$Y$TEMK LINKAGE. INFLUENCE AND CONTROL N A HOSPETAL DECl$iON-MAKlNG STRUCTLERE: S-VAUDATKON STUDY A EROS Thesis {'or We Degree 0% D“. D. WCMGM’Q STATE WW‘ERSETY Robert G. Hofioway 1962 I i 'A {Eng-i.) This is to certify that the thesis entitled Systemic Linkage, Influence and Control in a Hospital Decision-Making Structure: A Cross-Validation Study presented by Robert G. Holloway has been accepted towards fulfillment of the requirements for Ph.D. Sociology and Anthropology degree in QMWQW flajor professor Date August 31 I962 0-169 Michigan State University ABSTRACT SYSTEMIC LINKAGE, ‘INFLUENCEAND CONTROL IN A HOSPITAL DECISION-MAKING STRUCTURE: A. CROSS-VALIDATION. STUDY by Robert G. Holloway This study was concerned with the consequences of systemic linkage of community influentials on the decision-making process of a general hospital. . Systemic linkage was Operationally defined as the extent of overlapping representation of status occupants in the major decision- making units of the hospital (the board of trusteesl-ethe working-committees of the board and administration). The design to assess the consequences of linkage of community influentials in the organization was essentially twofold in nature. First it provides a methodological procedure by which the particular administrative decision-making process of the major units may be analyzed by observation of the interaction concerned with policy making. The design also provides‘a test of cross-validation of two measurement systems commonly used to identify community power structure and to predict control, the "reputational" and "positional" techniques, by means of a third and completely independent measurement system--Ba1es' "interaction process analysis. " The second aspect of the design is the testing of a central hypothesis at different levels of the heirarchy of decision-making units of the organization. This hypothesis is that community influentials are the most active participants in the decision- making process of the hospital and exert more control over the non- influentials than they in turn exert over the community influentials. A com- munity influential was defined as either a top influential, an economic influential, or both. Robe rt G . Holloway Interaction was observed, unitized, and attributized by the author and categorized from electrical recordings of meetings of the major units. Bales' Index of Directiveness of Control was used as the measurement of control. The Appendix contains a critical analysis of the reliability pro- cedures commonly used in profile analysis and recommended by Bales and others in this area, and suggests an alternative statistical procedure for estimating inter-rater reliability. .A 50 year historical analysis of the participation pattern of economic influentials rejected the Schulze hypothesis of "withdrawal" for this organi- zation (by means of the sign test) and gave some support to the Form-Miller hypothesis of increased representation of influentials at strategic intra- organizational levels. The central hypothesis of influential control over the board was supported. Influentials exerted most control over the administration, less over the non-influential board members, and the least amount among them- selves. They were the most active, also. Contrary to the pattern found on the board, at'the working committee level, the non-influentials exerted more control activity over the influentials. . Antagonism and hostility were readily visible in the attempt to resolve issues at this level. Faced with such conflict, the administrators tended to withdraw from interaction. . The influentials were able to direct policy by moving the issue up to higher levels of the organization where they had greater representation, power, and where overt conflict was less likely to occur. Both influentials and the administrators spent much of their time supplying information to their unit. The non-influentials spent mo st of their time in the emotional areas of interaction. . At the administrative level, the chief of the medical staff directed most of‘his control efforts toward the hospital director, who in turn was more concerned with control over his entire staff. The director of nursing Robert G . Holloway expended most of her control efforts toward the chief of staff. Both the physician and nurse spent more of their time in the problem-solving area, whereas the director and his assistants. spent most of their time in the emotional areas of behavior. Comparisons between the administrative conference of a psychiatric hospital, as cited in the literature, and the general hospital were made also. SYSTEMIC LINKAGE, INFLUENCE AND CONTROL IN A HOSPITAL DECISION-MAKING STRUCTURE: A CROSS-VALIDATION STUDY By Robert CIT-Holloway A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOC TOR OF PHILOSOPHY Department of Sociology and Anthropology 1962 ACKNOWLEDGMENTS Boundaries of indebtedness pathetically exceed my ability to express appreciation for the joys (and pain) of graduate socialization. To Dr. Jay W. Artis, my advisor, and Dr. Walter E. Freeman, co- directors of the hospital-community relations project, deep appreciation is extended for sustaining the threshold of frustration over umnet deadlines, for the freedom allowed me, and for the courage to see to its conclusion a very laborious, expensive, and perhaps unusual design with questionable payoff. Many ideas were also contributed by the other members of the project, James P. Harkness and Alex J. Muntean, which were appreciated. Thanks are also due to Drs. Richard N. Adams, J.. Allan Beegle, and Donald W. Olmsted for their committee work and helpful suggestions. I am particularly grateful to Dr. Charles P. Loomis for his constant encouragement and for ”norming in” the importance of codification; to Dr. A. O. Haller for his insistence on measurement and validation; and to Dr. William H. Form, for imparting some of his knowledge of (and for the inspiration for the analysis of) the community as a sociological unit. It is absolutely impossible to describe my full gratitude to Eugene C. Erickson for his extensive contribution to my thinking-~professional and personal. His unselfish devotion to my research design even in the most pressing of moments will never be forgotten. He is also equally responsible for the contribution in the Appendix. ' To my chief therapist and social-emotional leader, Verl R. W. Franz, I can only say "thanks friend for pulling me through. " I am especially indebted to Donald A. Clelland for providing me with the economic influential data, to Warren L. Sauer for the top influential data, to Carole E. Wolff for data on the women on the Woman's Board, and to Genie Lewis for editing and typing assistance. I I I l I l I I :,< >,: 3:: :{z >5 >,< z}: >,< >,: >,: >,: z}: ii TO MARGARET iii TABLE OF CONTENTS CHAPTER I. INTRODUCTION ................... II. THE PARTICIPATION PATTERN OF ECONOMIC INFLUENTIALS ON A HOSPITAL BOARD OF TRUSTEES. . .................. III. THE EMPIRICAL MEASUREMENT OF CONTROL OF ECONOMIC INFLUENTIALS AND TOP INFLUENTIALS ON THE HOSPITAL BOARD OF TRUSTEES: A CROSS-VALIDATION ANALYSIS IV. THE EMPIRICAL MEASUREMENT OF CONTROL OF ECONOMIC INFLUENTIALS AND TOP INFLUENTIALS ON TWO WORKING COMMIT- TEES OF THE BOARD OF TRUSTEES V. THE EMPIRICAL MEASUREMENT OF CONTROL AND PARTICIPATION IN THE ADMINISTRATIVE CONFERENCE .................. VI. SUMMARY AND CONCLUSION ........... BIBLIOGRAPHY .............. . ......... APPENDIX-.-INTER-RATER RELIABILITY OF CATEGORI- ZATION: A METHODOLOGICAL CRITIQUE OF THE PROFILE METHOD, A PROPOSED ALTERNA- TIVE, AND THE LEVEL OF RELIABILITY OF THE STUDY ........................ iv Page 10 26 51 7O 84 91 96 LIST OF TABLES TABLE Page 1. Representation of Economic Influential and Non- influentials on a Community Hospital Board of Trustees from 1910-1959 .......... . . . . . . . . . . . . 15 2. Number of Economic Influentials by 25 Year Periods and by Board Positions per Year ........... . . 16 3. Number of Economic Influentials in a Midwestern Com- munity and on a Hospital Board of Trustees, 1910-1959. 17 4. Representation of Economic Influentials as Office Holders, Committee Memberships and Chairmanships on a Hospital Board of Trustees by Positions per Year, 1910-1959 ...................... . . . 19 5. Number and Percentage of Hospital Board Meetings Attended by Economic Influentials and Non-influentials, 1910-1959 ..... . . . . ................ 22. 6. The Degree of Overlapping Classification of EI's and TI's of Men in the Community and Men on the Hospital Board . . . . . . . .................. . . 31 7. Board Members by El and TI Classification, Table Codes, Primary Occupation, and Economic Unit . . . 33-34 7a. Average Percentage of Board Meetings Attended by Each Board Member from 1950-1959, by Community Influential Classification and by Attendance at the Board ~Meeting Analyzed . . ................ . . . 35 8. Set of Bales' Categories . . . . . . . . . . . ...... 37 9. Index of Directiveness of Control of Values ...... . 4O LIST OF TABLES - Continued TABLE 10 11. 12. 13. 14. 15. 16. 17. 18. 19 20. 21. ., Mean‘IDC Value for El' 5 and Efs by Hospital Board Membership ........... . ........... Mean IDC Values for E1 and as .......... . . . Index of Direct__i_veness of Control Values o_f__ Each TI V__e_rsus Other TI's; Each TI Ver_s__us other TI's; Ea_c_h TI Versus other TI's; and Each TI Versus Other TI's by Hospital Board Membership ............. Mean IDC Values for TI's and T-I's by Board Member- ship ........................ Mean IDC Values for TI's and TI's ........ Rank order of Means between EI-fi and TI-TI Com- parisons on Index of Directiveness of Control ..... Comparison of IDC Values of EI's with TI's Who Do Not Share Overlapping EI-TI Classification ...... Profiles of Distribution of Interagtion by Bales' Cate- gories for EI'S,. EI's, TI's, and TIES, by Board Membership ....... . . . . . ........ Percentage Distribution of Ir_1__teraction in EaLh of the Bales' Categories for EI's s, EI's, TI's s, and TI's by Board Membership. . . . . .............. .. Executive Committee Members by ElzTI Classification, Occupation, and Economic Unit . . . . ........ Means of Index of Directiveness of Control of EI'S and TI'S by Committee Membership on the Executive Committee ......... . . . ........... Mean Differences Between the IDC Values of the Indi- viduals on the Executive Committee by EI:TI Classifi- cation. . ................... vi 41 42 43 44 44 45 46 47 54 55 LIST OF TABLES - Continued TABLE 22. 23. 24. 25. 26. 27. 28. 29. 3O 31. 32. Profile of Percentage Distribution of Interaction by , Bales' Categories by El and TI Classification for Executive Committee Members and Administrators. . . List of Topics Covered by the Executive Committee in TwoMeetings.......... ...... Percentage Distribution of Interaction in Each of the Bales' Categories by El and TI- Classification for Executive Committee Members and Administrators Operations Committee Members by El: TI Classifica- tion, Occupation, and Economic Unit ........... Means. of Index Directiveness of Control of EI's and TI's by Committee Membership on the Operators Com- mittee ........................ Mean Differences Between the IDC Values of the Indi- viduals on the Operations Committee by EI:TI Classifi- cation .......................... Profile of Percentage Distribution of Interaction by Bales' Categories, By EI and TI Classification for Operations Committee Members and Administrators. . , List of Topics Covered by the Operations Committee in Four Meetings. . . . . ............ .. Percentage Distribution of Interaction in Each of the Bales' Categories by El and TI Classification for Operations Committee Members and Administrators Members of the Administrative Conference by Table Code and Position ........ . ......... Means of the Index of Directiveness of Control for the Director, Chief of Medical Staff, and Assistant Admin- istrators. . .......... . . . ...... vii Page 56 57 59 61 62 63 64 65 66 7O 72 LIST OF TABLES - Continued TABLE 33. 34. 35. 36. 37. 38. 39. Mean Index of Directiveness of Control Values of Positional Influentials and Administrative Staff ..... Difference Between the Mean IDC Values of Each Participant in the Administrative Conference ...... Profile of Percentage Distribution of the Positional , Influentials and Administrative Staff ........... Profile of Percentage Distribution of the Hospital Director, Chief of Medical Staff, and the Administra- tive Staff ..................... The Rank Order of Responsibility Among the Adminis- tration and Their Activity. Rate ..... . ...... Summary of Profile Percentages by Area of Inter- action for Administrative Conference ......... Comparison of the Percentage Distributions of Inter- action by Bales' Four Areas for the Daily Administra- tive Conference of a Psychiatric Hospital and the Administrative Conference of a Community General Hospital for Several Role Groups ........ viii Page 72 73 75 76 77 79 80 FIGURE LIST OF FIGURES Page Schematic Diagram of the Organization of the 7 Governing Board, Administrative and Medical Units of the Hospital ..................... 6 Attending Membership of the Major Decision-Making Units of the Hospital and the Extent of Overlapping Status Linkages .................... 8 ix CHAPTER I INTRODUCTION TO THE PROBLEM AREA This study was undertaken as part of a project1 concerned with the ways in which a hospital was structurally linked with the community in which it was situated. Loomis has succinctly phrased an appropriate introduction to this problem area. To understand a social system it is always necessary to know how it is linked to other systems and to any larger systems of which it may be a part. To understand social change it is necessary to understand how agents of change link themselves to the target sys- tems which are changed.2 One aspect of the project was devoted to an analysis of the linkage between external demands (e. g. , population increase, the expansion and development of technology, the spread of communication and transporta- tion networks, increased demand for medical service, etc.) upon the com- munity facilities (including the hospital) and the community power structure.3 0 1This project was the Hospital-Community Relations Project under the co-directorship of Jay W. Artis and Walter E. Freeman, Department of Sociology and Anthropology, Michigan State University, and was spon-L sored by funds from the National Institutes of Health. ' 2Charles P. Loomis, Social Systems: Essays on Their-Persistence and Change, D. Van Nostrand Company, Inc., Princeton, 1960, p. 34. . In this work systemic linkage is examined primarily in terms of the dif- ferentiation of the Gemeinschaft-Gesellschaft dichotomy. . For a study of the relationship between systemic linkage and the establishment of norms, see Robert C. Hanson, "The Systemic Linkage Hypothesis and Role Con- sensus Patterns in Hospital-Community Relations, " unpublished manuscript, Department of Sociology, University of Colorado. 3Alexander J. Muntean, "Community Change and Hospital Develop- ment: A Case Study of Community Power Structure, " unpublished Master's thesis, Michigan State University, 1959. Muntean found, through the analysis of community issues, that the "paternalistic power structure" changed from an "exclusive elite" type to a "fluid influentials" type as the community experienced a transition from a one-industry community to a "commuter's town. "4’ A second aspect of the project was concerned with the linkage effects of external demands on the hospital organization pg s25 Harkness found that under conditions of increasing external demands on the hospital as a facility in the community, structural differentiation and functional specialization occurred, and as a consequence dissociation between the goals and norrris of the three main segments of the hospital was observed.6 Nevertheless, analysis of the organization demonstrated that its output was independent of the dissociation between the goals and the normative systems of the three major decision-making units of the hospital, and in effect was the consequence of systemic linkage. Introduction to the Present Study A third aspect of the project was the present study which was con- cerned with the consequences of systemic linkage of community influ- entials for the decision-making process of the hospital.7 Systemic linkage ‘Ibid., pp. 186-213. ,sJames P. Harkness, "Hospital Organization in Transition: A- Secio- logical Analysis of Interlocking Social Systems, ” unpublished Ph. D. dis- sertation, Michigan State University, 1961. 6The three main units were the board, the administration, and the medical staff. .7Loomis has presented several types of systemic linkage systems. See the appropriate sections on "systemic linkage, " 22. git. , pp. 32-326. Also, see C. P- Loomis and J. A. Beegle, Rural Sociology-~The Strategy of Change, Englewood Cliffs, N.J.: Prentice-Hall, Inc. , pp. 231-232 where the process was called "social-cultural linkage"; C. P- Loomis, "Toward A Theory of Systemic Social Change, " Rural Sociology in a Chang- iggSocietyzAProceedinis of a North Central Rural Sociology Committee (NCR-S) Seminar, Columbus, Ohio; The Ohio Agricultural Extension Service, was operationally defined in this study as the extent of overlapping repre- sentation of status occupants in the major decision-making units of the ' hospital}. In this case the major decision-making units were operationally defined as the board of trustees of the hospital, the working. committees of the board and of the administration. The administrative committee in- cluded the chief of the medical staff and the director of nursing, but the decision-making units of the medical and nursing services were not analyzed in this study. The design to assess the consequences of linkages of community influentials in the organization was essentially twofold in nature. . First it provides a methodological procedure by which the particular administrative decision-making process of the board of trustees, its two working com- mittees, and one administrative committee of the hospital may be analyzed. More specifically, it examines the observed dynamics or process of con- trol of community influentials within a specific organization in the community-- a non-profit general hospital. The design will allow a test of cross- validation of two commonly used measurement systems for identifying November,l959, pp. 12-48; C. P. Loomis,”Tentative Types of Directed Social Change Involving Systemic Linkage, " Rural Sociology, 24 (December, 1959), pp. 54-57; and C.- Redekop and C. P. Loomis, “The Development of Status-roles in the Systemic Linkage Process, 3? Journal of Human Relations, 8(March,1960), pp. 278-283. . For a discussion of several types of linkage systems in organizations see Robert Dubin, "Stability of Human Organiza- tions, ” in M.. Haire (ed.) Modern Organization Theory, New York: John Wiley and Sons, Inc. , 1959, pp. 218-253. , For a discussion of several mechanisms involving the articulation of the role-set, another form of systemic linkage, see Robert K. Merton, Social Theory and Social Structure, Glencoe: The Free Press, rev. and enlarged ed., 1957, pp. 368-379. 8This definition is an operational form of the analytical definition by Loomis, Social Systems, gp. c_il” p. 32, where it is defined as: ". . .the process whereby one or more of the elements of at least two social systems is articulated in such a manner that the two systems in some ways and on some occasions may be viewed as a single unit. " community power structure and prediction of control in the community-— the "reputational" and the "positional" techniques9 by means of a third and completely independent measurement system--Bales' "interaction process analysis. "10 The second aspect of the design is represented by the formu- lation of a central hypothesis that will be examined at different levels of the heirarchy of decision-making units in the hospital, and which in its operational form will allow a test of the validation of the influential and positional techniques. This hypothesis is that community influentials are the most active participants in the decision-making process of the hospital and exert more control over the non-influentials than they in turn exert over the community influentials. 11 A community influential was operationally defined as either a ”tOp influential” (TI), an "economic influential" (E1), or both. R Finally, this study contains a critical analysis of the reliability pro- cedures commonly used in profile analysis and recommended by Bales, his co-workers, and various other researchers in this area, and suggests an alternative statistical procedure for estimating inter-rater reliability. . Since this discussion is more technical and does not readily fit into the main text, it is set forth separately in the Appendix. 9A discussion of these techniques will be presented in Chapter III. 10Robert F. Bales, Interaction Process Analysis: A Method for the Study of Small Groups, Cambridge, Mass.: Addison-Wesley Press, Inc. , 1951. 11This hypothesis is an operational form of the analytical hypothesis by Loomis, where it is stated: "The status-roles which control the power in the power centered systems, also control most of the important decision- making for that system, " _i_b_i_£1., p. 22. 