The relationship between user participation and system success:
a simultaneous contingency approach
Winston T. Lin
a,*, Benjamin B.M. Shao
a,b,1 aDepartment of Management Science and Systems, School of Management,State University of New York at Buffalo, Buffalo, NY 14260, USA bSchool of Accountancy and Information Management, College of Business,
Arizona State University, Main Campus, Tempe, AZ 85282, USA
Received 12 April 1999; accepted 23 September 1999
Abstract
The relationship between user participation and information system (IS) success has drawn attention from researchers for some time. It is assumed that strong participation of future users in the design of IS would lead to successful outcomes in terms of more IS usage, greater user acceptance, and increased user satisfaction. However, in spite of this, much of the empirical research so far has been unable to demonstrate its bene®ts. This paper examines the participation±success relationship in a broader context, where the effects of user participation and two other factors, user attitudes and user involvement, on system success occur simultaneously. Other contingency variables considered here are: system impact, system complexity, and development methodology. The theoretical framework and the associated hypotheses are empirically tested by a survey of 32 organizations. Empirical results corroborate the positive link between user participation and user satisfaction and provide evidence on the interplay between user attitudes and user involvement.#2000 Elsevier Science B.V. All rights reserved.
Keywords:User participation; System success; User satisfaction; User attitudes; User involvement; System impact; System complexity; Outsourcing; Contingency; Simultaneity
1. Introduction
User participation has long been regarded as an important factor to improve the chances of the success
in developing an information system (e.g.,
[6,31,36,68]). User participation refers to `the various design related behaviors and activities that the target users or their representatives perform in the systems development process' [5]. Through participation,
users of an information system (IS) can interact with system designers in the stages of planning, analysis, design, testing, and implementation and, hence, aid in many aspects of the system development process. A variety of development methodologies, such as code-velopment, participative design (PD) and joint appli-cation design (JAD), have been proposed to operationalize user participation [3,10].
There are a number of bene®ts which can be expected of such user participative behaviors. Ives and Olson [37] have pointed out that user participation in system development can enhance system quality through a more accurate and complete identi®cation of user information requirements [60,63], knowledge and expertise about the organization the system is
*Corresponding author. Tel.:1-716-645-3257;
fax:1-716-645-6117.
E-mail addresses: mgtfewtl@acsu.buffalo.edu (W.T. Lin), ben.shao@asu.edu (B.B.M. Shao)
1Tel.:1-480-727-6790; fax:1-480-965-8392.
intended to support [49], avoidance of unacceptable or unimportant system features, and a better user under-standing about the system. User participation is also believed to increase user acceptance of the system with a more realistic expectation about system cap-abilities [27], an opportunity for users and designers to resolve con¯icts about design issues [41], users' feel-ings of ownership toward the system, a decrease in user resistance to possible changes incurred by the system, and greater commitment from users [51]. In consequence, user participation has been extensively sought and encouraged by practitioners in developing IS [32].
In the meantime, researchers also have made great efforts to ®nd the empirical evidence of ef®cacy of user participation in improving system success. How-ever, the reported results have been mixed and frag-mented. In an earlier comprehensive review of 22 empirical studies conducted from 1959 to 1981, Ives and Olson were able to identify only eight (36%) that show a positive link between user participation and some measure of system success, including system quality [9,24], system usage [68], user behavior/atti-tudes [1,35], and information satisfaction [22,30,38]. The results from other studies are either inconclusive or contrary to expectations.
Cavaye [11] has reviewed 19 more recent studies published from 1982 to 1992, but also found only seven (37%) of them can substantiate a positive
participation±success relationship
[2,7,14,16,23,40,42]. Despite the fact that 10 more years of research have lapsed, the progress seems to be limited and slow. The ®ndings are still mixed and fragmented.
Researchers have identi®ed several possible causes of the ambiguity and contradiction. Poor research methods and the omission of important contextual factors are possible causes [53]. Also a simple direct causal relationship between user participation and system success may not be suf®cient. DeBrabander and Edstrom [13] suggested that the context in which the system is being developed has to be considered. Hence, the contingency approach has been widely utilized to address this issue (e.g., [17,43,48,50,54,64,65,69]). This is helpful in identi-fying factors that may alter the consequences of a process. However, most of these models are sequen-tial, with assumed direct causal relationships.
Furthermore, it is suggested that several confound-ing constructs have to be carefully de®ned to reduce the possibility of confusing results (e.g., [6,44]). Hwang and Thorn [34] meta-analyzed 25 studies and found `both user involvement and user participa-tion are bene®cial, but the magnitude of these bene®ts much depend on how involvement and its effects are de®ned'.
The purpose of this paper is to explore the relation-ship between user participation and system success in a broader context where all major effects on system success are simultaneouslyandjointly deter-mined. Due to the complexity of this scenario, we believe it is bene®cial and appropriate to study the relationship by means of a simultaneity and contin-gency approach.
