The role of context and structure in radical and incremental logistics innovation adoption

ELSEVIER The Role of Context and Structure in Radical and Incremental Logistics Innovation Adoption Richard Germain OKLAHOMA STATE UNIVERSITY An analysis of manufacturers was undertaken to examine the adoption of lo~stics process innovation. A typology of innovation was created on innovation cost and radicalness. The results reveal that size and environmental uncertainty directly predict expensive, radical but not low-cost, incremental innovation. Specialization predicts both. Decentralization of logistics pro- cess innovation adoption decision-making predicts low-cost, incremental innovation but not expensive, radical innovation, whereas decentralization of manufacturing operations does not predict logistics process innovation. Finally, although integration does not predict low-cost, incremental innovation, it inversely predicts high-cost, radical innovation, j BUSN RES 1996. 35.117--127 ogistics can be defined as the organizational-wide flow of goods and related information encompassing physical distribution and materials management. This research focuses on the organizational adoption of logistical process innovation. Relevant innovations include both software and hardware such as inventory control and transportation sched- uling software systems, robotics, and bar codes. Logistics costs among U.S. manufacturers average 20% of sales (Glaskowsky, Hudson, and Ivie, 1992), suggesting that innovations in logistics can improve organizational performance. Indeed, the decline in the ratio of business inventories to gross national product from about 25% in 1980 to 20% in 1990 can be attributed to pull marketing systems that are reliant on process innovation (Coyle, Bardi, and Langley, 1992). The proclivity to innovate is affected by factors external and internal to the innovation. External factors include organiza- tional structure (Hage and Dewar, 1973; Moch, 1976), manage- ment traits (Howell and Higgins, 1990; Kimberly and Evanisko, 1981), context and strategy (Ettlie, 1983), production process (Collins, Hage, and Hull, 1988), other innovation adoption Address correspondence to: Richard Germain, 216 College of Business Administration, Oklahoma State University, Stillwater, OK 74078-0555. Journal of Business Research 35, 117-127 (1996) © 1996 Elsevier Science Inc. 655 Avenue of the Americas, New York, NY 10010 (Damanpour, Szabat, and Evan, 1989), and culture (Hoffman and Hegarty, 1993). Internal factors address adoption phase (Zmud, 1982), innovation radicalness, risk, compatibility, cost (Meyer and Goes, 1988; Moch, and Morse, 1977), and an ad- ministrative versus a technological innovative nature (Daman- pour, 1987). The purpose of this research is to examine the effect of a set of internal and external factors on logistical process inno- vation. The internal factors of innovation cost and radicalness are modded. Radical innovations are important because of their impact on performance (Nord and Tucker, 1987), yet only a handful of empirical studies exist that address this subject (Dewar and Dutton, 1986; Etthe, 1983; Etthe, Bridges, and O'Keefe, 1984). Two of the most widely studied predictors of innovation, context and organizational structure, are modeled as external factors (e.g., Damanpour, 1991). Figure 1 presents the theoretical framework of the study. In the framework: (1) context has a direct effect on innovation; and (2) structure mediates context --" innovation rdationships. In the section that follows, a hypothesis relating innovation cost and radicalness is proposed. Then, hypotheses are developed connecting structure with innovation, context with innovation, and context with structure. Subsumed within the framework is the structuralist perspective that organizational structure fosters or impedes innovation (Collins, Hage, and Hull, 1988). Relevant Literature Innovation radicalness can be defined as a trait that results in fundamental or significant change in inputs, outputs, or pro- cesses (see Hage, 1980, pp. 188-199). Hage (1980) explicitly noted that: (1) cost, radicalness, and divisibility are the three most important innovation properties; and (2) because radical innovations embody new knowledge and represent dear depar- tures from past practice, they tend to be both costly and risky. For example, automated storage and retrieval systems (AS/KS), which cost in the hundreds of thousands or millions of dollars ISSN 0148-2963/96/$15.00 SSDI 0148-2963(95)00053-U 118 J Busn Res R. Germain 1996:35:117-127 1 H5, H6 T CONTEXT size environmental uncertainty 1 ! H7, H8 ORGANIZATIONAL STRUCTURE specialization decentralization integration I H2-H4 1 INNOVATION low cost, incremental innovation high cost, radical innovation Figure 1. Theoretical framework. and handle over 100,000 units of inventory, result in fundamen- tal, radical change in warehouse and production processes. These systems improve service (e.g., fewer order errors, quicker order picking) and cut costs (e.g., elimination of staff, reduc- tion of floor space by as much as 80%) (Forger, 1993a, 1993b; Rees, 1994). Because the cost involved is high and the benefits sensitive to the unique context of a firm are uncertain, AS/RS technology is subjected to pre-adoption simulation to demon- strate how adoption will shift costs and operations (Appliance Manufacturer, 1994). This illustrates the uncertainty surround- ing the adoption of costly, radical innovation. The first hypoth- esis we propose is straightforward: innovation cost and radi- calness are associated positively. HI: Logistics innovation cost and radicalness are associated positively. H1 is critical relative to the development of further hypoth- eses because it implies a single cost/radical continuum exists. Given such, two sets of innovations may exist at opposite ends of the continuum: i.e., a set of expensive, radical and a set of low-cost, incremental logistics process innovations. For the sake of parsimony, the former is hereafter referred to as radical in- novation and the latter is referred to incremental innovation. Effect of Organizational Structure on Innovation Literature reviews have identified four latent dimensions of or- ganizational structure: specialization, decentralization, integra- tion, and formalization (Champion, 1975; Miller and Dr6ge, 1986; Mintzberg, 1979). Specialization can be defined as the level of expertise or knowledge within the firm. An organiza- tion with at least one employee dealing exclusively with a par- ticular knowledge area (such as sales forecasting) is assumed to be more specialized than the firm with no employees devoted exclusively to that knowledge area. Decentralization refers to the vertical locus of decision-making authority. The lower that decisions are made in an organization, the more decentralized the structure. Integration can be defined as lateral communica- tion that spans divisions, functions, or departments. Integra- tion can be accomplished through a