.,[,!.!,! MIT LIBRARIES 3 9080 02239 1178 DUPL ,.," ;'/ v.".'!',' '.;,'''.'. :;;;:,f:;;i;;:tei|i;ji|S|;i|;!;g v.v' " ^aMtitiir jmrtm ifhMtbtm* <* -uukk, jjt ro28 .M414 DESIGN TEAM PERFORMANCE: A CONTROL THEORY PERSPECTIVE l/S John C. Henderson Soonchul Lee December 1989 CISR Sloan WP WP No. 198 No. 3104-89 Center for Information Systems Research Massachusetts Institute of Sloan School of Technology Monogement 77 Massachusetts Avenue Cambridge, Massachusetts, 02139 NOV 1 4 2000 libraries" n \110 DESIGN TEAM PERFORMANCE: A CONTROL THEORY PERSPECTIVE l/S John C. Henderson Soonchul Lee December 1989 CISR Sloan ©1989 WP WP J.C. No. 198 No. 3104-89 Henderson, S. Lee Center for Information Systems Research Sloan School of Management Massachusetts institute of Technology I/S Design Team Performance: A Control Theory Perspective Abstract The control relationship between project managers and aspect of the working of any I/S design team. team members is a central This paper uses control theory to develop measures of managerial control and team member control from both output and process perspectives. The control relationship is studied for 41 actual design teams. indicate that high performing teams exhibit both strong process control by strong outcome control by peers. Results managers and 1.0 I/S DESIGN TEAM PERFORMANCE: A CONTROL THEORY PERSPECTIVE The increased use well of information technology as a key element of strategy has been documented (Bakos and Treacy 1984, Parsons 1983. Rockart manage Kemerer (I/S) yet, the ability to effectively still represents decade and a high percentage of systems paper we examine the performance of In this And and Scott Morton 1984). and Learmonth Ives a significant (1989), for example, noted that backlogs in I/S have been increased substantially over the last late. Cash and Konsynski 1985, development of information systems the challenge. 1986, I/S finish over budget and design teams from the perspective of control theory. While relationship performance clearly not the only theory that in this project-oriented task different model of effective I/S design. among members on a (1987) argued that interactions roles, One including manager, among these roles is a central to effective I/S 1978, design teams are designers, and critical step in aspect of these relationships team (Boland among is environment. Because most understanding the underlying relationships influence applicable to this issue, the control between the project manager and the team members composed of a number of ing a is is Robey and Farrow users, develop- the pattern of 1982). Salaway these roles occur at the intersection where user knowledge (business knowledge) and designer knowledge (I/S knowledge) meet. Each kind of knowledge must influence the other effectively in order to design information systems that can better skills, meet user needs. In addition, designers expectations, and motivations to the design process. and users bring different goals, Thus, a traditional task of the manager of their is to influence both designers own to work toward the team's goals instead goals. The above Flamholtz and users et al. discussion fits quite well with a control perspective of performance. (1985) defined control as the organization's attempts to increase the probability that employees will behave in ways that lead to the attainment of organizational goals. Weber (1947) defined control as a process of creating and monitoring rules through In his view, control systems regulate patterns of interaction to hierarchical authority. restrict employees' behavior. Ouchi (1979) saw control as an evaluation process which is based on the monitoring and evaluation of behaviors or outputs. In the leadership literature, control has been defined as a leader's taking actions that induce a subordinate to adopt organizational goals and Berkes 1979, Schriesheim et al. 1976). Thompson (House and Dessler 1974, Jermier (1967) and Reeves and Woodward (1970) defined control as a cybernetic process of testing, measuring, and providing feedback with respect to a defined goal structure. We will argue for a perspective of control that includes both managerial control and collegial control. That is, for the purpose of studying I/S design teams we will consider both control as exercised by a project manager and control exerted by members of the team on each other. This perspective example, Boland (1978), is reflected in many De Brabander and studies in the MIS design literature. For Thiers (1984), Henderson (1988), and others investigated different patterns of influence between designers and users in the design ot They found information systems. that strong mutual influence of designers and users produces an environment that better matches the way design actually takes place. Keider (1984) and Schmitt and Kozar (1978) found that most projects that lacked the basic management principles of such planning and control had failed. Similarly, Mills (1971) and others have focused on leadership as a predictor of design team performance. In the following we synthesize a number of of the team process in I/S planning and design. that managerial control variation in from 41 2.0 I/S design A We test the model central proposition of the and team-member control can explain a team performance. model control theories to provide a is significant portion of the proposition using empirical data gathered teams and discuss the implications for the I/S function. ALTERNATIVE CONTROL MODELS Mantei (1981) claimed that the two predominant control structures in I/S planning and design are the chief programmer team proposed by Mills (1971) and the egoless programming team proposed by Weinberg authority belongs throughout the to the (1971). chief programmer, team membership. programmer design team, but while In the former, the decision-making in Communications are latter, control centralized are decentralized in Weinberg's team. programmer team can be characterized as Weinberg's team team-member is the characterized by strong That having strong managerial control. in is diffused the is, chief a chief control while March and Simon (1958) proposed type of control structure hierarchical control possible; when is effective tasks become is (1950) and Leavitt (1951), When appropriate. is coordination by feedback that task predictability largely determines tasks are and work sequencing variable necessary and decentralized control in their programmer team as structure position we centralized First, as individual; to more take in this paper is is is to predict, effective. Bavelas difficult more to more time and generate Such research suggests many that design teams often fail In this paper and design. in most real software out, the effective chief lauded for to finish their tasks we propose This team structure team-member is control. chief approach that neither the purely hierarchical development programmer is its is a rare unlikely Second, communication and decision-making complexity. although the decentralized group a types of I/S design tasks, while a most so-called chief programmer teams are headed by someone who the twice as effective for other types. appropriate Yourdon (1976) pointed adequately handle strong groups. would be suited nor the purely decentralized approach situations. is is experiments on centralized and decentralized problem- decentralized team structure might be The routinized and predictable, because coordination and control by plan and schedule solving behavior, found that decentralized groups take many communications what open communication channels, such on time (Mantei 1981). a third alternative for modeling effective I/S planning is characterized by both strong managerial control and Strong team-member control