College and Research Libraries BEVERLY LYNCH Participative Management in Relation to Library Effectiveness This paper reviews a recent study on the influence of participative management on library performance. Because most of the recent the- oretical and empirical research being done in this area is ignored and an invalid measure of participation in decision making is used~ the study provides no basis for the generalization that an increase in the library staff's participation in decision making will increase the li- brary's effectiveness. IF RESEARCH IN LmRARIANSHIP is to pro- gress toward its objective of "extending the existing body of factual knowledge concerning the values and procedures of libraries in their many aspects," re- searchers must make every appropriate use of insights, concepts, and methods of other disciplines. 1 One advantage of the affiliation of library schools with universities is that it facilitates the use of theories developed in other fields. The library researcher who borrows a theory, however, must fully understand its assumptions and limitations and must be thoroughly familiar with the empirical evidence which tends to sup- port or to limit the application of the theory to the problems of librarianship. Naive or uninformed use of approaches found useful in other disciplines can be damaging, particularly if library ad- ministrators act on the basis of the in- valid generalization. It is therefore im- portant that studies that borrow from other fields be monitored critically so Ms. Lynch is the newly appointed execu- tive secretary of the Association of College and Research Libraries. At the time this paper was written she was a visiting lec- turer at the University of Wisconsin Library School, Madison. 382 I that only well-founded research will be accepted. This paper examines a study by Mau- rice P. Marchant, entitled, "Participa- tive Management as Related to Person- nel Development," which is based upon his dissertation, "The Effects of the De- cision Making Process and Related Or- ganizational Factors on Alternative Measures of Performance in University Libraries."2 Marchant's investigation was designed to measure the influence of the professional librarian's partici- pation in decision making upon the li- brary's effectiveness. Although Marchant found no statistically significant rela- tionships between these variables, he re- ports significant relationships between participation in decision making and staff satisfaction. Marchant borrows theory from be- havioral science and deals with a sub- ject which is of current interest to li- brarianship. His study is likely to be cited in library literature as evidence for the desirability of change in man- agerial style. Marchant's work is better than many recent studies on library or- ganization and management in its struc- ture, reporting, and use of statistical methods, but it reflects an insufficient knowledge of the theoretical and em- pirical work which has been done in participative management. There are al- so several faults in method and data analysis. THE THEORY The major theoretical basis for Mar- chant's work stems from Rensis Likert's theory of participative management, which was presented in Likert's New Patterns of Management and then elab- orated in his The Human Organiza- tion. 3 According to Likert a work group that accepts the "principle of suppor- tive relationships" will achieve a higher rate of productivity than one that does not. Likert relies heavily for support of this theory on the experiment conducted by Morse and Reimer in four clerical divisions of a large company over a pe- riod of a year. 4 Morse and Reimer re- port significant increases in productivity for both "'participative" and "hierarch- ical" units, with a slightly higher in- crease in the hierarchically controlled division. Despite these contrary results the researchers felt that over a longer period of time the adverse effects on morale, which they observed in the hier- archial groups, would reduce their pro- ductivity. They assumed, however, that no self-corrective measures would be in- troduced into the hierarchial groups as production fell. According to two recent comprehen- sive reviews research results relating par- ticipative management to productivity have been inconclusive. 5 Day and Ham- blin report democratic supervision to be more effective, while Shaw reports au- thoritarian treatment to be more effec- tive.6 Studies by McCurdy and Eber, Sales, and Spector and Suttell, indicate no relationship between leadership styles and productivity. 7 Current re- search in a government organization fails to support a positive relationship of participative management to produc- . tivity.8 Furthermore, Carey, Korman, and Vroom have suggested that studies which consider the causal effect of man- agerial style on subordinate perform- Participative Management I 383 ance can be interpreted as measuring the reverse effect, that is, the impact of performance on managerial style.9 More recently, another experiment provides strong evidence to support this hypothe- sis.10 It confirms an earlier study which reported that organizational behaviors are sensitive to prior organizational ef- fectiveness.11 Vroom and Mann, French et al., Foa, Gibb, Patchen, and Pelz suggest that the effects of participative management on productivity may depend upon what ex- actly is being measured in the study, participative supervision, considerate su- pervision, or closeness of supervision, and upon the needs and expectations of subordinates. 