Paving the path of good intentions
social psychological factors influencing the use of project management practices
Russell K. Thornley
The objective of the current study is to facilitate a better understanding of the social psychological factors that influence the use of project management practices, and to do so by integrating three popular theoretical frameworks. The Theory of Reasoned Action (TRA; Fishbein & Ajzen, 1975) and the Theory of Planned Behavior (TPB; Ajzen, 1991) from social psychology have shown remarkable heuristic value in predicting and explaining a broad range of behaviors. Similarly, the Technology Acceptance Model (TAM; Davis, 1986; Davis, 1989; Davis, Bagozzi, & Warshaw, 1989; Davis, 1993), from information systems research, has been used to explain and predict the use of a wide variety of information technologies.
Model comparison studies have shown that the revised TAM (cf. Szajna, 1996) predicts software usage intention moderately better than TRA (Davis et al., 1989); that TAM predicts intention only somewhat better than TPB (Mathieson, 1991); and, in a comparison of TAM against a “pure” (in contrast to “decomposed”) TPB, that TAM and the “pure” TPB predict intentions equally well (Taylor & Todd, 1995). In short, a review of the literature suggests, at best, that all three models are essentially equivalent in predictive ability, and at worst, that the research is inconclusive (see also Karahanna & Straub, 1999; Venkatesh & Davis, 2000; Mathieson, 1991; Taylor & Todd, 1995a, 1995b, 1995c).
Researchers have attempted to integrate these three frameworks in at least two different ways. The first approach leverages the attitude construct common to all three models and achieves a rudimentary integration (Figure 1). A second approach decomposes all the belief structures in all three models, resulting in a more complete integration, but with the added complexity of multiple “cross-over effects” (Figure 2).
The rudimentary integration decomposes the attitude construct from TRA. In the original expectancy-value conception of TRA, behavioral beliefs (bbi) and evaluations (evi) are multiplied and summed: AB = Σi = 1, n (bbi)(evi). As a derivative of TRA, TAM simply decomposes TRA’s context-specific behavioral belief component (bbi), and instead posits two distinct, generalized belief constructs – perceived usefulness and perceived ease of use. The evaluation term (evi) is either dropped or becomes the measure of attitude, with the relative contributions of these new generalized beliefs estimated by multiple regression (Davis, 1986; 1989).
In their model comparison study, Taylor and Todd (1995c) tested the key relationships suggested by this rudimentary integration in a model they called the “Augmented TAM” (p. 561). The results showed that a simple integration of TAM and TPB fit the data better than did either model alone. Similarly, in a series of model comparison studies, Taylor and Todd (1995a; 1995b) compared TAM and TPB with a “decomposed theory of planned behavior” (DTPB; see Figure 2) and found that the models were fairly comparable, with the decomposed model explaining intentions better than the other two.
Figure 1. A Simple Integration of TAM and TPB (cf. Taylor & Todd, 1995c)
Figure 2. An Integration of a “Decomposed” TPB and TAM
Moreover, Davis, Bagozzi, and Warshaw (1989) further explored their results by conducting a post hoc factor analysis of the seven behavioral belief items from TRA, along with the four usefulness and four ease of use items from TAM. The results suggested the existence of four dimensions, two of which pertained to “usefulness” and “ease of use.” Using the same data, they re-estimated the path coefficients using a 7-item “usefulness index,” and found that it accounted for the variance in behavioral intentions better than did the original models. They concluded that “combining the beliefs of TRA and TAM into a single analysis may yield a better perspective on the determinants of [behavioral intention] than that provided by either model by itself” (p. 994).
These results suggest that some of the items used to measure TAM’s usefulness construct may be measuring the same underlying dimension as TRA’s behavioral belief items. We should not find this surprising. In its original form, the “belief-evaluation term,” Σ(bbi)(evi), was part of the operational definition of an expectancy-value conception of attitude. Indeed, it has no intuitive interpretation of its own, and was not meant to stand alone as a distinct construct that served as a “determinant” of attitude. The same is true for the other belief-based terms in TRA and TPB – Σ(nbi)(mci) and Σ(cbi)(pfi); they have no independent interpretation, and are meant to be expectancy-value-based measures of subjective norm and perceived behavioral control, respectively. As such, returning to the original theorizing suggests another possible approach to integrating the three models. Rather than these belief-based measures serving as determinants of attitude, subjective norm, and perceived behavioral control, they are better conceived as manifest measures or indicators of those three key latent constructs, as shown in Figure 3 (see Byrne, 2001; Anderson & Gerbing, 1988).