software development practice because tasks in I/S is necessary planning and design are in modern difficult in nature; strong managerial control must be managed in is necessary because diverse, often competing, goal sets order to produce products as rapidly as possible to meet an organizational set of goals. We at the believe that both managerial control and same In general, past studies of small time. view of control in team-member group behavior assume a zero-sum design teams (Cartwright and Zander 1968, zero-sum assumption, an increase accompanied by a reduction in in the the McGrath control exercised by control control can be increased exercised 1984). team members must be by the manager. assumption, these past studies have indicated which design team structure terms of task contingency. in Given the is With more this effective This perspective suggests that a Weinberg group would function well in difficult I/S development projects that are not time-constrained and/or have an existing set of shared goals. involves users from the literature many In a software project with a tight deadline that functional areas, these assumptions are often incorrect. shows a major pattern of weakness in relation to Further, these assumptions. Although a great deal has been learned about the potential influence of the leader's behavior, leader behavior alone accounts for only a small portion of performance variance in most empirical studies (Kerr 1977). In a view that is contrary to a zero-sum perspective we maintain that strong managerial control can be achieved without sacrificing team-member control. Tannenbaum (1968) claimed that both the manager and the employee can increase their influence together without a negative impact upon one another. He also reported that in his comparative analysis of several different organizations, performance was correlated with the sum (1982), of the managers' control and the subordinates' control. Clegg (1981), and Hofstede (1967) same time. Conceptually many al. arguments consistent with the provided hypothesis that strong managerial control and strong Similarly, Bartolke et team-member control can exist at the scholars recognize this fact, but research has tended to focus on either a bivariate relationship between managerial control and performance or between team-member control and performance (Ouchi 1979, 1977; Eisenhardt 1985, Mills 1983, Peterson 1984). form of control is Empirical studies have also tended to emphasize that only one desirable at a time depending on the contingency of tasks (Ouchi 1979, 1977; Kerr and Jermier 1978, Jermier and Berkes 1979). Tannenbaum (1968) claimed that a high degree of control by the manager necessary for the efficient administration of an organization and, at the degree of team-member control loyalty. is same is time, a high also necessary to foster identification, motivation, and Lickert (1961) has also suggested the importance of a high level of mutual influence within teams as the basis for effective coordination of organizational activity as well as for the integration of the goals of individual These conditions, leading persons at PI: all levels, the members and of the organization. to effective performance, entail significant control exercised by manager as well as all of the team members. Thus, we propose: Increased levels of both managerial control and team-member significant positive effect on the performance of design teams. control have a In the following sections specific variables into 2.1. we develop the theoretical basis for the inclusion of our model. Managerial Control Managerial control refers to the manager's attempts to influence employees behave in project manager includes Control-oriented behavior for an accordance with organizational goals. defining and documenting the team members; functional analysis and coding tasks to work establishing to to I/S be done; assigning performance guidelines through task feedback; comparing actual performance to performance standards; and initiating corrective action as necessary (Katz and Lerman 1985). These types of managerial behavior are managerial control (Flamholtz Flamholtz et al. et al. consistent with a range of research on 1985, Jermier and Berkes 1979, Schriesheim (1985) claimed that the core control system measurement system, feedback, and is composed of a reward system. Both the Jermier 1978). planning, a and Berkes (1979) and the Schriesheim (1978) studies claimed that a leader's behavior can be viewed control mechanism which encourages employees the organization to as a behave consistently with the goals of and discourages them from doing otherwise. Recent control theories (Eisenhardt 1985, Ouchi 1979 and 1977, Ouchi and Maguire 1975, Peterson 1984) claimed that managerial control can be established by either behavior-based control or outcome-based control. According to control refers to the extent that the this view, behavior-based manager monitors and evaluates team members' behavior in order to them. assist In contrast, outcome-based control is the degree to which the manager monitors and evaluates only the outcome produced by the team , members. Leadership behavior studies (House and Dessler 1974, Howell and Dorfman 1981, House and Mitchell 1974, Jermier and Berkes 1979, Schriesheim 1978) provide a robust characterization of managerial behavior control. dimensions of managerial behavior control: expectations of subordinates specific tasks, methods. and work, (2) work assignment: in their we have adopted manager's behavior control To (1) role clarification: in I/S identified three clarifying management assigning subordinates to enforcing rules, procedures, and work (3) procedure specification: In this paper These studies have these three dimensions to characterize the planning and design projects. characterize the manager's outcome control, we have adopted perspectives from communication and cybernetics theories (Campion and Lord 1982, Flamholtz Sorensen and Franks 1972). According to these theories, outcome feedback relevant dimension of outcome-based control. be operationalized design activities. in The is 1985, the most for I/S design teams can terms of team goal achievement and the achievement of interim latter reflects a goals into a series of milestones. control (Katz and Outcome feedback et al. Lerman common practice in 1/S design of These milestones then provide a 1985, Keider 1984). 8 decomposing team basis for outcome 2.2 Team-Member Control Recently several researchers have pointed out that managerial control accounts for only a small portion of the criterion variance, such as in Two most empirical studies of leadership. findings (Howell performance and job explanations have been provided for these and Dorfman 1981, Kerr 1977, Kerr and Jermier 1978). noted, managerial control is satisfaction, First, as we have exercised through the manager's intervention process. There- fore, the actions or style of control of a given manager might be perceived differently by different employees. Second and more important, most empirical studies ignore team member's which may be acting self-control, in such a way as to prevent or neutralize the manager's influence. In this self-control process a person faced with response alternatives chooses what Mahoney 1974, otherwise would be regarded as a low-probability response (Thoresen and pp. 12). Self-control is differentiated from freedom or laissez faire in that it is related to organizational effectiveness. Unlike managerial control which process, team-member One common de Ven cannot et exercised through the manager's intervention control refers to self-control or self-management by perspective found in the literature, aL (1976), adequately procedures. is is that self-control measure