number of mechanisms including cross-functional committees, liaison personnel, and task forces. Finally, formalization refers to written communi- cations and rules. A meta-analysis of structure's effect on inno- vation demonstrated that specialization and decentralization, but not formalization, predict innovation (c.f. Damanpour, 1991). For this reason, formalization is excluded from the study. Because of its infrequent inclusion in innovation research, Damanpour's (1991) meta-analysis did not examine integra- tion. We include integration in this research, however, because it may differentially impact radical versus incremental inno- vation. SPECIALIZATION. The greater the specialization, the greater the firm's knowledge base and cross-fertilization of ideas and the more likely the firm is to innovate (Damanpour, 1991; Kim- berly and Evanisko, 1981; Aiken and Hage, 1971). Although specialization differentially predicts compatible versus incom- patible (Moch and Morse, 1977) and technological versus ad- ministrative innovation (Damanpour, 1987; Kimberly and Evanisko, 1981), it equally predicts radical and incremental in- novation. Dewar and Dutton (1986) expected specialization to be unrelated to incremental innovation as it requires limited knowledge and information. Such a requirement provides specialists and technicians with little incentive to tackle iden- tification and implementation problems. However, their results did not support this expectation. Rather, specialization was found to have a pervasive impact on innovation. Specialization is included in the model in order to: (1) demonstrate that not all dimensions of structure differentially impact incremental and radical innovation; and (2) determine the consistency of the empirical finding that specialization universally predicts inno- vation regardless of cost/radicalness. H2: Specialization is positively related to incremental and radical innovation. DECENTRAUZATION. The greater the decentralization, the greater the innovativeness because participatory structures foster or- ganizational member involvement, awareness, and commitment. Despite meta-analytic support for the decentralization ~ in- novation relationship (Damanpour, 1991), previous research is characterized by weak effects and inattention to the decen- tralization domain. Decentralization domain refers to different vertical loci for varying decision-making areas (Carter and Cul- Context and Structure in Logistics J Busn Res 119 1996:35:117-127 len, 1984; Jennergren, 1981; see also Mintzberg, 1979, p. 187, who called it "selective decentralization"). The effect of selec- tive decentralization on production process innovation has been examined. For example, line versus operating decentralization has been studied in relation to production process innovation (Collins, Hage, and Hull, 1988) but not in relation to radical versus incremental innovation. Dewar and Dutton (1986) argued that decentralization pro- motes the initiation phase of adoption by exposing decision- makers to information (Hage and Aiken, 1970), whereas cen- tralization promotes innovation by dampening ambiguity and conflict (Zaltman, Duncan, and Holbek, 1973). They suggested that the more centralized the firm, the more likely it would be to adopt radical innovation and the less likely it would be to adopt incremental innovation. If innovations are incremental (low in cost and in risk), and involve little new knowledge, then a minimum of resistance may emerge from supervisors. The adoption of radical innovation, resulting in a fundamental work pattern reconfiguration, may alienate affected organizational con- stituencies. Centralized decision-making may be required to overcome the resultant resistance. But, Dewar and Dutton (1986) found that decentralization predicted neither radical nor incremental innovation. The aforementioned argument must be weighed against Hage's (1980, p. 189) position that radical change meets resistance in centralized organizations because "vested interests" fear losing power, prestige, or status. We propose that modeling more than one decentralization domain can overcome the deficiencies of previous research. In particular, we examine innovation adoption decentraliza- tion and manufacturing operations decentralization. No cross- over effect is hypothesized between decentralization in manufac- turing operations and innovation (regardless of innovation cost or radicalness) because of the divergence of underlying domains (i.e., manufacturing operations versus logistics innovation). In contrast, decentralized responsibility for innovation should positively predict incremental innovation while failing to pre- dict radical innovation. The effect of decentralization on inno- vation when domains converge should not extent to radical in- novation, because control over finance is the most centralized domain in organizations (Thompson, 1969). When the domain of decentralization and behavior converge, radical innovation may not follow because centralized financial control may stand between the decision-maker and implementation. In contrast, little resistance may emerge from centralized financial planners over incremental innovation. H3: Innovation adoption decentralization is positively re- lated to incremental innovation and unrelated to radi- cal innovation, whereas manufacturing operations de- centralization is unrelated to logistics innovation regardless of its radicalness. INTEGI~TION. With the exception of early research on organiza- tional innovation (e.g., Aiken and Hage, 1971), the effect of in- tegration on innovation has largely been ignored. Radical in- novations, despite their ability to create a sustainable competitive advantage, are apt to meet with resistance precisely because of the cost and uncertainty that enshroud adoption (Hage, 1980). In addition to resistance from centralized financial control, re- sistance may also emerge from other organizational functions. Lateral communications can be used to preserve a function's power, to conserve scarce resources, and to limit the ability of radical innovations to affect cross-functional operations. For example, manufacturing may oppose bar code adoption by logis- tics if the innovation is perceived of as siphoning away re- sources, impacting decision-making in production, or increas- ing the expert power of logistics. Because of a concern for power and prestige, infighting may be evidenced by disagreements over which division will introduce and control new, innovative prod- ucts (Burns and Stalker, 1961). In contrast, integration should play no role in incremental innovation, because this type of innovation has little impact on relative cross-functional finance, power, status, and prestige. H4: Integration is unrelated to incremental innovation and inversely related to radical innovation. Effect of Context on Innovation Context defines the short-run operating situation of an organi- zation, and the text that follows will examine two of the most common variables. Size