12"The identification of community influentials will be discussed in Chapter III. The Hospital Organization and Extent of Systemic Linkage Figure l is a schematic diagram of the organization of the hospital, exclusive of the multitude of committees at the lower echelons of the organization. The board of trustees represents the highest controlling unit in the hospital. It is ultimately and legally responsible for the entire operation of the hospital. It appoints the two working committeesnthe executive committee and the operations committee. , Representatives of the woman's board of managers sit on both of these committees. The woman's board of managers is primarily concerned with the operation of the school of nursing, the voluntary auxiliary, housekeeping and dietary, and also the budget. The board of trustees is primarily concerned with the hospital physical plant, expansion, fund raising, operating finances, appointment of the medical staff, and the general operation of the hospital. Within the last four years, however, more and more responsibility and authority have been placed in the two working committees. With the clos- ing of the school of nursing in 1961, the woman's board has been largely confined ‘to the auxiliary,-‘, housekeeping, nursing services, student nurses'scholarships, and various philanthropic projects they devise periodically. The administrator and his staff are directly responsible to the board of trustees, and are responsible for the administration of all aspects of the organization. The chief of the medical staff is not a member of the administrative staff and this is represented by a broken line in the diagram. The test of the central hypothesis was made at the level of the board of trustees, the executive committee and the operations committee all of which have community influential representation. . As an introduction to the central hypothesis, a historical analysis of the 50 year participation pattern of economic influentials on the board since the founding of the FIGURE 1. SCHEMATIC DIAGRAM OF THE ORGANIZATION OF THE GOVERNING BOARD, ADMINISTRATIVE, AND MEDICAL UNITS OF A COMMUNITY GENERAL HOSPITAL BOARD OF TRUSTEES, [WOMAN'S BOARD OF MANAGERSJ J f EXECUTIVE COMMITTEE} I OPERATIONS COMMITTEE } I --------- - — —{DmECTofij (ALSafety Engineer l==fPurchasing Agent I Engineering Con- _, Central Stores 7 7, sultant _ and Receivin l I Director of Administrative | Medical Educ. '4 Resident I I ] I .CHIEF OF ASSISTANT DIREC- ASSISTANT ASSISTANT ‘ COMP- MEDICAL STAFF TOR 81 DIR. NURSES DIREICTOR "o"* ‘ DIBEQFQB "s"* ROLLER; ' l l Medical Section Nursing Service Dietary Maintenance Credit 81 Billing Internal Medicine Medical Medical Personnel . . . . . Admitting Ped1atr1cs Surgical Records Radiology , , Laundry School of Nursmg Rad1ology Accounts Pathology P bl Housekeeping aya e Surgery Section Surgery,. Emer- Laboratories 81 Pay- ency Room 81 roll General Surgery g Pharmacy GenitO-Urinary Central Supply OT 81 PT Switch- Surgery In-service Edu- Building Machine Room board Orthopedic cation Program Opthamalogy . , Cashier Ear, Eye, Nose Obstetrics-Nursmg and Throat ~Pediatrics-Nur sing Dentistry 81 Oral Surgery Anesthesiology Neuro- surgery General Practice Section Obstetric s and Gyne- cology Section * Since these two individuals are not differentiated by title, the code letter that will be used in the analysis is used to differentiate the two. hospital will be presented in Chapter II. The sub-hypothesis to be tested there is that economic influentials have been withdrawing from the active participation of the decision-making units over the 50 year period. Chapter III will examine the central hypothesis for the board of trustees. Chapter IV will examine the central hypothesis for the two working committees of the board--the executive committee and the Opera- tions committee. Since there were no community influentials participating at the administrative level Of the organization, the influence structure of the director, his assistants, the director Of nursing, and the chief of the medical staff will be examined. The hypothesis to be tested in Chapter V is: the director and the chief Of the medical staff are the most active participants in the decision-making process and exert more control over the nursing and administrative staff than they in turn exert over the two . positional influence figures in this unit. Figure 2 presents the list Of individual participants in the major decision-making units analyzed in this study. The capital letter identifies the community influentials and the asterisk denotes the chairman of the unit. It can be seen that four classifications of participants are represented in these units: community influentials and non-influentials on the board, and non-influential representatives of the administration and the woman's board of managers. The extent of status linkages (e. g.,overlapping membership) is also presented. Community influential "B" was a member Of 3 units: the board and both working committees. He was also chairman of the board and executive committee. Individual "e" was chairman of the Operations committee and was a member of two units as were "C“ and "D. " Board members "g, " "h, " and "l, " were members of 2, 3 and 1 units respectively. More will be said about the occupational composition of the board and each unit in each chapter. Only the members who were in attendance during the meetings analyzed in this study are recorded in this figure. Hmflaogmcu 3355500 cmfibmAOse a. N m m w m N N N m woman-Mum mcwmmmioer mo Hung—A X X X X X muadumflmmax X madam 30:58 mo 330 *X nopomfin— .m ouconownou u>wumhumwcfi§< ream X X X X wuunguawdfloe-ucoz X X XX *X X muongoz w ouufigoo mcoflmnumo X X X when—508 wagon/1:02 X X X X X X *X mucagcgl N ovum—«5800 0330on X X X muonfiufi mnflo>unoz x x x x x on ex on 33802 H mucundnfi mo pumom B > o a 8 U. m .H v a 03% n w H E wirm mien 12.0 **m 3.4 poniddd flab enema-no.2 mmduw mow arcane-HR. mo vudom ,mmdflccg mogdguaowmwuofl mo pudom madam—03 lion-HOD pom ooflmtmdfiafizs 05. mo mozaduaomcumom Z .. mo mmzuducom unmom ii NEH. 977* J H. = + Board Pos1t10n Per1od I Per1od II If I < 11. 2 _ 1910-1934 1935-1959 If I =—.- II = 0 1 13 0 + 2 l3 0 + 3 25 25 0 4 17 25 .. 5 6 25 .. 6 25 11 + 7 12 25 - 8 0 0 O 9 13 25 - 10 24 20 + 11 17 0 + 12 7 0 + 13 0 0 O 14 12 0 + 15 11 21 - 16 19 13 + 17 25 9 + l8 13 25 - Totals 252 225 + = 9 - : 6 : x Sign Test: (a) Ho: I = II 0 = 3 (b) p(xfi4orx310, N: 15):.059, two-tailed test. (c) x = 6 (d) Accept Ho since p (x) = 6 = . 61> .059 level of significance. :1: A nonparametric s1gn test was used because assumpt1ons necessary for a parametric test for related samples could not be met. See: Sidney Siegel, Nonparametric Statistics for the Behavioral Sciences, New York: McGraw-Hill, 1956, pp. 68-71, and Table D, p. 250. The 50-year period was divided into two equal time periods in order to equalize the number of possible board positions per period. However, since the data in Table 1 indicates a decline only for the last decade, a sign test was computed on the average influential representation for the first 40-year period and the last 10-year period. Again the null hypothesis (H0: 1: II) was accepted [ p(x) = 5 == .30, two-tailed test, since . 30 > 005 level 17 Table 3. .Number of Economic Influentials in a Midwestern Community and on a Hospital Board of Trustees, 1910-1959 Ec ono 'c Influential s Time Perioda In Community On BoardC N N % 1910-19 30 14 47 1920-39 57 24 42 1940-59 73 18 25 Total 160 56 35 a . . . There are several cases who were economic 1nf1uent1als and who were tabulated in more than one time period for both the community and the board of trustees. Data on community influentials were taken from Clelland, 22° Sit: CData in Tables 1 and 2 utilize as a unit of analysis the number of influentials per 18 board positions per year. The unit of analysis for this table is the absolute number of actors who are influentials for each speci- fied time period. Hence, the middle period reflects a high turnover of influentials on the board. 18 number of economic influentials in the community under analysis. 11 There appears to be no corresponding increase in the number of economic influentials on the board but rather a rough bell-shaped distribution was found, with the highest absolute number of influentials represented on the board during the 1920-1939 period (N = 24). There was a steady de— cline in the percentage of influentials in the community who were board members, from 47% to 25%, but then this might be expected in light of a fixed number of board members and an increasing number of economic influentials in the community. The second set of evidence with which to examine Schulze's hypothesis is. contained in Table 4, which traces the historical pattern of Offices on the board, and committee memberships and Chairmanships. Both Schulze and Clelland find a general decline in the number of economic influentials serving as executive officers of the organizations and public offices they examined. . Like most organizations, the hospital board changed its executive structure over the years. For the first decade, the executive structure had an average of 5 Officers: president, first and second vice- presidents, secretary, and treasurer. During the 1920's and 1930's the number was reduced to 3 on the average, and from 1943 to 1959, the average number increased to 7, including a chairman and a vice-chairman-- both primarily honorary positions. Inspection of Table 4 shows that for all board offices held, there was a general decline in the percentage of llClelland points out that there are no objective standards which one may employ to assess the increase or decrease of economic dominants in the community. Obviously, the number of influentials in any one period will be a reflection of the cutting points of the criteria utilized. 7 These points shift for each time period (decade) to compensate for shifts in in- creased property evaluation and the growth of industries. It is conceivable that one could set the cutting points to maintain a constant number of in- fluentials for each period. Only on the basis of the limits used by Clelland is an increase demonstrated. It is an open question as to whether utili- zation of different cutting points would affect the analytical relationship between the number Of influentials on the board and in the community, but there are no apparent reasons that would lead one to suspect that different points would affect this relationship. 9 1 81m?» Acme n3 oGOaumnoflfimom unopfimopm Eamon. cm can when» not» .pumon 05 mo >HOOmE can. no>o mucogmonm Eamon. my coma 3E0 9.5: 0.8:» JESSE Mom .053 HO msumcg wcfinhdcr new mdoflwmom 252.9? cwmdooo :3? mnopom 13638:“ 9.9.3 msogflo on, pasonm OH : Cadet» Mom Gofizmom: mo 8.8m 93 Ga ma £93.25 usuuauaoaucwommm .Ho cofioofio >n. McAOMOIuEHdSGQm :EOSG: 3.633 mg“ a“ cofifimom >uo>o oocflm .0. Am mw mm um .mc 92 3... mm; Ztm «um. hm cm on ago New 18.08 mm 2 mm 2. ON hm wv mm mu ooH OH OH Hm MN on omuomma no OH ma we ZN mm mo om we on «C CH ow mm mm 31.0on mm o A: 3 m mm mm NH mm 00 O 3 mm 0H Hm omuomma mm o 3 5m w 1: Se 0N my om m 0H Na. 3 mm omuommfi om w A: we Mg m; Hm mm mm ow m 0H om mm om «Luci: gum 3mm, ofinfimmonm p33 p33 odnwmmona Bum EOE manwmmona 363 303 uswmmom Bum Bum ozwmmonm as Z Z es Z Z o5 Z Z as Z Z es Z Z mmflmcmguflmgu ouOfiEEOU 033086” mmfimuuagcz «cogmounm mooflmO SAN pofiuom mafia. 0335800 :4. . moofizggoo no EHMSEODGGH OZEOGOOM 0330 9:303. mfimfldodfifi Owgocoom i omoanoaoa doc» Hum mcoflfimom .5 mucumnfih mo venom Hdfimmom e do demceEfidEO one mQEmuonEoz ccufiaaoo .mWOEOE 03me mm mfimflcogmfi ofigoaoom mo coflmunomcumom - JV 0138 20 available offices which were held by influentials, although the trend leveled off somewhat after 1930. The percentage of influential representation dropped from 50% for the 1910—1919 period to 31% for the 1950-1959 period with some fluctuation in between. The data in Table 4 also reveal interesting support for the Form and Miller position that economic influentials are moving to the more strategic organizational positions. For the entire historical period, one can see a rough U-shaped distribution of influential representation for the office of the president of the board, seemingly the most important board position. Six of the 9 men who were presidents of the board were economic influ- entials. When cast in the form of "position per year" as in Table 4, slightly over half of the offices were occupied through the years by influentials; but for the last 16 years, the president has been an economic influential. The representation of influentials on committees fluctuated from 51% of) committee positions for the first decade (1910's) to a low of 23% during the 1930's and back to a high of 63% and 48% for the last two decades. The same fluctuating pattern was found for committee Chairmanships, again with the low percentage of influential representation during the 1930's (38%) and higher percentages for both an earlier period (50% in the 1910's) and later periods (67% and 59% in the last two decades). A U-shaped distribution was found for the influential representation on the executive committee, which had power to act in place of the board. It is interesting to note that for all categories--board president, all committee chairmen and members, executive committee members, and attendance (see Tables 4 and 5)--the decade of the 1930's exhibited the lowest influential representation. Even for all offices held, this period is second lowest in representation, second to the last decade by virtue of 1% difference. It is tempting to suggest that the reason for this lower repre- sentation during the depression was a consequence of influentials' concern for the practical affairs of the general economy plus the fact that this period 21 was marked by a low capital outlay for expansion by the hospital (an area of prime importance and concern to the board). However, wielding Occam's Razor, we must explore more proximate explanations. _ Since the com- mittee members were generally appointed by the president of the board, it is important to establish that association, if any, exists between the president's status and the status Of the committee members he selected. By comparing the number of economic influential presidents by "office per year" for each decade with the number of influentials who were selected as committee members, a definite association was found (particularly for the most powerful of all committees--the executive committee). When the time period was characterized by a high percentage of presidents who were in- fluentials, there was a high percentage of committee members who were influentials (see Table 4). 7 Conversely, when the time period was character- ized by a high percentage of presidents who were not economic influentials, a marked reduction in the percentage of committee members who were influentials was found. This association may be accentuated by pointing out that for _a_1_1 those years in which the president of the board was an economic influential, 78% of the executive committee was composed of influentials. Finally, although the balance of influentials to non-influentials has been fairly even on the board, there has been a definite increase in influential representation on the executive committee and committee chairmenships over the last 30 years (see Columns 12 and 15). _ Data in Table 5 provide the third set of evidence with which to examine the influential withdrawal hypothesis, and also allow a comparison of attendance of influentials with non-influentials. The attendance of each board member was tabulated up to his time of separation from the board. . NO evidence of economic influentials declining in participation was found, as measured by attendance (or a decline in interest i_f attendance can be taken as an index of this attribute). In fact, attendance increased for both economic influentials and non-influentials for the last 30 years. When the .ooamoflwcmwm mo H03: :3. A mm . u H u can :5 H H an aUv . .38 333.95 .28. u 3 u z .e IAx oo o w sod 3G a u H "em 3 amps swam 22 m N Z c u 0 Mn "TH H I v n + No mmo . oom .2 om Sm to .2 233. in we hm: mum mo vmm :m omuomofi + on a: own cm of doom deuce: + om we of em .2. me omuomoa + mm I: own me 03 Sum amuomz , .. em as Hmm we 03 m; 2.02; mummucoouom wooden—«e popcofi< pdoufiw op couscofiw poonoufiw pcotssou cook-3m uncouonm Z odfimfim uncouom Z ofinwmflm swam Z Z @3qu o n H n H H, mwcflooz mmcfiooz oEfiH. .. u 3 ,V. H NH + u H A Hm 3330:qu ufiaocoom mfimficooflfiudoz emnm E H omzuoaoa .mHmflnoaaw-OMIGOZ can mHmSCoOQdH OZEOGooH >0. woodcuts mmcflooz Unmom Hmfimmom mo owmunoouonm can ponEdZ . .m 3an 2, 3. attendance of influentials with non-influentials was compared by means of the sign test, the hypothesis of no differencewas accepted, although, on the average, the non-influentials attended 14% more meetings thaninflu- ential s . Summary and Conclusions This chapter has examined a hypothesis proposed by Schulze that economic influentials are withdrawing from local decision-making partici- pation by means of an analysis of the historical pattern of economic influential representation on a hospital board of trustees over a 50-year period. . In the last decade, economic influentials declined sli htl in representation on the board of trustees, but the decline was not statistically significant. . For the last 30 years, a small but definite decline in influ- ential representation in all board offices combined. was noted. - Even so, it should be pointed out that these declines in representation were more gradual than for any organizational trend found in Cibola by Schulze as well as by Clelland in the present midwestern city. This pattern suggests that perhaps hospitals, as well as other organizations involved in community service, will remain the last organization Of influential participation. In other words, the rate of withdrawal is not uniform for all community organizations, and of course may not be complete for any given community. By examining other intra-organization statuses such as the office of president, committee Chairmanships and memberships, and also attendance patterns, evidence to support the Form and Miller hypothesis of increased participation at the strategic organizational level was found. . Similarly, since the representation of influentials on the board has continued to be substantial, this suggests that the hospital organization is a strategic com- munity organization, if Form and Miller are correct. . In this light, one might posit several propositions which will need further testing. 24 (1) Organizations representing community service areas, such as hospital boards, are "strategic" organizations in that they represent both a symbol of prestige in the community and a mechanism by which the economic influential can. fulfill his obligations of "community stewardship"--part of which specifies expectations of management of community resources. 12 This proposition has some related, though unsystematic, empirical support. From observation, several board members appeared to take great pride in discussing their role as actors “interveningl' on behalf of their friends in the community or of physicians concerning problems revolving around admission of patients, nursing service, parking facilities, etc. . Form and Sauer found that the most frequent community "issue" identified by the "top 40" community influentials in this same community was a recent hospital expansion fund campaign (which was headed by the president of the hospital board reported here). - Eight of these "top 40" were members of the present board. 13 (2) One might expect that leadership and control of nonprofit community general hospitals shall be in the hands of economic influentials longer than most other community organizations, but that the pattern of “withdrawal" will be a withdrawal to the occupation of strategic statuses within the organizational structure.14 It is suggested that this is a function of (1) above plus the increasing bureaucratic complexity of the hospital operation and hence an increasing importance of decision-making at the committee level. 12The role of hospitals as significant industrial units should be noted. The present hospital, although not exceptionally large as hospitals run, employs over 800 persons and operates on an annual budget exceeding four million dollars. I 13'William H. Form and Warren L. Sauer, "Business and Labor Images of Community Power Structure: A Comparative Analysis, " a paper read before the Fifty-fifth Annual Meeting of the American Sociological Associ- ation, August, 1960. ' 14The determination of the scope of generality of these propositions must be solved by further comparative research. 25... .It is suggested that the combination of a concern with economic matters "external" to the community, a period of low capital outlay by the community hospital, and more important, the—type of-status occupant who controls the selection process of key board positions produced a low representation of economic influentials in board participation. . Should these conditions persist, and should the presidency revert toa non- influential, one-would expect a lessened participation pattern by economic influentials in the next decade. _ What are the implications of such findings for the, "dynamics" of the actual decision-making process in the hospital? One might expect that by virtue of the kind of representation on the board (about. half are economic elite in a corporation and financial sense), their primary concern would be centered on financial problems. . One might also expect that economic influentials would more actively control the top level decisions of the hospital. . These problems are examined in the next chapter. CHAPTER III THE EMPIRICAL-MEASUREMENT OF CONTROL OF ECONOMIC ‘ INFLUENTIALS AND TOP‘INFLUENTIALS ON THEHOSPITAL BOARD OF TRUSTEES: A CROSS-VALIDATION ANALYSIS Introduction This chapter will analyze a meeting of the governing body of the hospital--the board of trustees. - Chapter II was concerned with the his— torical pattern of participation of economic influentials on the board over a 50 year period. . However, little may be said about the actual influence or control of community influentials in the current operation Ofthe hospital. This chapter will examine that problem. The hypothesis to be examined is that community influentials1 are actively involved in the decision-making process Of the board and have greater control over the non-influentials and administration of the hospital than they, in turn, have over the com- munity influentials. There have been several approaches to influence or power structures in the community, and as a consequence, various methodological techniques have been devised to identify the power structure and measure their influ- ence in the community. This chapter will attempt to cross-validate the use of two techniques in current use in sociology: the "reputational technique" and the "positional technique. " A brief discussion of these approaches will introduce the data and the analysis. 