is Slocum and Sims (1980) and is Van be implemented when the organization performance behavioral In other words, self-control any other choice. likely to i.e., team members. resorted to In contrast, a second view such as or standardize transformation when management does not have Manz and Sims (1980) and Mills implemented by the team members' own and that managerial control and team-member control can be operative at the same time. As (1983), is that self-control can be discussed above, we will investigate the the context of I/S design. We extent to which this second view define team member behavior used to define managerial control. that guides the member 3.0 team towards achievement of organizational is applicable is control in terms of the The key notion will in same types of that the control behavior goals is exercised by a team rather than a manager. RESEARCH DESIGN This section describes the design and data collection of the to explore our hypotheses concerning control in I/S planning Our plane analysis as of observation is that of the teams. As conducted and design teams. such, an appropriate research method (Seidler 1974). As field study we chose key Phillips informant and Bagozzi (1981) have noted, the measurement of team-level properties has often entailed the use of a key informant method. on a social The key informant method setting by is a technique for collecting information interviewing (or surveying) a selected Although the use of key informants has traditionally participants. been associated with methodology (Lofland 1971), several researchers (Seidler 1974, Phillips number of Silk and Bagozzi 1981) have used key informant methodology qualitative and Kalwani 1982, in conjunction with procedures for collecting survey data to obtain quantifiable measures on organizational characteristics. 10 In these situations, survey respondents provide information at aggregate the or role of key assume the collective unit of analysis informants and team or (e.g., organizational properties), rather than reporting personal feelings, opinions, and behaviors we In our research, (Campbell 1955). utilized key informants selected from the team to project manager, I/S designer, and report on the control behaviors of three key roles: domain representatives (see Table At 1). least one informant from each surveyed, enabling us to have two informants per role plus a self report (e.g., a was role manager informing on the managerial role.) We administered questionnaires to 432 individuals We organizations. for designers, used two basic types of instruments: domain representatives, questionnaire for stakeholders, That covered the design process. I/S (i.e., The team process questionnaire for designers position an obvious We it Teams : The project size 10 first and performance a type of questionnaire surveyed control structures with questions on the anchored the questions to the role being assessed. for project manager and the team process questionnaire slight changes in wording to in reflect their order to avoid bias. In choosing design team which had worked together 6 months). managers, used stakeholders to assess team performance common method Selection of Design project is, and domain representatives had on the team. project in a team process questionnaire non team members). The We planning and design process. and 48 design teams in teams we limited group inclusion to a for a significant period of time (approximately was controlled by number of 11 individuals (5-10) and duration Project manager Designers The person : who manages the focal project. Professionals whose expertise and duties are primarily m the area of l/S technology, system development, programming, etc. for the focal : project. Domain representatives Stakeholders: : Professionals whose primary expertise and duties are in the function/business of the customer/user. They should be team members of the focal project and often are customer representatives. Professionals who are not formal members of the focal project but are affected by the output of the team or can affect the performance of the team. Table 1 Role Definitions (approximately 12 months). A teams were working on designing business applications. All survey of design methodologies and geneous and standard approach to CASE I/S design. technology usage suggested a These guidelines helped fairly homo- to ensure that task complexity was similar. Selection of Respondents and Informants designers, For each design team one project manager, two : and two domain representatives, plus four stakeholders were surveyed. our design teams were composed of 5 to 10 people, individuals on the team to enabled a high percentage of be surveyed. Questionnaire Administration directly to the project this The survey process involved sending : the questionnaires managers and following up with them as appropriate. questionnaires were mailed directly back to us to ensure confidentiality. procedures were described Since in the Completed Confidentiality questionnaire in order to increase response rate and increase the reliability of reporting. Participation of individuals was voluntary and so noted in the questionnaire. Altogether 432 individuals asked to participate 72% for role of the total some teams is shown in in the study. number to about Table in 48 I/S Among planning design teams for 10 companies were these, 310 usable replies were received, or about of questionnaires sent out. 44% for the lowest teams. 2. 13 The response rate ranged from 100% The breakdown of response rate by Since each team has only one project manager, the response rate for the project manager was 100% Since response at the individual level was voluntary . individual could return a blank (i.e., an form and no follow up would occur), the response rate did vary. Of the 48 teams that received questionnaires, 41 returned at least the following: one project manager, one designer, one domain representative and two stakeholders. These 41 teams provide an informant from each for role and therefore serve as the database our analysis. 4.0 MEASUREMENT MODEL This section discusses the measurement model used in this study. The model provides the relationships between the measured variables and the theoretical constructs we chose 4.1 to study. Measurement Model As Phillips and Bagozzi (1981), Huber and Power (1985) and others have noted, key informant analysis is illustrated own their by for Control Silk is subject to method and Kalwani (1982). role's influence with respect to a bias. One critical bias relevant to this They found informants tended buying process. Similarly, work to exaggerate Henderson (1988) Some teams initially contacted chose not to participate due to the implied workload. These teams are not included in the response rate and may be a source of selection bias. ^ 14 found evidence of systemic role bias for the analysis self-report reporting on user influence in I/S design. discussed by For details on other Huber and Power tactics the reader in Thus, we differentiated order to test for a self of managerial control and team members' control by both team members and the project manager report bias. 4.1.1. in used to address the range of informant bias is referred to Lee (1989). Managerial Control As discussed, and outcome-based we measured two types of managerial control: behavior-based control Behavior control was broken down into three categories, control. based on leadership behavior theory (Jermier and Berkes 1979, Kerr 1977, Kerr and Jermier 1978, Schriesheim 1978, Schriesheim clarification et al. 1976). (ROLE), work assignment (ASSI), and procedure These categories are specification role (PROC). Each category was assessed by asking informants to respond to multiple items using seven point