2. Theoretical model and hypotheses
The termsuser participationanduser involvement
in virtually all MIS research were used interchange-ably in the past. To align research with the relevant concepts in such disciplines as psychology, marketing, and organizational behavior and, hence, to take advan-tage of their experience, Barki and Hartwick [6] felt is was necessary to differentiate between these two constructs, with user participation referring to `the behaviors and activities that the target users or their representatives perform in the systems development process,' and user involvement to `a subjective psy-chological state of the individual,' depending on the importance and personal relevance that users attach to a particular system or to IS in general. Fig. 1 presents our theoretical model.
2.1. Participation±success relationship
User participation in the IS development process has been regarded as a special case of participative decision making (PDM), which refers to group deci-sion making. Participation from users can be classi®ed into type and extent. The type may be consultative, representative, or consensus. The extent increases in degree from consultative to consensus [57]. Because the degree of participation is a more general concept than the type, it is used in this study to measure user participation.
System outcome, though dif®cult to be measured in economic terms, has a number of surrogates [6]. Garrity and Sanders [26] have suggested considering multiple perspectives. System use is frequently used to measure system success when system use is discre-tionary or voluntary [15]. However, user satisfaction is the most widely used measure for system success in the literature. In this study, user satisfaction is used as the surrogate for system success.
Signi®cant empirical results from PDM suggest a positive relationship between the degree of participa-tion in decision making and decision acceptance (e.g., [33,47,66]). This would suggest that there is a positive relationship between the degree of user participation and system success. Accordingly, it is hypothesized that:
Hypothesis 1. User satisfaction (X1) of using the information system increases with higher degree of user participation (X2) in the development process.
2.2. Simultaneous relationship: user participation, attitudes, and involvement
There is a difference between user involvement and user participation. User involvement is a subjective psychological state re¯ecting the importance and
personal relevance that a user attaches to a given system. At the same time, user involvement is different from user attitude, which has been de®ned as an affective or evaluative judgment toward some object or behavior, measured by a procedure that speci®es the individual on a bipolar scale (e.g., good or bad) [21]. Therefore, in IS development, user attitude re¯ects a psychological state that shows the user's feelings about a new IS.
Thus, user participation is conceptualized as a
behavioralconstruct (the degree of participative beha-viors of users during the development process), while user involvement and user attitudes arepsychological
constructs. The relationships among user participa-tion, user attitudes, and user involvement appear to be interrelated and complex. This leads to the hypoth-eses:
Hypothesis 2A. User participation (X2) increases with more favorable user attitudes (X3) toward the proposed system.
Hypothesis 2B. User participation (X2) increases with more user involvement (X4).
Many early researchers claimed that users who are actively participants in the system development
process are likely to establish the feeling that the system is good. Participants become acquainted with the new system. Their uncertainty and fear is alle-viated through the participative process.
Information systems which are thought to be both important and personally relevant to users are also likely to result in positive affective or evaluative feelings of the users. Similar supports for this argument can be drawn from relevant research in other disciplines. Per-sons who are highly involved with an issue have been found to possess more favorable attitudes about the issue (e.g., [67]). In marketing, customers who are highly involved with a product have been found to develop more positive feelings toward this product (e.g., [25,62]). Organizational behavior research shows that employees who are highly involved with their jobs have been found to appreciate their jobs (e.g., [39]). It is, therefore, hypothesized that:
Hypothesis 3A. User attitudes (X3) become more favorable through more user participation (X2).
Hypothesis 3B. User attitudes (X3) become more favorable with more user involvement (X4).
A number of psychological theories are able to elucidate the positive in¯uence of user participation on user involvement. Cognitive dissonance theory [20] and attribution theory [8] suggest that a person's beliefs will be aligned with his or her behaviors. Robey and Farrow [63] have claimed that participative users may change system attributes to meet their needs and desires, and help develop a system deemed as important and personally relevant.
The correlation between user involvement and user attitudes is bi-directional. Persons with extreme attitudes toward an issue tend to develop belief that the issue is both important and personally relevant. Millman and Hartwick [55] have argued that users with positive attitudes toward the new systems tend to become more involved. As a result, two hypotheses are made:
Hypothesis 4A. User involvement (X4) increases with more user participation (X2).
Hypothesis 4B. User involvement (X4) increases with more favorable user attitudes (X3).
2.3. Contingency factors
The contingency factors which have been studied in previous research include system complexity (e.g., [52]), stage development, resource constraints (e.g., [18]), communication, user attitudes, degree of user involvement, management styles, top management support, system impact, etc. It would be unwieldy and imprudent to incorporate all of these factors into one model. This study considers three other important contingency factors: system impact (X5), system com-plexity (X6), and development methodology (X7).
System impact refers to the potential changes in the organization brought about by the system. A new IS may incur organizational changes, such as jobs, inter-personal relationships, and organizational structure. Positions in the organization may be eliminated or created, and the job descriptions of people may be altered to incorporate different responsibilities.
Based on organizational change theory, it is known that people are used to the status quo and any disrup-tion of the organizadisrup-tional balance would evoke human resistance. It is, therefore, hypothesized that:
Hypothesis 5. System impact (X5) has a negative influence on user attitudes (X3).