can be defined as the operating scale of the organization and environmental uncertainty as external dynamism and unpredictability (Duncan, 1972). Size has consistently predicted innovation among nonprofit institutions such as educational establishments (Baldridge and Burnham, 1975), hospitals (Kimberly and Evanisko, 1981), libraries (Damanpour, 1987), and profit-oriented organizations such as newspaper publishers (Carter, 1984) and manufacturers (Blau et al., 1976; Germain, 1993). Size is associated with the presence of slack resources, specialists, engineers, and techni- cians- each contributing to the ability of larger organizations to invest in innovation (Thompson, 1969). However, larger or- ganizations possess slack resources that obviate research failures associated with radical innovations and risky departures from past practice. In contrast, organizations, regardless of their size, can afford the cost of incremental innovation and can absorb the consequence of adoption failure. Consistent with Dewar and Dutton's (1986) findings, size is expected to predict radi- cal but not incremental innovation (Dewar and Dutton, 1986). Uncertainty also fosters innovation. Frequent external change results in organizations that are more accustomed to adjusting to turbulence, more oriented toward the future, and more aware of external innovations and cues (Gatignon and Robertson, 1989; Ettlie, 1983; Pierce and Delbecq, 1977). External shocks make organizations more amenable to radical innovations (Hage, 1980). Piecemeal solutions or those of an incremental nature provide limited gains against variability and limited opportu- nity for creating a sustainable competitive advantage. Environ- mental uncertainty is thus expected to predict radical but not incremental innovation. 120 J Busn Res R. Germain 1996:35:117-127 H5: Size is unrelated to incremental innovation and posi- tively related to radical innovation. H6: Environmental uncertainty is unrelated to incremental innovation and positively related to radical innovation. Effect of Context on Structure In order to test structure as a mediator of context ~ innova- tion relationships, it is necessary to generate hypotheses relat- ing context to structure. Larger organizations are more special- ized, integrated, and decentralized. Specialization increases with size because of the subdivision of labor (Mintzberg, 1979). Decision-making delegation reduces control costs in larger firms by reducing senior manager burdens, and larger firms thus tend to be more decentralized (Child, 1972, 1973; Moch, 1976). Larger firms are more compartmentalized, and one means of overcoming vertical communication and isolationism is to in- stitute lateral integration (Mintzberg, 1979). Organic structures that are more specialized, decentralized, and integrated result from environmental uncertainty (Burns and Stalker, 1961; Galbraith, 1973; Lawrence and Lorsch, 1967; Mintzberg, 1979). Unpredictability leads to task complexity, nonrepetetiveness, and an emphasis on exception processing as opposed to rote or rule processing. Skill levels and special- ization increase under uncertainty. Decentralized structures arise to provide managers with the leeway to navigate the firm through a rapidly shifting and unpredictable environment. En- vironmental uncertainty is associated with integrated structures as subunits increasingly must coordinate their response to en- vironmental change. H7: Size is related positively to specialization, innovation adoption decentralization, manufacturing operations de- centralization, and integration. H8: Environmental uncertainty is related positively to spe- cialization, innovation adoption decentralization, manufacturing operations decentralization, and in- tegration. Method Sample A manufacturer's membership of 3,280 names was provided by the Council of Logistics Management (CLM). This profes- sional organization was selected because its members would be particularly knowledgeable about logistics innovation. The CLM list was culled by selecting only members residing in the United States, and then 1,000 members were randomly selected. A questionnaire, pretested by 10 managers for clarity, was mailed in two waves, followed by telephone callbacks. Of the questionnaires, 261 were returned, of which 44 were non- deliverable, leading to a 22.7% response rate (i.e., 217/956). Of the 217 returned by CLM members, 34 were discarded be- cause either too many values were missing or the firm oper- ated as a distributor or distributor subsidiary of a manufac- turer. The latter was necessary because the instrument was specifically designed for manufacturers. The analysis is based on 183 manufacturing organizations. Mean annual sales were $1.98 billion, with a median of $450 million and a standard deviation of $4.47 billion. The average number of employees per manufacturer was 7,400. The majority of respondents held director (54.7%) or manager (27.4%) title levels and represented a wide range of industries: food, bever- age, tobacco, 24.2%; chemicals, 17.0%; pharmaceuticals, health and beauty aids, 11.0%; electronics, computers, 8.2%; trans- portation, motor equipment, 6.0%; metals, minerals, petroleum, rubber, 5.5%; and building materials, related supplies, 4.9%. The sample is thus biased toward large manufacturers with a reliance on food and chemical suppliers. Nonresponse bias was tested by comparing the last quartile to respond (obtained through telephone callbacks) to earlier ones. If late respondents are similar to nonrespondents, non- response bias may not be severe (Armstrong and Overton, 1977). Using t-tests, no differences were found between early and late respondents in any of the variables in the study in- cluding sales, employees, and levels of innovation (at 0.05). In addition, no differences were found in industry, respondent title level, or functional responsibility area distributions (using X 2 tests with p ~ 0.05). Measurement of Context and Organizational Structure Following convention, size was measured by the natural loga- rithm of the number of employees (e.g., Blau, 1970; Kimberly and Evanisko, 1981). A logarithmic operationalization means that size has a diminishing effect on structure and innovation as size increases. Miller and Dr6ge's (1986) scales, traceable to Khandwalla (1974) and Inkson, Pugh, and Hickson (1970), were used for measuring environmental uncertainty and struc- ture. The Appendix details these scales and a discussion of them follows. Environmental uncertainty was measured by summing four seven-point ratings on the dynamism of marketing prac- tices, competitor actions, customer demands and tastes, and production processes (Miller and DrOge, 1986). The Appen- dix provides left hand anchors for the items. Specialization can be measured by the count from a list of knowledge areas dealt with exclusively by at least one individual in the firm (Inkson, Pugh, and Hickson, 1970). A list of 10 areas related to logistical operations was generated; the number