Figure 3. A Hypothesized Latent Variable Integrated Framework
Project managers from among the members and those affiliated with the Northern Utah Chapter of the PMI were recruited by ads in the chapter newsletters, along with a direct e-mail recruitment message to the email distribution list of 734 persons. 182 responses were received, resulting in a response rate of about 24.8%. Most participants were senior professionals, with 64.5% reporting being 35 years or older and 56.5% over the age of 45, and a majority (61.3%) being currently employed fulltime as a “project leader.” 68.3% of respondents were male. All respondents reported that they reside in Utah.
The survey instruments were developed according to established guidelines and adapted from several published instruments that have been validated in a number of studies of technology adoption (Ajzen, 1991; Ajzen, 2002); Davis, 1989; see also Taylor & Todd, 1995). The measures of project planning behavior and intention were developed in the same format as the Maturity Questionnaire (Zubrow, Hayes, Siegel, & Goldenson, 1994), but with questions created based on the project planning key process area of the CMMI models, and in accordance with the guidelines in the Software Engineering Institute’s “Standard CMMI Appraisal Method for Process Improvement” (SCAMPISM; see CMU/SEI-2001-TR034).
The endogenous variables in the hypothesized framework are project management-related constructs that have been operationally defined in the CMM models. Specifically, the “key practices” (i.e., “specific practices” in CMMI) from the Project Planning key process area were turned into 14 questionnaire items similar to those used in the original Maturity Questionnaire version 1.1 (Zubrow et al., 1994). These 14 items represent three key dimensions of project planning practices that correspond to the three goals of the project planning process area: four estimating practices, six plan development practices, and four plan commitment practices (cf. CMMI-SE/SW/IPPD, v1.1, Continuous Representation, p. 192 – 200). Based on previous process improvement pilot studies, the project planning practices were identified and scored as 5 = “best practice” (BP); 4 = “good practice” (GP); 3 = “conventional practice” (CP); 2 = “marginal practice” (MP); and 1 = “no practice/don’t know,” as shown in Table 1.
Data Analyses and Results
The measurement model is comprised of selected measurement indicators for eight exogenous variables: general and belief-based measures of each of attitude (ATT; BBEV), subjective norm (SN; NBMC), and perceived behavioral control (PBC; CBPF); measures of usefulness (USE) and ease of use (EOU); and three endogenous variables, as shown in Table 2.
The idea of internal consistency suggests that the items selected should in fact assess the same underlying construct. Initial estimates of the internal consistency of the TPB measures of ATT, SN, and PBC suggested generally marginal to adequate internal consistency, with α = .93, .74, and .80, respectively. Similarly, internal consistency of the TAM measures of USE and EOU were α = .86 and .75, respectively. In general, the distributions of these variables were slightly negatively skewed, although none of the measures of skew exceeded 2.0 (Byrne, 2001).
|plan Estimating||est_b01||Was a top-level work breakdown structure (WBS) established to estimate the scope of the project?||5 BP|
|est_b02||Were estimates of the attributes of the work products and tasks established and maintained?||4 GP|
|est_b03||Were the project life-cycle phases defined, upon which to scope the planning effort?||3 CP|
|est b04||Were the project effort and cost for the work products and tasks estimated based on estimation rationale?||2 MP|
|Plan Development||dev_b01||Was the project’s budget and schedule established and maintained?||2 MP|
|dev_b02||Were project risks identified and analyzed?||3 CP|
|dev_b03||Was the management of project data planned?||3 CP|
|dev_b04||Were resources to perform the project planned?||4 GP|
|dev_b05||Were knowledge/skills for the project planned?||5 BP|
|dev_b06||Was the overall project plan established and maintained?||4 GP|
|Plan Commitment||com_b01||Was the involvement of stakeholders planned?||5 BP|
|com_b02||Were all plans that affect the project reviewed to understand project commitments?||3 CP|
|com_b03||Was the project plan reconciled to reflect available and estimated resources?||2 MP|
|com_b04||Were commitments obtained from relevant stakeholders who were responsible for performing and supporting plan execution?||4 GP|
Table 1. Project Planning Behavior Categories and Scoring
I also estimated the reliability of the belief-based measures. It is important to note that Ajzen (1991) insists that no assumption need be made that accessible beliefs themselves are internally consistent. According to the expectancy-value model, it is in their aggregate that these measures provide the manifest indicators of the latent constructs. When aggregated in the multiplicative manner required by Ajzen (1991), however, potential problems arise.
