Project Contract5

download Project Contract5

of 17

Transcript of Project Contract5

  • 8/4/2019 Project Contract5

    1/17

    Management Accounting Research 20 (2009) 129145

    Contents lists available at ScienceDirect

    Management Accounting Research

    j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / m a r

    Determinants of contract terms for professional services

    Carsten Homburg a,, Peter Stebel b

    a Department of Management Accounting, University of Cologne, Albertus-Magnus-Platz, D-50923 Cologne, Germanyb PricewaterhouseCoopers AG WPG, Friedrichstr. 14, D-70174 Stuttgart, Germany

    a r t i c l e i n f o

    Keywords:Cost contracts

    Customer integration

    Double moral hazard

    Incentives

    Professional services

    a b s t r a c t

    This study provides evidence on the determinants of contract terms between professionalservices firms and their clients. Because professional services are typically characterized by

    a high degree of transactional uncertainty and a double moral hazard risk, contracts can be

    essentialfor thecreationof incentives to control the behaviorof thoseinvolvedin theservice

    encounter. Based on both agency and organizational theory, hypotheses on the determi-

    nants of implementing a specific type of cost contract are developed and tested empirically

    with respect to management consulting firms. Our results indicate that (1) service char-

    acteristics exert a significant impact on the chosen contract type, (2) performance-based

    contracts may not be optimal, even if service output is measurable and verifiable, and (3)

    experience-based trust and reputation impact on the choice of controls used in short-term

    contracts. These results contribute to the field of management accounting by providing

    insights into the design of management control systems in service organizations.

    2008 Elsevier Ltd. All rights reserved.

    1. Introduction

    Because co-productionby service employees and clients

    is at the heart of many services (Larsson and Bowen, 1989;

    Solomon et al., 1985), clients themselves often exert a con-

    siderable impact on both the cost of the service process

    and its output (Glckler and Armbrster, 2003, p. 277).

    Despite this insight, there is little research on how service

    organizationsactually design management control systems

    to govern transactions at the organizationclient inter-

    face (Chenhall, 2003; Hopwood, 1996; Modell, 1996; Otley,

    1994; Shields, 1997).In this paper, we focus on professional services, in par-

    ticular management consulting services, and examine the

    determinantsof contract terms between management con-

    sulting firms and their clients. Contracts are formal mecha-

    nisms of management control (Anderson andDekker, 2005;

    Kirsch et al., 2002) and theanalysis of contract terms yields

    insights into how transactions between both contracting

    Corresponding author. Fax: +49 221 470 5012.E-mail address: [email protected] (C. Homburg).

    parties are governed. Contractual relationships between

    management consulting firms and their clients have sev-

    eralrelevant andimportantcharacteristics.First,consulting

    services are often complex and involve a high level of

    both transactional uncertainty and risk for the client, as

    he does not purchase a ready-made product (Glckler

    and Armbrster, 2003, p. 269; Mitchell, 1994). Second,

    because of the interdependent and interactive character of

    co-production between consultants and clients in service

    delivery, both contracting parties can behave opportunis-

    tically, which results in a double moral hazard risk that

    needs to be considered in the design of a contract. Third,these contractual relationships are generally rather short-

    term in nature. It has been established that, in long-term

    inter-organizational relationships such as international

    joint-ventures (Groot and Merchant, 2000), outsourc-

    ing relationships (Anderson et al., 2000; Langfield-Smith

    and Smith, 2003; Van der Meer-Kooistra and Vosselman,

    2000), strategic alliances (Dekker, 2004) and integrative

    buyersupplier arrangements (Frances and Garnsey, 1996),

    the social context, i.e., trust and reputation, in which these

    business exchanges are embedded, is highly relevant as

    a means of mitigating potential opportunistic behavior.

    1044-5005/$ see front matter 2008 Elsevier Ltd. All rights reserved.doi:10.1016/j.mar.2008.10.001

    http://www.sciencedirect.com/science/journal/10445005http://www.elsevier.com/locate/marmailto:[email protected]://dx.doi.org/10.1016/j.mar.2008.10.001http://dx.doi.org/10.1016/j.mar.2008.10.001mailto:[email protected]://www.elsevier.com/locate/marhttp://www.sciencedirect.com/science/journal/10445005
  • 8/4/2019 Project Contract5

    2/17

    130 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    However, little is known on whether this effect is also rele-

    vant to short-term inter-organizational relationships, such

    as those between service organizations and their clients.

    In addition to defining the agreed-upon service, con-

    tracts also specify the monetary terms of the contract, that

    is, how the management consulting firm is to be remu-

    nerated. Contracts can be classified into two broad types,

    according to whether the remuneration of the consulting

    firm and therefore the cost to the client receiving the ser-

    vice, depend on the realized service output or not. Under

    a fixed-cost contract, the remuneration of the manage-

    mentconsulting firm is invariant withregardto the realized

    service output, while under a variable-cost (performance-

    based) contract, it is contingent on the realized service

    output. The literature usually differentiates between two

    types of controls that can be used to direct the behavior

    of a contracting party: behavior-based and outcome-based

    controls (Eisenhardt, 1985; Ouchi, 1979). Following Eisen-

    hardt, we operationalize a control strategy based on the

    monetary terms in the contract, that is fixed-cost contracts

    are a form of behavior-based control, while variable-cost

    (performance-based) contracts constitute outcome-based

    control (Eisenhardt, 1985, p. 144).

    The purpose of this paper is to add to the limited body

    of knowledge on the design of management control sys-

    tems in service organizations. Based on both agency and

    organizational theory, hypotheses on the determinants of

    implementing a specific typeof cost contract are developed

    and tested empirically in the context of management con-

    sulting. This paper contributes to management accounting

    in several ways. First, we present empirical evidence with

    regard to the circumstances under which service compa-

    nies use behavior- and outcome-based controls to govern

    transactions at the organizationclient interface. Second,

    our results indicate that both the characteristics of the ser-

    vice and the characteristics of the contracting relationship

    exert a significant impact on the chosen contract type. This

    indicates that trust and reputation can be effective in the

    provision of incentives and the mitigation of moral hazard

    risks also in contractual relationships that are short-term

    in nature. Third, the paper provides evidence that in con-

    tracting situations characterized by a double moral hazard

    risk it might be, under specific conditions, optimal not to

    tie the remuneration of a service provider to actual per-

    formance. This adds to our knowledge of the cybernetic

    process of monitoring and rewarding performance in inter-

    organizational relationships.

    The remainder of the paper is organized as follows. In

    Section 2, the contracting problems specific to professional

    services are discussed with reference to management con-

    sulting. In Section 3, hypotheses on the determinants of

    implementing a specific type of cost contract in the pres-

    ence of double moral hazard are developed, and tested

    empirically in Section 4. Section 5 concludes the study with

    a discussion of the results.

    2. Contracting problems for professional services

    Professional services firms, such as management con-

    sultancies and their clients, need to take a number ofpotential problems into account when designing a con-

    tract. First, the client does not purchase a ready-made

    product and is thus not able to evaluate its characteris-

    tics before signing the contract (Glckler and Armbrster,

    2003, p. 276). Second, in order to produce a service, a (writ-

    ten or oral) agreement between the service provider and

    the client, which details the service to be performed, is

    necessary. However, in consulting projects, the details of

    the required service are often unclear at the beginning

    and many relevant details are thus not conclusively estab-

    lished when a project commences. Contracts are therefore

    frequently incomplete with respect to the service output

    as well as to the input of the contracting parties. Third,

    because of their complexity and intangibility, it is often

    difficult to observe the effort of a contracting party and

    to define verifiable performance measures for evaluating

    the service output. Verifiability not only requires that out-

    put be observed and evaluated by the contracting parties,

    but also by an independent third party. The intangibility of

    services, however, often renders it difficult to define objec-

    tive quality and quantity measures for services (Dornstein,

    1977, p. 119; Mitchell, 1994, p. 325). Fourth, professional

    services are typically characterized by a high level of cus-

    tomer involvement in the provision of the service (Larsson

    and Bowen, 1989; Solomon et al., 1985). In this paper, we

    use the term integrativity to refer to the level of customer

    involvement. Integrativity means that the agreed-upon ser-

    vice cannot be finalized without the integration of external

    factors. External factors are defined as production factors

    relating to the client that, temporarily and restricted to a

    specific service process, enter into the service providers

    domain, where they are combined with the latters inter-

    nalproduction factors in order to produce theservice (Flie

    and Kleinaltenkamp, 2004, p. 394).1 While simple services

    often require only a low level of interaction between the

    service provider and the client, professional services like

    thoseof management consultancies aretypically character-

    ized by a high level of interaction. In many cases, the client

    therefore has a considerable impact on both the cost of the

    service process and its output (Glckler and Armbrster,

    2003, p. 277).

    In the case of asymmetric information between the ser-

    vice provider and the client with regard to each others

    inputs, both contracting parties can behave opportunis-

    tically before and during their contractual relationship

    (Martin et al., 2001). Before signing the contract, the service

    provider can overstate his qualifications and experience in

    order to secure the contract, whereas the client can down-

    play the difficulty of the task or overstate his cooperation.