1A community influential is operationally defined as an‘EI, T1, or both. 26 27 The Reputational and Positional Approach to Power For approximately the past 10 years, social scientists have become more and more interested in the phenomenon called "community power structure, " after the pioneering study by Hunter.7‘ Sociologists have generally followed his approach to the measurement of community power, which, inbrief, is based on a sociometric-like procedure ofzasking individuals to nominate or in some cases rank. other individuals whom they feel are influential in the affairs of the community. Generally, nominations are first taken from a panel of "knowledgeables, " and .the nominees in turn are asked for further nominations until some degree of consensus of overlapping nominations is reached. The technique has come to be knOwn as the - "reputational technique for determining community influentials. "3 .ZF. Hunter, Community Power Structure: A Study of Decision-Makers, Chapel Hill: The University of North Carolina- Press, 1953. ' 3Specific research studies which have relied on this technique are: Hunter, ibid.: F. Hunter, R..C. Schaffer, and-C. G. Sheps, Community giganizams, Action and Inaction, Chapel Hill: 7 The University of North Carolina Press, 1956; J. M.. Foskett and R. Hohle, "The Measurement of Influence in Community Affairs, ” Research Studies of the State College of Washigngton, XXV (1957), 148-154; R. L. Agger and V. Ostrom, "The Political Structure of a Small CoWunity, " Public OpinionQuarterly, 20 (1956) 81-89; R.. E. Agger, "Power Attributions in the Local Community; Theoretical and Research Considerations, " Social Forces, 34f(1956), 322- 331;. R. 0.. Schulze and L. U. Blumberg, "The-Determination of Local Power . Elites, " The American Journal of Sociology, 63 (1957), 290-296; D. C. Miller,. "Decision-Making Cliques in Community Power-Structures: A Com- parative Study of an American and an English City, " The American Journal of Sociology, 64 (1958), 299-310; D.. C. Miller,. "Industry andCommunity , Power Structure: A Comparative Study of an American andan English City, " American Sociolggical Review, 23 (1958), 9-15; R. O- Schulze, 'lEconomic Dominants in Community, Power Studies, " American Sociolojical Review, 23 (1958), 3-9; R- E. Agger and D. Goldrich, “Community Power Structures and Partisanship, " American Sociological Review, 23‘ (1958), 81-89; E. A. T<--Barth and B. Abu-Laban, "Power Structure and the Negro Sub- ~ Community, " American Sociological Review, 24 (1959), 69-76; 0.. E- Klapp and L. V. Padgett, "Power Structure and Decision-Making in a Mexican 28 The individuals with the highest number of mutual choices are called "top influentials. " Perhaps one of the most promising developments in the reputational method has been the work. of Miller"r and Hanson5 who have attempted to predict the outcome of community issues based on the type of power structures in the community, the "solidarity of the top influentials, " and the degree of involvement of "parts of the institutional power structure.-" As has been pointed out in the literature, 6 two other techniques for determining influentials are the "positional method" and the "is sue approach" method. Thel'positional method" operationally defines an influential as such by virtue of the fact that he occupies a seemingly important status in the community (typically in business, finance, civics or politics). This approach is best typified by Schulze, Blumberg7 and Polsby‘.8 The Schulze and Blumberg technique is of interest because they combine the "reputational" with the "positional" technique. . Schulze reports that inCibola the economic influentials, reputational influentials, and Border City, ” The American Journal of Sociology, 65 (1960), 400-406; W- H.. Form and W. L. Sauer, ”Community Influentials in. a Middle-sized 'City, " General Bulletin No. 5 of the Institute for Community Development, Michigan StateUniversity, 1960;. E. C. Erickson,. "The Reputational Tech- nique in a Cross-Community Perspective: Selected-Problems of Theory and Measurement, " Unpublished Ph. D. dissertation, Michigan State Univer- sity, 1961. i 4'D. C- Miller,. "The Prediction of Issue Outcome inCommunity Decision-Making, ” Research Studies of the State College of Washington, 25 (1957), 137-147. 5R- C.. Hanson, "Predicting aICommunity, Decision: A Test of the Miller-Form Theory, " American SocioloLical Review 24 (1959), 662-671. 6See for example: P- H- Rossi, "A Theory of Community Power, " a paper presented to the 1960 Annual Meetings of the American Sociological Association and R. O. Schulze and L. U. Blumberg, pp. §1_t.. .7Schulze and Blumberg, pp. gi_t_. 8N. W- Polsby, ”Three Problems in the Analysis of Community Power, " American Sociological Review, 24 (1958), 796-803. 29 heads of voluntary associations were all in essential agreement that "substantially the same set of persons" were "most influential in the affairs of the community. "9 However, there was very little overlap between those individuals who were nominated as ”top influentials, " and those who. occupied the seemingly important economic statuses in the community, the "economic influentials. " Schulze concludes that "EI's" are withdrawing from the power structure of the community. Polsby, in his New Haven study, drew up "pools of leaders" for three issue areas: education, political nominations and urban redevelopment, and assessed the degree of participation of themembers in the pool in each of the issue areas. However, contrary to his stated aim of eliminating 3 priori criteria for determining leadership pools, he chose all the formal status occupants within each issue area. . For example, in the area of political nominations he drew 497 names--all the political office holders in the city of New Haven. . With such large lists, it is of little surprise that he finds only two per cent of the economic elite participating in the political nomination issue. In a recent Notre Dame symposium on power, Dahl also criticizes most users of the reputational technique, and their findings that there is a single group of power wielders who are the (decision makers in the com- 10 munit . He, Wolfin er, 11 and Polsb 12 maintain that there is ,no eneral Y 8 Y g scope of influence, but rather the American community is made up of a .9Schulze and Blumberg, 313. c_i_t_., p. 295. ”R. A. Dahl, ”Equality and Power in American Society, .. in W. v. D'Antonio and H.. J.. Ehrlich, eds. ,. Power and Democracy in America, Notre Dame: University of Notre Dame Press, 1961. “R- E. Wolfinger,”Reputation and Reality in the Study of 'Community Power,” American. Socioleical Review, 25 (October, 1960), pp. 636-644. l“LN, Polsby, _o_p. c_it. 30 number of segmented "power structures" and that an individual's influence is limited to certain sectors of the community. The scope of influence is presumably limited to the institutional area within which the influential is identified or in which he presumably has skill. Thus, the critics of the reputational technique have maintained that the technique does not measure influence. Rather, they say it measures things other than influ- ence: e. g. , status, money, wealth, formal leadership, personality traits, etc. One of the best defenses in support of the reputational technique has been made by Erickson, who has attempted to determine the relationship between a general index of influence and several indexes of influence on specific issues using the nominational technique in six communities in the United States and Mexico. He found the general influence index to be closely related to indexes constructed on a specific community issue in all six communities. . He also found that the nomination technique was a fairly reliable instrument in identifying influentials in one community on the basis of a three year test period. 13 It is not the purpose of this chapter to elaborate on the finer points of the arguments, both pro and con, of the ”reputational technique" versus the "positional" and "issue-area techniques, " however worthwhile that may be. . The purpose of this chapter is to present the analysis of data designed to analyze how the hospital is structurally linked to the community in which it operates and determine the extent of control of community influentials in the decision-making process in the hospital. Since Form and Sauer have identified the key and top influentials by 1 the reputational technique, 4' and Clelland has identified the economic l3E. C..Erickson, 92' c_i_t_. 1‘W- H. Form and W. L- Sauer, "Community Influentials in a Middle- sized City, " General Bulletin No. 5 of the Institute for Community Development, Michigan State University, 1960. 31 influentials15 for the community in which we were studying the actual decision-making process of the hospital, the opportunity to validate the relative merits of the "reputational" (e.g. .TI) and "positional techniques" (e. g- EI) for measuring influence or control was made possible. Seven of the 18 men on the board of trustees of the'hospital are economic influentials and 8 are top influentials during this 1959 period. Four are both EI's and TI's,. but only two of these were present at the meeting analyzed in this chapter. It can be seen in Table 6, that the men on the hospital board share to a greater extent an overlapping EI-TI classification than do their counterparts in the community. On the board 57% of the EI's are also TI's, whereas in the community only 31% overlap. . For TI's on the board, 50% are also EI's, whereas in the community only 31% overlap. . It would seem that the board has the elite of the two classi- fications. Table 6. The Degree of Overlapping Classification of EI's and TI's of Men in the Community and Men on the Hospital Board m Overlapping With Other Classification Number .Number Per Cent Community Economic Influentials 39 12 31 Top Influentials 39* 12 31 Hospital Board (18) . Economic Influentials 7 4 57 Top Influentials 8 4 50 Forty men were nominated as top influentials, but one individual was not interviewed and was dropped from the study. See Warren. L. (Sauer, "Labor-Business Images of Community Power: Convergences and Diverg- ences, " unpublished Ph. D. dissertation, Michigan State University, (1960), pp. 25-28. 15Clelland, g. c_i_t_. The criteria for economic influentials during the 1958-1959 period were those individuals who occupied the top formal 32 Table 7 lists the 18 members of the board, the El and TI classifi- cation, the code letters which identify the individuals in the tables which follow, the primary occupational identification and the corresponding economic unit in the community for each member. It can be seen that 4 of the 7‘ EI's and 3 of the 8 TI's were present at the annual meeting analyzed. A legitimate methodological question is the extent of repre- sentativeness of this single meeting of the board. From a substantive viewpoint this problem is extremely difficult to solve unless one makes observations over a long period of time (e. g. , probably 10 years or the mean number of years of service per member, which was 13 years) since the board devotes the majority of its time to varying substantive problems over time. For example, during the 1959-1961 period the board devoted most of its energies toward building and expansion problems. In late 1961 and 1962 it was primarily concerned with labor union negotiations. One might speculate the neither of these problems will consume the same proportion of time that it did in these two periods for another 10 years or more. . However, the representativeness of the findings of this meeting may be partially assessed in terms Of the participation patterns of the influentials and non-influentials in attendance. If the community influentials who were not in attendance at the meeting were in general absent from board meetings, then the meeting is "more representative" than if the meet- ing analyzed happened to be one in which most individuals who generally attend were absent. The same may be said for the non-influentials. Table 7a- was prepared in order to assess the participation patterns of the board members who were in attendance and who were absent at the statuses of: industrial or commercial units having an assessed evaluation of $750, 000 or 250 employees; banks with resources of 6 million; savings and loan companies with gross assets of 6 million; and insurance companies with total assets of 6 million. Overlapping directorship was also used as a criterion. 33 ACC: . ~ ~.\.T. .t-unv-snuu. -. o no. .~.nIO.V--uh/~ nun... tin... -~ .db HR. Hm COMPMHOQHOU 0..-dies .uflvauhdflva HO HOHUOHMQ .ma m m >ucow< oucdudmda .ucopwmounw .4: HR. Hm 30560 none-H .JCoEmounm 03> .2 HR. HM Bod. as Kenyon-fia- .NH HR. HM noxcmm Josefina-h....umflmfinuwdpfi .60:qu .: E. E 3633:: 88m rapeseed .2 HR. Hm Amootwoago ooo .NH €03.25 up "dogma-H85 commommm ammo: :oflmuomuoU 325- 339282 3.3280 can unopflmounm oofir .o e e presences ooo .23 ”soaps-:95 commommm omoS Essa-COO H332 .ucogmounm 03> .w w w .uofimom .h a c 3338 A ”Hetero omo: room adopted to; .o o e c8222 .3. Ernesto wmo: foo ewe-Emma E5 duodenum .peooaopm 85 .m p Q $693380 mno mooo .mwwa Eon—9:96 communes mom: H0335 Hm "Hoodoo mmo: Essa-£00 oucsudmcH .ucogmounm .w. o o senator-65 soup-85h .poxeem .oopnom .m m m 30335 A "sou-3.396 communes ammo: mundane/oz mo nowmcsz impede—U was .uofiom ...Ho£mEo—Dnm .N < .4. Amootwoacco 005 .m 2.5338 NZ "Gofiddamx-o communes omoC “838.8980 oxenm mo nommcsz Hone-80 can unogmouna .H usofimofiflmmgu coflmoflfimmmau flab Osgocoom can coflmmsOoO muoaaog venom Zmflcosfifl QOH. HMS-Hogwfi ofifiocoom a JED ofiEocoom can .EofimadooO twang-Ana 6350 3an .GOSmOflfimmsHU HR. was Hm >3 muongoz venom .> cans- 34 .Hongm HIHI .8 HM .m >3 HuvHHHHHHmHVH mum mHmHHGmHHHHHHHuHHoZ .coHHmoHHHmmmHo .HmHHHuHquHH mHHH 0H :oHHHHUHum CH vauzmcm mcHHmoE QMIH Hm mmlcmvcvfim HHH Ho: mm? HmHHHUH>HHuGH 05 .55 moumoHHuHHH coHHmonHmmmHo :HB: .8 :5? :< .HmHHGdeHHHH uHHoc m mmHMGmHmvHv .832 ammo .533 23 "SB .HO HA cm . .m. 3 HmHuGoHHHHHHH cm mvamcmHmwv .5»on HMHHQMU oHHBN . .wHumH mmwmm .oH oHoHHuooH mum . .mvoudom Hang/mm SOHH mpoHoo mHHHHHH UHHHHOHHouo 93 HHHOQM GOHHMEHOHGH 03H; x x HovHBoH< H. .n HHHMHHHHwGOO H H Hfifimom mfi Ho H8025 m oHHHvaHHvfiHw HHH muonfiogudoz HR. Hm @HOHW wGwflmH—OHU .HOHOwHQOHflH ”and HCGUMmMHm ow." E. E 88 .23 6232?». 3333 $3: I ll >cmm§oo HooH .HmmmquZ monm Sumuouovm 3:0chon muH> .: F Hm Ammo>onEo mmH mooo .owma E33395 Hummmmmmw omoi 33:80 no: 83232 12050 can ”>5chon .3 35 meeting analyzed. . The attendancepattern of each boardmember for the past 10 years was tabulated for each of the influential classifications and the average number of meetings per member was computed. It can be seen that the economic influentials who were in attendance at the meeting generally attend 73% of all meetings held, whereas the economic influentials who were absent only attend 15% of the meetings held. A less striking difference was found between. the top-influentials present, who attend 79% of the board meetings compared to the influentials who were absent, who attend only 42% of the meetings, on the average. Table 7a“... Average Percentage of BoardMeetings Attended by Each Board Member from 1950-1959, by Community Influential Classifica- tion and by Attendance at the Board Meeting Analyzed _._--,-__-__, --—- - —-»—- -----— —-- ,____ == The Board Meeting Analjzed Board Member Members in Attendance Memb‘ersAb‘sent « Classification (Percentage of Meetings Attended 1950-1959) Economic Influentials 73 15 ~ Non- influentials 75 58 Top~ Influentials 79 42 Non-influentials 71 50 It would seem safe to say that the community influentials in attendance at the meeting analyzed are generally in attendance at meetings of the board. Although the non-influentials in both classifications who were in attendance are the most active participants, the non-influentials n_ot in attendance are generally more active than their absent counterparts among the influ- entials. . Taking the members in attendance at the meeting analyzed as a whole it would seem that they certainly are not atypical participants in board meetings in terms of a 10 year attendance pattern. 36 Procedure and the Measurement of Control Verbatim electrical recordings of the board and committee meetings were made in 1959-1960 after a 3 month period of non-participative observation on the part of the author. . Good rapport was believed to be obtained as evidenced by the highly confidential material, which was allowed to be recorded. The meetings were recorded and observations were attributized andunitized in order to submit them to Bales "interaction process analysis. "16 Table 8 contains the set of 12 categories of the Bales system. The activity of small groups, role partners, etc. , can be classified according to the 12 categories which may be combined to provide various measures of indexes of behavior of the group. The measurement of control or influence which shall be used to cross-validate the "reputational" and "positional" techniques is Bales' Index of Directiveness of Control.” The assumptions underlying this index are as follows. Of the twelve categories, three types of activity-- categories 6, 5 and 4 are all concerned with directing the activity of actors in the group and are used in the Index. Of the three, category 6--"giving orientation, information, repeating information, clarification of a point, or confirming a fact, " is the most non-directive. An example would be an actor who says "There are two points I would like to make. . The first one concerns our hospital charge offs. " Both of these would be scored as category 6. The second category, number 5, is more directive than number 6, and, as one can see from Table 8,. is composed of the "giving of opinions, evaluations, analyses of the situation, the expression of feelings, wishes, “R. F. Bales, 32. c_it_. . (For a critique of the reliability procedure using interaction process analysis, and estimates of reliability, see Appendix A. ) ”Ibid. ,. p. 145. '37 2:: Table 8' . Set of Categories Problem Areas: Social-Emotional Area: Positive Reactions Task Area: Probl em-Solving Attempts Task Area: Questions Social-Emotional Area: Negative Reactions E . Shows solidarity, raises other's status, gives help, reward 2. Shows tension release, jokes, laughs, shows satisfaction 3. Agrees, shows passive acceptance, under stands, concur s, complie s ab< \ 4. Gives suggestion, ‘direction, implying autonomy for other B 5. Gives opinion, evaluation, analysis, expressed feeling, wish 6. Gives orientation, information, repeats, clarifies, confirms 7. Asks for orientation, information repetition, confirmation C 8. Asks for opinion, evaluation, analysis, expression of feeling 9. Asks for suggestion, direction, possible ways of action 10. Disagrees, shows passive rejection, formality, withholds help D 11. Shows tension, asks for help, withdraws out of field 12. Shows antaggronism, deflates other's status, defends or asserts self A sub-classification of problems to which each pair of categories is most relevant: a. Problems of orientation b. Problems of evaluation c.. Problems of control d.. Problems of decision e.. Problems of tension—management f. Problems of integration * a Robert F. Bales, "A Set of Categories for the Analysis of Small Group Interaction, "American Sociological Review, 15 (April, 1950), pp. 257-63. 38 desires, " etc. For example, if an actor says "I wish we could beat this Blue Cross formula, " this would be scored in category 5. The third category in the index is the most directive of the three and is comprised of the process of "giving suggestions or directions. " For example, if an actor says "The first thing we should do is to provide a new parking lot for our patients, " this would be scored as category 4. The Index~ of Directiveness of Control is based on the preponderance of 4 and 5 in the total interaction process as compared to 6. In other words, the more the actor makes direct attempts to control the decisions of the group (by direct suggestions or expression of evaluations) in proportion to the amount of factual information he provides, the higher the index of control value. 18 The formula is as follows: 4 5 4 + 6 + 5 + 6 IDCX‘ y = 2 where: x and y 4, 5, 6 2 two given actors Bales' categories constant The value the index may assume may range from O to 1. 0, from no control to maximum control. It should be pointed out that the index does not differentiate between two actors who have no interaction and actors who interact but have no frequencies in categories 4 and 5. In both cases their IDC value would be 0. Since means will be computed for each role group in the study, the IDC values which contain any interaction inany of the three categories but which yield a . 00 value will be identified by an asterisk in the tables. . Means computed for each role-group will be computed on the basis of the number of index values which contain any 18Presumably it might be argued that other categories might be in- cluded as in index of control and perhaps even all 12 could be so arranged to estimate control. Bales ADR Index is a theoretical and methodological approach to such a question. IE. , p. 167 ff. 39 interaction. This will be explained below. Also, since this study is concerned with the comparison of community influentials with non- influentials, the activity of any actor which is directed toward the group as a whole is deleted from the IDC value since such activity cannot be differentiated in terms of whether the actor was directing his control over influentials or non-influentials. The IDC values are presented in Table 9 for board members as well as for the 3 individuals present who were not board members. . The latter included the director of the hospital, an architect, and a hospital con- struction consultant, one of whom was either an E1 or a TI. . The board members were classified as an EI or non-EI (EI), and their index values computed. . Tables 10 and 11 contain the mean values for these scores. The means are computed in the following manner. The mean IDC value of the EI's in Table 10 (. 24), is the sum of the IDC values among the EI's (A, B, C, D) divided by the total number of indexes containing any interaction. Thus: . 50 (B vs. C) + .00* (B vs. D) + .13 (C vs. A) + .34 (C vs. B) = .97; hence the EI:EI mean IDC = 3.327. = .24. It can be seen from Table 10 that as one moves across the table the values increase for every cell. The EI's exerted most control over the FI's who were not members of the board (.42), followed by the E-I's who were members of the board (. 35), and exerted the least amount of control over their counterparts on the board (. 24). This pattern was also true for the Ef's who were members of the board. They, too, attempted to control the actions of their counterparts who are not members of the board as evidenced by an index value of . 35, followed by a mean of . 