System complexity arises in the developer's envir-onment and refers to the ambiguity and uncertainty that surround the practice of development. A complex IS is considered dif®cult to develop due to the large number of interdependent parts and the lack of a model structure. It is hypothesized that
Hypothesis 6. Higher degree of system complexity (X6) results in more user participation (X2).
Development methodologies range from traditional system development life cycle, prototyping, off-the-shelf software, and end-user computing, to outsour-cing. Among these development methodologies, out-sourcing has drawn much attention recently and is now in vogue in the IS domain [56].
When outsourcing shifts the burden of developing a system to outside parties, it also appears to reduce user responsibility for developing and validating the com-plex evolving IS functions as well as the need to learn how to apply new technologies. Such a mitigation in
the responsibility as a consequence of outsourcing may mislead users to the belief that the outsourced IS is less important and less personally relevant. More-over, one major disadvantage of outsourcing is the loss of control [46]. Therefore, it is hypothesized that
Hypothesis 7. Outsourcing (X7) leads to less user involvement (X4).
3. Estimation models
3.1. The structural form
The simultaneous contingency model can be expressed as a simultaneous-equations system of the following form:
X1ib10b12X2iv1i; i1;2;. . .;n (3.1)
X2ib20b23X3ib24X4ib26X6iv2i;
i1;2;. . .;n (3.2)
X3ib30b32X2ib34X4ib35X5iv3i;
i1;2;. . .;n (3.3)
X4ib40b42X2ib43X3ib47X7iv4i;
i1;2;. . .;n: (3.4)
Three contingency variablesX5,X6 andX7 (system impact, system complexity, and outsourcing) are the exogenous (independent) variables which are deter-mined outside of the model. VariablesX1,X2,X3 and
X4 (system success, user participation, user attitudes, and user involvement) are the endogenous (jointly dependent) variables that are interdependent and should be determined jointly. Since there are four endogenous variables, there are exactly four equations in the simultaneous-equations system.
The hypotheses established from the relationships in the simultaneous contingency model as well as the associated coef®cients are summarized in Table 1, in which the dependent variable is highlighted in bold-face and an independent variable is expressed in normal font.
A model is identi®ed if it is in a unique statistical form, enabling unique estimates of its parameters to be subsequently made from sample data. An equation is under-identi®ed if its statistical form is not unique; otherwise, it is exactly- or over-identi®ed. It is
neces-sary that the order condition for identi®cation be satis®ed for each equation in the system. The order condition stipulates that the number of variables (exo-genous) excluded from an equation should be greater than (over-identi®cation) or equal to (exact identi®ca-tion) the number of endogenous variables in the equation less one [28,29].
The application of the order condition to the simul-taneous Eqs. (3.1)±(3.4) reveals that Eq. (3.1) is over-identi®ed, and Eqs. (3.2)±(3.4) are exactly-identi®ed. None of the equations is under-identi®ed and, conse-quently, there is no problem of identi®cation.
3.2. Estimation issues
The ordinary least squares (OLS) estimation method is not appropriate because the estimators of the structural coef®cients are biased and inconsistent due to the so-called simultaneity bias or simultaneous-equation bias [45]. Instead, the methods of two-stage least squares (2SLS) and three-stage least squares (3SLS) should be used.
The major difference between 2SLS and 3SLS lies in the assumptions underlying the random errorsv1,v2,
v3 and v4 of Eqs. (3.1)±(3.4). If the v's of these
equations are correlated (i.e., E vjvk 60, forj,k2
{1, 2, 3, 4} andj6k), then 3SLS is more appropriate than 2SLS because it produces more ef®cient
Table 1
Hypotheses tested and associated coefficients
Relationship Hypothesis Coefficient Direction
System success (X1)
User participation (X2) 1 b12
User participation (X2)
User attitudes (X3) 2A b23
User involvement (X4) 2B b24 User attitudes (X3)
User participation (X2) 3A b32
User involvement (X4) 3B b34
User involvement (X4)
User participation (X2) 4A b42
User attitudes (X3) 4B b43
System impact (X5)
User attitudes (X3) 5 b35 ÿ
System complexity (X6)
User participation (X2) 6 b26
Outsourcing (X7)
User involvement (X4) 7 b
estimates. Such correlations among the random errors could be present if other possible contingency vari-ables are unintentionally omitted from the simulta-neous contingency model, leaving the in¯uence of these omitted variables to be absorbed by the random errors of the equations and rendering the random errors correlated.
3.3. The reduced form
The reduced form of the structural form can readily be obtained through an algebraic operation. First, the structural-form equations are written in matrix for-mulation, tor of the endogenous or jointly dependent variables,
Xi 1;X5i;X6i;X7i the 41 column vector of the random disturbances,
B
is the 44 matrix of coefficients of the simulta-neously dependent variables, and
G
is the 44 matrix of coefficients of the exogenous variables.