present in the firm is the measure of specialization (see the Appendix). Manufactur- ing operations decentralization was measured by an abbreviated form of Miller and DrOge's (1986) scale: seven-point scales were used to measure the decentralization of five operational deci- sions, and a sum was taken. Using the same scale endpoints, respondents were asked to rate the vertical locus of decision- making over seven innovations. A sum was then taken (inno- vation adoption decentralization). Integration was measured by slightly adapting Miller and DrOge's (1986) scales. Integrative committees measure reliance on cross-functional committees Context and Structure in Logistics J Busn Res 121 1996:35:117-127 Table 1. Innovation Items, Means, Correlations, and Tests of Cluster Means Innovation Cluster/Innovation Item Mean Radicalness Cost Correlation Adoption of Cost and Rate Radicalness Radical innovation cluster Robotics (L) Automated storage/retrieval systems (L) Automated material handling equipment (L) Intermediate innovation cluster Optical scanners (L) E[ectronic data interchange Distribution modeling software Distribution requirements planning software Handheld data entry devices (L) Materials requirements planning software Bar codes (L) Local area networks (L) Direct product profitability software Order processing software Incremental innovation cluster In-process inventory control software Sales forecasting software Vehicle routing/scheduling software Warehouse on-line receiving software Warehouse short interval scheduling software Warehouse workload balancing software Warehouse order selection software Warehouse merchandiser locator software Order entry software Raw material inventory control software Finished goods inventory control software Supporting financials software Freight consolidation software Freight audit/payment software Cluster means Radical innovation (A) Intermediate innovation (B) Incremental innovation (C) Repeated measures Wilkes' lambda Post-hoc contrasts 4.19 4.68 0.126 0.44 3.77 4.53 0.208 0.28 3.43 3.97 0.426 0.50 3.40 3.31 0.399 0.50 3.41 3.01 0.770 0.32 3.22 3.16 0.541 0.43 3.10 3.21 0.437 0.38 3.22 2.98 0.514 0.44 2.83 3.30 0.699 0.46 3.29 2.87 0.634 0.36 2.97 2.92 0~656 0.52 2.92 2.95 0.492 0.48 2.52 3.21 0.979 0.43 2.68 2.88 0.628 0.50 2.71 2.81 0.776 0.49 2.68 2.72 0.443 0.43 2.64 2.75 0.557 0.37 2.71 2.69 0.301 0.69 2.68 2.65 0.322 0.51 2.49 2.80 0.727 0.49 2.59 2.69 0.601 0.41 2.33 2.85 0.934 0.57 2.45 2.73 0.809 0.48 2.45 2.61 0.891 0.46 2.42 2.62 0.653 0.46 2.51 2.47 0.656 0.48 2.33 2.47 0.820 0.32 3.79 4.40 0.253 3.11 3.01 0.629 2.54 2.70 0.653 0.36 0.14 0.36 A>B>C A>B>C A 122 J Bush Res R. Germain 1996:35:117-127 Table 2. Correlation Matrix and Summary Statistics Variable xl x2 yl y2 y3 y4 y5 y6 Mean SD Reliability xl. Natural logarithm of employees x2. Environmental uncertainty 12 - yl. Specialization 28 08 - y2. Innovation adoption decentral. 28 -03 15 - y3. Mfg. operations decentralization 18 17 -01 36 y4. Integrative committees 15 14 22 18 y5. Integrative mechanisms 26 18 24 17 y6. Incremental innovation 09 -01 17 18 y7. Radical innovation 24 21 19 09 09 15 56 - 01 10 10 - -01 02 03 35 7.67 1.77 - 4.19 0.99 0.62 6.83 2.32 0.76 2.80 0.81 0.92 3.57 0.76 0.77 4.44 1.45 0.81 4.37 1.36 0.71 4.88 2.91 0.86 3.32 4.92 0.69 r /> 0.13, p < .10; r a 0.17, p < .05; r ~ 0.20; p < 01. All reliabilities are Cronbach a's except for specialization, which is a Kuder-Richardson 20 estimate. a wide spectrum of operations (handheld data entry devices, optical scanners, bar codes). Cluster centroids were estimated and cluster centroid means were analyzed using repeated mea- sures tests (presented at the bottom of Table 1). Post hoc con- trasts reveal that all three cost and radicalness pairs differ sig- nificantly across clusters. Thus, innovations within the radical cluster are more costly and more radical than those within the intermediate innovation cluster which, in turn, are more costly and more radical than those within the incremental duster. The post hoc contrasts also showed that the adoption of radical in- novation is significantly less than the adoption of the other in- novation types (which themselves do not differ). Adoption was then weighted by radicalness and cost- that is, the binary adoption scale was multiplied by perceived radi- calness and cost. A firm received a score between zero and 25 for each innovation (i.e., a score of zero for nonadoption and a score between one and 25 for adoption). A mean was then estimated across the innovations within each cluster for each subject. This is similar to Dewar and Dutton's (1986) approach in that they weighted binary adoption by radicalness. Our scale differs from theirs by the inclusion of cost and the manner in which weighting ratings were obtained. Their radicalness weights were obtained by a panel independent of respondents and were constant across all adopters regardless of how inten- sively any firm implemented the innovation. To maximize the difference between innovation types, only the radical and in- cremental innovation clusters were retained for further analy- sis, which is consistent with Dewar and Dutton (1986). These researchers retained only the three most and the three least radical innovations from an initial list of 12. Reliability and Summary Statistics Table 2 presents the correlation matrix of the variables, and the mean, standard deviation, and reliability for each variable. A Kuder-Richardson 20 reliability is provided for the number of specialists and Cronbach c~ values for the remaining multi- item scales. With the exception of the two decentralization meas- ures, separate principal components analysis were conducted on the summed nonbinary scales. In each case, only the first eigenvalue was greater than one. The 12 decentralization meas- ures were simultaneously subjected to a principal components analysis. Two factors were selected because: (1) only the first two eigenvalues exceeded one (5.44 and 2.05, respectively); (2) the percent variance explained was satisfactory (62%); (3) the factors were as expected (i.e., the dimensions reflected in- novation adoption and manufacturing operations decentrali- zation); and (4) the scree test indicated two factors were ap- propriate (the third through fifth eigenvalues were 0.82, 0.75, and 0.60). 