A double moral hazard model of organization design Elazar Berkovitch, IDC Herzliya Ronen Israel, IDC Herzliya Yossi Spiegel, GSB Tel Aviv University

A double moral hazard model of organization design

Elazar Berkovitch, IDC Herzliya

Ronen Israel, IDC Herzliya

Yossi Spiegel, GSB Tel Aviv University


2

Introduction We develop a model in which the organization structure is

chosen to mitigate MH problems in: The selection of projects (managerial MH) The implementation of projects (MH in teams)

We consider a firm with 2 agents, 2 projects, and 2 tasks (Multi)divisional structure (M-form): each agent gets a full

responsibility for one project and performs both tasks on this project

(Unitary) functional structure (U-form): each agent specializes in one task and performs it on both projects

The main idea: DIV is more efficient ex post but also more susceptible to

managerial MH problem in projects’ selection The optimal organizational structure trades off ex post

efficiency and managerial MH


3

The organizational structure

manager

Division 1 Division 2

p2 m2p1 m1

manager

Departmentp

m1 m2p1 p2

Departmentm

Divisional structure Functional structure


4

Related literature Chandler (1962): "Structure follows strategy:" as firms grow and

adopted more diverse product lines, the difficulties in coordinating functions across different product lines induces a switch from U-form to the M-form

Williamson (1970, 1975): Diversified firms must follow the M-form to alleviate communication overload

Aghion and Tirole (EER, 1995): Under the U-form agents specialize, so the firm saves on training cost. However, the M-form enables agents to better signal their ability to the external job market and therefore strengthens their incentives to exert effort

Rotemberg (JEMS, 1999): Under the U-form agents who perform similar tasks are grouped together so the firm gains more info. about the optimal way to perform tasks. Under the M-form, agents who work on different tasks are grouped together so the firm can better coordinate different tasks with each other.


5

More related literature Qian, Roland, and Xu (2006): The U-form allows coordination between tasks on a company-

wide basis so there is a need for fewer managers to coordinate tasks. The M-form saves on costly cross-division communication

Maskin, Qian, and Xu (RES, 2000): The U-form exploits economies of scale by grouping similar tasks in the same division, but the M-form provides better incentives by promoting yardstick competition between different self-contained divisions

Besanko, Régibeau, and Rockett (2005): Under the U-form, compensation can be tied directly to the agents’ performance. Under the M-form, compensation is tied to additional profitability measures and hence it is more costly to motivate risk-averse agents. This result however may be reversed if there are asymmetries in the effect of tasks on profits, or significant positive externalities across tasks

Corts (2007): Under “individual accountability” (akin to the M-form) agents are compensated on the basis of (noisy) performance measures that depend only on their own efforts. Under “teams” (akin to the U-form), compensation is based on (noisy) performance measures which depend on the agents' joint effort. Hence, agents bear more risk under teams but on the other hand are rewarded on the basis of more performance measures

Harris and Raviv (2002): Under the U-form the firm hires 2 middle managers, each coordinates two similar tasks. Under the M-form, each middle manager coordinates two different tasks. Which is better depends on the likelihood of various cross task interactions and on the CEO’s opportunity cost of coordinating company-wide interactions.


6

The model The firm has a three-level hierarchy:

A manager - screens projects and recommends them to the board of directors

Board of directors – approves/rejects the manager’s recommendations

Two agents (middle managers, business units, or simply employees) – implement projects

If the board rejects the manager's selection, the game ends and all agents get a payoff of 0


7

Projects

All projects yield R if they succeed and 0 o/w H-type projects can always be discovered and require

an initial investment I < R L-type projects are discovered with prob. and

require 0 investment

The manager can conceal L-type projects When the manager recommends H-type projects the board does not know if he did or did not discover L-type projects

measures the importance of managerial ability or the importance of managerial MH


8

Preferences The board of directors is a perfect agent for outside investors and

hence prefers L-type projects

The manager prefers H-type projects: The manager likes to manage expensive projects (“empire building”) H-type projects enhance the manager's general human capital whereas