Secondly, premultiplying the structural form (3.5) through by Bÿ1, the inverse of B, and rearranging
yields the reduced form in matrix terms:
YiPXiWi
p40 p41 p42 p43
2
is the 44 matrix of the reduced-form coefficients,
andWiBÿ1Vi is the 44 matrix of the
reduced-form random disturbances.
Alternatively, we can write out the reduced-form system (3.6) algebraically, analogous to the structural-form Eqs. (3.1)±(3.4):
X1ip10p11X5ip12X6ip13X7iw1i (3.8) X2ip20p21X5ip22X6ip23X7iw2i (3.9) X3ip30p31X5ip32X6ip33X7iw3i (3.10) X4ip40p41X5ip42X6ip43X7iw4i (3.11)
Each of the reduced-form equations shows the dependence of an endogenous variable onall of the exogenous variables and a random disturbance. Thus, it enables us to ®nd the total effect (sum of the coef®cient estimates) of the exogenous contingency variables on an endogenous variable.
4. Research method
4.1. Data collection
We conducted a survey of 154 organizations ran-domly selected throughout the US to gather the data. The survey was directed to organizations that were under or had just completed the implementation of a new IS. Most of the organizations surveyed belong to the manufacture and service industries, and some were non-pro®t entities. The response rate was 54%. Excluding those returned blank or with incomplete information requested, the valid responses from 32 of the surveyed organizations were used to estimate the simultaneous contingency model and test the hypoth-eses presented. From the viewpoint of statistics, the sample size is considered large enough (n25). Users and system designers were surveyed separately to elicit data on different measures.
4.2. Measurement of variables
Several previous and useful studies were used as guidelines to design the survey questionnaire and to measure the relevant variables. They are listed in Table 2.
The survey involves 53 questions. Most of the questions contain ®ve to six bipolar adjective pairs in a question. For each pair, the respondents were
requested to check or circle the numbered box that is considered most suitable in the seven interval scale (from 1 to 7). The weights used to quantify the scaling of the seven intervals (boxes) went fromÿ3 through
3. During computation, these were assigned weights 1, 0.85, 0.70, 0.55, 0.40, 0.25, and 0.10. The score was then computed using
SiX
n
j1
RijWij
whereRijthe reaction to factorjby individuali, and Wijthe importance of factorj to individuali[4].
4.3. Data analysis
The Cronback alpha reliability coef®cient [12] was computed to assess the reliability of the responses to
all instruments. The alpha coef®cients for system success, user participation, user attitudes, user invol-vement, system impact, and system complexity are 0.91, 0.86, 0.83, 0.89, 0.78, and 0.72, respectively. These alpha reliability coef®cients are favorably com-parable to those reported in the literature and exceed the commonly accepted level of 0.70 used in social sciences studies. The descriptive statistics of the vari-ables and the corresponding correlation matrix of the variables are shown in Tables 3 and 4, respectively.
4.4. Estimation results
Three estimation methods, OLS, 2SLS, and 3SLS, are applied to estimate Eqs. (3.1)±(3.4) of the simul-taneous contingency model. The system estimates by 3SLS are more ef®cient than the 2SLS estimates if the random errors v's of structural Eqs. (3.1)±(3.4) are correlated. The estimation results by OLS, 2SLS and 3SLS are summarized in Table 5, which shows that the OLS estimation yields one estimate (b35, which is underlined) with the sign contrary to the expected direction, while the 2SLS and 3SLS estimates corre-spond to all the theoretical expectations. This abnor-mal sign may be attributed to the simultaneous-equation bias resulting from the inappropriateness of the OLS applied to estimate Eqs. (3.1)±(3.4) of the simultaneous-equations system. Therefore, we conclude that, for the simultaneous contingency
Table 2
Instruments to measure variables
Variable Instrument
User satisfaction (X1) [4,36] User participation (X2) [53,61] User attitudes (X3) [19]
User involvement (X4) A Likert-type scale from 1 to 7 System impact (X5) [52]
System complexity (X6) [52]
Outsourcing (X7) Dummy variable (1: outsourced; 0: inhouse)
Table 3
Descriptive statistics of the variables studied
Variable X1 X2 X3 X4 X5 X6 X7
Mean 0.986 1.276 1.521 1.661 1.057 ÿ0.024 0.607
Std. dev. 0.459 0.824 0.759 1.093 0.633 0.427 0.497
Table 4
Correlation matrix of the variables under consideration
X1 X2 X3 X4 X5 X6 X7
X1 1 0.368 0.596 0.634 0.655 0.347 0.129
X2 0.368 1 0.181 0.009 0.090 0.355 0.211
X3 0.596 0.181 1 0.470 0.555 0.010 0.324
X4 0.634 0.009 0.470 1 0.753 0.095 0.148
X5 0.655 0.090 0.555 0.753 1 0.208 0.132
X6 0.347 0.355 0.010 0.095 0.208 1 0.067
model developed above, 2SLS and 3SLS are more appropriate than OLS to obtain the estimates of the structural coef®cients in Eqs. (3.1)±(3.4).