12 These studies suggest that a participative, considerate, or sup- portive leadership style may be most ef- fective when the decisions are nonrou- tine in nature, when the information required for effective decision making cannot be centralized nor standardized, or when, because rapid decision making is not required, there is time for sub- ordinates to be involved in the process. Whether subordinates feel a need for independence, regard their participation in decision making as legitimate, con- sider themselves capable of contribut- ing to the decision making, or are con- fident enough to work without close su- pervision also may influence the effec- tiveness of .a participative leadership style. 13 Unfortunately, none of this lit- erature is presented by Marchant, nor does the design or methodology of his study reflect it. Previous research has not demonstrat- ed that participative management causes high productivity. Most of the research treats small groups with only incidental references to the organization as a whole. Most of the hypotheses regard- ing the relationships of participative de- cision making to other organizational variables have been too gross to be proved or disproved. Marchant defines his purpose as an at- 384 I College & Research Libraries • September 1972 tempt "to test the application of Lik- ert's participative management theory within academic libraries."14 Presum- ably he intends not to reevaluate Lik- ert's theory, but rather to apply it to a library setting. Despite objections in the management literature to the grossness of previous hypotheses, Marchant has not refined his own. He states his prin- cipal hypothesis as "the greater the ex- tent to which the professional personnel on the staff are involved in the library's decision making processes, the more ef- fective will be the library's perform- ance."15 Marchant does not define par- ticipation in decision making. Whether he means actual decision making, per- ceived decision making, actual or per- ceived influence in decision making, in- creased communication, supportive re- lationships, or something else remains unclear. Furthermore, while purporting to ap- ply Likert's theory, Marchant changes Likert's independent variables, adds de- pendent variables, and entirely omits the intervening variables of the Likert model, inserting his own control vari- ables, some of which are treated as in- tervening variables in his analysis. (See Figures 1, 2. ) Although Marchant introduces the important variable of job satisfaction, he neglects to cite any literature on the subject, although more than four thou- sand articles have been published since the 1930s.16 Despite the large amount of research, experts still do not agree on the causes of job satisfaction, nor has the question whether job satisfaction influences productivity or vice-versa been answered. 17 Marchant ignores the controversy. The omission reflects only one exam- ple of the theoretical and empirical re- search that Marchant might usefully have brought into the library literature. For instance, he might have described the characteristics of the decision mak- ing process, defined such concepts as par- ticipation, satisfaction, or productivity, or given us a comprehensive review of the literature in one or more areas. Since he fails to do so, his contribution is restricted to the merits of his particu- lar investigation, and even that is seri- ously flawed for lack of evidence of awareness of previous work. MEASUREMENT In his search for ways of measuring independent variables Marchant adopts Likert's "Profile of Organizational Characteristics," which has been validat- ed as a measure of managerial style.l8 A researcher who borrows another's in- strument must either use it in the con- text of the same theory or establish its validity in the context of his own or different theory; Marchant does neither. Moreover, he simply extracts three of Likert's eighteen questionnaire items and designates them as his "Decision Simplified Version of the Likert Model Causal Variables Supportive relationship Group decision making High performance goals Intervening Variables Favorable attitudes toward supe- rior High confidence and trust High reciprocal influence Excellent communication; up, down and lateral High peer-group loyalty High peer performance goals at all levels re: productivity, quality, scrap Source: Likert, The Human Organization, p. 137 Figure 1 End-result Variables Low absence and turnover High productivity Low scrap Low costs High earnings Participative Management I 385 Marchant's Research Model Independent Variables 1. Decision making 2. Organizational pro- file . . . . Control Variables 1. Doctoral degrees granted 2. Prerequisites available to librarians 3. Library expenditures 4. Decentralization of collection 5. Library autonomy 6. Beginning librarian salary 7. Staff size and composition 8. Collection size and growth 9. Staff breadth of education 10. Service time Performance Measurements Faculty evaluation Circulation Long-range planning Uniformity of evaluation Staff satisfaction Source: Marchant, "Participative Management .