'<m HD28 .M414 TO, Hol- ALFRED P. WORKING PAPER SLOAN SCHOOL OF MANAGEMENT The Effect of Interoffice Network Structure on Individual Experience of Problems Gordon Walker September, 1985 WP #1701-85 MASSACHUSETTS INSTITUTE OF TECHNOLOGY 50 MEMORIAL DRIVE CAMBRIDGE, MASSACHUSETTS 02139 'The Effect of Interoffice Network Structure on Individual Experience of Problems Gordon Walker September, 1985 WP //1701-85 1 ^• -<-• s ABSTRACT The present etur?y examines the effect of an interoffice network on the quantity quality and communications demonstrated that office problems which The network is composed offices in the network receive. interoffice of and position problem in referrals. th^^ network It has of i a primary effect on problems received and that the location of the position in the network structure a lesser effect. Also, the results show the advantage of viewing structure in network terms. The recent spate of interest in network approaches organizations has frequently been criticized technical than by substantive issues. metaphor, network the that (Coleman, to the being as study driven more by One reason for this response may be which is essentially relational in character is often used to explain differences in individual, 1959), of group or organizational attributes, and the connection in theory between specific relational and attributional phenomena is unconvincing study relatively a unexamined nature in organizations, approached from a but pervasive phenomenon of arrival the network of problem a perspective. interoffice referrals and communications in explain the In , The an network relational a office, is structure of used to an at present organization is differences in the quantity and quality of problems encountered by office members. A major assumption of the present study is that the number of problems received by an office and the appropriateness of these problems for the office should be determined by the amount and quality about the office available in study i s information the the problem source no matter whether the source lies inside or outside the interoffice system. present of The focus of the on the effect of the interoffice system on the number and quality of problems an office encounters. The quality of a problem here means the extent to which an estimates that the problem can be solved by his/her office. may be confronted with few unsolvable with are, however, success problems current means; but those they get individual Some offices are virtually other offices may get many problems which relatively easy to solve. Since the skills necessary for in problem solving are maintained by obtaining a sufficient number Page of appropriate problems, the quality and quantity of problems office members relevant is received by for their performance and the performance of In the present study the interoffice network is their office. 2 assumed to be the means by which appropriate problems are acquired. In a recent review of network models in sociology Burt (1980) makes between distinction (not types of network approach and levels of analysis (see He identifies two types of approach: Table 1). a positional and relational to be confused with the general relational approach, mentioned above, apparent in all network studies) . The present interoffice network of an organization from concerned with this network as a a approaches study the positional perspective and is structured The system. composed in this study of two types of relations: network is interoffice communicaton and problem referral. Intraorganizational communication networks have been studied in of relationship their Tushman, 1978; level from theory. a group effectiveness (Tushman, 1977; O'Reilly and Roberts, 1977), primarily and Jacobsen communication examines research examined (1977) network, individuals who linked different groups. present the at Katz and individual relational persective consistent with information processing Schwartz interpersonal to terms focusing the on structure the of an attributes of In contrast to these interoffice as opposed to studies, the interpersonal communications and is concerned with grouping offices on the basis of similarity of their relationships with other offices in the network. of offices which are similar in their relationships should know be known by other distribution, such a offices similarly; given the A set about and this kind of information set of offices should receive problems comparable in Page 3 as a their quantity and quality. Intraorganizational referral networks have been proposed coordination mechanism in Galbraith's (1974) information processing view of organization design; problem solvers in hierarchical over conflict research no assumption is made between offices other than office has done all it can to Referral relationships referral being potential about the the solutions. reasons for resort last for In the present problem referral the rather simple premise that the referring solve between problem a offices number of problems offices receive. by and passing is on. it definition contribute to the Offices which have referral similar relationships with other offices in the network should receive a comparable number of problems. A second major assumption of the present study is that are offices which classified together according to the similarity of their communication and referral relationships with other offices will receive a similar number of problems of comparable quality. The structure of communications and referrals between sets of corresponds offices to the structured system of subgroups in Burt's typology. The third major assumption of the present study is that sets of offices in this structure will have relationships with each other of unequal intensity; effect, this implies that sets of offices which have more in intense relationships with other offices should receive more and better referrals. From the assumptions stated above the following made: propositions can be Page There will be differences in the number and 1. different members organizational by encountered according sets problems of whose offices are classified into similarity the to quality 4 their of interoffice communications and referral relationships. The last assumption of the present research concerns the vary in their degree of access to interoffice system and high degree of access have offices distance from structure low problems which originate outside the that sets of offices which contain offices with encounter should access. high should of offices in the network:it is assumed that offices will of sets different location addition, In access more offices sets the in problems than sets whose offices located at of communication a a referral and encounter few problems of relatively poor quality. This assumption leads to the following proposition: There will be differences in the quality and quantity of 2. encountered by members organizational problems whose offices belong to different sets which are in different locations in the communications and referral and kind structure. The influence of the