In addition, one can observe that the 3SLS estimates are more ef®cient than their 2SLS counterparts. This then suggests that some other contingency variables were unintentionally omitted from the simultaneous contingency model, as it is theoretically and practi-cally impossible to consider all contingency variables. The reduced form (3.6) in matrix formulation or Eqs. (3.8)±(3.11) can be estimated by using the rela-tionship (3.7) between the structural coef®cients (B
and G) and the reduced-form coef®cients (P). The estimates ofPare given in Table 6 and re¯ect the total effect of one exogenous contingency variable on each endogenous variable.
From the hypothesized directions, it is expected that the total effects of system impact on system success, user participation, attitudes, and involvement (i.e., p11, p21,p31andp41, respectively) must be all negative. All the 3SLS estimates match, and all the 2SLS estimates violate, the expectations. This serves as a further evidence that 3SLS is more appro-priate than 2SLS in estimating the simultaneous contingency model. Similarly, the total effects of system complexity and outsourcing on the endogen-ous variables must be positive and negative, respec-tively. Again, the 3SLS estimates correspond to, but the 2SLS estimates fail to meet, the anticipated directions. Thus, we may conclude that the 3SLS method is indeed a better choice than the 2SLS method.
5. Discussion
5.1. Effect of user participation on system success
One of the main hypotheses of this study is empiri-cally to corroborate the positive in¯uence of user participation on the successful outcome of system implementation. Based on the results, Hypothesis 1 is con®rmed.
It should be pointed out that this ®nding does not come from a hypothesized simple direct causal equa-tion of user participaequa-tion and system success, esti-mated by the classical OLS. Instead, it is implied by the simultaneous contingency model estimated by the more rigid methodologies of estimation. The simultaneous contingency model appears to be a better and more appropriate alternative to examining the relationship between user participation and system success.
Table 5
Estimates of the structural-form coefficients of the model
Hypothesis 1 2A 2B 6 3A 3B 5 4A 4B 7
Coefficient b12 b23 b24 b26 b32 b34 b35 b42 b43 b47
Sign ÿ ÿ
OLS 0.334 0.331 0.007 0.593* 0.255 0.033 0:346 0.025 0.531 ÿ0.061 2SLS 1.088* 0.340 0.042 0.608* 1.061 0.995 ÿ0.998 1.953 1.005 ÿ0.212 3SLS 1.269** 0.354 0.186 0.407* 0.871 0.181 ÿ0.024 0.922 0.986* ÿ0.082
* significant at the 0.05 level; and ** significant at the 0.01 level.
Table 6
Estimates of the reduced form
P Method
2SLS 3SLS
p10 0.527 1.399
p11 0.361 ÿ0.133
p12 0.000 3.453
p13 0.076 ÿ0.212
p20 0.859 1.608
p21 0.332 ÿ0.105
p22 0.000 2.721
p23 0.070 ÿ0.167
p30 0.578 2.086
p31 0.798 ÿ0.162
p32 ÿ1.591 3.437
p33 0.192 ÿ0.229
p40 ÿ0.018 2.423
p41 1.451 ÿ0.256
p42 ÿ1.599 5.898
p43 0.118 ÿ0.462
5.2. Simultaneous relationship: user participation, attitudes, and involvement
The positive effects of user attitudes and involve-ment on user participation are consistent with our theoretical expectations. But both are statistically insigni®cant (and hence Hypotheses 2A and 2B are not supported by the data).
The in¯uences of user participation and user invol-vement upon user attitudes are consistent with the hypothesized directions but the estimates are not sig-ni®cant, and Hypotheses 3A and 3B are not con®rmed. There is a relatively large and positive association of user participation with user involvement but is again not statistically signi®cant and Hypothesis 4A is not con®rmed. User attitudes have been found to exert a signi®cant and positive effect on involvement (b430.986 from 3SLS which is signi®cant at the 5% level) and, hence, Hypothesis 4B is supported by the 3SLS estimate.
5.3. Effects of contingency factors
System impact as a contingency factor has been found to have a negative effect on user attitudes but it
is not signi®cant. Hence, Hypothesis 5 is not sup-ported.
The association of system complexity with user participation is positive and signi®cant at the 5% level (b260.608 from 2SLS and 0.407 from 3SLS) and, hence, Hypothesis 6 is supported. IS managers ask users to participate more in systems which are technically complex, in order to facilitate the devel-opment process and increase the likelihood of system success.
Outsourcing has a negative effect on user involve-ment but the result is not signi®cant. Therefore, Hypothesis 7 is not supported. This may suggest that outsourcing is rarely simple. Which aspects of out-sourcing are signi®cant to users is an interesting research question.
Fig. 2 presents the simultaneous contingency model along with the hypotheses tested and the structural coef®cient estimates obtained by 3SLS.
5.4. Total effects of exogenous variables on the jointly dependent variables
Based on the estimates of the structural form and the reduced form, we are able to compute the total effects
of exogenous variables on the endogenous variables, as presented in Table 7.
