S^ FACULTY WORKING PAPER NO. 1129 Planning System Success: Towards Developing and Testing an Operational Model N. Venkatraman Vasudevan Ramanujam tHEWA%$£.n&' ONlVntlfcitT'S -3F ILLINOIS AT' -site College of Commerce and Business Administration Bureau of Economic and Business Research University of Illinois, Urbana-Champaign BEBR FACULTY WORKING PAPER NO. 1129 College of Commerce and Business Administration University of Illinois at Urbana-Champaign March, 1985 Planning System Success: Towards Developing and Testing an Operational Model N. Venkatraman, Visiting Lecturer Department of Business Administration Vasudevan Ramanujam Case Western Reserve University Please do not quote without permission. Comments and criticisms welcome. Digitized by the Internet Archive in 2011 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/planningsystemsu1129venk -1- PLANNING SYSTEM SUCCESS: TOWARDS DEVELOPING AND TESTING AN OPERATIONAL MODEL ABSTRACT A measurement model of Planning System Success is proposed and validated using Joreskog's analysis of covariance structures approach and data from 202 leading North American corporations. Two dimensions — viz., improvements in the capabilities of the planning system and the extent of fulfillment of key planning objectives — are developed and their convergent and discriminant validities are demonstrated. Validated measurement schemes for these dimensions are offered for use in future research on the effectiveness of strategic planning. KEY WORDS: policy/planning, statistics, measurement models, and scales for strategic planning effectiveness. -2- In much of the research on strategic planning systems, the attention given to operationalization and measurement issues has been woefully inadequate. The degree to which a firm is "formalized" in its strate- gic planning practices, for example, has been typically operationalized in terms of categorical variables such as "planner vs. non-planner" (cf. Thune & House, 1970; Karger & Malik., 1975) or "programmed vs. impoverished" planner (cf. Fulmer & Rue, 1973). Such classifications have neither the required discriminatory power (Kudla, 1980) nor are generally reliable and valid (Nunnally, 1973). Similarly, the benefits of strategic planning have been typically evaluated using financial criteria such as Return on Investment, Return on Equity, etc. (cf. Thune & House, 1970), although many con- ceptual writings on strategic planning have emphasized the non- financial benefits (cf. Camillus, 1975; Steiner, 1979) or the "process" benefits of planning (cf. King & Cleland, 1978; King, 1983). As Wood and LaForge (1979) remarked, "It is time to... abandon the smorgasbord use of financial measures as dependent variables and to try to match up the appropriate performance criteria with the primary objectives of the organization being studied" (p. 526). It is increasingly recognized that more rigorous operationalizations of the complex constructs involved in strategic planning systems research is a necessary prere- quisite for theory development and testing in this area. This paper reports the results of a study aimed at developing and testing an operational model of the benefits or success of strategic planning. Development of the model, which includes a broad array of indicators reflecting planning system success is first discussed. Next, -3- the results of testing this model using data on the strategic planning practices of 202 planning units are presented. Finally, the potential use of this model for other researchers interested in furthering stra- tegic planning systems research is elaborated. DEVELOPING AN OPERATIONAL MODEL OF PLANNING SYSTEM SUCCESS Planning System Success is conceptualized in terras of two distinct, but interrelated dimensions — one, the extent of improvement in the capabi lities of the planning system to effectively deliver the support for strategic decision-making, and the other, the extent of fulfill- ment of key planning objectives . The theory underlying these two interconnected dimensions of the model are discussed in the following paragraphs, while Figure 1 depicts the overall operational model. INSERT FIGURE 1 ABOUT HERE Improvement in the Capabilities of Planning System (CAPABILITIES) A planning system can be visualized as a broadly-defined admin- istrative system which provides support for the efficient and effective management of the enterprise. The capabilities