Articles

Wall of Fame

Forum

Quarterly Archive

Return to APS

Leadership
and the Decision-Making Process

Jim Burns is an emergency response manager in a large company specializing in ecological control systems. His work runs the gamut from removal and disposal of toxic waste to cleaning up spills of oil and other contaminants. Typically, his firm works on contracts with organizations both public and private, but occasionally Jim is called upon to deal with situations not covered by existing contracts.

This morning Jim received a phone call from the police in a nearby town. They asked for his firm's assistance in dealing with an oil spill which threatened a nearby river. Jim drove to the site with four of his associates, and within an hour the team of five had obtained the following picture of what happened.

While filling an oil tanker, the driver had gone into the cab and had fallen asleep. Before it was noticed, 10,000 gallons of crude oil had escaped and began making its way five miles downstream and was within four hours of reaching a wildlife sanctuary.

While the potential for environmental damage is clear, the liability is not. The driver was an employee of a small subcontractor who was uninsured and who would be forced into bankruptcy, if deemed liable. The oil company contacted their insurance company who denied any responsibility for claims that might be made against them. Representatives of the State Environmental Protection Agency and Department of Fish and Games were contacted, and they offered their moral support, but neither had the half million dollars that you estimate would be necessary to contain and clean up the spill.

A decision must be made soon as to whether to risk the company's money in a matter in which reimbursement, if any, may have to be decided by the courts. The decision is Jim's to make, and he is experienced in making the difficult judgments which are called for. While conscientious, the members of his team lack this experience and are likely to look to him for direction. Nonetheless they will have to carry out any action which is called for and Jim has found that their involvement in decisions helps them to work together as a team.

Jim Burns' challenge raises two general issues relevant to solving problems and making decisions in organizations. The first issue involves determining what solution or decision should be adopted. In this case should Jim begin the cleanup or defer any action pending resolution of the liability issues? It is this facet of decision making that is the focus of most business school curricula and of the optimization models developed by management scientists. (To be sure, the nature of this particular problem complicates matters further by the introduction of a potential conflict between organizational goals and broader social concerns.)

The second issue revolves around not what should be decided but how and with whom it should be decided. Should Jim decide himself or should he involve the team in some way in determining what decision would be made? In this second issue theories of decision making intersect with theories of leadership style.

It is the latter perspective that has interested me and my colleagues and has become the focus of a large scale program of research at Yale. We are interested in what happens between a leader and the leader's associates in decision-making situations. This interest was inspired by an article by Bob Tannenbaum and Warren Schmidt. Their work distinguished seven different styles varying in influence by the manager and the size of the area of freedom afforded subordinates.

Being a believer in parsimony, I have collapsed some of their alternatives resulting in five styles which are labeled Decide, Consult Individually, Consult Group, Facilitate, and Delegate. Definitions of each of these styles were given to forty specialists in the field of organization development who were asked to locate them on a ten-point scale corresponding to the relative opportunities to influence the decision which they were likely to provide to group or team members. The definitions of these five processes, adapted from Tannenbaum and Schmidt, and the mean scale values, assigned by the OD professionals, are shown in Figure 1.

This language for describing leadership styles can be used in two distinctly different ways. It can be the starting point for the development of a normative model that would help managers or leaders to select the style that best fits a given situation. Like our predecessors, we are convinced that each of the styles is appropriate to certain kinds of situations and that an effective leader is one who explicitly tailors his or her style to demands of the immediate problem at hand.

The taxonomy of leadership styles in Figure 1 can also be used to describe what people do. A common vocabulary, independent of its normative uses, may be helpful in communication and setting of expectations between leaders and their colleagues. Furthermore, these concepts can be used by social scientists in developing a descriptive model aimed at understanding how managers actually decide whether and when to share their decision-making power.