Lickert-type scales. These items are included in Appendix 1 and are denoted by the acronyms ROLEl, ROLE2, ASSIl, ASSI2, PROCl, PR0C2. The two items for outcomebased control are denoted by of the item scores are also OUTPl shown in and 0UTP2. Appendix The means and standard deviations 1. Internal Consistency of Operationalizations To assess the internal consistency of these items, all 184 responses from the project managers, designers and domain representatives were tested for 15 reliability using Cronbach alpha. A factor analysis control dimensions were was performed distinct. The Cronbach alpha Reliability: 0.858, 0.786, 0.696, and coefficients for State Leadership Questionnaire (Schriesheim 1978). role clarification items PROC were to those of Even though the Ohio reliability for was achieved". in two factors This analysis supports the proposition that there are two dimensions of managerial control. and level were similar factor analysis using a varimax rotation resulted accounting for 65.5% of the variance. control items ASSI, and items were based on the Ohio State reliability coefficients procedure specification was low, the minimum A OUTP, ROLE, Our 0.534, respectively. Leadership Questionnaire and our Factor Analysis: outcome control and behavior to ensure that the However, the combine to results suggest that the form one factor, outcome and work assignment and procedure specification items form the second. One reason why the role clarification items covary with found by examining the contents of the items. level of performance that other, the project performance". (Schriesheim manager is In one, the project control can be manager "explains the expected of our project team members' work" and "lets our project team members know what is in the considered good These items were taken from work on leadership behavior theory 1978) in which role clarification was theorized to be one part of the manager's behavior-based control. However, " outcome The zero-order correlations for all in the context of I/S design, these role clarifi- items are available from the authors upon request. 16 cation items reflect what should be expected from closely related to manager's desired outcome. team members. That they are is, Interviews with teams supports the view that the role clarification items were interpreted as outcome control indicators most often linked to definition of milestones for the project. Based on control. we this analysis, In the following analysis include role clarification items as part of outcome-based outcome-based control and behavior-based control are an average for the respective items and are denoted as OUTPM and BEHAM. Convergent and Discriminant Validity Since there were three types of informants, we can examine discriminant validity of these measures using a multitrait-multimethod in which the informant type Bagozzi 1981). control, where Table 3 OUTPM is is the and viewed as a method MTMM BEHAM (Silk the convergent and (MTMM) and Kalwani 1982, approach Phillips and matrix (Campbell and Fiske 1959) for managerial are viewed as traits, and the project manager informant ("M"), designer informant ("D"), and domain representative informant ("U") as different methods. Convergent correlation validity was not achieved was 0.144 (p>0.10). correlation informants the OUTPM variable. The smallest This analysis suggests that there was an insignificant correlation between designer informants and representative in manager informants, and between domain and manager informants. However, there was a strong between the domain representative and the designer informants. 17 Convergent correlation validity between significantly different the for behavior control responses from zero. The from (BEHAM) informants strong. is The was 0.320 (p<0.01), which and the domain representative informants. There are two possible reasons why we did not achieve convergent First, since the informants, there might be self-report validity for weakest correlations were due to the manager method bias. However, this does not explain why Second and more probably, convergent validity was achieved for behavior control. may be is strongest correlation can be found again between the responses by the designer informants outcome-based control. smallest this explained by differences between managerial control given and managerial control Many received (Ouchi 1977). relationships between researchers have assumed control social actors. Wrong control need not be exercised to exist. is realized only in dyadic (1968), however, asserted that managerial Even though control is usually defined as the a distinct difference between the capacity to influence others, he emphasized that there is capacity to control and the actual practice of control. This distinction was noted by (1984) and Provan members et al. (1980), as well. Cobb Since designers and domain representatives are of the I/S planning and design team and subordinates of the project manager, their perceptions of managerial control can be viewed as managerial control received. The project manager's perception of his managerial control can be viewed as managerial control given. Outcome-based control might not be perceived by the team members they are performing well from the perspective of the project manager. project manager might closely monitor the outcomes of 19 his if In this case the team members and may claim that he is exercising strong outcome-based control, but his team members may not perceive strong managerial control. This reasoning can explain why we have achieved convergent validity for BEHAM. Since behavior control involves manager's behavioral interaction with team members, the distinction between control given and received From the above we discussion is not strong. separate managerial outcome control given (managerial outcome control perceived by the manager) from managerial outcome control received (managerial outcome control perceived by the team members). outcome control received the is assessed by two categories of key informants, the designer and domain representative informants, we can approach. validity. No an test discriminant validity using violations occurred in this case indicating convergent MTMM and discriminant However, managerial outcome control given can not be tested (we have only one informant); hence we must recognize the potential weakness of this measure. In the case of behavior based control, convergent achieved. Since managerial Applying the same logic of the above was tative informants. an analogous manner In was argument to behavior control, the between responses by designer and domain represen- strongest correlation also found validity and discriminant to the outcome control dimension, we separate behavioral control into behavioral control received (measured by the average of designer and domain representative responses) and behavioral control given (measured by the managers response). TTiis separation has the benefit of enabling subsequent analysis to 20 directly examine the relationship between a project manager's perceptions of control and team performance. In summary, we have convergent and discriminant validity for control received and managerial behavior control received. behavior control given show internal reliability, measured based on only a we can we examine following section, 4.1.2. is Managerial outcome and single informant. While the items not assess the convergent or discriminant validity. the measured validity for team member In the control. Team-Member Control The existing literature Van de Ven et al. 1976, (Manz and Sims Tannenbaum by the roles of team members. into managerial outcome differentiating these quite differently. 