•• ," p. 51 Figure 2 Making Index." Marchant then uses all eighteen Likert items to form what he calls "Profile Index." Marchant's use of a portion of the Likert questionnaire to measure one variable and then his use of the whole questionnaire, including the extracted portion, to measure an- other variable seems extraordinarily naive, particularly since Marchant re- ports the entirely anticipated correla- tion, .9696.19 This strong relationship merely demonstrates that the "Decision Making Index" and the "Profile Index" are different measures of the same char- acteristic, managerial style. Although the validity of the "Deci- sion Making Index" cannot be deter- mined from the single correlation ma- trix Marchant provides, some assessment of its validity can be made. 2° First, the average intercorrelations computed for these items in the "Decision Making In- dex" were computed and compared with the average correlations computed for these same items with the items in the "Profile Index." The within-cluster aver- age ( .6966) is higher than the between- cluster average ( .5898) as it should be if different constructs are being measured. As a comparison, two other three-item clusters were extracted from the "Pro- file Index." The items grouped under the heading, "Leadership process," have an average correlation of .69; they have an average correlation with items in the "Profile Index" of .5588. Three items listed as a "communication process" group have a within-cluster correlation of .64; the average correlation of these items with items in the "Profile Index" is .5735. Although the within-cluster av- erages are higher than the between-clus- ter averages for each of these groups, the difference between the correlations is dubious evidence to support the con- clusion that more than one construct is being measured. A simple examination of the correlation matrix indicates that the items of the "Decision Making In- dex" correlate about equally well with each other and with the items in the "Profile Index." It is unlikely that the Likert instrument, as used by Marchant, is measuring more than one construct. Even if the validity of Marchant's "Decision Making Index" were unas- sailable, questions would arise as to its reliability. Marchant makes no effort to 386 I College & Research Libraries • September 1972 support this characteristic. The demon- stration would have been difficult. A three-item scale is rarely reliable, for it requires that all three items measure a single concept. What is considered to be a unidimensional construct, as is as- sumed by Marchant in the "Decision Making Index," may be two-dimension- al or more. Three questions form the index: whether problems are discussed at various levels in the organization, whether decision makers are aware of the problems at the lower levels in the organization, and whether subordinates are able to make their own work deci- sions. Marchant maintains that each question taps a discrete dimension of the decision making; it is more likely that each question is measuring one as- pect of managerial style. Before the relationships between spe- cific variables can be tested there must be some confidence in the measurement. Since Marchant's "Decision Making In- dex'' inspires little confidence, his con- clusions based upon it are correspond- ingly shaky. Likert validated the "Pro- £Ie Index" to some extent, making it a more acceptable measure. But it is a measure of the general concept of man- agerial style, not a measure of participa- tive decision making. The evidence Marchant uses to assess his dependent variables consists of cir- culation statistics and responses to a va- riety of questionnaires. The use of cir- culation statistics as a measure of li- brary effectiveness has well-known limi- tations, but the measure has been ac- cepted in the literature as being the best one so far developed. 21 In the case of the questionnaire measures of satisfac- tion, faculty evaluation of effectiveness, long-range planning, and uniformity of evaluation, however, Marchant neglects to establish their validity or reliability. DATA ANALYSIS Marchant aggregates the data from the individual questionnaires to deter- mine a library's score. The library scores ( N = 22) are used in a step-wise regres- sion procedure in order to calculate the proportionate variance in the dependent variables accounted for by the inde- pendent and control variables. Variables accounting for the largest proportion of the variance, in most cases, are en- tered then into a regression equation. The coefficient of determination, R2, is often used as a convenient measure of the success of the regression equation in explaining variations in