Over the last two decades, my colleagues and I at Yale have conducted a program of research designed to provide us with a normative model which can be used by managers in evaluating specific decisions that they face and in selecting the most effective leadership style for each. The result has been the development of an Aexpert system@ which shows substantial promise in helping managers through the myriad of factors that need attention in deciding when and how to involve their associates in making decisions. In addition, we have made progress in developing a descriptive model of what managers= decision making practices actually do. Our studies, which now involve over 100,000 managers, have been aimed at understanding the factors that actually influence what managers do. Specifically, we have looked at such factors as organizational level, cultural influences, and the role of gender in leadership style. This article outlines the normative model first and then examines our progress in understanding its similarities and differences to what managers actually do.

Decision Quality: Let us first examine what is at stake in the choice of how much and in what way to involve others in solving problems and making decisions. The first, and undoubtedly the most important, is the quality of the decision. Above all we want wise, well reasoned, and analytically sound decisions which are consistent with the goals to be achieved and with potentially available information about the consequences of alternative means of achieving them.

What happens to decision quality as one moves from the autocratic process to the more participative processes? Undoubtedly the nature of the decision and its quality will change as we move across the scale. But will decision quality increase or decrease? A conservative answer, and one which we believe to be consistent with the available research evidence, is that the effects of participation on decision quality depend on certain observable features of the decision-making situation. It depends on where the relevant knowledge or expertise resides, i.e., in the leader, in the group, or both. It depends on the goals of the potential participants, particularly on the extent to which group or team members support the organizational objectives embedded in the problem. Finally, the amount of synergy exhibited in team-based processes depends on the skills and abilities of team members in working together effectively in solving problems.

Decision Implementation: While the quality of the decision may be the most important component of its effectiveness, it is not the only component. Many high-quality decisions have been ineffective because they were not effectively implemented. The effectiveness with which a group or team implements a decision can be shown to depend on the extent to which they are committed to its success. Here the evidence is clearer and less equivocal. People do support what they help to build. Under a wide range of conditions, increasing participation leads to greater "buy-in," commitment to decisions, and motivation to implement them effectively. To be sure, there are some situations in which the motivational benefits of greater commitment are nonexistent or irrelevant to the implementation of the decision. Sometimes the team may not be playing a role in implementation; in other situations the team may view the leader as the expert or as a person with the legitimate right to make the decision and, as a result, may fully support whatever decision the leader might make.

Costs of Decision Making: Apart from considerations of decision quality and implementation, which determine the effectiveness of the decision, there are considerations of efficiency relevant to the decision process. Use of any decision-making process consumes resources, and at the same time can add to resources, albeit, of a different kind. The resources consumed are costs and principally involve the time "used up" in the decision-making process. Increasing the amount of participation will increase the elapsed time to make the decision and, to an even greater degree, increase the number of hours consumed by the process. Both of these meanings of time constitute liabilities of participative leadership styles. Seeking consensus slows down the process and consumes substantially more hours than the directive or even consultative methods of decision making. The first of these costs, increasing the time interval between the occurrence of a problem and obtaining a solution, is most relevant in emergencies where a quick or immediate response is necessary. The second consideration, the hours consumed, is more generally relevant.

Development: Potentially offsetting these costs are developmental benefits of increased participation. Moving from the autocratic to highly participative styles increases the potential value of the group or team to the organization in three ways: 1) It develops the knowledge and competence of individual members by providing them with opportunities to work through problems and decisions typically occurring at higher organizational levels, 2) it increases teamwork and collaboration by providing opportunities to solve problems as part of a team, and 3) it increases identification with organizational goals by giving people "a voice" in making significant decisions in their organizations. These developmental benefits may be negligible when the decision lacks significance, i.e., when the issue being decided is trivial and lacks consequences to the organization. Furthermore, the development benefits may be of negligible value if the group or team members have a nonexistent or tenuous future within the broader organization.

We term a style inefficient when it wastes time without a commensurate return in development. Conversely, it is efficient when it is used judiciously in precisely those situations in which sufficient developmental benefits are realized. It is interesting to note that costs (time) and development, the two components of efficiency, are realized at different points in time. The time costs are immediately realizable. The slowness of response and the number of hours consumed in a group meeting have immediate effects. In contrast, the growth and development of individuals and teams may not pay off for a substantial period of time.