For instance, these measures are assessed on seven-point scales ranging from one (extremely unlikely) to seven (extremely likely). When each behavioral belief, for example, is multiplied by each evaluation term, and then summed in the typical manner, however, the resulting scales can range from four to 49, placing them on an underlying metric that differs substantially from the other measures in the model (cf. means in Table 2).
|Measure||# of Items||Mean||Std. Dev.|
|Subjective Norm (SN)||3||6.34||.771|
|Perceived Behavioral Control (PBC)||3||4.26||1.51|
|Perceived Usefulness (USE)||4||6.03||.771|
|Perceived Ease of Use (EOU)||4||5.05||.920|
|Belief-based Attitude (BBEV)||4||3.07||.490|
|Belief-based Subjective Norm (NBMC)||4||5.74||.820|
|Belief-based Perceived Behavioral Control (CBPF)||4||4.14||.910|
|Plan Estimating Practices (EST_BEHV)||4||10.40||3.14|
|Plan Development Practices (DEV_BEHV)||6||17.36||3.58|
|Plan Commitment Practices (COM_BEHV)||4||10.77||3.22|
Table 2. Summary of Exogenous and Endogenous Measures
As such, the measures of BBEV, NBMC, and CBPF were first multiplicatively combined as per Ajzen (1991), and then subjected to a square-root transformation to return them to an underlying metric that is more consistent with the other measures. The resulting measures showed internal consistency of α = .79, .83, and .81, respectively. All analyses were carried out using SPSS software, version 9.0 for Windows, or AMOS, version 4.0, developed by Arbuckle (1999).
To assess theory-specific construct validity – that my created measures were adequately assessing the constructs of the different theories – I performed three separate confirmatory factor analyses, first on the global measures of ATT, SN, and PBC (from TPB); then on the multiplicatively-combined belief based measures (from TPB); and finally on the USE and EOU items (from TAM). Tables 3, 4, and 5 contain the estimated loadings for these measurement items, respectively (factor loadings <; .30 are left blank for clearer presentation). In general, the factor loadings supported the construct validity of the theory-specific measurements.
The fact that SN02 did not load distinctly on its primary factor suggests that it is not an adequately clear indicator of the latent construct that it was intended to measure. One possible explanation for this is the ambiguity of this item. Participants may have been confused as to whether or not the question was asking about stakeholders, customers, and/or clients; and approval of one may not have been identical with approval of the others. As such, this item was dropped, bringing the internal consistency of the subjective norm indicator to α = .80.
|Attitude||ATT01||good – bad||.942|
|ATT02||harmful – helpful||.911|
|ATT03||positive – negative||.937|
|ATT04||foolish – wise||.665|
|Subjective Norm||SN01||overall, people important to me||.800|
|Perc.Behav. Control||PBC01||easy – difficult||.908|
|PBC02||under control – out of control||.650|
|PBC03||simple – complicated||.857|
Table 3. Factor Loadings: TPB General Measures of ATT, SN, & PBC
|Belief-based ATT [Σ(bbev)]||BBEV1||reduce overall costs of doing business.||.873|
|BBEV2||improve relationships with customers.||.720|
|BBEV3||improve communication with customers.||.795|
|BBEV4||keep pace with the competition.||.753|
|Belief-based SN [Σ(nbmc)]||NBMC1||Upper/top management||.771|
|NBMC2||Other project managers||.801|
|NBMC3||Other members of the PMO||.787|
|NBMC4||Your immediate supervisor||.878|
|Belief-based PBC [Σ(cbpf)]||CBPF1||Training it would take to get up to speed||.770|
|CBPF2||Appropriate tool knowledge/skill||.832|
|CBPF3||Cooperation of resources/technical people||.841|
|CBPF4||Cooperation and approval of management||.879|
Table 4. Factor Loadings: Multiplicative Measures of Σ(bbev), Σ(nbmc), & Σ(cbpf)
|Usefulness||USE01||improve my job performance||.833|
|USE02||accomplish tasks more quickly||.799|
|USE03||accomplish more work||.834|
|USE04||enhance my effectiveness||.865|
|Ease of Use||EOU01||require mental effort||.591|
|EOU02||easy for me to become skillful||.876|
|EOU03||easy for me to remember||.825|
|EOU04||overall, easy to use||.791|
Table 5. Factor Loadings: TAM Measures of USE & EOU
Next, I examined “goodness of fit” for the model, beginning with an assessment of the measurement model. According to Byrne (2001), the first evidence of poor fit is found when examining the adequacy of parameter estimates – their feasibility, the appropriateness of standard errors, and their statistical significance. In this regard, the hypothesized model evidenced poor fit as manifest by non-significant regression weights between the Attitude latent construct and BBEV, as well as the Personal Control latent construct and CBPF and EOU. For the current data, these three are not significant indicators of the expected latent constructs.