22 with TED—1's who are members of the board, and followed by a mean of . 17 with the EI's who are members of the board.. Finally, the IE-I-‘s who are not members of the board have the highest mean index of control among them- selves (. 50), followed by a mean of . 24 with E's who are members of the board, and a . 12 with EI's who are members of the board. 4O .GOMHUMHflHGH mGAMHGOU NOUCH #53. .QHJHNK/ OO. * N O w. W m -- om. oo. oo. oo. oo. 84 oo. oo. 8....8. x 9m 0. cm. -- om. oo. oo. 8. mm. 8. oo. oo.*oo. H m 8. oo. -- S...oo.*oo.*oo. oo. 2. oo. 3. n 8. om. mm. -- oo. 8.3? 8. oo. co. «m. x mm. mu 8. oo. *8. oo. -- cm. 8. 8. 8.1.8. 8. m m 8. oo. oo. oo. oo. ..... om. oo. oo. 8. 8. a p 8. $1 8. so. oo.*co. .. oo. *oo. $.35. m m m o. oo. om. oo. 8. oo. oo. oo. .. oo. 8. oo. o m 8. oo. om. oo. oo. om. oo. oo. .. em. 2. 0 FE s 8. S. om. om. oo. 3.30. *8. cm. -- co. m 8. oo. oo. oo. oo. 8. cm. 8. oo. oo. .. < x H z a m c o o o m < 88¢. deHucofiHdeiaoz. mHmedosHHGHucoZ mHmfidodHHGH UHEocoomH mammzmzioz mmmmzmz om0. 91le .5230 mdmuo> HmH Aomm Homo .m.HmH HoguO mdmho> Hunlm. Aomm .m..H|.m.H .8530 mamuo> HM nodm .m.Hm.H|I .5230 mamh0> HmHII aode Ho mosHm> Houudoo Ho mmmlzoxrfloonwfll Ho #065 .o 3an ** 41 Table 10. Mean IDC Value):< for EI's and FIT—'8 by Hospital Board Membership BOARD MEMBERS NON-MEMBERS A‘ctors - EI's EI's E's (A, B,C,D) (8: £9 g: 11) (jujak) H 3 EI's .24 .35 .42 E (D E _ 'g EI's .17 . 22 . 35 (d O m H —— g EI's . 12 .24 .50 8 <1) 2. c: 0 Z * . . Means are computed on indexes With any interaction There is clearly a pecking order of attempts at control. The as who are not members of the board received the greatest amount of control, which makes sense in terms of their functions in the management structure of the organization. ,However, the as on the board also received more control attempts from the EI' s than they in turn attempted with the EI's. Table 11 contains the simple relationships between EI's and ES, disregarding board membership. If the mean scores of EI's (. 38) are sub- tracted from EI's (.15), a mean difference of . 23 in favor of the EI's’ is obtained. 42 Table 11. Mean IDC Values for El's and 1735s Actors 'EI's EI's E1' 8 '. 24 . 38 .Eil s . 15 .30 Tables 12, 13, and 14 contain the same datafor "top influentials" who attended the meeting analyzed. These values are computed in the same manner as in Tables 9-11; and the same letter refers to the same person, except the capital letter here refers to the "top influential" classification. In Table 13, the same general pattern found for EI's also holds true for TI' 3, with one exception. The T—I board members have a higher mean over TI's on the board (. 26) than they do among themselves (. 03). When one looks at the simple relationship between. TI's and —T—I's in Table 14, however, one sees that TI's have a higher mean index over TI's (. 38) than the converse (. 20). The mean differenceis ,. 18 in favor of the TI's. Table 15 shows the heirarchy of control for bothEI-EI-comparisons and TI-TI comparisons. . The pecking order runs as follows: EI's over EI's,. F7118 with themselves, EI's with themselves, and finally E—Ils over ' EI's. The same heirarchy was found for the TI's. At this point it may be said that an independently derived measure of the degree of control exerted in a decision-making unit of a community organization has been established. It was used for purposes of cross- validating boththe "reputational technique" as represented by the TI classification, and the "positional technique" as represented by the E1 classification. 43 Gowuodnoufi mafimpdoo x323 «do. .oaHm> ,oo . * N m . -- om. oo. oo. 8. 84 oo. oo. 8.. oo. eoo. x m om. .. om. oo. 8. mm. 8. oo. 8. co. *8. .n arm. m 8. oo. -- S. .48. *8. oo. 2. *8. 8. en. a m. S cm. om. mm. .: oo. *8. 8. oo. 8. oo. 3.. a s 8. oo. *oo. oo. .2. oo. oo. 8. cm. *8. 8.. m m oo. 3. oo. 8. oo. -- oo. *oo. *8. S. *8. e 9.5 m. 8. cm. 8. co. co. co. .. oo. 8. oo. 8. a W 8. oo. om. oo. oo. oo. oo. -- 8. en. 2. e m o. oo. 8. 8. oo. 8. cm. 8. oo. .. co. 8. a. m 8. cm. om. cm. 8. *oo. *8. cm. mm. .. co. m 92. s 8. oo. oo. oo. 8. cm. 8. oo. oo. oo. -- < x. w H a a m e e o .m. m < 2894 MHdmuGQQEWGOZN ..de«uG03GGHIGO—IH: :dequ0ng anH... mammzmiioz mmmmzmz om3 m.H.H. .385 35.34, HH. SommH Head "m.H.H.. nonuo mama»; H.H. nomm "PFHQAHO mumps.» H.H. nomm "PHH. H.330 mdMMo> H.H. nomfl Ho “53> 33:00 Ho mmodozuoofln NMJBHEH . .NH 3nt 44 Table 13. Mean IDC Values* for TI's and T-I's by Board Membership ,= 3 BOARD MEMB_E_RS NON-MEMBERS TI'S TI'S ’TI'S Actors (A, B, F) (c, d, e, g, h) (i, j, k) S TI'S . 25 . 33 . 50 ..o E o __ 2 TI' S . 26 . 03 . 39 '0 a (U o In S __ .0 TI' 3 . 09 . 26 . 50 E a) 2. G 0 Z at: » Means are computed on indexes With any interaction. Table 14. Mean IDC Values for TI's and TI‘_I's Actors TI's TI" 5 TI'S . 25 . 38 TT'S‘ .20 .26 45 Table 15. Rank Order of Means Between E1- E1 and TI- TI Comparisons of Index of Directiveness of Control ll EI : —E-I Comparisons TI : TI Comparisons 35 Rank Rank SE EI'sz-Ei's .38 1 1 .38 TI'szTI's E'szfis .30 2 2 .26 fi's:TI's EI's : EI's .24 3 3 .25 TI's : TI's l—BI'S:EI's .15 4 4 .20 l—T—I's:TI's The problem of measuring the relative difference of control between EI's and TI's is complicated by the fact that 2 of the 4 EI's were also TI's and conversely, 2 of the 3 TI's were EI's, as mentioned earlier. . It is possible to compare the means of the two distributions in Table 15, but they do not appear to differ too significantly. Perhaps the TTS are slightly more active in control attempts than their counterparts, the E's. Another comparison possible is between the EI's who are not TI's and the TI who is not an El. Table 16 contains these comparisons, and for the one set of values we have, the E1 has the higher index score. . This one case, obviously, does not allow extended generalization. . Nevertheless, it is clear that for the meeting analyzed in this chapter, the hypothesis that community influentials exhibit greater control over non-influentials is supported.19 In order to more adequately assess the role of community influentials on the board, data in Table 17 is presented in the form of the percentage 19Two limitations must be observed in the use of this index of control. . First it is not an adequate measure of what is generally called "power. " Most sociologists working in this area generally follow the Weberian concept which includes an element of force or contest. . The second limitation is that this index should perhaps more accurately be labelled an index of attempts at control since it contains no measure of the acceptance of the suggestions or directions madein the meeting. 46 Table 16. . Comparison of IDC Values of EI's with TI's Who Do Not Share Overlapping EI-TI Classification EI (3'71) , TI (El) Actor C D F __ c -- .00 .50 BI (TI) D .00 .. .00 TI (E) F .00 .00 -- distribution of interaction of community influentials and non-influentials for the 12 categories in the Bales' system. The interaction of community influentials was clearly concentrated in category 6 (giving orientation, information, clarification and confirm- ation). What is very interesting to note is that the profile of interaction of the hospital administration and consulting staff was closer to the com- munity influentials than was the profile of the non-influentials who are members of the board. The latter were plagued by the greatest amount of tension or conflict and exhibited a higher incidence of disagreement and antagonism than the other two classifications of participants. This pattern was also true in terms of the domination of the amount of interaction that goes on in the board meeting. Table 17 also contains the mean number of acts per individual in attendance at the meeting. The community influentials had the highest number of acts, (especially the TI's with a mean of 122 and the EI's with 95), followed by the hospital staff (if = 90) and finally by the non-influentials on the board (3(- = 62 for EI'S and 52 for TI'S). The data in Table 17 were presented in such a form as to allow exami- nation of how members of the role groups distributed their total activity according to Bales' 12 categories. However, Table 18 presents the same 47 Na 2. Nm NNH Nm 2. Na 3 33802 3a 33.. no nongz m mam :N NoN 3m mam EN EN a: 39.. do nenfiaz Eon. an 2: 2: 2: an 2: SH 2: :80 new 389 m N S e m N 2 e Enaomfine. nsonm .NH 2 o a a 2 o o a census nsogm .2 s m ON N N. m S a $3985 .2 a o N N a o N N antenna .nonnemmsm now means. .a N a o a. N a N N nosnfigm .8220 new nan... .w e N N a e N a. e nonsense: 635225 new Bane. .e 3 2m 2 ow mm 3 2N ow aonnfinowfi £33335 32.6 a S M: mN 3 S 2 ON 3 consign .eoafiao 330 .m Na 2 Z 2 N2 2 w 2 8535 .aonnemmam $30 .1. w a w n m a S w . neenma. .N N z m N N N e a ennezem nonnaes 959% .N N o a e N o a m . 32828 36.6 .2 venom 2.73 an... 6.3 Ed .3 venom 2.3; 334.3 E.o.m.$ H.308 9m 9m. m.H.H. HmuoB m.HmH "3mm. m_HmH mowuomoumu .monm m uo3gozudoz . 9335on Unmom m uo3§oH>HlumoZ m .83802 pnmom mEmuo3§m2 venom >3 .m.H.H. was .93. 6me .mLH .HoH moflowoumu .monm >3 cofiomuoucH Ho cowgflflmfln owmaCoouonH Ho 3&0an .: 0H3mH. 48 m 2 S m NH a 82:388.. nsogm .2 a n S a a. S mosses 26$ .2 a N2 0 a. a: N newsmana .2 o m zN o w 2 8535 503833 new Sana. .a e o om e a S nosefinem .8820 new nan... .m w 0 ON 34 o A: coflmEuoHcH .coflmpcoCO .HoH mva< .5 3 N E E a. S connfieofi: .cosficeflo $30 a S a Z S w a nosnigm .8825 830 .m 3 a. B 2 m N2 303035 6.83333 $30 .4 m m 2 w a 2 33?. .N N. S S a. E a enneaem mosses nsogm .N N 2. N N N N fannezom nsonm .2 71.7: 2466.3 Ed .3 3323 £4.33 E.o.m.$ meern mam m.H,H. m..H.m.Hu Pm mHM mofiuowoumO .monm muo3EoH>HaGoZ muo3EoH>H Upmom muo3EoH>HucoZ . mpm3cfloH>H Unmom m33mso38o2 ppmom >3 .m.H.H. paw .m.H.H. .m_HmH .m_HmH 30H mmfiuowofimu amonm 03» Ho 30mm :3 303882 ponH :oflomuoadH Ho ommucooponlelonZ .wH 0H3mH. 49 data in a different form. . It presents the percentage of activity in each category that is accounted for by each member of the role group. In other words the relative proportion of activity within each category taken as a unit of analysis that is accounted for by each member of the role group. (on the average) is presented in Table 18. Here it is seen that community influentials accounted for the greatest percentage of interaction in showing solidarity (1), agreement (3), giving directions (4), asking for information (7), asking for evaluation (8), asking for suggestions (9), and showing tension (11). . Non-influentials accounted for most of the tension release (2), disagreement (10), and the showing of antagonism (12). Category 5, giving evaluation, is fairly evenly divided among all three classifications of participants. The administrators and the board members accounted for the largest proportions of giving information. . The conclusion one gleans from this analysis is that the data appear to lend further support to the hypothesis that community influentials are actively involved in the decision-«making process of the board. They have the highest rate of activity; they ask more questions, accept (and reject) more solutions to problems; and they show more tension. However, they are not as involved in conflict and are less antagonistic than the non- influentials on the board. Summary In summary, a third and completely independent measurement system has empirically validated the use of both the reputational technique (TI' 8) and the positional technique (EI's) as predictors of control or influence in the setting of the hospital board of trustees. A definite heirarchy of control on the board was found. The EI's on the board exerted the most control over the ES who are members of the hospital management staff, less con- trol over the EI-‘s who are members of the board, and least control was exerted among themselves. The same general pattern was found for the TI'S. 50 The community influentials clearly dominated the proportion of activity at the board meeting, and accounted for the largest proportion of interaction in the positive emotional area, the problem solving area, and in the area. of asking questions in order to solve problems confronting the organization. The non-influentials on the board reacted to this by exhibit- ing a higher proportion of negative emotions. The hypothesis that community influentials are actively involved in the decision-making process of the board and exerted a greater amount of control over the decision-making process is given considerable empirical support. CHAPTER IV THE EMPIRICAL MEASUREMENT OF CONTROL OF EI'S AND TI‘S ON TWO WORKING COMMITTEES OF THE BOARD OF TRUSTEES Introduction Chapter III examined the central hypothesis of this study--that com- munity influentials are actively involved in the decision-making process of the hospital and that they exert greater control over the non-influentials than they in turn exert over the community influentials, for the board of trustees. This chapter will examine the same hypothesis for two lower level units in the decision-making structure of the hospital--the executive committee and the operations committee. The Executive Committee As Figure 1 in Chapter I illustrates, 1 the executive committee is the most legally powerful committee in the hospital (outside of the board). It has legal authority to act in place of the board on many policy making decisions. It appoints physicians to the medical staff; it has authority to purchase land and enter into contracts; etc. However, in practice, it typically makes recommendations to the board which are seldom (if ever) reversed. The executive committee is composed of 10 persons, 5 from the woman's board of managers, and 5 from the board of trustees. Generally, two or three members of the administration are in attendance, but they are 1See page 6,. Chapter I. 51 52 not voting members of the committee. The committee members, their EI:TI classification, their primary occupational identification and economic unit are presented in Table 19. . The chairman of the committee, who is also the president of the board of trustees, is both an economic influential and a top influential. The second member is also an economic influential and a past president of the board. The representatives from the woman's board of managers, for the most part, are daughters and/or widows of socially prominent "old city-families. " They are also quite active in community activities. Data in Table 20 provide the test of the control hypothesis for the executive committee. Inspection of the table shows that the heirarchy of control favors the community influentials. The community influentials have a mean IDC value of . 25, followed by the non-influentials on the committee (5(- = . 23) and the administrators (i = . 15). The IDC means for the EI:TI increases for every~ cell'in the first row.. He exerted little control over his fellow EI who was not a TI (3(- = .06), moreover the fiz-TI'S who were members of the committee 0? = . 23), and the most control over the administrators of the hospital (i. = . 37). The same pattern was found for the EI:-'fi on the committee ()_('s = . 04, . 21, and . 33, respectively). The EI:TI'S on the committee exerted the least amount of control over the EI:TI (SE = . l9), concentrated their control efforts on the EI:TI-I (.46), less over their colleagues (35 = . 25), and exerted a modicum of control over the administrators (i = . 22). . Finally, the administrators exerted no control among their colleagues (5(— = . 00), somewhat more over the community influentials (3(- = . 12 for the EI:TI and . 06 for the EI:TI—I), and appeared to concentrate their control efforts on the EI:TI's on the committee (3(- = . 17). If the differences between the means are taken as the final index of control, the differences between the administrators and the community in- fluentials were in favor of the EI:TI (SE = .25) and the Eli“? (id = .27) d 53 683$on CH HomN>Hmcd mmcfloog $33 .9833 Ho: HUHHU #3615ch 03» 3.23 moumoHch :oHumoHHHm HmmmHo hdeofihmm m3 03 «53335 mmHo ..H.H.: so :5? 03H. .mcoflmo 138930 3303 CH HmHHCdeHGHuGoc .m moumcmHmov nouuoH omdo .833 03» Head 33333qu cm moudcwfimop .833 3339038 * a a uvHHofi—nHEoO o o nouoonHQ unwumHmm< .m H Hmfimmom e3» Ho .83th Am noguum 3:35.33 muo3EoH>HuGoZ .HI..H.I WM. uopuO Hmcumfimuh Ho newscmz Ho 3033 3 3 and: ooGMHSmGH .033 > > noumodpm .333 5 D Gunmen owned Ho .HopoHunHounH Ho 32633 a G noquoMH .3303? HR. HImHI 353500 mwcgmm. «ED HdHoGwcHh .JGoEmoHnH ooH> Goflduuchngae GofimnumeHEHflofl :oflmuum 31.53% muowmnmg Ho venom Pawns? . muowmcmz Ho Undom 95883 .. o o mummmcmz Ho pnmom Pumas? .w muvmmcmz Ho Houdom Pawns? .H. o mnommcmz Ho vumom mango? . momumsfih Ho venom . m 3 3 >stEoU 38.02 .ucopHmonnH moH> mmoumdufi Ho pumom .w w m uofimmm mooamduh Ho pumom .m o O pumom xcmm Ho GMEAHQHO .poufiom moeumduh Ho Unmom ..N m m AcmEHHNeHU moufiggouv Mundane/oz Ho ummmamHZ Heumaeo cam nofipm 1.3.333an mmvumafifi Ho Unmom .H *coHumUHHHmmmHO *aoflmoHHHmmmHO “ED oHEocoomH one coflmmdooo poucomoudom “ED Hmfimmom HmSGesHHGH HmfldogwfiH Hone .33802 moguHEEoO AHOH. UHEocoom HHGD UHEOGOUH Hoax coflmmdooo .COSMUHHHmmmHU HHLHMH >3 .muo3EoH2 ovuuHEEOU 0x33300034 {33 @HDMA. 54 Table 20. Means of Index of Directiveness of Control of EI's and TI's by Committee Membership on the Executive Committee (Two Meetings) Committee Members Administrators Actors EI:TI EI:T-f fizfi 733?sz Total (B) (G) (g. h. 11.11. V. W) (i. 0.13) 8 m EI:TI -- .01 .23 .37 .25 H _ .33 3 EI:TI .04 -— .21 .33 .25 5.5, fiz'fi' .19 .46 .25 .22 .23 ° 2 U "F371 "'1‘? 12 .oo 17 oo 15 Administrators (see Table 21). Although the EI:TI barely controlled the E—IzT—P-s on the committee (id = . 04) the EI:TI—i lost out to the fifis on the committee (id = 41.25). In order to examine the activity aspect of the central hypothesis, the percentage distribution of interaction of members of the executive committee was computed. . These data are presented in Tables 22 and 24. . Table 22 contains the percentage distribution of each EI:TI classifi- cation for committee members and the administrators. Almost one-half (49%) of the EI:TI's behavior was giving orientation to the committee. Ten percent of the time he was asking for information; 9% of the time he was agreeing; and another 9% of his behavior was concerned with giving directions or suggestions. The Elz—T—I- spent more of his time agreeing (15%), giving opinions (14%), and showing antagonism (12%) than did the EI:TI, 55 Table 21. Mean Differences Between the IDC Values of the Individuals on the Executive Committee by EI:TI Classification Committee Members Administrators Actor EI:TI EI:TI- EI:TI" E'iz'r—I' (B) (C) (g. h. n. II. V. W) (1.0.13) a, EI:TI -- -.03 .04 .25 8 1'3 ‘3 B EI:TI -- -.25 .27 E .8. _ _ 0 E :T -- . 04 o g I I EI.TI_ -- Administrators but he spent less of his time giving information (29%). The other members of the committee correspond very closely to the EI:-'I‘—I in each category but tended to spend somewhat more of their time giving information (39%). The administrators were largely concerned with giving information (57% of their time) and evaluations (13%). One of the most interesting findings was the high rate of activity of the EI:TI. His mean number of acts per meeting was 487. Although it appears to be extremely high in comparison with other classification means, this comparison is somewhat misleading. The other classifications (e. g. , EI:TI-'8) have more than one member in them, some members of whom are active and some of whom are not. . For example, one EI:TI on the committee had a mean of .440 acts per meeting (h). . The director of the hospital had a mean of 419 acts per executive committee meeting. It is apparent though, 56 oom NS Now NE 22 no :22. monooz no.2 2933252 Hem muo< mo 2232.922 M 3% .m 2.2 .N 232 .2 woo mom .2 32 2o Anmonoofi 3 39.. z 2392. 82 82 oo on 82 82 82 280 now 2892. o o a v m N2 o Emwcowmfifiw 9502.3 .NH 2 2 2 o N 2 2 28228.2. 9.626 .22 m m m m m a N 33925 .222 2 N 2 2 N. 2 2 282835 ..no2onommsm 82 923. .o N N 2 o m N 2 oo22oo2n>m .8325 .82 9232 .22 m o m N 22 m 2: 823882.22 .oosooooflo 82 32...... .2. 3 S. om on mm on 2. 28238822 533825 330 .o 22 N2 222 m2 22 2.2 2. 52339222 .8235 830 .m o m. m N o w o nonoofio .nonnommam $30 .2. 22 22 N2 22 22 2 o 33?. .m 2. m m m 2. m 2. 33232 82228.2. 9.65 .N 2 o N 2 o 2. 2 3232.228 259$ .2 95.5 nmuwmv P2222 . a .o .22 so .> .a .n .22 .3 6. 2m; awn—0H. HduOH. HMHOH mwmm IH.|H....mmHI WAN HEM—”m mvwhomwwmo .MOHNm . mnoumuumwcfiEp/w 98322232 ooufiEEoU Ammnflooz 03H; whom—duumflcwefiuafi pad whongoz 03235800 o>330mxm~ 20w GOSMUGfimmMHU HR. Ucm Hum >n mmwhowoumu .moamm >9 GowuomuounH mo GowudflmhumwQ mwmucoouonw mo ofiwohnm .NN 2.:nt 57 that the community influentials are more active than the committee non- influentials (SC-"s = 292 versus 121) and somewhat more active than the administration (E's = 292 versus 242). . Two possible reasons for the high activity rate of the EI:TI at these meetings are: first, he was the chairman of the committee; and second, he was proposing that the committee approve the acquisition of real estate adjacent to the hospital and the EI:TI-Is were against it (for the most part) since the hospital was going through an expansion program and was heavily indebted. The fact that the property was eventually acquired is an independent measure of the power of the ' EI:TI. . He was also the chairman of the united community hospital expansion board and a good proportion of his activity at both meetings was involved in explaining the plans and results of that board. Table 23 contains the list of major topics which were covered in these two meetings and the observer's evaluation of the extent of disagreement. Table 23. List of Topics Covered by the Executive Committee in Two Meetings Subject Presence of Disagreement Financial Report . No State and County Charge Offs Some Building Program No Control of Out-patient Traffic in Nursing Home Some Surgery Scheduling and Occupancy , No Foundations and Fund Raising No Acquisition of Real Estate Yes) 'Blue' Cross Formula Problem No 58 Table 24 contains the mean percentage of interaction per person in each category that is accounted for by each classification of membership on the committee. , It is interesting to note that the one EIETI accounted for a greater proportion of the activity in every category except one, where the EIzfi accounted for slightly more of the solidarity expressed. The EI:TI accounted for 40% of the directions, 37% of the questions asked, and ranged from 20% to 28% of each of the remaining categories except for requests for directions (13%). This is further evidence of his high rate of activity. Among the non-influentials, the administrators were, in general, highest in the positive emotional areas (categories 2 and 3), attempted answers (categories 5 and 6), and disagreement (category 10); the committee members showed slightly more solidarity, control, questions asked (cate- gories 7, 8, 9), and negative reactions (categories 11 and 12). . When the mean percentage of interaction in each category for influ- entials is compared with non-influentials, the influentials accounted for more interaction in each category than did the non-influentials except for asking directions. The Ope ration s C ommitte e Introduction The operations committee is the "work horse" of the board of trustees. It meets bi-monthly compared to the monthly meeting of the executive com- mittee and to the quarterly, meeting of the board as a whole. . It reviews in detail the Operation of the hospital--the financial reports, wage surveys, parking surveys, construction programs, insurance coverage, etc. It is directly responsible to the executive committee and makes reports and recommendations to it. 59 2. 