of the system then become the key influences on its effectiveness. In a review and criti- que of the appropriateness of various measures of planning effectiveness, Lorange noted that, "... many [of these] measures were based on some surrogate variable, when it probably would have been more relevant to measure effectiveness as a function of how well the formal planning system's capabilities were able to meet the specific planning needs ..." (1979, p. 230, emphasis added). -4- Ideally, the system's capabilities should be considered in relation to the specific needs of the context. However, a broad conceptualiza- tion of a system's major capabilities is developed here for large-scale comparative studies by focussing on a few generic capabilities of planning systems, which have been emphasized in normative and descrip- tive writings on strategy and strategic planning. These capabilities are required of nearly every formal administrative system. They include, but are not limited to, the system's ability to anticipate surprises and crises (Ansoff, 1975), its flexibility to adapt to a dynamic environment (Thompson, 1967), ability to facilitate effective manage- ment control (Anthony, 1965; Lorange & Vancil, 1977), its role in the identification of new business opportunities (Steiner, 1979), as well as its ability to enhance creativity and innovation (Taylor & Hussey, 1982). Based on a review of the literature on strategic planning, 12 key capabilities tapping the above requirement areas were identified. This list was presented to a group of 15 senior-level planning executives who participated in a seminar on strategic planning at the university. This enabled us to assess the "content" validity of the concept, as well as to ensure that these indicators were largely context-free. Such an exercise confirmed that the list was reasonably comprehensive as perceived by planning executives, and that the description of the items was understandable and unambiguous. The list of the 12 items of CAPABILITIES is provided in Table 1. INSERT TABLE 1 ABOUT riERE -o- Extent o£ Fulfillment of Planning Objectives (OBJECTIVES) While the degree of improvement in the system's CAPABILITIES reflect the process dimension of the concept of planning system suc- cess, this dimension is intended to tap the outcome benefits of planning Six key objectives of planning make up the OBJECTIVES dimension. Planning aims to fulfill both tangible and intangible objectives (King & Cleland, 1978; Lorange, 1980; Lorange & Vancil, 1977; Steiner, 1979). Using a goal model of planning success or planning effec- tiveness, the ultimate success of strategic planning can be expected to be reflected in the extent of fulfillment of key planning objectives. These include predicting future trends (Paul, Donavan & Taylor, 1978), enhancing management development through the educational value of the planning process (Hax & Majluf, 1984), evaluating alternatives based on more relevant information (King & Cleland, 1978), as well as improve- ments in financial performance. Here again, the focus was on iden- tifying context-free planning objectives with a balanced mix of both financial and non-financial objectives. The list of six important planning objectives is shown in Table 2. INSERT TABLE 2 ABOUT HERE TESTING THE OPERATIONAL MODEL In the previous section, an operational model of planning system success, with two interrelated dimensions, was conceptually developed. Such a model is not operationally useful unless it is tested against data to establish its measurement properties. The appropriateness of the proposed model's theoretical structure is evaluated using -6- Joreskog's analysis of covariance structures (Joreskog, 1969; 1971; Joreskog & Sorbum, 197b; 1979). Basically, the analysis of covariance structures enables one to test the degree of correspondence between the theoretical raodel(s) and its operationalization, and can be used to assess reliability and also different components of validity such as convergent and discriminant validity, predictive validity, etc. This analytical scheme has been employed to test a variety of measurement models in marketing (cf. Bagozzi , 1980) and in other disciplines (cf. Fornell, 1982). Increasingly, this analytical scheme is also being adopted in strategy research for testing measurement models (cf. Farh, Hoffman, & Hegarty, 1984) as well as substantive relationships (cf. Phillips, Chang, & Buzzell, 1983). Data The data for this study were drawn from a larger project on the changes and effectiveness of strategic planning systems of large North American Corporations. Data were collected using a structured self- administered mail questionnaire from 202 planning units between February and April 1984. This represents a response rate of nearly 33 percent of the 600 target planning units randomly selected from the Fortune 1000 list of manufacturing and service firms. Table 3 presents some key characteristics of the study sample. INSERT TABLE 3 ABOUT HERE Overview of Model Testing The testing of the operational model involved two steps. First, the adequacy of the two dimensions was independently assessed. Next, -7- the relationship between the two dimensions was evaluated. Four models were evaluated in this two-step process. The first test (Model 1) aimed at ascertaining the extent to which the 12 indicators reflect the theoretical dimension CAPABILITIES. The second test (Model 2) was a similar examination of the theoretical dimension, OBJECTIVES. Thus, Models 1 and 2 explored the convergent validity of the two dimensions. The third test (Model 3) examined whether these dimensions are indeed distinct dimensions, and this is a test of discriminant validity. Finally, Model 4 examines the nature of the relationship between the two dimensions, i.e., it tested the predictive validity of the two dimensions. The analytical details of testing these models and the results are provided below. Model 1: Convergent Validity of the CAPABILITIES Dimension Following Joreskog's work and conventions of structural equation modeling, this model for convergent validity is written as: X - A£ + <5 (1) where X is a vector of P measurements, £ is a K < P vector of traits, 5 is a vector of unique scores (random errors), and A is a PXK matrix of factor loadings. With the assumptions of E(£) = E(6) = 0; E(££') = , and E(66') = i|>, the variance-covariance matrix of X can be written as 2 = AA' + \\> (2) where I is the variance-covariance matrix of observations, is the matrix of intercorrelations among the traits, and \\i is a diagonal -8- matrix of error variances (6,) for the measures. For Model 1, K=l . and P=12 as shown in Figure 2. INSERT FIGURE 2 ABOUT HERE 2 Maximum likelihood parameter estimates (mLE) for A, ,^, and a x goodness-of-f it index for the null model implied by equations (1) and (2) can be obtained from the LISREL Program (Joreskog & Sorbura, 1978). 2 The probability level associated with a given x statistic indicates 2 the probability (p) of attaining a larger x value given that the hypothesized model (Figure 2) is supported. The higher the value of p, the better is the fit, and as a rule of thumb, values of p > 0.10 are considered as indications of satisfactory fit (Lawley & Maxwell, 1971). The base model (Figure 2) was estimated using LISREL, and the 2 1 resulting statistics were: x~( df;5 ^) = 189.1616; p = 0.00. This indicates that the model as hypothesized in Figure 2 should be re- 2 jected. However, exclusive reliance on the x statistic is criticized for many reasons (cf. Fornell & Larcker, 1981), and researchers increasingly complement this statistic with Rentier and Bonnett's (1980) incremental fit index A — which is an indication of the practical significance of the model in explaining the data. The A index is represented as follows A = 'F - F )/F (3) 1 k y A matrix of zero-order correlations of the 18 indicators can be obtained by writing to the first author. -9- where F rt = chi-square value obtained from a null model specifying mutual independence among tne indicators, and F = chi-square value for the specific model. The A value for this model was 0.83, indi- cating that the model should be rejected, since as a rule of thumb A should be greater than 0.90 (Bentler & Bonnett, 1980), although some argue that it should ideally exceed 0.95 (Bearden, Sharma & Teel, 1982). The rejection of the model shown in Figure 2 implies that all the variation and covariation in the measurement of the underlying construct cannot be represented as trait variance plus random error variance only 2 (cf. Bagozzi, 1980). However, an examination of the residual matrix (the difference between the sample variance-covariance matrix and the model-fitted variance-covariance matrix) indicated that other nonrandom factors may be causing variation in the measurement. As