There is at least one important explanation for these results. It is likely that issues of personal control (and ease of use) are not very meaningful factors when it comes to project planning. Recall that TPB was intended to be used in cases when behavior is not entirely under a person’s volitional control. These results suggest that project planning behavior may be perceived as very much under the volitional control (and within the skills) of project managers. As such, perceived behavioral control is not a strong factor in project planning.
Removing the non-significant paths would retain only significant and appropriate parameter estimates, but still retain key constructs from all three of the theories. Moreover, it would acknowledge that personal control issues might be only slightly involved in explaining project-planning practices. Therefore, for theoretical as well as empirical reasons, these constructs were removed from the hypothesized measurement model.
With the non-significant paths removed, the evaluation of the fit of this measurement model was accomplished by examining a few important fit indexes (Byrne, 2001). A fit index that provides a quick overview of model fit is the discrepancy statistic (CMIN), which is distributed as X2, and represents the discrepancy between the sample covariance matrix and the covariance matrix implied by the hypothesized model. More important is the “relative chi-square” [also called the “X2/df ratio”], which is the ratio of the CMIN and the degrees of freedom. Various researchers have recommended using ratios around 2.00, or as high as 5.00 to indicate a reasonable fit (e.g., Marsh & Hocevar, 1985; cf. Byrne, 2001).
Additionally, similar to the “comparative fit index” (CFI), the “goodness of fit index” (GFI) is a measure of the relative amount of variance and covariance in the sample matrix that is jointly explained by the hypothesized model matrix. It ranges from zero to 1.00, with values close to 1.00 indicating good fit. Next, the “root mean square error of approximation” (RMSEA) takes into account the error of approximation in the population, and provides an index of how well the hypothesized model (with optimally chosen parameter values) would fit the population covariance matrix if it were available. An RMSEA value of about 0.05 or less would indicate a close fit of the model in relation to the degrees of freedom, (exact fit would produce an RMSEA = 0.0). A value of about 0.08 or less indicates a reasonable error of approximation (Byrne, 2001).
Finally, “modification indices” (MI) are used to detect specific areas of misfit in the hypothesized model, perhaps suggesting places where residuals may reflect a patterned relationship. In contrast to the “omnibus” model fit statistics discussed so far, the modification indices can be conceptualized as a X2 statistic with one degree of freedom. Few meaningful modification indices, therefore, suggest a model that is not plagued by specific misfitting parameters.
With these guidelines in mind, overall goodness of fit for the hypothesized measurement model is very good, with X2(3) = 6.43, a CMIN/df ratio of 2.145, (p = .092). The comparative fit index (CFI) = .992, GFI = .986, and RMSEA = .079, which is well within the 90% confidence interval between .000 and .164 (PCLOSE = .221). Finally, no theoretically meaningful modification indices were computed for this model, suggesting no meaningfully patterned residuals. As such, the measurement model not only provides a very good fit to the data, and retains only significant and appropriate parameter estimates, but also retains key constructs from all three of the theories, recognizing that personal control issues may be only slightly involved in explaining project planning practices.
Next I evaluated the fit of the overall model, measurement and structural components simultaneously. The high-level index of overall model fit revealed that the full structural model provided only fair-to-poor overall fit, with X2(15) = 66.984, a CMIN/df ratio of 4.466, (p < .000), CFI = .924, GFI = .921, and RMSEA = .137 (PCLOSE < .000). A further examination of the covariance as well as the residual covariance matrices, however, suggested some theoretically interesting relationships, as shown in the correlation matrix in Appendix A. The existence of these relationships was also suggested by modification indices computed by the Amos software. Joreskog and Sorbom (1984) describe modification indices as estimates of the amount by which the discrepancy function would decrease if the analysis were repeated with the constraints on a specific parameter removed, or with an added path that does not currently appear in a model.