data, and Marchant adopts it as a measure in his study. However, an increase in R2 can be the result, not of a real significance of the variable added to the model, but of the fact that the number of parame- ters in the model is getting close to the number of observations.22 Since there are seventeen variables in the Marchant model and twenty-two observations, his reported coefficients of determination must be viewed with caution. Further- more, in deleting each variable with the highest partial correlation from the sub- sequent regression equations, Marchant violates the assumption required in re- gression analysis that the error terms are randomly distributed. A result of this violation can be that variables are re- tained erroneously in the model and their significance overestimated. Unfortunately, the table reporting partial correlations between managerial style variables and performance mea- sures in Marchant's article is misleading. It implies that the measures of the re- lationships between the performance measures and the "Decision Making In- dex" and "Pro£le Index," as reported in the table of partial correlations, are independent of the effects of all of the control variables. That is not the case. The partial correlation reported be- tween the "Pro£le Index" and staff sat- isfaction was calculated by controlling for ratio of librarians to staff, while the partial correlation b etween the "Profile Index" and faculty evaluation was cal- culated by controlling only for the ef- fect of number of librarians per full- time student.23 Furthermore, book ex- penditure per student explained the highest proportion of the variance in overall satisfaction when all the control variables and the "Profile Index" were accounted for, but it was omitted when the partial correlation was computed. While book expenditures per stu- dent was identified as the most impor- tant variable affecting staff satisfaction it would be improper to assume that redistributing available funds so as to increase the book budget would result in higher staff satisfaction. The book budget variable stands for the entire wealth factor set of variables, and to divide the budget in some ratio not typical of the libraries studied would insert a condition beyond the predic- tive capability of the equations. Further evaluation of the data . . . indicated two other important facts. First, variables representing several di- verse sources of influence contributed to the cumulative proportionate vari- ance. Second, variables from three dif- ferent organizational areas were the primary contributors. The first [book expenditure per student] represents the university's financial support. The second [Profile Index] represents man- agerial style, which is largely deter- mined by library top management. The third [breadth of education] characterizes the staff composition. It is apparent that no one ingredient de- termines staff satisfaction. 24 A more serious flaw in Marchant's analysis stems from his use of partial correlations to determine the causal pri- ority of the variables. Partial correla- tions do not demonstrate causal priori- ty; they assume knowledge of it. Robert Gordon, in his discussion of the partial- ling fallacy in multiple regression, com- ments on this directly: A somewhat more subtle version of the partialling fallacy is likely to be committed in multivariate studies that Participative Management I 387 present all of the possible highest-or- der partials between each one of a large set of independent variables and the same dependent variable. Appar- ently, this practice also draws inspira- tion from Kendall and Lazarsfeld al- though the procedures they advoc~ted are actually quite different in logic. Kendall and Lazarsfeld' s procedures assume knowledge concerning the pre- sumed causal priority of the variables -they are not intended to provide that knowledge. Roughly, they address the question, "Is variable A causally prior to B, or is it irrelevant?" and not the question, "Is variable A causally prior to B, or is B causally prior to A?" Yet it appears to be the latter question that is being posed when re- searchers calculate all possible highest- order partials to see which variable will emerge with the largest par- tial. ... There is no statistical rule for at- tributing controlled covariation to the influence of one rather than another of the independent variables, regard- less of the disparity in the size be- tween their partial correlations. The question (of whether variable A is prior to B or B is prior to A) is simply not answerable by these means.25 The Simon-Blalock method for infer- ring causal relationships from correla- tion data, which Marchant adopts, in- volves use of partial correlations in the circumstances where certain combina- tions of correlations are expected to dis- appear.26 Thus the Simon-Blalock meth- od can be used in some cases to deter- mine which variables might be deleted from a theory. Essential to the appro- priate use of the Simon-Blalock method is the existence of a theory. That is, if it is theorized that no relationship exists, between managerial