Joreskog and 2 Sorbum (1979) noted, "...the x goodness-of-f it-values can be used as follows. If a value of x * s obtained which is large compared to the number of degrees of freedom, the fit may be examined by an inspection of the residuals, that is the discrepancies between observed and repro- duced variances and covariances. The result of an analysis in conjunc- tion with subject-matter considerations may suggest ways to relax the model somewhat by introducing more parameters. The new model yields a 2 2 smaller x • A larger drop in x > compared to the difference in degrees 2 of freedom, supports the changes made. On the other hand, a drop in x 2 ^Residual matrices for this model as well as other models tested in this study are not presented here; interested readers may contact the first author. -10- which is close to the difference in number of degrees of freedom indi- cates that the improvement in fit is obtained by capitalizing on chance" (emphasis added). Theoretical justifications can be provided for only eight sets of covariation in error terms, where the entries in the residual matrix exceeded 0.10. These are indicated by (2,1), (3,2) (10,2) (8,3) (6,4) (8,5) (12,6) and (8,7), where numbers refer to the indicators of Exhibit 1. By referring to Exhibit 1, one can readily see that each of these sets of items share a common theme. As an illustration, items 2 and 1 both refer to environmental shifts, while items 3 and 2 reflect a firm's ability to exploit opportunities presented in the environment by adapting to environmental changes. The rationale for introducing such correlated errors into the model is that the original assumption of treating the 12 indicators as independent of one another may be too restrictive, and does not truly represent the underlying model structure (cf. Joreskog & Sorbum, 1979). The model presented in Figure 2 was re-estimated by incorporating the additional specification of these eight sets of correlated errors. This model provided a better fit to the data, with the associated model 2 2 statistics of x (df:46) = 62.2686; p = 0.0551; A = 0.94. Tne x d value was 126.893, statistically significant at p < 0.01. A p-value of 0.055 indicates a "marginal" fit and has been previously used to accept models (cf. Bagozzi, 1981; Phillips, Chang, & Buzzell, 1983). The p-value of 2 0.055, a significant value of x. > an 0.10 (Lawley & Maxwell, 1971) and A > 0.95 (Bearden et. al , 1982), are all satisfied indicating the acceptance of the model shown in Figure 3 with corre- lated errors between indicators 6 and 5. Table 5 presents a summary of INSERT TABLE 5 ABOUT HERE the model statistics, the ML estimates for the parameters, as well as the value of p for the model. All the individual model parameters are c statistically significant as indicated by the corresponding t-values, being larger than 1.96. 3 An alternative representation to the base model, hypothesizing that OBJECTIVES is a two-dimensional model, with financial objectives and non-financial objectives modeled as separate, but correlated dimen- sions. The estimation of this model yielded x (df:8) = 18.6781; p = 0.0167; A = 0.930. The difference between this model and the base model was x5^ df:1 ^ = °» 547 3, not significant. Hence the alternative model of separately specifying financial objectives and non-financial objectives was rejected. -13- Model 3: Discriminant Validity of the Two Dimensions Thus far, we have treated the hypothesized two dimensions of the model separately and evaluated whether the different indicators reflect the respective dimensions or not. A rival explanation which could be raised at this stage is that these two dimensions are merely sub-dimensions of an overall construct, and that they should not be considered as distinct dimensions. Since the indicators have shades of common meaning, one could conceivably argue that the improvement in system's capabilities and objective fulfillment are not distinct dimensions. In other words, a test of discriminant validity is necessary for rejecting this rival explanation. As noted by Bagozzi (1980), the strongest evidence of discriminant validity is obtained when maximally (conceptually) similar traits are used. Since the two dimensions appear to be conceptually similar, a test of discriminant validity should provide strong support for rejecting the rival explanation that these two dimensions are the same. Discriminant