1968) does not differentiate team members' control However, two types of in this study team-member control roles is that their tasks have The was divided rationale for been viewed by researchers Designers are expected to build an information system. on the other hand, are expected Therefore, in our study representative control. we used (BEHAD), domain Domain to supply relevant business knowledge. separate items to measure designer control and domain For each dimensions, giving four variables: control Slocum and Sims 1980, designer control and domain representative control. two types: representatives, 1980, Mills 1983, role, team member control was measured by two designer outcome control representative 21 (OUTPD), designer behavior outcome control (OUTPU), and domain representative behavior control in Appendix 2 with their (BEHAU). The means and standard items used to measure them are shown deviations. Internal Consistency' of the Operationalizations Since most of the empirical literature does not differentiate the type of team members, we need to test whether our differentiation is justified. This can be tested by examining the correlations between designer control and domain representative control the item level; that is, whether items intended to at measure domain representative control are significantly different from items intended to measure designer control. Table 4 shows the zero order correlations between items used Scanning this matrix shows that all in the questionnaire. the correlations between the items for measuring designer outcome control and domain representative outcome control are significant (p<0.001). In addition, all the correlations between the items for measuring designer behavior control and domain representative behavior control are significant (p<0.001). On the other hand, the items measuring behavior-based control of any role were not correlated with items measuring outcome-based control at the p=0.01 level. Therefore, we can conclude that items used in the questionnaire can differentiate the theoretical constructs, outcome-based control and behavior-based control there is no difference between designer control and domain representative further clarify our argument Factor analysis: we performed but that control. To a factor analysis. In order to verify that the data reflects the outcome and behavior -- control, a factor analysis -)"> two primary dimensions of was performed using varimax rotation. Two 59% factors resulted accounting for interpretation of outcome and behavior of the variance and are consistent with the control. Factor loadings for behavior control items ranged from .696 to .786 while those for outcome ranged from .738 to .801. These results support the premise that we measured two distinct dimensions. Since Reliability: we two dimensions of team-member control, we average find only designer outcome-based control and domain representative outcome control (denoted by OUTPT), and BEHAT). designer behavior and domain representative behavior control (denoted by The Cronbach alphas respectively. We can also test to BEHAT for determine if and OUTPT were 0.7440 and 0.7934, any given item does not merit inclusion by Results show examining corrected item-total correlations (Carmines and Zeller 1983). reliability could not be increased by removing any single item. This reinforces our decision In sum, to aggregate the items. control items, the items for if we aggregate team-member outcome designer and domain representative control and the items for team-member behavior control show good internal consistency of the operationalizations. Convergent and Discriminant Validity We MTMM can again use the fact that analysis and assess convergent team-member control, informant, designer, i.e., we have where OUTPT and multiple informant types to perform a validity. BEHAT Table 5 is the are viewed as traits, domain representative and project manager, 24 MTMM matrix for and the type ot as different methods. that neither convergent validity nor discriminant validity Note of the manager informants' responses. If we eliminate the could be related representatives. be strong to the working relationship between responsibility for input between these members. and design of the system. explanation of and domain roles has prime However, project managers are often et al. 1979). and domain representatives have more opportunity with the project manager, they would be more accurate sources to To the extent work together than of information concerning control behavior. However, since we cannot test the above argument with survey data (although this idea was confirmed through interviews), (1) designers Each of these concerned with a wider range of management issues (Allen team member One Since designers and domain representatives work as a team, there should interactions that designers achieved because manager informants' responses, both convergent validity and discriminant validity would be achieved. this is team-member outcome we separate scales to assess team-member control as perceived by the team members, (2) control: team-member behavior control as perceived by the team members, (3) team-member outcome control as perceived by the manager, and (4) manager. In effect, we team-member behavior control as perceived by the separate the perception of the project manager in a manner similar to that for the managerial control variables. each variable have good reliability, As a result, the underlying measurements for and for the team members we show convergence across multiple informant types. 25 4.2 Measures of Performance We used subjective measures to assess performance because of the substantial problems involved 1988). in using objective measures for this task (Kemerer 1989, Henderson Since our study involved teams from multiple organizations, use of internal accounting systems data to measure performance was inconsistent. Each organization and teams within organizations not only varied terms of the integrity of in terms of data collected but varied widely data collection process. Thus, expert judgement this in became our best source of performance data. The performance of the design teams, in terms of their efficiency (EFFE), and elapsed time (TIME) was assessed by non-team stakeholders. effectiveness Stakeholders are individuals who were affected by the output of the management responsibility for The number of stakeholders from two to five Appendix performance. its who shows the (I/S and members of the project but were directly team's performance. directly affect the line) who sponsored the project or had successful completion, implementation, or ongoing usage. filled with an average of 3 not formal team or could Examples include senior executives out the performance questionnaire per team ranged 2.6. questionnaire measures used for each dimension of Except for one item, the measures for EFFI and EFFE were the same those developed and validated by Henderson (1988). Note however that measures: (EFFI), "the team's adherence to budgets" the efficiency dimension. and "the team's adherence However, there were several instances 26 in we as separated two to schedules" for which the question concerning the team's adherence to budgets was not answered. stakeholders suggested that Therefore, we dropped reason for might not have known specific details of the 0.750, 0.723, and 0.736, respectively, suggest an adequate These values are lower than those obtained by Henderson (1988). this difference may be the increased While somewhat low, these study. budgets. the budget item to improve the reliability of our measure. Cronbach Alphas of of reliability. Follow-up interviews number of reliabilities are still level One organizations included in this within an acceptable range (Nunnally 1967). To be sure that our 41 teams represent independent sample points, i.e., variance in the dimensions of performance arises from genuine differences design teams rather than the organizations, that the we ran three ANOVAs that the among to test the hypothesis performance variables could be explained by an organization variable. multivariate Fs were 1.075, 1.881, 1.581 for efficiency, effectiveness, that there were no and time, significant differences across the organizations (P>0.10). variations in performance variables within-organization variations, among the The indicating Since the the organizations were not greater than the we have support for the general notion that variables varied due to differences across teams. In the follov^ng analysis, therefore, design teams are treated as individual sample points. 