1 Results HI: Innovation Cast and Radicalness The correlation between cost and radicalness for each innova- tion is provided in Table 1. They range from 0.28 (for AS/RS) to 0.69 (for warehouse short interval scheduling). All are sig- nificant at 0.01. In addition, the correlation between mean cost and radicalness across innovations is significant (r = 0.48, p < .01). H1 is supported: innovation cost and radicalness as- sociate positively. H2-H8: Structure as a Mediator of Context -+ Innovation Relationships LISREL was used to test H2 through H8. This affords several advantages: the entire model is simultaneously estimated; corre- lations between pairs of endogenous (or exogenous) constructs can, if necessary, be modeled; and statistical tests are available on the mediating role of structure. Using LISREL on the covariance matrix (J6reskog and S6r- bom, 1989), a full model with structural paths hypothesized to be significant and nonsignificant was estimated. Allowed to covary were several pairs of endogenous constructs: (1) the 1The decision concerning the number of factors to retain may also be aided by a confirmatory factor analysis of the 12 decentralization variables. Models of two correlated constructs (X 2 - 176.20; df- 55; p - .000), two uncor- related constructs (X 2 - 201.87; df- 54; p - .000), and a single construct (X 2 - 314.62; df- 54; p - .000) were not encouraging. A number of three- dimensional models were also examined, but none of these was encouraging as well. Context and Structure in Logistics J Busn Res 123 1996:35:117-127 two innovation constructs reflecting an overall propensity to innovate; (2) the two decentralization constructs indicating that decentralization in one domain is correlated with decentrali- zation in another; and (3) specialization with integration from the unique role of indirect labor specialization to provide in- formation necessary for cross-functional integration. A good fit was obtained (X 2 = 11.45; df - 11; p = 0.407). A series of six partially nested models was studied. In the partially nested models, each of the six zero-hypothesized paths was dropped from the full model one at a time. Under the"nested model" heading, Table 3 presents the X 2, degrees of freedom, and p-value of each of the six nested models. The test of the difference between each partially nested model and the full model is summarized under the "difference" heading in Table 3. The decrease in overall model fit is not significant for each of the six partially nested models. A fully nested model was then examined wherein all six zero-hypothesized paths were set to zero. The overall model results are presented in Table 3, and the difference between it and the full model is not sig- nificant (Ax2 = 2.25; Adf= 6; p > 0.10). Thus the fully nested model fits the data no worse than the full model and is preferred because it is more parsimonious. The remainder of our atten- tion is focused on the fully nested model. Other indicators also suggest that the fully nested model fits well: GFI = 0.982; AGFI - 0.963; the largest modification index is 2.71 (i.e., none exceed 5); none of the standardized residuals exceed + 2; and the Q-plot is almost linear. The total coefficient of determination is 0.308. The individual coefficients of determination are: 0.076 for specialization; 0.092 for innovation adoption decentraliza- tion; 0.049 for manufacturing operations decentralization; 0.104 for integration; 0.047 for incremental innovation; and 0.108 for radical innovation. Standardized parameter estimates and t-values for the structural paths, correlated endogenous con- structs, and measurement component of the model are provided in Table 4. HYPOTHESIS TESTING. H2 said specialization would predict in- cremental and radical innovation. The hypothesis is supported as specialization is a significant predictor of both innovation types (j85,1, 86,1). H3 stated that innovation adoption decen- tralization would predict incremental, but not radical innova- tion, and operations decentralization would not relate to logis- tics innovation. Innovation adoption decentralization predicts incremental innovation (~5,2). The paths from innovation adop- tion decentralization to radical innovation and from operations decentralization to both types of innovation were not signifi- cant in the full model and were individually dropped in nested models with no significant decrease in fit. H3 is supported com- pletely. H4 proposed that integration would inversely predict radical while failing to predict incremental innovation. Integra- tion inversely predicts radical innovation (86,4). The path from integration to incremental innovation was not significant in the full model, nor did setting it to zero in a nested model signifi- cantly decrease overall model fit. H4 is supported. H5 and H6 respectively said size and environmental uncertainty would pre- dict radical, but not incremental, innovation. Size and environ- mental uncertainty are significant predictors of radical innova- tion (3,5,1, 3'5,2). The paths from size and environmental uncertainty to incremental innovation were set to zero in indi- vidually nested models. No significant decrease in fit was found and both hypotheses are supported. H7 and H8 related size and environmental uncertainty to structure. Size predicts spe- cialization (3' 1,1), decentralization of innovation adoption and manufacturing operations (3`2,1,3"3,1 ), and integration (3`5,1). Environmental uncertainty predicts manufacturing operations decentralization (3'3,2) and integration (3`4,2), but not special- ization (3`1,2) or innovation adoption decentralization (3,2,2). Finally, LISREL statistics on mediating effects were exam- ined from the fully nested model. Structure is a significant medi- ator of the size -+ incremental (t - 2.58, p < 0.01), but not the size "+ radical innovation relationship (t = 0.17, p > 10). In the case of the latter, specialization and integration, both predicted by size, predict radical innovation in opposing man~ ners. Structure is not a significant mediator of the environmental uncertainty ~ incremental (t ~ 0.20, p > 0.10) or the environ- mental uncertainty ~ radical innovation relationship (t ~ 0.82, p > 0.10). Although a number of structural paths between con- text and structure and between structure and innovation are significant, structure is a significant mediator of only the size incremental innovation relationship. Table 3. Tests of Nested Models Nested Model Difference Fit Statistic with Full Model Path Set to Zero X 2 df p AX2 Adf p Size "* incremental innovation (3"5, 1) Uncertainty "* incremental innovation (',/5, 2) Innovation adoption decentralization ~ radical innovation (/~6, 2) Operations decentralization ~ incremental innovation (/~5, 3) Operations decentralization ~ radical innovation (~86, 3) Integration ~ incremental innovation (O5, 4) All of the above 11.46 12 0.491 0.01 1 >0.10 11.48 12 0.488 0.03 1 >0.10 12.37 12 0.417 0.92 1 >0.10 11.83 12 0.459 0.38 1 >0.10 12.89 12 0.377 1.44 1 >0.10 11.82 12 0.460 0.37 1 >0.10 13.70 17 0.688 2.25 6 >0.10 124 J Busn Res R. Germain 1996:35:117-127 Table 4. LISREL Model Results Standardized Parameter Estimate t-Value Structure ~ innovation Specialization ~ incremental innovation Specialization ~ radical innovation Innovation adoption decentralization ~ incremental innovation Integration ~ radical innovation Context ~ innovation Size ~ radical innovation Uncertainty '+ radical innovation Context ~ structure Size ~ specialization Size ~ innovation adoption decentralization Size ~ manufacturing operations decentralization Size "+ integration Uncertainty ~ specialization Uncertainty '~ innovation adoption