validity is achieved when the measures of each dimen- sion converge on their corresponding true scores which are unique from other dimensions. Stated differently, it is the degree to which a theoretical dimension in a theoretical system differs from other dimensions in the same theoretical system. This will be achieved when the correlations between the dimensions (<|>s) are significantly lower than unity. This requires a comparison of a model shown in Figure 4 with a similar model with the correlation (<£ 91 ) constrained to be equal 2 to unity. A significantly lower x value for the model with the -14- uncons trained correlation when compared with the constrained model pro- vides support for discriminant validity. A x difference value (x ) d with an associated p-value less than 0.05 (cf. Joresltog, 1971) supports the discriminant validity criterion. Figure 4 represents both models (i.e., constrained and unconstrained) with their model statistics. INSERT FIGURE 4 ABOUT HERE As indicated in Figure 4, the x H va lue of 94.1868, p < 0.001 strongly supports the discriminant validity hypothesis and thus rejects the rival explanation that the two dimensions are to be treated as one composite dimension. Figure 4 also presents the results of an additional test conducted to eliminate this rival explanation. In this test, an overall composite model represented by 18 indicators was com- pared with the unconstrained model of Figure 4 that they are two separate, 2 and related dimensions but not one composite dimension. A x ,(df:l) value of 104.51, p < 0.001 further rejects the rival explanations of a composite model. These tests provide strong support to the concep- tualization of planning system success in terras of the two separate dimensions as shown in Figure 1. Model 4: An Examination of Predictive Validity While a two-dimensional operational model of planning system suc- cess has been developed and tested based on criteria of convergent and discriminant validity, the nature of the relationship between the two dimensions has not yet been specifically examined. This can be tested by hypothesizing that an improvement in system's CAPABILITIES will result in higher levels of OBJECTIVE fulfillment, and is termed as an -15- exami nation of predictive validity. The theoretical support for expecting such a relationship can be derived from discussions on the central role of strategic planning in realizing organizational objec- tives (see especially, King & Cleland, 1978; Lorange & Vancil, 1977) as well as the specific notions of system's capabilities (Lorange, 1979) and strategic capability (Lenz, 1980) which influence an organization's strategic actions, which in turn results in the attainment of organiza- tional objectives. Predictive validity is tested using the model shown in Figure 5. The structural equation for this model is written as: n * re + ? (5) where, n = endogenous theoretical construct, r = matrix of structural coefficients relating exogenous theoretical construct to endogenous theoretical construct, z, = residuals of endogenous theoretical construct. The standardized gamma (y) value of the impact of CAPABILITIES on OBJECTIVES is 0.631 lending strong support to the positive effect of 2 CAPABILITIES on OBJECTIVES. The relatively high value of x ? df 125 ) = 237.1167, p = 0.00, A = 0.85 indicates that there are factors in addi- tion to CAPABILITIES which influence the fulfillment of objectives. This is consistent with the theoretical expectation that many facets of strategic planning have important roles in ensuring planning effective- ness. However, since the present focus is on examining the relation- ship between these two dimensions, rather than modeling planning effec- tiveness, we focus on the significance of y . and not on the overall model fit. -16- INSERT FIGURE 5 ABOUT HERE DISCUSSION In this study, we attempted to develop and test an operational model of Planning System Success. The model includes two concepts, viz., (i) improvements in the strategic planning system capabilities and (ii) the extent of fulfillment of key planning objectives. Generic and context-free indicators of CAPABILITIES and OBJECTIVES to develop and test a model which can be applied in large sample studies. The discussion in this section focuses on four issues. First , the results provide strong support for the measurement properties of the two dimensions. Specifically, the