L-type projects only contribute to his firm-specific human capital L-type projects are “traditional,” whereas H-type projects are more

“innovative” and the manager likes “cutting edge” technologies The manager is emotionally attached to H-type projects

Monetary incentive are insufficient to induce the manager to recommend L-type projects the manager’s wage is a constant which we normalize to 0

The manager will recommend 2 H-type projects if he expects that the board of directors will approve them. O/w he will recommend L-type projects (if he discovers them)


9

Projects' implementation The 2 agents need to perform 2 tasks on each project to

enhance their chances to succeed

For concreteness, we refer to the tasks as production (p) and marketing (m)

Each agent can perform at most 2 tasks the firm can at most adopt 2 projects

The prob. that a project succeeds is

mpmp eeeeq ,


10

Divisional structure Each project is assigned to one agent who performs

both tasks on that project The firm has 2 divisions, each is fully responsible for

one project Agent j's cost of implementing its project:

≥ 1, and ρ (-1, 1) measures the complementarily between tasks p and m 0 < ρ < 1 - the joint production of the two task

exhibits economies of scope -1 < ρ < 0 - there are diseconomies of scope

,ee

eeeeCmjpj

mjpjmjpjd

2, 22


11

Functional structure Each agent specializes in one task and performs it on

both projects The two agents are “production department” and

“marketing department” Agent i’s cost when performing task i on projects 1 and 2

(-1, 1) measures economies of scale 0 < < 1 - economies of scale -1 < < 0 - diseconomies of scale

,ee

e+eeeC iiiiii

f

2, 212

22121


12

The timing

The manager recommends 2 projects to the board of directors

The board accepts or rejects If no projects are recommended, the

game ends If the board accepts the manager's

recommendation, it signs incentive contracts with the two agents

The agents implement the selected projects and payoffs are realized


13

Contracting under the divisional structure


14

Contracting under the functional structure


15

Expected profits

Solving the contracting problems, the expected profits per project, gross of the cost of investment, are:

Divisional:

Functional:

d > f due to MH in teams

4

2Rd

42

2Rf


16

The optimal structure

I

f d

Divisional

Functional

H-typeL-typeprob. H-type

L-typeprob.

DivisionalDivisional


17

Implications Suppose that f I d. All else equal,

the functional structure is optimal for a wider set of parameters as

I (H-type projects are more costly) - ex ante considerations are more important

R (firms operate in more competitive environment) – ex post considerations are less important

ρ (there are weaker economies of scope) σ (there are stronger economics of scale)


18

Implications

Relative to firms with a functional structure, firms with a divisional structure:

have less restrictive standards for project evaluation (approve H-type projects even if f I d)

adopt more projects than firms (under the functional structure the prob. of adopting projects is )

have projects that are more likely to succeed (qd > qf) pay a higher expected compensation to their agents

(qd wd > qfwf)


19

Suppose that for each firm, R is drawn from some distribution function h(R)

The functional structure is optimal iff f I d and > (d–I)/f, or z0 ≤ R ≤ z1, where

Relative to firms with the functional structure, firms with a divisional structure: may or may not have higher expected net returns than firms with a

functional structure - consistent with Mahajan, Sharma, and Bettis (1988)

have projects with a smaller variance of gross returns

Implications

I

I

I=zI=z

42,

4241

min4 10


20

Discussion The firm may wish to choose an ex post inefficient functional

structure in order to mitigate the manager's incentive to recommend H-type projects

The idea that firms adopt ex post inefficient actions to provide incentives ex ante is not new: Debt may lead to costly bankruptcy ex post but may mitigate

managerial MH ex ante (e.g., Grossman and Hart, 1982) Borrowing from 2 creditors may lead to inefficient monitoring activity

and may prevent refinancing ex post but may improve projects' selection ex ante (Dewatripont and Maskin, 1995)