The structural total effect on a jointly dependent (endogenous) variable includes the effect of other jointly dependent variable(s) and the contingency variable(s) appearing in the same equation, while the reduced total effect on a jointly dependent variable is the sum of the effects of all contingency variables serving as the exogenous variables in the simultaneous contingency model. The difference between 2SLS and 3SLS is comparatively larger in the reduced form than in the structural form; this is because the
reduced form shows the compound effects of all
the exogenous contingency variables on each endo-genous variable, while the structural form basically re¯ects the sum of the direct effects of relevant variables in the equation corresponding to this endo-genous variable.
For system success, the structural total effect (1.088 from 2SLS and 1.269 from 3SLS) is the same as the direct effect of user participation since user participa-tion is the only dependent variable in the system success equation.
5.5. Managerial implications
Several managerial implications can be drawn for IS practitioners. First and foremost, this study con-®rms the positive contribution of user participation to successful system outcomes. Therefore, IS managers should encourage (or stipulate) active participative behaviors from users during the system development process.
Such a need for user participation is especially important when the systems to be developed are technically complex. User participation may be pro-moted as a social process of interaction between users
and designers through which both parties can learn about each other's expectations and requirements, and hence resolve their con¯icts [58,59].
Managers should also pay attention to the psy-chological states of users toward the systems. The feelings of the users for or against the systems are interrelated with the importance and personal relevance users perceive about the systems. User participation should have more ef®cacy if such behaviors originate from the underlying favorable attitudes and spontaneous involvement, instead of from manager's forcible orders. Therefore, manage-ment may want to foster an atmosphere that helps users perceive the importance of the system and enhances their favorable attitudes toward the system, in order to facilitate user participation in the devel-opment process.
6. Conclusions
In the design and development process of a new IS, the effect of user participation on system outcome is supposedly positive but should not be taken for granted. Such an effect must be scrutinized by considering the contextual environment. It is neces-sary to include the relevant contingency factors that may affect user participation and system success both directly and indirectly. However, it is equally important to notice that the scenario may not be as sequential. A research model based on the contin-gency approach and the simultaneity analysis was proposed and empirically tested to investigate the proposition that the relationships among the contin-gent factors are determined simultaneously, rather than sequentially.
The estimation results af®rm the positive link between user participation and system success, and suggest that user participation, user attitudes, and user involvement form a circular relationship. It implies that getting users involved in the development process may improve their attitudes toward the system and enhance the importance and relevance users perceive about the system. Furthermore, the empirical results show that system complexity has a signi®cant effect on user participation.
We conclude by indicating that the simultaneity or interdependence methodology is in sharp contrast with
Table 7 Total effects
Jointly dependent variable Structural form Reduced form
2SLS 3SLS 2SLS 3SLS
System success (X1) 1.088 1.269 0.437 3.108 User participation (X2) 0.990 0.947 0.402 2.449 User attitudes (X3) 1.058 1.028 ÿ0.601 3.046 User involvement (X4) 2.746 1.826 ÿ0.030 5.180
the sequential causality methodology and that the former may be superior in applied MIS research. This research may answer the call of methodological rigor in IS empirical studies. We hope that more answers to this call are forthcoming.
Acknowledgements
The authors are indebted to two anonymous refer-ees, the Editor, Professor Edgar H. Sibley, of this journal, Larry Sanders, and Bonnie Glassberg for their constructive and valuable suggestions and comments. The usual disclaimer applies.
References
[1] S. Alter, Development patterns for decision support systems, MIS Quarterly 2 (3), 1978, pp. 33±42.
[2] P. Allingham, M. O'Connor, MIS success: why does it vary among users? Journal of Information Technology 7, 1992, pp. 160±169.
[3] W.L. Anderson, W.T. Crocca, Engineering practice and codevelopment of product prototypes, Communications of the ACM 36 (4), 1993, pp. 49±56.
[4] J.E. Bailey, S.W. Pearson, Development of a tool for measuring and analyzing computer user satisfaction, Manage-ment Science 29 (5), 1983, pp. 530±534.
[5] H. Barki, J. Hartwick, Rethinking the concept of user involvement, MIS Quarterly 13 (1), 1989, pp. 53±63. [6] H. Barki, J. Hartwick, Measuring user participation, user
involvement, and user attitude, MIS Quarterly 18 (1), 1994, pp. 59±79.
[7] J.J. Baroudi, M.H. Olson, B. Ives, An empirical study of the impact of user involvement on system usage and information satisfaction, Communications of the ACM 29 (3), 1986, pp. 232±238.
[8] D.J. Bem, Self-perception theory, in: L. Berkowitz (Ed.), Advances in Experimental Social Psychology, Vol. 6, 1972, pp. 1±62.
[9] R.J. Boland Jr., The process and product of system design, Management Science 24 (9), 1978, pp. 887±898.