    During the service process, there is the risk that both con-

    tracting parties choose a lower than agreed upon level of

    effort. In order to minimize costs, the consulting firm can,

    for example, useresults from other projects without adapt-

    ing them sufficiently to the specific needs of the client.

    Similarly, a client can minimize costs by not assigning his

    best-qualified employees to the project. After the service is

    finalized, there is therisk that theservice provider will take

    advantage of any leeway in billing, for example, by charg-

    1

    External factors can take various forms: people (e.g., the customerhimself), objects, rights, money and information.

  • 8/4/2019 Project Contract5

    3/17

    C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145 131

    ing separately for alleged additional services. Likewise, the

    client might attempt to lower the bill by pretending to be

    dissatisfied with the projects results.

    A major objective of contracts is to (at least) restrict

    the potential for such opportunistic behavior by the man-

    agement consulting firm as well as by the client. In the

    following section, building on agency and organizational

    theory, we develop hypotheses with regard to the determi-

    nants of the choice of a particular type of cost contract as a

    means of mitigating double moral hazard risks.

    3. Research framework and hypotheses

    We consider a setting in which a management consult-

    ing firm is hired by a client. Due to the integrativity of the

    service, the efforts of both parties are assumed to be com-

    plementary and necessary for the productionof the service.

    Because we take the perspective of the service provider,

    it is he who offers the client a contract. This represents

    reality quite closely, as it is often the service provider who

    offersthe clienta contracton thelatters request. Based on a

    single-perioddouble moralhazardmodelby Bhattacharyya

    and Lafontaine (1995), we start by analyzing a contractual

    relationship that takes place only once.2 In this situation,

    because the misbehavior of a contracting party has no

    effect on the future, only the characteristics of the ser-

    vice have an impact on the chosen contract type. Then,

    building on the game-theoretic results of Fudenberg et al.

    (1990), we extend the analysis to a multi-period situation

    in which a consulting firm (long-run player) interacts with

    a sequence of clients (short-run players), each of whom

    plays only once, but observes all previous play. In this situ-

    ation, because the market is neither temporally limited nor

    anonymous, the social context (i.e., trust and reputation) in

    which the contractual relationship is embedded, becomes

    relevant.

    As a benchmark, we start by considering the first best

    solution that can be achieved if the behavior of each

    contracting party is verifiable, meaning that there is infor-

    mational symmetry. Verifiability requires that the behavior

    can be observed and evaluated not only by the contracting

    parties, but also by an independent third party, such as a

    court. In this case, the service provider and the client can

    conclude a forcing contractwith regard to their respective

    efforts. Incentives are not necessary. The service provider

    receives only a fixed-fee for his effort and the client retains

    the residual profit.3

    If, however, the efforts of the contracting parties are

    not verifiable (second best case), due, for example, to pro-

    hibitively high monitoring costs, a fixed-cost contract is no

    2 Double moral hazard models have been used to analyze incen-

    tive problems and contracts between franchisors and franchisees

    (Bhattacharyya and Lafontaine, 1995; Lal, 1990; Rubin, 1978), landowners

    and tenants (Eswaran and Kotwal, 1985; Reid, 1977), partners in corpo-

    rate ventures (Chi, 1996) and to determine the optimal level of product

    guarantees (Cooper and Ross, 1985; Dybvig and Lutz, 1993).3 While the utility of the service is not transferable and accrues only

    to the client, we assume that this utility can be represented in monetary

    terms. After paying the service provider, the client retains the residual

    profit deriving from the difference between the monetary utility of theservice, and the price paid to the service provider.

    longer optimal. While the client, who retains the residual

    profit, has an incentive to choose his first best effort, the

    service provider has an incentive to provide only his mini-

    mum effort. In order to prevent this, the service providers

    compensation must be made contingent on the realization

    of a verifiable performance measure that allows inferences

    with regard to his chosen level of effort. Bhattacharyya and

    Lafontaine show that, in the case of asymmetric informa-

    tion, (a) the service output must be shared between both

    contracting parties, and (b) the share of a contracting party

    increases relative to its contribution to the service output

    (Bhattacharyya and Lafontaine, 1995, p. 770). The client

    therefore needs to pay the service provider, in addition to

    a fixed-fee, a variable fee contingent on the integrativity

    level of service.4 Output sharing between the contracting

    parties, however, results in a trade-off between high levels

    of effort from the service provider and high levels of effort

    from the client. As a result, the combined effort in the sec-

    ond best case is strictly lower than in the first best case, i.e.,

    the agency faces an efficiency loss. This follows also from a

    general result from Holmstrm (1982, p. 326), who showed

    in his analysis of optimal team incentives, that a sharing

    rule that distributes the entire output among contracting

    parties (budget balancing constraint) cannot implement

    the first best result.

    The model has a number of implications, which can be

    formulated as empirically testable hypotheses. In a dou-

    ble moral hazard setting, sharing occurs as a result of both

    parties need for incentives. Implementing a variable-cost

    (performance-based) contract, however, results in an effi-

    ciency loss for the agency. The only means of avoiding

    this, given a certain level of integrativity, is to agree bind-

    ingly on the degree of effort of the service provider. This

    would, however, require its verifiability. If this were possi-

    ble, the client could conclude a forcing contract with the

    service provider, agree on the latters first best effort and

    compensate him with a fixed-fee. This leads to the hypoth-

    esis that the probability of using fixed-cost contracts (i.e.,

    behavior-based controls) increases, while the probability

    of using variable-cost (performance-based) contracts (i.e.,

    outcome-based controls) decreases, the greater the verifi-

    ability of service provider behavior:

    H1. As the verifiability of service provider behav-

    ior increases, the implementation of variable-cost

    (performance-based) contracts becomes less likely

    and the implementation of fixed-cost contracts more

    likely.

    In order for a client to be able to directly control the

    behavior of the service provider and implement the first

    best solution, it is firstly necessary that he understands

    what constitutes appropriate behavior and secondly, that

    he can identify deviations from this norm. In the organiza-

    4 In the model of Bhattacharyya and Lafontaine, it is assumed that both

    contracting parties are risk neutral (Bhattacharyya and Lafontaine, 1995,

    p. 761). Their results would not change qualitatively in the case of risk-

    averse contracting parties. The optimal share would, however, then also

    depend on the difference in risk aversion of both contracting parties and

    not only on the integrativity level of the service. The case of a risk-neutralprincipal and a risk-averse agent is considered in Kim and Wang (1998).

  • 8/4/2019 Project Contract5

    4/17

    132 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    tional theory literature, this is termed knowledge of the

    transformation process (Kirsch, 1996, p. 4; Ouchi, 1979, p.

    843). The greater the complexity of service, the more diffi-

    cult it becomes to define and verify appropriate behavior.

    The hypothesis is therefore:

    H2. As thecomplexity of service increases, the verifiability

    of service provider behavior decreases.

    The relative importance of client effort in the produc-

    tion of service output depends on the integrativity level of

    service. Comparative statics show that the efficiency loss

    for the agency increases when the required levels of effort

    of both contracting parties become more equal, i.e., the

    higher the integrativity level of service. The reason is that,

    in this case, the output needs to be shared equally, which

    results in weak incentives for both parties. This means that

    variable-cost (performance-based) contracts become less

    attractive, compared to fixed-cost contracts, the higher the

    integrativity level of service:

    H3. As the integrativity level of service increases, the

    implementation of variable-cost (performance-based) con-tracts becomes less likely and the implementation of

    fixed-cost contracts more likely.

    The intangibility of services, however, often makes it

    difficult to define objective and verifiable performance

    measures for evaluating the service output. The greater

    the degree of verifiability of service output, the more likely

    the implementation of variable-cost (performance-based)

    contracts:

    H4. As the verifiability of service output increases, the

    implementation of variable-cost (performance-based) con-

    tracts becomes more likely and the implementation of

    fixed-cost contracts less likely.

    The implementation of a variable-cost (performance-

    based) contract requires the verifiability of service output.

    This is, however, often problematic, especially for complex,

    professional services. The higher the complexity of service,

    the more difficult the verifiability of output:

    H5. As thecomplexity of service increases, the verifiability

    of service output decreases.

    The greaterthe level of knowledge of the transformation

    process, the easier it becomes to verify the behavior (i.e.,

    the level of effort) of the service provider and the client.

    In addition, this knowledge also makes it easier to definesuitable performance measures for the service output:

    H6. As the verifiability of service provider behavior

    increases, the verifiability of service output increases.

    H7. As the verifiability of client behavior increases, the

    verifiability of service output increases.

    If neither the output nor the efforts of the contracting

    parties are verifiable, it is no longer possible to imple-

    ment a service contract in a static, single-period model.

    In a contractual relationship that takes place only once

    and in which misbehavior has no effect on the future, it

    would be rational for the client (who retains the residualprofit from the contractual relationship) to claim that the

    service output is inadequate, in order to reduce his pay-

    ments to the service provider. Because the service provider

    would anticipate this, in the extreme, no services would be

    provided, leading to market failure in the sense of Akerlof

    (1970). In reality, however, markets and contractual rela-

    tionships are usually not temporally limited, nor are they

    anonymous. Particularly for professional services, a good

    reputation plays a major role and is often a prerequisite for

    arranging contracts between two parties (Kollock, 1994).

