The Determinants of Managerial Decisions Under Risk

1
The Determinants ofManagerial Decisions Under
Risk

• Martin G. Kocher
• University of Innsbruck
• Ganna Pogrebna
• Columbia University
• Matthias Sutter
• University of Innsbruck

Roma, 30 June 2007
2
Introduction

• In hierarchical organizations the role of a team
  leader often requires making decisions, which do
  not necessarily coincide with the majority
  opinion of the team. However, these decisions are
  final and binding for all team members.
• We are especially interested in decisions on
  projects with a stochastic outcome, i.e., risky
  decisions.
• While business practices provide substantial
  evidence that leaders make managerial decisions
  (e.g. Lieberman et al., 1990, or Knott, 2001),
  the question which factors motivate such
  decisions remains largely unanswered.
• We study managerial decision making in a
  controlled laboratory experiment we identify the
  conditions and reasons why team leaders may
  overrule the majority opinion of their team.

3
Related literature

• Economic research has largely concentrated so far
  on one particular type of leadership, i.e.
  leading by example.
• Theoretical basis for research on leadership with
  different information structures e.g. Hermalin,
  1998 Rotemberg and Saloner, 1993 and 2000
  Kobayashi and Suchiro, 2005 Komai et al., 2007.
• Experimental studies have confirmed the
  efficiency-increasing impact of leading by
  example, e.g., in public goods games (e.g.
  Vesterlund, 2003 Potters et al., 2005, 2007
  Andreoni, 2006 Güth et al., 2007).
• But we are not aware of any paper that studies
  managerial decisions for choices under risk.

4
Basic experimental design I

• The experiment consists of two stages an
  individual stage and a team stage (design
  balanced for order effects).
• We design 17 (34) pairs of lotteries with
  different payoff scheme and risk coefficients s
  for leaders and ordinary team members. In the
  individual stage, we elicit individual
  preferences of participants over these pairs of
  lotteries (lotteries are in random order).
  Subjects are also exposed to a risk test (Holt
  and Laury, 2002).
• In the team stage subjects are randomly assigned
  to teams of three people. Members of a team are
  asked to elect a leader (through self-proposals).
  Team members can communicate using an interactive
  chat (z-tree). Every team has three chat periods
  for discussion. Each chat period lasts three
  minutes, after which team members have to cast
  their votes. If a simple majority is not reached,
  the team proceeds to the next chat period (3
  times and, then, leader at random).

5
Basic experimental design II

• After the election of the leader, teams receive
  17 decision problems. In each problem they have
  to choose between paired lotteries. Although
  these decision problems are identical to the
  problems used in the individual stage, they are
  shown in a different order and framed
  differently.
• For each decision problem, all three team
  members, including the leader, are requested to
  cast a vote for one of the lotteries in a lottery
  pair.
• The team leader is informed about the lottery,
  chosen by at least two team members, and, then,
  has an option to either confirm the majority
  decision or to pick an alternative lottery. This
  decision is final and determines the payoff of
  the entire team.

6
Risk attitude and leader selection

• We consider the choices of 92 out of 108
  participants, who have made their decisions
  consistently in the Holt-Laury procedure. These
  subjects have switched from the safe choice to
  the risky choice at most one time. Overall,
  consistent subjects made 5.64 safe choices,
  indicating a slight degree of risk aversion in
  the aggregate.
• Leaders were significantly less risk averse than
  ordinary team members (p-value 0.05
  Mann-Whitney-U-test).
• There were 31 out of 36 teams that were
  successful in electing a leader. Out of them 25
  (81) chose the leader in the first vote (by
  simple majority voting), 5 (16) needed two
  voting rounds, and 1 team (3) agreed on a leader
  only in the third voting round. 5 teams could not
  agree on a team leader.

7
Managerial decisions I

• In 35.3 of cases leaders overrule the majority
  vote in their team. We use logit regression
  analysis to determine the impact of different
  explanatory variables on the likelihood of a team
  leader making such managerial decisions.
• More than 70 of the variability in the data can
  be explained by decision contradiction (whether
  the leader has voted for a different lottery than
  the two other members in the team during the team
  stage of the experiment). The variable preference
  contradiction is also significant (if the two
  ordinary team members vote for a different
  lottery than the leader has chosen in the
  individual stage).
• Risk seeking leaders are more likely to resort to
  managerial decisions that overrule the other team
  members. We also find that two personal
  characteristics gender and age are
  significant.

8
Managerial decisions II
9
Brief conclusion

• The main contribution of this paper is twofold
• First, we analyze leadership in a risky
  environment, which is one of the key elements of
  managerial decision making.
• Second, we consider leadership as a formal
  authority where the leader makes the ultimate
  decision for his group.
• These two important aspects of leadership have
  been largely understated in the previous
  research.
• Our design is extendable in several dimensions.

10
Thank you!