In this regard, several of the largest modification indices (and estimated parameter changes) reported by Amos were between the three exogenous variables (Attitude, Social Influence, and perceived behavioral control) and one, but not necessarily all, of the planning indicators (i.e., estimating behavior, plan development behavior, and/or plan commitment behavior). These patterns are not surprising. According to the CMM/I, these three types of planning activities represent groups of related practices that bring about the three “quality goals” of the project planning key process area. As such, this result confirms the conceptual distinction between these three types of planning activities, whereas the correlations between them support the expectation that they are all an important part of the planning key process area.
Overall, the model would fit the data better by removing the latent “planning behavior” construct and allowing the three indicators to serve as three separate, albeit correlated, endogenous variables. Moreover, because the correlations and patterns of residuals suggest that the three types of planning activities – estimating practices, development practices, and commitment practices – are predicted differentially by different exogenous constructs, the removal of the latent variable allows the estimation of more interesting “specialized effects.” As such, the model was modified and re-estimated, removing non-significant parameters to arrive at the “trimmed” model shown in Figure 4.
Figure 4. A “Trimmed” Model of Project Planning
The “trimmed” model of project planning provides very good fit: Χ2(13) = 18.13, a CMIN/df ratio of 1.395 (p = 0.153); CFI = .992, GFI = .978, and RMSEA = .046, which is well within the 90% confidence interval between .000 and .092 (PCLOSE = 0.503). Finally, an examination of the residual covariances suggested no additional patterns, supported by the fact that Amos computed no theoretically meaningful modification indices for this model.
The trimmed model has the benefit of revealing some informative “specialized effects.” Specifically, estimates of this model resulted in a significant path coefficient between Attitude Toward Planning and Estimating Practices, but neither plan Development nor Commitment practices. Likewise, a significant path coefficient was estimated between the one-item measure of perceived behavioral control (from TPB) and Commitment practices, but neither Estimating nor plan Development practices. By contrast, overall, all three types of project planning practices were predicted the most by Social Influence, the implications of which will be discussed next.
The TRA, the TPB, and the TAM suggest that project planning behavior is influenced by three things: the degree to which project managers have a positive attitude toward project planning; the degree to which people who are important to project managers are perceived to be positively disposed toward project planning; and the degree to which project managers perceive that they have control over project planning. Generally speaking, these relations are confirmed in the current study.
Specifically, Attitude did explain a particular type of project planning behavior – project estimating practices. Project managers who have a positive attitude toward project planning are more likely to carry out the important activities of creating a top-level work breakdown structure (WBS) upon which to estimate the scope of the project; estimate the attributes of the work products and tasks; define project life-cycle phases upon which to scope the project; and estimate project effort and cost according to some estimation rationale. The corollary is that, if project managers are failing to engage in these important estimating practices, managers and consultants would be well-advised to engage in persuasive interventions whereby project managers’ attitudes toward estimating are improved.
In doing so, managers and consultants would be wise to leverage social influences (e.g., Cialdini, 1993). Social influence, in the current study, provided the strongest influence on all types of planning practices – estimation, plan development, and plan commitments. This should come as no surprise. Project managers, as leaders of dynamic social groups called “project teams,” are engaging in activities in environments that 10 are fundamentally social in nature. As such, they not only leverage social influences to build teams and collaborate with key stakeholders, but they are also strongly influenced by these same social processes. As such, the importance of project managers’ involvement with peers in professional associations such as the PMI cannot be overestimated. Through the networking and career-development opportunities afforded them by these professional associations, project managers are more likely to engage in important project planning practices that can dramatically affect the likelihood of successful project outcomes.
The current results are also consistent with previous research as regards the influence of perceived behavioral control on project managers’ engagement in industry-recognized good practices. Specifically, there remains some question as to whether or not the PBC construct adds significant predictive ability beyond that provided by TRA. According to a meta-analysis conducted by Conner and Armitage (1998), the benefits of adding PBC to TRA seem to add only about 4% to 5% to the variance explained in intention and only about 1% to the variance explained in actual behavior, above that explained by attitude and subjective norm. So it is with the current results. The degree to which project managers perceive that they have control over project planning seems not to predict estimation and plan development practices at all, and adds very little to predicting the establishment of project commitments. This makes some sense inasmuch as the control that project managers typically have over their projects is related to the authorization and commitments that they receive from key stakeholders. According to the current results, this appears to be a relatively minor issue.