    Various investigations of market mechanisms for man-

    agement consulting show that reputation is one of the

    most important factors in the acquisition of new clients

    (Glcklerand Armbrster, 2003, p. 284). If a potential client

    has no personal experience with a particular management

    consulting firm, then recommendations from trustworthy

    persons within his network play a major role in reducing

    uncertainty with regard to the capabilities of the consul-

    tants.

    In order to analyze the role of trust and reputation in a

    short-term contractual relationship, we need to extend the

    analysis to a multi-period model in which a management

    consulting firm (long-run player) interacts with a sequence

    of clients (short-run players). Fudenberg et al. (1990) have

    shown that results similar to the folk theorem for repeated

    games with two long-run players5 can arise in a game in

    which a single long-run player faces a sequence of short-

    runplayers,eachofwhomplaysonlyonce,providedthat(1)

    the players in each period are aware of all previous play, (2)

    the game is infinitely repeated or alternatively concludes at

    an uncertain date, and (3) the discount rate of the long-run

    player is sufficiently close to one (Fudenberg et al., 1990, p.

    555).

    Because such a game will often have multiple equilibria,

    we focus on the equilibrium that induces an efficient out-

    come (Fudenberg and Levine, 1989, p. 759). The peculiarity

    of such a contractual relationship is that a short-run player

    in a particular period can make his decision contingent on

    thedecisions of thelong-run player in theprevious periods.

    In particular, it becomes possible for a short-run player to

    penalize the long-run player by not accepting a contract

    offer if the long-run player has behaved non-cooperatively

    in the past. In our context, this means that a manage-

    ment consulting firm can build a reputation for cooperative

    behaviorand for alwayschoosingthe firstbest effort. Taking

    this into account, a fixed-cost contract becomes incentive

    compatible and can implement the first best solution. The

    particularly interesting property of a fixed-cost contract is

    that a client, through retaining the residual profit, has the

    maximum possible incentive to choose his first best effort,

    whereas the service provider is disciplined by the potential

    utility that can be derived from having a good reputa-

    tion and therefore the possibility of securing future client

    relationships. With a given fixed-fee, the service provider

    deliberates between his utility for non-cooperation, which

    correspondsto theeconomized cost of effortfor oneperiod,

    5 In repeated games with two sufficiently patient long-run players,

    (almost) any equilibrium can be implemented, subject to the condition

    that each player achieves at least his minimax utility level ( Fudenbergand Maskin, 1986).

  • 8/4/2019 Project Contract5

    5/17

    C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145 133

    and his opportunity costs, which correspond to the antic-

    ipated lost utility of having contractual relationships with

    future clients. Cooperation is thus more likely, the higher

    theprobability of future client relationships and thesmaller

    the service providers discount factor for future revenues.

    A contract that meets these conditions is self-enforcing,

    sustained by the value of future client relationships. Ver-

    ifiability of effort or of service output by a third party is no

    longer required, as long as an existing client can observe

    the chosen effort of the service provider and a potential

    new client obtains this information either by word-of-

    mouth or from written sources (Fudenberg and Levine,

    1989, p. 761).

    Assuming that a management consulting firm wishes

    to remain in the market over the long-term, it thus has

    an incentive, even for one-off projects, to choose a high

    level of effort in order to have a reputation for cooperative

    behavior in the future. The strength of such an incentive

    depends on the transparency of the market. The easier it

    is for a potential client to learn about the previous behav-

    ior of the consulting firm (i.e., its reputation) the stronger

    the incentive for the consulting firm to cooperate. In addi-

    tion to incentives from reputation effects, clients can use

    another lever to stimulate effort. It was implicitly assumed

    above that the service was performed essentially in one

    step, and that a client pays the entire agreed-upon fixed-

    fee after the performance of the service. A characteristic

    of many professional services, however, is that they are

    performed in several steps (project phases). In contrast

    to services that require only a brief interaction period

    between the contracting parties and where the output is

    produced after a short time period, both contracting par-

    ties have the opportunity to learn about each other. The

    potential for opportunistic behavior is thus at least par-

    tially restricted. Interim results and additional information

    can be used to adjust available resources or to counter-

    act any discerned misbehavior. When the continuation

    of the project depends on the successful conclusion of a

    project phase or milestone, then, the mere prospect of

    being awarded the consequent project phases, can create

    the incentives described above.

    If the conditions for cooperation in a particular period

    are not met, that is, if the expected utility of future client

    relationships is not sufficiently high, it can be shown that

    the service provider will already choose not to cooperate

    in the first contracting period. Since a client would antici-

    pate this, a fixed-cost contract would no longer be feasible.

    In this case, the optimal cost contract in the multi-period

    model would be identical to that in the one-period model.

    Only a variable-cost (performance-based) contract would

    be feasible, provided that the service output is verifiable.

    In summary, the following hypotheses can be formu-

    lated:

    H8. As the level of expected utility of future client rela-

    tionships increases, the implementation of variable-cost

    (performance-based) contracts becomes less likely and the

    implementation of fixed-cost contracts more likely.

    Fig. 1. Hypothesis model.

  • 8/4/2019 Project Contract5

    6/17

    134 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    H9. As the reputation of a service provider improves,

    the implementation of variable-cost (performance-based)

    contracts becomes less likely and the implementation of

    fixed-cost contracts more likely.

    Fig. 1 provides an overview of all hypotheses.

    In the empirical study presented below, all the above

    hypotheses are tested. It is important to note that, as

    we assume equilibrium conditions here, we do not testwhether the contracts used by the respondents actually

    are efficient in a specific situation. This is assumed implic-

    itly, because of survival-of-the-fittest conditions (Chenhall,

    2003, p. 134).

    First of all, the data collection and composition of the

    sample are described below. Subsequently, all constructs

    that have been identified as determinants of implement-

    ing a specific type of cost contract are conceptualized and

    operationalized.

    4. Empirical study

    4.1. Data collection and data basis

    The data for this study was obtained from German

    management consulting firms by means of an online ques-

    tionnaire. Prior to the study, a preliminary version of the

    questionnaire was pre-tested with partners and managers

    from three management consultancies. The remarks and

    comments fromthe pretest were incorporatedintothe final

    version of the questionnaire. A total of 700 management

    consulting firms in Germany were contacted by e-mail,

    from which the largest 350 management consultancies

    were contacted by telephone prior to e-mailing the ques-

    tionnaireand asked to participate in the survey. The contact

    data were obtained frompubliclyavailable databases (Hop-

    penstedt Company Profiles and Registry of Management

    Consulting Firms from the German Association of Manage-

    ment Consultants BDU e.V.). In the e-mail, the background

    of the survey was described and a link provided to the

    online questionnaire. The survey was conducted anony-

    mously, and a company-specific access code ensured that

    no firm could participate in the survey more than once.

    A total of 81 management consultancies participated in

    the survey. After discarding incomplete questionnaires, 76

    usable questionnaires remained. This represents a total

    return rate of 11.6%, or 10.9% excluding incomplete ques-

    tionnaires.

    The low response rate raises the potential for non-

    response bias, as non-respondents might differ in impor-

    tant aspects from respondents (De Vaus, 1995). In order

    to detect potential problems with non-response bias, we

    divided the data set into thirds, according to the num-

    ber of days from the initial e-mail until receipt of the

    returned questionnaire. We conducted a comparison of

    the means of all variables for the first and last thirds

    of respondents. The results showed no significant differ-

    ences (p < .05) between responses, which indicates that

    non-response bias does not seem to be an issue in our

    study.

    Table 1 gives an overview of the returned question-naires, broken down by turnover and number of employees

    Table 1

    Surveyed management consultancies by turnover and number of

    employees.

    Turnover (in million D) Sampl e ( % of MCs) Return (% o f MCs)

    1.00 13.8 12.01.015.00 47.2 60.0

    5.0110.00 17.2 8.0

    10.0120.00 11.8 8.0

    20.01 10.0 12.0Total 100 100

    n 700 75a

    Employees (number) Sample (% of MCs) Return (% of MCs)

    15 7.6 8.0

    610 13.7 13.3

    1125 23.7 34.7

    2650 28.0 24.0

    51100 13.4 12.0

    101500 12.1 5.3

    501 1.5 2.7

    Total 100 100

    n 700 75a

    MCs= management consultancies.a Oneconsultancydid notprovidedetailsfor eitherturnoveror number

    of employees.

    at the surveyed management consultancies. The distribu-

    tion of those management consultancies that participated

    in the survey indicates that there is no problematic bias

    with regard to the sample.

    A total of 34.2% of the consulting firms operate only

    in one business sector, whereas 65.8% operate in at least

    two. The most frequently cited business sectors are strategy

    consulting (41 listings) and organizational consulting (39listings). Table 2 provides an overview by business sector.

    Altogether, 69.3% of the respondents are partners or

    vice presidents and 18.7% are consultants including project

    Table 2

    Surveyed management consultancies by business sector and number of

    business sectors.