222 o n 22 2N 8222289332 36$ .N2 22 2.2 2 N2 2 2.N 2822.32. 36$ .22 22 m2 2. 2. o NN $3935 .2 2o 22 o 22 N 2 nonoona .nosnommom .82 329.. .o a 2 N 2: o mN oo2ooa2n>m .8823 .82 923.. .m n 2N N 22 2. 2.2 2823682222 .aosfiooflo .82 929.. .s a 2. 2 m 2. 2... 82222882222 53288220 330 o w 2.2 N2 2 22 ON ao2ooa2o>m 62222220 330 -m o 2.N N o 22 222. 28228.25 .nosnommsm $30 .2. o 2.2 S o m oN noonm... .N 2. 22 N2 m o mN 88232 28228.2. n3o22m .N o N 2. 2.. 2.N NN 3232.228 n3o22m .2 nbmnlm new 222 .o .22 23 .> .o .a .22 .3 202 22: 2.308 2.308 mnHlmw MWHM: HIRIQH HBQM mofiuomoumo .mofimm mucumbmdfigpaq 22.822.5on oofifihfioO Amman—0oz 025.5 22.203223222228222. was muongoz ovuuMEEoO o>flfiooxm 20m GowudoflwmmmHO HH. pad Hum >9. mofiuomoumu .mofimm 05 mo 220mm 2222 2203092323 mo ommuaoouonm Geog . .23 03mm. 60 Our central hypothesis will be examined again within the context of this committee. . Since there is only one influential on the committee, it is hypothesized that the influential is more active than, and exerts a greater degree of control over, the non-influentials on the committee. The Committee Table 25 lists the members of the committee, their EI:TI classification, their primary occupational identification and economic unit. . The committee is composed of four representatives of the board of trustees, two repre- sentatives of the woman's board of managers, and in addition, is attended by three or four members of the administrative staff. They are not voting members of the committee, however. The chairman of the committee is a vice-president of a financially prominent economic unit in the community. The chairman of the board. (and chairman of the executive committee), who is an EI:TI, is also on the committee. The two representatives of the woman's board of managers are active business women in the community and are both widows. One is the president of a construction supply firm. ! . Table 26 presents the data for the test of the control aspect of the central hypothesis. It can be seen that the heirarchy of control pattern remains the same. The EI:TI had the highest mean index of control (3(- = .40); the non-influentials on the committee had the next highest mean (i = . 38), followed by the non-influential administrators (D? = . 31). The EI:TI exerted the most control over the administrators (5(- = . 56) and less over the non- influentials on the committee 0? = . 26). The non-influentials on the com- mittee concentrated their greatest efforts of control on the EI:TI (SE = . 57), followed by the administrators (i = . 39) and their colleagues on the com... mittee (i = . 34). The administrators similarly concentrated their efforts on the EI:TI (SE = . 35), followed by the non-influentials on the committee 61 .maoflmo tflwmmgo 25022 222 Adflaofiflcfltcoc m moumcmfimop .23on omwo .2033 on» 9293322033: 2.2m moudamwmop .2323 2.3.3220 <2. v w HOHmR—mwcwgaoa «copwmom 22023233252222 m n 9029392200 nowaduumeEU/w o o .2300an “2533922. 22023232222594. 2 2 2.332.203 0:» mo .2300an 22023532225222 30232322225233 m Honaoztdoz G 2.2 uofimom mo 30pm? paw .uofimom muowmnmz mo pumom P252220? -0 82 E .00 >Hnafim 220302232200 . .ucogmoumtxm. mo 30.33 was unopwmounm muowmcmz mo vumom 952503 .m H H .00 .GOHH uommnmg 3.8280 pom «Gopfimonm mooamdnB mo Uuoom .2. A A .00 2.22232 .Empwmounm 03> movemdufi mo Unmom .m o o 22.22.23.920 ooHfiEEoO was 700 mmcgdmv 2.2223 Homucmowh .Jcopwmounm 03> mooumdufi mo «Zoom .N m m Mundane/oz mo nomads; 222.2280 was .23pr auogmflannm movumfiuh mo Uumom .2 *Gofldoflwmmgu *Goflmoflwmmmdu . 22229 ofiaocoom poacomonmom 322D Hmfiucofiflfifi 33222.52. paw 22032322000 Hmuwmmom paw 209802 032.222.2500 90H. ofigoooonm E 22:3 ofigoaoom cam 503222222000 .coflmoflfimmgu HBHHM >2. muonfiog 32325800 983.9890 .mN 0368 62 (i = . 17), and true to form, expend the least amount of control efforts among their fellow administrators. Table 26. . Means of Index of Directiveness of Control of EI's and TI's by Committee Membership on the Operations Committee (Four Meetings) Committee Mgrriers Admyslzrators Actors EI:TI EI:TI EI:TI Total (B) (89h: 1,111, n) (i, or P: Q) 3 :3 EI:TI -- .26 .56 .40 ‘3 3 E g — — E o EI:TI . 57 . 34 . 39 . 38 ° 2 0 ET? . 35 . 17 . 1o . 31 Administrator* The mean differences, the final measurement of control, are pre- sented in Table 27. Inspection of that table shoWs that the community influential lost out in control to the non-influentials on the committee (id = -. 31). . However, he maintained his control pattern over the adminis- trators (i = . 21), as did the non-influentials on the committee (i = . 22).. In order to examine the activity aspect of the central hypothesis for this committee Tables 28 and 30 were prepared. ~ Table 28 contains the percentage distribution of each. EI:TI classification of committee members and the administrators. The community influential concentrated his inter- action in giving information (35%), giving opinions and evaluation. (18%), and giving directions or suggestions (11%). The non-influentials on the committee spent a disproportionate amount of time in agreeing with the 63 Table 27. Mean Differences Between the IDC Values of the Individuals on the Operations Committee by EI:TI Classification (Four'Meetings) Committee Mil-pars Admi_13_ist_r_ators Actors 'ElzTI . EI:TI . EI:TI (B) (e. h. 1. m. n) (i. 0. p. q) EI:TI -- -. 31 . 21 o m t :3 f: .o E E 131' I -- 2?- o 2 U Administrator 3 other members (19%)—-although they tended to have a high percentage in this category generally. In the four meetings of the committee, they were low in giving orientation (18%). The administrators followed true to their pattern; highest in giving information (46%), followed by giving Opinions and evaluations (20%) and agreement (12%). - Conflict between the EI:TI and the committee non-influentials is evident by the 9% activity rate in showing ' antagonism, compared to only 3% for the administrators. Again,'the EI:TI had the highest rate of activity (7: = 424), compared with the non-influentials (J? = 257) andthe administrators (5(— = 187). Although the non-influential chairman of the committee had the highest mean number of interactions (802), nevertheless, as a variable it appears that the EI:TI had a higher mean number of acts than non-influentials on the average (i's = 424 versus 257). The non-influentials, in turn, had a higher mean than the administrators (is = 257 versus 187). 64 mem 2: SN 3.. 9.332 3a 2.3339: 8a 39.. mo 3.25.2 m 3m .e . Nmo.~ 03 .m 2; .2 .emaseefi a 8.3. 2 Each ma 2: 2: on: «sou nonm 1308 m m a e Seacomfiae. 36% -2 A o A N :oflmsoh 959% -3 m a. e a. $2330 .2 H H N o Gofloouwn— ..aoflmomwfim MOM 934 .w M a m m acreagm .8880 e8 9.9.. .m o m 2 m con—manowfi ..aofl.muao30 yo“ mxm< .n @N 3. 3 mm coflmfiuomfi .sofimunofluo $3.10 .0 S 8 2 S aotefieé .8380 330 .m m o 9 3 25300.30 .Goflmommsm mo>w0 J. 2 . .2 S a 33m... .m m w m m ommofiom c3388 959% ..N a a N a 323:8 952$ .2 3.9.0 .3 3.84313 Amy 939:0 mum Imwwm HBHHM moiomoumu .moamm H308 mucumuummgafiexw muonfioz ooufiEEoU Ammcflooz 55.5 muofiwuumficflnpaea pad 989502 oouuwgou .mcowumHoQO no“ cofiumoflflmmmfiu HR. paw Hum >0. .mowuowoumu .mofimm >2. aoflomuoufi mo cofldfluumwfl owmucoouom mo 330nm .mm .3an 65 . Table 29 contains the list of topics which were discussed in the four meetings analyzed. Three topics aroused some "heated" discussions, and again an extended argument centered around the acquisition of real estate between the EI:TI and the chairman of the committee ("e") and actor ”h" another non-influential on the committee. Both non-influentials were against the acquisition when it was first brought up and "1;" continued the opposition when the resolution was reported in the executive committee meeting. Table 29. List of Topics Covered by the Operations Committee in Four Meetings m 7 Presence of Extensive Subject Disagreement Financial Reports ‘ No Acquisition of Real Estate ’ Yes ‘Expansion of Fund Campaign - No County and State Charge Offs No Creation of a New Subsidiary. Corporation Some Building Program No Rennovation of Nurses Home - Some Table 30 contains the mean percentage of interaction per person in each. category that is accounted for by each classification of membership on the committee. What was most striking at this level was the fact that the EI:TI accounted for the largest share of the negative emotional reactions (categories 10, 11, and 12). . For example, he accounted for 40% of the tension shown, 27% of the antagonism and 20% of the disagreements. This was a consequence of the conflict over the acquisition of real estate with the non- influentials on the committee . 66 The non-influentials also had a high proportion of interaction in emotional areas, but less than the single influential. . The influential had the highest proportion of activity in 9 of‘ the 12 categories, again-further evidence of his activity. . The non-influential administrators, in the face of even greater con- flict over the real estate issue participated very little in the meetings of the committee. Table 30. Mean Percentage of Interaction in Each of the Bales' Categories By EI and TI Classification for Operations Committee Members and Administrators (Four Meetings) Committee Mega-tars Adinilis_t_r_ators Bales' Categories EI:TI "EI:TI EI:TI (B) (e,h,l,m,n) (i, o, p, q) 1.. Shows Solidarity 11 14 5 2. Shows Tension Release 21 10 8 3. Agrees l3 l3 6 4. Gives Suggestion, Direction 29 12 5 5. Gives Opinion, Evaluation 23 9 8 6. Gives Orientation, Information 27 6 10 7. Asks for Orientation, Information 14 15 2 8. Asks for Opinion, Evaluation 25 14 2 9. Asks for Suggestion, Direction 5 13 7 10. Disagrees 20 12 5 11.. Shows Tension 4O 12 0 12.. Shows Antagonism 27 . 12 3 67 Summary and Interpretations This chapter examined the central hypothesis of this study, that community influentials are actively involved in the decision-making process of the hospital and that they exert more control over the non-influentials in the two working committees of the board--the executive committee and the operations committee. Contrary to the pattern found for the board, the community influentials did not exhibit a greater amount of attempts at control at the two committee levels of the decision-making structure. In the smaller committee and more informal setting, 3 where the influential was not chairman, the non- influentials actually had a higher mean difference of Index of Directiveness of Control over the community influentials as well as over the administrators. On one major issue resolved during these four meetings, the acquisition of real estate, they lost out to the recommendation of the community influential.‘ Is is apparent from observation at each of the several levels of the decision- making heirarchy of the organization, that when issues were brought up, they were brought up at the lower levels where hostility, antagonisms, and pressures were openly brought to bear. However, the non-influentials were "out-maneuvered" as the issue went up the heirarchy of the organization 3The meetings of the operations committee were held in the director's office as opposed to the larger lounge of the nursing home for the executive committee, and the conference room of the local newspaper for the board meetings. ‘They were able to raise enough opposition to bargain with, however. They agreed to the acquisition of the property providing they could get a bank note to purchase the property rather than paying for it out of operating expenses as suggested by the economic influential. (This was not really a "bargaining factor" since the influential apparently had already informally contacted a local banker and had a verbal committment to that effect. However, it apparently was treated in this manner as a face saving device.) 68 where the influentials had greater representation. At the board level, hostility and opposition were not generally brought out in the open. . It is interesting to note that no matter how intense the disagreement was at the lower levels of the organization, when the final vote was called for (either at the committee or board level), rarely was a negative vote cast.5 a For example, one issue concerning the plans for parking at the hospital was commented upon by anElzTI on the board: "Their suggestions [the operations committee's chaired by "e", a non-influential] I don't agree with, but I'm perfectly willing to accept your suggestion. ,[The EI:TI president of the board] [Laughter] Because I don't think in any institution of any kind today you can go to without free parking, whether it's retail, wholesale, manufacturing, or hospital. . No matter what it is, I don't think it's possible. But I agree with you [the EI:TI president of the board]; I'll go along with it [on pay parking]. "6 In terms of the profiles of interaction, the community influentials consistently dominated the interaction 3313 _a: v_i§ the other members of the committees. When a decision was challenged they reacted negatively. The non-influentials on the committees also accounted for a fairly large share of both positive and negative emotions but less than the influentials. The administrators, in the face of conflict between the influentials and non-influentials, tended to reduce their amount of interaction e_n_ t_ot_o. The central hypothesis was not clearly supported at the lower levels of the decision-making structure of the hospital. Although the community influentials were rigorously active they were out-manned by the control efforts of the non-influentials. The influentials were only able to control 5A reading of the minutes over a 50 year span appears to> support this finding as a definite pattern. .Not a single negative vote on any resolu- tion was found recorded in a 50% random sample of minutes scanned. 6Taken from the verbatim electrical recordings of the board of trustees meeting, June 26, 1959. 69 the decisions of the committees by taking the issue to higher levels in the organization where they have greater support and where the non- influentials were more reluctant to engage in open opposition to the pro- posals. CHAPTER V THE EMPIRICAL MEASUREMENT OF CONTROL AND PARTICIPATION IN THE ADMINISTRATIVE. CONFERENCE The individuals who participate in the administrative conference are listed in Table 31. This committee is composed of the director of the hospital, the two assistant directors, the director of nursing, the comp- troller, the resident administrator,1 and on specified days, the chief of the medical staff. Table 31. Members of the Administrative Conference by Table Code and Position Title or Position Table Code Director of the Hospital i Assistant Director "o"* 0 Assistant Director "s"* s Comptroller p Director of Nursing r ‘Chief of Medical Staff t Resident Administrator q :5: See Figure 1, page 6 for the departments and the areas of responsibility of these individuals. 1Resident administrators are graduate students in hospital adminis- tration who serve a residency for one year as a condition for the partial fulfillment of the requirements of a master's degree in administration. 70 71 The hypothesis to be tested in this chapter is that the director and the chief of the medical staff, by virtue of their positions, will be the most active participants in the decision-making process and will exert more control over the administrative staff than they in turn will over the two positional influentials. Tables 32, 33 and 34 present the data to test the control aspect of this hypothesis. The mean Index of Directiveness of Control values for thefive meetings analyzed are presented for each in Table 32. The director of the hospital exerted the least amount of control ()7 = . 09) over the chief of staff which has been a common pattern among high status occupants at all levels of the organization. However, the physician's control mean over the director was . 36. .What is quite interesting is that the director of nursing had a high mean index of control in her interaction with the chief of staff (35 = . 81). Some of the extended interaction between these two individuals involved discussion about the hospital (e. g. , nurses) being caught between two conflicting orders from two physicians concerning the discharge of a patient. Both she and the assistant director "0" were offer- ing suggestions in an attempt to push for a decision which would clarify the discharge procedure. . Problems involving patient identification and practices which might lead to possible lawsuits were also discussion topics. Table 33 contains the means for the two positional influentials and the non-influentials (the administrative staff). , It can be seen that the positional influentials exerted more control over the administrative staff than they did among themselves ()—('s = . 36 and . 23 respectively), and more than the administrative staff expended uponrthe positional influentials (3(- = . 27). The administrative staff expended more control efforts among themselves 0? = . 33) than they did over the positional influentials (3(- = . 27). If the difference between the means is taken as the final measure of control, it is apparent that the positional influentials had the higher control scores than did the administrative staff (id = . 36 - . 27 = . 09). When this table is 72 Table 32. Means of the Index of Directiveness of Control for the Director, Chief of Medical Staff, and Assistant Administrators (Five Meetings) Positional Administrative Position Influentials Staff Actors . i t, o s p r q Director of the Hospital i -- . O9 . 37 . 36 . 42 . 31 . 31 Chief of Medical Staff t . 36 -- . 50 . 00 . 00 . 23 . 40 Assistant Director 0 .12 . 25 -- . 21 . 36 . 25 . 38 Assistant Director 3 .16 . 00 . 23 -- 31 . 37 13 Comptroller p . 37 . 00 . 32 . 36 -- . 42 . 00 Director of Nursing r . 28 . 81 . 22 . 27 . 25 -- . 36 Resident AdJninistrator q . 13 . 00* . 00 1. 00 . 00 . 15 - - . 00 value, but index contains interaction. * Table 33. Mean Index of Directiveness of Control Values of Positional Influentials and Administrative Staff Positional Administrative Influential 5 Staff (1. t) (0. S. p. r. q) Positional Influentials . 23 . 36 Administrative Staff . 27 . 33 )3 Means are computed on indexes With any interaction. 73 broken down by each participant some interesting results are found. Table 34 contains the mean differences for each of the participants in this unit . Table 34. _ Difference Between the Mean IDC Values of Each Participant in the Administrative Conference Positional Influentials ,Administrative Staff Individuals i t o s p r q Director i -- -.27 .25 .20 .05 .03 .18 Chief of Staff t - - -- . 25 . 00 . 00 - . 58 . 27 Assistant Director 0 -- -- -- -.02 .04 .03 .38 Assistant Director 3 -- -- -- -- -.05 .10 -.87 Comptroller p -- -- -- -- —- .17 . 00 Director of Nursing r -- -- -- -- -- -- . 21 Resident Adminis- trator q -- -- -- -- -- -- -- The director is controlled by the chief of staff (id = -. 27) but he con- trolled the rest of his administrative staff. The chief of staff controlled two of the five administrative assistants but is controlled by the director of nursing (id = -. 58). Assistant director "0" had a small margin of control over the comptroller (id = . O4) and the director of nursing (id = .03), lost by a small margin to assistant director "8", and expended a higher rate of control over the resident (id = . 38). The resident, in turn, was controlled by everyone except that he controlled the assistant director "s" (whom he finally replaced when he completed his residency). The director of nursing concentrated her control efforts on the physician (id = . 58), but is then directed by everyone else in the conference. The comptroller exerted very little control one way or another (his mean differences ranged from -. 05 to . 05) except with the director of nursing (id = . 17). 74 It may be concluded that on the basis of these five meetings, the positional influentials did exert more control over the administrative staff than the converse. . However, by computing the differences between the means of control for each individual, it was found that the director of nursing expended more of her control efforts over the chief of staff, and he in turn directed his control efforts over the director of the hospital. , Tables 35 and 36 present the data to test the activity aspect of the hypothesis and the profile of behavior of each participant in the administra- tive conference. Table 35 contains the profile of behavior for the two positional influentials (the director and chief of staff) and the administrative staff. The two classifications did not differ inarea A (positive reactions). Both the positional influentials and the administrative staff expended 21% of their activity in this area. The administrative staff expended slightly more interaction in attempting to answer questions (59% versus 51% for the influentials), but both classifications concentrated the same amount of activity in asking questions (10%). The influentials tended to exert more negative reactions than did the assistant administrators (18% versus 11%). According to this table, the influentials were the most active participants. Their mean number of acts per individual per meetings was 316 compared to 237 for the members of the administrative staff. This comparison is somewhat misleading. Table 36 contains a finer breakdown for each individual. The director is by far the most active participant 0? number of acts = 465), and his high activity rate substantially raised the mean for influentials. . The chief of staff had the lowest activity rate with the exception of the resident administrator who played a- passive "student role" (he was the recording secretary for the administrative and operation's committee meetings). Assistant administrator "0" was the next most active person (if = 337) followed by the director of nursing services (SE = 328). It would seem that the activity hypothesis of the chief of staff (as a positional influ- ential)“was not supported. 