decentralization Uncertainty ~ manufacturing operations decentralization Uncertainty "-+ integration Correlated endogenous constructs Specialization "-'~ integration Innovation adoption decent. "--~ manufacturing operations decent. Incremental innovation "--~ radical innovation Measurement model Size Environmental uncertainty Specialization Innovation adoption decentralization Manufacturing operations decentralization Integration Integrative committees Integrative mechanisms Incremental innovation Radical innovation /35, 1 0.154 2.044 b /36, 1 0.146 1.874 b ~5, 2 0.142 1.993 b /36, 4 -0.129 -1.481 c 3'6, 1 0.213 2.8738 3'6, 2 0.201 2.852 a 3'1, 1 0.271 3.667 a 3'2, 1 0.295 4.018 ~ 3'3, 1 0.161 2.141 b 3'4, 1 0.267 2.702 a 3"1, 2 0.044 0.601 3'2, 2 -0.071 -0.966 3'3, 2 0.152 2.022 b 3'4, 2 0.181 1.995 b • 1, 4 0.201 2.208 b • 2, 3 0.322 4.261 ~ g'5, 6 0.302 4.091 a kxl, 1 1.773 kx2, 2 0985 ~.yl, 1 2.314 Xy2, 2 0.812 Xy3, 3 0.758 Xy4, 4 0.941 - Xy5, 4 1.163 4.009 a Xy6, 5 2.898 - Xy7, 6 4.903 - Significance levels are based on one-tailed tests. a p < .01. b p < .05. p < 10. Conclusion This research examined empirically the adoption of logistics process innovations. As expected, innovation cost and radical- ness correlate, and the subsequent classification of the 27 in- novations yielded three distinct clusters of innovation: in- cremental, intermediate, and radical innovation. The cluster of radical innovations consisted entirely of prod- uct handling hardware: robotics, automated storage and retrieval equipment, and automated material handling equipment. These innovations are, on average, expensive and radical in their im- pact on operations. As already mentioned, these systems drasti- cally alter material handling landscapes. But, because of high cost (and probably risk), they are infrequently adopted: the av- erage adoption rate within the cluster is 0.253, significantly less than the adoption rate of the two remaining clusters. Size has a direct effect on radical innovation because larger organiza- tions possess the financial muscle to overcome cost hurdles as- sociated with adoption. Environmental uncertainty, as well, has a direct impact on radical innovation as dynamism engenders a more future-oriented organization aware of external change and hence more receptive to original solutions. When it comes to organizational structure, a factor under a high degree of management control, radical innovation is positively predicted by specialization, inversely predicted by integration, and not predicted by decentralization (neither in innovation adoption nor in manufacturing operations domains). Structure is not a significant mediator of context ~ radical innovation relationships. In other words, it does not transmit the effect of size and environmental uncertainty to radical in- novation. Concerning size, larger organizations are more special- ized and integrated, but specialization promotes whereas in- tegration stifles radical innovation. Large organizations are more reliant on specialists, but they may be purposefully introduc- Context and Structure in Logistics J Busn Res 125 1996:35:117-127 ing lateral links as a means of counterbalancing the proclivity of specialists to adopt innovation regardless of cost or impact. Face-to-face integration, a more egalitarian source of influ- ence compared to centralized financial control, may counter specialists in their quest for expensive, radical innovation and may be a more palatable source of control to functional super- visors. Structure plays no role in transmitting the effect of en- vironmental uncertainty to radical innovation because the mul- tiplicative effect of uncertainty on integration and integration on radical innovation is simply too weak. The incremental innovation cluster consists of 14 innova- tions low in cost and radicalness. The entire set of innovations consists of computer software applications that are applied to single functions (e.g., sales forecasting, finished goods inven- tory control). These innovations are significantly more likely than radical innovations to be adopted (an average adoption rate of 0.653 compared to 0.253), and are significantly less costly and less radical than radical innovations. The low cost and lack of radicalness probably derives from incremental innovations being applied to an arena of operations limited in scope. Despite the distinction between incremental and radical in- novation, adoption of the two innovations is correlated (r = 0.35, p < 0.01). This stems from efforts by organizations to cre- ate unified material handling/warehousing systems. Some or- ganizations are combining AS/RS technology with automated guided vehicles, bar codes, radio frequency communications, carrousel stacking systems, and a host of specialized warehouse software (e.g., short interval scheduling, workload balancing) into a single overarching material handling technology that spans the entire gamut of incremental, intermediate, and radical in- novation (e.g., Muller, 1993). Size and environmental uncertainty have no direct effect on incremental innovation: organizations across the scale spec- trum possess the financial resources for adoption and organi- zations facing dynamic environments possess no "future- oriented" advantage. In addition, it is possible that prior research on the mediating effect of structure in context ~ innovation relationships reporting a weak direct effect for size (e.g., Moch, 1976) examined innovations moderate in cost and radicalness. Regarding structure, specialization and innovation adoption de- centralization predict incremental innovation, whereas opera- tions decentralization and integration do not. In this case, struc- ture is a significant mediator of the effect of size. However, it is not a significant mediator of environmental uncertainty. Larger firms are more specialized and decentralized regarding inno- vation adoption, and specialization and decentralization in this domain influence incremental innovation adoption. Uncertainty fosters operations decentralization and integration, but neither of these affect incremental innovation. Examining both innovations types simultaneously leads to three observations. First, specialization predicts both innova- tion types. This consistency with previous research (Dewar and Dutton, 1986) suggests that the specialization ~ innovation rela- tionship is robust across studies, and the relationship is inde- pendent of innovation cost/radicalness. Second, that two decentralization domains relate differen- tially to two innovation types helps explain why past efforts to relate measures of global decentralization to innovation, al- though significant across studies (Damanpour, 1991), have in many instances been nonsignificant, or, if significant, statisti- cally weak. The relationship between global decentralization and innovation may be significant only to the extent that it corre- lates with innovation adoption decentralization and incremental innovation. A global decentralization measure