Contract flexibility and negotiator incentives
Tridas Mukhopadhyay
Eric Walden
Tepper School of Business
Carnegie Mellon University
Abstract: IT outsourcing contracts are plagued by inflexibility. We develop an economic model to explain how the contract negotiator’s incentives influence contract flexibility. We show that neither wage nor one time commission type pay give any preference for flexibility. However, a promotion incentive, where the bonus is an ongoing bonus, does give the negotiator a preference for less flexible contracts. The preference for less flexible contracts increases as the discount rate increases. The preference for less flexible contracts increases as the number of competitors for the promotion increases.
Key words: Tournament theory, outsourcing, contracting, incentives
Introduction
Information Technology (IT) contracts are plagued by inflexibility that leads outsourcing deals to fail at an alarming rate. Some estimates of failure rates are as high as 70% (Overby, 2007) or 80% (Gartner, 2005), but in general, the accepted wisdom is that outsourcing projects have a 50/50 chance of success (Robinson and Iannone, 2007). Lack of flexibility is the main reason for renegotiating contracts (Gartner, 2005). With nearly half of a trillion dollars’ worth of IT services being outsourced every year (Mulligan et al., 2008), it seems incumbent upon us to investigate this issue. In this paper, we develop an economic theory of why IT contracts might be inherently plagued by inflexibility. Specifically, we identify the conditions under which a negotiator will prefer a less flexible contract. When a negotiator might receive a promotion (as opposed to a onetime bonus) based on the outcome of the contract, he prefers an inflexible contract. This preference increases as the discount rate increases, and as the number of competitors for the contract increases.
Literature Review
Flexibility was first formally studied as an economic problem by Stigler, who proposed that one plant was more flexible than another if the slope of the marginal cost curve was smaller (Stigler, 1939). In general, a firm may build a plant that produces at a very low cost for some level of output, but in which changing output is costly. Or a firm may build a plant that produces at a medium cost over a large range of production. Flexibility is a measure of the cost of change. A plant with a steeper marginal cost curve has a greater cost of change. Note that flexibility is a concern of the shape of the marginal cost curve rather than the absolute location.
The notion of flexibility has been well used by economists (Jones and Ostroy, 1984, Marshak, 1962, Mills and Schumann, 1985) and operations researchers (Bish et al., 2005, Graves and Tomlin, 2003). However, the focus tends to be on the physically constrained flexibility of a plant or machine.
Contracting is an important consideration in outsourcing, and as such, has received a good deal of consideration. Early research examined how uncertainty about cost and technology gave rise to opportunistic behavior on behalf of the client and vendor, and what sorts of incentives could be created to ameliorate uncertainty based problems (Whang, 1992). This perspective was used to propose that companies can benefit from contracting for pilot projects before committing to full projects (Gopal et al., 2003, Snir and Hitt, 2004).
In addition to viewing contracts as sets of incentives, researchers have viewed contracts as mechanisms to join two organizations in a semi-hierarchical arrangement (Ang and Beath, 1993). This view of contracts proposes that organizations cannot merge with one another fully, but write contracts to simulate a merger in certain areas.
Another perspective on contracts is the incomplete contract perspective, in which there is no uncertainty but both client and vendor are only willing to make investments that directly impact one or the other of them. This perspective takes ownership of property as the solution to contracting problems (Bakos and Brynjolfsson, 1993). This perspective has been updated to include ownership of intellectual property, in which the use of property by one entity does not put limits on the use of the property by another entity (Walden, 2005). This perspective also proposes that companies can do better contracting for both stages of a two stage contract together rather than separately (Richmond and Seidmann, 1993).
Recently, there has been a great deal of interest in different types of contract terms. One important question is the difference between a fixed price contract and a time and materials contract (Gefen et al., 2008). A fixed price contract allows the vendor to choose the nature of the work and provides high power incentives for efficiency. A time and materials contract allows the client to choose the nature of the work, but provides low power incentives to the vendor. Thus, in general, a fixed price contract is preferred (Gopal and Sivaramakrishnan, 2008).
A reoccurring theme in all this literature is the consideration of client and vendor as the only entities of interest. We change that by considering that a contract is actually negotiated by an agent of the client and an agent of the vendor and these agents have their own incentives. The client and vendor are bound to the contracts, and certainly gain or lose based on them, but the negotiating agents are not bound by the contracts; an employee of a firm is a distinct legal entity from the firm itself. To the best of our knowledge, the incentives of the negotiators have not been considered.
The study of incentives to employees in economics is covered by principal-agent theory. This branch of economics studies how one agent can incent another agent to act on his behalf. It is assumed that the agent acts to maximize his own welfare instead of that of the principle. For our purposes, that means that the contract negotiator strives to do what is best for the contract negotiator, rather than what is best for the agent. Presumably, the contract negotiator faces a set of incentives that aligns her behavior with the principle.