interoffice network on the problems encountered by office members should individual and situational characteristics. member has of other offices and of number of be moderated by several First, the knowledge an office his/her/ own office and official position should affect the perception of the quality of the referral. Such knowledge would naturally increase the longer an individual was employed by the organization or in associated with a particular office or achievement may position. Informal status also increase an individual's access to information about the organization through contact with members of other . Page Second, the general position of an office member within an office offices. quality should affect the number and receive should 5 problems fewer than of problems his encountered: deputy, problems should likewise increase his deputy fewer and problems than the front-line secretary of receptionist. boss a The difficulty of as they are passed up the intra-offce hierarchy. A model describing the general relationships between the variables this study is presented in Table in 2. DATA AND METHODS The data for this study consist of responses to a 86 members of 42 offices in a large eastern university. the largest number of each office responded; office was six and the smallest number one. basis of a Problems snowball which sample from members encountered size was 49; a single The offices were chosen on the sample begun in several offices which were office from Not all members of respondents possibly and defined as students who arrived at the office with snowball questionnaire a well-known. referred were problem. The final seven of these offices did not complete the questionnaire Office members were asked to indicate the number of students who to with them a came problem daily, the percentage of these problems which were solvable only by the respondent's office, the percentage partially solvable by the respondent's his/her office. office, and the percentage totally inappropriate for Respondents were also asked to how indicate frequently students were referred to other offices in the sample, to nominate at least five offices with which communications were frequent and degree of frequency of to indicate the face-to-face, telephone, and memo communications Page 6 Respondents also provided infonnation with the offices nominated. number their offices and in the university; they indicated their and their intra-office position (e.g. the positional analysis of networks. commonly method used (Breiger, Boorman and Arabie, a approach general several reasons: representing network 1) the members with developed above; subgroup analysis, terms of in a a sets, construction blockmodels of 1979). As blockmodel ling was chosen for structure of interrelated positions, number of network offices); are the 3) system into blockmodel 1 ing techniques are specifically directed at networks consistent offices Arabie, Boorman and Levitt, 2) structural equivalence, behind blockmodels as is classify The classification algorithm used was for 1^75; network to each position consisting of case degree positions in conformity with the technical literature on called a highest secretary, assistant administrator). The first stage of the analysis was to CONCOR, the years and months they had worked in their current positions, in of hereafter on theoretical among positional the a members (in present the primary principle, called representation of perspective approaches on referral networks networks networks to at the level the technique of blockmodels has the most extensive and well developed published literature. Responses regarding interoffice relationships were the responses of dichotomized, the members of an office were combined to form indicator of office to office relationships. The input to CONCOR a and single was an array of four stacked binary matrices, one matrix for interoffice referrals and one for each mode of interoffice communication. Page The output from CONCOR is always input a bifurcation of the columns the CONCOR was run on the two groups produced by the initial run array. and twice more on groups produced by these runs so that in all sets or positions of offices were derived. The degree to is presented in Table 3. students refer of 7 to communicate and six disjoint The composition of these groups which offices each in position with each other and offices in other positions is represented in the set of density matrices in Table The 4a. blockmodels for these matrices are shoewn in Table 4b. To test the proposition encountered office by that members number the quality and whose offices were in different positions would not be the same, analyses of covariance were run with the and students, problems of independent inappropriate variables problems his/her of level as dependent variables. The were the position of the respondent's office in the network, his/her position in the office and number percentage of solvable problems, of partially solvable the and problems of of education { polychotomized into three levels), (polychotomized into five levels). The covariates were the total number of months employed in the university, the office, and in his/her current position in the office. The resutlts of the analyses of covariance are shown in Table 5. The proposition regarding the effect of the network structure number in on and quality of problems was tested in a two part procedure. the First, although the blockmodels constucted from the density matrices indicate some consistency in the structural location of each position, these locations could not be interpreted precisely enough to lead to the second proposition. Consequently, in order a confident testing of to determine positions were structurally similar, each density matrix was appended which with Page 8 its transpose so that the first six columns constituted densites to the six positions and the last six columns densities from the six positions. The appended matrices were then stacked, and the intercolumn correlation matrix of the stacked array was analysis analysis interpreted analysed. found in Table 6. are factor the factor as a results of positions location within within these factor the groups structure. covariance were then run using the same variables as in Table the the Three groups of positions were taken from results; sharing The were Analyses of 5 except that office positions were replaced by groups of office positions. Table 7 contains the results of this analysis. RESULTS The classification of offices into easily interpreted. offices and I student school dean's are service functions, offices. Financial across spread shown in Position II consists composed of the plus one residence and 3 primarily is of and Position III of the undergraduate and lesser administrative Positions II and III. service Religious groups and operational offices of the university constitute Position IV. is Table is composed of three top administrative powerful dean's office. a direct functions Position positions Position V offices which administer residential life on campus, a student-run