TRA, TPB, and TAM were also expected to integrate well by their indicators loading meaningfully on three latent constructs. This expectation was only partially realized. The measurement model that provided the better fit integrated ATT measures from TRA, and USE measures from TAM, but EOU from TAM and PBC played either no role or a very minor role in the overall model. In this regard, it is particularly noteworthy that only one of the multiplicative, belief-based terms – NBMC – contributed meaningfully in this research. Despite the fact that the methodologies followed in the current study were those used in all TPB/TRA studies, neither BBEV nor CBPF provided meaningful indicators of an underlying construct. As such, the resulting measures should have supported, but did not, the plausibility of a three-dimensional structure underlying the antecedents of project planning behavior.
The most likely explanation for these results is that, overall, project managers’ perceptions of their personal control over project planning activities may not play a significant role in their actual use of planning practices. This is understandable given that the theory of planned behavior was posited for those conditions in which the behavior of actors is not under their volitional control. The mild degree to which personal control (and ease of use) may be involved in these processes was either appropriately captured in the current study, or was too weak for the power of this design to detect. In the latter regard, although numerous studies have supported the multidimensionality of the factor structure for attitude, subjective norm, and perceived behavioral control, as well as usefulness and ease of use from TAM, a meaningful latent variable integration was not achieved as regards the EOU and PBC constructs from TAM and TPB. Therefore, further research is certainly needed on the validity of these measurement instruments and the approaches to their development.
The study makes several important contributions to research and practice. First, this research provides the basis for the development of measures of integrated theoretical constructs that can be used to predict good practice intentions and behavior. Through this study, we learn whether or not an integrated model of the social psychological factors that influence technology acceptance applies not just to traditionally-defined “technology,” such as computer software, hardware, and related applications; but whether it also applies to the prediction of the use of “good practices” as an important form of technology. With improved understanding of the factors that influence the use of good practices, executives, managers, consultants, and other change agents can increase their chances of capitalizing on the benefits of organizational change initiatives.
Second, we learn that two of the three key constructs of the integrated model – attitude and social influence – are reliable, parsimonious constructs for prediction in this arena. As such, managers, consultants, and practitioners can leverage this model to predict the “readiness” of an organization for adopting specific good practices by administering pre-project questionnaires that assess practitioners’ attitudes, social influence, and control beliefs. Armed with this information, change agents can plan to mitigate risks to the success of their improvement initiatives by emphasizing techniques that differentially influence practitioners’ attitudes, perception of social influences, and/or perceptions of personal control, as needed.
These types of mitigation plans can be handled in a number of ways. For instance, if a consultant determines that attitudes toward a set of good practices are negative, he / she can include persuasive information activities that emphasize the “usefulness” of the good practices. Similarly, if preliminary surveys indicate social influences that are working against the adoption of good practices, a consultant can build more participative activities into the project approach, perhaps identifying influential members of the practitioner community and recruiting them to serve as key mentors and “power users” of the good practices. He or she might also provide “testimonials” and other information that indicates the social acceptability and/or popularity of the good practices.
Additionally, unlike most research in this area, this research is not only grounded in three well-established theories, making it the first study of its kind to apply these theories to project management practices, but is also a field study of those who actually manage real-world projects. By contrast, most of the previous research has been conducted on samples of university students (e.g., Davis et al., 1989; Davis, 1989; Davis, 1986; Mathieson, 1991; Taylor & Todd, 1995). As such, the results of the current study have more external validity and are more likely to tell us something about how these well-established theories work when applied to real-world settings.
Some Limitations of the Current Study
Despite the degree of external validity of this study, as with many studies, the results and conclusions are very likely to be limited to the particular sample, variables, and time frame represented by the design. The sample size did not allow for cross-validation (Browne & Cudeck, 1989). Conclusions must be regarded as tentative until further research can confirm or disconfirm similar findings. Moreover, these findings are limited to project managers in Utah. Extrapolating beyond that population must be done tentatively and with care, as the Utah Chapter of the PMI may or may not be strongly representative of the nationwide or worldwide population of those who manage projects.