    Business sectors Return (number of MCs)

    Strategy consulting 41

    Organizational consulting 39

    IT consulting/services 28

    Human resources consulting 24Other 31

    Total a

    n 76

    Number of business sectors Return (% of MCs)

    1 34.2

    2 28.9

    3 28.9

    4 2.6

    5 5.3

    Total 100

    n 76

    MCs= management consultancies.a Because multiple responses were allowed, the total is not given.

  • 8/4/2019 Project Contract5

    7/17

    C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145 135

    managers.6 It is therefore reasonable to assume that the

    majority of respondents have an adequate understanding

    of the contract types used by their respective company.

    4.2. Operationalization and measurement of the

    constructs

    The constructs used in this study were measured bymultiple indicators with two exceptions (reputation of

    service provider and fixed-cost contracts), which were

    measured by a single indicator. Scales already available in

    the literature were utilized where possible and adapted

    to the management consulting context. All scales and

    indicators are provided in Appendix A. In the following

    descriptions of scales for the constructs, the respective

    indicators are listed in parenthesis. Two types of measure-

    ment models were used depending on the hypothesized

    causal relationshipbetween a constructand its correspond-

    ing indicators. A reflective measurement model was used

    if observed indicators were manifestations of the under-

    lying construct and can be expected to have significantintercorrelations. In that case, the direction of causality is

    from the construct to its indicators, and changes in thecon-

    struct cause changes in all its indicators (Jarvis et al., 2003,

    p. 200). In contrast, if observed indicators cover different

    facets and jointly definethe characteristics of the construct,

    a formative measurement model was used. In that case,

    the direction of causality is from the indicators to the con-

    struct and indicators cannot be expected to have significant

    intercorrelations (Jarvis et al., 2003, p. 201).

    The scale for measuring the complexity of service com-

    prises five indicators based on Van de Ven and Ferry

    (1980, p. 392). Complexity is measured on the basis of

    structurability (com1 and com2), variability and speci-ficity of a service (com3com5). Structurability refers to

    the knowledge that the contracting parties have about the

    relationship between the required effort (input) and the

    resulting service output. This knowledge is the prerequi-

    site for breaking a task down into specific steps, so that the

    desired output can be achieved with sufficient certainty.

    For services with good structurability, the required solu-

    tion steps are known in advance, whereas this is not the

    case for services with poor structurability. Variability refers

    to the number of exceptions arising during the service pro-

    cess, which require different methods and approaches if

    they are to be solved (Van de Ven and Ferry, 1980, p. 392).

    In particular, it is the specificity of a service that leads tovariability. The more specific a service, i.e., the more a ser-

    vice needs to be adapted to the specific requirements of a

    client, the less a consulting firm is able to rely on previous

    experience, and the greater the uncertainty and complex-

    ity associated with the service (Gresov, 1989, p. 452; Van

    de Ven and Ferry, 1980, p. 392).

    The scale for measuring the integrativity level of service

    comprises six indicators. Despite its long tradition in the

    literature and despite being a constitutive characteristic of

    services, to the best of our knowledge, there is no scale

    6

    12% have another position in theconsulting firm(e.g., head of control-ling and finance or employee in HR, marketing or public relations).

    in the literature that encompasses all aspects of integra-

    tivity relevant to our study.7 In developing the scale, we

    reviewed the literature to identify potential indicators and

    tested them with experts from various management con-

    sulting firms. As the identified indicators cover different

    aspects of integrativity, a formative measurement model

    was chosen. In the literature, the integrativity level of ser-

    vice is characterized with reference to theinfluence a client

    exerts on the service process, as well as on the service out-

    put (e.g., Fitzsimmons and Fitzsimmons, 2001, p.5; Larsson

    and Bowen, 1989, p. 214). With reference to Van de Ven and

    Ferry (1980, p. 402), we capture the level of dependency

    of a consulting firm on the clients cooperation and col-

    laboration. The influence of a client on the service process

    (int1int3) is assumed to be greater, the more information

    and data required from him and the more coordination of

    activities required with him during the service process. The

    influence a client exerts on the service output (int4 and

    int5) is assumed to be greater, the larger the scope of his

    tasks and the higher the degree to which the quality of

    service output depends on his cooperation. In interviews

    with experts, it was indicated that the level of integrativ-

    ity should be differentiated with regard to the hierarchy

    level in the client firm (senior management, middle man-

    agement and administrative staff). In order to account for

    this, three indicators (int1, int2 and int4) were measured

    separately by hierarchy level. For these three indicators, the

    value is determined by averaging the values per hierarchy

    level. The final indicator measures the amount of time the

    consulting firm works onsite at the client (int6).

    The scale for measuring the verifiability of service

    provider behaviorcomprises three indicators. Firstly, it cap-

    tures how difficult it is for a client to evaluate service

    provider behavior (vsb1). Secondly it measures whether

    a client has the required knowledge and information for

    this evaluation (vsb2), and thirdly the level of resources

    required by the client to verify service provider behav-

    ior (vsb3). The scale for measuring the verifiability of

    client behaviorcomprises two indicators. These capture the

    degree to which the requirements made on the client are

    defined in advance (vcb1) and whether compliance with

    these requirements can be observed by the consulting firm

    (vcb2).

    The scale for measuring the verifiability of service output

    comprises three indicators. Verifiability of service output

    requires that performance measures be available that cap-

    ture all relevant performance dimensions of the service

    (ver1).8 The way in which the results of these measures

    are to be interpreted and used by the contracting parties,

    as well as by anindependentthird party, shouldbe as objec-

    tive as possible,in order to avoid any future disputes (ver2).

    The verifiability of service output also depends on the time

    7 Available scalesin theliterature only cover specific aspects of integra-

    tivity, such as the intensity of client contact (Kellogg and Chase, 1995).8 Analogously to the one-sided moral hazard case, it is assumed here

    that incentive problems arise when some relevant performance dimen-

    sions are not verifiable. Whether this is also a problem in the presence

    of double moral hazard, has not yet been proven formally. For incen-

    tive problems in the multitasking case with one-sided moral hazard, seeHolmstrm and Milgrom (1991).

  • 8/4/2019 Project Contract5

    8/17

    136 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    interval between theconclusion of a project andthe obtain-

    ment of project results (ver3). The longer this period, the

    more difficult it becomes to differentiate between the per-

    formance of the consulting firm and other external factors

    that impact on the service output (Kotter, 1995, p. 67).

    The scale for measuring the expected utility of future

    client relationships comprises three indicators. One aspect

    is the probability of being awarded future assignments by

    a client. Another aspect is the relevance and importance

    of an existing client in the acquisition of new clients. As

    these are different aspects of expected utility, we used

    a formative measurement model for this construct. The

    probability of acquiring future projects is captured by the

    proportion of those clients in a consulting firms portfo-

    lio with whom there is a long-term client relationship

    (exp1). Empirical studies indicate that consulting firms

    achieve approximately 6070% of their turnover by per-

    forming subsequent projects for existing clients (Glckler

    and Armbrster, 2003, p. 285). The value of a client for

    the acquisition of new clients is measured by how rele-

    vant recommendations are to the consulting firm (exp2).

    Several empirical studies show that recommendations by

    business partners are one of the most important decision

    criteria for clients in choosing a consulting firm (Glckler

    and Armbrster, 2003, p. 286). If a client is lost as a ref-

    erence due to poor performance of the consultancy (even

    if this might only be due to the clients subjective mis-

    perception), this can have considerable negative economic

    consequences for the consultancy. This is especially true if

    market transparencyis highand clients exchange their con-

    sulting experiences with one another (exp3). The last two

    indicators are based on Nooteboom et al. (1997, p. 337).

    Following Banerjee and Duflo (2000, p. 994) the rep-

    utation of service provider is measured by the age of the

    consulting firm (rep). The longer a consulting firm has been

    in business, the more likely it is to have established a good

    reputation.

    For fixed-cost contracts, we differentiate between two

    manifestations: fixed-cost contracts where the total remu-

    neration of the consulting firm is fixed (fix-total), and

    contracts where only the daily or hourly rate is fixed

    and the remuneration of the consulting firm is time and

    materials-based (fix-tim). Both contract types are indepen-

    dent of performance. For those consulting firms, which

    agree on variable-cost (performance-based) contracts with

    their clients (var1), in addition, the percentage of total

    compensation that depends on the achievement of con-

    tractually defined goals (var2) was captured. In the causal

    analysis, variable-cost (performance-based) contracts were

    measured by two indicators (var1 and var2), while both

    types offixed-cost contracts were each measured by a single

    indicator (fix-total or fix-tim).

    Two control variableswere included in the analysis. First,

    the size of a consulting firm, measured in terms of turnover

    and number of employees, can have an effect on the type

    of cost contract selected. The larger a consulting firm, the

    more likely it is to have sufficient market power to avoid

    contract types that it deems disadvantageous. Second, the

    focus of a consulting firm, i.e., the consulting firms service

    spectrum, might also have an impact on the chosen typeof cost contract. Consultancies which are highly special-

    ized might be able to evaluate potential performance and

    transactional risks more accurately, or, thanks to the lower

    heterogeneity of their services, face a lower output uncer-

    tainty than firms with a wider service spectrum. The focus

    of a consulting firm is measured by the number of business

    sectors in which it is active.