[10] E. Carmel, R.D. Whitaker, J.F. George, PD and joint application design: a transatlantic comparison, Communica-tions of the ACM 36 (4), 1993, pp. 40±48.
[11] A.L.M. Cavaye, User participation in system development revisited, Information and Management 28, 1995, pp. 311± 323.
[12] L.J. Cronbach, Coefficient alpha and the internal structure of tests, Psychometrika 16 (3), 1951, pp. 297±334.
[13] B. DeBrabander, A. Edstrom, Successful information system
development projects, Management Science 24 (2), 1977, pp. 191±199.
[14] B. DeBrabander, G. Thiers, Successful information system development in relation to situational factors which affect effective communication between MIS users and EDP specialists, Management Science 30 (2), 1984, pp. 137± 155.
[15] W.H. DeLone, E.R. McLean, Information systems success: the quest for the dependable variable, Information Systems Research 3 (1), 1992, pp. 60±95.
[16] W.J. Doll, G. Torkzadeh, A discrepancy model of end-user computing involvement, Management Science 35 (10), 1989, pp. 1151±1171.
[17] A. Edstrom, User influence and the success of MIS projects: a contingency approach, Human Relations 30 (7), 1977, pp. 589±607.
[18] P. Ein-Dor, E. Segev, Organizational context and the success of management information systems, Management Science 24 (10), 1978, pp. 1064±1077.
[19] D. Elizur, L. Guttman, The structure of attitudes toward work and technological change within an organization, Adminis-trative Science Quarterly 21 (4), 1976, pp. 611±622. [20] L. Festinger, The Theory of Cognitive Dissonance, Row,
Peterson, Evanston, IL, 1957.
[21] M. Fishbein, I. Ajzen, Belief, Attitude, Intention and Behavior: an Introduction to Theory and Research, Addi-son-Wesley, Reading, MA, 1975.
[22] C.R. Franz, Contingency Factors Affecting the User Involve-ment Role in the Design of Successful Information Systems, Ph.D. dissertation, University of Nebraska, 1979.
[23] C.R. Franz, D. Robey, Organizational context, user involve-ment, and the usefulness of information systems, Decision Sciences 17 (4), 1986, pp. 329±356.
[24] C.A. Gallagher, Perceptions of the value of a management information system, Academy of Management Journal 17 (1), 1974, pp. 46±55.
[25] M.P. Gardner, A.A. Mitchell, J.E. Russo, Low involvement strategies for processing advertisement, Journal of Advertis-ing 14 (2), 1985, pp. 4±12.
[26] E.J. Garrity, G.L. Sanders, Information Systems Success Measurement, Idea Group Publishing, Hershey, PA, 1998. [27] H.L. Gibson, Determining user involvement, Journal of
Systems Management 28, 1977, pp. 20±22.
[28] A.S. Goldberger, Econometric Theory, Wiley, New York, 1964.
[29] W.H. Greene, Econometric Analysis, 2nd Edition, Macmillan, New York, 1993.
[30] A. Guthrie, Attitudes of user-managers towards management information systems, Management Informatics 3 (5), 1974, pp. 221±232.
[31] J. Hartwick, H. Barki, Explaining the role of user participa-tion in informaparticipa-tion system use, Management Science 40 (4), 1994, pp. 440±465.
[32] R.A. Hirschheim, Assessing participative systems design: some conclusions from an exploratory study, Information and Management 6, 1983, pp. 317±327.
influence, and satisfaction among members of problem-solving groups, Psychological Reports 16, 1965, pp. 661±667. [34] M.I. Hwang, R.G. Thorn, The effect of user engagement on system success: a meta-analytical integration of research findings, Information and Management 35, 1995, pp. 229± 236.
[35] R.H. Igersheim, Management Response to an Information System, AFIPS Conference Proceedings 1976, pp. 877±882. [36] B. Ives, M.H. Olson, J.J. Baroudi, The measurement of user
information satisfaction, Communications of the ACM 26 (10), 1983, pp. 785±793.
[37] B. Ives, M.H. Olson, User involvement and MIS success: a review of research, Management Science 30 (5), 1984, pp. 586±603.
[38] K.M. Kaiser, A. Srinivasan, The relationship of user attitudes toward design criteria and information systems success, National AIDS Conference Proceedings, 1980, pp. 201±203. [39] R.N. Kanungo, Work Alienation: an Integrative Approach,
Praeger, New York, 1982.
[40] L. Kappelman, E. McLean, The respective roles of user participation and user involvement in information system implementation success, Proceedings of the International Conference on Information Systems, New York, NY, 1991, pp. 339±349.
[41] P.G.W. Keen, Information systems and organizational change, Communications of the ACM 24 (1), 1981, pp. 24±33. [42] E. Kim, J. Lee, An exploratory contingency model of user
participation and MIS use, Information and Management 11 (2), 1986, pp. 87±97.
[43] W.R. King, T. Lee, The effects of user participation on system success: toward a contingency theory of user satisfaction, Proceedings of the International Conference on Information Systems, New York, NY, 1991, pp. 327±338.