style and size, the correlation between these variables holding all other variables constant' would be expected to be zero. If th~ correlation is not zero, the model is ad- justed accordingly, and a test of the ad- justed model then is made. 388 I College & Research Libraries • September 1972 Marchant's Preliminary Model of Causality Explaining Faculty Evalua- tion ( see Figure 3) shows faculty evalu- ation being directly influenced by staff satisfaction and collection size. Variables of managerial style, wealth, and breadth of education (measured by "counting up the number of different disciplines in which a given staff reported having a degree and giving each discipline a weighted value depending upon the high- est level of degree reported") explain faculty evaluation indirectly, through their effect upon collection size and staff satisfaction. 27 Marchant's Preliminary Model Xt = Managerial style X2 = Wealth Xt X2 X := Breadth of education X , 1/ '\..X, X4 = Collection size \ / Xr. = Staff satisfaction X~Xn Xo = Faculty evaluation Source : Marchant, " Partic ipative M a na gem ent .. . ," ]1 .• '54 Figure 3 In this model Marchant assumes that no relationships exist between collection size and staff satisfaction ( x4 and x5)' nor between size and breadth of educa- tion or managerial style ( x 4 and x 3 or X1 ). However, the simple correlations reported between the "Profile Index" and some measures of size suggest that some relationships between these vari- ables may exist: the correlation between the "Profile Index" and collection size is -.1706, although it is not significant; the correlation between number of li- brarians ( another measure of size) and the "Profile Index" is - .4285 (significant at the .05 level). 2 8 These correlations are interesting in light of Marchant's attempt to control for size by studying Association of Research Libraries' mem- bers with collections under three million vplumes. However, the size variables are deleted from the data analysis along with several other control variables be- cause of their potentially confounding nature. Variables related to size and growth of collection and staff are heavily rep- resented [in the group of potentially confounding variables]. It is possible that the managerial style of the larger and older libraries tended to develop when classical theory of administration was current and that smaller libraries tended to be more influenced by later, more participative theories. If this is true, these variables may simply reflect the managerial style which accompa- nies them. 29 Aside from the intriguing assumption that managerial style is a static variable, this passage suggests that age and size of library might be causally prior variables, but Marchant dismisses them as mere re- flections of managerial style. There is no convincing theoretical argument sup- porting his exclusion of these variables from his model predicting academic li- brary effectiveness. If several initial as- sumptions about the ordering of the variables are equally plausible, the sta- tistical techniques used by Marchant provide no basis for deciding among them. SuMMARY Marchant, in his study of the effects of the decision making process on li- brary performance, ignores most of the recent theoretical and empirical research being done on these variables. He at- tempts to test hypotheses others were testing in the 1940s and 1950s. His re- sults, like theirs, are inconclusive. Marchant fails to establish the validi- ty and reliability of his decision making index. Because the only independent variable being measured in his study is the general concept of managerial style, he is unable to test his hypothesis-"the greater the extent to which the profes- sional personnel on the staff are in- volved in the library's decision making processes the more effective will be the library's performance." Inconclusive though Marchant's study is, he does suggest some variables that might be determinants of satisfaction or dissatisfaction in academic libraries. He provides no convincing theoretical reasons, however, to explain the inclu- sion of these variables rather than any others. He bases his choice solely on the results of the step-wise regression pro- cedure, which attributed to them the largest proportion of the explained var- iance. Unfortunately, the step-wise re- gression procedure warrants no such conclusions. Had Marchant presented the assump- tions and limitations of Likert's theory Participative Management I 389 and offered empirical evidence that sup- ported or limited the application of this theory to problems of librarianship, li- brary science might have profited. Had Marchant used measures developed by Likert in the context of the same theory or developed valid and reliable mea- sures of his own, his research would have added to the knowledge of the re- lationships of participation in decision making and library effectiveness. Given the inadequacies of the theoretical de- velopment and the invalid measure of decision making, however, the study pro- vides no basis for the generalization that an increase in the library staff's par- ticipation in decision making will in- crease library effectiveness. REFERENCES 1. Douglas Waples, "The Graduate Library School at Chicago," Library Quarterly 1:26 (Jan. 1931). 