So far our inquiry has led us to identify four outcomes of participation, each of which is contingent on one or more situational factors. To be useful to leaders, we must supplement our analysis with a suitable tool for synthesizing the effects that we have postulated. Figures 2 and 3 depict decision matrices which constitute such a tool. In Figure 2 we show the Time-Driven Model. It is short term in its orientation, being concerned with making effective decisions with minimum cost. No attention is placed on employee development.

In contrast, Figure 3 shows the Development-Driven Model. It may be thought of as a long-term model since it is concerned with making effective decisions with maximum developmental consequences. No attention is placed on time.

To use one of these two models, you must have a decision problem in mind which has two properties. First, it must fall within your area of freedom or discretion, i.e., it must be up to you to decide. Second, there must be some identifiable group of others who are potential participants in the decision.

One enters the matrix at the left hand side at "Problem Statement." Arranged along the top of the matrix are seven situational factors, each of which may be present (H for high) or absent ( L for low) in that problem. To obtain the recommended process, you first ascertain whether the decision to be made is a significant one. If so, you select H and answer the second question concerning the importance of gaining commitment from the group. Continuing this process (while avoiding the crossing of any horizontal line) will bring you to a recommended process. Sometimes a conclusive determination can be made based on as few as two factors (e.g., L,L) others require three (e.g., L,H, H) four (e.g., H,H,H,H,) or as many as seven questions (e.g., H,H,L,L,H,H,H).

Submitting the same problem to both the Time-Driven and Development-Driven Model can be instructive. Sometimes the two models yield identical recommendations. Where they differ, the Development-Driven Model recommends a higher level of participation. Occasionally, the difference may be greater than one position on the participation scale. For example, in the Jim Burns case with which this article began, the Time-Driven Model recommends Decide (H,H,H,H) and the Development-Driven Model recommends Consultation with the Group (H,H,H,H,L).

While the situational factors which identify the columns in Figures 2 and 3 are sufficient for the experienced user of these matrices, a less experienced user may wish to refer to explanations of the situational factors in Figure 4. To practice using the models, read each of the four cases in Figure 5. Underneath each case, we have shown the recommended actions made by the two models along with the path by which these recommendations are obtained.

The model is an outgrowth of 25 years of research on leadership and decision-making processes. We began by collecting cases from managers of successful and unsuccessful decisions and ascertaining which decision process they used on each. If the decisions were unsuccessful, we wanted to find out why, if it could have been avoided and if so, how. Our goal was to build a model which would maximize the frequency of successful decisions while avoiding as many of the unsuccessful ones as possible. Early on, we were joined by social scientists operating in various parts of the world who helped us to test our concepts. We were somewhat encouraged by the findings (based on six separate studies) conducted in three different countries, that decisions made in accordance with a decision tree on which we were working at the time were almost twice as likely to be successful as were decisions that were inconsistent with the model.

But these investigations also made it clear that we had a long way to go, so we continued our efforts to extend and refine our early work. Now we have developed a complex set of equations which show great promise in forecasting the consequences of participation on quality, implementation, cost, and development. The decision matrices shown in Figures 2 and 3 are derived from the use of these equations and are the simplest way in which the implications of the model can be shown on paper. However, the full power of the model is better revealed in a computer program which allows much more complexity and precision while, at the same time, is easier to use. Contained on a CD-ROM, the program has a number of features not possible in a decision matrix. These include: 1) use of eleven situational factors rather than the seven shown in the matrices, 2) permitting five possible responses corresponding to the degree to which situational factors are present, 3) incorporating the Value of Time and Value of Development as situational factors, rather than portraying them as separate matrices, and 4) guiding managers through the process of analyzing situations that they face with definitions, examples, and other sources of help.