As a cross-sectional (rather than longitudinal) design, the current study was limited to a particular occasion of measurement. As such, directional influences posited within the hypothesized models must be interpreted with caution. Gollob and Reichardt (1991) have pointed out that directional effects in structural equation models require three conditions: First, directional effects take some finite amount of time to operate. Second, a variable may be influenced by the same variable at an earlier point in time, an effect called “autoregressive.” And, finally, the magnitude of an effect may vary as a function of the time lag. Strictly speaking, then, there is no single true effect of one variable on another with a cross-sectional design. In this regard, subsequent phases of this line of research will provide more longitudinal data for making more appropriate “causal” inferences, and assessing the dynamic nature of many of the variables under consideration.
The results of the current study may also be limited by the choice of “technology” – general project management practices and, in particular, the “best,” “good,” “conventional,” and “marginal” practices based on pilot studies using the SEI’s Capability Maturity framework of models. As such, similar results may not be obtained when measuring project management practices using some other standard for project management practices.
Because continuous improvement implies that there is an ongoing intention to continue to engage in good practice discipline, the approach taken in the current study to integrate the extant intention models would be well suited to longitudinal investigations of project management process improvement initiatives. This is especially important in light of the fact that the purpose of the current phase of this study is to take a single-occasion snapshot of a system of variables and constructs. As such, the design is cross-sectional (see MacCallum & Austin, 2000), and not designed to allow time during which the formation of intentions can mediate between social psychological constructs and actual behavior. Future research should allow an examination of these relationships over time, including an examination of autoregressive influences (see Gollob & Reichardt, 1991; MacCallum & Austin, 2000).
KEY VARIABLES CORRELATION MATRIX
|Note: Estimates were based on the covariance matrix; this matrix is shown for interpretive convenience (Boomsma, 2000). See Table 2 for means and standard deviations of these variables.|
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211.
Ajzen, I. (2002). Perceived behavioral control, self-efficacy, locus of control, and the theory of planned behavior. Journal of Applied Social Psychology, 32, 665-683.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3), 411-423.
Arbuckle, J. L. (1999). Amos 4.0 [computer software]. Chicago: Smallwaters.
Boomsma, A. (2000). Reporting analyses of covariance structures. Structural Equation Modeling, 7(3), 461-483.
Byrne, B. M. (2001). Structural equation modeling with Amos: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum.
Conner, M., & Armitage, C. J. (1998). Extending the theory of planned behavior: A review and avenues for further research. Journal of Applied Social Psychology, 28, 1429-1464.
Davis, F. D. (1986). A technology acceptance model for empirically testing new end-user information systems: Theory and results. Doctoral Dissertation, Sloan School of Management, MIT, 1-291.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13, 319-339.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: A comparison of two theoretical models. Management Science, 35(8), 982-1003.
Davis, F. D. (1993). User acceptance of information technology: System characteristics, user perceptions and behavioral impacts. International Journal of Man-Machine Studies, 38(3), 475-487.
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley.
Joreskog, K., & Sorbom, D. (1993). LISREL 8 user’s reference guide. Chicago: Scientific Software, Inc.
Karahanna, E., & Straub, D.W. (1999). The psychological origins of perceived usefulness and ease-of-use. Information and Management, 35(4), 237-250.
MacCallum, R. C., & Austin, J. T. (2000). Applications of structural equation modeling in psychological research. Annual Review of Psychology, 51, 201-226.
Mathieson, K. (1991). Predicting user intentions: Comparing the technology acceptance model with the theory of planned behavior. Information Systems Research, 2(3), 173-191.
Szajna, B. (1996). Empirical evaluation of the revised technology acceptance model. Management Science, 42(1), 85-92.
Taylor, S., & Todd, P. (1995a). Understanding information technology usage: A test of competing models. Information Systems Research, 6(2), 144-176.
Taylor, S., & Todd, P. (1995b). Decomposition and crossover effects in the TPB: A study of computer adoption. International Journal of Research in Marketing, 12, 137-155.
Taylor, S., & Todd, P. (1995c). Assessing IT usage: The role of prior experience. MIS Quarterly, 19(4), 561-570.
Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2), 186-204.
Zubrow, D., Hayes, W., Siegel, J., & Goldenson, D. (1994). Maturity questionnaire. (Special Report CMU/SEI-94-SR-7), Software Engineering Institute, Carnegie Mellon University, Pittsburgh, PA.
©2006 Project Management Institute