75 .mcoflomom o>flmwoz M Q "muoflmoa u U "whogmcafl woumgouu< u m “2330de ”we/330nm u <4. L! me hmm 3m wnfiooz mom acmnonm Hem 304 nofigz X 1mm .N. com .m mmo .N muo< nongz #308 2:. I: on: . acoO Menu 1308 c m w Emwnommuefi marosm .N v m o :oflmcofi 959.5 . .2 Q m . m w mooanmwQ .oH H H H Gofloeflfl .cowumemmdm no“ mxm< .m N N N Goflmfiflmkm 503me no“ mxm< .w 0 n N. N. cofimEnowcH .5333“..qu no“ 33.4 .b mm mm mm coflmgnoflfi .Gofimucofino moZD .o f f f c2339“ 6353 $30 _ . .m m w m n Gofloouwn— .soflmowwdm mo>m0 .v S 2 a 333. .m a w : mmmofiom cowmcofi m30£m .N. < N N m 328:8 28% 4.. 3 .u d .m .3 3 .3 95.30 3.95 mgmflcmdficw mofiuomofimu .mofimm «0.2% 130R. o>3dnumficw§< Hmaoflflmom * E Ammcflmoz 9,th madam o>flm5mwcflbp< pad mHmSQodGfi Hmcofifimonw 65 mo cofisflbmflm ommucoouom mo oflwounm .mm oEmB 76 «N mNm vvm omm nmm 9: m3. mcfimoz mom GOmumnH pom 39% Z X «so AH .ON ma a god ouv PH OH OH w H om om HM Ma 0 H NLDLOO‘ ow MM m CH ed M OOH mfi‘m om Md w Ha NH N Hod mm mm OH H o 2: wN ma «GOO hem H308 , Emflcommfidfl 950nm Gofimcoh mBOJm mooanmwQ ”“0300an .coflmomwdm no“ mxm< Goflmgmcrm 88330 MOM mxm< :oflmfiuofifi , 5333030 90m mxm< coflwEuoEH . .aoflmucofiuo .0630 dogma—Hmzm .853me mofiO Gofloouwfl ..:oflmowmdm woe/“O moohm< mmmofiom Gowmcoh maronm 32328 39% .NA .HH .oH U .H m m 0 u H nounmbmfiggp’w mammhdz popcouMQ uaogmom mo acuoouwfl .3393 nQEoO pogoouwfl acmumfimmmw youoofifl ”Edam am m < 33m mmfiuomoudu .moamm 3066qu Ho mo 30 Ammcfloo: 95.5 madam o>3m3mwcwé< ms» was .wwmum Hmoflooz mo mofiaU .HOHUoHMQ Hmfimmom 9.3 «o Gomunnfinuma ommucoohmnm mo mfimounm .om mannmh. 77 To digress momentarily from the analysis of the profiles, Table 37 presents an interesting relationship. Whenever the director of the hospital and one or more of his assistants were out of town, it was noted who was "in charge" of the hospital during their absence. On one "lucky" occasion (from the standpoint of this research design) three of the administrators were away at one time which allowed the following rank order comparison to be made. 7 Table 37 presents the heirarchy of authority for the hospital and the activity rates of the administrators. Table 37. The Rank Order of Responsibility Among the Administration and Their Activity Rate Order of Responsibility Mean Activity Rate for the Hospital per Meeting Director 465 Assistant Director "0" 337 Director of Nursing 328 Comptroller 244 Assistant Director "s"* 230 or Resident Administrator’i‘ ' 24 *The occasion never arose where the director had to choose between assistant director "8" and the resident administrator. Thus, the rank between these two is not differentiated. . The resident, however, did re- ~" place "8" after he completed his degree, and the former was released. Clearly, the heirarchy of authority among the administration was closely correlated with their activity rate.z zThere was also a perfect correlation between the activity rates of the top influentials on the board of trustees and the number of votes they received by the panel of "knowledgeables. " Several studies have shown that 78 How do members of the administrative conference spend their time? Since Table 36 is socomplex, Table 38 was prepared to compare the inter- action areas of each participant. Table 38 shows that the director spent 50% of his time in the area of attempting to answer questions (category 6 = 29%; category 5 = 15%; category 4 = 7%). Only 10% of his time was spent in seeking the answers to problems (category 7 = 7%; category 8 = 2%; and category 9 = 1%). The social-emotional areas were almost balanced. _ He expended 22% of his interaction in the positive emotional area (category 2 = 12%; category 3 = 7%; and category 1 = 3%), and 19% in the negative emotional area-(category 12 = 8%; category 11 = 7%; category 10 = 4%). . The chief of the medical staff expended a greater proportion of his time in attempting to answer questions (62%), and is exceeded in this area only by the director of nursing (67%). The comptroller had the next highest proportion in this area (59%) followed by the assistant director (58%). Since medical, nursing and financial problems were of most concern to this committee, the association between proportion of activity by area and the specific status occupant seems to have a direct relationship. It is also interesting to note that assistant director "0", who is "second in command" had a profile quite similar to the director's, 3 except the director exhibited a slightly higher proportion of his activity in higher status persons tend to have a higher participation rate in group dis- cussions. Caudill found that in the daily administrative conference of a psychiatric hospital, the senior staff participated more than the resident physicians who in turn participated more than the nurses and other personnel. . See- William Caudill, The Psychiatric Hospital as A Small Society, Cambridge, Mass.: Harvard University Press, 1958, pp. 243-251. 7 Strodtbeck,. James and Hawkins also found a- close association between occupational status and participation rate in mock jury deliberations. See F. L- Strodtbeck, R. .M. . James, and C. Hawkins, "Social Status in Jury Deliberations, " in E. ,Maccoby, T. Newcomb and E.. Hartley, (eds.) Readings in Social Psycholog, 3rd ed. ,. New York: Henry Holt and Company, 1958, pp. 379-388. 3They are also quite close personal friends. 79 0mg .2: mo moo cod Hod HOH ”.an Mona $3.08 2 S e .2 .2 3 2 A2 use 2 .2 320338 983.38% 9,34“.me .Q om a E e .2 w 2 a was w .e mecomeeeov mcoflmoa .O S, S 3 mm 3 Ne om 3 see m .w meiomeeeov meek/melee poumEoufiw .m S 2 om MN mm 2 MN A... was N .H meioweeeov mcofiumom o>wfim0nm .< v .H a m o u w youmsmwcfigwxw mawmndz .5305 uofioopwfl Roaming wwmpm 15632 pogooflfl GOSUMHBGH mo mou< “seamed mo HOHUouMQ nQEOO acmumwmm‘ea acmumfimmifl mo EEO museuomcoO o>3m3mficflhp< .HOH coflomumucH mo define >3 mommamoumm 33on mo enumegsm .wm 3an 80 antagonistic reactions to the proceedings of the committee. The resident administrator, again playing the student role, exhibited the highest pro- portion of his activity in asking questions (20%), and exhibiting "good cheer, " (31%) of his activity was in the positive emotional area. Table 39 compares the percentage distribution of interaction by area for Caudill's administrative conference in a psychiatric hospital with the distribution of interaction of the administrative conference of the community general hospital in the present study for several role groups. It should be pointed out that comparisons are between several actors for each role group in the psychiatric hospital committee and one physician (the chief of staff), one nurse (director of nursing), and one director (who has no counter- part in the psychiatric committee, since the head of that hospital was a physician). Several overuall differences are immediately apparent. Much more of the time of the administrative committee of the general hospital was spent in the social-emotional areas. The average percentage of time the role groups in the general hospital spent in positive reactions was 20% compared to 8% for the role groups in the psychiatric hospital. The role groups spent 13% of their time in negative reactions compared to 8% in the psychiatric hospital. The difference between positive and negative reactions was in favor of the general hospital committee (7% versus 0%). When role groups within the two committees are compared, the physician in the general hospital spent more of his time in positive reactions (17% versus 7%, respectively) and negative reactions (14% versus 8%, respectively), and less time in attempting to answer questions (62% versus 72%, respectively). The comparisons between the nurses' profiles are remarkably similar for each area of interaction. In fact, the two profiles did not differ by more than 3% for any one area. . Some interesting between-group comparisons are present, however. The nurses in the psychiatric hospital committee 81 0003800 3.30 00000000 000 002000 0000HU0H uw new 6323 .00 5-2 .262 $33. :40 mm .0630 .3 8o: 22830... as S: 2: S2 2: so 2: 2: so 200 .80 0.89 2 Z S 2 E w 2 .2 w 33880 033002 .0 S Z S a w w 2 2 a 28530 .0 em. mm om S we on A: E E. Seamus. 2080883 .0 om mm MN 2 2 w 2 2 e 300380 03030 .< 0.850.: :3 .0 EN .8 $05. z 2800; .m>< m000000m0 000.00.2Q 00002 0000 .m>< 00000 0000.92 .0300me 003000000H mo 00.3% £00000 .umm< ufim>£nw .8500 nwm>£nm 0320005 #000000 2320000000 0.2003003 02000200>mnm HI 000000 000m 2000>0m 000 1320003 0000000 50000000 0 mo 000008000 03000000000034 05 000 2030003 02000200>mnm 0 mo 000003000 0300300020004 >300 000 00% 0000< 000m .m020m >0. 003000005 mo 000300200020 0m0~00000n~ 000 mo 003009000 .om 030M. 82 spent proportionately less of their time in attempting to answer questions in comparison to other role partners in the committee, whereas the nurse in the general hospital committee spent proportionately more of her time in attempting to answer questions. This may be because the nurse in the general hospital administrative committee is responsible to the lay adminis— trator and took an active role in resolving hospital decisions. In this case, she was often supported by the administrators in taking issue with the chief of staff. In the psychiatric hospital, the nurses are responsible to the physicians and the chief of staff was also a physician. In such a milieu, nurses are perhaps more reluctant to provide answers to problems raised. There is no counter role partner in the psychiatric hospital to the lay director and the administrative staff of the general hospital, nor are there counterparts to the "other" roles of the psychiatric hospital in the administrative conference in the general hospital. .Nevertheless, it can be seen that the administrators devoted proportionately less time to the problem-solving or task areas (B and C) than the physicians and nurses in both types of hospitals, and proportionately more time in the social- emotional areas (A and D). Summary and Conclusion It was hypothesized that the director of the hospital and the chief of the medical staff, by virtue of occupying two seemingly important statuses in the organization, would be the most active participants in the decision— making process and would exert more control over the administrative staff than the latter, in turn, would exert over the two positional influentials. . The hypothesis was supported in that the positional influentials had a higher mean difference score (id = By examining the mean differences between each participant, it was found . 09) over the administrative staff. that the chief of staff directed his control efforts over the director; the director in turn directed his control over all the administrative staff, fi- 83 and the director of nursing concentrated her control efforts on the chief of the medical staff. The positional influentials also were more active participants in the decision-making process than the administrative staff (3(- number of acts = 316 versus 237, respectively). However, this difference was largely attributable to the extremely high rate of activity of the director (i = 465) rather than to the chief of staff whose mean number of acts was next to the lowest in the committee (166). A direct relationship between the rank of the members of the committee and their activity rate was observed. The director of nursing and the chief of staff spent p’roportionately more of their time attempting to answer questions (67% and 62%, respectively) compared to the director (50%) and the administrative assistants (51%, 58%, 59%, and 31%). The resident administrator, the assistant administrators, and the director of the hospital spent proportionately more of their time in the social-emotional areas than did the physician and nurse. When the administrative conference of a psychiatric hospital was com- pared with the administrative committee of the general hospital, it was found that the role groups in the general hospital spent more of their time in the social-emotional areas, and less of their time in the task or problem- solving areas. The physician in the general hospital spent less time in problem-solving areas than the physicians in the psychiatric hospital setting. The nurses in both settings had similar profiles but, in comparison with the physicians in the group, the nurse in the general hospital spent more of her time in the task area than her counter role partners in the psychiatric setting. This was interpreted in terms of differences in the composition of the role-partners in the committee (lay administrators in the general setting; physicians in the psychiatric setting). The administrators spent proportionately less time in the task areas and more time in the social- emotional areas than the physicians and nurses in both types of hospitals. CHAPTER VI SUMMARY AND CONCLUSIONS This study has been concerned with the consequences of systemic linkage of community influentials on the decision-making process of a community general hospital. The central question which guided the research design was "to what extent do community influentials actively participate in and control or influence the decision-making process of one community organization--a non-profit general hospital. The extent of overlapping representation of influentials in the major decision-making units (the board, two working committees, and the administrative committee) was examined. The design to assess the consequences of linkage in the organization was essentially twofold. First it provided a methodological procedure by which the particular process of decision-making could be observed at various levels of the organization-~Bales' interaction process. It provided a. cross- validation of two commonly used measurement systems for identifying community influentials and predicting their control—-the reputational and positional techniques, by using Bales' Index of Directiveness of Control based on observations of interaction rather than "observations" elicited from structured or unstructured questionnaires about who said. "who did what" on a specific decision. In order to provide a test of validation, a hypothesis concerning what the reputational and positional techniques are supposed to predict was formulated. . This hypothesis took (the operational form that community influentials are the most active participants in the decision-making process of the organization and exert more control over the non-influentials than the latter, in turn, exert over the influentials. 84 85 Since community influentials interact at different levels of decision-making in the organization the hypothesis was tested at each level of the decision- making heirarchy. In that manner the consequences of linkage within the organization were examined. . Chapter II presented a historical analysis of the participation pattern of economic influentials over a 50 year period (1910-1959). That chapter examined the "withdrawal" hypothesis of Schulze that economic influentials were withdrawing from local decision-making participation. It was found that economic influentials declined slightly in representation on the board of trustees, but the decline was not statistically significant. . For the past 30 years, a small but definite decline in influential representation in all board offices combined was observed but the trend was more gradual than for any organization found in Cibola by Schulze or in Wheelsburg by Clelland. This pattern suggests that hospitals may remain the last organi- zation of influential participation or that the rate of withdrawal is not uni- form for all community organizations and of course may not be complete in any given community. By examining other intra-organization statuses such as the office of president, committee Chairmanships and membership, and also attendance patterns, evidence was found to support the Form and. Miller hypothesis of increased influential participation at the strategic organizational level. Chapter III examined the central hypothesis of control in the board of trustees. A definite heirarchy of control in the board was found. . The economic influentials on the board exerted the most control over the non- influentials who were members of the hospital management staff, less con- trol over the non-influentials who were members of the board, and least control was exerted among themselves. The same pattern was found for the top influentials. The community influentials clearly dominated the proportion of activity at the board meeting, and accounted for the largest proportion of 86 interaction in the positive emotional area, the problem solving area, and in the area of asking questions. The non-influentials reacted to this by exhibiting a higher proportion of negative emotions. The central hypothesis was supported and the use of Bales' interaction process analysis validated the reputational and positional instruments for the board. Chapter IV examined the central hypothesis of the study in two working committees of the b0ard--the executive committee and the operations committee. . Contrary to the pattern found in the board, the community influentials did not exhibit a greater amount of control at the committee level. In the smaller committee and more informal setting where the community influential was not chairman, the non-influentials had a higher mean of con- trol over the community influentials and administrators. However, on the major "issue" resolved during this period of observation, they "lost out" to the decision of the community influential. It was apparent from observ- ation at each level of the organization that when issues were brought up, they were brought up at the lower levels where hostility, antagonisms, and pressures were openly brought to bear. As a consequence of systemic linkage, the non-influentials were "out-maneuvered” as the issue was moved up the heirarchy of the organization where the influentials had greater representation and power. At the board level, hostility and opposition were not brought out in the Open. The community influentials concentrated on providing the information as they did in the board meetings, and they clearly dominated the activity of the committee including engagement in conflict with the non-influentials, contrary to the pattern on the board. . The non-influentials expended less of their time in providing information and were high in the proportion of their time spent in positive and negative reactions. . In the face of such conflict, the administrators showed little activity. Chapter V examined the hypothesis that the director of the hospital and the chief of the medical staff, by virtue of occupying two seemingly 87 important statuses in the organization, would be the most active partici- pants in the decision-making process and would exert more control over the administrative assistants than the latter, in turn, would exert over the two positional influentials in the administrative conference. The hy- pothesis was supported in that the positional influentials had a higher mean difference score over the administrative staff. . By examining the mean differences between each status occupant it was found that the chief of staff directed his control efforts over the director, the director in turn directed his control efforts among all the administrative staff, and the director of nursing concentrated his control efforts on the chief of staff. The two positional influentials were also more active participants than the administrative staff, however, this difference was largely attribut- able to the extremely high rate of activity of the director rather than the chief of staff. A direct relationship between the rank of the administrators in the committee and their rate of activity was found. The director of nursing and the chief of staff spent proportionately more of their time attempting to answer questions compared to the director and the administrative assistants. The resident administrator, the assistant administrators, and the director of the hospital spent proportionately more of their time in the social-emotional areas than did the physican and nurse. . When the administrative conference of a psychiatric hospital was compared with the administrative committee of the general hospital it was found that the role groups in the latter spent more of their time in the social- emotional areas and less of their time in the task or problem-solving areas. The physician in the general hospital spent less time in the taskarea than his counterparts in the psychiatric conference. The nurses in both types of hospitals had very similar profiles but in comparison to the physicians in their groups the nurse in the general hospital spent more of her time in 88 the task area than her counterpart in the psychiatric conference. . This was interpreted as a consequence of the differing composition of the committee. _ The lay administrators in the general hospital often supported the nurse in opposition to the one physician on committee decisions, which may explain why she was more active in problem solving than the nurse situated in a conference run by physicians. The administrators spent proportionately less time in the task areas and more time in the socio-emotional areas than the physicians and nurses in both types of hospitals. . Finally, in the Appendix, a critical analysis of the reliability procedure recommended by Bales and his co-workers for testing inter-rater reliability of categorization and unitization was presented. . It was found that the use of chi-square as an index of goodness of fit was not an appropriate statistical procedure for estimating inter-rater reliability. An alternative statistical procedure was developed using the normal approximation of the binomial. Limitations of the Study and Future Research This study was primarily concerned with developing a methodological approach that would validate two measurement systems which identify com- munity influentials and which would examine their role in the decision- making process of the administrative system of a community general hospital, as has been pointed out. In this regard, the study provides a methodological contribution to the analysis of the role of any type of status occupant within any organization in the community. . However, there are both theoretical and methodological limitations to the study that should be briefly pointed out. . First, it must be observed that only 1 board meeting, 6 meetings of two working committees of the board, and 5 administrative committee meetings were analyzed in this study. Although 19, 070 units of interaction were observed (far more than most contrived small group laboratory experi- ments utilize) such observations really constitute a short period of time in 89 the dynamic process of decision-making in a complex organization. Under these limitations it would not be surprising if the findings did not stand the test of replication since possible conditions affecting such inter- action may change as the organizationcontinues to operate-.1 For example, it may be that on a board with less community influential representation, the non-influentials would exert more control than they would on a board with more influential representation (e. g. , a generalization based on our intra-organization finding). . It may be that community influentials sitting ona board of a hospital operated by a religious order may defer to fiscal policy set forth by that order. It may be that community influentials on a board of a hospital that is supported by federal, state or city funds may not control fiscal policy. Or if a community power structure is bifurcated and influentials are more concerned with regional and national interests, they may not control policy on boards dominated by active local leaders (again, particularly if funds were allocated from a city council, etc. ). These are just a few of the conditions that need further examination. It should also be noted that decision-making is far more complex than discussed in this study particularly when put in the social context of a complex organization (and what organization is not complex in terms of the capacity of our present measurements ?). Policy may be pre-determined before the observed meeting is ever called to order. The administration may have been consulted prior to the board meeting, but never asked during the meeting what their recom-‘v mendation might be for any given policy. Or conversely, the policy may 1Nevertheless, in all fairness to the study, the evidence does con- tribute to a growing body of empirical data which demonstrate that community influentials, as measured, are the most active participants in the community decision-making process and exert control over that process. It further pro- vides an example of how, within one community organization, the influentials are able to control policy by moving the issue up the heirarchy of the organi- zation where they have greater representation and power. 