27 performance all 5.0 A.N EMPIRICAL TEST OF THE TEAM PROCESS MODEL The underKing theme and behasioral. this \iev,' is v,idely advocated this The [see it in theor\ to the context of team members Performance IS planning and measured in outcome design. (OLTPM;. team-member and domain This research attempts to are managerial behavior control control (OLTPT).-^ (designers is (1968) for a review]. we measured managerial outcome control team-member outcome that control behavior, both organizational control literature and has strong empirical Tannenbaum control variables is performance of IS design teams. As discussed, positively related to the is evidence to support extend for our proposition behavior control (BEHAM), (BEHAT) and Each control variable has been assessed by representatives) and the project manager. terms of efficiency (EFFI), effectiveness (EFFE) and elapsed time (TIME). We ry-elve test hypotheses that reflect the relationships between each of these variables and our three performance variables. no relationship DUBERAM addition. and The results of these tests are DUOLTPT DL'BERAT was (p<0.05j. ^ In exists. In each case, the null h\pothesis were related to summarized significantly related to all EFFE The managers' assessment (p<0.05) and in Tables 6 and that the project Both performance variables. In DUOLTPM was related to EFFI of the control variables was not significantly related is preceded by DU or M, where DU indicates indicates and the domain representative assessed the variable, and our notation each control variable that the designer 7. that is manager assessed M the variable. 28 to performance variables. any assessment of ways. First, own his is discussed earlier, the results using a manager's (MBEHAM and MOUTPM) can be interpreted may be these measures are self-reports, so position bias natural conclusion he control As exercising. different things: is that a manager's own self-report overestimates in operating: how much two the influence Second, the two measures of managerial control may actually measure managerial control as assessed by the project manager manager and managerial control control given by the project as assessed by may refer to team members can be viewed as control received, as Ouchi (1977) has claimed. To assess the significant difference argument, first between team members (t(OUT) = was at least we argued 4.12, plausible. that, If ran two paired self-reports by the project reject the hypothesis that there Earlier we is p < .01, t(BEH) = t-tests. 4.01, p < .01). show a Thus, we cannot a significant self-report bias. based on the measurement model, the second intei-pretation we adopt this view, managerial control received performance variables. managerial control received may be more important than managerial individuals with different roles is more In the case of the I/S design teams, planning and design entails complex tasks which require explicit results managers and non-self-reports by significantly correlated with I/S The control given. much interaction Since among (Guinan and Bostrom 1986), team members may need reinforcement and coaching from the manager. have a manifest impact on team members' perception. 31 Thus, managerial control should We now analyze the results using assessments behavior-based manager's control and made by team members. When outcome-based behavior-based control was more significantly related to The designers and the domain representatives have control all different were compared, These data suggest the important providing explicit direction outcome-based The skills, role perceptions, relationships strongly related to among performance and this interaction by all the team members' control variables control variables, outcome-based control was of the performance variables than behavior-based control behavior control refers to self-control in as providing feedback on performance. strong collegial control, i.e., . executing a team member's whereas team-member outcome control can be viewed tasks, outcome design process, which cannot be done using only Among team members' Team-member own the managing final and control. were the opposite. more in role of the the performance variables. mental schemata which must be combined through their interactions into a (Boland 1978). the as collegial control, such These data indicate those teams that exhibit exerted peer pressure to achieve outcome commitments, were significantly higher performing. A regression analysis using the four variables (as assessed by team provides support for proposition PI and for the claim that "* We assessed team-member control I/S members) design tasks require both separately for both designers and domain representatives, and found that self-report and non-self-report measures were in agreement in this study (for the measurement model see the previous section). Thus, we can exclude the possibilit)' of self-bias in the aggregated measures, which include both self-report and non-self-report measures of the team member 32 control variables. (DUBEHAM) be to processes control efficient and and team outcome control Managerial effective. (DUOUTPT) behavioral 13% There are seem likely. First, as We inadequate. Second, tv^'o TIME However, these of the variance for the time dimension. discussed in Section 5.2, our subjective measures for somewhat among compared to those of past empirical work. their and ours studies differently. determining ... is difficult They measured policy?" team-members' control to team-members our Tannenbaum and (the designers A related to organizational effectiveness. is since they comparison strict operationalize and the the constructs the leader's control by asking survey respondents actually have in determining assessed team member's control by asking "how subunit. in projects for scarce resources. "how much influence does your leader in be (1968) tested the hypothesis that the degree of control, exercised both by representatives), between TIME may might be better explained by other theoretical variables not included leaders (in our case the project manager) and by domain variable that used only two items which had not been pretested by other studies. findings can be his colleagues TIME interpretations of our findings concerning the study, e.g., competitive pressure Our 44% and 38% together explain of variations in the efficiency and effectiveness of the I/S design teams. variables only explain control much influence They then added the values measure the total amount of Despite these differences, our results are 33 in ... policy", and they do team members have for leader's control and control in the organization or general agreement with theirs. In Table 7 them showed correlations we generated correlations with significant with several measures for the total performance (DUBEHAM + DUOUTPT). We of control variables the are can interpret found this in can control reflected in our variable, is show Team-member our study outcome is its impact on