We have found by observing manager=s use of the model on problems that they are currently facing on their jobs that the model=s recommendations can be affected by the way in which the problem is framed. For example, if the problem is seen as a deficiency within the team, efforts to find a solution are less likely to be effected than if the problem is located in the situation. Accordingly, we have provided a help screen for Atesting@ a manager=s framing of the problem to make sure that it has been defined in a way that is likely to be productive. In addition, we have found in this rapidly changing world that groups defined by a common manager may not be the most effective for solution of organizational problems. Thus, the software provides a help screen called Team Formation, which provides advice on making up a group to solve a particular problem or make a particular decision.

Toward a Descriptive Model: To study managerial styles (as they are, rather than how our models say they should be), we have developed an innovative measuring device which we term a problem set. It consists of a set of 30 cases, each depicting a manager faced with a decision to make. Figure 5 gives examples of some of the shorter cases from a typical set. Cases are based on real situations and each has been condensed to fit on a single page, providing information on the manager's role, the organizational context, the decision which has to be made, the group of persons which the manager is considering involving, and so on.

The set of cases cover the whole gamut of managerial decisions. Titles include ASaving a Savings Bank,@ ATrimming Expense Accounts,@ARelocating the Head Office,@ etc. However, the cases are not randomly selected. Rather, they vary with respect to the critical factors which the model deems highly relevant to choice of leadership style. Each of the factors contained in the decision matrices shown in Figures 2 and 3 is varied across the set of cases, and each is varied independently of each other factor. This latter feature, which statisticians refer to as a multi-factorial experimental design, is most important since it permits determining which of the relevant factors influences each individual manager, in what way, and to what degree. Thus, the problem set becomes a powerful diagnostic tool capable of revealing "the manager's model," i.e., the way in which each individual manager responds to decision-making situations.

While we originally developed the problem set as a research tool, we discovered early on that managers enjoyed and benefitted from the experience of thinking through how they would deal with highly different situations and of attempting to make sense out of the different choices they made. To aid in the learning that was resulting from this measuring instrument, we have developed a java-based computer program that could quickly analyze a manager's choices and produce a five page individualized report comparing him or her with peers and with the models, both time-driven and development-driven. Furthermore, due to the power of the statistical design underlying the problem set, we are able to show each manager how their choices were influenced by each of the situational factors used in the decision matrices. Each manager's individualized analysis also shows how well his or her choices are likely to result in decisions that are 1) of high quality, 2) would be effectively implemented, 3) were economical in use of time, and 4) would have favorable developmental consequences on their team.

Here we will focus not on what managers have learned from having a mirror held up to them but rather what we have learned from managers about how they go about deciding when and where to share their decision-making power. In a world in which it is common to label managers with terms like autocratic or participative or Theory X or Theory Y, it was instructive to see that managers make different choices in different situations. In fact, it makes somewhat more sense to talk about autocratic and participative situations than autocratic or participative managers. The differences in behavior among managers are about one-third of the size of the differences among situations.

Managers behave situationally. They adapt their behavior to the situations they face. Furthermore, the kinds of situations that evoke autocratic and participative styles are very similar to those in which the normative model would recommend such styles. Each of the seven situational factors shown at the top of Figures 2 and 3 affects the average manager in roughly the same way as they affect the behavior of the model. Managers make more participative choices on highly significant decisions, when they need the commitment of the group, when they lack the expertise, when the likelihood of commitment to their decision is low, when the group's expertise is high and when the group has a history of working together effectively.

But not all managers behave that way! Some are influenced by only one or two of these factors and appear to ignore the rest. Still others are affected by one of these factors but in what we believe to be the wrong way. For example, one fifth of the U.S. mangers that we have studied (and three quarters of all managers from Poland) are more likely to involve others in insignificant, trivial decisions.

