to meet specified performance requirements, a failure to do so willfully or inadvertently can impose considerable costs associated with business disruption on the client. Such standard problems of agency are compounded when asset ownership rights are transferred along with decision rights. Now, the client no longer owns the assets for service delivery making it more difficult for the client to bring service delivery back in-house and further raises the risk of hold-up by the vendor. Moreover, the vendor is now charged with choosing the level of ongoing investment in the assets, which may result in under-investment and decaying assets. From the vendor perspective, its exposure to holdup

15 13 risk also increases once it has made a large transaction-specific investment. Therefore, contracts governing such an arrangement should include measures to mitigate the risks to both parties. Even though contracts that govern complex ITO arrangements can be lengthy and specify decision and ownership rights, the measurement system, pricing provisions, incentive clauses, and numerous other provisions, they are still incomplete. Yet, just as in supply relationships in other industries, contracts have been argued to be...the only certain way to ensure that expectations are realized in IT Outsourcing (Kern and Willcocks 2001). The contract provides a detailed blueprint for the outsourcing relationship, and is critical because it specifies the precise terms of engagement. Tight or better-specified contracts are generally considered to lead to reduced opportunistic behavior (Lacity and Hirschheim 1993; Lacity and Willcocks 1998) and have been shown to predict IT outsourcing success (Saunders et al. 1997). Kern and Willcocks (2001) distinguish between the management control dimensions of a contract and its legal nature. They identify key elements of a post-contract management agenda: service description and exchanges, service enforcement and monitoring, financial exchanges, financial control and monitoring, key vendor personnel, dispute resolution, and change control and management. In this paper, we build on this conceptual framing to focus on 18 clauses that capture the management control aspects highlighted by Kern and Willcocks (see Table 3 for the set of clauses). Of these 18 clauses, we expect that some clauses will be present in all ITO contracts, independent of asset transfer. Given uncertainty in future business and technology trends, all contracts should specify which party has the right to make specified decisions, termed the formal specification of rights and responsibilities. Moreover, for the same reason, these contracts should also allow for scheduled renegotiation, which specifies the dates at which the two parties may seek to revise the terms of exchange. Service level agreements and the associated performance measurement clauses are likely to be present in most, if not all, contracts. Dispute escalation and dispute resolution clauses should also be specified in most contracts. Similarly, the case for a clause that specifies the terms and conditions that

16 14 govern termination for cause is self-evident. Contracts will also specify the terms and conditions for termination for convenience since it may be difficult to prove the non-performance of a vendor given the complexity of service provision, the intangible nature of IT services output, and the interaction between client and vendor effort. Of course, both clients and vendors would want a force majeure clause, which frees both parties from liability when an extraordinary event, such as war or a large earthquake occurs. Now consider our context of contracts with asset transfer wherein the assets are central to production and service delivery. While production assets, such as data center hardware and software, may seem to be general purpose, the dedication and customization of these assets to clients idiosyncratic needs makes them relationship-specific. That is, these assets cannot be redeployed or sold in the secondary market without considerable loss in value. For example, the deployment of these assets may reflect the enterprise architecture and specific software needs of the client firm. Moreover, redeploying these assets at potential clients is challenging, since these clients will already have significant production capacity of their own and the associated costs of redeployment can be significant. Importantly, given hardware price declines and the rate of obsolescence, even assuming that a contract is terminated relatively quickly after signing, say in three or four years, the vendor s investment in durable hardware assets is largely sunk. In the case of human capital, these assets typically represent a bundle of transferable, e.g. enterprise system software expertise, and firm-specific skills, e.g. knowledge of a client firm s processes. Of course, these vary with the job type. Data center operators are far less relationshipspecific than software professionals who have knowledge of a firm s idiosyncratic software and processes. That is, firm-specific capabilities are likely to be undervalued outside the client firm. Overall, in cases of asset transfer, the vendor bears the risk of a large sunk investment, which can lead to a stronger bargaining position for the client. For example, a client may choose to renege on the contract, or negotiate for better pricing after contract signing, now that the vendor has made a large sunk investment. Conversely, the client exposes itself to potential opportunism arising from the loss of ownership of assets essential to service delivery and its ongoing operations. Given the high switching costs for a