control refers to the et al. DUBEHAM, performance amount of most strongly reflected the last strongest row of Table total 7: amount (1968) defined managerial control team members' behavior. This since the managerial behavioral through team-members' control exercised by in the variable All of DUOUTPT, behavior. team members and i.e., a in team member's control. Tannenbaum et al. (1968) found that the correlation between the total control and effectiveness ranged from 0.29 to 0.45. effectiveness -I- The combined variable as the as the extent of a manager's influence in determining the relationship control. performance variables. Tannenbaum in a didactic relationship. amount of and the DUOUTPT) total amount of control amount of In our case, the correlation in the didactic relationship between (DUBEHAM was 0.569 (p<0.01)^. Given the support we found for the hypothesis that control variables can explain performance, a more detailed analysis of the pattern of performance between managerial control and and team-member control DUOUTPT, we Tannenbaum is provided. Using the median values for DUBEH.AM split the 41 I/S planning and design teams into four subgroups as et al. operationalized performance variables by effectiveness items similar to our items for the effectiveness variable. 34 shown in Figure 1. We EFFE, and TIME) then examined the average values of performance variables (EFFI, in the appropriate As would be expected, cell. the degree of managerial behavior control and team-member outcome control are clearly reflected performance variables. team-member outcome the best, and worst. First, when both in the high managerial behavior control and high control were operating (Cell One), the performance variables were when both were low (Cell Four), all the performance variables were the Second, when either low managerial control or low team-member control was achieved, the performance variables were not greater than those in Cell One, but were better than those in Cell Four. Although we can compare neither Cell nor to Cell Three because of the small sample Cell Four, using the t-statistic. were The high managerial behavior control and high compared team-member outcome 6.0 to teams (n=5), we can compare values of the t-statistics for EFFI, 3.93, 3.28, 1.72 (all significant at p=0.01), performers size with low One to Cell Cell Two One and EFFE, and TIME which supports the claim that teams with team-member outcome managerial behavior control are better control and low control. CONCLUSION This result supports the proposition that a control theory perspective can be used to explain the relative performance of I/S concepts of behavior and outcome control design teams. in the context Our research of I/S design. these to notions of control behavior to both project managers and 35 operationalizes the Further, we apply team members. Low managerial behavior control High managerial behavior control High team outcome member control EFFI EFFE TIME 5.364 5.427 4.913 ^^ " = Cell Low team member outcome control 1 The That results are consistent with previous research in control theory. outcome and behavioral control can co-exist and the greater the degree of is, both explicit control behavior perceived by the team, the better the team's performance. From an Our perspective, the findings are particularly interesting. 1/S results suggest high performance I/S design team have managers that exhibit behavioral control and team members that exhibit outcome control. The former highlights the importance Our interviews reinforced this of knowledge-based leadership in an I/S design context. Behavioral control reflected the finding with an important insight. to aid the design process. The knowledge needed restricted to I/S design practice. of the manager for this assistance, however, is not Rather, the design process requires an understanding of the business (from both a technical and a social perspective). Our interviews suggested managers often brought domain knowledge to bear on work assignments, effective procedure clarification and so on. training ability and selection To the extent that this single study criteria for I/S project managers must is validated, the reflect their ability to provide behavioral-based control. The in an I/S significance of collegial design context. easily distribute outcome based control For example, emerging performance data on a CASE real time basis to performance is important technology provides a means to throughout the team. These results team performance by enabling a peer-to-peer outcome control structure, rather than the project leader hierarchy that is suggest such practice could help to improve overall traditionally used. 37 Of course, this single study does not provide the basis to conclude definitively that managerial behavior control combined with team member outcome control is preferred. Additional research must consider factors such as task contingencies, experience of team members or technology assisted methodologies. Yet, this does provide one theoretically grounded lens with which the systematic study of high performing I/S planning and design teams can be conducted. Acknowledgements This research was funded by the and the Center Management for Information 38 in the 1990s project Systems Research. REFERENCES M.L. Tushman and D.M.S. Lee, 'Technology Transfer as a Function of Position the Spectrum from Research through Development," Academy of Management Journal, Allen, T.J., in 22, 4 (1979), 694-708. Bakos, J.Y. and M.E. Treacy, "Information Technology and Corporate Strategy: Research Perspective," Bartolke, K., Participation W. MIS A Quarterly, 10, 2 (1986), 107-119. Eschweiler, Flechsenberger, D. and AS. Tannenbaum, "Workers' as Perceived by and the Distribution of Control Members of Ten German Companies," Administrative Science Quarterly, 27 (1982), 380-397. Bavelas, A., "Communication Patterns in Task-Oriented Groups," Journal of Acoustical Society America, 22 (1950), 725-730. Boland, R.J., Jr., 'The Process and Product of System Design," Management Science, 24, No. 9 (1978), 887-898. D.N. and D.W. Fiske, "Convergent and Discriminant Validation by the Multitrait-Multimethod Matrix," Psychological Bulletin, 56, No. 2 (1959), 81-105. Campbell, Campbell, D.T., "The Informant 60 (1955), 546-553. in Quantitative Research," American Journal of Sociology, Campion, M.A. and R.G. Lord, "A Control Systems Conceptualization of the Goal-Setting and Changing Process," Organizational Behavior and Human Performance, 30 (1982), 265-287. Carmines, E.G. and R.A. Zeller, Beverly Hills, CA, "Reliability and Validity Assessment," Sage Publications, 1983. Cartwright, D. and A. Zander, "Power and Influence in Groups: Introduction," Cartwright and A. Zander (Eds.), Group Dynamics: Research and Tfteory, 3rd. Row, New York, 1968, 215-235. ed.. in D. Harper & Cash, J.I. and B.R. Konsynski, "IS Redraws Competitive Boundaries," Harvard Business Review, 63, 2 (1985), 134-'l42. Clegg, S, "Organization and Control," Administrative Science Quarterly, 26, No. 4 (1981), 545-562. 39 Cobb, A.T., "An Episodic Model of Power: Rtseavch," Academy of Management De Review, 9, Toward an Integration of Theory and No. 3 (1984), 482-493. Brabander, B. and G. Thiers, "Successful Information System Development in Relation Which Affect Effective Communication between MIS-Users and EDP-Specialists," Management Science, 30, No. 2 (1984), 137-155. to Situational Factors Eisenhardt, K.M., "Control: 31, Organizational Economic Approaches," Management Science, No. 2 (1985), 134-149. 