Incentives in principle agent problems generally fall into three categories. The first is salary, which is a fixed payment, and generally has very little marginal effect. The second is commission, which is a variable compensation that depends on the outcome of the activities. In this particular case, that would be a bonus based on the outcome of the contract. The third type of incentive is a promotion.
Promotions have been studied by a sub-discipline of principle agent economics known as tournament theory (Lazear and Rosen, 1981). In this theory, a promotion is considered to be the prize for a tournament. There are limited prizes and winning one depends not on absolute performance, but on relative performance. In general, the prize is awarded only to the person with the largest observed performance. Tournament theory helps explain why there are large discrepancies in pay between job categories, but little within a category. The idea is that everybody within a certain category is competing to get to the next category.
Economists noted that CEO compensation seemed way out of line with what could be considered reasonable, and proposed tournament theory as an explanation (Lazear and Rosen, 1981) for this. Researchers proposed that CEO compensation was actually a motivation for vice-presidents. The model suggested that if CEOs were highly compensated, then vice presidents would work very hard to become CEOs. However, there is only one CEO and many vice presidents, so the award of the CEO job is based on a tournament: The vice president that performs the best gets the job, and indeed data supports this (Eriksson, 1999). Similar logic applies at all levels of the organization, with managers striving to become vice presidents, production workers striving to become managers, and mail clerks striving to become production workers. Each level is a step up in power and money, but within any given level, the power and money are essentially equal even though simple observation shows a huge variation in performance within any given job.
The amount of reward that a promotion offers can be considerable. Research on law firms has found that as approximately 30% of associate salaries can be accounted for by the possibility of promotion (Ferrall, 1996). For most career paths, increases in earnings over time are the result of promotions rather than commission or salary. Salary usually changes at a fixed rate, close to inflation over time. Commissions, in principle could vary greatly over time. However in practice, most corporate career paths do not see large positive changes in commission earnings year after year. Still, when promotions are awarded, they do bring permanent and substantial increases to income.
The interesting difference between commission and promotions as a reward is that commission rewards performance, but promotion only rewards extreme performance. Promotions go to those who do the best work, not just those who do good work. The person who does the second best and the person who does the worst both get the same promotion. This becomes problematic if there is unobservable uncertainty (i.e. luck) involved. To the degree that promotion is important, individuals choose to take greater risks, because expected performance is not as important at the possibility of extreme performance.
Principle-agent models are usually on time models. Commission and/or promotions are awarded based on the outcome of a single event. However, flexibility is inherently a theory of changes over time. Flexibility is not concerned with the costs of production in any one period, but rather with the cost of changing levels of production over periods. To integrate these two economic perspectives, it is necessary to let a principle-agent model persist over time. For salary and commissions, this is a straightforward change, but for promotions, we have to recognize the permanence of promotions. Thus, we modify the standard tournament theory approach to be more of a promotion approach.
Tournament theory is generally concerned with rewards for relative performance, and looks at promotions as one of those rewards. However, when a tournament model is extended to multiple time periods and the reward is a promotion, one has to consider the duration of the reward. One could propose memory-less tournaments, where a reward was given in one period and had no effect on the next period. In this case, the question would be purely one of relative performance over time. If the reward was the employee of the month parking space then a one period reward makes sense. However, if the actual reward is a promotion, then one period rewards are nonsensical. Promotions tend to be permanent. Even if an agent later changes employers, the promotion generally provides value. It is easier to move from CEO of company 1 to CEO of company 2, than from COO of company 1 to CEO of company 2.
There is no single difference between negotiator incentives in IT outsourcing contracts and other types of contracts. However, there are differences in degree in many important areas. While the analysis here is geared toward understanding why hundreds of billions of dollars worth of IT contracts fail every year, the results may be useful for other industries that face similar environmental structures. We present several stylized facts about the IT contracting environment. This is not an exhaustive list, but rather a list that is relevant to the theory at hand.
The first stylized fact is that the demand for IT services changes a great deal over time. This is an artifact of the underlying development of information processing technology, and the massive digitalization of information that has been taking place over the last few decades. For example, in 1998 Google indexed 28 million pages, while in 2000 it indexed one billion pages, and in 2008 it indexed one trillion pages (Alpert and Hajaj, 2008), which is more than a 100% growth rate per year. More traditional firms are also growing their databases at a similar rate. For example, Wal-Mart had 583 terabytes of data in 2006 (Babcock, 2006) and in 2008 it had 2.5 petabytes of data (Lai, 2008), which is, once more, a greater than 100% annual growth rate. Similarly, computing capacity has demonstrated a 50% growth rate for some time. The point is that the amount of analysis and the nature of analysis of information changes dramatically over time, which gives rise to large changes in the demand for IT services. This is important because in the absence of changes in demand for IT services, flexibility is not an issue. It does not matter how flexible a contract is, if the demand for the thing supplied never changes. However, if large changes occur, flexibility becomes a large concern, and the changes in the demand for IT services over time are very large indeed.