17 15 client to move to another vendor, or to bring service delivery back in-house, the vendor s bargaining power has increased. In such cases, a vendor may provide services at a quality level lower than ideal in the knowledge that the client s switching costs make it difficult to terminate the vendor. In a more extreme case, a vendor s withholding of services in case of a dispute can impose substantial costs on the client firm. It is therefore likely that, while writing contracts, both clients and vendors will want the inclusion of additional clauses that mitigate the risk of the other party behaving opportunistically in the presence of asset transfer. Now, we systematically consider how asset transfer requires additional protections for both parties. Once a vendor makes a significant investment in durable transaction-specific assets, it will want a commitment that the client firm will buy at least a predetermined volume of services over the life of the contract, termed a specified purchase obligation, to ensure that it can be compensated for its investment. The client, on the other hand, will want a specified supply obligation, in which the vendor commits to supply a predetermined volume of specified services so as to ensure that it will receive the necessary level of services and that the vendor will not withhold any services. However, given business uncertainty, say about demand, both parties would prefer a flexibility clause that allows supply and demand to adjust to changes in economic conditions that specifies what deviations from initial commitments are allowable and at what cost. Moreover, the client, recognizing the uncertainty in future technology costs will want an assurance that it is receiving a guarantee of best price, sometimes called a most favored customer provision. Correspondingly, the vendor may require a preferred supplier clause, which requires the client to negotiate exclusively with the vendor in good faith for any new ITO business and further incentivizes the vendor to make incremental investments and not engage in opportunistic behavior. The client may also require a key people provision, which specifies in cases of human asset transfer that some number of named employees continue to be assigned to the client s account after transfer and not be assigned elsewhere. In the event of contract termination, the client may incorporate an employee return clause and retain the right to rehire certain employees. To the extent that arrangements with asset transfer are wider

18 16 in scope and perhaps more strategic in nature, the client firm may also restrict its vendor s relationships with its competitors using an exclusivity clause. To summarize, 5 of the 8 clauses protect the client, 2 protect the vendor, and 1 protects both parties. Contract Extensiveness Based on the above discussion, we test three hypotheses that govern the service enforcement and monitoring dimension of the management control function of the contract. Specifically, we hypothesize that contracts with asset transfer (CA) will include more explicit clauses than contracts with no asset transfer (CNA) to address the additional risks of opportunistic behavior. Further, we hypothesize that CA will include more clauses than CNA to safeguard the separate interests of clients and vendors. Hypothesis 1A: Asset transfer is positively associated with contracts that are more extensive in their clauses. Hypothesis 1B: Asset transfer is positively associated with the number of clauses that limit the risk of clients holding up the vendor. Hypothesis 1C: Asset transfer is positively associated with the number of clauses that limit the risk of vendors holding up the client. Contract Duration Renegotiation or haggling costs form a critical component of the ex-post cost of contracting (Williamson 1985). Long-term contracts can therefore lower transaction costs for both parties by reducing the need for repeated bargaining (Masten and Crocker 1985; Joskow 1987). Importantly, longterm contracts can create an environment of better sharing of the gains from trade. In an experimental setting, Brown et al. (2004) show that when third party verification is not possible, long-term contracts do in fact have more generous sharing of rent and higher effort levels of the agents. Vendors of ITO services naturally prefer long-term contracts as an assurance of a longer stream of revenue and expected profits. In the presence of asset transfer, longer contracts gain considerable