weakly correlated with innovation adoption decentralization may not predict in- novation, especially if the researcher does not control for in- novation cost and radicalness. Furthermore, decentralization must be targeted to the behavior. There are limits, however. A paired difference t-test of the mean locus of innovation adop- tion (2.80) minus manufacturing operations decentralization (3.60) is significant (t - -11.75, p < 0.01). Logistics innova- tion adoption is more centralized than manufacturing opera- tions. Organizations may not be decentralizing innovation adop- tion too far down hierarchies from fear of ceding power to ill-qualified subordinates. Managers should be aware that in- novative behavior requires decentralization in the targeted do- main, but that the desired behavior cannot be too costly or radical. Finally, the research demonstrated that integration, a vari- able that has received little attention in the innovation litera- ture, is inversely related to radical innovation and unrelated to incremental logistics innovation. Managers should under- stand that organizations apparently rely on integration to con- trol the liberal adoption of radical, expensive innovation. Limitations The findings of the study must be tempered by several short- comings. Despite the use of LISREL, a cross-sectional study can- not be used to infer causality. Although a context ~ structure innovation ordering was axiomatic to this research, alterna- tive causal orderings cannot be ruled out. For instance, radical innovation may promote growth and ultimately size. Our mea- sure of uncertainty was a perceived and not an objective one. The sample is biased toward large firms and is overly reliant on one or two industrial groupings. The binary nature of the innovation adoption scale, despite its subsequent weighting, does not take into account when an innovation was adopted or, if adopted, when upgraded. In addition, specification error, or the omission of variables, must be highlighted as a shortcoming. With the exception of formalization, the exclusion of which is justified by meta-analytic findings (Darnanpour, 1991), the set of latent organizational structure variables in the study is complete, but observable dimensions of structure are not in the study. Radical but not incremental innovation may be associated with flatter, wider structures (Child, 1988). Management traits may also impact innovation adoption; greater executive tenure may associate with resistance to radical but not incremental change (Daman- pour, 1991), and senior management cosmopolitanism may as- 126 J Busn Res 1996:35:117-127 R. Germain sociate with receptivity to innovation (Kimberly and Evanisko, 1981). Furthermore, the role of strategy was not studied. Strategy may mediate context ~ structure relationships and may directly impact innovation (Ettlie, 1983). The Miles and Snow (1978) strategic typology consisting of prospectors, defenders, reactors, and analyzers may prove valuable in this regard. Prospectors are adaptive organizations that invest in research and develop- ment and environmental monitoring systems. An innovative orientation is also part of their profile. Defenders are conser- vative organizations low in adaptive capability and high in cost control focus. Prospectors may thus be more likely than de- fenders to search for and implement radical innovations. The issue of strategy is certainly worthy of further research. Appendix. Operational Measures Environmental uncertainty: The sum of four seven-point items. Presented are the left-hand anchors indicating environmental sta- bility for: a. marketing practices: 'our firm must rarely change its mar- keting practices to keep up with the market and compe- titors" b. competitor actions: "actions of competitors are quite easy to predict" c. demand and customer tastes: "demands and customers' tastes are quite easy to predict (e.g., milk company)" d. production processes: "the production process is well es- tablished and not subject to very much change (e.g., steel production)" Specialization: The number of areas one or more full time specialists deals with: warehouse facilities design, plant facilities design, mate- rial handling, market research, sales forecasting, production schedul- ing, distribution equipment procurement, plant or warehouse facil- ity location, transportation scheduling, manufacturing quality control. Operations decentralization: Sum of five seven-point items on delega- tion of authority over operations. Endpoints of"decision made above chief executive (e.g., board of directors); and "individual below first level supervisor" endpoints. Intermediate points associated with spe- cific organizational level (e.g., 2 - divisional manager). a. the number of workers required b. allocation of work among available workers c. internal labor disputes d. overtime at the plant level e. plant machinery or equipment to be used Innovation decentralization: The sum of seven seven-point items on delegation of authority over innovation adoption. Same endpoints as operations decentralization. a. finished goods inventory control/order processing software b. materials management software; warehousing software c. transportation software d. electronic data interchange e. data handling hardware f. product handling hardware Integrative committees: The sum of four seven-point items. Respon- dents rated the extent to which decision-making at the top levels of the firm are characterized by participative, cross-functional com- mittees in which different departments, functions or divisions get together to decide specific strategies. Endpoints of"rare use" and "fre- quent use." a. distribution service strategy b. marketing/sales strategy c. capital budget decisions d. long-term strategies of growth and diversification Integrative mechanisms: The sum of three seven-point items on the extent to which the following are used in assuring the compatibility among decisions in one area (e.g., distribution) with those in an- other. "Rare use" and "frequent use" endpoints. a. interdepartmental committees b. temporary task forces c. liaison personnel References Aiken, Michael, and Hage, Jerald, The Organic Organization and In- novation. Sociology 5 (June 1971): 63-82. Appliance Manufacturer, Software Solves Whirlpool's Material Handling Quandary. 42 (April 1994): 51. Armstrong, J. Scott, and Overton, Terry S., Estimating Nonresponse Bias in Mail Surveys.Journal of Marketing Research 14 (August 1977): 396-402. Baldridge, J. Victor, and Burnham, Robert A., Organizational Innova- tion: Individual, Organizational and Environmental Impacts. Ad- ministrative Science Quarterly 20 (June 1975): 165-176. Blau, Peter M., A Formal Theory of Differentiation in Organizations. American Sociological Review 35 (April 1970): 201-218. Blau, Peter M, Falbe, Celcelia M., McKinley, William, and Tracey, Phelps K., Technology and Organization in Manufacturing. Administrative Science Quarterly 21 (March 1976): 20-40. Bowersox, Donald J., Daugherty, Patricia J., Dr6ge, Cornelia L., Get- main, Richard N., and Rogers, Dale S., Logistical Excellence: It's Not Business as Usual, Digital Press, Burlington, MA. 1992. Burns, Tom, and Stalker, G.M., The Management of Innovation, Tavistock, London. 