operational model was evaluated in terms of (i) reliability criterion (p ), (ii) convergent and discrimi- nant validity (models 1, 2, and 3), and (iii) predictive validity (model 4). Since all these criteria were found to be satisfied, the measure- ment scheme presented here could either be directly employed in future research on strategic planning or can be used as the basis for further refinement and extensions. Second , it needs mention that the analytical scheme employed here, viz., structural equation modeling approach (Joreskog & Sorbum, 1978) is not the only available analytical scheme. Although its advantages are apparent in certain research designs (see Bagozzi, 1980, Joreskog & Sorbum, 1979 for detailed discussions), other analytical schemes are available (e.g., partial least square estimation of Wold, 1982). Further, to aid readers to evaluate the measurement properties, the Cronbach a values for the two dimensions are provided. These are: -17- CAPA3ILITIES - 0.871, and OBJECTIVES - 0.748, which indicate accept- able levels of reliability (Nunnally, 1978). In addition, acceptable levels of factor loadings (viz, As reported in Tables 1 and 2) augur well for the use of these indicators in future research. However, use of the structural equation modeling approach enables researchers to explicitly model measurement error, correlate measurement errors when theoretically appropriate, and thereby evaluate relationships between theoretical constructs under less restrictive conditions than explora- tory factor analysis and ordinary least square regression approaches (see Bagozzi & Phillips, 1982 for a comparative discussion). The third issue relates to a limitation of the study in terms of employing a single respondent per unit of analysis. Although the respondents were senior-level managers such as Presidents, Vice Presi- dents - Corporate Planning, and Vice President of functional areas of large corporations (over 60% had sales in excess of $1 billion — see Exhibit 3), measurement focused at an organization-level of analysis would be better served if data were collected from multiple respondents to assess inter-judge consistency. This is noted as an area for future research. Fourth , it is believed that the two-dimensional measuring scheme for Planning System Success presented here should be of value and use to other researchers interested in the research stream of strategic planning effectiveness. Although the CAPABILITIES dimension emerged as a strong predictor of objective fulfillment, we urge that both dimen- sions be employed since they represent different, but related, notions of planning-success . However, measurement schemes are merely first -13- steps towards testing substantive relationships, and by presenting a set of reliable and valid scales for planning system success, we hope that we would have stimulated some interest among researchers to address a broader and a more important question: What are the key determinants of planning system success? Specifically, it would be interesting and useful to examine if the determinants of the two dimen- sions are same or not. While it was shown that the capabilities dimen- sion is distinct from the objectives dimension, further support for such a two-level scheme can be derived if the determinants of these dimensions are indeed different. CONCLUSIONS By noting that an appropriate operationalization of the theoreti- cal construct of Planning System Success is necessary for theory devel- opment and testing in the area of strategic planning systems, this paper developed and tested a two-dimensional measurement scheme. Based on data on the planning practices of 202 planning units, and adopting a data-analytic framework rooted in Joreskog's analysis of covariance structures, key measurement criteria for the operational model were found to be satisfied. This should serve as a useful guide for future strategy researchers interested in testing various propositions on strategic planning effectiveness, especially the question: What are the key factors that lead to planning system success? -19- REFERENCES Ansoff, H. I. Managing strategic surprise by response to weak signals. California Management Review , 1975, 18 (Winter), 21-33. Anthony, R. N. Planning and control systems: A framework for analysis Division of Research, Graduate School of Business Administration, Harvard University, 1965. Bagozzi, R. P. Causal models in marketing , 1980, NY: Wiley. Bagozzi, R. P. An examination of the validity of two models of atti- tude. Multivariate Behavioral Research , 1981, 16, 323-359. Bagozzi, R. P., & Phillips, L. W. Representing and testing organiza- tional theories: A holistic construal. Administrative Science Quarterly , 1982, 27, 459-489. Bearden, W. 0., Sharma, S., & Teel, J. E. Sample size effects on chi square and other statistics used in evaluating causal models. Journal of Marketing Research , 1982, 19, 425-430. Bentler, P. M. , & Bonnett, D. G. Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin , 1980, 88, 588-606. Camillus, J. C. Evaluating the benefits of formal planning. Long Range Planning , 1975, 8(3), 33-40. Farh, J. L. , Hoffman, R. C, & Hegarty, W. H. Assessing environmental scanning at the subunit level: A multitrait multimethod analysis. Decision Sciences , 1984, 15, 197-220. Fornell, C. (Ed.). A second generation of multivariate analysis , 1982, NY: Praeger Publishing. Fornell, C, & Larcker, D. F. Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research , 1981, 18, 39-50. Fulmer, R. M. , & Rue, L. W. The practice and profitability of long range planning , 1973, Oxford, Ohio: The Planning Executives Institute. Hax, A. C. & Majluf, N. S. Strategic management: An integrative perspective . Englewood Cliffs, N J : Prentice-Hall, 1984. Joreskog, K. G. A general approach to confirmatory maximum likelihood factor analysis. Psychometrika, 1969, 34, 183-202. -20- Joreskog, K. G. Statistical analysis of congeneric tests. Psychometri ka, 1971, 36, 109-133. Joreskog, K. G. , & Sorbum, D. Analysis of linear structural relation- ships by method of maximum likelihood , 1978, Chicago, 111.: National Educational Resources. Joreskog, K. G. , & Sorbum, D. (Eds.). Advances in factor analysis and structural equation models , 1979, Mass: Abt Books. Karger, D. W. , & Malik, Z. A. Long range planning and organizational performance. Long Range Planning , 1975, 8(6), 60-69. King, W. R. Evaluating Strategic Planning Systems. Strategic Manage - ment Journal , 1983, 4, 263-277. King, W. R. , & Cleland, D. I. Strategic planning and policy . New York: Van Nostrand, Reinhold, 1978. Kudla, R. J. The effects of strategic planning on common-stock returns. Academy of Management Journal , 1980, 23, 5-20. Lawley, D. N. , & Maxwell, A. E. Factor analysis as a statistical method , 1971, London: Butterworth. Lenz, R. T. Strategic capability: A concept and framework for analy- sis. Academy of Management Review , 1980, 5, 225-234. Lorange , P. Formal planning systems: Their role in strategy for- mulation and implementation. In D. Schendel and C. Hofer (Eds.), Strategic management: A new view of business policy and planning , 1979, 226-241, Boston: Little, Brown, & Co. Lorange, P., & Vancil, R. F. Strategic planning systems , 1977, NY: Prentice-Hall . Nunnally, J. C. Psychometric theory . New York: McGraw-Hill, 1978. Paul, R. N. , Donavan, N. B., & Taylor, J. W. The reality gap in strategic planning. Harvard Business Review , 1978, 56(4), 124-130. Phillips, L. W. , Chang, D. R. , & Buzzell, R. D. Product quality, cost position, and business performance: A test of some key hypotheses. Journal of Marketing , 1983, 47 (Spring), 26-43. Steiner, G. A. Strategic planning: What every manager must know , 1979, New York: Free Press. Taylor, 3., & Hussey, D. The realities of planning . NY: Pergamon Press, 1982. -21- Thorapson, J. D. Organizations in action . New York: McGraw-Hill, 1967. Thune , P. S., & House, R. Where long range planning pays off? Business Horizons , 1970, 81-89. Werts, C. E., Linn, R. L. , & Joreskog, K. G. Interclass reliability estimates: Testing structural assumptions. Educational and Psychological Measurement , 1974, 34, 25-33. Wold, H. Systems under indirect observation using PLS. In C. Fornell (Ed.) A second generation of multivariate analysis . Vol. I. Methods. 1982, New York: Praeger, 325-347. Wood, D. R. , & LaForge , R. W. The impact of comprehensive planning on financial performance. Academy of Management Journal , 1979, 22, 516-526. D/295 -22- TABLE 1 KEY CAPABILITIES OF PLANNING SYSTEM 3 1. Ability to anticipate surprises and crises. 2. Flexibility to adapt to unanticipated changes. 3. As a mechanism for identifying new business opportunities. 4. Role in identifying key problem areas. 5. As a tool for managerial motivation. 6. Role in the generation of new ideas. 