90 be determined by influentials in committee meeting but never implemented in the organization by administration. . Such problems are not controlled using the observational procedure of this study. A further limitation of this observational approach is that the norms of the particular role groups have not been mapped and thus the measurement of the extent of conformity or deviation from institutionalized control patterns was not possible. The effects of such variables must await further research before a very definite statement about the control of influentials in a community organization may be tendered with much confidence. A legitimate methodological criticism that might be raised is that the writer was aware of the identity of the community influentials, and might he not tend to categorize the behavior of the participants to favor the influ- entials as far as control behavior is concerned? Utilizing a very small sample of interaction (N = 327 acts), the writer's categorization was com- pared to one rater who was not aware of the identity of the community influentials. . In this case, the other rater categorized more behavior-in category 4, directs or controls, for the community influential than did the writer and similarly the mean index of control of the influential was higher by . 17, than the writer's. .In other words, for a very! small sample, it would appear that the writer was somewhat conservative in categorizing control behavior of the influentials, or at leas‘ was not exhibiting a great deal of bias in the same directionas the results reported. . BIBLIOGRAPHY Books Bales, Robert F. , InteractionProcess Analysis: A Method for the Study of Small Groups, Cambridge, Massachusetts: Addison-Wesley Press, Inc. , 1951. Caudill, William, The Psychiatric Hospital as a. Small Society, Cambridge, Massachusetts: Harvard University Press, 1958. . Dixon, Wilfrid J. , and Massey, Frank J. , Jr. ,, Introduction to Statistical Analysis, New York: McGraw-Hill B00k~Company,. Inc. , 1957. Form, William H. , and Miller, Delbert C. , Industry, Labor and Com- munity, New York: Harper and Bros. , 1960. Haggard, Ernest A. ,. Intraclass Correlation and the Analysis of Variance, New York: The Dryden Press, Inc. , 1958. ' Hunter, F. , Community Power Structure: A Study of Decision-Makers, Chapel Hill: The University of North Carolina Press, 1953. Hunter, F. ,. Schaffer, R. D. , and Sheps, C. G. , Community Organizations , Action and Inaction, Chapel Hill: The University of North Carolina Press, 1956. Loomis, Charles P. ,. SociaLSystems: Essays on Their~ Persistence and Change, D- Van Nostrand Company, Inc. , Princeton, 1960. Loomis, Charles P. , and Beegle,. J. Allan, Rural Sociology--The Strategy of Change, Englewood Cliffs, N. J.; Prentice-Hall, Inc. , 1957. Mills, C.. W. , The Power Elite, (4th Printing), New York: Oxford University Press, 1957. Merton, Robert K. ,. Social Theory and Social Structure, Revised and Enlarged Edition, Glencoe, The Free' Press, 1957. Siegel,.‘Sidney, Nonparametric Statistics for the Behavioral Sciences, New York, McGraw-Hill, 1956. 91 92 Articles Agger, R.. E. ,. "Power Attributions in the Local Community; Theoretical and Research Considerations, " Social Forces, V01. 34, (1956). Agger, R. E. , and Goldrich, D. , "Community Power Structures and Partisanship,, " American. Sociological Review, Vol. 23, (1958). Agger, R.- E. , and Ostrom, V. ,. "The Political Structure of a- Small Com- munity, " Public Opinion Quarterly, Vol. 20, (1956). Anderson, H. H. , "Domination and Social Integration in the Behavior of Kindergarten Children and Teachers, " Genetic Psychology Monograjh, Vol. 21,.(1939). Bales, Robert F. ,. "A Set of Categories for the Analysis of Small Group Interaction, " American Sociological Review, Vol. 15, (1950). Bales, R.. F. , and Strodtbeck, F. L. , "Phases in Group Problem Solving, " The Journal of Abnormal and Social Psychology, Vol. 46, (1951). Barth, E. A- F. , and Abu-Laban, B. ,. "Power Structure and the Negro Sub-Community, " American Sociological Review, Vol. 24, (1959). Borgatta,. E. F. , and Bales, R.. F. , "The Consistency of Subject Behavior and the Reliability of Scoring in Interaction Process Analysis, " American Sociological Review, Vol. 18, (1953). Carter, L., Haythorn, W., Meirowitz, Beatrice, and Lanzetta, J. R., "A Note on a New Technique of Interaction Recording, " Journal-of'Ab- normal and'Soc‘ialPsycholow Vol. 47, (1951). Carter, L. ,. Haythorn, L. ,. Meirowitz, Beatrice, and Lanzetta, J. , "The Relation of Categorizations and Ratings in the Observation of Group Behavior, " Human Relations, Vol. 4, (1951). Chapple, E. D. , ”The InteractionChronograph: Its Evolution and Present Application, " Personnel, V01. 24,. (1949). Cochran, W. G. ,7 "The Chi-Square Test of Goodness of Fit, "-Annals of Mathematical Statistics, V01. 23, (1952). Cochran, W. G. , "Some Methods for Strengthening the Common Chi- Square Tests, " Biometrics, Vol. 10,. (1954). 93 Dahl, R. A. ,. "Equality and Power in. American Society, " (Power and Democracy in America, V. W.. V. D'Antonio and H.. J. Ehrlich, (eds.), Notre Dame: University of NotreDame Press, (1961). Dubin, Robert, ”Stability of Human Organizations, " Modern Organization Theory, Mason Haire (ed.).- New York; John‘Wiley and Sons, Inc. , (1959). Form, W. H. , and Sauer, W.. L. ,. "Community Influentials ina Middle- Sized City, " General Bulletin No. 5 of the Institute for Communiiy Development, Michigan State University, (1960). Foskett, J.. M. , and Hohle,. R. , ”The Measurement of Influence in Com- munity Affair s, " Research Studies of the State College of Washington, Vol. 25, (1957). Freedman, M.. B., Leary, T. F., Ossorio, A. G., and Coffey, H. S., "The Interpersonal Dimension of Personality, " Journal of Personality, Vol. 20, (1951). Guetzkow, Harold, "Unitizing and Categorizing Problems in Coding Quali— tative Data, " Journal of Clinical Psychology, Vol. 7, (1950). Hanson, R.. C. ,. "Predicting a Community Decision: A Test of the Miller- ‘Form Theory, " American Sociological Review, Vol. 24,, (1959). Heinicke, C. , and Bales, R.. F. ,. "Developmental Trends in the Structure of Small Groups, " Sociometij, Vol. 16, (1953). Heyns, R. W. , and Lippitt, R. , "Systematic Observational Techniques, " in Gardner Lindzey (ed.) Handbook of Social Psychology, V01. 1, Cambridge, Massachusetts: Addison-Wesley Publishing Co. ,, (1954). Jack, Lois N. ,. "An Experimental Study of Ascendent Behavior in: Pre- school Children, " Universitl of Iowa Studies in Child Welfare, Vol. 9, (1934). Klapp, 0.. E. , and Padgett, L. V. , "Power Structure and Decision- ' Making in a Mexican Border City, " The American Journal of . Sociology, V01. 65, (1960). Loomis, C. P. , "Tentative Types of Directed Social Change Involving Systemic Linkage, " Rural Sociology, Vol. 24, (1959). 94 Loomis, C. P. ,. "Toward a Theory of Systemic Change, " Rural Sociolo in a Changing Society: * Proceedings of atNorth Central Rural Sociology Committee (NCR-S) Seminar, Columbus, Ohio: The Ohio Agricultural Extension Service, November (1959). Miller, D. C. ,. "Decision-Making Cliques in Community Power'Structures: A Comparative Study of an American and an English City, " The American Journal of Sociology, V01. 64,. (1958). Miller,. D.. C. ,. "Industry and Community Power Structure: A Comparative Study of an American and an English City, " American Sociological Review, Vol. 23, (1958). Miller, D. C. ,. ”The Prediction of Issue Outcome in Community Decision- Making, " Research Studies of the State College of Washington, Vol. 25,, (1957). Polsby,. H. W. ,. "Three Problems in the Analysis of Community Power, " American Sociological Review, Vol. 24,, (1958). Redekop, C. , and. Loomis, C. P. ,. "The Development of Status-Roles in the Systemic Linkage Process, " Journal of Human Relations, Vol. 8, (1960). Schulze, Robert O. , "The Role of Economic Dominants in Community , Power Structure, " American Sociolggical Review, Vol. 23, (1958). Schulze, R. O. , and Blumberg, L. U. ,. "The Determination of Local Power Elites, " The American Journal of Sociology, V01. 63, (1957). Snyder, W. U. , "An Investigation of the Nature of Non-Directive Therapy, " Journal of Genetic Psycholojy, Vol. 33, (1945). Steinzor, B. , "The Development and Evaluation of a Measure of Social 1 Interaction, " Human Relations, Vol. 2, (1949). Strodtbeck, F- L. , James, R.. J. , and Hawkins, C. ,. "Social Status in Jury Deliberations, " in E. Maccoby, T.. Newcomb, and E. Hartley , (eds.), Readings in Social Psycholfiy, 3rd ed. ,. New York: Henry Holt and. Company, (1958). Wolfinger, R. E. ,. "Reputation and Reality in the Study of Community , Power, " American Socioloflal Review, Vol. 25, (1960). 95 Unpublished Manus c ripts Clelland, Donald P. , "The Role of Economic Dominants in the Power Structure of a Midwestern Community, " unpublished Master's Thesis, Michigan State University, 1961. Erickson, E. C. ,. "The Reputational Technique in a'Cross-Community . Perspective: Selected Problems of Theory and Measurement, " unpublished Ph. D. dissertation, Michigan State University, 1961. Form, William H. , and Sauer, Warren L. , "Business and Labor Images of Community Power Structure: A Comparative Analysis, " paper read before Fifty- fifth Annual Meeting of the American Sociological Association, August,1960. Hanson, Robert C. , "The Systemic Linkage Hypothesis and Role Con- sensus Patterns in Hospital-Community Relations, ” unpublished Manuscript, Department of Sociology, University of C010rado. Harkness, James P. ,. "Hospital Organization in Transition: A Sociological Analysis of Interlocking Social Systems, " unpublished Ph. D. dis- sertation, Michigan State University, 1961. Muntean, Alexander J. , "Community Change and Hospital Development: A Case Study of Community Power Structure, " unpublished Master‘ 3 Thesis, Michigan State University, 1959. . Rossi, P- H. , "A Theory of Community Power, " paper presented to the 1960 Annual Meetings of the American Sociological Association. Sauer, Warren Louis, "Labor-Business Images of Community Power: . Convergences and Divergences, " unpublished Ph. D. dissertation, MichiganState University, 1960. APPENDIX INTER-RATER RELIABILITY OF CATEGORIZATION: A METHODOLOGICAL CRITIQUE OF THE PROFILE .METHOD, A PROPOSED ALTERNATIVE, AND THE LEVEL OF RELIABILITY OF THE STUDY Systematic observation of the interaction between members of groups has become commonplace in sociology. A number of methods have been devised, l and the scientific importance of these developments cannot be overstated. However, in an excellent summary of these methods Heyns and Lippitt have suggested that to be scientifically useful, the taxonomy must be clear, communicable, and subject to confirmation by other scientists} The researcher who applies a given set of cate- gories to an empirical situation must be able to do so with some evidence that his results will be comparable not only to those of other researchers, but to work he himself has done earlier. This is reliability which may be defined in this case as the consistent agreement between independent observers or observations over a specified period of time. Heyns and Lippitt maintain ”that there is only one score whose reliability need be assessed, and that is the score which is actually to be used in the analysis. "3 A similar view, apparently, has been held by a 1See R. W. Heyns and R.. Lippitt, "Systematic Observational Techniques, " in Gardner Lindzey (ed.) Handbook of Social Psycholog, Vol. 1,. Cambridge, Massachusetts: Addison-Wesley Publishing Co. , 1954, pp- 370-404, for a review of a number of these techniques. zIhic1., p. 397. 31bid. 96 97 number of persons who have devised classification systems for cate- gorizing interaction. 4' Any classification of interaction in a group from a set of categories of behavior can be viewed in at least three ways. .First, one can analyze the profile of observation of any given individual in the gr oup--that is, the frequency with which a given individual is observed to "act" or "behave" in each of the elements of the classification scheme. . It is possible to imagine specified situations in which agreement between raters might be high in this kind of observational scheme due to consistent (i. e. , patterned) behavior which any given individual in the group might follow. This might be especially true of a chairman following an agenda or "rules of procedure. " Second, one can analyze periods of interaction--ca11ed "phase analysis" by Bales.5 Third, one can analyze the total profile of interaction, that is, analyze the number of acts that occur in specified categories for all members of a group. The writer agrees with Heyns and Lippitt that the degree of inference required by an observer in classifying acts, the observational load which he confronts over a period of time, and the definition of the category units 4H. H. Anderson, "Domination and Social Integrationin the Behavior of Kindergarten Children and Teachers, " Genetic Psychology Monograph, 21, (1939) pp. 285-385; L. Carter, L.. Haythorn,. Beatrice Meirowitz, and J. Lanzetta,. "The Relation of Categorizations and Ratings in the Observa- tion of Group Behavior, " Human. Relations, 4,. (1951) pp. 239-254; L. Carter, W. Haythorn, Beatrice Meirowitz, and J. R.. Lanzetta, "A Note on a‘New Technique of Interaction Recording,"‘J.oua'na1‘oJfAbnormalhaiid Social Psy‘choloa, 47, (1951) pp. 258-260; E. D. Chapple,. "The Interaction. Chronograph: Its ‘Evolution and Present Application, " Personnel, 24,. (1949) PP. 295-307; M. B. Freedman, T.- F. Leary, A. G. Ossorio, and H.. S..Coffey, "The Interpersonal Dimension of Personality, " Journal of Personality, 20, (1951) pp. 143-161; Lois N. Jack, "An Experimental Study of Ascendent Behavior in Preschool Children, " University of Iowa Studies in Child Welfare, 9, (1934) No- 3; W.. U.. Snyder, "An Investigation of the Nature of Non-Directive Therapy, " Journal of Genetic Psychology, 33, (1945) pp. 193-223; and B. Steinzor, "The Development and Evaluation of a Measure of Social Inter- action, " Human Relations, 2, (1949) pp. 103-122. - 5R- F. Bales and F. L. Strodtbeck,. "Phases in Group Problem- ‘Solving, " The Journal of Abnormal and Social Psychology, Vol. 46, No. 4, (Oct., 1951) pp. 485-495. 98 are all important in determining the eventual level of reliability}. It may be that the various classification: systems cited above'may be subject, differentially, to criticisms as a result of each. of these points. One of the most ingenious classification schemes used in observ- ation of face-to-face interaction of the group has been that developed by Bales.7 The observer is expected to assign each unit act of behavior to one of twelve interaction categories. The categories andtheir corres- ponding numbers are: (1) showing solidarity, (2) showing tension release, (3) agreement, (4) giving suggestions, (5) giving opinions, (6) giving orientation, (7) asking for orientation, (8) asking for opinion, (9) asking for suggestions, (10) disagreement, (11) showing tensions, and (12) showing antagonism. 8 There are three problems of variation in inter-rater reliability which Bales delimits. . The first of these is the attributing problem. This involves the attribution or designation of the originator and .target for the particular unit act. The second is that of unitizing, or the division of the period of interaction into "subject-predicate units" to be scored. And the final problem involves categorization, or the assignment of acts to one of the twelve categories. There is no doubt that each of these three sources of variation might cause considerable difficulty to a researcher. Though these problems have been specifically cited by Bales, they are certainly not limited to his observational system. Heyns and» Lippitt, 6Heyns and Lippitt, gp: c_i_t., p. 397. .7R- F. Bales, Interaction: Process Analysis, 22. gi_t. 8A unit act is defined by Balesas ". . . the smallest discriminable segment of verbal or nonverbal behavior to which the observer, using the present set of categories after appropriate training, can assign a classifi- cation under conditions of serial scoring. . . [and] the single item of thought or . . . behavior. " Ibid., p. 37. .Fora discussion of the 12 cate- gories see I__b_i__d. ,. ChapterwII and the Appendix. Also, R. F.. Bales,"A Set of Categories for the Analysis of Small Group Interaction, " American Sociological Review, 15 (April, 1950) pp. 257-63. 99 for example, in reporting on the "Carter Observational Procedure, " describe the test of reliability of observation identical to that used by Bales.9 The most common form of checking the reliability of the observer's judgments is to compare the profile which each observer develops while using a given classification system of behavior. . It is this comparison of observers' results for which Heyns and Lippitt have noted substantially high reliability between observers. . It also characterizes the findings on reliability and categorization by Bales and his co-workers.10 A Critique of the Profile Method The problem with which this Appendix will deal is that of critically evaluating the use of two statistical procedures used by Bales and his co-workers as estimates of reliability of between- rater cate orization of interaction. In addition, an alternative form of estimating reliability will be proposed. The Bales and Strodtbeck procedure for estimating reliability is based on what we have called the total profile for the group. The problem of assessing the variation in inter-rater categorization is an important one and is one which may only be solved after the reliability of unitization 11 has been established first. This Appendix is limited to the estimate of 9Heyns and Lippitt, gp. gfiu p. 384. 10See E.- F. Borgatta and R.. F. Bales, "The Consistency of Subject Behavior and the Reliability of Scoring in. Interaction Process Analysis, " American Sociological Review, 18 (October, 1953) pp. 566-569, and 1C. Heinicke and R.. F. Bales, “Developmental Trends in the Structure of Small Groups, " Sociometry, 16. (February, 1953) pp. 7-38. 11For an excellent discussion of this problem see Harold Guetzkow, ”Unitizing and Categorizing Problems in Coding Qualitative Data, " Journal of Clinical Psychology, 7 (1950) pp. 47-58. 100 reliability of two or more observers or raters with respect to the assign- ment of an. act of one of the twelve categories of interaction. . It is not concerned with self-reliability,” "the consistency of the observed phenomenon, "13 and other such related problems.“I The two commonly recommended statistical procedures for estimat- ing inter-rater reliability of categorization are chi-square as an. index of goodness of fit, 15 and the Pearson product moment correlation coefficient. ‘6 Heyns and Lippitt found the latter to be the more common procedure. 17 In Interaction Process Analysis, Bales and Strodtbeck maintain that the chi-square test is the more appropriate test of the two since (a) it is more easily extended to analysis of more than, two raters; (b) "r tends to be relatively insensitive to variations in values with small densities"and to be‘ "predominantly determined by the large values of the distribution, " whereas chi-square is "more sensitive to the variations in the pairs of values of smaller magnitude"; and (c) "r is insensitive to the number of acts within categories so long as the proportion of acts within categories to the total acts is constant, " whereas chi-square ”permits a concomitant test of both categorization and unitizing. " 18 There are other limitations to the use of the Pearson "r. " One of these is that there is no random sampling procedure involved in categorizing ”Borgatta and Bales, _o_p. git” p. 567. 