[44] K. Klenke, Construct measurement in management informa-tion systems: a review and critique of user satisfacinforma-tion and user involvement instruments, INFOR 30 (4), 1992, pp. 325± 348.
[45] A. Koutsoyiannis, Theory of Econometrics, 2nd Edition, Macmillan, New York, 1977.
[46] M.C. Lacity, R. Hirschheim, Information Systems Outsour-cing: Myths, Metaphors, and Realities, Wiley, Chichester, 1993.
[47] E.E. Lawler, J.R. Hackman, Impact of employee participation in the development of pay incentive plans: a field experiment, Journal of Applied Psychology 53, 1969, pp. 467±471. [48] H.P. Lu, J.Y. Wang, The relationships between management
styles, user participation, and system success over MIS growth stages, Information and Management 32, 1997, pp. 203±213.
[49] H.C. Lucas Jr., Systems quality, user reactions, and the use of information systems, Management Informatics 3 (4), 1974, pp. 207±212.
[50] R.I. Mann, H.J. Watson, A contingency model for user involvement in DSS development, MIS Quarterly 8 (1), 1984, pp. 27±38.
[51] M.L. Markus, Power, politics, and MIS implementation, Communications of the ACM 26 (6), 1983, pp. 430±444.
[52] F.W. McFarlan, Portfolio approach to information systems, Harvard Business Review 59, 1981, pp. 142±159.
[53] J.D. McKeen, T. Guimaraes, J.C. Wetherbe, The relationship between user participation and user satisfaction: an investiga-tion of four contingency factors, MIS Quarterly 18 (4), 1994, pp. 427±451.
[54] J.D. McKeen, T. Guimaraes, Successful strategies for user participation in systems development, Journal of Management Information Systems 14 (2), 1997, pp. 133±150.
[55] Z. Millman, J. Hartwick, The impact of automated office systems on middle managers and their work, MIS Quarterly 11 (4), 1987, pp. 479±491.
[56] D. Minoli, Analyzing Outsourcing: Reengineering Informa-tion and CommunicaInforma-tion Systems, Mcgraw-Hill, New York, 1995.
[57] E. Mumford, Consensus system design: an evaluation of this approach, in: N. Szyperski, E. Grochla (Eds.), Design and Implementation of Computer-Based Information Systems, Sijthoff and Noordhoff, Groningen, Holland, 1979. [58] M. Newman, F. Noble, User involvement as an interaction
process: a case study, Information Systems Research 1 (1), 1990, pp. 89±113.
[59] M. Newman, D. Robey, A social process model of user-analyst relationships, Information Systems Research 3 (2), 1992, pp. 249±266.
[60] D. Norton, F.W. McFarlan, Product management, in: R.L. Nolan, F.W. McFarlan (Eds.), The Information Systems Handbook, Dow Jones-Irwin, Homewood, IL, 1975, pp. 517±528.
[61] M.H. Olson, B. Ives, User involvement in system design: an empirical test of alternative approaches, Information and Management 4 (4), 1981, pp. 183±195.
[62] R.E. Petty, J.T. Cacioppo, D. Schumann, Central and peripheral routes to advertising effectiveness: the moderating role of involvement, Journal of Consumer Research 10, 1983, pp. 135±146.
[63] D. Robey, D. Farrow, User involvement in information system development: a conflict model and empirical test, Manage-ment Science 28 (1), 1982, pp. 73±85.
[64] N. Saleem, An empirical test of the contingency approach to user participation in information systems development, Journal of Management Information Systems 13 (1), 1996, pp. 145±166.
[65] R.J. Schnoberger, MIS design: a contingency approach, MIS Quarterly 4 (1), 1980, pp. 13±20.
[66] I.S. Seeborg, The influence of employee participation in job design, Journal of Applied Behavioral Science 14 (1), 1978, pp. 87±98.
[67] C.W. Sherif, M. Sherif, R.E. Nebergall, Attitude and Attitude Change: the Social Judgment Ð Involvement Approach, Saunders, Philadelphia, 1965.
[68] E.B. Swanson, Management information systems: apprecia-tion and involvement, Management Science 21 (2), 1974, pp. 178±188.
[69] P. Tait, I. Vessey, The effect of user involvement on system success: a contingency approach, MIS Quarterly 12 (1), 1988, pp. 91±108.
Winston T. Lin is Professor in the School of Management at the State University of New York at Buffalo. He received his B.A. from National Taiwan University and Ph.D. from Northwestern University, Evanston, Illinois. His cur-rent research interests are in the areas of forecasting, information systems, and multinational finance. He has published 53 articles in refereed proceedings and leading academic and professional journals including Journal of the Association for Information Systems, Communications of the ACM, Information & Management, Journal of Business & Economic Statistics, The Financial Review, Journal of Financial and Quantitative Analysis, International Journal of Forecasting, Journal of Forecasting, etc. He has presented his papers at 48 national and international conferences. He has been awarded numerous significant research grants and awards.
Benjamin B.M. Shaois Assistant