19 17 added importance. Since clients almost always demand immediate cost reductions, and efficiency gains are achieved over a period of time rather than instantly, the increased duration facilitates the recouping of the large sunk investment in the production assets and also leads to better sharing of gains. For clients, in addition to the reduced costs of repeated haggling, their objectives of outsourcing are more likely to be met when vendors invest in the assets. We therefore predict that ITO contracts will be of longer duration when assets are transferred. Hypothesis 2: Asset transfer in ITO is positively associated with longer contract duration. Performance Outcomes: Payment Mechanisms We now turn our attention to payment mechanisms: pricing structures and performance incentives, to examine how they influence performance outcomes in case of asset transfer. As already discussed, the allocation of decision rights to the residual claimant of the assets does itself provide an implicit control mechanism. Yet, this instrument may not be sufficient especially when a typically riskaverse vendor is expected to make further investments (Holmstrom and Milgrom 1994). In cases of asset transfer, the choice of ongoing investment decisions made by the vendor can result in improved performance. Therefore, a client will incentivize these investments by rewarding decisions to invest in a manner consistent with its goals and by allowing the vendor to capture a share of the higher net value (Jensen and Meckling 1992). Therefore, beyond asset ownership, the compensation mechanisms, which include both the pricing structure and rewards and penalties for exceeding or missing performance targets, are also incentive instruments. Pricing Structure The pricing structure in large ITO arrangements is either variable (such as cost-plus) or predetermined for the duration of the contract. Variable pricing allows adaptation to the actual cost and effort levels in each period. Pre-determined prices are either pegged to a contracted deliverable (e.g. manage the data center for $x million per period) or are specified as unit prices (e.g. per utilized MIP) and the actual

20 18 compensation depends on the volume of services delivered. In the former case, the vendor assumes the risk of variability in the volume of work and its costs, while in the latter, it only needs to estimate unit costs accurately. The choice of variable or pre-determined pricing depends largely on task complexity (Bajari and Tadelis, 2001) and the nature of uncertainty in the transaction. Variable pricing is preferred in the presence of ex-ante uncertainty in cost (Kalnins and Mayer, 2004), requirements, project size, or resources (Gopal et al. 2003), and vendor ability (Banerjee and Duflo 2000); and ex-post uncertainty in quality (Kalnins and Mayer 2004) and outcomes (Eisenhardt 1989). Variable pricing provides both parties with flexibility in dealing with future contingencies. The risk associated with unforeseen contingencies, as discussed, is amplified when assets are transferred. Clearly, the need for unpredictable future investments increases the uncertainty in the transaction. Besides safe-guarding flexibility, clients are likely to prefer variable pricing when vendors are expected to invest in RSA, since fixed-price contracts can induce underinvestment (Gopal and Sivaramakrishnan 2008). For the vendor, profits are higher from variable price contracts compared to fixed-price contracts (Ethiraj et al. 2005), while generating less overruns and paying a smaller share of the overruns created (Banerjee and Duflo 2000). For example, when a vendor that has made an investment in an asset, increasing the fixed component of its cost structure, is confronted with an unanticipated decrease in demand, it will be affected more negatively when the contract specifies a pre-determined pricing structure relative to a variable pricing structure that can take these costs into account. This will reduce its incentives to invest in the production assets. Correspondingly, for a client, an unanticipated increase in demand will not result in lower fees even though the vendor may enjoy greater economies of scale. Therefore, it is likely that when assets are transferred, both clients and vendors would prefer to include provisions for flexible pricing.

21 19 Taken together, the effect of uncertainty, higher profitability and investment incentives lead us to hypothesize that when assets are transferred it is likely that clients achieve greater gains with flexible pricing. Hypothesis 3: Performance from the client perspective in ITO arrangements with asset transfer is higher when the pricing structure reflected in the contract is flexible. Performance Incentives It is well recognized that the riskiness of vendor investment can be compensated for by incentive payments for the achievement of objectives (Fama and Jensen, 1983). Moreover, Holmstrom and Milgrom (1991; 1994) have argued that the strong incentives provided by asset ownership are likely to be complemented by strong performance based incentives, which in multi-task settings have the additional benefit of prioritizing effort where value to the client is greatest. For example, a vendor s set of feasible actions include the options to postpone investments or only invest in those activities that pose limited risk even when riskier investments may be desirable (Holmstrom and Milgrom 1991). Therefore, beyond the compensation for the costs associated with improving the assets offered through the pricing scheme, which encourages investment but does not direct it, the client must also provide incentives to the vendor to select the investments and engage in effort that will result in performance outcomes that the client values. This leads us to the following hypothesis. Hypothesis 4: Performance from the client perspective in ITO arrangements with asset transfer is higher when the contract provides explicit IT-related performance incentives. 5. Data We conducted a comprehensive survey of outsourcing firms in 2005, collecting detailed information on numerous aspects of the arrangement. 291 North American firms that had outsourced for at least one year were identified from a dataset of outsourcing arrangements undertaken during the period