'Toward an Integrative Framework of Organizational Control," Accounting Organizations and Society, 10, No. 1 (1985), 35-50. Flamholtz, T.K. Das and A.S. Tsui, and R.P. Bostrom, "Development of Computer-Based Information Systems: Communication Framework," Data Base, 18 (1986), 3-16. Guinan, A E.G., P.J. Henderson, J.C, "Involvement as a Predictor of Performance in I/S Planning and Design." Center for Information Systems Research Working Paper #175, M.I.T., Cambridge, MA, August 1988. Hofstede, G.H., The Games of Budget Control, Von Gorcum, Assesn, Germany, 1967. House, R.J. and G. Dessler, 'The Path-Goal Theory of Leadership: Some post hoc and a priori Tests," in J.G. Hunt and L.L. Larson (Eds.) Contingency Approaches to Leadership, Southern Illinois University, Carbondale, IL, 1974, 29-55. and T.R. Mitchell, "Path-Goal Theory of Leadership," Journal of Contemporary Business, Howell, 3, No. 4 (1974), 81-97. J.P. and P.W. Dorfman, "Substitutes Academy of Management for Leadership: Test of a Construct," Journal, 24, No. 4 (1981), 714-728. Huber, G.P. and D.J. Power, "Retrospective Reports of Strategic-Level Managers: Guidelines for Increasing Their Accuracy," Strategic Management Journal, 6 (1985), 171180. and G.P. Learmonth, "The Information System Communications of the ACM, 27, 12 (1984), 1193-1201. Ives, B. as Jermier, J.M. and L.J. Berkes, "Leader Behavior in a Police Closer Look at the a Competitive Command Weapon." Bureaucracy: Quasi-Military Model," Administrative Science Quarterly, 24, No. (1979), 1-23. 40 A 1 Work Breakdown Katz, B. and N. Lennan, "Building a Three Dimensional Data Base, Summer 1985, 14-26. Keider, S.P., "Why Systems Development Management, Kemerer, 1, C.F., Structure," Journal of Information Systems Projects Fail?" No. 3 (1984), 33-38. "An Agenda Software Engineering for Research (CASE) Tool in the Managerial Evaluation of Impacts," Computer-Aided Proceedings of the Twenty-Second Annual Hawaii International Conference on System Science, Kailua-Kona, Hawaii, January 3-6, 1989, pp. 219-228. Kerr, "Substitutes for Leadership: S., and Administrative Organization and J.M. Jermier, Some Implications for Organizational Design," Sciences, 8, No. 1 (1977), 135-146. "Substitutes for Leadership: Their Human Organizational Behavior and Meaning and Measurement," Performance, 22, No. 3 (1978), 375-403. "Some Effects of Certain Communication Patterns on Group Performance," Journal of Abnormal and Social Psychology, 46 (1951), 38-50. Leavitt, H.J., Lee, S., "Managing I/S Planning and Design Teams: A Control Theory Perspective," MA Doctoral Dissertation, Sloan School of Management, M.I.T., Cambridge, Lickert, R., New Lofland, Analyzing Social J., Patterns of Management, McGraw-Hill, Settings, John Wiley Mantei, M., 'The Effect of Programming Communication of the ACM, 24, & Team New Sons, 1989. York, 1961. New York, 1971. Structures on Programming Tasks," No. 3 (1981), 106-113. Manz, C.C. and H.P. Sims, Jr., "Self-Management as a Substitute for Leadership: A Social Learning Theory Perspective," Academy of Management Review, 5, No. 3 (1980), 361-367. March, J.G. and H.A. Simon, Organizations, John Wiley McGrath, J.E., & Sons, New York, 1958. Groups: Interaction and Performance, Prentice-Hall, Englewood Cliffs, NJ, 1984. Mills, H.D., "Chief Programmer Teams: 71-5108, IBM Principles and Procedures," Fed. Syst. Div., Gaithersburg, MD., 1971. 41 IBM Rep. FSC Mills, P.IC, ?Topenies," "Self-Management: Its Academy of Management Control and Relationship to Other Organizational Review, 8, No. 3 (1983), 445-453. Nunnally, J.C., Psychometric Theory, McGraw-Hill, New York, 1967. "A Conceptual Framework for the Design of Organization Control Mechanisms," Management Science, 25, No. 9 (1979), 833-848. W.G., Ouchi, "The Relationship between Organizational Structure and Organizational Control," Administrative Science Quarterly, 22, No. 1 (1977), 95-113. , and M.A. Maguire, "Organizational Control: Science Quarterly, 20, No. 4 (1975), 559-569. Functions," Administrative A New Competitive Weapon," Sloan Management Parsons, G.L., "Information Technology: Review, 25, Two (1983), 3-14. 1 Peterson, K.D., "Mechanisms of Administrative Control over Managers in Educational Ovganizaiions," Administrative Science Quarterly, 29, No. 4 (1984), 573-597. Phillips, L.W. and R.P. Bagozzi, "On Measuring Organizational Properties: Methodological Use of Key Informants," Working Paper, Stanford University, Graduate Issues in the School of Business, 1981. Provan, K.G., J.M. Beyer and C. Kruytbosch, "Environmental Linkages and Power Resource-Dependency Relations between Organizations," Administrative Science 25, No. 2 (1980), 200-225. Quarterly, Woodward, 'The Study of Managerial Control," in Industrial J. Woodward, J. (Ed.), Oxford University, London, Reeves, T.K. and Organizations: Behavior and Control, J. in England, 1970, 37-56. Robey, D. and D. Farrow, "User Involvement in Information System Development: Conflict Model and Empirical Test," Management Science, 28, 1 (1982), 73-85. Rockart, J.F. and M.S. Scott Morton, "Implications of Changes for Corporate Strategy," Interfaces, 14, 1 in A Information Technology (1984), 84-95. Salaway, G., "An Organizational Learning Approach to Information Systems Development." MIS Quarterly, 11, No. 2 (1987), 245-264. Schmitt, J.W. and K.A. Kozar, "Management's Role Failures: A Case Study," MIS Quarterly, 2, in Information System Development No. 2 (1978), 7-16. 42 New Leadership Behavior Expectancy Research Doctoral Dissertation, Instruments," College of Administrative and Science, Ohio State University, Columbus, OH., 1978. Schriesheim, C.A., "Development, Validation, and Application of ^ R.J. House and S. Kerr, "Leadership Initiating Structure: Discrepant Research Results and Some A Reconciliation of Empirical Tests," Organizational Behavior and Performance, 15, No. 2 (1976), 297-321. "On Using Informants: A Technique for Collecting Quantitative Data and Controlling Measurement Error in Organization Analysis," American Sociological Review, Seidler, J., 39 (1974), 816-831. A. and M. Kalwani, "Measuring Influence in Organizational Purchase Decisions," Journal of Marketing Research, 19 (1982), 165-181. Silk, Slocum, J.W., Jr. and H.P. Organization, and Job Design," Sims, "A Typology Jr., Human for Integrating Technology, Relations, 33, No. 3 (1980), 193-212. Sorensen, J.E. and D.D. Franks, "The Relative Contribution of Ability, Self-Esteem and Evaluative Feedback to Performance: Implications for Accounting Systems," Accounting Review, 47, (1972), 735-746. Tannenbaum, Thompson, A., Control in Organizations, J.D., Organizations in Action, McGraw-Hill, McGraw-Hill, New New York, 1968. York, 1967. Thoresen, E. and M. Mahoney, Behavioral Self-Control, Holt, Rinehart, & Winston, New York, 1974. Van de Ven, A.H., A.L. Delbecq and R. Koenig, "Determinants of Coordination within Organizations," American Sociological Review, 41, No. 2 (1976), 322-338. Weber, M., The Theory of Social and Economic Organization, Free, Glencoe, Weinberg, G., The Psychology of Computer Programming, York, 1971. Wrong, D.H., Power: Yourdon, E., Its Van Nostrand Modes IL, 1947. Reinhold, New Forms, Bases, and Uses, Basil Blackwell, Oxford, England, 1968. Managing the Structured Technique, Prentice-Hall, Englewood 1976. 43 Cliffs, NJ, Appendix 1. Managerial Control Questions How accurate do you think the following statements are in describing your project. My project manager in the XXXXX project manager for the XXXXX Appendix 2. Team-Member Control Enablers Questions Please indicate the extent to which you agree or disagree with each of the following statements. For work relating to the XXXXX project Appendix 3. Performance Questions The following questions ask you to compare the XXXXX project team to other teams. other comparable project teams you have served on or observed, rate on each of the following. In relation how does the XXXXX project to team M9i 018 Date Due Lib-26-67 3 9080 02239 1178