The second stylized fact is that IT professionals have great opportunity for promotion. This is a byproduct of the rapid development of the field. Within the last fifty years, the need for IT professionals has gone from zero people worldwide to more than 3 million in the US alone; and the need for IT workers is projected to continue rising at double digit rates (Dohm and Shniper, 2007). Given this combination of size and growth, there are abundant opportunities for IT professionals to get promoted. Promotion may occur within a company or by moving from one company to another. Promotion may take the form of a more senior job title, the same title with a higher paying (usually larger) firm, or a job which is more applicable to the skills and desires of the employee.
Given that there is enormous growth in the IT profession, it follows that outcome based promotions will be an important factor in IT worker’s lifelong earnings. Of course, this is a matter of degree. Promotions are important everywhere, and performance matters. It just so happens that IT professionals have enormous opportunities for promotions and their individual performance matters a great deal.
Thus, the two fundamental elements that we incorporate into the model are the idea that IT demand growth changes over time and promotions are an integral part of an IT professional’s utility function. The first element means that flexibility is a major consideration in contracting. The second element means that our model needs to consider a utility function which offers some probabilistic reward based on performance.
Formulation of the model
Assume two firms: A client and a vendor enter into an outsourcing relationship. Both are risk-neutral. The client has an agent, the contract negotiator, who derives income from three sources: a fixed wage w, a bonus based on the outcome of the contract b, and a promotion reward based on the outcome of the contract x. The contract runs for two periods and everyone has the same discount value δ. Thus, the negotiator’s utility function is:
.
Notice that a promotion is a permanent change in income levels, while a bonus is a onetime award based on performance in a given period.
A contract specifies the prices in periods 1 and 2 and the quantities in periods 1 and 2 {p1,q1; p2,q2}. For now, assume q1 and q2 are known with certainty and q2 = q1 + k. Notice, that uncertainty in the outcome is not the issue in this model, so fixing q1 and q2 has no loss of generality. The uncertainty we are concerned with in this model is the uncertainty in the negotiator winning the tournament based on the contract’s outcome.
The client has a cost function cc(q) and the vendor has a cost function cv(q). Assume that cv(q)< cc(q) for all q so that trade makes sense. The client pays the agent based on the average savings of the contract (cc(q1) – p1) in period 1 and (cc(q2) – p2) in period 2. Two other metrics that could be used are the total savings (cc(q1) – p1) q1 or the percentage savings (cc(q1) – p1)/ cc(q1). For operator reduction purposes, define si = (cc(qi) – pi) for i in {1,2}.
The promotion reward x depends on how well the agent does relative to other agents. For the sake of simplicity, assume there is only one promotion available in any given period, and that the promotion is given to the agent who has the best performance in that period. Assume that the extreme value distribution of the highest performing other agent in a period is a random variable ai, which is measured in units consistent with the measure of contract outcomes so that they are comparable. This means that the probability of receiving xi is prob[si >ai]=f(si). Assume f’(si) > 0 so that the probability of receiving the promotion is an increasing function of the savings. Thus, the agent’s utility function is:
(1)
Allow the client to pay a linear bonus for contract savings so that bi = βsi. If the agent is risk neutral, the utility function can be rewritten as:
.(2)
Contract Preference
In this section we derive the conditions under which a negotiating agent prefers one contract to another.
Assume the vendor offers two contract options, each with the same average margin to the vendor. Contract A specifies sA1 and sA2, while contract B specifies sB1 and sB2, such that sA1 > sB1. We will now show two things. First, sA1 > sB1 => contract A is more flexible than contract B. Second, the agent prefers B to A.
Flexibility is the inverse of the cost of change (Marshak, 1962). A contract that offers a higher cost of change from quantity 1 to quantity 2 is less flexible. Flexibility can formally be defined in the case of only two known quantities as 1/[(s1 – s2)/(q1 – q2)]. Thus, contract A is more flexible than contract B when (sB1 – sB2) > (sA1 – sA2).
Given that contract A and B have the same value to the vendor, we have (cv(q1) – pA1)+ δ(cv(q2) – pA2) = (cv(q1) – pB1)+ δ (cv(q2) – pB2). By assumption, sA1 > sB1 which implies pA1 < pB1. To balance the equation then, pA2 > pB2 which implies sA2 < sB2.