    4.3. Data analysis

    The data were analyzed using causal analysis based on

    the partial least squares (PLS) method. PLS is a variance

    analysis method that has several advantages for this study,

    compared to covariance-based methods such as LISREL.

    While covariance-based methods require relatively large

    minimum sample sizes in order to provide stable param-

    eter estimates, PLS also works for relatively small sample

    sizes. Moreover, theevaluation of fit of a PLS model is based

    on such resampling procedures as jackknifing or bootstrap-

    ping, which do not make parametric assumptions (Efron

    and Tibshirani, 1993). In contrast, the estimation proce-

    dures in LISREL require a multivariate normal distributionof the manifest variables (Fornell and Bockstein, 1982, p.

    289).9 Moreover, the analysis of formative constructs in a

    covariance analysis is only possible under certain condi-

    tions, failing which the model might not be identifiable

    (Jarvis et al., 2003, p. 213).

    Because the sample size of 76 data sets in this study

    is smaller than the recommended sample size for LISREL10

    and formative constructs were used,PLS was chosen for this

    study. The data analysis was conducted with SmartPLS 2.0

    (Ringle et al., 2005). In the section below, the reflective and

    formative measurement models are evaluated, followed by

    an evaluation of the structural model.

    4.4. Evaluation of the measurement models

    In order to evaluate the reflective measurement models,

    the indicator reliabilities were evaluated on the basis of the

    factor loadings (see Appendix A). A factor loading of at least

    .7 is required, meaning that the underlying latent variable

    should account for at least 50% of the variance of an indi-

    cator. Reflective indicators should be eliminated when the

    factor loadings in the PLSmodel are lower than .4 (Hulland,

    1999, p. 198). With few exceptions, all factor loadings are

    higher than .7 and significant at the 1% level. Because all

    factor loadings are well above .4, these few exceptions

    need not be considered problematicand no indicators wereeliminated. The factor reliabilities, which capture the inter-

    nal consistency of a construct, are of greater importance

    9 It should be noted that the estimation methods in LISREL are quite

    robust with regard to violations of the multivariate normal-distribution

    assumption (Jreskog and Srbom, 2001, p. 26).10 A rule of thumb often applied with LISREL is that the minimum sam-

    ple size should be five to ten times the number of unknown parameters

    in the model (Fornell and Bockstein, 1982, p. 289). Unknown parameters

    are loadings and measurement errors of all indicators in the measure-

    ment models, as well as the path coefficients between the endogenous

    and exogenous variables in the structural model. Depending on the spec-

    ification of the model here, there would be approximately 30 unknown

    parameters. The required sample size for LISREL would therefore beapproximately 150 data sets.

  • 8/4/2019 Project Contract5

    9/17

    C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145 137

    Table 3

    Measurement information.

    Construct Measurement model Number of items Cronbachs alpha (standardized) Factor reliability AVE

    1. Complexity of service R 5 .78 .83 .50

    2. Verifiability of service provider behavior R 3 .62 .78 .54

    3. Verifiability of client behavior R 2 .68 .84 .73

    4. Verifiability of service output R 3 .72 .84 .64

    5. Integrativity level of service F 6

    6. Expected utility of future client relationships F 3 7. Reputation of service provider R 1 1.00 1.00

    8. Type of cost contract

    Var R 2 .82 .91 .83

    Fix-Total R 1 1.00 1.00

    Fix-Tim R 1 1.00 1.00

    Notes: R = reflective measurement model; F = formative measurement model; AVE= average variance extracted; all reliability indicators are given only for

    reflective constructs; Var = variable-cost (performance-based) contract; Fix-Total = fixed-cost contract with total costs fixed; Fix-Tim = fixed-cost contract

    that is time and materials-based.

    Table 4

    Correlations and square root of average variance extracted per construct for variable-cost (performance-based) contracts.

    Construct 1 2 3 4 5 6 7 8

    1. Complexity of service .71

    2. Verifiability of service provider behavior .15 .743. Verifiability of client behavior .10 .21 .864. Verifiability of service output .41*** .22 .29** .805. Integrativity level of service .31*** .05 .03 .29** (F)6. Expected utilit y of fut ure client relat ionships .05 .39*** .29*** .03 .20 (F)7. Reputation of service provider .11 .10 .22 .03 .07 .10 1.008. Type of cost contract= Var .09 .24** .13 .31*** .31** .45*** .04 .91

    Notes: In the diagonal the square root of average variance extracted is given in bold (

    AVE); Var = variable-cost (performance-based) contract.** p < .05 (two-tailed).

    *** p < .01.

    in assessing the reflective measurement models. On the

    basis of both Cronbachs alpha and factor reliability, it is

    determined whether the indicators of a construct show a

    strong correlation to one another. The requirements are.7 for Cronbachs alpha (Nunnally, 1978) and .6 for factor

    reliability (Bagozzi and Yi, 1988, p. 82). As can be seen in

    Table 3, these conditions are fulfilled by each construct,

    with two exceptions for Cronbachs alpha (verifiability of

    service provider behaviorand verifiability of client behavior).

    Because factor reliabilities, in contrast to Cronbachs alpha,

    consider the actual factor loadings in the weighting of indi-

    cators, and both are clearly higher than the threshold of .6,

    it can be concluded that the measurement of constructs is

    reliable.11

    The discriminant validity of the constructs was evalu-

    ated on the basis of the Fornell/Larcker criterion (Fornell

    and Larcker, 1981). The objective is to ensure that differ-ent constructs capturedifferent aspects in terms of content.

    For each type of cost contract and for all construct pairs, an

    evaluation was made of whether the square root of average

    variance extracted for each construct is greater than the

    correlation between these two constructs. It can be seen

    in Tables 46, that this condition is fulfilled for all (reflec-

    tively) measured construct pairs.

    When evaluating the two formatively measured con-

    structs (integrativity level of service and expected utility

    of future client relationships), it is important to note that

    11

    In calculating theCronbach Alpha, all indicators have an equal weight(Bagozzi and Yi, 1988, p. 82).

    standard reliability and validity criteria are no longer

    applicable (Bollen, 1989; Diamantopoulos and Winklhofer,

    2001).12 The reason for this is the inverse causal relation-

    ship between the latent variable and its indicators. Whilethe latent variable is assumed to determine the values of

    its indicators in reflective measurement models, the indi-

    cators in formative measurement models are assumed to

    determine the latent variable. Therefore, the evaluation

    of model fit, based on the internal consistency of indica-

    tors, is no longer meaningful, because the indicators are

    not necessarily highly and positively correlated with each

    other.13 In formative measurement models, the relevance

    of an indicator is, instead, evaluated on the basis of its

    respective indicatorweight,rather than its reliability (Chin,

    1998, p. 307). The indicator weights show which indica-

    tors contribute most to the formation of a construct. In

    order to evaluate the factor validity of formative measure-ment models, it is proposed in the literature to evaluate

    its nomological validity (Reinartz et al., 2004, p. 298). The

    evaluation is thus based on the significance, strength and

    direction of its relation to other latent variables in the pro-

    posed model.14

    12 Fora discussion of the differing validityrequirementsof formative and

    reflective constructs, see Bollen and Lennox (1991), Chin andGopal(1995)

    and Cohen et al. (1990).13 The indicators in formative measurement models can have a positive,

    negative or zero correlation (Bollen and Lennox, 1991, p. 307; Chin, 1998,

    p. 306).14

    The significance tests in PLS are based on approximated t-statisticsgenerated by resampling procedures (Chin, 1998, pp. 318320).

  • 8/4/2019 Project Contract5

    10/17

    138 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    Table 5

    Correlations and square root of average variance extracted per construct for fixed-cost contracts with total costs fixed.

    Construct 1 2 3 4 5 6 7 8

    1. Complexity of service .71

    2. Verifiability of service provider behavior .13 .743. Verifiability of client behavior .10 .23** .864. Verifiability of service output .39*** .25** .31*** .805. Integrativity level of service .00 .04 .16 .23** (F)6. Expected utility of future client relationships .05 .15 .15 .05 .11 (F)7. Reputation of service provider .11 .01 .22 .03 .11 .16 1.008. Type of cost contract = Fix-Total .15 .26** .22 .10 .33*** .22 .08 1.00

    Notes: In the diagonal the square root of average variance extracted is given in bold (

    AVE); Fix-Total= fixed-cost contract with total costs fixed.** p < .05 (two-tailed).

    *** p < .01.

    Table 6

    Correlations and square root of average variance extracted per construct for fixed-cost contracts that are time and materials-based.