Can the firm achieve this goal by other means? Threaten to fire the manager if he recommends H-type projects Hire two managers to screen projects and use yardstick competition to induce

them to recommend L-type projects Distort the agents' compensation such that they will implement H-type projects

inefficiently


21

Size effect Suppose that each agent can complete 2n tasks,

where n is a positive integer the firm can adopt 2n projects Divisional structure: the firm will assign n projects to

each agent who will perform tasks p and m on these n projects

Functional structure: each agent will specialize in one task and will perform it on all 2n projects

With prob. there are at least 2n L-type projects and with probability 1- , there are no L-type projects

Since each agent needs to perform tasks p and m on n > 1 projects, Cd is affected by both econ. of scope and econ. of scale


22

Size effect - n ( > 0)

I

f d

Divisional

Functional


L-typeprob.


n means that econ. of scale are more significant


23

Complexity effect Suppose that each project requires 2k tasks, where k is a positive

integer To ensure that the firm can adopt exactly two projects, we will also

assume that there are 2k agents, who can perform two tasks each To ensure that the prob. of success does not increase simply due

to the increase in k

Under the divisional structure, the firm will establish two divisions and will assign each division the full responsibility over one project Now there are k > 1 agents in each division, so there is MH in teams

even under the divisional structure Unlike the functional structure, here each agent performs two different

tasks on the same project, whereas under the functional structure, each agent performs the same task on two different projects

Under the functional structure, each of the 2k agents specializes in a single task and performs it on the two projects

k

ijk e

keeq

2

121

1,,


24

Complexity effect - k

I

f d

Divisional

Functional


L-typeprob.


k means that MH in teams is more severe


25

Cross-task asymmetries Suppose that task m has a bigger influence on the likelihood

of success than task p:

Divisional structure: h induces each agent to shift effort from task p to task m Shifting effort from m to p has a first order effect on the prob.

of success but second order effect on the cost of effort

Functional structure: since h > 1, it is more efficient to pay agent m an extra dollar at the expense of agent p; the firm will effectively shut down department p The firm only operates the m department whose marginal

productivity is increasing with h

1

1

2,

hh

heeeeq mpmp


26

Cross-task asymmetries h

I

f d

Divisional

Functional


L-typeprob.


only if I is sufficiently large


27

Narrow business strategies (NBS) Adopt a strategy that imposes a technological constraint

that prevents the firm from implementing H-type projects Advantage: the firm can mitigate managerial MH without

adopting an ex post inefficient organizational structure Disadvantage: the firm forgoes H-type projects even if they are

profitable and even if there are no L-type projects around

Suppose that with prob. μ, H-type projects require an initial investment I1 and with prob. 1-μ they require an initial investment I2, where I1 < f < I2 < d

I = I1: managerial MH under both organizational structures I = I2: managerial MH only under the divisional structure (the

board of directors rejects H-type projects)


28

Expected payoffs under NBS Under NBS, the firm focuses exclusively on L-type

projects

The firm might as well adopt the ex post efficient divisional structure and implement L-type projects efficiently

L-type projects are available only with prob. .

The expected per-project profit of the firm under NBS:

dN =O


29

Expected payoffs without NBS Under the divisional structure, the firm will

always have H-type projects:

Under the functional structure, the firm will have H-type projects with prob. and L-type projects with prob. (1-):

21 1 II=O dd

fff I=O 11


30

When is NBS desirable?

I1=1, I2=8, R=3, =0.5, =0, =0.4.

ON-Od

ON-Of

Of-Od ON with Of with

NBS

Funct.

Divisional


31

When is NBS optimal?

is high (info. MH is “important”)

is small (H-type projects are not “that bad”)


32

Conclusion The organizational structure of the firm is chosen by

trading off ex post efficiency against mitigating managerial MH

The divisional structure is more efficient ex post but more susceptible to managerial MH

The functional structure may overcome managerial MH

Adopting NBS and focusing exclusively on L-type projects could be profitable only when managerial skills are important and the managerial MH problem is severe

Documents

A double moral hazard model of organization design Elazar Berkovitch, IDC Herzliya Ronen Israel, IDC Herzliya Yossi Spiegel, GSB Tel Aviv University