1961. Carter, Nancy M., Computerization as a Predominant Technology: Its Influence on the Structure of Newspaper Organizations. Academy of Management Journal 27 (june 1984): 247-270. Carter, Nancy M., and Cullen, John B., A Comparison of Centraliza- tion/Decentralization of Decision Making Concepts and Measures. Journal of Management 10 (2) (1984): 259-268. Champion, Dean J., The Sociology of Organizations, McGraw-Hill, New York. 1975 Child, John, Organization Structure and Strategies of Control: A Repli- cation of the Aston Study. Administrative Science Quarterly 17 (June 1972): 163-177. Child, John, Predicting and Understanding Organizational Structure. Administrative Science Quarterly 18 (june 1973): 168-185. Child, John, Information Technology and Organization, in Innovation and Management: International Comparisons, K. Urabe, J. Child, and T. Kagono, eds., Walter De Gruyer, Berlin. 1988, pp. 225-301. Collins, Paul D., Hage, Jerald, and Hull, Frank M., Organizational and Technological Predictors of Change in Automaticity. Academy of Management Journal 31 (September 1988): 512-543. Coyle,JohnJ., Bardi, EdwardJ., and Langley, C. John, The Management of Business Lo~stics, West Publishing, New York. 1992. Damanpour, Fairborz, The Adoption of Technological, Administra- tive and Ancillary Innovations: The Impact of Organizational Fac- tors. Journal of Management 13 (4) (1987): 765-788. Context and Structure in Logistics J Bush Res 127 1996:35:117-127 Damanpour, Fairborz, Organizational Innovation: Meta-Analysis of Ef- fects of Determinants and Moderators. Academy of Management Jour- nal 34 (September 1991): 555-590. Damanpour, F., Szabat, Kathryn A., and Evan, William M., The Rela- tionship between Types of Innovation and Organizational Perfor- mance. Journal of Management Studies 26 (November 1989): 587-601. Dewar, Robert D., and Dutton, Jane E., The Adoption of Radical and Incremental Innovations: An Empirical Study. Management Science 32 (November 1986): 1,422-1,433. Duncan, Robert G., Characteristics of Organizational Environments and Perceived Environmental Uncertainty. Administrative Science Quarterly 17 (June 1972): 313-327. Ettlie, John E., Organizational Policy and Innovation among Suppliers to the Food Processing Sector. Academy of Management Journal 26 (March 1983): 27-44. Ettlie, John E., Bridges, William P., and O'Keefe, Robert D., Organiza- tion Strategy and Structural Differences for Radical Versus Incremen- tal Innovation. Management Science 30 (June 1984): 682-695. Forger, Gary, We Now Manage More Inventory Better with Automa- tion. Modern Materials Handling 48 (May 1993a): 46-48. Forger, Gary, AS/RS Trims Errors, Conserves Floor Space. Modern Materials Handling 48 (August 1993b): 46-47. Galbraith, Jay, Designing Complex Organizations, Addison-Wesley, Read- ing, MA. 1973. Gatignon, Hubert, and Robertson, Thomas S., Technology Diffusion: An Empirical Test of Competitive Effects. Journal of Marketing 53 (January 1989): 35-49. Germain, Richard, The Adoption of Logistics Process Technologies by Manufacturers. Journal of Business Research 27 (May 1993): 51-63. Glaskowsky, Nicholas, A., Hudson, Donald R., and Ivie, Robert M., Business Lo~stics, The Dryden Press, Fort Worth, TX. 1992. Hage, Jerald, Theories of Organization: Form, Process, and Transforma- tion, John Wiley and Sons, New York. 1980. Hage, Jerald, and Aiken, Michael, Social Change in Complex Organiza- tions, Random House, New York. 1970. Hage, Jerald, and Dewar, Robert, Elite Values Versus Organizational Structure in Predicting Innovation. Administrative Science Quarterly 18 (September 1973): 279-291. Hoffman, Richard C., and Hegarty, Harvey W., Top Management In- fluence on Innovations: Effects of Executive Characteristics and So- cial Culture. Journal of Management 19 (Fall 1993): 549-574. Howell, Jane M., and Higgins, Christopher A., Champions of Techno- logical Innovations. Administrative Science Quarterly 35 (June 1990): 317-341. Inkson, J.H., Pugh, Derek, and Hickson, David, Organization, Context and Structure: An Abbreviated Replication. Administrative Science Quarterly 15 (September 1970): 318-329. Jennergren, L.P., Decentralization in Organizations, in Handbook of Or- ganizational Design, vol. 2, P.C. Nystron and W.H. Starbuck, eds., Oxford University Press, New York. 1981, pp. 39-59. J6reskog, Karl G., and S6rbom, Dag, LISREL 7: A Guide to the Program and Applications, SPSS Inc., Chicago. 1989. Kimberly, John R., and Evanisko, Michael J., Organizational Innova- tion: The Influence of Individual, Organizational, and Contextual Factors on Hospital Adoption of Technological and Administra- tive Innovations. Academy of Management Journal 24 (December 1981): 689-713. Khandwalla, Pradip, Mass Output Orientation of Operations Technol- ogy and Organizational Structure. Administrative Science Quarterly 19 (March 1974): 74-97. Lawrence, Paul R., and Lorsch,Jay, Organization and Environment, Har- vard Graduate School of Business Administration, Boston. 1967. Meyer, Alan D., and Goes, James B., Organizational Assimilation of Innovations: A Multilevel Contextual Analysis. Academy of Manage- ment Journal 31 (December 1988): 897-923. Miles, Raymond E., and Snow, Charles C., Organizational Strategy, Struc- ture, and Process, McGraw-Hill, New York. 1978. Miller, Danny, and Dr6ge, Cornelia, Psychological and Traditional De- terminants of Structure. Administrative Science Quarterly 31 (De- cember 1986): 539-560. Mintzberg, Henry, The Structuring of Organizations, Prentice-Hall, En- glewood Cliff, NJ. 1979. Moch, Michael K., Structure and Resource Allocation. Administrative Science Quarterly 21 (December 1976): 661-674. Moch, Michael K., and Morse, Edward V., Size, Centralization, and Organizational Adoption of Innovations. American Sociological Re- view 42 (October 1977): 716-725. Muller, E. J., The Best Game Nintendo Built . . . . Distribution 92 (De- cember 1993): 30-35. Nord, Walter R., and Tucker, Spencer, Implementing Routine and Radi- cal Innovation, Lexington Books, Lexington, MA. 1987. Pierce, John L., and Delbecq, Andre. L., Organizational Structure, In- dividual Attitudes, and Innovation. Academy of Management Jour- hal 2 (January 1977): 27-37. Rees, Reg, Smooth Landing at Air Canada for 1,000,000 Aircraft Parts. Industrial Engineering 26 (June 1994): 28-29. Rogers, Everett M., Diffusion of Innovations, The Free Press, New York. 1983. Thompson, Victor A., Bureaucracy and Innovation, University of Ala- bama Press, University, AL. 1969. Zaltman, Gerald, Duncan, Robert, and Holbek, Johnny, Innovations and Organizations, John Wiley and Sons, New York. 1973. Zmud, Robert, The Diffusion of Software Practices: Influences of Cen- tralization and Formalization. Management Science 28 (December 1982): 1,421-1,431.