13Ibid. ‘ 14'Bales, "Phases in Group Problem-Solving," pp. gi_t. 15Bales, Interaction Process Analysis, 22" c_i_t. ,. Chapter IV. 16Heinieke and Bales, pp. c_it. 17Heyns and Lippitt, 22. g_i_t_,, p. 396. 18Bales, gp: c_it_,, pp. 102-03. 101 interaction. Acts which occur rapidly tend to be underrepresented. Similarly, errors in categorization presumably are the result of different, but hardly random "mental sets" of the raters. A short summary of the use of chi-square for the test of inter- rater reliability follows. Table 1 is a paradigm for tabulating the fre- quency distribution of categorized unit acts for the twelve Bales categories. The two-rater case will be employed to illustrate the procedure. The fre- quencies in the diagonal cells are the sum of those unit acts on which both raters were in perfect agreement. Thus, a unit scored in category 1 ("shows solidarity") by rater A and rater B would fall in cell AIBI. If rater A scored an act as category 4 ("gives suggestion") while rater B scored it as category 12 ("shows antagonism") then the act would fall in cell A4B12 and so forth. Table 2 illustrates the way in which Bales suggests tabulation he set up for chi-square analysis of rater reliability. . Column I is the list of the 12 interaction categories. Column II is the frequency of ratings by a category for rater A and Column III is the frequency of ratings by category for rater B. Column IV is the mean of the frequencies in Columns 11 and III--the average number of observations by rater A and rater B for each unit act in each category. The mean is taken as a theoretical value of the row (i. e. , category). Column V is the square of the observed frequency for raters (from ColumnsII and III) minus the mean of A and B's frequencies (Column IV) for each row (or category). The summation of the values obtained in Column V yields chi-square.. Data in Column II for rater A and Column III for rater B are the column totals for rater A ( ) and row totals for rater B (n ) in Table l. nAiBl BlAl . Bales has established an additional rule that if in any row the fre- quency of scores for either rater B or rater A is less than five that row will be collapsed with all other like rows. Recent findings with the regard 102 Table 1. Paradigm of Tabulation of Acts by Categories for Two Raters y Rater A Bales Total Category 1 2 3 4 ..... 12 nAiBi 12 1 A1 B1 ...... AlzBl EA 2B1 2.. A25, 1 3 Rater 4 ' B 12 12 AIBIZ A4B12 AIZBIZ 2A 2B 1 12 Total n 12 12 12 12 Table 2. . Frequency of Score by Rater by Category 103 m. ; I . . .. R +R aura-3?)" Categories Rater Rater Mean= A2 B .3?— f f (I) (II) (III) (IV) (V) 2 3 4 5 6 7 8 9 10 11 104 to characteristics of chi- square indicate that this may be overly stringent. ‘9 Thus, all such collapsed rows (categories) will be treated. as one row (categories). Degrees of freedom in this case are determined by r (c - l) where r is the number of rows (categories) and c is the number of columns (raters). . Thelevel of significance which Bales sets at . 50 is an arbitrary level. 2° (It would seem that one should set at level around . 90 or higher to yield a good estimate of agreement.) Three hypothetical examples of inter-rater categorization will be presented. . In each of the following three tables the categories have been reduced from twelve to four for illustrative purposes. In each example, the chi-square value is zero and would not allow rejection of the hypothesis that there is no difference between the observers' judgments with regard to placement of unit acts in specific categories. Hence, the researcher would accept the test result at an estimate of reliability, according to the procedure described by Bales. ' . Table 3 contains the first model which, on inspection, shows a considerable amount of inter-rater agreement on unit acts for each cate- gory. . Raters A and B agreed that ten acts should be placed in category 1. Table 3. AHypothetical Model of Agreement'of Inter-Rater Judgments of Unit Acts and Chi-Square: Test of Inter-Rater Reliability Chi-Square Test Bales' Rater A - Bales' Rater X Category 1 2 3 4 Total Category . A B (4233-39-3.- 1 1o 1 1 o 12 1 12 12 12 .o ‘2 2 1 1o 1 1 13 2 13 13 1.3. .0 +3 3 1 1 10 1 13 3 13 13 13 .o m 4 o 1 1 1o 12 4 12 12. 12 .0 Total 12 13 13 12 50 Chi-Square = . 0 Note: See the following page for footnotes 19 and 20. 105 However, rater A placed one set in category 1 which rater B placed in category 2, and so forth throughout the table. The really important point to notice is that the column and row totals are equal for each cate- gory for‘ both raters and hence, the chi-square value is exactly zero. The null hypothesis would not be rejected and the researcher would assume he had reliability. The second hypothetical model is illustrated in Table 4. Table 4. A Hypothetical Model of Uniform Disagreement of Inter-Rater Judgments of Unit Acts and Chi-square Test of Inter-Rater Reliability Chi-Square Test Bales' Rater A Bales' Rater 32 (A - i)? Category 1 2 3 4 Total Category A B i l 5 . 5 5 5 20 1 20 20 20 O ”2 2 5 5 5 5 20 2 20 20 20 o 0 :3 3 5 5 5 5 20 3 20 20 20 .0 m 4 5 5 5 5 20 4 20 20 20 0 Total 20 20 20 20 80 Chi-Square = .0 I In this case raters A and B only disagreed as frequently on eachclassifi- cation for each unit act as they agreed. As can be seen by the tabulation, the column and row totals are again equal for each category. . The chi- square is again exactly zero. . 19W. G. Cochran, "The Chi-Square Test of Goodness of Fit, " Annals 9£Mathematical Statistics, 23, (1952) pp. 315-345, and "Some Methods for Strengthening the Common Chi-Square Tests, " Biometrics, 10 (1954) pp. 417-451. z”Bales, c_>p_. c_i_t_., p. 103 and Table 2, p. 110. 106 In Table 5 the third model isillustrated by a. case which is even more extreme. . Raters A and B disagreed completely in the classification. of unit acts in the categories. . The cells showing agreement, namely the right diagonal, are vacuous. Yet, row and column totals are equal and yield a chi-square value of zero. . With only the row and column infor- mation, the researcher again assumes he has reliability between his raters. Table 5. A Hypothetical Model of Complete Disagreement of Inter-Rater Judgments of Unit Acts and Chi-Square Test of Inter-Rater Reliability Chi-Square Test Bales' Rater A Bales' Rater K (A - 702 Category 1 2 3 4 Total Category A B X 1 0 3 2 10 15 1 15 15 15 .0 CO 2 3 o 5 2 1o 2 10 1o 10 .0 $4 .3. 3 2 5 o 3 10 3 1o 1 1o 10 .o ‘9‘ 4 1o 2 3 o 15 4 15 15 15 .0 Total 15 1o 10 15 50 Chi-Square = .o It is apparent by the models shown in Table 3, 4, and 5, chi-square is completely insensitive to agreement (or disagreement) between raters on particular unit acts. Rather, it is sensitive to differences in the absolute number of acts scored in a single category by the two raters. Thus, the test results are the same for distributions of disagreement which produce equality of row and column totals as they are for complete agreement. Yet, the judgment of raters over each specific unit act is of interest in a test of agreement. Since this analysis has shown the serious limitation with regard to chi-square tests, (or indeed any test which relies on data 107 from the total profile) it would seem wise to discard this statistical procedure. The chi-square test described above is dependent upon the tenuous assumption that the pool of unit acts chosen by rater A to be classified in a particular category will be the same unit acts that rater B placed in that category. It has been shown that if the assumption is true (as is nearly the case in Table 3), the chi- square value obtained will reflect inter-rater agreement. .However, if the assumption is not true (as in Table 5), its chi-square value will be insensitive to inter-rater disagree- ment. In other words, the Bales-Strodtbeck procedure for testing reliability is a measure of the existence of agreement between observers that certain observed proportions of the total number of acts for a given time period falls into each of the twelve categories--a profile analysis. The procedure is insensitive to whether or not they agree that each unit act they are observing at a specific time falls into the same category. In an article by Heinicke and Bales, apparently the use of the Pearson product moment correlation follows the same estimation pro- ” Here too the correlation cedure and makes the same tenuous assumption. coefficient represent the degree of agreement between the raters, where the datum is the proportion of total acts which fall into the twelve cate- gories. Again, it does not assess the degree of agreement betweenrtwo observers of:a single” apt at'a single point in time. As if this were not enough, there is an additional source of contami- nation. . We have observed that raters soon become familiar with the over- all distribution of acts by categories (or what Bales calls profiles). (This refers to the proportion of acts in the row or column totals for each category in Table 1.) Thus, the raters may tend to have substantial agreement in the profiles without regard to specific unit acts. A statistic which only tells the similarity of the profiles is not con- fronting the central issue, viz. , the judgments by different raters of a 21Heinicke and Bales, <31. c_i_t: 108 single unit act. In effect as Model 5 illustrates, two raters may have very low rates of agreement for each category, but may categorize the total number of acts observed in a similar profile. .What Bales calls unitizing is not even remotely tested by means of chi-square, and categorization--insofar as it refers to the classification of the particular acts by two or more observers-~is_not accurately estimated for reliability. The reader should not interpret these comments to mean that Bales schema has little utility, for one can hardly come to that conclusion when using it for the analysis of group interaction. . Similarly, it is not implied that it is impossible to get reliability with Bales or any other procedure, although the problems involved in non-laboratory settings are far more legion than one might expect. . Thus, part of the purpose of this analysis has been to attempt to draw attention to an uncritical application of the reliability procedure recommended by researchers in a relatively new and stimulating area of social science research. In short, the writer is in particular disagreement with Bales and Strodtbeck for using the specific tests of reliability which they suggest. In addition, the findings of this report suggest Heyns and Lippett's justifi- cation of the comparison of profiles of behavioral acts to be invalid as a general thesis. At the outset, we quoted Heyns and Lippitt as saying, "that there is only one score whose reliability need be assessed, and that is the score which is actually to be used in the analysis. " The state- ment is, of course, valid. However, it has been shown that its authors have also ignored its implications. There are several statistical procedures available to help the researcher with his problem of estimating the differences in observer's classification of behavior. One alternative which is simple to compute, which is sensitive to all of the distribution problems discussed above, and which allows a test of significance based on a known sampling distribution, is the binomial. 109 A Proposed Alternative Test of Inter-rater Reliability For purposes of discussion this procedure will be called the "proportion technique. " First the observation of the two observers are tallied in a distribution sheet given as in Table l . The population para- meter, P, is fixed at whatever level meets an acceptable criterion of agreement for the researcher. . For example, if P is set equal to .9 the result tests whether or not 90 per cent of the judgments were in agreement. The probability associ- ated with finding a value equal to or greater than the amount of agreement on the basis of the sample value is examined. . The binomial is laborious to compute when the sample size (the number of unit acts) is larger than 25. _ In. addition, tables are not generally available for larger samples. . However, it can be shown that as N increases, the binomial distribution tends to be approximated by the normal distributions and therefore the _z_ statistic can be used as an approximation. The‘N for an hour' 5 observ- ation of a small group is usually quite large (around 2500 unit acts) and the _z_ statistic is a good approximation to the binomial in such cases. This is true even-if P is near 1 or 0, a fact which is generally a limiting assumption of the use of the binomial distribution. A rule-of—thumb indicates that if P is near 1 or 0,. NPQ (where Q = 1 - P) must be at least 9 for this statistic to have a normal distribution. Thus, if P = . 9, the total N must be at least 100 where N = 1392) or' N = Té-sz—l) , therefore, . N = 100. This condition is quite easy to meet. . It appears that the binomial is sensitive to the various models examined here and yields a correct estimate of reliability within specified- probability values. A brief example of the use of the "proportion technique" might be helpful at this point. The formula is: = (xi .5) -~ NP ~l NPQ 110 Let P = . 7, the proportion of agreement which w0uld be accepted as the degree of reliability of categorization for two raters (i.e. , 70 per cent). Let Q = . 3, the proportion of disagreements between two raters (i.e. , 30 per cent). Let X = 40, the observed number of agreements between two raters, and Let N = 50, the total- number of unit acts observed. Correcting for continuity, since we are using a distribution for a continuous variable to approximate a distribution for a discreet variable, a . 5 must be added to X.“ Sufficient data is now available to compute the value of_z_ and make a decision as to an estimate of reliability. .The null and alternative hypotheses are: Ho : P > . 7 and 1H1 : P < .7 With alpha at say .05, the null hypothesis is rejected if the prob- ability associated with the value is less than alpha. . For the data presented in Table 3, E = 1. 39. The probability associated with that value (taken from the tables for the normal distribution)23 is equal to . 082. . Since . 082 is greater than . 05 (alpha) the null hypotheses is not rejected and it is reasonable to assume that on the basis of this test we have an adequate estimate of reliability. . Several other procedures for estimating inter- rater reliability have been reviewed and rejected, but perhaps deserve mentioning. . In his book» Intra-Class Correlation, Haggard presents a “The . 5 is added since X is greater'than NP. A . 5 would be sub- tracted from X if X were less than NP. 23See for example Table A, p. 247 in Sidney Siegel,. Nonparametric Statistics for the Behavior Sciences, gp; cit. 111 number of intriguing possibilities.“ However, since it is questionable whether the assumptions underlying the analysis of variance are met with a dichotomous measurement ("agree" or "disagree"), rendering interpretation of the F's and R's highly tenable, and since random errors must be uncorrelated, the intraclass technique was rejected.“ Variations due to inferences drawn by observers about the behavior they are cate- gorizing and the appropriateness of the categories themselves are con- sistent rather than random. Another paper of substantial importance is that of Harold Guetzkow.Z6 He proposes a technique for estimating inter-rater reliability, but puts an exceedingly strict limitation to his formulation. He maintains that "the proposition of units upon which two coders agree may be conceived as the sum of those items which both coders correctly classify and those items which both coders incorrectly classify in the same incorrect way. "27 He thus assumes that the correct classification is identified and known, whereas the proportion technique merely measures the agreement between two raters. He also assumes that the probability of correct classification should be the same for all units. This is another exceedingly stringent assumption which probably cannot be met in the use of Bales'categories. The traditional marginal method of computing the chi-square (using the distributions in the margins to compute the theoretical fre- quencies) was reviewed and found to be insensitive to the distributions of the type exemplified in Table 4. “We are indebted to Bernard Lazerwitz for calling our attention to the possible use of intra-class correlation as an estimate of inter-rater reliability. See Ernest A. Haggard, Intraclass Correlation and the Analysis of Variance, N. Y.: The Dryden Press, Inc. , 1958. ”Ibid., p. 46 and p. 91. 26Guetzkow, gp_. cg. 27Ibid., p. 51. 112 Summary of C ritique A brief critical review of the commonly accepted procedure of testing inter-rater reliability by comparing the profiles of observed behavior categorized by raters was reviewed and found to be wanting. Specifically, the use of chi—square and the Pearson product moment correlation coefficient for tests involving the profile method, as used, do measure the existence of agreement between observers that certain proportions of the 1:29;]: number of acts of a groupfall into each category. However, they are insensitive to whether or not the raters agree that each unit act which they observe falls into the same category. The 5 statistic, the normal approximation of the binomial test-- corrected for continuity, is proposed as an acceptable alternative which is simple to compute, is insensitive to various types of theoretically possible distributions of disagreements, and provides a- statement of the probability associated with the level of estimation of reliability between two raters . The Level of Reliability of The Study It is important to point out that the writer's level of competence in using Bales' system is the sole consequence of study of Bales' works in the area of interaction process analysis. He was not trained by Bales nor by any of B les' co-workers. After a thzee month period of non-participant observation in the hospital, electrical recordings were made of the various groups in the hospital. 7‘8 At these meetings the writer simultaneously attributized the interaction ("who interacted with whom") on a stenographic pad, ”Since the boards and committees met at different intervals dur- ing the year, the length of time the writer was in the hospital before recording the meeting of any given group varied within this time period. 113 according to pre-coded identification of the participants. . In order totmatch the protocol with the recordings, the first two or three words spoken by each respondent were entered on the pad after each attributization. Transcriptions were then typed by matching the protocol with the record- ings. Two other raters were then asked to read Bales' Interaction Process Analysis. After a two week trial period 6 tests were made.29 Table 6 contains the percentage of agreement in categorization between the writer and raters B and C, and the percentage agreement in unitization between the writer and rater D. The writer and raters B and C unitized the tran- scriptions and listened to the recording before independently categorizing the transcription. The writer also checked his internal reliability by re- categorizing a transcription. after a six months interval. . Inspection of Table 6 shows that the writer agreed with his previous categorization 80% of the time. However, the extent of agreement between the writer and rater B was only 53%, and with rater C 52% on categorization. (The total for the writer and rater B represents the sum of three separate tests.) Raters B and- C were able to agree on 64% of the units categorized. The fact that the two raters were able to agree more frequently between themselves than with the writer was attributed to the writer's experience in the group meetings. The latter was able to bring to the transcription knowledge of the situation and personalities of the participants that was not shared by the two raters. . For example, humor was very often missed by the two raters who scored the interaction "as they saw it. " The writer, knowing what went onrin previous meetings was able to recognize the "subtleness of the straight line. " Tension is very difficult to identify from a transcription even though listening to the recording helps a great deal. As a general limitation, it is suggested that tension is particularly under- represented in these categorizations. The above problems, however, do a ”This is an exceptionally short "training period. " 114 Table 6. The Percentage Agreement inCategorization and Unitization Scoring from Electrical Recordings Categorization Unitization Raters A1 B2 -C D (N) 96 (N) % (N) _% (N) % A (197) 80 (842) 53 (170) 52 (2, 123) 81 B -- -- -- -- (165) 64 -- -- C -- -- s- -- -- -- -- -- 1Rater A is the writer. This test was based on a 6 month elapse period. ' 2A. summary of three separate tests. not entirely account for the low level of reliability. The three raters very often could not fundamentally agree on such seemingly simple dis- tinctions between when a unit of interaction was a fact or piece of infor- mation (e. g. , category 6) or an opinion (e. g. , category 5). The resolu- tion of such disagreements depended on the rater's evaluation of such things as the expertise of the participant, etc. Table 7 contains the reliability tests between raters based on the ”proportion technique. " Inspection of the table shows that eventhough the writer was not able to reach an acceptable level of agreement (e. g. , 70% was the pre-established criterion) with Raters B and C, (_z = -7. 0 and -5. 2, respectively) he was internally consistent (_z_ = 3. 1). He was also able to establish reliability in unitization with rater D (E = 11. 5). 115 Table 7. . _z_ Tests of the Hypothesis that Inter-Rater Reliability is Equal to 70% Agreement for Categorization. and Unitization w Raters Categorization Unitization A B C D A (The Writer) 3.1* —7.0 -5.2 11.5* B -- -- -1. 8 -- c -- -- -- .. * Significant at the . 05 level. The hypothesis of reliability is rejected if z < 1. 65. "'TITIi‘fijllifljMfitfitfliujfilfl:tifllflnjfgi'm‘m'ES