    Construct 1 2 3 4 5 6 7 8

    1. Complexity of service .71

    2. Verifiability of service provider behavior .11 .743. Verifiability of client behavior .09 .19 .864. Verifiability of service output

    .40*** .29** .31*** .80

    5. Integrativity level of service .08 .06 .01 .26** (F)6. Expected utility of future client relationships .02 .14 .36*** .04 .09 (F)7. Reputation of service provider .11 .11 .22 .03 .01 .13 1.008. Type of cost contract = Fix-Tim .12 .23** .11 .38*** .44*** .20 .07 1.00

    Notes: In the diagonal the square root of average variance extracted is given in bold (

    AVE); Fix-Tim = fixed-cost contract that is time and materials-based.** p < .05 (two-tailed).

    *** p < .01.

    Because formative measurement models are based

    on the principle of multiple regression analysis, their

    indicators must first be checked for multicollinearity. Mul-

    ticollinearity in formative measurement models leads to

    inflated standard errors in the coefficients, which can

    lead to biased individual indicators of the latent variable(Fornell and Bockstein, 1982, p. 292). The examination of

    both formatively measured constructs did not indicate any

    problematic multicollinearity.15 All indicators for the two

    formatively measured constructs, except one (int5), have

    significant weights with regard to at least one type of cost

    contract. Low and/or insignificant weights, however, do

    not clearly demonstrate the lack of relevance of an indi-

    cator, but merely that this indicator does not contribute

    to explaining the specific type of cost contract considered.

    One of the indicators (int5) for the construct integrativ-

    ity level of service always has weights close to zero and

    is never significant, which indicates that it is hardly rel-

    evant to this construct. It must be noted, however, thatbecause allformative indicators together define a construct,

    theyshould not be eliminated automatically because of low

    and/or insignificant weights (Rossiter, 2002, p. 315).

    15 The test of multicollinearity is based on thefollowingcriteria: (a) cor-

    relationsbetweenthe indicators of a construct(valuesclose toone indicate

    multicollinearity); (b) tolerance of the indicators (values less than .1 indi-

    cate multicollinearity (Hair et al., 1998, p. 208)); (c) condition index of

    the indicators (valuesbetween 10 and30 indicate moderate, values larger

    than 30, severe multicollinearity (Belsley et al., 1980, p. 117)). Integrativ-

    ity level of service: (a) between .05 and .69; (b) between .45 and .81; (c)

    between 5.76 and 21.06; expected utility of a long-term client relation-

    ship: (a) between .30 and .44; (b) between .75 and .85; (c) between 8.93and 15.48. All calculations conducted with SPSS 13.0.

    4.5. Results

    In evaluating the structural model, the PLS method has

    the disadvantage, compared to covariance-based methods,

    that no inferential statistical tests can be performedto eval-

    uate the overall model fit. This is due to the less restrictivedistribution assumptions with regard to the manifest vari-

    ables. Instead, the structural model can only be evaluated

    on the basis of non-parametric tests, i.e., the coefficient of

    determinationR2 of theendogenous variables as well as the

    signsand significanceof thepathcoefficients(Chin, 1998, p.

    316). The coefficient of determination R2 shows the fraction

    of explained construct varianceand measuresthe goodness

    of fit of a regression function to the empirically measured

    manifest variables. The predictive relevance of the model is

    evaluated on the basis of the non-parametric StoneGeisser

    test (Chin, 1998, p. 315; Fornell and Cha, 1994, pp. 7173).

    The StoneGeisser test criterion, which has a value between

    1 and +1, indicates how well a dependent variable can bereconstructed by the model. If this criterion is greater than

    zero, the model has predictive relevance.

    Theresultsof thecausal analysisbased on PLSare shown

    in Tables 7 and 8. The coefficients of determination R2

    are between .25 and .36, which means that between 25%

    and 36% of the variance of the implemented type of cost

    contract is explained by the model. Comparedto other stud-

    ies, this value can be considered satisfactory.16 With Q2

    16 Previous empirical studies on the determinants of different forms of

    behavior control have yielded coefficients of determination between .11

    and.65 (e.g., Abernethy et al., 2004; Eisenhardt, 1985; Kirsch, 1996; Krafft,1999).

  • 8/4/2019 Project Contract5

    11/17

    C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145 139

    Table 7

    Results of causal analysis with PLS for variable-cost (performance-based) contracts.

    Hypothesis (anticipated sign) Type of cost contract Standardized parameter coefficient t-Value Empirical support (Yes/No)

    H1 () Var .19** 1.66 YesH2 () Var .15 (ns) .89 NoH3 () Var .20** 1.98 YesH4 (+) Var .32*** 3.56 Yes

    H5 () Var .37*** 3.95 YesH6 (+) Var .12 (ns) 1.15 NoH7 (+) Var .23*** 2.39 Yes

    H8 () Var .34*** 3.90 YesH9 () Var .12* 1.47 YesSize (control variable) Var .02 (ns) .18 Focus (control variable) Var .13* 1.54 R2 Var .36

    Q2 Var .25

    Notes: ns= not significant; Var = variable-cost (performance-based) contract.* p < .10 (one-tailed).

    ** p

  • 8/4/2019 Project Contract5

    12/17

    140 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    Fig. 2. Summary of results.

    ifiability of service output (H4), the integrativity level of

    service, expected utility of future client relationships, ver-

    ifiability of service provider behavior and service provider

    reputation, all have a significant negative impact on the

    implementation of variable-cost (performance-based) con-

    tracts (H317, H8, H1, and H9). In particular, the effect of

    the integrativity level of service provides useful insights.

    As predicted, there is a negative relationship between the

    integrativity level and the implementation of variable-cost

    (performance-based) contracts, as the agencys opportu-

    nity cost increases with the integrativity level. However,

    there is also a negative relationship between the integra-

    tivity level and the implementation of fixed-cost contracts

    with total costs fixed. The only positive relationship is with

    fixed-cost contracts that are time and materials-based. This

    result can be interpreted as follows. Time and materials-

    based contracts often have the advantage for a client

    that they can be terminated within a short time frame.

    Therefore, in the short-term, time and materials-based

    contracts have the characteristic of fixed-cost contracts

    with total costs fixed, as the client has to pay for the

    17 To test H3 for variable-cost (performance-based) contracts, we addi-

    tionally calculated the correlation between the integrativity level of

    service and the average percentage of compensation that depends on the

    achievement of contractually defined objectives. The correlation is nega-

    tive (.283) and significant at the 5% level (two-tailed test) and providesadditional support for H3.

    work performed, regardless of the results achieved. In

    the long-term, however, they tend to have the character-

    istic of variable-cost (performance-based) contracts. The

    positive relationship between the integrativity level and

    the implementation of fixed-cost contracts that are time

    and materials-based indicates that, in practice, they serve

    as a compromise between variable-cost (performance-

    based) and fixed-cost contracts with total costs fixed. As

    expected, the relationship between the expected utility of

    future client relationshipsand the implementation of fixed-

    cost contracts, both with total costs fixed and time and

    materials-based, is positive (H8). The relationship between

    reputation and fixed-cost contracts (H9), however, is not

    significant. Fig. 2 provides a summary of the results.18

    The effect size of each construct can be calculated in

    two steps from the coefficient of determination of a depen-

    dent latent variable, oneincluding (R2incl

    ) and one excluding

    (R2excl

    ) the respective independent latent variable:

    f2 =R2

    incl R2

    excl

    1 R2incl

    .

    18 We also assessed whether the characteristics of the contracting rela-

    tionship potentially moderate the effects of the characteristics of the

    service on the type of cost contract. Following the approach suggested

    by Chin et al. (2003) we found that all moderating effects are small andnot significant.

  • 8/4/2019 Project Contract5

    13/17

    C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145 141

    Table 9

    Effect size of constructs for each type of cost contract.

    Construct Type of cost contract

    Var Fix-Total Fix-Tim

    Verifiability of service provider behavior .04 () .08 (+) .07 ()Verifiability of service output .09 (+) .00 () .05 ()Integrativity level of service .04 () .08 () .12 (+)Expected utility of future client relationships .11 (

    ) .03 (+) .05 (+)

    Reputation of service provider .02 () .01 (+) .00 ()Size (control variable) .00 () .01 (+) .00 (+)Focus (control variable) .02 () .06 () .03 (+)

    Notes: The direction of the relationship between a construct and the type of cost contract is given in brackets; Var= variable-cost (performance-based)

    contract; Fix-Total= fixed-cost contract with total costs fixed; Fix-Tim = fixed-cost contract that is time and materials-based.

    The greater the value of f2, the greater the respec-

    tive effect size. Values of .02, .15 and .35 indicate weak,

    medium and strong effects, respectively (Chin, 1998, p. 316;

    Cohen, 1988, p. 413). From Table 9, it can be seen that each

    type of cost contract has two main determinants with a

    moderate effect size. The implementation of variable-cost(performance-based) contracts is negatively affected by the

    expected utility of future client relationships and positively

    bythe verifiability of service output. This canbe regarded as

    an empirical confirmation of the hypothesis that contracts

    independent of performance are preferred, if the expected

    utility of future client relationships is sufficiently high,

    despite the availability of verifiableperformance measures.

    While the implementation of fixed-cost contracts that are

    time and materials-based is positively related to the inte-

    grativity level of service, fixed-cost contracts with total

    costs fixed are positively related to the verifiability of ser-

    vice provider behavior.