22 Forty-four firms responded positively to our request to be surveyed with a response rate of 15% while, depending on model specifications, up to four observations were discarded due to missing variables required in our analyses. Given the sensitivity of the information sought, this is an excellent response rate. A highly experienced survey firm administered the survey telephonically. All respondents were senior IT executives (above the level of IT director) who were knowledgeable about the objectives and outcomes of the ITO contract. 68% of our respondents are firmlevel decision makers, a fifth are at the business unit level and less than a tenth of our respondents are at lower levels. In terms of firm revenues, employees, industries and services outsourced, our data is representative of the top 100 worldwide outsourcing contracts in the same time frame (International Data Corp. 2006). Manufacturing firms account for a third of our sample and 46% are from the service sector. Firm size ranges from 750 to 195,000 employees. Excluding financial services, the average revenue was $9 billion 4 (see table 1). Table 1 Summary statistics Revenue 5 Number of Services Contract Value Employees ($Million) per contract ($Million) Mean 9,087 35, Minimum Maximum 27, , , Duration (years) We captured details regarding the outsourcing arrangement including the title and reporting structure of the outsourcing decision maker, the vendor selection process, the firm s objectives for outsourcing, its satisfaction levels, and the extent to which its objectives were achieved. Our survey also recorded contract-specific details such as the duration, included services and value of the contract. For the largest services, the respondents were asked to identify the included contract clauses from a comprehensive list of 18 such clauses (see Table 3). While outsourcing contracts include a large number of clauses, we focused on those that pertain to client and vendor protection. To derive the set of relevant 4 By comparison, the Fortune 250th company s revenue in 2005 was $8.9 billion 5 This value of revenue excludes firms in the Finance, Insurance, and Real Estate industries that often report holding assets as revenue.

23 21 clauses, we began with the seminal book on ITO contracts by Halvey and Melby (1996), which presents a comprehensive listing of contractual clauses and explains their use. We then used the Kern and Willcocks (2001) framework to extract the clauses that are relevant to our research. A brief description of these clauses is presented in section 3. Preliminary analysis Before proceeding with the formal analysis of the effect of asset transfer on contract structure and ITO performance, we conduct a general comparison of contracts with and without asset transfer to examine whether asset transfer does in fact matter. First, we compare the average size of CA to that of CNA. Since contract size can be measured in various ways, we examine four alternate specifications of size: total contract value, annualized value, the number of services, and the number of clauses in the contract. The average size of CA is $1,066 million compared to $386 million for CNA (see table 2). Accounting for contract duration, the annualized values are $118 million and $61 million respectively. The mean contract duration of the arrangements in our sample is 6.2 years, with a mean of 8.3 for CA and 5.3 for CNA. The average number of services included is five for CA and three for CNA. On average, contracts with asset transfer contain 4.75 more clauses than contracts without asset transfer. We see that CA are larger than CNA on all measures. Table 2- Contract differences when assets are transferred CNA CA Sig. (32 firms) (12 firms) Value $386 million $1,066 million *** Annualized value $61 million $118 million * Contract Duration 5.33 years 8.33 years *** Objectives (average score out of 5) *** Average number of services per contract 3 5 *** Average number of clauses ** Company level initiation 47% of firms 75% of firms ** Competitive bidding 53% 75% * *** significantly different at.01 level ** significantly different at.05 level * significantly different at.1 level