One of the most important functions of the feedback is to draw each individual's attention to those aspects of the situation that they are overlooking. We should make it clear that the model not only responds to the seven situational factors but does so configurally. Thus, the effects of one factor depends on the level of certain other factors. For example, where the knowledge resides (in the leader or in the group or both or neither) has more effect on leadership style in highly significant decisions than in those of lesser significance.

Of great interest to us is the fact that managers also behave configurally. There is evidence in our data that managers attend to combinations of factors rather than being influenced by each factor separately. However, these effects are less strong among managers than in the model, suggesting that only a small number of managers behave that way or that, in the typical manager, the configural effects are small in relation to those that are linear.

We said earlier that the situational effects dwarfed the differences among individual managers. While that statement is true, it does not imply that differences among managers in their typical or overall behavior are insignificant or inconsequential. If one averages the choices of a manager on the 30 cases, one obtains a mean score which reflects, on average, where he or she stands on the scale shown in Figure 1. We turn now to consider some of the things that we have found to correlate with differences in where people stand on our ten-point scale of participation.

The first factor is when people took the test! Our data have been collected over a 25 year period and throughout most of that period we have observed an increase in the use of the more participative processes on that scale. Something seems to be producing a move toward higher involvement, more participation, greater empowerment, and more frequent use of teams over time. We do not know precisely what is producing this but we suspect that it reflects changes in the 1) external environment of organizations (greater rates of change, greater complexity), 2) the flattening of the pyramid (greater spans of control resulting in difficulties in hierarchical control), 3) the growth of information technology making it easier to get information closer to the occurrence of problems, and 4) the changing nature of the labor force (higher education, higher needs for independence, etc.). Of the demographic factors, the culture in which the organization functions accounts for the greatest variance. High involvement managers are more likely to be found in countries with high per capita GNP, with a strong democratic tradition and with a highly educated work force.

We have also investigated gender differences and have found women managers to be significantly more participative than their male counterparts. Supporting this conclusion, we have found sizeable differences in the reactions to autocratic men and autocratic women managers. In general being participative is valued by direct reports but this is more true for women than for men. Participative men and women are equally valued but autocratic males are strongly preferred to autocratic women.

A third demographic factor which correlates with leadership style as measured by our problem sets is level in the organization. In each organization that we have studied in depth, the higher the level in the organizational ladder, the more participative the manager. To be sure, we have never carried our investigation up to the level of the CEO where we cannot rely on sample size and the law of large numbers to cancel out chance factors due to personality or to measurement.

We should point out that our findings are restricted to what managers say they would do on a standardized set of cases. While managers have no incentive to lie (since it will only decrease the accuracy of the computer feedback which they alone will receive), we have no guard against self deception. As a possible check on such tendencies, we have given the same problem sets to both the manager and to his or her direct reports. The latter are asked to describe how their manager would respond to each case. The result is striking. Virtually all managers are seen by their direct reports as closer to the left side of that scale than they see themselves. We have referred to this difference as the autocratic shift. We do not know whether the biases are in managers conceptions of themselves, the perceptions of them held by their by their direct reports, or both.

Historically, the people dimension of management has been viewed as basically intuitive, clinical, and "touchy feely." The kinds of analytical approaches which are customary in finance, operations, and to a lesser degree, strategy, have not been applied, or even viewed as applicable to issues of behavior. We have violated that norm and have sought to apply analytical methods to the development of better normative and descriptive tools for understanding leadership style. We will be the first to admit that our model is far from perfect. We ignore deliberately what style managers are "good at," what they are accustomed to practicing, and what they are encouraged to use in their "organizational culture." We do this because we believe that what worked in the past is no guarantee of success in the future. We believe that leadership styles deserve a fresh look.

At Yale and in other environments, when I teach the model presented in this paper and provide people with the computer feedback on their own leadership style, I stress that both are intended to stimulate reflection, and self examination. They are not tools to be slavishly embraced and used in all decisions. I believe that much behavior that is currently driven by habits needs to be converted back into choices. The changing demands of today demand that we reexamine the styles we used in the past and reassess their appropriateness to today's environment.