    5. Conclusions

    This study sought to develop a better understanding of

    the determinants of contract terms between professional

    services firms and their clients. The intention was to bet-

    ter understand the control mechanisms used to manage

    short-term inter-organizational exchanges characterized

    by a high level of transactional uncertainty and a double

    moral hazard risk.

    While we tookcare to address methodological concerns,

    several limitations of the study should be noted when con-

    sidering the empirical evidence presented here. As with

    all cross-sectional surveys, the results do not constitute

    proof of the relationships. Rather, the evidence presented

    can only be said to be consistent with the theoretical posi-

    tion developed in the paper. The major problem is the low

    response rate which limits the generalization of our find-

    ings to larger populations of professional services firms.

    While a test for non-response bias did not indicate any

    problems, a higher response rate could yield more gen-

    eralizable findings. Notwithstanding these limitations, the

    study does provide some useful empirical insights into

    formal controls used by service companies to govern trans-

    actions at the organizationcustomer interface.

    First, the empirical data indicate that service charac-

    teristics exert a significant impact on the chosen contracttype. Consistent with prior empirical results, the use of

    outcome-based controls (i.e., variable-cost (performance-

    based) contracts) is positively related to services for which

    the output can be measured and verified, while the use

    of behavior-based controls (i.e., fixed-cost contracts) is

    positively related to the verifiability of service provider

    behavior (Eisenhardt, 1985; Kirsch et al., 2002). The use

    of fixed-cost contracts that are time and materials-based,

    which can be considered a mix between behavior-based

    and outcome-based controls is positively related to ser-

    vices that are characterized by a high level of integrativity.

    This contract type has the advantage for clients of provid-

    ing them with a high degree of flexibility in their decision

    on whether to continue an ongoing project or not. Man-

    agement consulting firms therefore have an incentive to

    choose a high level of effort even in situations where nei-

    ther their behavior nor the service output is verifiable, in

    order to avoid risking a project being terminated, because

    a customer feels he is not being treated professionally and

    appropriately.

    Second, variable-cost (performance-based) contracts

    might not be optimal, despite the verifiability of ser-

    vice output. For professional services characterized by a

    high level of integrativity, fixed-cost contracts can pro-

    vide stronger incentives to all contracting parties than

    variable-cost (performance-based) contracts. The reason is

    that output sharing can weaken the incentives for all con-

    tracting parties, which can be circumvented, provided that

    the expected utility of future client relationships is high

    enough to prevent the service provider from engaging in

    opportunistic behavior.

    Finally, the results indicate that trust and reputation

    have an impact on the choice of controls used in short-

    term contracts. As markets and contractual relationships

    are usually not temporally limited nor anonymous, both

    contracting parties need to take into account that oppor-

    tunism often only pays off in the short-term. This offers

    the possibility to control the behavior of those involved in

    the service encounter,despitedifficulties in the verifiability

    of both effort and service output. Provided that a manage-

    ment consultancy wishes to remain in the market over the

    long-term, the prospects of either obtaining future assign-

    ments from an existing client or having valuable references

    for the acquisition of new clients can mitigate potential

    opportunistic behavior on their side. Behavioral controls

    can therefore alsobe applied effectively in situationswhere

    behavior is observable, but not verifiable.

  • 8/4/2019 Project Contract5

    14/17

    142 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    Our results contribute to the field of management

    accounting by providing insights on control mechanisms

    used to manage inter-organizational exchanges, in par-

    ticular for service companies. Management accounting

    research on service organizations often focuses on mea-

    suring the service output in order to increase the efficiency

    of service encounters. This study indicates, however, that

    it might be as beneficial for both service organization and

    client to raise the level of transparency of the service pro-

    cess and, thereby, the observability (not necessarily the

    verifiability) of the behavior of the contracting parties.

    This is particularly true for services characterized by a

    high integrativity level and for contractual relationships

    characterized by a high expected utility of future client

    relationships. Various measures in the area of project man-

    agement can be applied to achieve this. Examples are

    regular steering committee meetings or, for the client, the

    staffing of experienced client employees in a project team.

    It is importantto note, however, that it cannotbe concluded

    from the results, that the implementation of fixed-cost

    contracts that are independent of performance makes it

    unnecessary or even harmful for clients or consulting

    firms to measure and evaluate the service output. The

    service output should still be measured, but it may be ill-

    advised to attempt to provide an incentive to the service

    provider by making his remuneration contingent on this

    output.

    Acknowledgements

    We gratefully acknowledge the helpful comments and

    suggestions from the editor, Michael Bromwich, and the

    anonymous reviewers on previous drafts of this paper.

    Appendix A. Constructs and indicators

    (1) Reflectively measured constructs

    Construct/indicators M S Factor loading Indicator reliability

    Complexity of service

    In our projects, problems for which there are no direct or obvious solutions

    arise frequently. (17c) (com1)

    3.95 1.82 .78 .61

    Approximately how much of their work time do your consultants spend

    finding solutions to problems for which there are no direct or obvious

    solutions? (Solution time in % of the work-week time: 1 = 020%,

    2 = 2140%, 3 = 4160%, 4 = 6180%, 5 = 81100%) (com2)

    1.93 .93 .83 .68

    To what extent are the project tasks for your consultants similar from day to

    day? (1= almost all are the same, 2= many are the same, 3 = 50% are the

    same, 4= many are different, 5 = almost all are different) (com3)

    3.46 .79 .58 .34

    How often do difficulties and problems arise in a project that require your

    consultants to use substantially different methods and approaches

    compared to other projects? (15a

    ) (com4)

    2.78 1.04 .69 .48

    How often do unanticipated situations arise in your projects that have a

    substantial impact on the course of a project? (15 a) (com5)

    3.14 .98 .63 .40

    Verifiability of service provider behavior

    During an ongoing project, our clients can easily verify whether or not we

    are performing our work well. (17c) (vsb1)

    5.17 1.29 .88 .77

    Our clients usually have the required knowledge and information to evaluate

    our work during an ongoing project. (17c) (vsb2)

    3.92 1.80 .61 .37

    Our clients need to spend considerable resources in order to evaluate and

    monitor our work during an ongoing project. (17c) (recoded) (vsb3)

    5.47 1.47 .70 .50

    Verifiability of client behavior

    In our contracts, we establish precisely and in detail the performances and

    resources we expect from our clients in the course of a project. (17 c)

    (vcb1)

    4.96 1.68 .95 .91

    During an ongoing project, we can verify accurately whether or not a client is

    complying with his agreed-upon responsibilities. (17c) (vcb2)

    5.67 1.27 .75 .56

    Verifiability of service outputIn our projects, we frequently face the situation that major output

    dimensions are not objectively measurable. (17c) (recoded) (ver1)

    3.74 1.84 .87 .76

    After a project is finalized, it is possible to agree upon the level of success of

    the project and the quality of our service easily and without dissent.

    (17c) (ver2)

    5.22 1.28 .84 .70

    It often takes a long time after a project is finalized, before its success or

    failure can be determined. (17c) (recoded) (ver3)

    4.26 1.80 .69 .47

    Variable-cost (performance-based) contract

    How high is the relevance of the following forms of compensation in your

    projects? (15b)

    Variable-cost (performance-based) contracts: the client pays a fee, the size

    of which depends on the achievement of contractually specified goals.

    (var1)

    2.24 1.33 .94 .89

  • 8/4/2019 Project Contract5

    15/17

    C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145 143

    Appendix A (Continued )

    Construct/indicators M S Factor loading Indicator reliability

    If you conclude a variable-cost (performance-based) contract: what is the

    average percentage of your total compensation that is contingent on the

    achievement of contractually specified goals? (1 = less than 5%, 2 = 610%,

    3 = 1120%, 4 = 2140%, 5 = 4160%, 6 = 6180%, 7 = 81100%) (var2)

    2.77 1.91 .88 .77

    Fixed-cost contract with total costs fixed and fixed-cost contract that is time and

    materials-based

    How high is the relevance of the following forms of compensation in yourprojects? (15b)

    Fixed-cost contracts: The client pays a fixed cost for a contractually

    agreed-upon service. This cost does not include any performance-based

    components. (fix-total)

    3.26 1.30

    Time and materials-based contracts: The client pays an agreed-upon daily

    (or hourly) rate for the time spent on the project. This cost does not

    include any performance-based components. (fix-tim)

    3.87 1.27

    Reputation of service provider

    In what year was your management consultancy founded? (recoded into

    years) (rep)

    17.2 10.2

    Notes: M = mean; S = standard deviation.a 5-point scale: 1= very rarely, 2= sometimes, 3 = fairly often, 4= often, 5 = very often.b 5-point scale: 1= not applicable to us, 2 = low, 3= medium, 4 = high, 5 = very high.c 7-point scale: 1 = do not agree at all, 7 = agree totally and completely.