service group, and other undergraduate residences comprise Position VI. The analysis of covariance strikingly the influence results on all four Table 5 demonstrate rather of office position on the number and quality of problems encountered by office members. effect in Office position has dependent variables; a significant and only for the perception of inappropriate problems are the context factors significant. Page 9 The factor analysis of the stacked, appended density matrices produced three factors with eigenvalues greater than one, which accounted for 83% of A Varimax rotation produced factor loadings the variance. factor repersents a distinct group of positions. each that off loading cut of factor positions, factor .4, loading which represents one fourth, the two three the first position. single which With second the indicate a factor and third fifth and sixth positions, and factor Factors one and three are confounded each detracts from the otherwise by a perfect interfactor discrimination. The results of the analysis of grouped according to covariance similarity the of their structure are slightly different from the ANCOVA Variation positions. in the number of explained significantly by the groups of analysis of which in positions were location in the network results students positions. using ungrouped encountered was not As initial in the covariance with the positions ungrouped, no other independent variable or covariate explains variation in the number of students. The results for problem quality are similar to those of the ungrouped analysis. Groups of positions have solvability significant effect of student problems and, to of partial solvability. the a inappropriateness a on the perception the results for problem are identical to the first analysis. a Thus, the number of problems encountered varies significantly among groups positions but structure. Problem quality, on not among the lesser degree, on the perception In terms of level of significance, of of the of positions other positions and by groups of positions. DISCUSSION hand, determined is by explained office network both by Page 10 In the present study problems are however, inembers; the constructs encountered by individual office proposed as explanators are groups of offices and groups of groups of offices. The level of the office itself is in favor of the network characteristics of offices both in term? of jumped an initial clustering of similarity of offices interoffice positions on the basis of into positions relationships the their and on basis terms of in similarity the in the of the grouping of a interpositional structure. The results of both classifications were quite interpretable different position However, the is for The positions constructed in the first classification reasons. can be understood in terms of office functions. onto but an obvious empirical virtual The mapping function of isomorphism for the present data. classification of offices according their to communication and referral relationships should not be trivialized in favor of an which a priori functional understanding of the the present scheme produces in organization. One result much richer form than would an a a priori functional approach is the network structure. of In the present study positions were grouped in terms of the similarity their degree structural location with a relatively high discrimination, considering that network relations both given and by a position loaded on the same factor. types of relation. on the for all A second reason for the coherence of the factors is the distinct grouping of the positions themselves in the blockmodels received One reason for this result is the general symmetry of the network structure along the main diagonal four of modes of communication make clear in separation of the first and last three positions. In the network. a The visual way the blockmodel for Page group and three and the can see the tightly knit group of positions two one moreover, referrals, 11 positions of five and Apparently, six. the relationships of position one with the other positions on all four types of relation are sufficiently Considering the different membership of for to it this merit position, its such own group. finding a is understandable. The separation of positions into clearcut such relative degree of organizational stratification. groups suggests a The strata, or groups of positions explain variance in the qualitative aspect of individual referral experience but not other hand, explain the quantitative aspect. variance in both Unaggregated positions, on quantitaive the qualitative and aspectE.~ These results suggest that although an office may change the number of problems it receives by altering its relationships with other offices and thus altering its position in the interoffice network, it is less likely that the quality of problems encountered will change since an alteration of position does not necessarily mean a change in structural location. CONCLUSION The present research has focused on an interoffice referral network as distinguished from interpersonal relations. to most studies using universities as Salancik, 1977; non-academic considered concerns. Katz, offices; the communication result 1979), the consequently, of organizations sample present the study structure administrative as has presented opposed In this regard this study meets in part the need descriptions of administrative structure as a Contrary (Pfeffer used to and and primarily can be professional for empirical set of interrelated offices. Page 12 As defined in this study problems are students who arrive at an office with problems the office may or may not be able to solve. of this definition however, it to seems perceptions likely of task or job The relationship is not specified? that problems defined as an input to an office would tend to affect perceptions of the office probl em- solving system. In this sense the quality and quantity of problems referred mediate the effect of the interoffice network structure on individual perceptions of task job characteristics. Analogous types of problems in and non-educational organizations are employee complaints and problems initiated by clients. Finally, problem referral dependence since without a may be seen as a type of interoffice place to send what is locally unsolvable, an administrative unit decreases in value perceived by perhaps in actual effectiveness as well. its constituency and t B50^ U I k CONCEPTS OP NETWORK STKUCTORE WITHIN EACH OF SIX MOPrS OF NETWORK ANALYSIS (from Burt, 1930) ACTOR ASGREGATION IN A ONIT OF ANALYSIS MULYTXCAL APPROACHES RELATIONAL PERSONAL NETWORK AS EXTENSIVE, DENSE OR MULTIPLEX OCCUPANT POSITIONAL MULTIPLE ACTORS AS A NETWORK SUBGROUP MULTIPLE ACTORS/ SUBGROUPS AS A STRUCTURED SYSTEM PRIMARY CROUP AS A NETWORK CLIQUE: A SET OF ACTORS CONVECTED BY COHESIVE RELATIONS SYSTEM STRUCTURE AS DENSF AND/OR TRANSITIVE oa a. ^ 1^ fN -< B < O U D. ^> o 10 c < 4-) a a o i o h > o u m ^ ir> ^H ^^ t-i >> 0\ I MIT 3 TDfiD I ie,RA»!H 003 DSi 7' u- '^ l~ OJ Lib-26-67 hc^ cvdjL ^^ <^ hcuiJ^ cji^^tr~