7. Ability to communicate top management's expectation down the line. 8. As a tool for management control. 9. As a means for fostering organizational learning. 10. Ability to communicate line manager's concerns to Che top management . 11. As a mechanism for integrating diverse functions and operations 12. As a basis for enhancing innovation. Each indicator was measured using a five-point interval scale ranging from much improvement (+2) to much deterioration (-2), to cap- ture the general trend of changes. -23- TABLE 2 MAJOR OBJECTIVES OF PLANNING SYSTEM 5 1. Enhancing management development. 2. Predicting future trends. 3. Short-term performance. 4. Long-term performance. 5. Evaluating alternatives based on more relevant information. b. Avoiding problem areas. Each indicator was measured using a five-point interval scale ranging from entirely fulfilled (+2) to entirely unfulfilled (-2). -24- TABLE 3 Key Charactieristics of the Study Sample (n=202) 1 . Level of the Planning Unit Corporate level 81% Business unit level 19% 2. Title/Job Position of the Respondent Planning Responsibility (e.g., Vice President- 69.2% Corporate Planning) Operating (line) Responsibility (e.g., President, 30.8% Vice President of functional areas) 3. Range of Sales Less than $50 million 6.6% $51 - $100 million 4.6% $101 - $250 million 5.1% $251 - $500 million 10.2% $501 million - $1 billion 12.2% over $1 billion 61.4% 4. Business Category Consumer Goods 19.1% Capital Goods 19.1% Raw or semi-finished materials 13.1% Components 9.0% Service Sector 39.7% -25- TABLE 4 SUMMARY STATISTICS OF MODEL-TESTING FOR THE "CAPABILITIES" DIMENSION (A) Base Model (B) Model with Correlated Errors x 2 (df:54) - 189.1616 x 2 ( df:4 6) = 62.2686 p = 0.000 p = 0.0551 A = 0.83 A = 0.94 (C) ML Parameter Estimates Parameter ML Estimate t-value Standardized Solution h 1.00* — 0.504 X 2 0.996 7.527 0.502 V 1.112 5.888 0.560 X 4 1.293 6.406 0.651 l 1.431 6.771 0.721 X, 1.449 6.799 0.730 X 7 1.358 6.598 0.684 X 8 1.209 6.171 0.609 X 9 1.517 6.962 0.764 X 10 1.239 6.281 0.624 X ll 1.367 6.623 0.689 X 12 1.282 6.378 0.646 ♦ll 0.254 3.633 1.000 *Constrained parameter. -26- TABLE 5 SUMMARY STATISTICS OF MODEL-TESTING FOR THE "OBJECTIVES" DIMENSION (A) Base Model (B) Model with Correlated Errors X 2 (df:9) = 19.2254 x 2 (df:8) = 7.7814 P = 0.0234 P = 0.4551 A = 0.927 A = 0.97 p = 0.750 c (C) ML Parameter Estimates Parameter ML Estimate t-value St andardized Solution h 1.00* — 0.717 X 2 0.804 6.621 0.576 X 3 0.633 5.386 0.454 \ 0.927 7.334 0.665 X 5 0.751 6.157 0.539 X 6 0.779 6.363 0.559 ♦u 0.514 4.996 1.000 * Cons trained parameter. FIGURE 1 P LANNIN G SYSTE M SUCCES S: A SCHE MATIC REP RESENTATION OF THE TWO-DIMENSIONAL MODEL PLANNING SYSTEM SUCCESS IMPROVEMENT IN SYSTEM'S CAPABILITI E S {CAPABILITIES} EXTENT FULFILLMENT OF OF OBJECTIVES {OBJECTIVES} -28- FIGURE 2 A MODEL OF CONVERGENT VALIDITY OF THE "CAPABILITIES" DIMENSION 3 The notations ot structural equation modeling are followed in the diagram, where the latent (unobservable) variable or theoretical con- struct is drawn as an ellipse; observable indicators are presented as squares; measurement relations are shown as arrows; error factors are represented as arrows but without origin. \s represent the degree of correspondence between observed indicators and unobserved theoretical construct. -29- FIGURE 3 A MODEL OF CONVERGENT VALIDITY OF THE "OBJECTIVES" DIMENSION 3 FULFILLMENT ( OBJECTIVES / i For detailed explanation of the notations, see Figure 2. -30- FIGURE 4 A MODEL OF DISCRIMINANT VALIDITY OF THE TWO DIMENSIONS A. Unconstrained Model X (df:125) = 237.1167; p = 0.000; = 0.631 B. Constrained Model x (df:126) = 331.3035; p = 0.000; X (df:l) = 94.1868; p < 0.001 supports the unconstrained model C. Alternative Model CAPABILITIES + OBJECTIVES 18 indicators X (df:126) = 341.6312, p = 0.00 Only a skeletal diagram is drawn for schematic clarity. The respective models for the two dimensions are the same as shown in Figures 2 and 3 with relevant correlated errors discussed in the text. -31- FIGURE 5 AN EXAMINATION OF THE PREDICTIVE VALIDITY OF THE TWO DIMENSIONS CAPABILITIES r 11 OBJECTIVES n, 12 indicators 6 indicators x "(df :125) P A 237.1167; 0.00 0.85 Y n = 0.631 std. Only the skeletal diagram is drawn for schematic clarity; the respective models for the two dimensions are as shown in Figures 2 and 3 with relevant correlated errors discussed in the text.