    (2) Formatively measured constructs

    Construct/indicators M S Weight (t-Value) per type of cost contract

    Var Fix-Total Fix-Tim

    Integrativity level of service

    During their daily project work, how much do your consultants depend on

    the following persons to obtain the required data, information, materials,

    etc.? (15a) (int1)

    3.35 .66 .12 (.33) .12 (.41) .34* (1.37)

    Clients senior management (management board, managing director, area

    manager); clients middle management (department heads, supervisors);

    clients administrative staff

    How often do consultants need to coordinate their activities with thefollowing persons during the performance of their main tasks? (15 a)

    (int2)

    2.61 .81 .01 (.03) .03 (.11) .26*

    (1.29)

    Clients senior management; clients middle management; clients

    administrative staff

    In ourprojects, we are also able to perform our tasks successfully withoutthe

    cooperation of our clients (or their employees). (17 b) (recoded) (int3)

    5.88 1.46 .55* (1.34) .24 (.67) .05 (.17)

    When your consultants finalize their part of the work, how much do they

    have to rely on the following persons to complete the next steps before

    the service is completed? (15a) (int4)

    3.23 .75 .33 (1.03) .24 (.80) .30* (1.50)

    Clients senior management; clients middle management; clients

    administrative staff

    In our projects, we are also able to ensure high-quality project results

    without the cooperation of our clients (or their employees). (17 b)

    (recoded) (int5)

    5.43 1.72 .02 (.04) .18 (.49) .10 (.35)

    On average, how much of their working time do your consultants spend on

    site with the client? (1= up to 1 day a week, 2= 2 days a week, 3= 3 days aweek, 4= 4 days a week, 5= 5 days a week) (int6)

    2.52 1.42 .45** (1.68) .90*** (3.49) .70*** (3.69)

    Expected utility of future client relationships

    We have long-term business relationships with most of our clients. (17b)

    (exp1)

    5.89 1.53 .73*** (3.80) .02 (.06) .17 (.38)

    Recommendations from our clients are very important to us for the

    acquisition of new clients. (17b) (exp2)

    6.24 1.16 .23 (.84) 1.05*** (3.38) .29 (.54)

    Clients who are not satisfied with our services can cause us substantial

    harm, due to their business connections with other potential clients.

    (17b) (exp3)

    5.38 1.55 .65*** (3.28) .21 (.51) .80** (2.11)

    Notes: M = mean; S = standard deviation; Var= variable-cost (performance-based) contract; Fix-Total= fixed-cost contract with total costs fixed; Fix-

    Tim = fixed-cost contract that is time and materials-based.a 5-point scale: 1= not at all, 2= a little, 3= medium, 4 = greatly, 5 = very greatly.b 7-point scale: 1 = do not agree at all, 7 = agree totally and completely.* p < .10 (one-tailed).

    ** p

  • 8/4/2019 Project Contract5

    16/17

    144 C. Homburg, P. Stebel / Management Accounting Research 20 (2009) 129145

    References

    Abernethy, M.A., Bouwens, J., van Lent, L., 2004. Determinants of controlsystem design in divisionalized firms. Acc. Rev. 79 (3), 545570.

    Akerlof, G.A., 1970. The market for lemons: quality uncertainty and themarket mechanism. Quart. J. Econ. 84 (3), 488500.

    Anderson, S.W., Dekker,H.C., 2005. Managementcontrol for market trans-actions: the relation between transaction characteristics, incompletecontract design, and subsequent performance. Manage. Sci. 51 (12),

    17341752.Anderson, S.W., Glenn, D., Sedatole, K.L., 2000. Sourcing parts of complexproducts: evidence on transaction costs, high-powered incentives andex-post opportunism. Acc. Organ. Soc. 28 (8), 723749.

    Bagozzi, R.P., Yi,Y., 1988. On the evaluation of structural equation models.J. Acad. Market. Sci. 16 (1), 7494.

    Banerjee, A.V., Duflo, E., 2000. Reputation effects and the limits of con-tracting: a study of the Indian software industry. Quart. J. Econ. 115(3), 9891017.

    Belsley, D.A., Kuh, R.E., Welsch, R.E., 1980. Regression Diagnostics: Identi-fying Influential Data and Sources of Collinearity. John Wiley & Sons,New York.

    Bhattacharyya, S., Lafontaine,F., 1995. Double-sided moral hazard and thenature of share contracts. RAND J. Econ. 26 (4), 761781.

    Bollen, K.A., 1989. Structural Equations with Latent Variables. John Wiley& Sons, New York.

    Bollen, K.A., Lennox, R., 1991. Conventional wisdom on measure-ment: a structural equation perspective. Psychol. Bull. 110 (2),305314.

    Chenhall, R.H., 2003. Managementcontrol systems design within its orga-nizational context: findings from contingency-based research anddirections for the future. Acc. Organ. Soc. 28 (2/3), 127168.

    Chi, T., 1996. Performance verifiability and outputsharing in collaborativeventures. Manage. Sci. 42 (1), 93109.

    Chin,W.W., 1998. The partial leastsquares approachto structural equationmodeling. In: Marcoulides, G.A. (Ed.), Modern Methods for BusinessResearch. Lawrence Erlbaum, Mahwah, NJ, pp. 295336.

    Chin,W.W., Gopal,A., 1995.Adoptionintention in GSS:relative importanceof beliefs. Data Base Adv. Inf. Syst. 26 (2/3), 4264.

    Chin, W.W., Marcolin, B.L., Newsted, P.N., 2003. A partial least squareslatent variable modeling approach for measuring interaction effects:results from a Monte Carlo simulation study and an electronic-mailemotion/adoption study. Inf. Syst. Res. 14 (2), 189217.

    Cohen, J., 1988. Statistical Power Analysis for the Behavioral Sciences, 2nded. Lawrence Erlbaum, Hillsdale, NJ.

    Cohen, P., Cohen, J., Teresi, J., Marchi, M., Velez, C.N., 1990. Problems inthe measurement of latent variables in structural equations causalmodels. Appl. Psych. Meas. 14 (2), 183196.

    Cooper, R., Ross, T.W., 1985. Product warranties and double moral hazard.RAND J. Econ. 16 (1), 103113.

    De Vaus, D.A., 1995. Surveys in Social Research. Allen and Unwin, NorthSydney.

    Dekker,H.C., 2004. Control of inter-organizational relationships:evidenceon appropriationconcernsand coordination requirements.Acc. Organ.Soc. 29 (1), 2749.

    Diamantopoulos, A., Winklhofer, H.M., 2001. Index construction with for-mative indicators: an alternative to scale development. J. Mark. Res.38 (2), 269277.

    Dornstein, M., 1977. Some imperfections in the market exchangesfor professional & executive services. Am. J. Econ. Sociol. 36 (2),113128.

    Dybvig, P.H., Lutz, N.A., 1993. Warranties, durability, and maintenance:two-sided moral hazard in a continuous-time model. Rev. Econ. Stud.60 (3), 575597.

    Efron, B., Tibshirani, R.J., 1993. An Introductionto theBootstrap.Chapman& Hall, New York.

    Eisenhardt, K.M.,1985. Control: organizational and economic approaches.Manage. Sci. 31 ( 2), 134149.

    Eswaran,M., Kotwal, A., 1985. A theoryof contractual structurein agricul-ture. Am. Econ. Rev. 75 (3), 352367.

    Fitzsimmons, J.A., Fitzsimmons, M.J., 2001. Service Management: Opera-tions, Strategy and InformationTechnology,3rd ed. McGraw-Hill,NewYork.

    Flie, S., Kleinaltenkamp, M., 2004. Blueprinting the service com-pany: managing service processes efficiently. J. Bus. Res. 57 (4),392404.

    Fornell, C., Bockstein, F.L., 1982. A comparative analysis of two structuralequation models. LISREL and PLS applied to market data. In: Fornell,C. (Ed.), A Second Generation of Multivariate Analysis. Praeger, NewYork, pp. 289323.

    Fornell, C., Cha, J., 1994. Partial least squares. In: Bagozzi, R.P. (Ed.),Advanced Methods of MarketingResearch. Blackwell, Cambridge, MA,pp. 5278.

    Fornell, C., Larcker, D.F., 1981. Evaluating structural equation models withunobservable variables and measurement error. J. Mark. Res. 18 (1),3950.

    Frances, J., Garnsey, E., 1996. Supermarkets and suppliers in the UnitedKingdom: system integration, information and control. Acc. Organ.Soc. 21 (6) , 591610.

    Fudenberg, D.,Maskin, E., 1986. The folktheorem in repeated games withdiscounting or with incomplete information. Econometrica 54 (3),533554.

    Fudenberg, D., Levine, D.K., 1989. Reputation and equilibrium selection ingames with a patient player. Econometrica 57 (4), 759778.

    Fudenberg, D., Kreps, D.M., Maskin, E.S., 1990. Repeated games with long-run and short-run players. Rev. Econ. Stud. 57 (4), 555573.

    Glckler, J., Armbrster, T., 2003. Bridging uncertainty in managementconsulting:the mechanismsof trust andnetworked reputation. Organ.Stud. 24 (2), 269297.

    Gresov, C., 1989. Exploring fit and misfit with multiple contingencies.Admin. Sci. Quart. 34 (3), 431453.

    Groot, T.L.C., Merchant, K.A., 2000. Control of in