2. Maurice P. Marchant, "Participative Man- agement as Related to Personnel Develop- ment," Library Trends 20:48-59 (July 1971 ). --, "The Effects of the Decision Mak- ing Process and Related Organizational Factors on Alternative Measures of Per- formance in University Libraries." Unpub- lished Ph.D. dissertation (Education), Uni- versity of Michigan, 1970. 3. Rensis Likert, New Patterns of Manage- ment (New York: McGraw-Hill Book Co., 1961 ). --, The Human Organization (New York: McGraw-Hill Book Co. , 1967). 4. Nancy C. Morse and Everett Reimer, "The Experimental Change of a Major Organiza- tional Variable," Journal of Abnormal and Social Psychology 52:120-29 ( 1956). 5. Aaron Lowin, "Participative Decision Mak- ing: A Model, Literature Critique, and Pre- scriptions for Research,'' Organizational Behavior and Human Performance 3:68- 106 (1968); and Stephen M. Sales, "Su- pervisory Style and Productivity: Review and Theory," Personnel Psychology 19: 275- 86 ( 1966). 6. Robert C. Day and Robert L. Hamblin, "Some Effe cts of Close and Punitive Styles of Supervision," American Journal of Soci- ology 6~ :499L510 (1964). Marvin E. Shaw, "A Comparison of Two Types of Leader- ship in Various Communication Nets," Journal of Abnormal and Social Psychology 50:127-34 (1955). 7. Harold G. McCurdy and Herbert W. Eber, "Democratic versus Authoritarian: A Fur- ther Investigation of Group Problem-Solv- ing," Journal of Personality 22:258-69 ( 1953). Stephen M. Sales, "A Laboratory Investigation of the Effectiveness of Two Industrial Supervisory Dimensions," Un- published M.S. Thesis, Cornell University ( 1964). Paul Spector and Barbara Suttell, An Experimental Comparison of the Effec- tiveness of Three Patterns of Leadership Behavior. (Washington, D.C. : American Institute for Research, 1957). 8. Reed M. Powell and John L. Schlacter, "Participative Management-a Panacea?" Academy of Management Journal 14:165- 73 (June 1971). 9. Alex Carey, "The Hawthorne Studies: A Radical Criticism," American Sociological Review 32:403- 16 ( 1967). Abraham K. Korman, "Consideration," "Initiating Struc- ture," and " Organizational Criteria-a Re- view," Personnel Psychology 19:349- 61 ( 1966). Victor H. Vroom, Work and Moti- vation (New York: J. Wiley, 1964). 10. Aaron Lowin and James R. Craig, "The In- fluence of Level of Performance on Man- agerial Style: An Experimental Object-Les- son in the Ambiguity of Correlational Data," Organizational Behavior and Hu- 390 I College & Research Libraries • September 1972 man Performance 3:440-58 (1968). 11. George Franklin Farris, "A Causal Analysis of Scientific Performance," Unpublished Ph.D. dissertation, University of Michigan ( 1966). 12. Victor H. Vroom and Floyd C. Mann, "Leader Authoritarianism and Employee Attitudes," Personnel Psychology 13:125- 39 (1960 ). John R. P. French, Jr. et al., "An Experiment on Participation in a Nor- wegian Factory," Human Relations 13:3-19 (1960). Uriel G. Foa, "Relation of Workers Expectation to Satisfaction with Super- visor," Personnel Psychology 10:161-68 ( 1957). Cecil A. Gibb, "Leadership," in The Handbook of Social Psychology, ed. by Gardner Lindzey and Elliot Aronson. (Reading, Mass.: Addison-Wesley, 1969). Martin Patchen, "Supervisory Methods and Group Performance Norms," Administra- tive Science Quarterly 7: 27 5-94 ( 1962 ) . Donald C. Pelz, "Influence: A Key to Ef- fective Leadership in the First Line Super- visor," Personnel29:209-11 ( 1952). 13. Alan C. Filley and Robert J. House, Man- agerial Pro9ess and Organizational Behav- ior (Glenview, Ill.: Scott, Foresman, 1969). 14. Marchant, "Participative Management ... ," p. 51. 15. Marchant, "The Effects of the Decision Making Process . . . ," p. 26. 16. Edwin A. Locke, "What Is Job Satisfac- tion?" Organizational Behavior and Human Performance 4:309-36 (1969). 17. Donald P. Schwab and Larry L. Cum- mings, "Theories of Performance and Satis- faction: A Review," Industrial Relations 9: 408-30 (Oct. 1970). 18. Marchant, "Participative Management ... ," p. 52. Likert, Human Organization. p. 117- 22. 19. Marchant, "The Effects of the Decision Making Process ... ," p. 132. 20. Ibid., p. 298. 21. Leon Carnovsky, "Survey of the Use of Li- brary Resources and Facilities," in Library Surveys, ed. by Maurice F. Tauber and Ir- lene Roemer Stephens. (New York: Colum- bia University Press, 1967). 22. N. R. Draper and H. Smith, Applied Re- gression Analysis (New York: J. Wiley, 1966). 23. Marchant, "The Effects of the Decision Making Process ... ," p. 157 and p. 250. 24. Ibid., p. 158. 25. Robert A. Gordon, "Issues in Multiple Re- gression," American Journal of Sociology 73:592-616 (March 1968). 26. Herbert A. Simon, Models of Man (New York: J. Wiley, 1957); Hubert M. Blalock, Jr., Causal Inferences in Nonexperimental Research (Chapel Hill: University of North Carolina Press, 1964). 27. Marchant, "The Effects of the Decision Making Process ... ," p. 114. 28. Ibid., p. 134. 29. Ibid., p. 136.